Full opinion text
By RUBIN, GARZA and KING, Circuit Judges Jointly. Acting under the mandate of the Clean Water Act (CWA), 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. 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 wastes-treams. 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 Variances 1. 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 Exceedanc-es: 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: I. 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 BAT2 Limitations for Volatile Pollutants Based on Steam Stripping Technology: 1. PPG and Dow have failed to demonstrate that the BAT2 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 BAT2 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 BAT2 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 Dis-chargers: 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 SOC-MA’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 Subcate-gorize on the Basis of POTW Removal Credits: 1. The EPA is not required to subcate-gorize 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 Sau-get 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 was-testream 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 rule-making 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. Table of Contents Introduction: Page Statutory Background 195 Page Standards of Review. 197 I. Procedural Challenges. 200 1. The EPA’s Economic-Impact Study. 200 2. Limits on Metal-Bearing Wastestreams. 202 II. Best Practicable Technology (BPT) Issues. 203 A. The EPA’s Consideration of the Industry’s Costs of Complying with the BPT Limitations. 204 1. CMA’s Challenge Based on the “Knee-of-the-Curve” Cost-Effectiveness Test. 204 2. The BCT Cost-Effectiveness Test. 206 3. The Cost of Compliance with the BPT Limitations.207 B. The EPA’s Definition of the BPT Data Base. 207 1. The EPA’s Determination of the “Average of the Best” Dischargers. 207 2. The EPA’s Rejection of Sequential Treatment Options 208 C. The “Summer/Winter Issue”: The EPA’s Decision Not to Subcategorize Based on Climatic Differences.. 210 1. Diversity of the Data Base. 211 2. Winter Removal Efficiencies 213 Page D. BPT Subcategorization and the EPA’s Use of Standard Industrial Classification (SIC) Codes. 214 1. CMA’s Challenge to the BPT Subcategorization. 214 2.NRDC’s Notice and Comment Challenge E. Issues Concerning Waste-Stabilization Ponds. 217 1. The EPA’s Determination that Pond Algae Were Conventional Pollutants and Thus Subject to BPT Regulations. 217 2. The EPA’s Decision Not to Create a Subcategory for Plants Utilizing Waste-Stabilization Ponds. 218 3. The EPA’s Consideration of the Cost of Compliance for Plants Utilizing Pond Technology. 219 4. Copper Sulfate Treatment for Algae Control. 220 5. The Cost of the BPT Limitations.221 F. Plant-Specific Challenges to the BPT Limitations and the Availability of Fundamentally-Different-Factor Variances. 221 1. Union Carbide. 222 2. Borg-Warner. 223 3. DuPont’s Chambers Works Plant. 223 4. Monsanto and Ethyl. 223 5.FDF Variances 225 Page III. Best Available Technology (BAT) Issues. 226 A. The EPA’s Statistical Method of Developing the BAT Limitations. 227 1. The EPA’s Use of Weighted Averaging in Deriving the Long-Term Averages. 227 2. Averaging of Variability Factors. 228 B. Remedy for Unavoidable Exceedances. 228 C. The EPA’s Analytical Methodology. 230 1. Analytical Variability. 230 2. Complex Wastestreams. 230 3. Computation of Variability Factors When Data Were Insufficient (“Borrowed Data”). 231 4. CMA’s Challenge to the EPA’s Choice of the Minimum Analytical Value to Assign to Non-Detect Readings 231 5. Changes in Analytical Methods . 233 D. Use of Minimum Analytical Values for Enforcement Purposes. 233 E. The EPA’s Sampling Techniques. 233 F. The EPA’s Toxic Limitations. 234 1. Application of the Toxic Limitations to All OCPSF Dis-chargers . 234 Page 2. Courtaulds’ Notice-and-Comment Challenge. 235 3. The NRDC v. Train Consent Decree. 235 G. NRDC’s Challenge to the EPA’s BATi and BAT2 Subeategorization. 235 H. DuPont’s Chambers Works Plant. 236 I. BAT Limitations for Phenol. 236 1. Phenol-Dominated Wastestreams. 236 2. The EPA’s Cost Estimates for the Phenol Limitations 237 J. The BAT2 Limitations for Volatile Pollutants Based on Steam-Stripper Technology. 238 1. Achievability of the Limitations. 238 2. Wastestream Characteristics. 239 3. Steam-Stripper Maintenance .239 4. Dow Chemical’s Plant-Specific Claims. 240 K. The BAT2 Limitations for Priority Pollutants Based on In-Plant Biological Treatment. 240 1. Achievability of the Limitations. 240 2. The EPA’s Estimate of the Costs of Complying with the Limitations. 241 Page 3. Land Costs . 241 L. Compliance Deadline... 242 1. Lack of Sufficient Lead Time for Industry Members to Comply. 242 2. “Availability” of BAT Technology. 243 IV. Pretreatment Standards for Existing Sources (PSES) Issues_243 A. “Pass Through” Issues . 243 1. The EPA’s Definition of “Pass Through”. 243 2. The EPA’s Methodology in Assuming an Absence of “Pass Through” Based on POTW Averages. 247 3. The EPA’s Decision Not to Find “Pass Through” Based on Sludge Contamination. 247 B. The EPA’s Decision Not to Regulate Six Volatile Organic Pollutants on the Basis of Interference with POTW Worker Safety. 248 C. Application of PSES to Small Plants. 249 1. Economic Impact of PSES on Small Dischargers. 249 2. Validity of the EPA’s Cost Assessment. 252 3. Alternative Pretreatment Standards. 252 D.Application of PSES to Paint/Resin Plants 253 Page 1. The EPA’s Data Base. 253 2. Achievability. 254 3. The NRDC v. Train Consent Decree. 256 4. Subcategorization. 257 E. The EPA’s Decision Not to Subcategorize on the Basis of POTW Removal Credits. 257 1. Removal Credits. 258 2. GCWDA and Village of Sauget POTWs. 261 3. Crown Central Petroleum Consent Decree. 261 V. New Source Performance Standards (NSPS) Issues. 261 1. The EPA’s Cost Test in Establishing NSPS. 262 2. NRDC’s Challenge that the EPA Failed to Consider Technology Beyond BPT and BAT. 262 VI.The Economic Impact of the Montreal Protocol on the Chlorofluorocarbon Industry. 264 VII. Scope of the Regulations. 265 1. The EPA’s Reservation of Nonconventional Pollutants and Eight Priority Pollutants for Future Rulemaking 265 Page 2. Application of the Regulations to Laboratory Discharges . 266 Conclusion 266 Statutory Background The purpose of the Clean Water Act (CWA) is to restore and maintain the chemical and biological integrity of the nation’s waters. It was adopted to effectuate Congress’s declared “national goal that the discharge of pollutants into the navigable waters be eliminated by 1985,” a date later extended to March 31, 1989, and to prohibit “the discharge of toxic pollutants in toxic amounts.” Until 1972, the federal government relied primarily on state and local action to accomplish federal pollution-abatement goals. 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, 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. Recognizing the growing seriousness of the problems created by toxic pollution and the inadequacy of the 1972 FWPCA to deal with them, Congress amended that Act by adopting the CWA in 1977. 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. 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 dischar-gers based on BAT or more stringent criteria. 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. 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). 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. 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.” 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. 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 dis-chargers of toxic pollutants must comply with the even more rigorous effluent limitations based on the “best available technology economically achievable” (BAT). 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. 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 wastewa-ter. 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, the Act requires the EPA to establish “pretreatment standards for existing sources ” (PSES). These standards must also be based on BAT. 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. An FDF variance application may be based on supporting data submitted to the EPA during the rulemak-ing process or on information that the applicant did not have a reasonable opportunity to submit at that time. 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. 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. Standards of Review The 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.” 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.” 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. A party petitioning for review of an agency’s regulations bears the burden of overcoming this presumption. 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.” In statutory interpretation, the judiciary is “ ‘the final authority and we must reject administrative constructions which are contrary to clear congressional intent.’ ” 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. If, therefore, the statute is susceptible to more than one interpretation, we must accept that of the EPA if it is reasonable. 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.” According deference to an agency, however, does not imply “rubber stamping” its decision. In interpreting the Act, we do not lack precedent. The Supreme Court has twice considered issues involving its construction, and eight circuit courts have done so in at least twenty-seven cases. 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.” As a reviewing court, however, we must not “substitute [our] judgment for that of the agency,” but must "start with the assumption that the agency’s action is valid.” 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.” 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.” 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,” 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.’ ” 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.” Because judicial review “must be based on something more than trust and faith in EPA’s experience,” a court may not respond to claims of technical expertise by “rubber stamping” an agency decision as correct. 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.’ ” 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; 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; whether the issue turns on the agency’s interpretation of its own authorizing statute, which entitles it at least to guarded deference; and the validity of the reasoning upon which the agency relies to justify its actions. 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. I. Procedural Challenges 1. The EPA’s Economic-Impact Study The Clean Water Act explicitly requires the EPA to consider economic achievability in formulating pollutant effluent limitations. The Chemical Manufacturers Association (CMA) contends that, in preparing its economic-impact study, the EPA departed from its prior practice during the rule-making 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. The agency must then give interested persons an opportunity to participate in rulemaking through criticism and comments. 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. 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. 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 “[tjhese critical elements of the proposal did not change, and the final rule was a ‘logical outgrowth’ of the proposed rule.” In Community Nutrition Institute v. Block 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.”. 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 industry-wide “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. 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. 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. The APA provides that a reviewing court shall take due account of “the rule of prejudicial error.” 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.” 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.” 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 Wastestreams Appendix 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. 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 un-complexed metals, the source of metals in OCPSF wastestreams, whether each metal is complexed or uncomplexed, and, for com-plexed metals, a list of the complexing agents. 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. 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.’ ” 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) Issues The CWA authorizes the EPA to establish effluent limitations for direct dischar-gers 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). 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... . 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. 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. 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 Limitations CMA 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.” 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. 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. 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. 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 dischar-gers 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. 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 at an annualized compliance cost of 76.6 million dollars after a capital investment of 215.8 million dollars. 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 Test CMA 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. 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. 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. 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. 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.” 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-per-pound 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, 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. 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. In the current case, the regulation will require a 10% increase above current industry costs to remove 108 million additional pounds. The EPA reasonably concluded that these costs were not “wholly disproportionate” to the benefits. 2. The BCT Cost-Effectiveness Test CMA 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. 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... . 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), 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 — 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. 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 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.” 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. 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. 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 Limitations Finally, 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-pou