Chemical Manufacturers Association, Et Al., Petitioners, v. U.S. Environmental Protection Agency, Respondent., 870 F.2d 177 (5th Cir. 1989)
Theodore L. Garrett, Corinne A. Goldstein, Jay T. Smith, Covington & Burling, Washington, D.C., for Chemical Mfrs. Ass'n.Dov Weitman, Lee M. Thomas, Acting Admin., E.P.A., Michael Wenig, David J. Kaplan, Francis S. Blake, Gen. Counsel, U.S.E.P.A., Roger Marzulla, Acting Asst. Atty. Gen., Land & Natural Resources Div., U.S. Dept. of Justice, Beth S. Ginsberg, Asst. Atty. Gen., Margaret N. Stand, Chief, Environmental Defense Section, Land & Natural Resources Div., Dept. of Justice, Washington, D.C., for U.S.E.P.A.Ronald J. Wilson, Robert Wayne Adler, Washington, D.C., for intervenor Natural Resources, etc.William A. Anderson, II, A. Frank Smith, III, Washington, D.C., for Sterling Chemicals, Inc.John C. White, William A. Olson, Houston, Tex., for Gulf Coast Waste Disposal Authority.Theodore J. Garrett, Covington & Burling, Jay T. Smith, Washington, D.C., for Union Carbide Corp., Air Products Mfg. Corp., and Air Products and Chemicals, Inc., E.I. DuPont de Nemours & Co., Monsanto Co., Goodyear Tire & Rubber, Akzo Chemicals, Inc., et al., M & T Chemicals Inc., Sherex Chemical So., Inc.Ronald J. Wilson, Robert W. Adler, Jessica C. Landman, Washington, D.C., Sarah Chasis, Natural Resource Defense Council, Inc., New York City, for Natural Resources Defense Council.Lawrence R. Liebesman, Lisa Zannoni, Washington, D.C., for Nat. Paint & Coatings Ass'n.Charles M. Lanier, Timothy E. Kelley, Sheila T. Walet, New Orleans, La., for Rubicon, Inc.Robert Brager, Richard S. Davis, Washington, D.C., Cheryl B. Sikkema, The Dow Chemical Co., Midland, Mich., R. Dean Cooper, Plaquemine, La., for The Dow Chemical Co.James B. Harris, Elizabeth Ann Rabon, Thompson & Knight, Dallas, Tex., for Texas Eastman Co., a div. of Eastman Kodak Co.David M. Flannery, M. Ann Bradley, Kim Brown Poland, Robinson & McElwee, Charleston, W.Va., for Laroche Chemicals, Inc.Robert Brager, Washington, D.C., Susan Kuis, Pittsburgh, Pa., for PPG Industries, Inc.Beveridge & Diamond, Washington, D.C., David Landgraf, FMC Corp., Philadelphia, Pa., for FMC Corp.Dean A. Calland, Donald C. Bluedorn, II, Babst, Calland, Clements & Zomnir, Pittsburgh, Pa., for Koppers Co., Inc.Robert T. Stewart, Kevin R. Murray, Dallas, Tex., for The Lubrizol Corp.R. Kinnan Golemon, Sara M. Burgin, Brown, Maroney, Rose, Barber & Dye, Austin, Tex., for Hoechst Celanese Corp.Samuel O. Buckley, III, Michael A. Chernekoff, New Orleans, La., for Etheyl Corp.William A. Anderson, II, Mary E. Wall, Washington, D.C., for OO Chemicals, Inc.David M. Flannery, M. Ann Bradley, Charleston, W.Va., for Borg-Warner Specialty Chemicals, Inc.Leonard A. Miller, Robert S. Taylor, Swidler & Berlin, Washington, D.C., for Allied-Signal, Inc., and W.R. Grace & Co.Lynn L. Bergeson, Emily R. Paradise, Fox, Weinberg & Bennett, Washington, D.C., for Courtaulds Fibers, Inc.Petitions for Review of an Order of the Environmental Protection Agency.Before RUBIN, GARZA and KING, Circuit Judges.By RUBIN, GARZA and KING, Circuit Judges Jointly.Acting under the mandate of the Clean Water Act (CWA),[fn1] the Environmental Protection Agency (EPA), has promulgated final regulations limiting the discharge of pollutants into the nation's navigable waters by manufacturing plants in the organic chemicals, plastics, and synthetic fibers (OCPSF) industries. The regulations, which the statute requires to be implemented beginning March 31, 1989, cover both direct discharge and indirect discharge through publicly-owned treatment works (POTWs). The Chemical Manufacturers Association (CMA) and a number of companies affected by the regulations allege both procedural defects in their promulgation and substantive defects in various provisions, as well as defects in the application of specific provisions to particular plants. Intervening in some of the cases consolidated for review and appearing as amicus curiae in the others, the Natural Resources Defense Council (NRDC) also challenges the regulations, but on the different ground that they fail to require a sufficiently high degree of effluent pollution control. Although it contends that the regulations are invalid, the NRDC urges that they be enforced until more stringent standards can be adopted.The case is of such complexity that the parties have submitted briefs totalling more than 3,000 pages and a joint appendix 9,000 pages long distilled from a 600,000-page administrative record. To enable us to render a decision as promptly as possible, the members of the panel have divided responsibility for preparing portions of this opinion, as the District of Columbia Circuit did in Alabama Power Co. v. Costle.[fn2] Judge Garza prepared sections V, VI, and VII of this opinion, as well as all portions discussing issues raised by the NRDC; Judge Rubin prepared sections I and III; and Judge King prepared sections II and IV, except for those portions discussing issues raised by the NRDC.This is a summary of our rulings on the principal issues:I. The EPA did not violate the notice-and-comment requirements of the Administrative Procedure Act (APA) by: (1) utilizing an updated Dun & Bradstreet economic-impact study to supplement data it had previously disclosed without making the new data public either during the notice-and-comment period or in the public record except to this court under seal; or (2) by failing to publish its regulations for the control of toxic metals for public comment prior to final promulgation.II. Best Practicable Technology (BPT) Issues:A. The EPA's Consideration of the Costs of Complying with the BPT Limitations:1. The Act does not require the EPA to apply a "knee-of-the-curve" cost-effectiveness test in establishing BPT limitations; the EPA sufficiently considered the total costs of the BPT limitations in relation to the effluent-reduction benefits and this is all that the Act requires.2. The BCT cost-effectiveness test does not replace the BPT cost requirement.3. Although the BPT limitations will double industry's current costs for the removal of conventional pollutants, these costs are not sufficiently high to make the Administrator's decision arbitrary or capricious; identifying the point of diminishing returns is within the discretion of the Administrator.B. The EPA's Definition of the BPT Data Base:1. The EPA did not use unreasonably weak statistical editing criteria to determine the "average of the best" dischargers in defining the BPT data base.2. The EPA's rejection of polishing ponds and multimedia filtration as the BPT model technology was not arbitrary or capricious; the EPA determined that the benefits of polishing ponds relative to their costs were not substantial and that the effectiveness of multimedia filtration was not well demonstrated.C. The "Summer/Winter Issue":The EPA's determination that neither special subcategorization nor establishing special limits was necessary for plants utilizing biological treatment systems in colder climates was not arbitrary or capricious; the record shows that the performances of such plants were comparable to those in other regions and technological modifications are available to account for the limited instances in which cold weather may affect treatment performance.D. BPT Subcategorization:1. The EPA's use, in part, of Standard Industrial Classification (SIC) codes for the purpose of categorizing the OCPSF industry was reasonable; because SIC codes are based on product type, the EPA could reasonably assume that plants producing similar products would have similar wastestreams.2. The NRDC's contention that the EPA failed to provide public notice of its intent to limit the applicability of the regulations to certain SIC codes is without merit; as early as 1985 the EPA defined the OCPSF industries to include all facilities within specified SIC codes.E. Issues Concerning Waste Stabilization Ponds:1. The EPA's determination that pond algae are conventional pollutants is consistent with the Act, which defines a conventional pollutant as a biological oxygen-demanding substance or a suspended solid.2. The EPA's decision not to create a subcategory for plants employing waste stabilization pond treatment systems was reasonable, because the EPA determined that more effective technology was available; the EPA need not create separate subcategories for industry members that install less effective technology than the model technology.3. The EPA adequately considered the compliance costs that will be incurred by plants utilizing pond technology.4. The EPA's suggestion in the final regulations that copper sulfate could be used to control pond algae was in response to the industrial petitioners' comments and thus did not require further notice and comment under the Administrative Procedure Act.5. The costs of the BPT limitations are not wholly disproportionate to its benefits.F. Plant-Specific Challenges to the BPT Limitations and the Availability of Fundamentally-Different-Factor Variances1. The EPA reasonably concluded that the BPT limitations for biological oxygen-demanding substances were achievable for Union Carbide's plants through the use of biological treatment.2. The EPA reasonably concluded that the BPT limitations were achievable for Borg-Warner's plants and others with phenol-dominated wastestreams.3. DuPont's claim that the wastestream of its Chambers Works plant is uniquely complex, precluding compliance with the BPT limitations, is not a basis for invalidating the limitations.4. The EPA's decision not to create a separate subcategory, as urged by Ethyl and Monsanto, for plants with high concentrations of influent total-dissolved solids was reasonable; there was insufficient evidence that any plant has enough total dissolved solids to preclude effective treatment.5. The plant-specific challenges of Union Carbide, Borg-Warner, Dow, Monsanto, and Ethyl that their plants cannot meet the limits are without merit because the Act does not require the EPA to consider allegedly fundamentally different factors of individual plants in promulgating nationwide BPT limitations for an industry; the CWA provides that fundamentally different factors of plants that preclude compliance must be considered in a collateral variance proceeding.III. Best Available Technology (BAT) Issues:A. The EPA's Statistical Methodology:1. The EPA's use of weighted averaging to derive long-term averages was not an abuse of discretion.2. The EPA's averaging of variability factors in deriving effluent limits was not an abuse of discretion.B. Remedy for Unavoidable Exceedances:1. The EPA's decision to exclude extremely high discharges from the calculation of "variability factors" was reasonable, and the "upset defense" is a sufficient remedy for exceedances that result from exceptional circumstances beyond the control of the plant operator.C. The EPA's Analytical Methodology:1. The industrial petitioners failed to demonstrate that the EPA's statistical model did not account for analytical variability.2. This court defers to the EPA's determination that the presence of multiple pollutants in a plant's wastestream did not preclude accurate measurement of those pollutants.3. The EPA employed reasonable criteria for the use of borrowed data when performance data for a particular pollutant were not available.4. The EPA's error, if any, in determining "minimum analytical values" was accounted for by the EPA's conservative methodology which ensured that the limitations are achievable.5. The industrial petitioners were not prejudiced by changes in analytical methods in the development of the OCPSF limitations.D. Use of Minimum Analytical Values for Enforcement Purposes:The petitioners' claim that a non-detect reading should be considered zero for enforcement purposes is not ripe for review because the EPA has not adopted a formal policy on this issue.E. The EPA's Sampling Techniques:1. The industrial petitioners failed to demonstrate that the effluent discharge limits were based on inaccurate EPA sampling.F. The EPA's Application of the Limitations for Toxic Pollutants to All OCPSF Dischargers:1. The EPA did not abuse its discretion by subjecting all OCPSF dischargers to the limits for all toxics.2. The EPA provided adequate public notice that it was considering subjecting all dischargers to the limits for all toxics.3. Even if this is the proper court in which to raise the issue, the EPA's limitations for toxic pollutants do not violate the consent decree in NRDC v. Train.G. BAT Subcategorization:1. The EPA violated the notice and comment requirement of the Administrative Procedure Act by failing to publish for public comments its BAT subcategorization prior to final promulgation of the regulations.H. DuPont's Chambers Works Plant:1. DuPont's assertion that its Chambers Works plant will not be able to comply with the BAT limitations because of the plant's uniquely complex wastestream is not a basis for invalidating the limitations.I. The BAT Limitations for Phenol:1. Borg-Warner failed to demonstrate that the BAT limitations for phenol are unachievable for plants with phenol-dominated wastestreams; the record shows that such plants are capable of meeting the limitations.2. Borg-Warner failed to demonstrate that the EPA's cost estimates were not a reasonable approximation of the costs the industry will have to incur to meet the BAT limitations.J. The BAT Limitations for Volatile Pollutants Based on Steam Stripping Technology:1. PPG and Dow have failed to demonstrate that the BAT limitations for volatile pollutants are not achievable; the record shows that at least one of the plants in the data base was able to meet the limitations for each pollutant.2. Because the EPA determined that the limitations could be met by all plants regardless of their wastestream characteristics, the EPA did not abuse its discretion by failing to further subcategorize the plants subject to the limitations for volatile pollutants on the basis of wastestream characteristics.3. Because the EPA determined that steam-stripper maintenance could be conducted without discharging volatile pollutants, the EPA did not abuse its discretion by failing to make an allowance for such discharge in the limitations.4. Dow Chemical's claim that two of its plants will not be able to meet the limitations for volatile pollutants is not a basis for invalidating the limitations.K. The BAT Limitations for Priority Pollutants Based on In-Plant Biological Treatment:1. CMA has failed to demonstrate that the EPA's reliance on data from end-of-pipe biological treatment systems in deriving the BAT limitations was arbitrary or capricious.2. CMA has failed to demonstrate that the EPA's cost estimates for in-plant biological treatment systems are not a reasonable approximation of the type and cost of the technology that industry will have to utilize to meet the limitations.3. Contrary to CMA's assertion, the record shows that the EPA considered the land costs associated with in-plant biological treatment systems.L. Compliance Deadline:1. The EPA's post-deadline enforcement policy provides an adequate time extension for industry members that will not have sufficient time to install the technology required to meet the limitations.IV. Pretreatment Standards for Existing Sources (PSES) Issues:A. "Pass Through" Issues:1. The EPA properly selected pollutants that pass through POTWs for the pretreatment standards.2. The EPA's decision to define "pass through" based on POTW average removal, rather than actual reported removal, does not violate the CWA.3. The EPA's decision not to find pass through of chromium, copper, and nickel, which contaminate sludge, was not arbitrary or capricious because sludge pollutants will be subject to future rulemaking proceedings.B. The EPA's failure to establish PSES for six volatile pollutants was not arbitrary or capricious because some of these pollutants were sufficiently controlled by the limits for structurally similar compounds and the EPA did not have sufficient POTW removal data to fix limits for the other pollutants.C. Application of PSES to Small Dischargers:1. The EPA's decision not to exempt small indirect dischargers from PSES and not to establish a different standard for these dischargers was reasonable; the record supports the EPA's determination that the economic impact of PSES on these dischargers was not so disproportionate to the impact on the industry as a whole as to require a separate subcategory.2. The Synthetic Organic Chemical Manufacturers Association (SOCMA) failed to demonstrate that the EPA's cost projection for PSES was not reasonable.3. The EPA reasonably rejected SOCMA's proposal that the EPA regulate only small plants producing large discharges; such a plan would leave substantial pollutant discharges unregulated.D. Application of PSES to Paint/Resin Plants:1. The EPA's reliance on data from resin plants in promulgating PSES does not make the standards inapplicable to combined paint/resin plants; the EPA found that the wastestreams of resin plants and paint resin plants were equally treatable.2. The National Paint and Coatings Association has failed to demonstrate that PSES as applied to paint/resin plants are not economically achievable; the EPA identified cost-effective means of segregating the resin stream from such plants thereby making separate regulation of the resin stream possible.3. The EPA's regulation of the resin discharge of combined paint/resin plants does not violate the NRDC v. Train consent decree which exempts paint manufacturers from regulation.4. The regulations make it clear which resin manufacturers are subject to PSES.E. The EPA's Decision Not to Subcategorize on the Basis of POTW Removal Credits:1. The EPA is not required to subcategorize on the basis of POTW removal credits; Congress intended that credits for a POTW's pollutant removal would be available only after the EPA completes its comprehensive sludge regulations.2. Because Congress has suspended the removal credits regulations, this court need not address the claims of Gulf Coast Waste Disposal Authority and the Village of Sauget POTWs that they supplied the EPA with sufficient data to support their claims for removal credits or subcategorization.3. Gulf Coast Waste Disposal Authority is bound by the consent decree in United States v. Crown Central Petroleum Corporation which requires Gulf Coast to recognize and enforce the pretreatment standards.V. New Source Performance Standards (NSPS) Issues:1. The EPA's application of BPT- and BAT-costing methods to NSPS was reasonable.2. The EPA's failure to consider wastestream recycling as a model technology in promulgating NSPS was arbitrary; the record indicates that recycling is a well-demonstrated technology resulting in substantially higher pollutant removal.VI. Montreal Protocol Issue:A. The EPA's failure to consider the economic consequences of the "Montreal Protocol on Substances that Deplete the Ozone Layer" in the cost analysis of the chlorofluorocarbon industry was not arbitrary or capricious, because the issue was never brought to the EPA's attention and at the time the regulations were promulgated the Montreal Protocol was not yet in force.VII. Issues Concerning the Scope of the OCPSF Limitations:1. This court is without jurisdiction to review NRDC's claim that the EPA's reservation of nonconventional pollutants and eight priority pollutants for future rulemaking was contrary to the CWA; such a challenge must be raised in the first instance in the district court.2. The EPA reasonably concluded that the OCPSF limitations apply to research as well as manufacturing discharges.Statutory BackgroundThe purpose of the Clean Water Act (CWA) is to restore and maintain the chemical and biological integrity of the nation's waters.[fn3] It was adopted to effectuate Congress's declared "national goal that the discharge of pollutants into the navigable waters be eliminated by 1985,"[fn4] a date later extended to March 31, 1989,[fn5] and to prohibit "the discharge of toxic pollutants in toxic amounts."[fn6]Until 1972, the federal government relied primarily on state and local action to accomplish federal pollution-abatement goals.[fn7] Congress became dissatisfied, however, with the division of responsibility for setting standards between federal and state water-pollution-control agencies, with the EPA's dilatory pace, and with the ponderous federal enforcement procedure. Consequently, it enacted the Federal Water Pollution Control Act (FWPCA) Amendments of 1972,[fn8] which imposed greater federal regulatory responsibilities and set deadlines for the completion of limitations on pollutant effluents.Because the EPA had failed timely to fulfill its responsibilities under the 1972 Act, the Natural Resources Defense Council (NRDC) filed suit in NRDC v. Train challenging the EPA's failure to promulgate the effluent standards mandated for toxic substances. The consent decree entered into by the EPA, NRDC, and various industrial intervenors in that suit established a schedule for Agency promulgation of effluent limits, new source standards, and pretreatment standards for priority toxic pollutants.[fn9]Recognizing the growing seriousness of the problems created by toxic pollution and the inadequacy of the 1972 FWPCA to deal with them,[fn10] Congress amended that Act by adopting the CWA in 1977.[fn11] As thus amended, the FWPCA contains several distinct, though interlocking, regulatory schemes.First, Title II of the Act encourages the construction of publicly-owned waste-treatment works by providing federal grants-in-aid to states for the construction of such plants.[fn12] As a condition of receiving grants-in-aid, the Act requires states to establish area-wide management agencies with both planning and regulatory functions for waste treatment.The second regulatory feature of the FWPCA, as set out in Title III, authorizes the EPA to set and to enforce federal effluent standards. This part of the 1977 statute codifies the toxics consent decree issued by the United States District Court for the District of Columbia in which it required the EPA to develop BAT effluent guidelines by July 1, 1980 for 65 toxic pollutants listed in the decree. The Act also requires the EPA to promulgate pretreatment standards for indirect dischargers based on BAT or more stringent criteria.[fn13]Third, the Act requires the states to establish water-quality criteria and to set ambient quality standards for each of their rivers, subject to EPA approval. This continues the procedures under prior law pursuant to which the states submitted their ambient water standards for federal approval, submitted their own effluent limitations designed to meet these standards, and had responsibility for enforcement of the limitations.[fn14]To monitor compliance with the pollutant effluent limitations, the Act establishes a system for issuing pollution permits called the National Pollutant Discharge Elimination System (NPDES).[fn15] Each discharger, including POTWs, must obtain a permit from the EPA. No permit may be issued unless the effluent to be discharged meets federal effluent standards.[fn16]As Professor Frank P. Grad has noted in his Treatise on Environmental Law, these four "regulatory schemes ... are limited ... by a variety of general provisions dealing with enforcement, citizen suit, judicial review, and the like."[fn17]Congress again amended portions of the Act in 1987 to extend the compliance dates prescribed in the 1977 Act from July 1, 1984, to March 31, 1989; to permit modification of effluent limitations for certain specified "nonconventional" pollutants; to permit modification of secondary treatment requirements, to alter the guidelines applicable to facilities that are fundamentally different; and to effect other changes. Our references in this opinion to the Act refer to its provisions as finally amended in 1987.The Act requires direct dischargers to comply with technology-based pollutant-effluent limitations that, in time, will become more stringent.[fn18] First, it orders all direct dischargers of conventional pollutants to comply with effluent limitations achievable by application of the "best practicable control technology presently available" (BPT) by July 1, 1977. Second, it orders all direct dischargers of conventional pollutants to comply by March 31, 1989 with effluent limitations based on a more exacting standard, the "best conventional pollution control technology" (BCT). It mandates in addition that, by the same date, direct dischargers of toxic pollutants must comply with the even more rigorous effluent limitations based on the "best available technology economically achievable" (BAT).[fn19]The EPA must determine the BPT, BCT, and BAT requirements and announce them in regulations establishing "effluent limitations guidelines" for various classes and categories of dischargers. In establishing each set of standards Congress required the EPA to consider a number of factors including costs, although the cost factor is accorded less weight for facilities not yet constructed and for discharges more harmful to the environment.New plants constructed after the promulgation of the OCPSF Guidelines that discharge directly into navigable waters are subject to separate standards referred to as "new source performance standards" (NSPS). The new source performance standards are based on the "best available demonstrated control technology" (BADCT) as identified by the EPA.[fn20]Indirect dischargers rely on POTWs to treat their wastewaters, and the Act requires the EPA to set effluent limitations for POTWs engaged in the treatment of municipal sewage or industrial wastewater.[fn21] Although the POTW requirements are determined by separate regulations, they must be based on BAT to ensure that all final dischargers of toxic pollutants meet the same standards. The treatment usually accorded by POTWs, however, may not remove all pollutants discharged into their facilities by industrial users and their operation of these facilities may be damaged by some industrial discharges. Therefore, to regulate the discharge into POTWs of those pollutants determined "not to be susceptible to treatment" by POTWs or likely to "interfere with the operation" of POTWs,[fn22] the Act requires the EPA to establish "pretreatment standards for existing sources" (PSES). These standards must also be based on BAT.[fn23]The OCPSF limitations are technology-based and apply to plants grouped into categories based on their industrial characteristics. The EPA, with the concurrence of the state affected, may establish special provisions for a facility that is fundamentally different with respect to one or more of the factors relevant in developing the regulations other than cost. These provisions, which are known as "fundamentally different factor" (FDF) variances, are intended to adjust the general limitations and provide different ones for a plant whose individual characteristics prevent it from performing within the limits set for its industrial category.[fn24] An FDF variance application may be based on supporting data submitted to the EPA during the rulemaking process or on information that the applicant did not have a reasonable opportunity to submit at that time.[fn25]The regulatory process is not static. Various provisions of the Act require the EPA to review the guidelines periodically and to revise them when appropriate.[fn26]For the past eleven years the EPA has conducted studies and rulemaking proceedings for the purpose of establishing OCPSF pollutant-effluent limitations. The EPA has identified model technologies that in its view satisfy the development criteria for BPT, BAT, NSPS, and PSES. Based on these model technologies, the EPA has determined treatment performances and has established effluent limitations for conventional and toxic pollutants within the range of the performances achieved by the model technologies. The OCPSF limitations provide maximum daily and maximum monthly average limits for the discharge of designated pollutants from each "point source," that is, each discharge pipe.[fn27]Standards of ReviewThe Administrative Procedure Act (APA) fixes the standard for appellate review of agency actions. Agency actions may be set aside only if the "agency action, findings, and conclusions [are] found to be ... arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law"; or "in excess of statutory jurisdiction, authority, or limitations, or short of statutory right"; or "without observance of procedure required by law."[fn28] In making its determinations, "the court shall review the whole record or those parts of it cited by a party, and due account shall be taken of the rule of prejudicial error."[fn29]The many issues raised on appeal concern three aspects of the EPA's actions, each governed by different standards: (1) its rulemaking procedures; (2) its interpretation of the Clean Water Act; and (3) the validity of its regulations and its actions to enforce the Act.The standards of review for each of these types of issues have been stated and restated in a host of cases, not always consistently. The most frequently stated verbal formulae are these:Review of the validity of the challenged procedure is a question of law governed principally by the APA, under which our review is plenary. In determining validity, the Administrator's decision is entitled to a presumption of regularity.[fn30] A party petitioning for review of an agency's regulations bears the burden of overcoming this presumption.[fn31]When we turn to the EPA's interpretation of the statute, our review is again plenary, for we may not accept its interpretation if "contrary to Congress's intentions as revealed by the Act's language, structure, and legislative history."[fn32] In statutory interpretation, the judiciary is "`the final authority and we must reject administrative constructions which are contrary to clear congressional intent.'"[fn33] Nevertheless, we accord some deference to the Agency's interpretation of the statute whose enforcement is entrusted to it if Congressional intention is not pellucid.[fn34] If, therefore, the statute is susceptible to more than one interpretation, we must accept that of the EPA if it is reasonable.[fn35] We need not find that it is the only permissible interpretation, but merely that the "EPA's understanding of this very `complex statute' is a sufficiently rational one to preclude a court from substituting its judgment for that of the EPA."[fn36] According deference to an agency, however, does not imply "rubber stamping" its decision.[fn37]In interpreting the Act, we do not lack precedent. The Supreme Court has twice considered issues involving its construction,[fn38] and eight circuit courts have done so in at least twenty-seven cases.[fn39]After interpreting the statute according to these principles, we must next determine whether the EPA's findings or the regulations based on them are "arbitrary or capricious." In doing so we must conduct a "searching and careful review" of the facts to determine whether "the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment."[fn40] As a reviewing court, however, we must not "substitute [our] judgment for that of the agency,"[fn41] but must "start with the assumption that the agency's action is valid."[fn42] The court's proper function is only to determine whether the agency has "considered the relevant factors and articulated a rational correlation between the facts found and the choice made."[fn43]An agency rule is "arbitrary" 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."[fn44]Because "technological and scientific issues such as those presented in [reviewing effluent limitations] are by their very nature difficult to resolve by traditional principles of judicial decisionmaking,"[fn45] the court "`must look at the [agency's] decision not as the chemist, biologist or statistician that [it is] qualified neither by training nor experience to be, but as a reviewing court exercising ... certain minimal standards of rationality.'"[fn46]Even with regard to complex technical or scientific decisions, however, a reviewing court may not simply defer to an agency's expertise, but must "steep" itself in technical matters sufficiently to determine whether the agency "has exercised reasoned discretion."[fn47] Because judicial review "must be based on something more than trust and faith in EPA's experience,"[fn48] a court may not respond to claims of technical expertise by "rubber stamping" an agency decision as correct.[fn49]These formulae, however, ultimately may prove to be deceptive guides. The Supreme Court's decisions seem to embody two different approaches that are, "analytically in conflict `with the result that a court of appeals must choose the one it deems more appropriate for the case at hand.'"[fn50] In determining the degree of deference appropriate to an agency's decision, the factors to be considered include the fact-law dichotomy, despite the difficulty of drawing a precise line between fact and law;[fn51] whether the question to be decided is one concerning which the courts have a special competence, such as constitutional law, or one that turns on technical expertise of the kind the agency staff possesses;[fn52] whether the issue turns on the agency's interpretation of its own authorizing statute, which entitles it at least to guarded deference;[fn53] and the validity of the reasoning upon which the agency relies to justify its actions.[fn54]Either the application of these formulae or the balancing of factors praised by scholars leaves the reviewing court with a wide margin for decision. Any approach permits different possible decisions on the various issues. Our duty requires us to resolve them with the good judgment that the latitude afforded the court demands. In performing this duty, we recognize our own limitations in assessing policy decisions.[fn55]I. Procedural Challenges1. The EPA's Economic-Impact StudyThe Clean Water Act explicitly requires the EPA to consider economic achievability in formulating pollutant effluent limitations.[fn56] The Chemical Manufacturers Association (CMA) contends that, in preparing its economic-impact study, the EPA departed from its prior practice during the rulemaking proceedings and violated the APA's notice-and-comment requirement by relying on economic data obtained from Dun & Bradstreet that were never made available to the public for comment.The APA requires an agency to set forth in its notice of proposed rulemaking "the terms or substance ... or a description of subjects and issues involved" in the proposed rule.[fn57] The agency must then give interested persons an opportunity to participate in rulemaking through criticism and comments.[fn58] Although the APA does not specifically require notice and comment on the technical data that an agency considers, this court has recognized that fairness requires that the agency afford interested parties an opportunity to challenge the underlying factual data relied on by the agency.[fn59]If the final rules differ from the proposed rules or if new data are considered after the agency receives comments on the data it initially provides, the nature of the change in the proposed rule or in the newly-considered data determines whether it must again publish notice and invite additional comments. As the District of Columbia Circuit noted in Air Transport Association of America v. Civil Aeronautics Board:An agency adopting final rules that differ from its proposed rules is required to renotice when the changes are so major that the original notice did not adequately frame the subjects for discussion.... The agency need not renotice changes that follow logically from or that reasonably develop the rules it proposed originally.[fn60]In that case, the CAB had promulgated a final fee schedule relying on internal staff studies that had not been made available for comment but that had been prompted by prior public comments that had "strongly challenged" calculations in the proposed fee schedule. The court found no notice-and-comment violation because the agency had in its notice of proposed rulemaking both outlined the method that it proposed to use and indicated the type of actions it proposed to take. The District of Columbia Circuit concluded that "[t]hese critical elements of the proposal did not change, and the final rule was a `logical outgrowth' of the proposed rule."[fn61]In Community Nutrition Institute v. Block[fn62] the same court considered the situation in which the agency did not make a change in the proposed rule but considered new and unpublished data after giving notice. The court noted that the Secretary of Agriculture had requested further information on the subject of a proposed rule and that the petitioner in response then pointed out a possible methodological flaw in the studies relied on by the Secretary but did not provide the requested information. The Secretary then relied on two supplemental staff studies that were developed after the close of the comment period to address the alleged flaw. These studies were not made available for comment. The court said: "Rulemaking proceedings would never end if an agency's response to comments must always be made the subject of additional comments."[fn63]In a 1986 notice of availability the EPA announced that it intended to prepare an economic-impact study of the OCPSF pollutant effluent limitations. The EPA stated that in preparing the economic-impact study it would rely on an industrywide "FIN/STAT" data base that covered the period from 1976-81.Industry members commented that the FIN/STAT data base was outdated and did not have adequate data for plants whose sales exceeded $10 million annually. In the preamble to the final rules the EPA announced that, in response to industry comments, "the financial data base used to calculate discounted cash flow and liquidation values for OCPSF plants in the impact analysis was changed from FIN/STAT to Dun & Bradstreet," but it did not reveal the new Dun & Bradstreet data. These data, however, were edited in the same manner as the older Dun & Bradstreet data used in the FIN/STAT study.[fn64] The EPA further stated in the preamble that it used the later Dun & Bradstreet data to increase the size of the entire data base, to increase the number of plants in the "greater than $10 million sales" category (from 4 to 73), and to update the data base to cover the period from 1981 to 1986.[fn65]The new Dun & Bradstreet data were used to estimate plants' cash flow, liquidation value, and profits. Although these data were important in the economic-impact study, the EPA also relied on data that had earlier been obtained from other sources, including material from Robert Morris Associates, COMPUSTAT, Data Resources, Inc., the Bureau of Economic Analysis, the Bureau of Labor Statistics, the International Trade Commission, the Bureau of the Census, chemical-industry periodicals, and surveys submitted to plants under Section 308 of the CWA. The EPA did not supplant its economic-impact study, or replace its original data with completely new and different data, but, in response to industry criticisms, updated and expanded one of several data sources. Indeed, the data it originally relied on were not so scanty that a reasonable rulemaker could not have relied upon them.The 1986 notice adequately advised interested parties of the method the EPA had followed, the financial data it proposed to rely on, and its intention to develop an economic-impact study. The EPA's use of the updated and expanded Dun & Bradstreet data base was a logical and reasonable development based on industry comments and as such did not require further notice and comment.In any event, CMA has failed to demonstrate that it was prejudiced by the EPA's use of the new Dun & Bradstreet data.[fn66] The APA provides that a reviewing court shall take due account of "the rule of prejudicial error."[fn67] The rule applies "when a mistake of the administrative body is one that clearly had no bearing on the procedure used or the substance of the decision reached."[fn68] A petitioner who objects to an agency's failure to publish data for comment must "indicate with `reasonable specificity' what portions of the document it objects to and how it might have responded if given the opportunity."[fn69]The EPA offered to make the new Dun & Bradstreet data available to CMA during the post-rulemaking period on a confidential basis, explaining that it had failed to publish the data because its contract with Dun & Bradstreet required it to keep the data confidential. While a contractual provision of this sort cannot, absent unusual circumstances, relieve an agency of its duty to publish data, the EPA offer did make it possible for CMA to review the data, to comment on it on an "in camera" basis to this court, and to demonstrate that the data were sufficiently inaccurate or misleading to prejudice the affected industries. CMA declined the EPA's offer. In response to direct inquiry by this court, the only prejudice CMA has suggested is that it lacked an opportunity to show that the OCPSF effluent limitations were not economically achievable. Because CMA was fully able to make this showing even without the Dun & Bradstreet data, we fail to discern any substantial prejudice from the EPA's use of the 1981-86 Dun & Bradstreet data to supplement the other information on which it relied. We therefore decline to overturn the regulations because of the EPA's use of undisclosed supplementary economic data from Dun & Bradstreet.2. Limits on Metal-Bearing WastestreamsAppendix A to the effluent limitations establishes limits for the discharge of toxic uncomplexed metals. Appendix B lists "complexed metals," that is, metals bonded with an organic molecule, and states that limits for such pollutants will be established on a case-by-case basis.[fn70] CMA and DuPont contend that Appendices A and B were never presented to the public for comment, this violated the APA's notice and comment requirement, and the limitations in these appendices therefore are unlawful.In a 1985 public notice the EPA proposed BAT limits for all metals found in OCPSF wastestreams. In response to EPA's notice several industry members commented that few OCPSF wastestreams contained metals and that the model technology chosen by EPA to treat toxic metals was not effective in removing complexed metals. In a 1986 public notice the EPA requested additional OCPSF industry information concerning removal of complexed and uncomplexed metals, the source of metals in OCPSF wastestreams, whether each metal is complexed or uncomplexed, and, for complexed metals, a list of the complexing agents.[fn71] In Appendix A to the final rule, the EPA listed industrial processes likely to produce metal-bearing wastestreams and established effluent limitations for the discharge of such metals. In Appendix B the Agency listed the processes that produce complexed metals and stated that such metals would be regulated on a case-by-case basis when individual NPDES permits were issued.[fn72]The APA notice requirement is satisfied if the notice fairly apprises interested persons of the subjects and issues the agency is considering; "the notice need not specifically identify `every precise proposal which [the agency] may ultimately adopt as a final rule.'"[fn73] Though the EPA did not give the industry notice of Appendices A and B, the 1985 and 1986 notices informed the industry that the Agency was considering establishing limitations for complexed and uncomplexed metals, and this was all the APA demands. Moreover, the changes reflected in the final rule were instigated by industry comments and were based on data supplied by the industry; thus the final rule was a logical outgrowth of the comments received.II. Best Practicable Technology (BPT) IssuesThe CWA authorizes the EPA to establish effluent limitations for direct dischargers of conventional pollutants. The Administrator is required to establish effluent limitations for categories or classes of point sources discharging these pollutants based on the "best practicable control technology currently available" (BPT).[fn74] The CWA does not specifically define BPT, but does identify factors that the EPA should consider in determining it.Section 304(b) of the CWA, as amended, states that in assessing BPT, the EPA must consider:the total cost of application of technology in relation to the effluent reduction benefits to be achieved from such application, and shall also take into account the age of equipment and facilities involved, the process employed, the engineering aspects of the application of various types of control techniques, process changes, non-water quality environmental impact ... and such other factors as the Administrator deems appropriate....[fn75]BPT limitations are intended to represent the average of the best levels of performance by existing plants of various sizes, ages, and unit processes within the category or subcategory for control of conventional pollutants.[fn76] In promulgating the regulations, the Agency identified a model technology: biological treatment preceded by appropriate process controls and in-plant treatment followed by secondary clarification as necessary to assure adequate control of solids.[fn77] CMA argues that the EPA's data indicate that its limitations will require the installation of additional treatment equipment at a cost "wholly out of proportion" to the marginal effluent reduction that the equipment would achieve and that the limitations consequently fail to meet the cost-effectiveness test required by Section 304(b)(1)(B) of the Act and the "best conventional technology" (BCT) test enacted in 1977. The EPA asserts that the total cost of the BPT rules is justified by the total amount of pollutant that would be removed.A. The EPA's Consideration of the Industry's Costs of Complying With the BPT LimitationsCMA maintains that the cost-effectiveness of BPT rulemaking should be measured by a "knee-of-the-curve" test to determine the point at which costs rise steeply per pound of pollutant removed and that, under such a test, the BPT rules are not cost-effective.The CWA contains no specific statutory language establishing a BPT "knee-of-the-curve" test or any other quantitative cost-benefit ratio test for BPT. The statute simply requires that the EPA consider "the total cost of application of technology in relation to the effluent reduction benefits to be achieved from such application."[fn78] The courts of appeal have consistently held that Congress intended Section 304(b) to give the EPA broad discretion in considering the cost of pollution abatement in relation to its benefits and to preclude the EPA from giving the cost of compliance primary importance.[fn79]Senator Muskie, the principal Senate sponsor of the Act and the Chairman of the Senate Subcommittee on Air and Water Pollution, stated:The balancing test between total cost and effluent reduction benefits is intended to limit the application of technology only where the additional degree of effluent reduction is wholly out of proportion to the costs of achieving such marginal level of reduction for any class or category of sources.The Conferees agreed upon this limited cost-benefit analysis in order to maintain uniformity within a class and category of point sources subject to effluent limitations, and to avoid imposing on the Administrator any requirement to consider the location of sources within a category or to ascertain water quality impact of effluent controls, or to determine the economic impact of controls on any individual plant in a single community.[fn80]The EPA argues that the Administrator acted well within this broad discretion in concluding that the costs of the OCPSF BPT limitations were justified by the significant quantities of pollutants that would be removed.[fn81] The EPA notes that the OCPSF industry is currently a national leader in discharging conventional pollutants into our nation's waters. The industry has approximately 300 direct dischargers which annually discharge an estimated 61 million pounds of biochemical oxygen-demanding substances (BODS) and 100 million pounds of total suspended solids (TSS) for a total estimate of approximately 161 million pounds annually.[fn82] The EPA estimated that the BPT limitations would result annually in the removal of 108 million pounds of conventional pollutants from OCPSF discharges and consequently from our nation's waters[fn83] at an annualized compliance cost of 76.6 million dollars after a capital investment of 215.8 million dollars.[fn84] Thus, the EPA concluded that the total cost of BPT is warranted by the total pounds of pollutant removed.1. CMA's Challenge Based on the "Knee-of-the-Curve" Cost-Effectiveness TestCMA argues that Congress was concerned generally that the EPA's regulations not require expenditures that would pass the point at which costs escalate rapidly in relation to benefits - the "knee-of-the-curve" on a diagram depicting the cost curve. CMA conceives of the knee-of-the-curve test as a generally applicable cost-effectiveness test with the "knee" defining the most stringent level of regulation permissible. Thus, CMA asserts, whether the EPA labels its regulations BPT or BCT, the EPA is required to consider whether the marginal costs exceed the marginal benefits of the rule. Applying this test, CMA argues that increasing the removal of conventional pollutants from 96 to 99 percent as required by the limitations would cost the OCPSF industry almost twice as much per pound of pollutant removed as current treatment methods: The annual removal of 108 million pounds would cost 76 million dollars per year - 71 cents per pound - whereas industry efforts to date have required an expenditure of only 38 cents per pound. CMA concludes that the cost per pound for removal of pollutants is thus well beyond the knee-of-the-curve and that the regulations are therefore not cost-effective.The EPA argues, however, that even if the knee-of-the-curve test applies to any of its regulations, the test is applicable only to assess the cost-effectiveness of incremental increases in limitations beyond BPT - that is, only to BCT.[fn85] Representative Roberts, the author of the conference report on the 1977 amendments, emphasized that the additional technology requirements of BCT were to be imposed only to remove additional "cheap pounds" of conventional pollutants beyond BPT.[fn86] Congress, however, did not specify that initial BPT must be "cheap." In fact, Congress anticipated that initially BPT might cause many plant closures and the loss of 50,000 to 125,000 jobs.[fn87]The BCT provisions were intended to establish an intermediate level between BPT and the stricter BAT limitations for conventional pollutants by adding a cost-effectiveness test for incremental technology requirements that exceed BPT technology.[fn88] Under BCT, additional limitations on conventional pollutants that are more stringent than BPT can be imposed only "to the extent that the increased cost of treatment [would] be reasonable in terms of the degree of environmental benefits."[fn89]Thus, Congress intended that cost would occupy a different role in EPA's promulgation of BPT limitations than it would in the promulgation of BCT because of the different aims of the two standards. While Congress did not consider cost to be irrelevant to BPT, it clearly intended it to be a less significant factor than in the promulgation of BCT limitations. The EPA's interpretation of the Act is rational and supported by both the legislative history and the case law insofar as the EPA emphasizes that the BPT limitations are not subject to the type of stringent cost-benefit analysis required for BCT. The relevant inquiry with respect to BPT, as indicated above, is whether the costs are "wholly disproportionate" to the benefits.To the extent that CMA's claim is that "wholly disproportionate" is to be measured by a knee-of-the-curve test, the EPA responds that CMA misconceives the nature of the test. Rather than displaying the rate at which costs increase relative to pounds of pollutant removed, CMA's curve displays the rate at which the cost-perpound increases relative to the percent of pollutant removed, resulting in a misleadingly steep curve. While both the BPT and BCT tests require a comparison between costs and effluent reduction,[fn90] neither test requires the comparison of costs to the percentage removed, as implied by CMA's curve.CMA relies upon legislative history to justify its percent-removal approach to evaluating the reasonableness of costs. However, the statute does not require that a percent-removal approach be used to establish BPT regulations.[fn91] Almost all of the BPT regulations promulgated by the EPA since the 1972 enactment of the Clean Water Act are based upon either concentration limitations (as in the case of the OCPSF rule) or more stringent "mass limitations" which limit both concentrations and flow volumes.[fn92] In the current case, the regulation will require a 10% increase above current industry costs to remove 108 million additional pounds.[fn93] The EPA reasonably concluded that these costs were not "wholly disproportionate" to the benefits.2. The BCT Cost-Effectiveness TestCMA also argues that whether or not BPT rules are, as a general matter, subject to a knee-of-the-curve test, the EPA's BPT limitations for conventional pollutants must pass the BCT cost test which Congress enacted in 1977.[fn94]In promulgating BCT limitations, the Act directs the EPA to consider:the reasonableness of the relationship between the costs of attaining a reduction in effluents and the effluent reduction benefits derived, and the comparison of the cost and level of reduction of such pollutants from the discharge from publicly owned treatment works to the cost and level of reduction of such pollutants from a class or category of industrial sources....[fn95]CMA contends that this test governs the BPT rules because they represent an increase in regulation over the limitations established on a case-by-case basis by NPDES permits issued before 1977. In other words, CMA contends that the permit limitations established BPT for individual plants and that in enacting the BCT requirements in 1977 Congress intended that any subsequent, more stringent regulations must be evaluated according to the BCT standards.The EPA responds, however, that its authority to promulgate BPT regulations is not abrogated by the fact that, pursuant to Section 402(a)(1),[fn96] NPDES permits were issued prior to the promulgation of industry-wide BPT regulations. The EPA notes that, since 1977, it has promulgated BPT regulations limiting conventional pollutants in the iron and steel, metal finishing, coal mining, oil and gas, battery manufacturing, plastics molding and forming, metal molding and casting, coil coating, porcelain enameling, aluminum forming, copper forming, electrical and electronic products, and nonferrous metals forming industries[fn97] - notwithstanding the fact that most of these facilities had previously been regulated by permits. The oil-and-gas-pollutant effluent limitations were promulgated in 1979 and reviewed by this court in 1981 without any reference to the BCT cost test.[fn98]The EPA also maintains that Congress did not intend BCT to displace BPT. The EPA notes that Congress has never repealed the BPT factors as a vital and continuing requirement of the Act[fn99] and has not stripped the EPA of its explicit authority, under Section 304(b) of the Act, to revise or update BPT periodically. Section 304(b) directs the EPA to "publish ... regulations, providing guidelines for effluent limitations, and at least annually thereafter, revise, if appropriate, such regulations."[fn100] Thus, as the EPA interprets the Act, BCT standards, which place cost-effectiveness constraints on incremental technology requirements that exceed BPT technology, do not displace BPT or override the EPA's authority to promulgate BPT for conventional pollutants.As additional evidence that Congress enacted BCT to supplement, rather than to replace, BPT, the EPA points to the fact that, ten years after the enactment of BCT, Congress enacted a "stricter BPT" provision "requiring a level of control substantially greater or based on fundamentally different control technology" for BPT regulations promulgated after 1981.[fn101] This applies to all BPT regulations for all pollutants, including conventional pollutants, without limitation.As evidenced by numerous rulemakings, the EPA has consistently interpreted the Act to allow the promulgation of BPT limitations applicable to facilities operating under NPDES permits despite the enactment of BCT standards in 1977. We must accord "considerable weight" to an agency's construction of a statutory scheme it is entrusted to administer.[fn102] Finding the EPA's interpretation of the Act to be reasonable, we conclude that CMA's objections do not compel us to remand the limitations.3. The Cost of Compliance with the BPT LimitationsFinally, having concluded that the EPA construed the statute reasonably in declining to subject the BPT limitations to the BCT cost test, we find that the Administrator did not act arbitrarily and capriciously in determining that the costs of the limitations were justified by the significant amount of pollutants that would be removed.Although the cost per pound of 71 cents required to meet the BPT limitations is almost double the 38 cents per pound that the OCPSF industry presently spends to remove conventional pollutants, the 71-cents-per-pound figure is not so high as to make the EPA's decision arbitrary and capricious:[fn103] "The selection of the point of diminishing returns is a matter for agency determination."[fn104]B. The EPA's Definition of the BPT Data BaseNRDC challenges the EPA's BPT limitations on BODS and TSS as being too lenient. Specifically, NRDC argues that the BPT limitations are not based on the average of the best dischargers, and that the EPA improperly rejected sequential treatment options. These two arguments will now be more fully examined.1. The EPA's Determination of the "Average of the Best" DischargersThe EPA is required to promulgate BPT regulations based on the "average of the best" performers in the industry.[fn105] We accord some deference to the EPA's interpretation of its controlling statute; therefore, if the statute is capable of more than one reasonable interpretation, we must accept the EPA's interpretation, if reasonable.[fn106]NRDC claims that the EPA used data from 71 of 99 plants, approximately 72%, as representing the group of "best dischargers" for purposes of promulgating BPT regulations. How can the group of "best" dischargers encompass 72% of the industry, queries NRDC, leading it to argue that the EPA should have further tightened its editing criteria, which would have led to more stringent regulations.NRDC's argument is misleading, however. Out of 304 direct dischargers in the industry that will be subject to regulation, the EPA chose a particular technology, namely biological treatment with secondary clarification, which is used by 99 plants. Out of these 99 plants, the EPA then chose data from 71 plants to determine the "average of the best" for the purpose of promulgating its BODS regulations. The EPA defends its decision by noting that its initial edit reduced the field from 304 to 99. Thus, the NRDC's complaint that the EPA used 71 of 99 plants is mistaken because the edit in question was the second edit, 205 dischargers having already been weeded out.We hold, therefore, that the EPA's class of performers for determining the "average of the best" was not unreasonably broad.Another question is whether the CWA requires the EPA to consider the average of the best performers within an industrial category, or the average of the best performers that use a particular chosen technology within an industry. We hold that it is appropriate to extract a group of "best" performers from an industry category; this was done in this case when the EPA selected 99 out of 304 plants in its initial edit. In fact, the EPA went further by narrowing the 99 plants down to 71. The EPA was not required to take the average of the best 99 plants using a particular technology, but merely to take the average of the best 304 plants in the industrial category. The legislative history of the 1972 amendments to the CWA specifies that "[t]he administrator should establish the range of `best practicable' levels based upon the average of the best performance by plants of various sizes, ages, and unit processes within each industrial category."[fn107] Therefore, we find that the EPA's methods for setting the BPT standards for BODS were in compliance with the CWA.Lastly, NRDC argues that the EPA's editing criteria for its representative "best performers," which were 95% biological oxygen demand (BOD) removal and 40 mg/l concentration, are actually below the industry's median. In support of this contention NRDC quotes from the preamble to the final rule, which it interprets as saying that the industry median is 95.8% removal and 29 mg/l concentration. Again, NRDC mischaracterizes the statistics. The median 95.8% removal and 29 mg/l concentration statistics apply to the 99 plants using biological treatment methods, not the 304 plants comprising the industry.[fn108]2. The EPA's Rejection of Sequential Treatment OptionsNRDC argues that the EPA, in choosing biological treatment plus clarification (Option I) for BPT, improperly rejected other sequential options. Specifically, these options were polishing ponds (Option II) and multimedia filtration (Option III), either of which can be sequentially added to a system already utilizing biological treatment plus clarification.[fn109] NRDC makes specific arguments with regard to both technologies; these arguments will next be further explored.NRDC objects on several grounds to the EPA's failure to base BPT on Option II technology, which is biological treatment with clarification followed by polishing ponds. First, NRDC argues that the EPA's conclusion that polishing ponds are not currently used by a representative portion of the industry is incorrect. In support of this, NRDC points out that 64 plants in the industry use polishing ponds, and, of those, 17 meet the EPA's final editing criteria.[fn110] The EPA defends its conclusion by initially noting that the OCPSF industry is diverse and as a result plants have significantly different wastestream characteristics. Since only 17 plants out of 64 which had installed polishing ponds were able to meet the BPT editing criteria, the EPA in its judgment concluded that polishing ponds are not a sufficiently effective technology upon which to base the BPT limits. Moreover, the question is not merely whether the fact that 17 out of 64 plants using polishing ponds were able to meet the BPT editing criteria proves that polishing ponds are effective, but also whether the experiences of these 17 plants can be applied to the 304 plants subject to regulation. The EPA concluded that this was not possible due to the diverse nature of the industry.[fn111] The EPA also balanced the questionable feasibility of using polishing ponds against the cost of acquiring the large amounts of land needed for them and concluded that feasibility had not been sufficiently demonstrated to serve as a basis for BPT limitations.[fn112] We find the EPA's judgment in this regard to be rational and well supported by the record.NRDC next argues that the EPA's position that the majority of plants using polishing ponds do not meet the BPT criteria shows that the industry is inadequate as a whole. The NRDC quotes from the preamble to the final rule where the EPA noted that "plants using polishing ponds in the OCPSF industry have done so not to add another treatment step after effective Option I level biological treatment but rather to improve upon substandard biological treatment."[fn113] The EPA responds that this quote is taken out of context. Again noting the diverse nature of the industry to be regulated, the EPA stresses that the plants did not use ponds to remedy inadequate performance which can be attributed to the model technology, but rather because these particular plants had unique and complex wastestreams which required additional treatment.NRDC further alleges that the EPA's analysis of the data for Option II plants is flawed because the EPA compared the performance of Option II plants to all plants in the industry, which includes plants using Options I, II, and III. Thus, the EPA essentially compared plants using Option II against a class which also contained Option II plants; this tended to improve the performance of the latter class, masking the true benefits of ponds. NRDC contends that the EPA should have compared Option II plants to all other plants which do not use Option II. The EPA admits that NRDC's suggested approach is the better one, but notes that it corrected its error in the preamble to the final rule by comparing Option II plants to Option I plants.[fn114] The EPA corrected this error in promulgating the final rule, and the corrected data support the EPA's conclusion.NRDC claims, in similar fashion to its above argument regarding polishing ponds, that the EPA improperly rejected Option III as a BPT sequential treatment option. NRDC claims that multimedia filtration results in significant improvements in TSS removal, but the EPA cites data which indicate that NRDC's claim is unfounded.[fn115] Specifically, the EPA rejected filtration as unproven because, of the 11 of 28 plants with filtration that satisfied the EPA's editing criteria, all 11 had other processes in addition to filters, and there was no evidence that the plants would not have satisfied the BPT criteria even without filters.[fn116] Due to the diversity of the OCPSF industry, the EPA determined that the data from 11 plants did not demonstrate the effectiveness of filters throughout the industry, and we defer to the EPA's determination.C. The "Summer/Winter Issue": The EPA's Decision Not to Subcategorize Based on Climatic DifferencesPetitioners Goodyear, DuPont, M & T, Monsanto, and Union Carbide (hereinafter collectively "Goodyear") challenge the BPT limitations for failure to account for cold-temperature effects. Goodyear argues that because cold temperatures adversely affect biological treatment, the EPA should have created a separate class for cold-temperature plants, conducted a technical assessment and cost analysis for this class, and calculated the incremental cost-effectiveness of bringing this class into compliance.The EPA used biological treatment with secondary clarification as the model technology for formulating the BPT limitations.[fn117] Biological treatment involves treating wastewater with microorganisms that biodegrade and sometimes absorb the organic pollutants. Goodyear contends that the microorganisms used in biological treatment are adversely affected by cold temperatures, thus affecting the efficacy of biological treatment. Therefore, Goodyear argues, many plants in the northern and middle latitudes will not be able to comply with the BPT limitations even if they employ the model technology. Goodyear argues that these cold-temperature plants should be grouped into a separate subcategory and subjected to less stringent BPT limitations.The EPA decided not to provide separate summer/winter limitations or to subcategorize the OCPSF industry based on temperature. This decision is clearly supported by the record. The EPA determined that the best plants in the industry could comply with the BPT limitations throughout the year. The EPA also conducted a detailed analysis which demonstrated that the plants operating in the coldest temperatures and with the greatest seasonal temperature fluctuations dealt with these problems more effectively than southern plants, indicating that attention to appropriate design and operation features overcomes any temperature-related difficulties in treatment. The EPA concluded that cold temperatures had, at most, an insignificant effect upon the treatment of wastewater at plants that took appropriate measures. Finally, the EPA also considered the incremental costs of ensuring effective treatment during the winter and increased its cost estimates to reflect such treatment using a temperature-correction factor.1. Diversity of the Data BaseThe BPT regulations for the OCPSF industry are based upon a broad and geographically diverse population of plants. The EPA used data from 71 plants to establish the long-term averages, and data from 21 plants to establish the variability factors. Together, these data formed the basis for the BPT limitations.[fn118] The northern, southern, and middle latitude regions are all well represented in both sets of data.[fn119] Furthermore, the data from these plants were collected throughout the year and thus represent a level of treatment performance that may be obtained under all seasonal conditions.[fn120] Additionally, this large data base represents a wide range of production operations. The size and diversity of the data base ensures that the BPT limitations based upon this data can generally be achieved across the industry, regardless of climate.[fn121] This basis alone is sufficient to support the limitations.[fn122]Petitioners object to the EPA's exclusion of seven plants that had permits allowing compliance with less stringent limitations during the winter months from the set of 21 plants used to develop the variability factors. The EPA fully explained its decision to exclude these plants:[B]ecause these plants were subject to meeting two different sets of permit limits, they had no incentive to attempt to achieve uniform limitations throughout the year. Not surprisingly, then, the daily data from these plants exhibit a two-tier pattern. These data can be characterized by two means, and the variability of these data over a 12 month period is fundamentally different from the data from plants required to meet only one set of permit limits. Consequently, the data generated during these periods is not representative of well-operated biological treatment, which as noted above is capable of uniform treatment throughout the year as demonstrated by a number of plants.[fn123]These plants were in fact included in the 71-plant data base used to develop long-term averages. They were excluded only from the variability analysis.[fn124] Therefore, their lower average performance during the winter is reflected in the BPT limitations. It is reasonable for the EPA to choose not to rely on plants that have no regulatory incentive to achieve uniform performance, particularly where an ample data base of plants meeting uniform year-round permit limitations was available.[fn125] The exclusion of the data from the seven plants operating under special permits was adequately explained and is a determination "peculiarly within EPA's expertise." The decision is therefore entitled to judicial deference.[fn126] Moreover, the EPA did not selectively exclude from the variability data base those plants that performed worse in the winter than in the summer.[fn127] Thus, any difference between winter and summer variability at the best plants was incorporated into the variability factors used to develop the limitations. There is ample support in the record for the EPA's conclusion that lower removal efficiencies were the result primarily of factors other than cold temperature and that the BPT limitations are achievable in all climates on a year-round basis.[fn128] Goodyear's contentions to the contrary are therefore without merit.2. Winter Removal EfficienciesGoodyear also argues that the EPA was required to establish a separate subcategory for plants that could not achieve the BPT limitations due to cold-temperature effects. As we held above, however, the record supports the EPA's conclusion that cold temperatures alone did not have a significant effect on plants' abilities to meet the BPT limitations. Thus, the EPA could reasonably conclude that the effects of temperature on treatment effectiveness did not provide a basis for subcategorization. Goodyear contends that the EPA did not have sufficient data indicating that technology was available to overcome the effects of temperature. The EPA, however, did identify numerous practices that were used successfully by OCPSF plants to improve their winter performance.[fn129] The EPA is not required to analyze all possible causes of each plant's non-compliance with BPT and to recommend design and operating practices for each plant.[fn130] The EPA concluded reasonably that the limitations could be met during all seasons and identified several means of doing so.Finally, Goodyear argues that the EPA failed to consider whether the technology required to overcome the effects of temperature is cost-effective. The EPA is required by statute to consider costs in relation to benefits only for categories or subcategories of plants, not for individual plants.[fn131] As we held above, the EPA was not required to create a subcategory consisting of plants that had difficulty complying with the limitations due to temperature effects. There is no record basis for isolating the effect of temperature on treatment effectiveness from the effects of other, more significant, factors.[fn132] Consequently, as the EPA notes correctly, there is no class of plants for which temperature-related costs can be estimated and assessed against pollutant reduction benefits.We hold, therefore, that the Administrator concluded rationally that the BPT limitations are practicable for all plants in all seasons.D. BPT Subcategorization and the EPA's Use of Standard Industrial Classification (SIC) CodesCMA and NRDC both argue that the EPA's classification scheme in development of BPT limits for the OCPSF industry was improper because it was based in part upon Standard Industrial Classification (SIC) codes. There are two strands to this argument. CMA argues that there is no significant correlation between SIC codes and effluent levels, thus rendering the SIC-code-based classification scheme arbitrary. NRDC takes a different tack, arguing that EPA did not provide for sufficient public notice and comment regarding the proposed system. Based on the analysis below, we reject both arguments and uphold the EPA's SIC-code-derived classification.1. CMA's Challenge to the BPT SubcategorizationCMA challenges the EPA's division of the OCPSF Industry into seven subcategories for the purpose of establishing BPT limitations on the ground that the subcategories are based on SIC-product groupings rather than wastewater characteristics or treatability. CMA argues that the EPA has created an inequity by grouping together plants with substantially different influent concentrations and subjecting them to the same concentration limitations, thereby requiring, in effect, a higher percent removal at plants with higher influent concentrations.In Section 304(b)(1)(B), Congress listed several factors, in addition to cost, that the EPA shall "take into account" in determining BPT, including the age of equipment, the process employed, engineering aspects of treatment, process changes, and non-water-quality environmental impacts.[fn133] Based on these factors, the EPA determines whether plants within an industry should be assigned to a subcategory subject to more particularized regulations than the industry as a whole.However, the EPA is required to create a separate subcategory for a group of plants only when they are so fundamentally different from other plants on which the limitations are based that they cannot practicably achieve the effluent limitations achieved by the average of the best plants in the industry.[fn134] The EPA has considerable discretion in evaluating these factors; it is enough that the EPA considered the relevant factors and reached a rational conclusion about them.[fn135] The Agency's task is "to establish numerical standards limiting effluent pollution," and it should concentrate "on grouping plants that could meet the same limitations."[fn136] If plants can meet the same limitation, they need not be subcategorized simply because they are different.[fn137]The subcategories for the OCPSF industry are defined by generic types of OCPSF production. The EPA maintains that since a plant's chemical processes are closely related to the type of production in which it engages, the subcategorization bears a reasonable relationship to such chemical processes and the wastewater characteristics associated with those processes.[fn138]In making its decision to subcategorize the OCPSF industry for the purpose of its BPT regulations, the EPA analyzed several potential subcategorization factors, such as age of plant, temperature, and SIC-code classification of the plant, just to name a few. CMA disputes whether SIC codes actually have a relationship with BODS effluent concentration. They present a statistical analysis purportedly showing that SIC codes have no significant correlation with BODS effluent. CMA analyzed several categorization schemes, including the one retained by the EPA, and determined the value of R, the coefficient of determination, for each.[fn139] CMA argues that the categorization scheme which was retained by the EPA has an R of only 0.052, which is not significantly better than other rejected classification schemes. In fact, a random shuffling performed by CMA of BODS data with respect to SIC codes yielded an R of 0.030 in the CMA analysis. This, argues CMA, supports its position that the classification scheme adopted by the EPA is arbitrary and capricious, since it is only slightly better than a random categorization scheme. CMA also criticizes the EPA's addition of allegedly irrelevant dummy variables to the EPA's R analysis of its own model, the alleged effect of which was to artificially inflate the R value.[fn140]We are of the opinion that the R analysis presented by CMA is inconclusive. Though an R analysis can be informative, it cannot of itself conclusively prove or disprove the adequacy of a particular categorization scheme. A good model may nonetheless have a small R value if the spread of data points along the x-axis (horizontal axis) is small.[fn141] Thus, the fact that the R value attributed to the SIC-code-based classification is relatively small (0.052) is an insufficient basis for us to conclude that the EPA's industrial classification is arbitrary and capricious. Moreover, the coefficient of determination, R, is merely a limited indicator of model adequacy. Ultimately, we believe the EPA's explanation that SIC codes tend to be organized around the products produced by various segments of the industry, and that the type of product in turn influences the wastestream characteristics of those plants, to be a sufficient rationale upon which to uphold the EPA's classification scheme.[fn142]CMA further argues that analysis of its proposed subcategorization scheme based on influent BODS concentrations yields an R of 0.42, which is higher than the R value of the EPA's SIC-code-based plan (although not higher than the EPA analysis when the dummy variables are included). Therefore, argues CMA, the EPA should have subcategorized the OCPSF industry based on influent levels, as opposed to SIC codes. Again, we find this argument to be insufficient for the following reasons. First, it should be remembered that a large R value does not conclusively prove that the variables are causally related.[fn143] Second, CMA provides a graph of its data points showing influent levels (x-axis) versus effluent levels (y-axis). There are two remote points in x-space on this graph, both with large influent and effluent levels. Data plots with such remote points tend to have much higher R values even though the model is not necessarily superior, since those points exert a greater influence on the slope of the regression line than the points which are clumped together nearer the origin.[fn144] Moreover, if these points are bad values (due to error), their deletion may reduce the resulting value of R. Although these points may just as well be valid measurements, this determination requires the exercise of discretion on the part of the analyst. It is unclear how CMA exercised its editing discretion in developing the model they presented, and in any event, the limited nature of CMA's regression analysis precludes us from mandating that EPA adopt CMA's model.2. NRDC's Notice and Comment ChallengeNRDC argues that the EPA failed to provide notice of its intent to limit the applicability of the OCPSF regulations to certain SIC codes. NRDC complains that, as a result of this lack of notice, it was denied the opportunity to comment on this part of the regulation.We find NRDC's argument to be lacking in merit. During the rulemaking, EPA invoked SIC codes to define which facilities would be covered by the final rule.[fn145] Our review of the record persuades us that NRDC was fairly apprised of the issues at stake, and that the EPA's decision to limit the rules to the five SIC codes identified in its original notice was a logical outgrowth of the rulemaking proceeding.E. Issues Concerning Waste-Stabilization PondsPetitioners DuPont, Texas Eastman, Union Carbide, and Air Products incorporate the use of waste stablilization ponds ("WSP systems") for the treatment of industrial wastewater. WSP systems consist of tanks where organic matter in wastewater is broken down through bacterial action. During the treatment process, algae form which increase the TSS in the effluent. Petitioners raise several claims regarding the application of the BPT limits to plants employing WSP systems. First, petitioners contend that the EPA acted arbitrarily and capriciously in determining that pond algae are a conventional pollutant subject to the BPT limitations for BODS[fn146] and TSS.[fn147] Second, petitioners contend that the EPA violated the Act by failing adequately to account for algae-related problems in setting BPT limits for BODS and TSS. Third, petitioners argue that the EPA acted arbitrarily and capriciously in specifying copper sulfate treatment as a means of controlling algae without adequate notice and comment on the practicality of that treatment technology.We find that the EPA's implicit regulation of algae as a component of BODS and TSS had a rational basis and that the EPA was not required to create a separate subcategory for plants utilizing WSP systems. These four plants perform the same types of operations as other OCPSF plants and generate the same types of OCPSF process wastewaters and are thus properly subject to the same BPT limitations. Finally, we conclude that the EPA did not act arbitrarily and capriciously in designating copper sulfate treatment as a means of controlling algae.1. EPA's Determination that Pond Algae Were Conventional Pollutants and Thus Subject to BPT RegulationsPetitioners argue that the EPA should not have included algae as a component of BODS and TSS because algae do not contribute to the environmental problems associated with BODS and TSS. We find, however, that the EPA's regulation of algae as a component of BODS and TSS had a rational basis.The EPA notes correctly that Congress, without limiting the definition of the terms, includes BODS and TSS in the broad definition of conventional pollutants.[fn148] The EPA concludes that because algae are both oxygen-demanding organisms and suspended solids, algae are encompassed within both BODS and TSS as those terms are used in the Act, and the EPA is thus authorized to include algae in the measurement of both TSS and BODS. The EPA also notes that Congress has specifically recognized that algae are a significant cause of water quality problems.[fn149]In the OCPSF rulemaking, the EPA found that algae can indeed present significant water-quality problems:The control of algae growth in ponds, lakes, and reservoirs can be a serious problem in water quality management. Among the nuisances created by the often sudden blooming of one or more algal genera are: odors and tastes; fish kills; poisoned water fowl; shortened filter runs in water-purification plants; growths in pipes and other water conduits; and interference with industrial water uses. Therefore, proper control of algae growth in ponds is necessary to avoid the potential nuisances in the ponds and/or receiving waters.[fn150]While petitioners take issue with some of the EPA's specific conclusions regarding the effects of algae on water quality, they have failed to demonstrate that the Administrator acted arbitrarily and capriciously in regulating algae as a conventional pollutant. The EPA has authority to regulate the discharge of a pollutant even where its effects are subject to some uncertainty[fn151] We will not substitute our judgment for the Administrator's in this regard.2. The EPA's Decision Not to Create a Subcategory for Plants Utilizing Waste-Stabilization-PondsPetitioners argue next that the EPA violated Section 304(b)(1)(B) of the Act by failing to adequately account for algae-related problems in setting BPT limits for BODS and TSS. As explained more fully above, Section 304(b)(1)(B) requires the EPA to "take into account" several factors, in addition to cost, in determining BPT, including the age of equipment and facilities, the process employed, the engineering aspects of the application of various types of control technologies, process changes, and non-water quality environmental impacts.[fn152] Petitioners assert that based on those considerations, the EPA should have created a separate subcategory for plants utilizing WSP systems, subject to less stringent BPT limitations on BODs and TSS than those applicable to the rest of the industry.Petitioners first claim that the EPA failed to consider the extent to which high effluent TSS concentrations are a function of the type of wastewater in the treatment system. The EPA maintains, however, that petitioners' plants do not differ materially from other OCPSF plants with respect to manufacturing operations or wastewater characteristics. Rather, the EPA argues, their plants differ only with respect to the type of treatment petitioners voluntarily chose to employ. Petitioners did not choose to employ the type of biological treatment system that is most commonly used by good performers in the OCPSF industry and that was used by the EPA for cost-estimating purposes: activated sludge, followed by secondary clarification.[fn153] Instead, petitioners chose to employ WSP treatment systems which generate additional pollutants in the form of algae. The EPA asserts that it therefore reasonably required petitioners' plants to meet the limitations demonstrated to be achievable by the average of the 71 "best" plants in the OCPSF industry,[fn154] rather than providing less stringent limitations for these plants simply because they chose to employ less effective treatment technology.The EPA asserts that, as required by the Act, it considered in detail "the engineering aspects of the application of various types of control techniques" and used this factor to identify the data base from which it derived the BPT limitations. Those limitations properly reflected the results achieved by more effective treatment systems than those employed by petitioners. Petitioners cannot now be excused from meeting the BPT limits on the grounds that their present method of treatment is less effective than the average of the best - in this case, because WSP systems create additional pollutants. Such a construction of Section 304(b)(1)(B) would defeat the very purpose of establishing BPT limits. The EPA is not required to base BPT upon a technology that is less effective than the best practicable technology demonstrated in the industry.[fn155] Rather, in establishing BPT, the EPA "must forbid the level of effluent produced by the most pollution-prone segment of the industry, that segment not measuring up to `the average of the best existing performance.'"[fn156] We therefore hold that the EPA was entirely reasonable in declining to create a subcategory for plants employing WSP treatment systems.3. The EPA's Consideration of the Cost of Compliance for Plants Utilizing Pond TechnologyPetitioners argue next that the EPA's failure to consider waste treatability resulted in the designation of unworkable upgrades to achieve BPT. The EPA does not seriously dispute petitioners' claim that the algae growth associated with petitioners' present treatment systems may not be remedied by relatively simple and inexpensive upgrades. Rather, the EPA notes that it "is entitled to look at costs on an industry-wide basis as opposed to plant-by-plant," and may use industry averages to develop "rough estimates" to help the Agency determine whether the cost is "wholly out of proportion" to the benefit.[fn157]The EPA concedes that petitioners may be required to install entirely new treatment units consisting of activated sludge and secondary clarification. However, the EPA notes that it estimated the costs of such steps for nearly half of the plants in the industry requiring treatment improvements to comply with BPT.[fn158] Thus, even if Texas Eastman, DuPont, and Air Products are required to install new activated-sludge systems in order to comply with BPT, the costs of these systems would be within the range generally estimated for the industry as a whole. We have already held that the costs of the BPT limits for the OCPSF industry are not "wholly out of proportion" to the benefits. Because petitioners' compliance problems are not attributable to their production process, raw wastewaters, or other relevant factors, but only to their treatment systems, their costs in constructing and operating a biological system complying with the regulations will not differ materially from those incurred by many other industry members.[fn159] We therefore agree with the EPA that even this worst-case scenario does not provide a basis for exempting petitioners from the BPT limits that apply to the rest of the industry.4. Copper-Sulfate Treatment for Algae ControlFinally, petitioners contend that although the EPA proposed copper sulfate as a means of controlling algae at a reasonable cost, this method was not considered in the rulemaking and would in fact impair the efficiency of their ponds and possibly result in excess discharges of copper, thereby causing water-quality problems.While petitioners do raise legitimate concerns regarding the practicability of employing copper-sulfate treatment as a means of complying with BPT, the EPA did not rely on copper-sulfate treatment alone in estimating petitioners' compliance costs.The EPA's cost estimates for Union Carbide, Texas Eastman, and DuPont were based on far more substantial treatment than the application of copper sulfate. For Union Carbide, the EPA costed an entirely new activated sludge system, which generates no algae. For both Texas Eastman and DuPont, the EPA estimated the costs of installing an additional chemically-assisted clarifier to enhance solids removal, and, for Texas Eastman, the EPA also costed an upgrade to its biological unit to enhance BODS removal.[fn160]Again, while the EPA concedes that the treatment methods costed for Texas Eastman and DuPont may not in fact enable their WSP treatment systems to achieve the BPT limits, the EPA reasonably concluded that there is no basis on which petitioners may be excluded from the BPT rules.Petitioners also argue that the BPT limitations should be remanded because the cost of copper sulfate as an algae treatment was not specifically discussed in the EPA's proposals. However, the EPA notes that the control of algae by using copper sulfate was not the basis for BPT and was considered as part of a subsidiary analysis that specifically responded to petitioners' comments.[fn161]Thus, to the extent that the costing of copper-sulfate treatment for algae played a role in the rulemaking proceeding, it was a "logical outgrowth" of the proceedings in general and of petitioners' own comments in particular.[fn162]5. The Cost of the BPT LimitationsPetitioners argue that the cost of BPT, even apart from the issue of algae-control costs, is wholly disproportionate to its benefits. We held above that the cost of BPT was not wholly disproportionate to its benefits and therefore need not further address petitioner's claims to the contrary.F. Plant-Specific Challenges to the BPT Limitations and the Availability of Fundamentally-Different-Factor VariancesSeveral petitioners, including Union Carbide, Borg-Warner, DuPont, Monsanto, and Ethyl, claim that the BPT limitations are arbitrary because the EPA failed to account adequately for wastestream characteristics. Petitioners assert that the wastestream characteristics of certain of their plants preclude compliance with the OCPSF industry limitations. These claims are considered together because several of these petitioners raise highly individualized objections to the limitations.We will address each petitioner's claim in turn. As an initial matter, however, we note that the EPA is not required to consider fundamentally different factors of particular plants in the national BPT rulemaking. Both Congress and the Supreme Court have expressed concern that the process of formulating nationally applicable water-quality standards would be unduly impeded by requiring EPA to address the idiosyncracies of individual plants in the context of a national rulemaking. The Supreme Court has held that the fundamentally-different-factors (FDF) variance procedure provides an entirely acceptable alternative to subcategorizing an industry to account for plant-specific characteristics.[fn163] Congress has codified the FDF procedures in the CWA, encouraging the EPA not to complicate and delay unduly the promulgation of national effluent-limitation guidelines and standards where the FDF procedure could be employed to address the concerns of individual facilities claiming to be unique.[fn164]The Supreme Court held in 1977 that the EPA may establish categorical BPT limitations, "so long as some allowance is made for variations in individual plants, as EPA has done by including a variance clause in its 1977 limitations."[fn165] Relying heavily on legislative history that demonstrated Congress' intent to replace the site-specific approach to water-quality regulation with technology-based limitations that apply uniformly to categories of dischargers,[fn166] the unanimous Court reasoned that the alternative view "would place an impossible burden on EPA" contrary to the legislative purpose of the Act.[fn167]Addressing the EPA's identical FDF variance procedure for pretreatment standards, the Supreme Court has approved the procedure "as a mechanism for insuring that [EPA's] necessarily rough-hewn categories do not unfairly burden atypical plants."[fn168] The Court explained:EPA and CMA point out that the availability of FDF variances makes bearable the enormous burden faced by EPA in promulgating categories of sources and setting effluent limitations. Acting under stringent timetables, EPA must collect and analyze large amounts of technical information concerning complex industrial categories. Understandably, EPA may not be apprised of and will fail to consider unique factors applicable to atypical plants during the categorical rulemaking process, and it is thus important that EPA's nationally binding categorical pretreatment standards for indirect dischargers be tempered with the flexibility that the FDF variance mechanism offers....[fn169]The Court stated that the FDF variance procedure was authorized by Congress in significant part to ensure that the national rule would not be overturned simply because of the Agency's failure to consider unique plants.[fn170] Several courts of appeal have subsequently relied upon the availability of an FDF variance procedure as the basis for rejecting challenges to BPT regulations that are based upon allegedly facility-specific factors.[fn171]In codifying the FDF variance procedure in the CWA,[fn172] Congress specifically emphasized that the procedure serves as a "safety valve" to the categorical statutory scheme, allowing EPA to address plant-specific variations through a separate administrative process, outside of the national rulemaking. The House Report stated:There are two approaches for responding to a facility with valid grounds for arguing that it is fundamentally different from other facilities in its category. One possibility is to develop a separate subcategory within the regulation, undertake a separate data collection and analysis effort and then repropose and issue the final rule. The other alternative is to leave the national rule in place and use the FDF determination procedure to establish alternative technology-based limitations for the facility that accurately reflect its situation. The subcategorization approach would add further complications and require potentially substantial additional time in developing what are already extraordinarily complex and detailed national regulations. By contrast, the FDF determination procedure allows both implementation of the national rule and consideration of individual petitions claiming unique factors.[fn173]Given Congress' clear intent that the national rulemaking process not be unduly impeded by highly individualized objections to the regulations that would be more appropriately addressed in an FDF proceeding, we address the following claims with this concern in mind.1. Union CarbideUnion Carbide argues that high influent BODS concentrations at its Taft, Louisiana, plant preclude effective treatment by the model BPT technology. Union Carbide thus contends that the EPA has neither costed nor identified any BPT technology which will enable plants with high-BOD influent to comply with BPT. The EPA costed the addition of a biological treatment unit followed by a secondary clarifier to Taft's existing treatment system.[fn174] This additional treatment system will remove almost four million pounds of conventional pollutants at an annualized cost of $1,242,200.[fn175]The EPA reasonably concluded that the Taft plant was not so fundamentally different from the industry as a whole as to warrant exclusion from the rule simply because the Taft plant had the highest levels of BODS influent in the data base used to develop the limits. The EPA has concluded reasonably that the limitations are both achievable and practicable for the Taft plant.2. Borg-WarnerBorg-Warner seeks a separate subcategory for wastestreams containing significant amounts of phenol, alleging that high phenol concentrations adversely impact biological treatment by inhibiting biodegradability. The EPA specifically considered the effects of phenol concentrations and found that several plants with high phenol influents achieved low levels of biological oxygen demand.[fn176] The EPA found that phenol-dominated wastestreams were treatable and therefore declined to create a subcategory for phenol-dominated wastestreams. The EPA's conclusion is supported by the record and is not arbitrary or capricious.3. DuPont's Chambers Works PlantDuPont claims that the BPT limits for TSS are not attainable by its Chambers Works plant through the use of the technology relied upon in establishing the regulations. DuPont maintains that the EPA failed to take into account the nature of the wastewaters subject to regulation and specifically failed to take into account the fact that "[d]ue to the complexity and unique aspects of its manufacturing process" the wastestream of the Chamber Works plant contains TSS and mixed liquor suspended solids at significantly higher levels than did the wastestreams of the plants in the EPA's data base. DuPont argues that its wastestream is also fundamentally different from other plants in that its TSS/BODS ratio is nearly four times greater than the average.While DuPont asserts that the EPA failed in general to adequately consider wastestream characteristics in its subcategorization of the industry, it does not propose any specific basis on which the EPA should have created a subcategory that would address the characteristics of the Chambers Works plant wastestream.[fn177] Rather, DuPont asserts that the limitations should be set aside because the EPA failed to take into account "unique" characteristics of the wastewater at the Chambers Works plant and rejected data submitted by DuPont that demonstrate the "unachievability" of the TSS limitations.As we note above, the EPA is not obligated to address in its national rulemaking the fundamentally different characteristics of an individual plant; such concerns are appropriately raised in an FDF variance proceeding.[fn178] DuPont has filed for an FDF variance, and the claims raised here are more properly addressed in the first instance in that forum. Accordingly, we express no opinion on whether such a variance would be appropriate.4. Monsanto and EthylPetitioners Monsanto and Ethyl similarly claim that the EPA failed to adequately account for wastewater characteristics in setting BPT limits and that the EPA's subcategorization scheme is therefore unlawful. Monsanto claims specifically that plants using BPT technology may not be able to achieve the EPA's TSS limits where their wastewaters contain high total-dissolved-solids (TDS) levels and that the EPA should have created a separate subcategory for plants that have high levels of TDS in their wastewater. Monsanto maintains that two of its plants will be unable to comply with the TSS limits for this reason. Similarly, Ethyl asserts that its Elgin plant will be unable to comply with the TSS limits because the plant's wastestream contains high levels of brine. Ethyl claims that dilution is necessary to treat wastewater with high brine content but that the BPT effluent limits are based on process flow only, without allowing for dilution. Ethyl notes that it may be possible to meet the standards based on effluent only but that the EPA has not identified the appropriate technology for doing so and therefore has not accounted for the cost of such technology in its BPT limitations.In response to these objections, the EPA asserts that Ethyl failed to submit any comments during the rule-making proceeding that would establish that a plant's TDS levels would preclude compliance or greatly increase the cost of compliance and that only one company in the industry, Monsanto, claimed during the rulemaking to have a compliance problem caused by high TDS levels in its wastestream.[fn179] The EPA states that it therefore declined to create a subcategory based on TDS levels because there was not sufficient information in the record to demonstrate that any plant in the OCPSF industry could not comply with the TSS limits as a result of elevated TDS levels.The EPA notes that while CMA also commented on the relationship between TDS and TSS, its comments weighed against the creation of a separate subcategory based on TDS influent levels.[fn180] The EPA determined that technology does exist to clarify wastestreams with especially high levels of solids and that facilities use a variety of methods to ensure the effective biological treatment of unique wastestreams that contain pollutants that impede biological treatments. The EPA noted, for example, that technologies such as reverse osmosis can eliminate materials in a plant's wastewater which may inhibit or upset biological treatment systems.[fn181]In specific response to Monsanto's comment, the EPA stated that three facilities which have TDS levels exceeding 5,000 mg/l have nevertheless achieved good TSS removal.[fn182] The EPA further concluded that few, if any, OCPSF plants have TDS levels of sufficient magnitude to impair TSS removal.[fn183] Accordingly, the EPA rejected Monsanto's request that a correction factor for high TDS levels be incorporated into the final TSS limits.[fn184]Although Monsanto takes issue with the EPA's responses, we conclude that the EPA's decision not to establish a special subcategory based upon TDS levels was reasonable. The EPA found, based upon the record before it, that it was uncertain at best whether any plant in the industry had TDS levels that precluded effective treatment and that if there were any such problems, they would be unique to Monsanto, the only company to claim TDS problems. Therefore, Monsanto's concerns would be more properly addressed through an FDF variance proceeding than through the national rulemaking.The EPA notes that, unlike Monsanto, Ethyl never submitted comments - in response to either the proposed regulations or the three subsequent public notices - to inform the EPA that it believed its Elgin plant would experience TSS compliance problems as a result of the level of TDS in the plant's wastewater. Ethyl asserts that it does not suggest that the EPA should have created a separate subcategory for its Elgin plant, but rather submits the argument that the Elgin plant cannot meet the TSS limits as evidence that the EPA failed adequately to consider wastewater characteristics in establishing BPT. To the extent that Ethyl challenges the overall validity of the EPA's subcategorization approach, based on the EPA's alleged failure to adequately consider wastestream characteristics, that issue is addressed above. Furthermore, as we also conclude above, the fact that a single plant may have difficulty in meeting BPT requirements due to unique characteristics of that plant, does not render the entire rulemaking invalid. To the extent that Ethyl does raise concerns unique to its Elgin plant, those issues are properly raised through an FDF variance proceeding rather than through the national rulemaking.5. FDF VariancesThe highly individualized claims of DuPont, Monsanto, and Ethyl are more appropriately addressed in an FDF administrative proceeding. DuPont and Ethyl have filed for variances, and Monsanto may certainly do so in the near future. Petitioners argue, however, that we may not decline to address their claims because the EPA has not yet ruled on the applications. Thus, petitioners claim that the EPA will leave them in "administrative limbo" while the limitations go into effect. To the extent, however, that petitioners DuPont and Ethyl seek to compel an Agency decision on their FDF applications, such relief is outside the scope of this court's limited jurisdiction under CWA Section 509(b)(1) to review the EPA's effluent limitations guidelines, pretreatment standards, and new source performance standards.[fn185]This court would have jurisdiction to review the FDF claims only after the EPA has ruled on petitioners' applications and after review by the district court.[fn186] The Act's regulatory scheme is consistent with the prudential doctrine of primary jurisdiction which holds that complex scientific and technical issues, such as those presented here, should be resolved in the first instance by the EPA, the entity best suited to pass on these issues.[fn187]Congress was aware of the difficulty and corresponding delays in processing FDF variance applications for individual plants.[fn188] Accordingly, Congress attempted to expedite the process by specifically requiring the EPA to determine the merits of applications for FDF variances for individual plants within 180 days of the submission of the application.[fn189] Thus, the WQA ensures that all FDF applications on which the EPA had not previously ruled would henceforth be subject to the Act's 180-day time limit.Notwithstanding Congress' 180-day deadline, the EPA admits that historically it has taken, on average, three years to process an FDF application. The EPA believes that it may require more than 180 days to complete the review of these highly technical applications. However, the fact that the Agency has exceeded the statutory time limit for issuing its decision on the OCPSF FDF applications does not permit this court to order the EPA to produce a schedule for rendering its decisions. Congress provided an explicit statutory deadline for these decisions which specifically contemplates that efforts to compel timely Agency action would be heard exclusively in the district courts.[fn190] Nevertheless, petitioners argue that because of this delay, and because the OCPSF limitations will require them to install costly control technology, this court should stay application of the OCPSF limitations pending the EPA's consideration of their FDF applications.Section 509(b)(1)[fn191] of the CWA authorizes the courts of appeal to review the effluent-pollution limitations promulgated by EPA. Only the rulemaking proceedings are subject to this court's review under Section 509(b)(1), however, and the petitioners do not challenge the promulgation of the regulations. Rather, they challenge the implementation or application of the regulations and such a challenge is not subject to this court's review under Section 509(b)(1). The industrial petitioners' claim that the EPA has failed to consider their FDF applications in a timely manner, therefore, even if true, does not undermine the legality of the regulations because an FDF proceeding is collateral to the rulemaking proceedings.A challenge to the implementation or application of the regulations may be brought in a civil action under CWA Section 505(a)(2), which provides:any citizen may commence a civil action ... against the Administrator where there is alleged a failure of the Administrator to perform any act or duty under this Act which is not discretionary with the Administrator.[fn192]A civil action under Section 505(a)(2) must be commenced in the district court.In addition, the relief requested by the petitioners is precluded by the Act. The Act provides:An application for an alternative requirement under this subsection [i.e., an FDF application], shall not stay the applicant's obligation to comply with the effluent limitation guideline or categorical pretreatment standard which is the subject of the application.[fn193]Staying the regulations pending petitioners' FDF applications would be contrary to Congress' expressed intent that the effluent limits be enforced notwithstanding a pending FDF application.Finally, the petitioners argue that denial of a stay will require them to install millions of dollars of technology that may prove to be redundant if they are later granted FDF variances. This seems to be the possible result of the Congressional pattern in requiring enforcement of the regulations before final action on applications for FDF variances. The impact of this requirement is, however, lessened by the likelihood that, even if each of the petitioners is eventually granted an FDF variance, each will likely be required to meet standards that require the installation of some new methods of pollution control. An FDF variance would not exempt petitioners from the limitations; it would merely subject them to less stringent limits using the best technology.[fn194]III. Best Available Technology (BAT) IssuesSection 301(b)(2) of the Clean Water Act[fn195] requires the EPA to establish effluent limitations for toxic substances ("toxics") based on the "best available technology economically achievable" (BAT). Congress intended these limitations to be based on the performance of the single best-performing plant in an industrial field.[fn196] The BAT standards promulgated by the EPA separate regulated plants into two subcategories and fix different limitations for each of these subcategories. Subcategory 1 (BAT) consists of plants that utilize end-of-pipe biological-treatment systems, that is, systems in which biological processes are used to treat wastewater after all plant processes have been completed and immediately before discharge of the effluent into a navigable stream. Subcategory 2 (BAT) consists of plants that remove pollutants while the wastestream is still in the plant; in-plant methods include biological treatment, steam stripping for volatile pollutants, activated carbon absorption for organic pollutants, chemical precipitation for metals, and alkaline chlorination for cyanide.[fn197]A. The EPA's Statistical Method of Developing the BAT Limitations1. The EPA's Use of Weighted Averaging in Deriving the Long-Term AveragesTo derive the BAT limitations the EPA averaged over a long term the amount of the toxic discharged by what it considered to be the best plant or plants in the data base using BAT technology. It then multiplied this "long-term average" by a "variability factor," a number always greater than one, to account for its estimate of the reasonable variation from the average that could be expected by the several best-performing plants in the data base. The product was the BAT limitation for that toxic.A toxic may be present in a discharge in such a small amount that it cannot be detected by state-of-the-art analytical methods, that is, the amount of a toxic may be less than the "analytical minimum." Generally, the analytical minimum for the detection of a toxic is 10 parts per billion (ppb), a proportion that, following an analogy employed by CMA at oral argument, is comparable to one minute in two thousand years. For the purpose of calculating long-term averages the EPA assumed that the amount of an undetected substance contained in a discharge (the "non-detect value") equalled the analytical minimum; in other words, if the toxic might be present, but only in such a small concentration that it could not be detected by currently-used analytical methods, the EPA assumed the toxic was present at the analytical minimum, generally 10 ppb. None of the petitioners objects to the EPA's use of non-detect values in the calculation of long-term averages, presumably because assigning the analytical minimum to a reading in which a specific toxic could not be detected had the effect of raising the long-term average and thus raising the effluent limits. CMA objected to weighted averaging of such values, however, because this lowered the limitations.The long-term averages were calculated by averaging detectable values and non-detect values. The EPA calculated the average of the assigned non-detect values and the average of detectable values for the data base and then weighted the averages according to the proportions of all non-detect and detectable values reported for plants in the data base for a given pollutant. The long-term average for a particular plant was calculated as the weighted sum of the averages of detect and non-detect values reported for plants in the data base.[fn198]CMA contends that the EPA's weighted averaging of non-detect values significantly lowered the long-term averages it calculated from measured effluent concentrations, and thus lowered the OCPSF effluent limits for at least 27 pollutants. CMA concludes that this was without "justification or support," hence arbitrary.The choice of statistical methods is committed to the sound discretion of the Administrator.[fn199] Weighted averaging is a recognized statistical method for adjusting a data set when the data for particular units, plants in this case, are not represented by a comparable number of readings or samples. When some units in a data set are represented by fewer readings than other units in the set, the units represented by fewer readings should be given less weight in the average, i.e., discounted through weighted averaging. Conversely, those units in a data set based on a larger number of readings should be given greater weight in the average. Including a unit for which there is a smaller number of readings in a calculation, without discounting it by weighted averaging, would distort the average.[fn200]CMA correctly contends that use of a different averaging method would yield different results. Indeed, several different methods might have been used, each producing a different result. As the First Circuit has noted, however, "[t]he choice of any given method may mean that an alternative method would yield different results. The necessary corollary ... is that any other system chosen would be open to the same criticism. We will not leave the Agency so vulnerable."[fn201] We agree, provided, of course, that there is no demonstration that the Agency's method was chosen arbitrarily or merely for the purpose of achieving a predetermined prejudicial effect. The EPA's use of weighted averaging to determine the long-term averages was not an abuse of discretion.2. Averaging of Variability FactorsThe same plant using the same treatment method to remove the same toxic does not always achieve the same result. Tests conducted one day may show a different concentration of the same toxic than are shown by the same test the next day. This variability may be due to the inherent inaccuracy of analytical testing, i.e., "analytical variability," or to routine fluctuations in a plant's treatment performance.The EPA attempted to take this variability into account. To do so, it calculated the BAT limitations by multiplying the long-term average concentration of each pollutant by a variability factor that reflected observed variations in treatment performance experienced by plants that had attempted to remove that pollutant. The EPA computed the variability factor for each plant in the BAT data base for that pollutant or, if there were two or more plants with a significant number of data points in the detectable range, it averaged the variability factors for all the plants.CMA asserts that averaging variability factors from different plants assumes that the plants are similar, an assumption that is not supported by the record, and that the EPA has failed to show that the different plants can reduce the amount of variability to the factor computed by the EPA. CMA further contends that the EPA should have created separate subcategories to account for the different variability experienced by the various types of plants using BAT.As the EPA notes, averaging variability factors inures to the benefit of the industrial petitioners because it yields a greater variability factor than the factor that would have resulted from using data from the single plant that experienced the least variability. The EPA decided not to subcategorize BAT plants on the basis of variability because the EPA determined that OCPSF plants utilizing BAT could achieve uniformly high levels of removal of toxics.[fn202]The reasonableness of the variability factors used by the EPA is supported by the record.[fn203] Moreover, CMA has failed to demonstrate that the greater variability measured at some plants is due to uncontrollable factors rather than plant inefficiencies or that the fact of such greater variability shows that the limitations are not achievable.B. Remedy for Unavoidable ExceedancesPetitioners PPG and Dow assert that the EPA's statistical model demonstrates that a well-operated plant using BAT can be expected to perform within the daily effluent limitations only 99 percent of the time and within the monthly effluent limitations only 95 percent of the time. They contend that the regulations are unachievable, hence arbitrary, because the model implies that exceedances, i.e., toxic discharges in excess of the BAT limitations, can be expected to occur an average of one percent of the time on a daily basis and five percent of the time on a monthly basis. The regulations, they assert, do not provide a remedy for such unavoidable exceedances.This argument is based on an apparent misunderstanding of the EPA's statistical model. The EPA developed two variability factors for each pollutant - a maximum daily variability factor and a maximum monthly variability factor. To develop the variability factors, the EPA fit the data for each pollutant at each plant to a statistical distribution curve that graphed discharge levels (represented on the horizontal axis) against the probability of the particular discharge occurring (represented on the vertical axis). For the daily data distributions, the EPA estimated both the mean and the level below which 99 percent of the discharge measurements fell for each pollutant at each plant; for the monthly data, the EPA estimated both the mean and the level below which 95 percent of the discharge measurements fell for each pollutant at each plant. The EPA then divided the 99th and 95th percentile figures by the respective means of daily and monthly distributions to determine plant-specific daily and monthly variability factors.[fn204] Finally, the EPA averaged plant-specific variability factors for all of the plants in the data base to determine a daily and monthly variability factor for each pollutant.[fn205]The data points reflecting discharge levels in excess of the 99th and 95th percentiles represent extreme departures from the mean and were not used by the EPA in calculating the variability factors. The industrial petitioners apparently argue that, because the model technology produced these data points and because these points were not used in determining the variability factors, the limitations are not achievable using the model technology.The EPA did not use the data points in excess of the 99th and the 95th percentiles in calculating the variability factors because it thought that these extremes were due to plant quality-control problems, not unavoidable exceedances. It determined that these variations could be controlled by quality-control methods such as frequent inspection and repair of equipment, use of back-up systems, operator training and performance evaluations, management control, careful communications and coordination among production and wastewater treatment personnel, spill diversion and holding systems, and equalization basins to make wastewater flow and quality more uniform. The EPA also determined that "use of these techniques should result in compliance at all times apart from instances of upsets."[fn206]The industrial petitioners argue that the EPA has not adequately demonstrated that the data points exceeding the 99th and 95th percentiles represent controllable rather than uncontrollable variability. The EPA's conclusion that these data points result from quality-control problems is, however, reasonable because these points are isolated and extreme departures from average performance. "The purpose of these variability factors is to account for routine fluctuations that occur in plant operation, not to allow for poor performance."[fn207] The data points exceeding the 99th and 95th percentiles, by definition extreme, do not reflect routine performance, and were reasonably excluded.In any event, the EPA has provided an exception for unavoidable exceedances. The regulations provide that an unavoidable exceedance caused by an "upset" is an affirmative defense in an action for noncompliance.[fn208] The regulations define an "upset" as:an exceptional incident in which there is unintentional and temporary noncompliance with technology-based permit effluent limitations because of factors beyond the reasonable control of the permittee. An upset does not include noncompliance to the extent caused by operational error, improperly designed treatment facilities, inadequate treatment facilities, lack of preventive maintenance, or careless or improper operation.[fn209]It is not clear whether upsets were included in the data used by the EPA to calculate the variability factors. At one point the record states that upsets were edited from the data; however, the record also states that the EPA used all data collected from the data-base plants.[fn210] We need not resolve this discrepancy. The EPA reasonably concluded that the data points exceeding the 99th and 95th percentiles represent either quality-control problems or upsets because there can be no other explanation for these isolated and extremely high discharges. If these data points result from quality-control problems, the exceedances they represent are within the control of the plant. If, however, the data points represent exceedances beyond the control of the industry, the upset defense is available.C. The EPA's Analytical Methodology1. Analytical VariabilityThe analytical methods for measuring pollutants become unreliable at the low concentrations the EPA has established in the limitations. Two different laboratories, each using acceptable methods, may measure the pollutant in a given sample and reach different results, yet neither lab may be demonstrably wrong. This has been referred to above as "analytical variability." Monsanto, Dupont, and CMA contend that the BAT limits cannot be achieved because analytical variability will in some instances inevitably result in a detection reading (or laboratory measurement) in excess of the limitations even though the "true" value of the sample is within the limits.The EPA responds by asserting that analytical variability was adequately accounted for in its derivation of the BAT limitations. As previously discussed, the variability factors were derived from empirically-observed variations in measurements due either to analytical variability or to routine fluctuations in a plant's performance. To calculate the BAT limits, the EPA multiplied the long-term averages by the variability factor, and because it did so, the EPA asserts, its statistical model necessarily accounts for analytical variability.The record shows, and the EPA concedes, that certain pollutants are subject to substantial analytical variability. In review proceedings, however, the burden is on the petitioner to show that the Administrator's determination was arbitrary, capricious, or an abuse of discretion.[fn211] The industrial petitioners have not shown that the variability factors the EPA assigned to the respective pollutants failed to account for analytical variability, and therefore the petitioners have failed to meet their burden.2. Complex WastestreamsMany of the OCPSF plants use a variety of processes to produce a number of different organic chemicals or plastics, and, as a result, the wastestreams from these plants include several pollutants, making the wastestreams "complex." Monsanto and Dow contend that current analytical techniques cannot reliably measure low pollutant concentrations in a complex wastestream due to interference from other chemicals and that, therefore, the low effluent levels required by the BAT limitations cannot be achieved by their plants.The EPA has developed comprehensive "Guidelines Establishing Test Procedures for the Analysis of Pollutants," known as "Part 136."[fn212] The EPA has determined "minimum levels" at which these analytical techniques can reliably measure the concentration of a pollutant without interference from other pollutants through a calibration process by which the known concentration of each pollutant is analyzed. The EPA has determined that the limitations are all well above the minimum levels for reliable detection.[fn213]The EPA purports to have accounted in this fashion for the problems of analytical uncertainty and interference from other compounds. Monsanto's and Dow's contentions amount to the argument that their studies are contrary to the EPA's and that their studies show the BAT limitations to be within a range that will be affected by analytical uncertainty and interference. This issue turns on a question of analytical chemistry that this court could not resolve without performing independent laboratory work, and that is beyond both its power and its technical capacity. When reviewing an agency's scientific determinations in an area within the agency's technical expertise, a reviewing court must be at its most deferential.[fn214] We defer to the EPA's determination of this issue.3. Computation of Variability Factors When Data Were Insufficient ("Borrowed Data")When the EPA had not obtained sufficient plant variability data to calculate the variability factor for a pollutant, it used the average of those variability factors that it had established for other similar pollutants or, if no such average had been calculated, for all of the pollutants subject to the BAT limitations. CMA contends that the EPA's use of this "borrowed data" to determine the variability factor for a pollutant when data were not otherwise available was "sheer guesswork."For the BAT limitations, when data were not otherwise available, the EPA determined the variability factor by using data from pollutants exhibiting similar "chemical structure and characteristics" that bear on treatability. This was based on the reasonable assumption that similar compounds react similarly to similar treatment. For BAT, when direct data were not available, the EPA computed variability factors by averaging the variability factors for all other pollutants that the EPA found to be treatable by the same technology. The EPA has established a reasonable basis for its use of borrowed data.[fn215] Its assumptions are logical and there is nothing in the record to establish that they led to scientifically inaccurate results.4. CMA's Challenge to the EPA's Choice of the Minimum Analytical Value to Assign to Non-Detect ReadingsTo calculate the long-term averages for plants in the data base the EPA included non-detect values. As previously discussed, the value assigned to a non-detect reading was the minimum analytical value. To derive the limitations for the polynuclear aromatics, the EPA assigned to these pollutants the published minimum analytical value of 10 ppb that appears in Part 136.[fn216]The BAT limitations for the polynuclear aromatics were based on data from plant 1293T.[fn217] CMA notes that the EPA lab reported an average minimum analytical value of 13.3 ppb for this plant. CMA asserts that the published Part 136 minimum analytical values were determined based on pollutant samples in "pure reagent water" free of other interfering chemicals and do not account for real-world interference from other compounds that raise the minimum analytical value, such as the reported minimum analytical value at plant 1293T. CMA argues that the EPA therefore erred by using the published minimum analytical value, rather than the lab-reported average minimum analytical value for plant 1293T, and this error of 3.3 ppb made the limitations more stringent.The EPA's methodology for deriving the effluent limitations included several conservative assumptions that created a margin for error and ensured that the limitations are achievable.[fn218] The first of these conservative assumptions is that when a data base plant reported a non-detect reading, the EPA assigned the minimum analytical value to such a reading.[fn219] By assigning the minimum analytical value, rather than zero, to plants that reported non-detect readings, the EPA raised each plant's long-term average above the "true" treatment level achieved by the plant. This resulted in higher, less stringent, limitations than would have been imposed if the EPA had assigned a zero to non-detect readings.[fn220] Second, in calculating the variability factors for each pollutant, the EPA did not use data from plants that consistently reported non-detect readings.[fn221] Consequently, data from the best performing plants were not used to derive the BAT variability factors, thus avoiding higher BAT limitations. The EPA's error, if any, in using the published Part 136 minimum analytical value, rather than the wastestream-specific value reported for plant 1293T, was, therefore, offset by the EPA's conservative methodology used to develop the BAT limitations.The record shows that the BAT limitations based on data from plant 1293T are achievable.[fn222] The lowest of the effluent limitations based on the performance of this plant is well above the minimum analytical level whether 10 ppb or 13.3 ppb is used. For instance, the BAT maximum daily limitation for benzo(k)fluoranthene is 47 ppb, while the maximum monthly average is 19 ppb.[fn223] The data show that plant 1293T performed within these limits whether the minimum analytical value is 10 ppb or 13.3 ppb.[fn224]CMA also asserts that at least one plant reported minimum analytical values of 100 ppb for toluene and chloroethane and 60 ppb for chlorobenzene. CMA contends that the EPA erred by assigning the Part 136 minimum analytical value of 10 ppb to toluene and chlorobenzene and 50 ppb to chloroethane in promulgating the limitations.The EPA asserts that the reported minimum analytical values of 100 ppb for toluene and chloroethane and 60 ppb for chlorobenzene were not representative of the data as a whole and that these data should have been excluded from its calculation of the limitations. The EPA attributed these extremely high minimum analytical values at one plant to dilution due to interference from other chemicals. The EPA's data-editing criterion was to exclude data from plants whose samples could not be measured down to the minimum levels listed in Part 136. Thus the data associated with the extremely high analytical minimums identified by CMA should have been excluded. If these data had been excluded the limitations would have been the same, based on the remaining data, because in calculating the limitations the EPA relied on the median performance of the best plants and the vast majority of the readings for toluene, chlorobenzene, and chloroethane were non-detect readings with minimum analytical values equal to those assigned in Part 136.[fn225] Because the limitations based on the other data would have been the same even if the EPA had excluded the objectionable data, the EPA's error in including these data in the data base was harmless.5. Changes in Analytical MethodsThe EPA's analytical methods for determining pollutant concentrations were developed during the rulemaking period. Part 136 was later amended to provide newer, more accurate analytical methods. Monsanto argues that applying one analytical method for the development of the OCPSF limitations and another more accurate standard for enforcement produced anomalous results and that, therefore, the limitations are arbitrary.Current Part 136 outlines several approved analytical methods reflecting both old and new methods for measuring pollutants. An industry member that believes it can record lower measured pollutant-effluent values using the older methods is permitted to do so.[fn226] Because Part 136 provides industry members with a choice of several analytical methods for reporting purposes, the industrial petitioners were not prejudiced by changes in analytical methods in the development of the limitations.D. Use of Minimum Analytical Values for Enforcement PurposesDuPont and Hardwicke assert that the BAT limitations for some toxics are very close to their minimum analytical values. The petitioners argue that a non-detect reading should be considered zero for enforcement purposes.Although the EPA states in its brief that, to be consistent with its approach in calculating the BAT limitations, "it would be reasonable" to treat non-detect values as the analytical minimum, it is not clear that the Agency has adopted this position as its policy, for neither the regulations nor the development documents address it.The issue presented is therefore not ripe for review. In Abbott Laboratories v. Gardner,[fn227] the Supreme Court stated that the doctrine of ripeness is intended to prevent the courts from entangling themselves in premature abstract discussions of administrative policies, and "to protect agencies from judicial interference until an administrative decision has been formalized and its effects felt in a concrete way by the challenging parties." To further these purposes, the Supreme Court held that whether a controversy is "ripe" requires a court "to evaluate both the fitness of the issues for judicial decision and the hardship to the parties of withholding court consideration."[fn228]Because the EPA has not formally adopted a position, our discussion of the validity of what it might do would be merely advisory. Withholding court consideration will not result in substantial hardship to the petitioners because enforcement proceedings have not taken place. If there is any merit to their argument, the issue can be addressed in enforcement proceedings. Finally, "the possibility that the petitioner may have to make capital budgeting decisions under a cloud of uncertainty" does not amount to the requisite hardship when the petitioners do not face sanctions for non-compliance.[fn229]E. The EPA's Sampling TechniquesCMA asserts that in developing the BAT data base the EPA used sampling points that contained non-OCPSF process water, and thus the resulting limitations are based on lower concentrations due to dilution. In its petition for review CMA identifies five plants where it claims samples were taken at points including non-OCPSF flow. CMA contends that the limitations should be adjusted to account for this dilution.To accurately measure BAT-treatment performance wastestream samples must be taken before (influent sampling) and after (effluent sampling) BAT treatment. If dilution occurs after influent sampling, but before effluent sampling, then the effluent sample represents dilution as well as treatment performance and the long-term average is inflated by dilution. If, however, dilution occurs before influent sampling, the dilution has no effect on the measure of treatment performance, provided the influent to the treatment system contains a sufficient "threshold" concentration of a pollutant.[fn230]The EPA states that to sample influent it established sampling points only at locations where OCPSF process wastewater flows were not later significantly diluted by non-process flows.[fn231] The EPA determined that the dilution at the five plants identified by CMA occurred before the BAT treatment stage and before the points at which influent samples were taken. The record also shows that the influents at these five plants contained sufficient toxics to meet the "threshold" for meaningful measurement of each plant's treatment performance.[fn232] Thus the samplings at the five plants identified by CMA served as an accurate measure of BAT treatment performance.F. The EPA's Toxic Limitations1. Application of the Toxic Limitations to All OCPSF DischargersThe effluent limitations subject each discharger to the BAT limits for all regulated toxic pollutants.[fn233] Pursuant to the EPA's existing NPDES regulations, the permit writers will establish monitoring requirements for toxics on a case-by-case basis, but each plant must be monitored for each pollutant at least once per year.[fn234]Courtaulds and Ethyl argue that requiring all dischargers to monitor for compliance with the BAT limitations for all toxic pollutants is arbitrary because this will require expensive monitoring and some toxics are not likely to be present in particular wastestreams. They urge that they should be required to monitor only for those toxics likely to appear in their wastestreams.The EPA contends that this challenge to the monitoring requirement is not ripe for review because monitoring requirements will be, but have not yet been, established on a case-by-case basis. The EPA's response misses the point. The petitioners are not directly challenging the frequency of monitoring but the requirement that all dischargers must monitor their effluents for all toxics, not only for those that could be expected to be found. This issue is therefore ripe for review.On the merits the EPA states that its requirement that all dischargers must comply with the effluent limitations for all priority pollutants is based on evidence and comments indicating that OCPSF plants engage in such diverse production processes that they cannot guarantee against the existence of priority pollutants in their wastestreams.[fn235] Therefore, the EPA argues, its regulations must be industry-wide. Furthermore, the EPA points out that the regulations provide that monitoring frequencies will be established on a case-by-case basis by the individual NPDES permit writer, taking into account the likelihood that certain pollutants will or will not be found in a particular discharger's wastestream.[fn236]The EPA has provided a rational reason for the need for an industry-wide rule and has provided a flexible mechanism for the mitigation of any harsh economic consequences of the rule. The rule is therefore not arbitrary.2. Courtaulds'-Notice-and-Comment ChallengeCourtaulds claims that the EPA failed to give adequate notice that it intended to subject all dischargers to the limitations for all toxics. In a 1985 notice of proposed rulemaking the EPA announced:EPA has decided to promulgate a single set of BAT limitations which would be applicable to all OCPSF facilities. (However, permits would tailor these requirements somewhat to account for the fact that most OCPSF plants routinely discharge only a subset of pollutants covered by the BAT limitations ...).[fn237]The notice listed three options the EPA was considering for establishing monitoring requirements. The first of these options, and the one ultimately adopted by the Agency, stated:(1) Limiting all pollutants covered by the regulation in the permit and as a pretreatment standard, but requiring frequent monitoring only for pollutants of concern. Other pollutants would be monitored for compliance only occasionally (e.g., once or twice per year).[fn238]The notice made it clear that the EPA was considering subjecting all dischargers to the limits for all toxics. Courtaulds' challenge is without merit.3. The NRDC v. Train Consent DecreeCourtaulds also claims that the EPA violated section 8(a)(iii) of the consent decree in NRDC v. Train, which provides that the EPA "may" exclude from regulation pollutants found only in trace amounts.[fn239] Although the decree provides that the EPA "may" exclude such pollutants from regulation, nothing in the consent decree requires the EPA to do so. Even if, therefore, this is the proper court in which to complain of noncompliance with the decree, a point we do not decide, no such violation has been shown.G. NRDC's Challenge to the EPA's BAT and BAT SubcategorizationThe EPA divided the BAT limitations into two subcategories, BAT and BAT. BAT limits are based on a model technology that added various in-plant treatment systems to end-of-pipe biological treatment. EPA recognized that some plants with low influent BODS levels do not have enough waste "nutrients" to support their treatment biota; such plants, therefore, will be unable to maintain an end-of-pipe biological treatment system.[fn240] In response to this problem, the EPA derived the BAT limits based on a variety of in-plant treatment systems. The EPA stated that the plants subject to the BAT limits are those plants that do not employ end-of-pipe biological treatment. The result of this rule is that plants that can find a way to comply with the BPT limitations without installing end-of-pipe biological treatment systems are subject to the less stringent BAT limitations.[fn241]NRDC argues that the EPA's adoption of the BAT and BAT subcategories violated the APA notice and comment requirement because this subcategorization was presented for the first time in the final rule. NRDC also asserts that this type of BAT subcategorization violates the CWA because it allows a discharger who chooses not to employ end-of-pipe treatment to be subject to less stringent BAT limits. NRDC states that, if it had an opportunity to comment, it would have urged the EPA to establish a single numerical BODS influent parameter which would define whether a plant was in the BAT or BAT subcategory. This number would presumably be based in part upon the EPA's judgment regarding the required influent BODS concentration required to effectively operate an end-of-pipe biological system; plants falling below the required BODS influent level would qualify for BAT subcategorization.We agree that the EPA failed to present its BAT subcategories for comment as required by APA § 4(c). The EPA has failed to direct this court's attention to any public notice announcing its BAT subcategories prior to promulgation of the final limitations. In fact, the record indicates that prior to final promulgation of the limitations the EPA announced that it did not intend to establish any subcategories for BAT. The record states:EPA considered whether the industry should be subcategorized for BAT purposes by evaluating the same subcategorization factors which were considered for BPT. EPA has decided to promulgate a single set of BAT limitations which would be applicable to all OCPSF facilities.... The available data for BAT show that plants ... can achieve similar low toxic pollutant effluent concentrations by installing the best available treatment components.... Therefore, the Agency believes that BAT subcategories do not appear to be necessary for effective, equitable regulation. However, EPA will continue to explore the possibility of subcategorizing the industry for BAT purposes and invites comments and supporting data on appropriate approaches.[fn242]The last sentence notwithstanding, this notice was not sufficient to fairly apprise interested parties that BAT subcategorization was still a live issue; rather the notice gives every indication that the issue had been laid to rest.We remand this issue to the EPA for notice-and-comment proceedings.[fn243] In the interim, the limitations will be given effect[fn244] for three reasons: First, we recognize Congress' concern for limiting the discharge of toxic pollutants within the statutory deadline. Second, the notice-and-comment proceedings may disclose that the BODS parameter urged by the NRDC is neither necessary nor feasible. Finally, the industrial petitioners are not prejudiced by being subjected to BAT limitations which, if anything, may be too lenient.H. DuPont's Chambers Works PlantDuPont claims that the wastewater of its Chambers Works plant is highly concentrated, with 50 of the 63 priority pollutants subject to the limitations, and the EPA failed to consider the unique nature of this plant in determining whether the BAT limitations for volatile organic pollutants are achievable. DuPont contends that the BAT limitations were derived from a data base with plants that had relatively simple wastestreams and therefore the limitations cannot be achieved by plants with complex wastestreams such as its Chambers Works plant. DuPont seeks either a subcategory containing only the Chambers Works plant or an exemption from the BAT limitations.The CWA directs the EPA to establish BAT pollutant "effluent limitations for categories and classes of point sources."[fn245] In promulgating nationwide pollutant effluent limitations the EPA need not consider the hardship faced by a particular plant. Congress intended "fundamentally different" characteristics of particular plants to be considered by the EPA in a Section 301(n) FDF variance proceeding.I. BAT Limitations for Phenol1. Phenol-Dominated WastestreamsBorg-Warner asserts that the BAT limitations for phenol are unachievable for its Westmar plants and other plants with wastestreams containing high concentrations of phenol. Borg-Warner argues that the EPA should have established a separate subcategory for such plants.The EPA determined that phenol is "highly biodegradable," and is therefore readily susceptible to biological treatment.[fn246] To derive the BAT limits for phenol the EPA used data from 23 plants and took 178 samplings.[fn247] As Borg-Warner notes, four of the data-base plants had influent phenol concentrations higher than those at Borg-Warner's Westmar plant. None of these four plants reported an exceedance of the limitations; the vast majority of the samplings from these plants were below the detectable level.[fn248] The data from these four plants establish that the phenol limits are achievable for plants with phenol-dominated wastestreams.Borg-Warner argues that the limits are based on unrepresentative data because the EPA took very few samples from the four data-base plants with phenol-dominated wastestreams over a relatively short period of time. The record shows that for database plants 306 and 1494 the EPA took three samples at each plant over a three-day period; for plant 2711 the EPA took two samples over a two-day period; and for plant 1293 the EPA took fifteen samples over a nineteen-day period.[fn249] While Borg-Warner contends that the samplings at the first three of these plants are not representative it has neither provided a bench mark nor demonstrated why sampling over a two-or three-day period is not sufficient. Merely asserting that the sampling was inadequate does not make it so.In any event, the data from the last of these four plants were taken over a substantially longer period of time and these data are consistent with the performances of the other three plants; this confirms that the data on the whole are representative. The number of data points relied on by the EPA was not insignificant, and "there must exist some reasonable termination point in the process of data collection."[fn250] The petitioners have failed to demonstrate that the EPA's choice of that termination point was unreasonable.2. The EPA's Cost Estimates for the Phenol LimitationsThe EPA designated biological treatment as the model technology for the removal of phenol. To estimate the industry costs of the BAT limitations, the EPA estimated the costs of end-of-pipe biological treatment systems. To estimate the costs of the BAT limitations the EPA projected costs for in-plant biological treatment systems. To project the size of the biological treatment system needed by a particular plant the EPA utilized a "k" rate. The "k" rate, or "biokinetic constant," is one of the factors affecting a pollutant's rate of biodegradation: the lower the k rate for a pollutant the more expensive it will be to treat that pollutant.Borg-Warner contends that the EPA erred by averaging k rates on an OCPSF industry-wide basis rather than determining the k rate for plants with phenol-dominated wastestreams. Borg-Warner argues that the k rate for phenol is substantially lower than the k rate for the industry as a whole, and therefore by averaging k rates the EPA has underestimated the costs of compliance for plants with phenol-dominated wastestreams.The Act requires the EPA to "take into account" the costs of BAT;[fn251] it does not require a precise calculation. The EPA "need make only a reasonable cost estimate in setting BAT"; it is sufficient if the EPA develops "`a rough idea of the costs the industry would incur.'"[fn252]The EPA averaged k rates on an industry-wide basis because k rates varied from plant to plant depending on each plant's mix of pollutants. Most plants in the industry produce multiple pollutants and have complex wastestreams, and thus the k rate for a single pollutant will rarely accurately reflect the rate of biodegradation at a particular plant. Like any other averaging, averaging k rates might result in either an over-or an under-estimate of a particular plant's actual costs.Borg-Warner has failed to demonstrate that the EPA's cost estimates were not a reasonable approximation of the costs the industry will have to incur to meet the BAT limitations. Though Borg-Warner asserts that the EPA used an inappropriate k rate, it has failed to demonstrate how this error, if any, affected the EPA's over-all cost estimates for plants with phenol-dominated wastestreams; in other words, Borg-Warner has failed to demonstrate how much greater actual costs will be than the costs projected by EPA. Borg-Warner's claim is too vague and conclusory to serve as a basis for invalidating the limitations.In estimating compliance costs, the EPA determined that a plant with influent phenol concentrations of 900 ppb or less could meet the BAT limitations with an end-of-pipe biological treatment system. Borg-Warner asserts that it submitted data to the EPA showing that an end-of-pipe biological treatment system would not be adequate for the Borg-Warner plant and that Borg-Warner's compliance costs would be greater than those projected by the EPA.The EPA is required to consider costs only for a category or subcategory of plants; it is not required to determine costs on a plant-by-plant basis.[fn253] Thus it is irrelevant that Borg-Warner's individual compliance costs would be greater than the costs projected by the EPA.J. The BAT Limitations for Volatile Pollutants Based on Steam Stripper Technology1. Achievability of the LimitationsThe EPA designated steam stripping as the model technology for 28 volatile and semi-volatile pollutants in establishing the BAT limitations. Steam stripping is a water reclamation process in which wastewater is run through a distillation column where it is injected with super-heated steam causing the volatile pollutants to vaporize and collect so that the volatile pollutants can be extracted from the water.PPG and Dow contend that the BAT limitations are unachievable and therefore arbitrary because the data show that some of the data-base plants could not achieve the limitations. The EPA used data from PPG plant 913 and Dow plant 415 to develop the limitations for volatile pollutants including chloroform and trichloroethylene. The record indicates, however, that PPG plant 913 exceeded the monthly average limit for chloroform, and Dow plant 415 exceeded the daily maximum limitation for trichloroethylene on one occasion. Relying on the Fourth Circuit's decision in Tanners' Council of America, Inc. v. Train,[fn254] the petitioners argue that data showing that the data-base plants "[met] the limitations for some, but not all, of the pollution parameters" is not sufficient to demonstrate that the limitations are achievable.The single exceedance of the daily limitation for trichloroethylene by Dow plant 415 does not make the limitation unachievable. As previously discussed, the EPA could rationally exclude extreme departures from average performance in calculating the limitations. The petitioners' own data show that Dow plant 415 had a long-term average discharge of 16.1 ppb of trichloroethylene. The EPA set the daily maximum limitation at 69 ppb. The Dow plant's single exceedance of 85 ppb is an extreme departure from average performance and the EPA could reasonably conclude that this poor performance was either an "upset" unlikely to recur or due to quality-control problems within Dow's control because on all other occasions the Dow plant was able to perform within the limitations. Thus, the industrial petitioners have failed to demonstrate that the Dow plant is incapable of meeting the BAT limitations.Second, we reject the petitioners' premise that the limitations are unachievable unless all plants in the data base have met the limitations. The legislative history of the CWA indicates that the "best available technology" refers to the single best performing plant in an industrial field.[fn255] The EPA urges that because the Act and the legislative history do not provide more particular guidance, it was free to determine the "best" plant on a pollutant-by-pollutant basis. The Supreme Court has stated that "[i]t is by now commonplace that `when faced with a problem of statutory construction, this Court shows great deference to the interpretation given the statute by the officers or agency charged with its administration.'"[fn256] This court defers to the EPA's interpretation of the Act. The EPA's interpretation of the Act is rational and is not precluded by the legislative history.[fn257]Though PPG plant 913 exceeded the monthly average limitation for chloroform, the petitioners' own data show that Dow plant 415 performed well within the limits. On at least one occasion Dow plant 415 exceeded the daily limit for trichloroethylene; however, the data show that PPG plant 913 performed within the limits. Given the EPA's interpretation of the Act, which is entitled to deference, an exceedance by one of the data-base plants is irrelevant so long as another data-base plant demonstrates that the limitations are achievable.2. Wastestream CharacteristicsPPG and Dow also assert that the EPA's limitations for volatile pollutants failed to account for the wastestream characteristics of OCPSF plants and that the wastestream matrix of these plants makes the limitations unachievable. The petitioners urge that the EPA should have established separate subcategories to reflect the different wastestream characteristics of different plants.As previously discussed, the EPA determined that no further subcategorization was necessary because all BAT plants could achieve uniformly high levels of toxic removal and all plants could achieve the limitations regardless of wastestream characteristics.[fn258] With respect to the BAT limitations for volatile pollutants, the EPA determined that steam-stripper technology is a versatile technology that can be modified to account for a particular plant's wastestream matrix.[fn259] The EPA therefore determined that properly-designed steam-stripper technology would make the limitations for volatile pollutants achievable for all OCPSF plants.3. Steam-Stripper MaintenanceSteam strippers must be shut down and started again weekly for periodic cleaning and maintenance and these operations result in greater discharges of toxics. The petitioners contend that the EPA edited the high discharge readings associated with such operations from the data base. They argue that this exclusion of data representing routine maintenance is arbitrary, and by excluding such data the EPA has insured that the limits are unachievable.The EPA considered this problem and determined that a surge tank with a 24-hour detention time could be used to store wastewater during shut-down for routine maintenance; following maintenance the wastewater could be recirculated through the steam stripper without discharging the volatile pollutants into navigable waters. The EPA's decision to exclude the high discharge readings associated with routine maintenance was therefore neither arbitrary nor capricious.4. Dow Chemical's Plant-Specific ClaimsDow states that its chlorinated methane plant cannot meet the BAT limitations for methyl chloride because of the plant's complex wastestream. Dow also states that its two Olefin plants cannot meet the limitations for benzene. Dow urges that these plants demonstrate that the limitations are unachievable; alternatively Dow urges that these plants are entitled to separate subcategories subject to different limits.Dow's inability to meet the limitations at these plants is irrelevant because the data show that they are achievable by other plants. The EPA is not required to establish separate subcategories for single plants. Congress intended the EPA to consider individual plant characteristics that preclude compliance in an FDF variance proceeding.K. The BAT Limitations for Priority Pollutants Based on In-Plant Biological Treatment1. Achievability of the LimitationsCMA contends that the EPA designated in-plant biological treatment "with a 24-hour detention time" and "no pre- or post-biological treatment" as the "model" technology for the BAT and pretreatment limitations for twenty priority pollutants. However, the three data-base plants used to derive the limitations used end-of-pipe biological treatment systems with detention times of 17.2 days, 3.5 days, and 38 hours respectively. These three plants also had extensive pre- and post-biological treatment systems including settling tanks, equalization basins, oil separation, clarification, filtration, and neutralization systems. CMA argues that because the data-base plants used far more extensive technology than the EPA-designated "model" technology, the EPA has not demonstrated that the limitations are achievable using the model technology.The petitioners have failed to demonstrate that end-of-pipe biological treatment systems are sufficiently different from in-plant systems to make the EPA's reliance on end-of-pipe data irrational. For all we can tell from the parts of the record that have been cited, the only difference between the two systems is that they are installed at different positions in the production process.The record indicates that the EPA designated in-plant biological treatment as the BAT model technology for the treatment of priority pollutants.[fn260] The EPA based the BAT limitations on data from three end-of-pipe biological treatment plants because these plants achieved exemplary treatment and they treated wastestreams comparable to the segregated wastestreams that would enter an in-plant treatment system.[fn261]There is no indication in the record that the EPA limited the model technology to one with a "24-hour detention time" and "no pre- or post-biological treatment" as CMA asserts. In-plant biological treatment is the generic technology chosen by the EPA as a basis for the BAT biological treatment limitations. In addition to in-plant biological treatment, an OCPSF plant may have to install one or more additional technologies to account for the plants' wastestream characteristics. For instance, plants with wastestreams containing high concentrations of suspended solids might require settling tanks, filtration, or clarification processes.[fn262] Some plants may need an equalization process to reduce the variability of the raw waste load; this technology makes the waste more uniform and makes treatment more efficient.[fn263] Others may need to neutralize their wastestreams to adjust either an acidic or a base wastestream to facilitate effective biological treatment.[fn264] Plants that have wastestreams containing oil or grease may need an oil separation system.[fn265]The Act requires the EPA to establish effluent limits "for a category or class of point sources."[fn266] It does not require the EPA to identify the specific technologies that a particular plant must install to meet the limitations.2. The EPA's Estimate of the Costs of Complying with the LimitationsTo determine the costs of in-plant biological treatment systems the EPA used a computer costing model developed by the U.S. Army Corps of Engineers called "CAPDET." The CAPDET model generates capital, operation, and maintenance costs for a biological treatment system based on the characteristics of a particular system. For plants with wastestreams of less than 0.5 million gallons per day (MGD), the CAPDET model provided cost estimates for an "extended aeration activated sludge" treatment system consisting of an equalization tank, an aeration tank, secondary clarification, and a control system.[fn267] This system requires a detention time of 18 to 36 hours.[fn268] For plants with wastestreams greater than 0.5 MGD, the CAPDET model estimated the cost of the larger "completely mixed activated sludge" treatment process;[fn269] this system requires a detention time of 50.6 hours.[fn270]As CMA notes, the detention times for two of the data-base plants were greater than the longest detention time used to calculate the costs of biological treatment systems. CMA argues that this error makes the limitations unachievable. The EPA's error, however, if any, would affect the accuracy of its cost projections and not the achievability of the limitations.Though the detention times of two of the plants used to derive the limitations were substantially longer than the detention times projected for the costed technology, the EPA determined that the other factors in a biological treatment system could be modified to achieve a particular level of pollutant removal. These factors include organic loading in the influent, the concentration of biodegrading organisms in the aeration basin, and the length of time those organisms remain in the aeration basin.[fn271] The EPA has demonstrated that the technology used to estimate costs is a reasonable approximation of the type and cost of technology that must be used to meet the limitations.3. Land CostsThe pretreatment standards for the priority pollutants, which are based on the BAT limitations, apply to indirect dischargers who are typically located in urban areas where land is scarce and expensive. In-plant biological treatment systems require additional land for the large primary treatment, equalization, and aeration tanks that are part of the system. CMA argues that the pretreatment standards are arbitrary because the EPA failed to consider the land costs associated with in-plant biological treatment.The record shows that the EPA considered the land costs necessary for in-plant biological treatment systems.[fn272] The EPA determined that the costs of the pretreatment standards were warranted, despite substantial economic dislocations, because the standards would prevent the discharge of substantial amounts of toxic pollutants into navigable waters.[fn273] The EPA considered the costs of the pretreatments standards, including land costs, and determined that these costs were reasonable; this was all that was required under the Act.L. Compliance Deadline1. Lack of Sufficient Lead Time for Industry Members to ComplyThe Water Quality Act (WQA) Amendments of 1987 directed the EPA to promulgate BAT limitations by December 31, 1986.[fn274] The WQA Amendments also established March 31, 1989, as the final compliance date for the OCPSF BAT limitations.[fn275] The EPA promulgated the OCPSF limitations on November 5, 1987. Borg-Warner argues that Congress intended industry to have at least three years lead time before it would be subject to the limitations and that the EPA's failure to promulgate regulations with sufficient lead time for compliance constitutes a denial of substantive due process.We first look to Congressional intent. Because the BAT limitations had not been promulgated on February 4, 1987, the date Congress passed the 1987 WQA Amendments, Congress obviously understood that substantially less than three years would elapse before the compliance date. The Act and its legislative history reflect Congress' intention that good faith non-compliance due to the Administrator's failure to promulgate regulations by the prescribed date would be accommodated by the EPA's post-deadline enforcement policy.[fn276]The Act, thus interpreted, does not deny industry members due process. Section 309 of the Act provides that, if a discharger fails to comply with a "final deadline," the Administrator shall schedule a "reasonable" time for compliance "taking into account the seriousness of the violation and any good-faith efforts to comply with applicable requirements."[fn277] The Conference Report accompanying the WQA Amendments further explains:With respect to the establishment of an outside date of March 31, 1989, for compliance with technology based requirements, the conferees note that prompt promulgation of the relevant effluent limitations will be essential to allow dischargers sufficient time to come into compliance. Therefore, the conferees direct that the Administrator promulgate such limitations as expeditiously as possible.If dischargers in an entire category are unable to meet the March 31, 1989, deadline provided in the conference substitute as a result of the Administrator's failure to promulgate effluent limitations in sufficient time to allow for compliance by such date, non-compliance resulting from the Administrator's delay can be dealt with under EPA's current post-1984 deadline enforcement policy. That policy calls for the Agency, at the same time a permit containing the statutory deadline is issued, to issue an administrative order to the non-complying company which specifies a schedule of compliance as expeditiously as practicable, but not later than three years after permit issuance.[fn278]Through Section 309 Congress has adequately dealt with the dilemma that may confront an industry member due to the Administrator's delay in promulgating the limitations.[fn279]Borg-Warner argues that reliance on the EPA's prosecutorial discretion is not an adequate remedy for the Administrator's delay in promulgating the regulations because of the risk of selective enforcement. This argument presumes that the EPA will act in bad faith and contrary to congressional intent. If that should occur, an industrial petitioner's remedy is to seek judicial review of the EPA's illegal actions or to invoke that illegality as a defense to an enforcement action.2. "Availability" of BAT TechnologyBorg-Warner asserts that at least three years lead time is necessary for the installation of BAT technology and many plants in the industry will not be able to install the technology in time for the compliance deadline. Borg-Warner argues that without sufficient lead time the technology needed to satisfy the BAT limitations is not "available" within the meaning of the Act.This argument is based on a misreading of the Act. The Act directs the EPA to establish effluent limitations "which shall require application of the best available technology economically achievable for such category or class."[fn280] It requires that the effluent limitations be based on "available" model technology; it does not require the EPA to consider the temporal availability of the model technology to individual plants.[fn281] Section 304(b)(2)(B) of the Act[fn282] lists those factors the EPA should consider in identifying the BAT model technology; these factors do not include consideration of an individual plant's lead time. Moreover, as previously discussed, Congress separately provided an exception under Section 309 for plants without sufficient lead time to comply with the regulations. This exception would be redundant if Congress intended the EPA to consider lead time in determining whether technology is "available."IV. Pretreatment Standards for Existing Sources (PSES) IssuesA. "Pass Through" Issues1. The EPA's Definition of "Pass Through"In addition to authorizing the EPA to regulate the discharge of pollutants directly into the nation's waters, the CWA authorizes the EPA to establish pretreatment standards for "indirect dischargers" - industrial dischargers who release wastes into publicly-owned treatment works (POTWs) rather than directly into navigable waters.Congress "recognized that the pollutants which some indirect dischargers release into POTWs could interfere with the operation of the POTWs, or could pass through the POTWs without adequate treatment."[fn283] Accordingly, the Act authorizes the EPA to identify those pollutants which "pass through" POTWs without adequate treatment, or which "interfere with" the operation of POTWs, and to promulgate pretreatment standards which require indirect dischargers to reduce the levels of such pollutants before they are released into public sewage systems.The Act provides:The Administrator shall, within one hundred and eighty days after October 18, 1972, and from time to time thereafter, publish proposed regulations establishing pretreatment standards for introduction of pollutants into treatment works ... which are publicly owned for those pollutants which are determined not to be susceptible to treatment by such treatment works or which would interfere with the operation of such treatment works.... Pretreatment standards under this subsection shall specify a time for compliance not to exceed three years from the date of promulgation and shall be established to prevent the discharge of any pollutant through treatment works ... which pollutant interferes with, passes through, or otherwise is incompatible with such works.[fn284]In 1977, when Congress amended the CWA to strengthen the provisions for controlling toxic pollutants, Congress provided that "an indirect discharger ... had to `pretreat' its waste waters so as to achieve, together with the [POTW] that treated the waste before final discharge into navigable waters, the same level of toxics removal as was required of a direct discharger."[fn285]Pursuant to this statutory mandate, the EPA has developed pretreatment standards for specific categories of dischargers that limit the types and amounts of pollutants which may be discharged to POTWs by facilities in each industrial category. These categorical pretreatment standards are technology-based and are analogous to the BAT effluent-limitation guidelines for the removal of toxic pollutants - that is, they are intended to represent the best available technology that is economically achievable by indirect dischargers.[fn286] In order to determine when a pollutant "passes through" a POTW and should therefore be subject to pretreatment, the EPA has adopted a "BAT comparison" approach. A pollutant is deemed to "pass through" for purposes of the categorical pretreatment standards "if the nation-wide average percentage of the pollutant removed by well-operated POTWs achieving secondary treatment is less than the percent removed by the BAT model treatment system."[fn287]CMA asserts that the PSES that the EPA has promulgated for the OCPSF industry are arbitrary and capricious because EPA's "BAT comparison" approach for identifying pollutants that "pass through" POTWs and are thus subject to pretreatment regulations does not take into account the actual performance of particular POTWs and therefore requires redundant treatment. Consequently, CMA claims, the PSES lead to "absurd results."Specifically, CMA argues that the EPA has established pretreatment standards for some pollutants which are already receiving exemplary treatment by POTWs across the country and that it has furthermore established pretreatment standards for some pollutants which cannot even be detected in the effluent of POTWs by present analytical methods. With respect to this last group of pollutants, CMA contends that the EPA has unlawfully assumed pass through, without any evidence that the POTW is actually achieving less effective removal than a direct discharger complying with the BAT standards.The statute itself simply directs the EPA to promulgate pretreatment standards for pollutants "which are determined" to pass through and does not specify a method by which the EPA is required to make that determination.[fn288] "Where a statute is silent with respect to the precise question at issue, the question for the reviewing court is whether the agency's interpretation is based on a permissible construction of the statute."[fn289] As the EPA interprets the statute, the Agency is not required to assume that no pass through takes place in the absence of full-scale data to the contrary.[fn290] Because we find the EPA's interpretation of the statute to be reasonable and consistent with Congress' intent, we decline to invalidate the OCPSF PSES on the grounds proposed by CMA.[fn291]In adopting the "BAT comparison" approach to defining "pass through," the EPA explained that it sought to satisfy two competing congressional objectives:That standards for indirect dischargers be equivalent to standards for direct dischargers, and that the treatment capability and performance of the POTW be recognized and taken into account in regulating the discharge of pollutants from indirect dischargers.[fn292]The EPA explained that it had determined, after considering alternative methods, that its approach of comparing average percent removal rates to determine pass through was the best solution to the inherent difficulty of measuring the effectiveness of POTWs' treatment of toxic pollutants. The difficulty stems from the fact that the concentration of toxic pollutants in POTWs' influent is much lower than that in industry treatment systems because the industrial dischargers' wastewater mixes in the POTW system with wastewater from other sources that does not contain toxic pollutants. As a result of this dilution, the POTWs' influent concentrations of toxics may already be nearly undetectable by present methods - even though the mass of toxic pollutants has not been reduced. The concentration of toxic pollutants in effluent may therefore be undetected, even if the POTWs' treatment is not, in actuality, very effective in reducing the mass of toxic pollutants.The EPA explained that it had rejected the alternatives of direct comparison of effluent concentrations[fn293] and a percent differential approach[fn294] to measuring pass through because both approaches tended to overestimate the effectiveness of POTW treatment and therefore created an unacceptable risk that toxic pollutants were in fact passing through POTWs.[fn295]The EPA concluded that the approach of comparing average percent removal rates "is unbiased in that it does not favor either overregulation or underregulation in determining which pollutants are regulated at PSES."[fn296] CMA objects that the EPA if applying this approach, it would find pass through, and therefore impose pretreatment standards, even if POTWs achieved a 96.2% removal rate for a particular toxic pollutant, so long as the EPA predicts that plants complying with BAT can achieve slightly greater removal - even 96.3%.Given the EPA's well-founded concern that the effects of dilution may cause the effectiveness of POTW treatment to be overstated, we cannot conclude that the EPA's method of determining pass through is arbitrary. As we have stated, "At first blush, it may be unclear why dilution can be a problem. One might believe that the proper goal of a treatment system is to produce water `so clean' that pollutants are present in only immeasurably small amounts. This is usually, but not always the case. Certain pollutants are dangerous even in immeasurably small concentrations."[fn297]CMA also maintains that the EPA arbitrarily assumed that certain pollutants that were not detected in POTW effluent were present at the minimum analytical level. Because the minimum analytical level is then used to calculate removal, CMA asserts that some pollutants may be subject to PSES even if the POTWs' actual percent removal is higher than the BAT percent removal of the same pollutant.The EPA has explained, however, that its practice of assigning the minimum analytic value to certain pollutants that did not appear at detectable levels in POTW effluent is also designed to compensate for the effects of dilution. The Agency explained:The conservative approach of adopting the `detection' limit or the analytical threshold as the effluent value for such measurements has the effect of underestimating the POTW's percent removal.... In many cases, in fact, both POTW and BAT treatment systems with relatively [low] influent concentrations yielded effluent measurements below detection, and the resulting percent removals were not true measures of treatment effectiveness, but rather were functions of influent concentrations.... The POTW might be achieving as high a percent removal as the BAT level technology, but there was no basis for determining whether this was so or not.[fn298]The EPA reasonably concluded that due to the effects of dilution on influent concentrations, the "non-detects" or "ND" values (indicating effluent concentrations too low to measure) derived from POTWs with low influent concentrations of priority toxic pollutants did not necessarily demonstrate that the pollutants had been effectively treated and removed from the effluent.[fn299]We hold that the EPA reasonably adopted the conservative methodology of assigning the minimum analytical value to certain pollutants that yielded "ND" values in effluent concentrations in order to account for the possibility that, because of the effects of dilution on influent concentrations, priority toxics are present in POTW effluent at levels far greater than can presently be measured. We further note that the EPA has taken reasonable steps to reduce the likelihood of underestimating the effectiveness of POTW removals.[fn300]2. The EPA's Methodology in Assuming an Absence of "Pass Through" Based on POTW AveragesNRDC argues that the EPA's pass-through methodology in practice tends to overestimate the amount of toxic pollutants removed by POTWs. Originally, the EPA had proposed to find pass through whenever it had insufficient POTW data to make a comparison with BAT removal; this position was supported by NRDC. The EPA then revised its position and decided to assume an absence of pass through for three toxic pollutants - bis(2 chloroisopropyl)ether, acrylonitrile, and 3, 4 benzofluoroanthene - on the basis of POTW removal averages. NRDC argues that the EPA's new position is inconsistent with § 307(b)(1) of the Clean Water Act, which reads: "Pretreatment standards ... shall be established to prevent the discharge of any pollutant through treatment works ... which are publicly owned, which pollutant interferes with, passes through, or otherwise is incompatible with such works."[fn301] NRDC further alleges that the EPA committed analytical errors in its calculation of average POTW removal. The EPA decided to calculate the overall average of BAT and POTW removal rates rather than average daily rates. This, NRDC alleges, tends to merge spike concentrations of individual pollutants into the base flow and thus determines a plant's average removal, rather than a plant's consistent, daily removal capability.The EPA responds to NRDC's argument that determining the average POTW removal rate for purposes of determining pass through is inconsistent with the Clean Water Act by correctly noting that its regulations meet Congress' mandate that the EPA establish pretreatment standards for pollutants which pass through. Since Congress provided no criteria by which the EPA is to determine when a pollutant "passes through" a POTW untreated, the establishment of pass-through criteria is left to the Administrator's discretion. In reviewing the Administrator's interpretation of the phrase "pass through," we must accord the Administrator a presumption of regularity.[fn302] While we must "reject administrative constructions that are contrary to clear congressional intent,"[fn303] we must nonetheless accord the EPA deference if it is within the limits of its authority to interpret that statute.[fn304]The EPA has chosen to define pass through by reference to the average amount of toxic pollutants removed by the POTW. NRDC would rather have the EPA determine pass through on a daily consistent basis. We hold that the EPA's decision to define pass through based on POTW average removal does not violate the CWA. Due to the fact that industrial waste entering a POTW is mixed with other municipal wastes, such as sewage, it is impossible to trace a given influent stream through the POTW. It would therefore be impossible for the EPA to accurately determine what percentage of waste a POTW was removing on a daily basis, since this requires a direct comparison of influent concentration to effluent concentration.[fn305] As a result of this analytical difficulty, the EPA has chosen the reasonable and available step of evaluating POTW performance on an average basis. Given the factual situation and analytical uncertainty in this area, we find this to be an acceptable interpretation of § 307(b)(1) of the CWA.3. The EPA's Decision Not to Find "Pass Through" Based on Sludge ContaminationThe EPA determined that chromium, copper, and nickel do not pass through POTWs, and therefore the EPA did not promulgate pretreatment standards for these metals. NRDC contends that these metals contaminate the sludge, which is not treated by wastewater technologies, and that sludge contamination was not considered by the EPA. The result of this, NRDC argues, is that the EPA's calculations tend to overestimate POTW removal effectiveness because pollutants transferred to sludge are counted as removed from the wastewater when, in effect, they are merely transferred to a different medium, in this case, sludge. NRDC therefore requests this court to order the EPA to develop pretreatment standards for these three pollutants. The EPA responds, first, that neither NRDC nor any other petitioner commented on this issue. While this does not necessarily bar NRDC from challenging the rule now, we will nonetheless accord NRDC less latitude in their argument.[fn306]EPA notes, secondly, that it plans to regulate sludge separately from wastewater, so presumably the problem of these three pollutants will be addressed in the forthcoming sludge regulations, which are now in the developmental process. Congress' intent was that the presence of pollutants in sludge would be regulated under § 405(d).[fn307] As further evidence of Congress' intent, in 1977 Congress rejected a proposed amendment which would have required the EPA to regulate pollutants which contaminate the sludge of POTWs.[fn308] Therefore, we hold that the EPA acted in accordance with its statutory mandate when it decided not to find pass through of chromium, copper and nickel, which contaminate sludge, since sludge will be the subject of future regulations.B. The EPA's Decision Not to Regulate Six Volatile Organic Pollutants on the Basis of Interference with POTW Worker SafetyNRDC argues that EPA's decision not to regulate six toxic pollutants violated the Clean Water Act. Specifically, the EPA declined to find pass through, and therefore did not establish pretreatment standards for, six pollutants which, according to NRDC, volatilize or evaporate into the air. If a pollutant volatilizes, pollution is transferred from one medium to another - in this case, from water to air. NRDC stresses that the EPA must determine that a pollutant is physically degraded or treated in order to find that it has been removed.The EPA responds to NRDC's challenge by noting that three of the six pollutants not regulated - 2, 4 dinitrophenol, benzo(k)-fluoroanthene, and acenaphthylene - would be sufficiently treated by in-plant controls on the same wastestream for other toxic pollutants that have been determined to pass through.[fn309] This is due to the fact that these three compounds are structurally similar to the other compounds for which PSES are established, thus ensuring sufficient control.[fn310] We find this reasoning to be rational and therefore uphold the EPA's judgment in this regard.The EPA's decision to reserve three other pollutants - acrylonitrile, bis(2-chloroisopropyl)ether, and benzo(b)fluoroanthene - was based on the fact that they do not volatilize extensively and that EPA lacks POTW removal data.[fn311] The EPA points to the record which indicates that benzo(b)fluoroanthene does not volatilize.[fn312] Acrylonitrile volatilizes at the low rate of approximately five percent.[fn313] This is compared to a volatilization rate of 80-95% for the pollutants that were regulated by the EPA on the basis of volatilization.[fn314] Thus, we are persuaded that the EPA's decision to reserve acrylonitrile and benzo(b)fluoroanthene was reasonable due to the low amounts of volatilization that actually occur.NRDC finally argues that the EPA should have regulated additional pollutants because certain volatile toxic pollutants interfere with POTW worker safety. NRDC does not, however, direct us to any portions of the record which adequately support this contention. The EPA solicited comments regarding the regulation of volatile organics on the basis of interference with worker safety, but ultimately rejected this option due to a lack of record support.[fn315] Though NRDC did, in 1985, comment generally regarding the problems that volatile organics may cause in POTW sewers, it never provided the EPA with documented support for a finding of interference based on worker safety.[fn316] Due to the lack of any evidence confirming NRDC's assertions of dangers posed to workers' safety, we must uphold the EPA's decision not to regulate on this basis.C. Application of PSES to Small PlantsThe Synthetic Organic Chemicals Manufacturers Association (SOCMA) claims that the pretreatment standards are arbitrary and unlawful as applied to small indirect-discharging plants or those which produce less than 15 million pounds of OCFSF products annually. SOCMA argues first, that PSES will have a disproportionately severe impact on small plants as compared to large plants and that, absent the creation of a subcategory for such plants, the EPA has failed to meet the requirement that PSES be economically achievable. Second, SOCMA maintains that the EPA's analysis of the economic impact on small plants is invalid because the EPA improperly relied on vendor-supplied package biological plants as a model technology. Finally, SOCMA argues that in refusing to create a subcategory for small plants, the EPA improperly failed to consider that only a few plants and a few pollutants account for most discharges by small plants. SOCMA maintains that the EPA should have exempted small plants from the PSES or established less stringent standards for such plants.The EPA responds that the economic impact of PSES on small direct dischargers is not so severe and disproportionate as to warrant the creation of a special subcategory subject to less stringent regulations. The EPA emphasizes that indirect dischargers have not been subject to national regulations over the past fifteen years, and that many have not employed any treatment at all. EPA concludes that small dischargers cannot be exempted from PSES or subject to relaxed standards without allowing large quantities of pollutants to go, unregulated, into public sewage systems.1. Economic Impact of PSES on Small DischargersAs noted above, PSES are equivalent to BAT standards.[fn317] Accordingly, PSES are to be set in accordance with the directives of Section 304(b)(2)(B), which dictates that the EPA:take into account the age of equipment and facilities involved, the process employed, the engineering aspects of the application of various types of control techniques, process changes, the cost of achieving such effluent reduction, non-water quality environmental impact (including energy requirements), and such other factors as the administrator deems appropriate.[fn318]Congress did not specify a particular method of evaluating the costs of compliance with BAT or specify how those costs should be considered in relation to other factors, but simply directed the Agency to consider whether the costs associated with pollutant reduction are "economically unachievable."[fn319] Both Congress and the Supreme Court have made clear that in setting BAT, the EPA is not required to compare the costs against the benefits of pollution reduction in the same manner as the EPA is required to do in setting BPT standards.[fn320]Because standards based on BAT, like BAT itself, reflect the intention of Congress to push industries toward the goal of eliminating the discharge of pollutants as quickly as possible, this goal is factored into determinations of the reasonableness of the costs associated with the regulation.[fn321] As long as the required technology reduces the discharge of pollutants, a court's "inquiry will be limited to whether the Agency considered the cost of technology, along with the other statutory factors, and whether its conclusion is reasonable."[fn322]In evaluating the costs associated with PSES compliance, the EPA performed a detailed cost assessment[fn323] and found that the projected capital and annualized costs associated with compliance with PSES for the OSPCF industry as a whole would be $291,500,000 and $204,300,000, respectively.[fn324] The EPA did not find these costs so significant as to render the PSES economically unachievable.[fn325]In assessing the economic effects of the PSES, the EPA also evaluated the plant closures, associated job losses, and other significant impacts that may result from compliance with PSES. The EPA estimated that compliance with the PSES would force 14% of all indirect discharging plants to close and cause a 1.2% reduction in total industry employment. The EPA evaluated these estimates as follows:While these impacts are significant, the Agency does not believe they constitute economic unachievability for the indirect discharging segment of the OCPSF industry. Eighty-six percent of the indirect discharger segment of the industry will not suffer either plant or product line closures, and 69 percent of the indirect discharging plants will not be significantly impacted under any measure. A very large number of pounds of toxic pollutants (22.5 million pounds) from discharges to POTWs will be removed by PSES. EPA has therefore concluded that promulgation of PSES as described above is therefore warranted with respect to OCPSF indirect dischargers.[fn326]SOCMA claims, however, that the costs of PSES fall disproportionately on plants producing less than or equal to 15 million pounds of OCPSF product annually. Specifically, SOCMA maintains that plants of this size produce only 1.37% of the OCPSF products produced by all OCPSF dischargers, but will bear 35% of the capital costs. Moreover, SOCMA notes, the proportion of small plants that will be severely impacted by PSES is 50%, compared to 14% for those plants producing more than 15 million pounds of OCPSF production. SOCMA concludes that the PSES are economically unachievable for plants under the 15 million pound production level, and that the EPA's failure to create a subcategory for these plants thus renders PSES arbitrary as applied to this group.Our role in reviewing the EPA's decision is not to undertake our own economic study, but to determine whether the EPA "has established in the record a reasonable basis for its decision."[fn327] We find that the EPA has done so. Specifically, the EPA compared the projected plant closures and associated job losses for groups of plants categorized according to the amount of OCPSF production manufactured annually. The EPA analyzed the impacts on plants producing less than 15 million pounds of OCPSF goods annually and further compared subsets within that larger group. The EPA found that plants producing less than 5 million pounds of OCPSF products annually are not impacted significantly more than plants producing less than 10 or 15 million pounds annually and thus do not warrant recognition as a distinct class of plants.[fn328] Of the 106 plants producing less than 5 million pounds annually, 27 or approximately 26% of those plants were projected to close, resulting in a loss of 800 jobs.[fn329]The EPA explained that it focused on those plants under the 5-million-pound cutoff because it found that the 10- and 15-million-pound cutoffs did not capture impacts as accurately as the 5-million-pound cutoff. While both groups included approximately the same percentage of significantly affected plants, the EPA determined that there are 22 more plants under the 10-million-pound cutoff and 47 more plants under the 15-million-pound cutoff that are not significantly affected. The EPA concluded that any relief for plants of this size would be inappropriate because of the number of unaffected plants, and because the amount of pollutant discharged by plants under the 10- or 15-million-pound production cutoffs is far greater than that discharged by plants under the 5-million-pound cutoff.[fn330] The EPA concluded that the closure rates for these three subgroups under 15 million pounds of production annually (under 5 million, 5 to 10 million, and 10 to 15 million pounds) - approximately 25%, 20%, and 12%, respectively - are close enough to warrant similar treatment and that none of these three groups alone exhibited an impact so severe as to warrant an exemption.[fn331] The EPA found that the alternative of granting an exemption for all indirect dischargers under the 15-million-pound cutoff would exempt from PSES almost 70 percent of all indirect dischargers that are not significantly affected and would forego the removal of 5.42 million pounds of toxic pollutants annually.[fn332]While we agree that PSES will have a serious economic impact on OCPSF indirect dischargers under the 15-million-pound production level, we conclude that the EPA has provided a sufficient record basis for its decision not to create a subcategory for these plants. Accordingly, we cannot conclude that the EPA's decision was arbitrary.[fn333]We must view the EPA's decision in the light of Congress' intent regarding the role of economic factors in setting water quality standards. Congress clearly understood that achieving the CWA's goal of eliminating all discharges[fn334] would cause "some disruption in our economy," including plant closures and job losses.[fn335] Moreover, Congress was aware that the economic impact of environmental regulations would be most severe for small plants.[fn336]In the light of Congress' judgment that society must bear such costs as the price of achieving the long-term benefits of eliminating pollutants from our nation's waters, courts have been exceedingly reluctant to invalidate environmental regulations on grounds of cost - even in cases where the economic impact was more severe than in the instant case.[fn337]2. Validity of the EPA's Cost AssessmentSOCMA further disputes the EPA's use of vendor-supplied package biological treatment plants in estimating the cost of PSES to small plants. SOCMA's argument is premised on the contention that none of the model technologies selected by the EPA - let alone the package biological treatment plants - will be able to achieve PSES. Thus, SOCMA argues, small plants will in fact have to bear the greater cost of more extensive treatment than the EPA has estimated to be necessary.We held above, however, that the EPA's choice of model technologies for BAT (and thus for PSES) was reasonable and that the record supported the EPA's conclusion that BAT limits could generally be achieved using the model technologies. SOCMA has failed to point to any record evidence that would undermine our earlier conclusion. We therefore hold that PSES are not arbitrary due to EPA's reliance on package biological treatment as a model technology for PSES.3. Alternative Pretreatment StandardsFinally, SOCMA claims that the EPA's data indicate that a few small plants account for a majority of the pollutant loadings and that a limited number of pollutants account for most discharges by small plants. SOCMA argues that the EPA could have promulgated PSES based on these factors that would not have had such a severe impact on small dischargers. SOCMA's alternative approach would impose a limitation on the mass or pounds of pollutants that a small plant may discharge to a POTW. SOCMA seeks to limit individual dischargers to a total of 25 pounds per day, or alternatively, to 100 pounds per day, of only 16 "dominant" pollutants. SOCMA asserts that this approach would result in the worst polluters bearing a more proportionate share of the costs of compliance.The EPA responds that SOCMA's proposal would result in unacceptably high annual loadings of 9,125 and 36,500 pounds per plant, respectively. Such a result, the EPA argues, would be inconsistent with Congress' intent that indirect dischargers achieve the same levels of removal as their direct-discharging counterparts,[fn338] performing in this case as well as those direct dischargers that comply with BAT standards. The EPA notes that pretreatment is not limited to "significant" removals, "dominant" pollutants or "large dischargers" but is intended to achieve a uniform standard of removal throughout the OCPSF industry.It is well established that the Agency's regulatory approach may not be invalidated simply because petitioners purport to have identified an alternative, less burdensome approach.[fn339]We conclude that SOCMA has not demonstrated that the EPA's refusal to create a subcategory, subject to less stringent standards, for small dischargers was arbitrary, capricious or contrary to law.D. Application of PSES to Paint/Resin PlantsThe National Paint and Coatings Association (NPCA) challenges the inclusion of paint manufacturers with subsidiary resin operations (paint/resin plants) in the scope of the national PSES rule for the OCPSF industry. NPCA argues first that the EPA failed to consider the unique characteristics of paint/resin plants and relied, in setting PSES, on data from other types of plants which are different in several crucial respects from paint/resin plants. As a result, NPCA maintains, the PSES are not technologically or economically achievable for combined paint/resin plants and are therefore not lawful absent the creation of a special subcategory. NPCA also argues that the EPA has in effect withdrawn the exemption from national effluent-pollution guidelines granted to paint manufacturers under the consent decree in NRDC v. Train. Lastly, NPCA argues that the final rule is unclear as to which paint/resin operations will be subject to PSES.1. The EPA's Data BaseNPCA maintains that the EPA's data base was inadequate to establish PSES for combined paint/resin plants because the EPA included no such plants in its data base, but relied instead on data from plants employing continuous rather than batch production processes and from large chemical/polymer plants with larger and less variable wastewater flows.With regard to NPCA's first claim, it is well-established that, in identifying model technologies, the EPA may use performance data from one part of an industry for another part of that industry if there is sufficient basis in the record to support the EPA's conclusion that the data are applicable.[fn340] In this case, we find sufficient basis in the record to support EPA's conclusion that its data were applicable to combined paint/resin facilities.The EPA notes that, while it did not sample any wastewater from combined paint/resin facilities,[fn341] it did sample the wastewater from six plants manufacturing resins typically associated with paint production (five of them, like paint/resin plants, employed batch processes).[fn342] The EPA expressly found that wastewater characteristics from resin manufacture would be the same regardless of where the resin was manufactured:The resin manufacturing polymerization processes (e.g., bulk, addition, and condensation) are essentially identical chemical processes whether conducted in a paint formulation facility as captive or in secondary OCPSF manufacture or in a primary OCPSF facility for use in secondary paint formulation at that same facility or for sale to a paint formulation facility. Likewise, the characteristics of the wastewaters produced in resin manufacturing are essentially identical.[fn343]Raw wastewaters from resin-manufacturing plants were found typically to contain organic pollutants in the same concentration as the untreated wastewaters of the plants that the EPA used to establish PSES.[fn344] In many cases, raw resin wastewater contains significantly lower concentrations of pollutants, including organic pollutants, than the raw wastewater of the plants in the EPA's data base.[fn345]The EPA also found that pollutants in untreated resin wastewater are within the concentration ranges of pollutants with the same determinative physical properties that are present in the untreated wastewaters of the plants used to establish PSES[fn346] and are therefore susceptible to those model treatment technologies for PSES which depend more on the pollutants' physical properties than on pollutant concentration.[fn347]We find that the EPA has adequately considered whether its data for establishing PSES are applicable to paint/resin plants. We cannot conclude on this record that the EPA acted arbitrarily in subjecting combined paint/resin plants to PSES.2. AchievabilityNPCA argues that combined paint/resin plants will be unable to achieve PSES because their wastestreams combine wastewater from the paint production process - which is not subject to OCPSF standards - with wastewater from the resin production process - which is subject to OCPSF standards. The EPA proposed two methods of treating OCPSF wastewater from paint/resin manufacture: First, the wastewater can be treated on a combined basis - with the resin portion of the flow receiving an allowance based on the OCPSF rule and the paint portion of the flow receiving an allowance based on the combined wastestream formula.[fn348] Second, the wastewater from resin manufacture can be segregated from the wastewater from paint production and either treated separately on site or hauled on contract to a different treatment facility. In most cases, the EPA assessed the compliance costs of this latter option (separate treatment) and found the costs to be economically achievable.NPCA asserts that neither method will make PSES achievable for combined paint/resin plants. First, NPCA maintains that the combined wastestream formula will operate to impose an impossibly stringent standard for treating combined wastestreams. Second, NPCA claims that altering the plants to allow for separate treatment of the two wastestream components would be too costly.With respect to NCPA's first objection, the EPA responds that NPCA has misinterpreted the formula: the combined wastestream formula adjusts the discharge limit established by the relevant categorical pretreatment standard (here the OCPSF standard) where the wastestream regulated by that pretreatment standard is combined with other wastewaters prior to pretreatment by the indirect discharger.[fn349] If the wastestream is combined with specified types of dilution wastestreams, the formula provides that the allowance be reduced to reflect the portion of the flow attributable to the dilution stream.[fn350] If, on the other hand, a regulated stream is combined with an unregulated stream, the concentration limit for the regulated "non-dilute" pollutant in the combined stream remains unchanged.[fn351]The EPA asserts that NPCA's objections rest on the incorrect assumption that the EPA would treat the paint stream as "dilution flow" for purposes of applying the formula. Thus, NPCA contends, the allowance which would otherwise apply to the regulated, OCPSF resin portion of the wastestream would be reduced by an amount proportionate to the unregulated paint flow, so that the combined stream would have to meet the OCPSF standards, thereby imposing an impossible standard on plants choosing to treat combined streams. The EPA argues, however, that paint flows would not be treated as dilution flows for purposes of the formula. Dilution flows receive no allowance under the EPA's combined wastestream formula while unregulated but "non-dilute" streams (such as those from paint manufacture) would receive the same allowance as the regulated flow (i.e., the OCPSF resin flow). The allowance otherwise applicable to the resin portion of the wastestream would therefore remain unchanged for paint/resin plants treating combined wastestreams. That is, plants treating combined streams would not be required to meet a more stringent standard than plants treating the paint and resin wastestreams separately.Because NPCA's objections to the combined wastestream formula are based upon an interpretation of the formula that the EPA has explicitly rejected, the EPA concluded reasonably that the PSES were economically achievable for paint/resin plants treating combined wastestreams. The EPA, therefore, determined reasonably that the PSES could be achieved by at least one of the two methods proposed for paint/resin plants.NPCA further asserts that the EPA failed to consider other "unique" characteristics of combined paint/resin plants that will make PSES unduly costly and therefore unachievable. NPCA argues that combined paint/resin plants are different from primary OCPSF production plants with respect to processes employed, wastewater volumes, wastewater characteristics, and age of equipment and facilities. The EPA responds that it not only took these factors into account, but also did so on a plant-by-plant basis and determined that the PSES were economically achievable for combined paint/resin facilities.The EPA notes that as part of its 1985 proposal, it estimated compliance costs for every plant covered by the OCPSF PSES that had responded to the Agency's request for information pursuant to Section 308 of the CWA. EPA next determined the costs that each plant would incur as a result of the proposed rule.[fn352] The Agency then determined, based on these cost estimates, that the proposed PSES were economically achievable for each subcategory.We find that the record supports the EPA's contention that in evaluating costs on a plant-by-plant basis, it considered all of the factors that NPCA claims the Agency disregarded: the age of the facilities, the amount and cost of land needed to install treatment equipment, wastewater volumes, the type of treatment equipment needed, and wastewater characteristics.[fn353] In addition, the EPA developed specific cost estimates for monitoring, capital, operating and maintenance, sludge disposal, and land.[fn354]The record similarly contains ample evidence that the EPA's determination that PSES are economically achievable rests on a thorough analysis of compliance costs. In assessing compliance costs for combined paint/resin plants, the EPA assumed that the resin flow would be treated separately from the paint flow and assessed the cost of treating the resin flow accordingly. For smaller plants that have OCPSF flows of under 500 gallons per day and OCPSF production that is less than 20% of the total plant production, the EPA assumed the resin wastewater would be hauled on contract to a different wastewater treatment facility.[fn355] For plants with larger wastewater flows, the EPA assumed that a separate system would be constructed to treat the resin portion of the flow.