Weyerhaeuser Company v. Douglas M. Costle, Administrator, Environmental Protection Agency, No. 76-1674

McGOWAN, Circuit Judge: Under the aegis of the Federal Water Pollution Control Act Amendments of 1972 (the Act), Pub.L. No. 92-500, 86 Stat. 816, 33 U.S.C. §§ 1251-1376, as amended, Clean Water Act of 1977, Pub.L. No. 95-217, 91 Stat. 1566, the Environmental Protection Agency has embarked upon a *1019 step-by-step process of issuing effluent limitations for each industry that discharges pollutants into the waters of the United States. By these consolidated petitions, members of one such industry, American pulp and paper makers, challenge the validity of EPA regulations limiting the 1977-83 effluent discharges of many pulp, paper, and paperboard mills. See 42 Fed.Reg. 1398-426 (1977). We are satisfied that EPA properly construed and rationally exercised the authority delegated to it by Congress — and that, with one exception, it did so according to the appropriate procedures. Accordingly, we uphold the resulting effluent limitations in all but one instance. I. THE STATUTE After several years of judicial experience with the Federal Water Pollution Control Act Amendments of 1972, and, in particular, with industry-wide challenges to effluent limitations promulgated thereunder, little further explanation of the statutory framework is necessary. See, e. g., E. I. duPont de Nemours & Co. v. Train (duPont), 430 U.S. 112, 116-21, 97 S.Ct. 965, 51 L.Ed.2d 204 (1977); American Paper Inst. v. Train, 177 U.S.App.D.C. 181, 186-89, 543 F.2d 328, 333-36, cert. dismissed, 429 U.S. 967, 97 S.Ct. 398, 50 L.Ed.2d 335 (1976); American Frozen Food Inst. v. Train, 176 U.S.App.D.C. 105, 113-122, 539 F.2d 107, 115-24 (1976). As now authoritatively interpreted by the Supreme Court in duPont, supra, 430 U.S. at 126-36, 97 S.Ct. 965, section 301(b) of the Act, 33 U.S.C. § 1311(b), 1 authorizes the Environmental Protection Agency (EPA or the Agency), after orchestrating a process of study, proposal, notice, and comment, to issue two sets of progressively more stringent regulations precisely limiting the effluent discharges of every “category [or] class of [existing] point sources,” i. e., generally, every industry that pollutes the Nation’s waters. According to section 301(b), the first set of regulations must limit discharges between July 2, 1977 and July 1, 1983, inclusive, to levels characteristic of point sources utilizing “BPCTCA,” i. e., the “best practicable control technology currently available.” Section 301(b)(1)(A). The second set applies thereafter and is defined in terms of the more restricted levels of discharges from point sources, using “BA-TEA,” i. e., the “best available technology economically achievable.” Section 301(b)(2)(A). Section 304 of the Act, 33 U.S.C. § 1314, 2 establishes the minimum cross-industry cri *1020 teria that EPA must use in developing each of the two sets of industry-specific limitations, and requires EPA to identify other factors as well as the model technology (BPCTCA for 1977, BATEA for 1983) relevant to each affected industry. Jurisdiction to review both the 1977 and 1983 sets of industry-wide regulations is conferred upon the United States Courts of Appeals by section 509(b)(1) of the Act, 33 U.S.C. § 1369(b)(1). Meanwhile, the permit-issuing system established by section 402 of the Act, 33 U.S.C. § 1342, provides a procedure whereby the general effluent limitations for each class of point sources are transformed by EPA, or some EPA-approved state agency, into an authorization for a specific plant or mill to discharge effluents up to specified limits. The statute contemplates a very close correlation between the general effluent limitations promulgated by EPA and the specific discharge authorizations allowed each mill by the permit-issuing agency. The only exception to this rule is the Act’s provision of a “modification” or variance procedure under which EPA may relax the general standards somewhat in individual permits under certain limited circumstances. Section 301(c), 33 U.S.C. § 1311(c). See duPont, supra, 430 U.S. at 128, 97 S.Ct. 965. These permit-related proceedings are also judicially re viewable under the Act. Section 309(b)(1)(F), 33 U.S.C. § 1369(b)(1)(F). Finally, the regulatory circle is closed by stringent criminal and civil penalties — the latter of which are enforceable both by the government and by private citizens — for any effluents discharged in excess of those provided for in the permit. Sections 301(a), 309 of the Act, 33 U.S.C. §§ 1311(a), 1319. II. THE PAPER INDUSTRY REGULATIONS A. The regulations at issue in this case are the result of a rulemaking process developed by the Agency over the past six years for promulgating industry-wide effluent limitations under sections 301(b) and 304(b) of the Act. See notes 1 and 2 supra. See generally, La Pierre, Technology-Forcing and Federal Environmental Protection Statutes, 62 Iowa L.Rev. 771, 810-13 (1977). The procedures for these regulations began in early 1973 when EPA divided the American pulp and paper industry into two segments for purposes of establishing 1977 and 1983 effluent limitations. In “Phase I” of its rulemaking effort for the industry, it proposed, received several tiers of comments on, and promulgated 1977 and 1983 limitations for the “unbleached” segment of the industry, which produces unbleached pulp and paper. 39 Fed.Reg. 18742 (1974). *1021 This Court reviewed and upheld those regulations in full in 1976, and they are not in issue here. See American Paper Inst., supra. Promulgation of “Phase II” regulations for the apparently larger, “bleached” segment of the paper industry did not proceed with the same dispatch as in Phase I. 3 Accordingly, along with many of EPA’s responsibilities under the Act, its rulemaking for Phase II of the paper industry became the subject of law suits and settlement agreements between environmental action groups and the Agency, resulting in the establishment of court-enforced time-tables for the promulgation of each industry’s limitations. See Natural Resources Defense Council (NRDC) v. Train, 6 E.R.C. 1033 (D.D.C.1973), rev’d in part & remanded, 166 U.S.App.D.C. 312, 510 F.2d 692 (1975), timetable extended on remand, see NRDC v. Costle, 183 U.S.App.D.C. 11, 17, 561 F.2d 904, 910 n.32 (1977). These efforts apparently put EPA under a judicially enforceable obligation to issue the 1977 limitations by January 30, 1976 which it chose to accomplish by promulgating those limitations separately and more quickly than their 1983 counterparts. Consequently, it is only the 1977-related regulations that are before us. The actual promulgation procedures used in devising the challenged regulations were not unlike those used in Phase I and delineated in American Paper Inst., supra, 543 F.2d at 334-36. EPA commenced the 1977 part of the Phase II rulemaking process by commissioning two consulting firms in July 1973 to prepare a draft study suggesting possible industrial subcategories, as well as BPCTCA and achievable effluent reduction levels for each. About a year later, in August 1974, EPA made the draft study public and some 22 organizations, including many of the petitioners now before us, made it the target of comment and criticism. 4 Another year passed as the Agency responded to this new wave of solicited information by revising the draft report and by appending it, described as a “draft Development Document,” to an advance notice of proposed regulations. That notice was issued in the fall of 1975, and comments were again solicited. 40 Fed.Reg. 41300 (1975). It was followed a few months later by the issuance of a third contractor’s study of the economic impact of the proposed regulations (“Economic Analysis”). Comments on this complete package of proposals and supporting analyses were processed and reflected in “Interim Final” regulations published on February 19, 1976 and accompanied by revised drafts of the Development Document and Economic Analysis. The Agency’s obligation to comply with the January 30, 1976 deadline accounted for this quasi-final format as well as for the relatively quick turn-around time between this and the preceding public notices. See NRDC v. Train, supra, 510 F.2d at 711. This procedure also allowed petitioners in these consolidated cases to set the judicial review mechanism in motion even as they were provided — and utilized— one final opportunity to urge the Agency on its own to revise the limitations before their final issuance. The Agency’s efforts during these intermediate stages attracted thousands of *1022 pages of public comment. The industry, moreover, was offered the opportunity to introduce further testimony at a formal hearing, but it declined. The “Final Limitations” issued on January 6, 1977, with final Development Document and Economic Analysis attached. These regulations, like each of the Agency’s intermediate proposals during the rulemaking process, reflected significant changes — generally favorable to industry — in response to comments received. The present litigation commenced in early 1976 with a petition for review filed in the United States Court of Appeals for the Third Circuit by petitioners in No. 76-1.675. Subsequently, other review petitions were also filed in the Ninth Circuit but were transferred by that court to the Third Circuit under 28 U.S.C. § 2112(a). EPA then convinced the Third Circuit to transfer all of the cases to this Circuit in light of its experience with the related issues raised in the Phase I litigation, American Paper Inst., supra. This court consolidated the petitions and established briefing and oral argument schedules providing for a single presentation of the issues raised in common by all of the petitioners, as well as for supplemental presentations by individual petitioners, or groups thereof, on specialized issues. Although for ease of discourse we will generally refer collectively to the challengers before us as “petitioners”, it is important to note that the limitations have somewhat different impacts on, and accordingly have elicited some separate challenges from, different segments of the industry. In order to understand the challenges as they reflect both the common concerns of all members, and the separate concerns of individual subcategories, of the American paper industry, we turn next to a brief discussion of the manufacturing and pollution control processes typical of the industry. B. To make paper from trees is an old art; to do it without water pollution is a new science. In papermaking, logs or wooden chips must be ground up or “cooked” in one of several processes until only cellulose pulp is left. The pulp is bleached and made into various types and grades of paper. The cooking solutions and wash water that are left contain a variety of chemicals produced during “cooking” and other processes, including acids and large quantities of dissolved cellulose-breakdown products. Indeed, in some pulping processes, more of the wood is discarded in the waste water than is used to make paper. Appendix (App.) 2116. EPA has selected three parameters for measuring the pollutant content of the industry’s effluent, all of which have been used extensively in this and other industries’ measurements: total suspended solids (TSS), biochemical oxygen demand (BOD), and pH. 5 TSS reflects the total amount of solids in solution, while BOD reflects the amount of biodegradable material in solution, and pH measures the acidity of the solution. 6 *1023 EPA has divided this segment of the industry into 16 subcategories, and further subdivided it into 66 subdivisions, for the purposes of its rulemaking effort. As noted, some of petitioners’ challenges concern all of the regulations for the whole industry, while other challenges are directed to regulations for particular industry subcategories. Actually, of the 16 subcategories in the whole industry, only three — the three that use some form of the “sulfite process” —have evoked particularized challenges. 7 The reaction of sulfite mill operators stems from the limitations’ greater economic impact on them. 8 That impact in turn results from the fact that the sulfite process creates one of the highest pollution loads of any industrial process, and certainly the highest within the pulping industry. 9 In fact, the Act’s legislative history focused in particular on injury to shellfish in rivers and bays caused by sulfite wastes. A Legislative History of the Water Pollution Control Act Amendments of 1972 718-21 (1973) [hereinafter referred to as Legislative History]. Because the sulfite process is central in this case, we describe it in greater detail. In the sulfite process, wooden chips are “cooked” in hot solutions of sulphurous acid and other chemicals. The cooking dissolves the binding agent in the wood (lignin) and also a good deal of the cellulose. The end product is cellulose pulp and an acid solution called spent sulfite liquor (SSL). There are two types of sulfite processes. In the papergrade sulfite process, which produces cellulose pulp for paper, the cellulose does not need to be very pure, and moderate steps suffice for cooking and separating pulp from SSL. In the dissolving sulfite process, aimed at making the raw cellulose base for materials such as cellophane and rayon, the cellulose must be pure, so the cooking and separation of the SSL are more complete and extra bleaching of the pulp is performed. The sulfite process results in SSL and other potential pollution solutions containing acids, dissolved cellulose break-down products, and many types of organic compounds. These solutions receive various kinds of waste treatment. SSL is evaporated and burned in “SSL recovery,” eliminating the water pollution potential and producing usable heat and sometimes reusable *1024 chemicals. 10 The non-SSL waste solutions are pumped into sedimentation tanks where some of the suspended solids settle out of solution. The solutions are then pumped into other tanks and lagoons for secondary or “biological” treatment, in which bacteria are grown in the solutions to feed on and break down the waste. Eventually, the water is drained from the bacteria-laden solutions by a variety of methods. After draining, the caked solid or “sludge” that is left is incinerated or used for landfill. III. SCOPE OF REVIEW Before turning to the merits of petitioners’ challenges to the 1977 effluent limitations, some mention must be made of the appropriate scope of review under the circumstances of this case. Generally, informal rulemaking such as produced the regulations involved herein is reviewed under section 10(e)(2) of the Administrative Procedure Act (APA), 5 U.S.C. § 706(2). Pursuant to this provision, we must “set aside” any portion of the 1977 effluent limitations that is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law,” is “in excess of statutory . authority ... or short of statutory right,” or is “without observance of procedure required by law.” See American Paper Inst., supra, 543 F.2d at 338; American Frozen Food Inst., supra, 176 U.S.App.D.C. at 130-31, 539 F.2d at 132-33. See generally Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 415-17, 91 S.Ct. 814, 28 L.Ed.2d 136 (1971). In form, section 10(e)(2) embodies a list of various adjectives or adjectival phrases- — any one of which, if found applicable, requires our disapproval of the administrative action in question. These formulations are far from being entirely discrete as a matter of the ordinary meaning of language, and, indeed, are in some respects cumulative rather than differential in their applicability. This is especially true of the words “arbitrary,” “capricious,” and “abuse of discretion.” In its totality, at all events, section 10(e)(2) is indicative of a multifaceted review function committed to the courts. In the present context, for example, that function is divisible into three categories — statutory, procedural, and substantive. In the case before us, we first must determine whether the EPA regulations involved herein are “not in accordance with law,” or, more particularly, whether they are “in excess of statutory . . . authority.” APA §§ 10(e)(2)(A), (C), 5 U.S.C. § 706(2)(A), (C). We must also consider whether the process used in arriving at those regulations afforded those affected in their procedural due. More specifically, in the informal rulemaking context involved herein, this inquiry asks whether the agency gave “interested persons an opportunity to participate in the rule making through submission of written [or other] data” and whether it “incorporate[d] in the rule adopted a concise general statement of their basis and purpose.” 11 5 U.S.C. § 553. *1025 See Vermont Yankee Nuclear Power Corp. v. National Resources Defense Council, 435 U.S. 519, 98 S.Ct. 1197, 1202, 55 L.Ed.2d 460 (1978). Finally, we are charged by section 10(e)(2) of the APA to consider whether the agency “abuse[d its] discretion” (or was “arbitrary” or “capricious”) in exercising the quasi-legislative authority delegated to it by Congress, or, on the other hand, whether its “decision was based on a consideration of the relevant factors and [was not the product of] a clear error of judgment.” Citizens to Preserve Overton Park, Inc., supra, 401 U.S. at 416, 91 S.Ct. at 824. See also Vermont Yankee, supra, at 1214. Due concern both for the intent of Congress in drafting the particular statute at issue, and, more generally, for the “boundaries between the legislative and the judicial function,” Industrial Union Dep’t v. Hodgson, 162 U.S.App.D.C. 331, 339, 499 F.2d 467, 475 (1974), often demands that we exercise certain aspects of our review function with more circumspection than is appropriate to others. Turning to the case at hand, therefore, we are initially confronted with a “complicated and lengthy statute,” American Frozen Food Inst., supra, 176 U.S.App.D.C. at 111, 539 F.2d at 113, aimed at achieving the monumental “national goal” of eliminating “the discharge of pollutants into the [nation’s] navigable waters” within the short span of 13 years. Section 101(a)(1) of the Act, 33 U.S.C. § 1251. The immensity of the planning task thrust upon the Agency by Congress was heightened by the drafters’ insistence upon industry-by-industry uniformity of effluent limitations and control techniques. E. g., Legislative History, at 170 (statement of Sen. Muskie). Not only would the Agency’s regulations have to work almost immediately, but they also would have to achieve cross-industry applicability despite the geographical, technological, and economic diversity that characterizes almost every discrete sector of manufacturing and agriculture in this country. Yet, ambitious as was their goal, the drafters of the Act labored under no illusions about the uncertain state of current knowledge concerning the creation, effects, and control of water pollution. E. g., Legislative History, at 1332 (remarks of Sen. Buckley). Their intent, therefore, was to rely on EPA’s ingenuity, see, e. g., C & H Sugar Co. v. EPA, supra note 6, 553 F.2d at 286-87 — backed up by the Act’s comprehensive system of administrative rulemaking and stiff penalties — to force each industry on its own to develop the technology necessary to achieve the Act’s aspiring goal. See generally La Pierre, supra. Congress’s commitment to that goal, as well as the severe impact its achievement may have on those immediately affected, is further illustrated by the drafters’ realization that enforcement of the Act would probably shut down some plants around the nation. E. g., Legislative History, at 231 (remarks of Rep. Jones). See American Iron & Steel Inst. v. Train, 526 F.2d 1027, 1052 (3d Cir. 1976). In light of the structure and aims of the Act, and the breadth of authority delegated by it to the EPA to identify highly sophisticated control technology in an area fraught with scientific uncertainty, our review function encounters significant limitations in the substantive aspect where the given statutory standards are “arbitrary,” “capricious,” or “abuse of discretion.” First, it is elementary that our function is not to weigh de novo the available evidence and to substitute our judgment for that of the Agency. Second, an expansive concept and exercise of the review power in the eleven Courts of Appeals charged with that function could easily impede accomplishment of the Act’s ambitious pollution-ending aspiration as well as its goal of industry-by-industry uniformity. See gen *1026 erally, Currie, Judicial Review Under Federal Pollution Laws, 62 Iowa L.Rev. 1221, 1261-71 (1977). This problem looms larger in the substantive area of review than it does in those areas involving statutory interpretation and procedural integrity because of the technological and scientific uncertainty-that EPA must overcome as best it can in making the discretionary judgments delegated to it by Congress. There are also obvious limitations upon-the capacity of courts to deal meaningfully with arcane areas of knowledge of this kind. EPA has taken its responsibility quite seriously, employing no less than three highly qualified private consulting firms to augment its own technological expertise. Accordingly, we must not be too quick to draw conclusions, differing from those of the Agency, in this necessarily imprecise area of knowledge. E. g., Industrial Union Dep’t, supra, 162 U.S.App.D.C. at 338-39, 499 F.2d at 474-75 n. 18 (“Where existing methodology or research in a new area of regulation is deficient, the agency necessarily enjoys [a] broad discretion to attempt to formulate a solution to the best of its ability on the basis of available information”). See Permian Basin Area Rate Cases, 390 U.S. 747, 811, 88 S.Ct. 1344, 20 L.Ed.2d 312 (1968); Ethyl Corp. v. EPA, 176 U.S.App.D.C. 373, 396, 399, 541 F.2d 1, 24, 27 (en banc), cert. denied, 426 U.S. 941, 96 S.Ct. 2663, 49 L.Ed.2d 394 (1976); id. 176 U.S.App.D.C. at 438-439, 541 F.2d at 66-67 (Bazelon, C. J., concurring); Society of the Plastics Indus., Inc. v. OSHA, 509 F.2d 1301, 1308 (2d Cir.), cert. denied, 421 U.S. 992, 95 S.Ct. 1998, 44 L.Ed.2d 482 (1975). Indeed, the mere fact that the counsel on both sides in this suit could draw upon the opinions of diverse experts sitting virtually at their elbows during preparation of their briefs, 12 while we are forced to decide between their resulting arguments on the basis of our “generalist” judicial backgrounds, argues for restraint' on our part. 13 In sum, unless we are quite certain of our basis for doing so, we must be slow to overturn the Agency’s judgment, whichever way it may incline, when Congress has required it to act quickly and decisively despite the lack of exact data.. The likelihood of that degree of certainty on our part, we might add, almost inevitably decreases when our review is based on the record of informal rulemaking before the Agency. Because of the nature of such rulemaking, again exaggerated in this case by the complexity of the subject matter, the record’s portrayal of facts is essentially unsystematic (i. e., unaffected by any rules of evidence, or by face-to-face adversarial presentation). See Vermont Yankee, supra, 98 S.Ct. at 1217. Moreover, the greater part of the record will often transcend even an eclectic collection of facts and will instead be oriented towards aiding the Agency in making “pure[ly] legislative judgments.” Industrial Union Dep’t, supra, 162 U.S.App.D.C. at 338, 499 F.2d at 474. As such, it is relatively unhelpful to judges who have no way of testing the veracity of its “extensive and often conflicting” contents, or of analyzing it in the light of the neutral principles in which judicial decisionmaking is grounded. Id. 162 U.S.App.D.C. at 338-340, 499 F.2d at 474-76; accord FCC v. National Citizens Comm. for Broadcasting, 436 U.S. 775, 813-814, 98 S.Ct. 2096, 2121-2122, 56 L.Ed.2d 697 (1978); see Ethyl Corp. v. EPA, supra, 176 U.S.App.D.C. at 392, 541 F.2d at 20; Society of Plastics Indus., Inc., supra, 509 F.2d at 1304; Automotive Parts & Accessories Ass’n v. Boyd, 132 U.S.App.D.C. 200, 208, 407 F.2d 330, 338 (1968). In these circumstances, therefore, we will be content in carrying out our substantive review (that is, assuming the statute and *1027 the requisite procedures have been followed) first, to insist upon an explanation of the facts and policy concerns relied on by the Agency in making its decision; second, to see if those facts have some basis in the record; and finally, to decide whether those facts and those legislative considerations by themselves could lead a reasonable person to make the judgment that the Agency has made. See Citizens to Preserve Overton Park, Inc., supra, 401 U.S. at 416, 91 S.Ct. 814; Amoco Oil Co. v. EPA, 163 U.S.App.D.C. 162, 180-81, 501 F.2d 722, 740-41 (1974); Industrial Union Dep’t, supra, 162 U.S.App.D.C. at 339-40, 499 F.2d at 475-76. On the other hand, the Act — its history, intent, and the nature of the duties it delegates to the Agency and the judiciary — does not imply any derogation of the courts’ traditional primacy in interpreting statutory directives and enforcing procedural rectitude. Although calling upon the Agency to make technical judgments, the provisions of the Act that are controlling in this case use common-sense terms that are thoroughly treated in the legislative history, so that their meaning is not inherently more accessible to the Agency than to ourselves. See Texas Gas Transmission Corp. v. Shell Oil Co., 363 U.S. 263, 270, 80 S.Ct. 1122, 4 L.Ed.2d 1208 (1960); Lubrizol Corp. v. EPA, 562 F.2d 807, 816-17 & n.23 (1977). Accordingly, in this statutory area of review, and particularly in the absence of some special indication that the Agency interpretation may coincide with that of Congress, cf. Udall v. Tallman, 380 U.S. 1, 16-18, 85 S.Ct. 792, 13 L.Ed.2d 616 (1965), we are less hesitant to reject the decision of the Agency, if the latter does not comport with the conclusion dictated by established principles of statutory construction. See Adamo Wrecking Co. v. United States, 434 U.S. 275, 98 S.Ct. 566, 570 n.5, 54 L.Ed.2d 538 (1978); ASARCO, Inc. v. EPA, 188 U.S. App.D.C. 77, 83, 578 F.2d 319, 325 (1978), discussing Train v. Natural Resources Defense Council, 421 U.S. 60, 75, 95 S.Ct. 1470, 43 L.Ed.2d 731 (1975). Even more so than our review of EPA’s statutory interpretations, our review of its procedural integrity in promulgating the regulations before us is the product of our independent judgment, and our main reliance in ensuring that, despite its broad discretion, the Agency has not acted unfairly or in disregard of the statutorily prescribed procedures. E. g., National Asphalt Pavement Ass’n v. Train, 176 U.S.App.D.C. 296, 304-5, 539 F.2d 775, 783-84 (1976). Our assertion of judicial independence in carrying out the procedural aspect of the review function derives from this country’s historical reliance on the courts as the exponents of procedural fairness. 14 E. g., Hurtado v. California, 110 U.S. 516, 4 S.Ct. 111, 28 L.Ed. 232 (1884); Palko v. Connecticut, 302 U.S. 319, 58 S.Ct. 149, 82 L.Ed. 288 (1937); Goldberg v. Kelly, 397 U.S. 254, 90 S.Ct. 1011, 25 L.Ed.2d 287 (1970); Goss v. Lopez, 419 U.S. 565, 95 S.Ct. 729, 42 L.Ed.2d 725 (1975). Recently, this reliance has transcended cases arising under either of the due process clauses and has infused modern notions of administrative law, in particular in the area of informal rulemaking. E. g., Rodway v. Department of Agriculture, 168 U.S.App.D.C. 387, 391-95, 514 F.2d 809, 813-18 (1975); Portland Cement Ass’n v. Ruckelshaus, 158 U.S.App.D.C. 308, 326, 486 F.2d 375, 393 (1973), cert. denied, 417 U.S. 921, 94 S.Ct. 2628, 41 L.Ed.2d 226 (1974); see Independent Broker-Dealers’ Trade Ass’n v. SEC, 142 U.S.App.D.C. 338, 350, 442 F.2d 132, 144, cert. denied, 404 U.S. 828, 92 S.Ct. 63, 30 L.Ed.2d 57 (1971). Our reliance on careful procedural review, moreover, derives from an expectation that if the Agency, in carrying out its “essentially legislative task,” has infused the administrative process with the degree of openness, explanation, and participatory democracy required by the APA, it will *1028 thereby have “negate[d] the dangers of arbitrariness and irrationality in the formulation of rules . . .,” Automotive Parts & Accessories Ass’n, supra, 132 U.S.App.D.C. at 208, 407 F.2d at 338, quoted in National Asphalt Pavement Ass’n, supra, 176 U.S.App.D.C. at 304-05, 539 F.2d at 783-84. See Ethyl Corp. v. EPA, supra, 176 U.S.App.D.C. at 438, 541 F.2d at 66 (Bazelon, C. J., concurring), quoting International Harvester Co. v. Ruckelshaus, 155 U.S.App.D.C. 411, 448, 478 F.2d 615, 652 (1973) (Bazelon, C. J., concurring). Furthermore, such procedures will maximize the susceptibility of the record to judicial review. Dry Colors Mfrs. Ass’n v. Department of Labor, 486 F.2d 98, 105-07 (3d Cir. 1973). In short, we are willing to entrust the Agency with wide-ranging regulatory discretion, and even, to a lesser extent, with an interpretive discretion vis-a-vis its statutory mandate, so long as we are assured that its promulgation process as a whole and in each of its major aspects provides a degree of public awareness, understanding, and participation commensurate with the complexity and intrusiveness of the resulting regulations. 15 See American Frozen Food Inst., supra, 176 U.S.App.D.C. at 133, 539 F.2d at 135; note 11 supra. Even here, however, beyond the notice, comment, and explanation requirements of section 553 of the APA, it is generally up to the Agency to select among the myriad available techniques to accomplish the goal of public understanding and participation. Vermont Yankee, supra, 98 S.Ct. at 1211-17. Cf. duPont, supra, 430 U.S. at 131-32 nn.22-23, 97 S.Ct. 965; Texaco Inc. v. FEA, 531 F.2d 1071, 1082 (Em.App.), cert. denied, 426 U.S. 941, 96 S.Ct. 2662, 49 L.Ed.2d 394 (1976). 16 IV. THE AGENCY’S PROCEDURES Turning first, then, to our procedurally oriented review task, we note that in general the Agency appears to have bent over backwards to accommodate public participation in, and understanding of, the promulgation of the effluent limitations before us. No less than four opportunities for public comment on parts or all of the proposed regulations and supporting data were provided by the Agency and liberally utilized by the industry. That EPA made the hoped-for use of these procedures is evidenced by its willingness at each stage to reflect the comments in revisions of its conclusion, 17 and by the fact that members of the industry apparently exhausted most of their comments and criticisms before the Agency exhausted its procedures. Hence, the industry’s trade association rejected EPA’s offer to hold a formal hearing. The one procedural inadequacy attributed to EPA in this case is of very limited scope. It concerns EPA’s derivation of one effluent limitation (out of three in all) for one subdivision of the industry (out *1029 of 66 in all): the BOD limitation for acetate grade dissolving sulfite mills. The genesis of this challenge is complicated and somewhat obscure. It revolves around EPA’s computation of the “secondary waste load” 18 that must be treated by typical acetate grade dissolving sulfite mills using BPC-TCA. To derive a typical secondary waste load, the Agency was forced to adjust the limited data available to it, see note 70 infra, to make it representative of the appropriate mills. The dispute has arisen out of the Agency’s identification and computation of the appropriate adjustments. Petitioners contend that, properly adjusted, the figures would show that these mills produce large amounts of waste that are costly to treat. Consequently, they urge that a high cost figure must be utilized in the Agency’s cost-benefit balancing. EPA responds that there is not so much waste to treat at a typical mill, and thus that it justifiably used a lower cost estimate than the one argued for by petitioner. EPA originally computed the secondary waste load for these mills at 487 pounds of waste to be treated per ton of product (lbs/ton), App. 1514, a figure to which the industry did not object. 19 Between the publication of the Interim Final Limitations and the Final Limitations, however, the Agency apparently secured new data and recalculated the waste load as lower, i. e., 404 lbs/ton. 20 EPA purportedly arrived at the latter figure after it (1) studied two other factors that it realized would require adjusting the figure upwards and downwards respectively, but then (2) decided to disregard the two factors on the assumption that they cancelled each other out. 21 We say “purportedly” because the record includes no mention of the two factors, much less of their mode of being considered. Finally, in making its cost calculations, EPA took the 404 lbs/ton figure and reduced it to 314 lbs/ton on a further assumption that it now admits is erroneous. 22 Because the industry had no opportunity to comment on either the 314 or 404 lbs/ton figures after the Final Limitations were issued, it made its objection to them in its brief on appeal. 23 It concluded these objec *1030 tions by recalculating secondary waste load and arriving at a figure close to the Agency’s original 487 lbs/ton calculation. Brief for Petitioners in Nos. 76-1688 and 76-1690, at Appendix F. Included in its recomputations were some new data that had not been . placed in the record. Although counsel for EPA conceded in response that petitioners’ objections to the secondary waste load figures were partly correct, they also noted that petitioners’ new data were favorable to EPA’s conclusions. 24 Counsel asserted that if EPA recalculated secondary waste load based on both the new data, and on the two factors previously disregarded because they were assumed to cancel out, see note 21 supra and accompanying text, it would arrive at secondary waste load figure of 292 lbs/ton — even below its earlier computations. In sum, EPA conceded some errors and deletions but urged that by remaking its computations from scratch, it could justify the figures used in the Final Limitations. 25 The labyrinthine trail that must be followed merely to set forth this dispute— without even venturing a guess as to how to resolve it — amply illustrates the inadequacy of the procedures followed by the Agency. Most obviously, the defect stems from the inadequacy of the Agency’s final published explanation. That explanation includes, and we presume the Agency relied upon, computations now admitted to be erroneous. See note 22 supra and accompanying text. Further, it deletes mention of a phantom set of not-quite-offsetting adjustments that Agency attorneys now deem crucial to the viability of the EPA’s cost determination. See note 21 supra and accompanying text. Absent a coherent discussion — in the record — of the factual “basis” and legislative “purpose” underlying EPA’s conclusion, see 5 U.S.C. § 553, we are unable to rely on our usual assumption that the Agencywhen relying on supportable facts and permissible policy concerns and when obligated to explain itself, will rationally exercise the duties delegated to it by Congress. It is for this reason that we take no solace in the fact that the Agency’s counsel, after the fact, may be able convincingly to rationalize the Agency’s decision. E. g., Dry Color Mfrs. Ass’n, supra, 486 F.2d at 106. See Vermont Yankee, supra, 98 S.Ct. at 1214, quoting Camp. v. Pitts, 411 U.S. 138, 143, 93 S.Ct. 1241, 36 L.Ed.2d 106 (1973). Moreover, the Agency’s procedures in this limited instance improperly denied petitioners the opportunity to comment on a significant part of the Agency’s decisionmaking process as required by section 553. See American Frozen Food Inst., supra, 539 F.2d at 135, 176 U.S.App.D.C. at 133. Thus, the 404 and 314 lbs/ton figures variously relied upon by the Agency must be validated, if at all, by reference (1) to certain data on two mills obtained by the Agency after the opportunity for public comment had lapsed; (2) other data from one mill that is not reflected in the record at all but was included by petitioners in their brief; 26 and *1031 (3) calculations, also not reflected in the record, based on information scattered throughout the record. This denial of an opportunity for comment on these facts further undermines our usual assumption that notice and comment rulemaking, by virtue of its accessibility to public scrutiny, will achieve rational results. Our conclusion does not imply any dissatisfaction with the rule that the Agency need not subject every incremental change in its conclusions after each round of notice and comment to further public scrutiny before final action. E. g., International Harvester Co. v. Ruckelshaus, 155 U.S.App.D.C. 411, 424, 478 F.2d 615, 632 n.51 (1973); South Terminal Corp. v. EPA, 504 F.2d 646, 659 (1st Cir. 1974). But in this case, the Agency’s final conclusions are far from the “logical outgrowth” of the preceding notice and comment process, id., and instead are the result of a complex mix of controversial and uncommented upon data and calculations. Given the lengths that the Agency must travel to justify its revisions between the interim and final stages, we cannot be sure that further and ultimately convincing public criticism of those changes would not have been forthcoming had it been invited by the Agency. 27 See Marathon Oil Co. v. EPA, 564 F.2d 1253, 1271-72 n.54 (9th Cir. 1977). For all of the foregoing reasons, therefore, we must remand the regulation setting forth the BOD limitation for acetate grade dissolving sulfite mills. On remand, the Agency should conduct notice and comment proceedings aimed at reassessing and more fully explaining its conclusions concerning that limitation’s validity in light of the cost of complying therewith. See id. at 1271-72. V. THE AGENCY’S INTERPRETATION OF THE STATUTE Petitioners raise two major statutorily premised arguments in challenging these effluent regulations. First, they claim that the Act, as interpreted by the Supreme Court in duPont, requires a greater degree of flexibility to vary the general effluent limitations in individual permits than the regulations allow. Second, it is contended that EPA has too severely narrowed the group of factors relevant in determining allowable effluent limitations and in so doing has contravened the will of Congress as expressed in section 304(b)(1)(B), quoted in note 2 supra. These two arguments merge to the extent that EPA refuses to allow variances based on the same factors that it also refuses to •consider in setting the overall limitations. We find the Agency’s statutory interpretation correct in both instances. A. The Act explicitly includes a variance procedure for use by the permit-granting agencies in deciding how to reflect the EPA’s overall 1983 limitations in the effluent permits of specific mills. Sectioii 301(c). No similar variance clause with reference to the 1977 limitations is explicit in the Act. Nonetheless, the Agency consistently has included variance provisions in its regulations setting 1977 limits for the various industries, see United States Steel Corp. *1032 v. Train, 556 F.2d 822, 844-45 (7th Cir. 1977), and recently the Supreme Court held that under its understanding of the Act, such a variance provision is legally required: We conclude that the statute authorizes the 1977 limitations as well as the 1983 limitations to be set by regulation, 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. duPont, supra, 430 U.S. at 128, 97 S.Ct. at 975 (emphasis added and footnote omitted). See United States Steel Corp., supra, 556 F.2d at 844-45. Despite the Court’s closing phrase, which seems to validate the Agency’s actual variance practice, that question was left open in a footnote: We agree with the Court of Appeals, [E. I. duPont de Nemours & Co. v. Train, 541 F.2d 1018, at 1028 (4th Cir. 1976)], that consideration of whether EPA’s variance provision has proper scope would be premature. duPont, supra, 430 U.S. at 128 n. 19, 97 S.Ct. at 975. Thus, while duPont clearly establishes the necessity of including some provision for variances in the 1977 limitations, it raises a question as to the justiciability of the provision’s specific validity in the context of review of the general set of limitations. Our reading of duPont and of the cited portion of the lower court opinion in that case leads us to conclude that, while final review of the variance provision must await individual permittees’ attempts to utilize it, see note 29 infra, a .threshold review of the provision is a prerequisite to validation of the general limitations. This conclusion is mandated by the narrowness of the rationale relied upon by the Fourth Circuit, and adopted by the Supreme Court, as a basis for avoiding the specifics of the variance question in duPont : The administration of these [variance] provisions in practice is a matter of speculation at the present. The question will arise when a claim for a variance is made in a permit application. 541 F.2d at 1028. Accord, Natural Resources Defense Council, Inc. v. EPA, 537 F.2d 642, 647 (2d Cir. 1976). In the three years that have now elapsed since duPont was briefed and argued in the Fourth Circuit, however, enough indicia of the Agency’s attitude toward the 1977 variance provision under the Act has accumulated so that its administration is anything but “a matter of speculation.” See especially In re Louisiana-Pacific Corp., 10 E.R.C. 1841 (1977) (Decision of the Administrator of EPA). Petitioners have alleged that these recent indicia contradict the rebuttable presumption of administrative regularity that motivated the Supreme Court and Fourth Circuit in duPont to allow the limitations to go into effect in advance of final proof that the variance provision was meaningful. In petitioner’s view, the indicia point to the complete emasculation by the Agency of its 1977 variance provision. Accordingly, because the Supreme Court premised the Agency’s power to set 1977 limitations by regulation on the availability of a meaningful variance provision, see pp.--of 191 U.S.App.D.C., pp. 1033-1034 of 590 F.2d infra, and because the presumption that its variance will be applied meaningfully is no longer necessitated by a lack of concrete information, we cannot approve the regulations without finding that they include a sufficiently flexible variance provision. 28 Since its own decision in duPont, the Fourth Circuit has reached a similar *1033 conclusion and has undertaken to review the variance clause in certain effluent limitations for electric generating plants. Appalachian Power Co. v. Train, 545 F.2d 1351, 1358-60 & n. 22, modified, 545 F.2d 1380 (4th Cir. 1976). We stress, however, that our view of the variance provision in the paper industry limitations before us, while indispensable in reviewing those limitations, is quite narrow. It seeks only to establish that the provision can be applied with enough flexibility to support the general rulemaking effort. We consequently take no position on its application in specific cases, or on its precise interpretation, beyond insisting that it meet this minimum-flexibility requirement. 29 The limited question before us, therefore, is whether the variance provision included in the 1977 paper industry limitations now under review has a capacity for the degree of flexibility that duPont deemed crucial to the legality of any general set of industry-wide effluent limitations under the Act. Answering that question requires an examination of the Court’s analysis in duPont. Prior to that case, the circuits had disagreed over whether the Act even permitted EPA to bind individual point sources with general regulations under section 301 or whether the binding limits were to be set in every case by the permit-issuing agencies with jurisdiction in each of the 50 states. Compare CPC Int’l, Inc. v. Train, 515 F.2d 1032, 1038 (8th Cir. 1975), with, e. g., American Frozen Food Inst., supra. The Supreme Court upheld the Agency’s assertion of authority to issue binding, general limitations. It started with the observation that as to the 1983 limitations, section 301 of the Act “leaves no doubt that these [general] limitations are to be set by regulation.” duPont, supra, 430 U.S. at 126, 97 S.Ct. at 974. Noting, however, that “[different language is used in § 301 with respect to the 1977 limitations,” id. at 127, 97 S.Ct. at 9741, see id. at 133-34 n. 24, 97 S.Ct. 965, the Court nonetheless decided that two sets of limitations could be established by the same procedures, “so long as some allowance is made for variations in individual plants” in the 1977 limitations. Id. at 127-28, 97 S.Ct. 965. The Court apparently insisted upon reading a variance requirement into section 301’s provisions with respect to the 1977 limitations in order to make those provisions coextensive with the section’s 1983 provisions — which do include a variance requirement, section 301(c) — and to “unambiguously” invest the Agency with rulemaking authority. 430 U.S. at 127, 97 S.Ct. 965. The importance that the Court assigned to a meaningful variance as a prerequisite to valid general limitations may be seen first in its use of the mandatory phrase “so long as.” Moreover, in the same case the Court unequivocally refused to read a variance requirement into the Act’s provisions governing effluent regulations for newly built plants. 430 U.S. at 137-39, 97 S.Ct. 965. Here the Court found “that Congress intended these regulations to be absolute prohibitions,” id. at 138, 97 S.Ct. at 980 — a *1034 finding it did not make with respect to the 1977 limitations despite the Act’s failure explicitly to provide for variances therefrom. Finally, the Court emphasized the need for a meaningful variance when it argued that giving EPA the power to set general regulations would not undermine the Act’s intent to give the states, through the permit-granting agencies, an important role in administering the Act. The Court subscribed to the view that, by leaving the granting of variances to the state agencies in the first instance, the significance of their role would be preserved. Id. at 134 n. 24, 97 S.Ct. 965, citing American Meat Inst. v. EPA, 526 F.2d 442, 452 (7th Cir. 1975). Accord, American Paper Inst., supra, 177 U.S.App.D.C. at 190, 543 F.2d at 337. See generally Train v. Natural Resources Defense Council, Inc., 421 U.S. 60, 78-87, 95 S.Ct. 1470, 43 L.Ed.2d 731 (1975). Because the states can only grant variances that conform to EPA’s interpretation of its variance provision, see In re Louisiana-Pacific Corp., supra, the states’ role could still be severely undercut if the Agency were to interpret the variance allowance so narrowly as to remove its effect. The motivating force in duPont was the Court’s desire to overcome the “highly anomalous” result that would attend a reading of section 301 to require a different pattern of promulgation for the 1977 limitations than the one so clearly laid out for the 1983 limitations. Hence, the Act was read, despite its nonparallel language, to establish the same promulgation procedures for both sets of limitations and, accordingly, to require a variance provision under both. At minimum, therefore, duPont indicates that the Agency must give permittees the ability to secure variances from the 1977 limitations analogous to their statutorily provided ability to secure the same with respect to the 1983 standards. 30 Section 301(c) of the Act allows a modification of the general 1983 limitations in section 301(b)(2)(A) upon a showing . . . that [a variance] (1) will represent the maximum use of technology within the economic capability of the owner or operator; and (2) will result in reasonable further progress toward the elimination of the discharge of pollutants. 31 The crucial language here appears to be “technology within the economic capability of the owner.” This language clearly harks back to the “best available technology economically achievable for such category [of point sources]” language in section 301(b)(2)(A) which provides for the establishment of general 1983 limitations. 32 In *1035 other words, section 301(c) (if properly invoked) authorizes the Agency to relieve a particular point source operator from any demands that the Act does not allow the Agency to make of the industry generally. That is to say, it may relieve him from the requirement of using more than the best available technology economically achievable (BATEA). Even the second qualification in section 301(c), that the variance may not halt progress toward eliminating pollution, tracks language in section 301(b)(2)(A) requiring that BATEA “result in reasonable further progress toward the national goal of eliminating the discharge of all pollutants.” See note 32 supra. Furthermore, the Act quite explicitly lays out the minimum factors that the Agency must consider in identifying BATEA. Those factors, referred to in section 301(b)(2)(A), are found in section 304(b)(2)(B). 33 As such, it seems clear that in deciding under section 301(c) whether the variance sought by a mill owner “representé] the maximum use of technology within the economic capability of [that] owner,” the permit-granting agency, and the EPA in supervising that agency, must consider the factors laid out in section 304(b)(2)(B). Analogously, the 1977 variance provision must at minimum allow a petitioning mill operator to seek a dispensation from any limitation that, as a whole, demands more of him than section 301(b)(1)(A), the 1977 limitation provision, allows EPA to demand of the industry as a whole. Although this formulation ensures a meaningful variance, it should be noted that it is not a license for avoidance of the Act’s strict pollution control requirements. It simply allows individual operators to argue, that, given the overall impact of an effluent limitation on their operations, they are faced with stricter requirements than the Act authorizes EPA to place on the industry as a whole. 34 More specifically, section 301(b)(1)(A), as interpreted in duPont, requires “the application of the best practicable control technology currently available” (BPCTCA) for each industrial subcategory. See note 1 supra. Moreover, it, too, refers to a portion of section 304 that enumerates the factors relevant in setting BPCTCA. Section 304(b)(1)(B); see note 2 supra. Under duPont, therefore, a variance provision should allow the state agencies and EPA to excuse mill operators from making more than the maximum use of technology practicably available to them. Pursuant to section 304(b)(1)(B), the outlines of practicability in each case depend upon “the total cost [to the operator] of application of technology in *1036 relation to the effluent reduction benefits to be achieved” by that technology as well as upon “the age of [the mill operator’s] equipment . . . the process employed [by him], the engineering aspects of the application [at the mill] of various types of control techniques, process changes [and] non-water quality environmental impact [of the technology specified for his mill] (including energy requirements).” The precise interrelationship of these factors in making variance decisions, moreover, basically should track the scope and interrelationship that we have assigned them in developing the general, industry-wide limitations. See pp.---of 191 U.S.App.D.C., pp. 1041-1053 of 590 F.2d infra. As will be explained more thoroughly infra, this approach clearly rules out the necessity of considering local receiving water quality in making variance decisions. See In re Louisiana-Pacific, supra; pp.- - of 191 U.S.App.D.C., pp. 1041-1044 of 590 F.2d infra. A more difficult question surrounds the relevance and importance of economic hardship. This issue is crucial, of course, because those mill operators who are most hard pressed economically will be the most likely to pursue vigorous variance demands. Moreover, when faced with the ultimate threat of economic hardship — plant closure, with attendant unemployment and regional economic dislocation — the local permit-granting agency will find it difficult to resist a plea for a variance. We have explored this issue carefully, and we express our conclusion emphatically: Although the “total cost" of pollution control at the petitioning mill must be considered under a satisfactory variance provision, it is only relevant “in relation to the effluent reduction benefits to be achieved” at that mill, section 304(b)(1)(B); so long as those costs relative to the pollution reduction gains are not different from those that may be imposed on the industry as a whole, the difficulty, or in fact the inability, of the operator to absorb the costs need not control the variance decision. 35 We reach this conclusion under the statute only after satisfying ourselves that the legislative intent is as clear as the result is harsh. Most prominently, the Act’s supporters in both Houses acknowledged and accepted the possibility that its 1977 requirements might cause individual plants to go out of business. E. g., Legislative Histo *1037 ry, at 231 (remarks of Rep. Jones); id. at 1282 (remarks of Sen. Bentsen). They self-consciously made the legislative determination that the health and safety gains that achievement of the Act’s aspirations would bring to future generations will in some cases outweigh the economic dislocation it causes to the present generation. See American Iron & Steel Inst., supra, 526 F.2d at 1052. They accordingly authorized EPA to impose effluent restrictions that they knew might shut down parts of regulated industries — including, specifically, paper mills. See Legislative History, at 718-21 (statement of Rep. Meeds). The Agency, in turn, has projected that its limitations for the paper industry may shut down eight marginal mills, see p. - of 191 U.S.App.D.C., p. 1047 of 590 F.2d infra, and the variance provision need not protect these or other individuals from impacts authorized for the industry as a whole. Even more specifically, at least one legislator considered and rejected “giving variances to pollution controls based on [such purely] economic grounds” as [w]hen the otherwise healthy small independent firm is forced to close, the tax base of the community is further weakened, and workers and families are forced back into already congested metropolitan areas . Legislative History, at 1355 (remarks of Sen. Nelson). Despite the appeal of economic hardship variances in circumstances such as those described, Senator Nelson argued, they could become “a tool used by powerful political interests to obtain so many exemptions on the flimsiest of pretenses [that] tragic delay in stopping the destruction of our environment” would result. Id. Hence, he opposed such variances and chose to rely on other means of protecting businesses from effluent-control-induced bankruptcy. See id., discussing section 8 of the Act, amending Small Business Act, § 7,15 U.S.C. § 636, now codified in 15 U.S.C. § 636(g) (authorizing low interest loans to aid small businesses in meeting the Act’s requirements). 36 We come, then, to the question of whether the actual variance provision included in the Phase II paper industry limitations, as interpreted, is capable of the minimum degree of flexibility just describ *1038 ed. 37 That provision notes that the EPA was at pains to account for all available data “such as age and size of plant . energy requirements and costs ” relevant to effluent levels but acknowledges the inevitable possibility “that data which would affect these limitations” have not been considered (emphasis added). It accordingly allows a discharger to secure a permit containing less stringent limitations than those established by the EPA, if the discharger can show “that factors related to the equipment or facilities involved, the process applied, or other such factors related to such discharger are fundamentally different from the factors considered in the establishment of the guidelines” (emphasis added). Not surprisingly, EPA’s likely interpretation of these terms, and particularly the parts of it italicized above, have occasioned some dispute. The Agency itself at one point accepted public comments on this type of 1977 variance under the Act, looking toward a regulation definitively interpreting it. 39 Fed.Reg. 28926 (1974). No such regulation has been forthcoming, however. Nonetheless, in 1974, the Agency’s General Counsel instructed Agency personnel that the “other such factors” language in the typical 1977 variance provision quoted above did not envision the consideration of any economic factors. He accordingly limited the provision’s application to cases involving fundamentally different “factors of a technical and engineering nature.” Memorandum to Regional Administrators of EPA, 39 Fed.Reg. 30073 (1974) (emphasis added). It was this interpretation that the Fourth Circuit disapproved of in Appalachian Power Co., supra, discussed in notes 30 & 35 supra. Because the General Counsel’s 1974 opinion relied exclusively on Congress’s failure to include a variance procedure for the 1977 limitations we, too, would be inclined to find the opinion inconsistent with the Act, as now authoritatively interpreted in duPont to include a meaningful variance requirement. See Currie, Congress, The Court, and Water Pollution, 1977 S.Ct.Rev. 39, 53-56. Nonetheless, we need not reach that question in light of the Agency’s recent reinterpretation and expansion of the variance provision. As of September 13, 1977, when EPA filed its brief in this case, the Agency still adhered to its 1974 interpretation of the variance provision to exclude consideration of economic factors. Brief for Respondent at 17-20. Moreover, that brief suggested that the provision’s “fundamental difference” language required that the applicant demonstrate a qualitatively different factor than any considered by the Agency in its rulemaking before a variance could be granted. Id. at 12-17. That is to say, if the Agency had considered data relevant, for example, to climate, or to energy needs for sludge disposal, the brief indicated that no variance based on those factors could subsequently issue, even if the climate or energy situation facing the applicant was quantitatively different from the range of such situations considered by the Agency. *1039 Nonetheless, two days after the brief was filed in this case, the Administrator of the EPA handed down a decision in In re Louisiana-Pacific, supra, which seems significantly to have liberalized the Agency’s approach to the variance provision. 38 In essence, this decision tracked the analysis in duPont and thereby expanded the range of factors recognized by EPA as relevant to the granting of a variance to include all of those factors listed in section 304(b)(1)(B): My authority to provide for variances from [BPCTCA] flows from, and is inherent in, my authority to promulgate effluent limitations guidelines under Sections 301(b)(1)(A) and 304(b)(1) . . .. Variances can only be based on fundamental differences in factors which are appropriate to [the general] technology-based regulations and limitations derived through the variance process must still meet the congressional definition of best practicable control technology. . This does not mean that where a fundamental difference can be shown with respect to a factor other than [one not listed in section 304(b)(1)] a variance may not be appropriate. Id. at 1851. See also id. at 1847 (describing section 304(b) of the Act as the source of “factors which must be taken into account in developing [the general] industrial effluent limitations . . ..”). Consistent with his reliance on section 304(b), the Administrator clearly broke with the 1974 opinion of the General Counsel and cited such nontechnical factors relevant to the 1977 variances as energy requirements, id. at 1851, raw materials, id. at 1846 n. 9, non-water quality environmental impacts, such as land and air pollution from sludge disposal or burning, id. at 1850-51 & n. 22, 1852-53 & n. 30, and even such economic factors as compliance costs, id. at 1051-52 & n. 27. 39 As a matter of the potentially relevant factors, therefore, the Agency’s interpretation has been brought into line with what we believe are the dictates of duPont, and on that score we find the variance clause capable of the requisite minimum degree of flexibility. The Louisiana-Pacific decision also clarifies the Agency’s position with respect to the meaning of the “fundamentally different” language in the variance clause. First, it makes clear that, if an individual operator demonstrates a “substantial[ ]” id. at 1851, or “fundamental” difference in a section 304(b)(1)(B) factor vis-a-vis the Agency’s regulatory findings about the factor “on a national basis,” a variance will be allowed. Id. at 1851 n. 25, 1852-53. Since EPA must, and claims that it did, consider all of the section 304(b)(1)(B) factors in setting the general limitations, the Administrator’s decision indicates that factors already considered during rulemaking can, and in fact must, be considered during variance proceedings, so long as the requisite *1040 difference is shown. Here again, the degree of flexibility necessary to justify EPA in setting general 1977 limitations is achieved, and variances are not restricted to qualitatively different circumstances from those considered by EPA during rulemaking. The remaining question, therefore, concerns the requirement that the difference between the individual and national situations (whether qualitative or quantitative) is fundamental. None of the petitioners in this case takes issue with the “fundamental difference” language so long as it is not used to preclude consideration of any of the section 304(b)(1)(B) factors. Petitioners realize that such a requirement, along with the allocation of the burden of proof to the variance applicant, assures that the pin-hole safety valve envisioned in the Act and duPont does not become a yawning loophole. E. g., Joint Brief for Petitioners Addressing Common Issues, at 24. We agree with the parties to these petitions that the fundamentality requirement does not deprive the variance provision before us of the minimum potential for flexibility required by duPont. Although the variance must prevent the regulations from having a greater overall impact on an individual mill than the Act authorizes the general regulations to have on the industry, the one designed by EPA for use in its industry-wide limitations for 1977 accomplishes this goal. Because EPA, in devising the limitations, undertook a meticulous effort to obtain all relevant information from all available sources including the industry itself, and attempted to account for that information in all its diversity, the Agency has built a significant degree of flexibility into the regulations themselves. This flexibility is reflected in the 16 subcategories and 66 subdivisions thereof, as well as in the establishment of maximum single-day, and 30-day-average, limits that are much higher than the yearly-average limits that the Agency has found within the technological reach of the industry. Thus, to a great degree, the Agency has accounted for cross-industry, and even “cross-subcategory,” differences in establishing the limits. Allowing for variances based on slight or moderate differentials at individual plants would accordingly ignore the liberality that is already built into the system. It would allow for variances, when the impact on an individual did not exceed the range of impacts considered by the Agency for the industry generally. Moreover, without the fundamentality requirement the rulemaking process could be shortcircuited. As discussed earlier, Congress has placed EPA under the burden of almost instantaneously establishing comprehensive regulations aimed at achieving a monumental goal. That effort must proceed in advance of the availability of much relevant data. If the regulations could be ignored every time a mill owner — who did not produce data from his operation during rulemaking — suddenly develops the facts and finds them somewhat different from the ones available to the Agency, the achievement of the congressional goals would be pushed far beyond the time periods established in the Act. Finally, the Agency, without destroying the necessary flexibility, may insist upon a fundamental difference as to one or even several individual factors, in order to account for the fact that each limit is the product of consideration of myriad factors, each of which may be expected to vary from plant to plant. As noted earlier, a variance only need be granted when the overall situation facing an individual operator differs from the overall situation of the industry. See note 34 supra and accompanying text. The Agency has properly adopted an approach that focuses on only one or a few of the relevant factors — but sets a high differential standard therefor— in order to relieve itself of having to analyze all of the relevant factors in every variance case to achieve the overall differential picture. In sum, the most recent delineation by the Agency of its 1977 variance policy under the Act convinces us that its policy is capable of sufficient flexibility to buttress its claim of authority to limit 1977 effluent *1041 discharges by way.of general regulations. We emphasize, once more, however, that this conclusion is inextricably linked to the Agency’s current interpretation of its variance provision, and that, while crucial to our affirmance of the regulations as a whole, it implies no opinion on the specifics of individual variance applications. B. EPA’s consideration of the factors bearing on “the best practicable technology currently available” (BPCTCA) has inspired several challenges from petitioners. Some of these challenges concern the Agency’s refusal to consider receiving water quality, while others concern EPA’s manner of assessing the factors that all agree must be considered: cost and nonwater environmental impacts. We uphold the Agency’s interpretation and application of the statute against both sets of challenges. 1. Some of the paper mills that must meet the effluent limitations under review discharge their effluents into the Pacific Ocean. Petitioners contend that the ocean can dilute or naturally treat effluent, and that EPA must take this capacity of the ocean (“receiving water capacity”) into account in a variety of ways. 40 They urge what they term “common sense,” /. e., that because the amounts of pollutant involved are small in comparison to bodies of water as vast as Puget Sound or the Pacific Ocean, they should not have to spend heavily on treatment equipment, or to increase their energy requirements and sludge levels, in order to treat wastes that the ocean could dilute or absorb. 41 EPA’s secondary response to this claim was that pollution is far from harmless, even when disposed of in the largest bodies of water. As congressional testimony indicated, the Great Lakes, Puget Sound, and even areas of the Atlantic Ocean have been seriously injured by water pollution. 42 Even if the ocean can handle ordinary wastes, ocean life may be vulnerable to toxic compounds that typically accompany those wastes. 43 In the main, however, EPA simply asserted that the issue of receiving water capacity could not be raised in setting effluent limitations because Congress *1042 had ruled it out. We have examined the previous legislation in this area, and the 1972 Act’s wording, legislative history, and policies, as underscored by its 1977 amendments. These sources, which were thoroughly analyzed in a recent opinion of the administrator of the Agency, fully support EPA’s construction of the Act. 44 They make clear that based on long experience, and aware of the limits of technological knowledge and administrative flexibility, Congress made the deliberate decision to rule out arguments based on receiving water capacity. The earliest version of the Federal Water Pollution Control Act was passed in 1948 and amended five times before 1972. Throughout that 24 year period, Congress attempted to use receiving water quality as a basis for setting pollution standards. W. Rodgers, Environmental Law 355-57 (1977). At the end of that period, Congress realized not only that its water pollution efforts until then had failed, but also that reliance on receiving water capacity as a crucial test for permissible pollution levels had contributed greatly to that failure. EPA v. State Water Resources Control Board, 426 U.S. 200, 202, 96 S.Ct. 2022, 48 L.Ed.2d 578 (1976). 45 Based on this experience, Congress adopted a new approach in 1972. Under the Act, “a discharger’s performance is measured against strict technology-based effluent limitations — specified levels of treatment — to which it must conform, rather than against limitations derived from water quality standards to which it and other'polluters must collectively conform.” Id. at 204-05, 96 S.Ct. at 2024 (footnotes omitted). See Save the Bay, Inc. v. EPA, 556 F.2d 1282, 1284 (5th Cir. 1977); American Frozen Foods Inst., supra, 176 U.S.App.D.C. at 113, 539 F.2d at 115. This new approach reflected developing views on practicality and rights. Congress concluded that water pollution seriously harmed the environment, and that although the cost of control would be heavy, the nation would benefit from controlling that pollution. Yet scientific uncertainties made it difficult to assess the benefits to particular bodies of receiving water. Even if the federal government eventually could succeed at the task at which had failed for 24 years and thus could determine benefits and devise water quality standards, Congress concluded that the requisite further delay was too long for the nation to wait. Note, The Federal Water Pollution Control Act Amendments of 1972: Ambiguity as A Control Device, 10 Harv.J.Legis. 565, 571-72 (1973). Moreover, by eliminating the issue of the capacity of particular bodies of receiving water, Congress made nationwide uniformity in effluent regulation possible. Congress considered uniformity vital to free the states from the temptation of relaxing local limitations in order to woo or keep industrial facilities. 46 In addition, national uniformity made pollution clean-up possible without engaging in the divisive task of favoring some regions of the country over others. *1043 More fundamentally, the new approach implemented changing views as to the relative rights of the public and of industrial polluters. Hitherto, the right of the polluter was pre-eminent, unless the damage caused by pollution could be proven. Henceforth, the right of the public to a clean environment would be pre-eminent, unless pollution treatment was impractical or unachievable. The Senate Committee declared that “[t]he use of any river, lake, stream or ocean as a waste treatment system is unacceptable” — regardless of the measurable impact of the waste on the body of water in question. Legislative History at 1425 (Senate Report). The Conference Report stated that the Act “specifically bans pollution dilution as an alternative to waste treatment.” Id. at 284. This new view of relative rights was based in part on the hard-nosed assessment of our scientific ignorance: “we know so little about the ultimate consequences of injection of new matter into water that [the Act requires] a presumption of pollution. . . . ” Id. at 1332 (remarks of Sen. Buckley). It also was based on the widely shared conviction that the nation’s quality of life depended on its natural bounty, and that it was worth incurring heavy cost to preserve that bounty for future generations. The Act reflects the new approach in a number of provisions. As noted, its goal, was zero discharge of pollutants by 1985, section 101(a)(1), 33 U.S.C. § 1251(a)(1), not discharges at acceptable or tolerable levels for receiving water. The rest of the statute “authorize[s] a series of steps to be taken to achieve [that] goal,” duPont, supra, 430 U.S. at 116, 97 S.Ct. at 969. It defines “pollution,” “pollutant,” “discharge of a pollutant,” and “effluent limitation” in terms of any addition to water that alters its “chemical, physical, biological, [or] radiological integrity;” it does not specify additions that diminish the quality of the receiving water. 47 In only one limited instance, thermal pollution, is receiving water capacity to be considered in relaxing standards, and the section allowing such consideration was drafted as a clear exception. Section 316(a) of the Act, 33 U.S.C. § 1326(a). Otherwise, receiving water quality was to be considered only in setting “more stringent” standards than effluent limitations otherwise would prescribe. Section 301(b)(1)(C) of the Act, 33 U.S.C. § 1311(b)(1)(C) (emphasis added). The Act was passed with an expectation of “mid-course corrections,” Legislative History, at 175 (statement of Sen. Muskie), and in 1977 Congress amended the Act, although generally holding to the same tack set five years earlier. Pub.L. No. 95-217, 91 Stat. 1584. Notably, during those five years, representatives of the paper industry had appeared before Congress and urged it to change the Act and to incorporate receiving water capacity as a consideration. See, e. g., Hearings before the Subcomm. on Environmental Pollution of the Senate Comm, on Environment and Public Works, 95th Cong., 1st Sess., pt. 3, at 193, 195, 540. Nonetheless, Congress was satisfied with this element of the statutory scheme. Except for a provision specifically aimed at discharges from “publicly owned treatment plants,” section 301(h) of the Act, 33 U.S.C. § 1311(h), 48 it resolved in the recent amend *1044 ments to continue regulating discharges into all receiving waters alike. 49 Our experience with litigation under the Act, and particularly with this case, emphasizes the weight of Congress’ policies. Even without receiving water capacity as an issue to delay it, EPA was late in promulgating these regulations. See pp.---of 191 U.S.App.D.C., pp. 1020-1024 of 590 F.2d supra. We have wrestled with the problems of weighing technological imponderables and can understand the greater difficulties that would have arisen if the receiving water issues involving even greater imponderables had also been involved. Historically, the paper industry itself, and particularly the sulfite process sector, avoided the impact of regulation because of the difficulty of proving that its discharges adversely affected receiving water. 50 Under the new statutory scheme, Congress clearly intended us to avoid such problems of proof so that a set of regulations with enforceable impact is possible. The dangers of ignoring this congressional mandate are clearly revealed by the one experiment in the Act with allowing consideration of receiving water capacity. As we have noted, thermal pollution regulation is the only area where the 1972 Act explicitly allowed receiving water capacity to continue as an issue. In reviewing the results of that experiment during consideration of the recent amendments to the Act, Congress found that the water capacity issue had led to a regulatory breakdown. “Heat has thus become an unregulated pollutant, clearly not the intent of the Congress. That limited exemption has been turned into a gaping loophole.” S.Rep. No. 370, 95th Cong., 1st Sess. 8, reprinted in [1977] U.S.Code Cong. & Admin.News, pp. 4326, 4334. Given the clarity of Congress’ desire not to allow the receiving water capacity loophole to engulf its overall regulatory efforts in this area, we affirm the Agency’s refusal to consider water quality in setting its limitations. 2. Petitioners also challenge EPA’s manner of assessing two factors that all parties agree must be considered: cost and non-water quality environmental impacts. They contend that the Agency should have more carefully balanced costs versus the effluent reduction benefits of the regulations, and that it should have also balanced those benefits against the non-water quality environmental impacts to arrive at a “net” environmental benefit conclusion. Petitioners base their arguments on certain comments made by the Conferees for the Act, see notes 52 & 67 infra, and on the fact that the Act lists non-water quality environmental impacts as a factor the Agency must “take into account.” *1045 In order to discuss petitioners’ challenges, we must first identify the relevant statutory standard. Section 304(b)(1)(B) of the Act, 33 U.S.C. § 1314(b)(1)(B), identifies the factors bearing on BPCTCA in two groups. 51 First, the factors shall include consideration of the total cost of application of technology in relation to the effluent reduction benefits to be achieved from such application, and second, they 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 (including energy requirements), and such other factors as the Administrator deems appropriate^] See note 2 supra. The first group consists of two factors that EPA must compare: total cost versus effluent reduction benefits. We shall call these the “comparison factors.” The other group is a list of many factors that EPA must “take into account:” age, process, engineering aspects, process changes, environmental impacts (including energy), and any others EPA deems appropriate. We shall call these the “consideration factors.” Notably, section 304(b)(2)(B) of the Act, 33 U.S.C. § 1314(b)(2)(B), which delineates the factors relevant to setting 1983 BATEA limitations, tracks the 1977 BPCTCA provision before us except in one regard: in the 1983 section, all factors, including costs and benefits, are consideration factors, and no factors are separated out for comparison. Based on our examination of the statutory language and the legislative history, we conclude that Congress mandated a particular structure and weight for the 1977 comparison factors, that is to say, a “limited” balancing test. 52 In contrast, Congress did not mandate any particular structure or weight for the many consideration factors. Rather, it left EPA with discretion to decide how to account for the consideration factors, and how much weight to give each factor. In response to these divergent congressional approaches, we conclude that, on the one hand, we should examine EPA’s treatment of cost and benefit under the 1977 standard to assure that the Agency complied with Congress’ “limited” balancing directive. See note 52 supra. On the other hand, our scrutiny of the Agency’s treatment of the several consideration factors seeks to assure that the Agency informed itself as to their magnitude, and reached its own express and considered conclusion about their bearing. More particularly, we do not believe that EPA is required to use any specific structure such as a balancing test in assessing the consideration factors, nor do we believe that EPA is required to give each consideration factor any specific weight. Our conclusions are based initially on the section’s wording and apparent logic. By singling out two factors (the comparison factors) for separate treatment, and by requiring that they be considered “in relation to” each other, Congress elevated them to a *1046 level of greater attention and rigor. Moreover, the comparison factors are a closed set of two, making it possible to have a definite structure and weight in considering them and preventing extraneous factors from intruding on the balance. By contrast, the statute directs the Agency only to “take into account” the consideration factors, without prescribing any structure for EPA’s deliberations. As to this latter group of factors, the section cannot logically be interpreted to impose on EPA a specific structure of consideration or set of weights because it gave EPA authority to “upset” any such structure by exercising its discretion to add new factors to the mix. Instead, the listing of factors seems aimed at noting all of the matters that Congress considered worthy of study before making limitation decisions, without preventing EPA from identifying other factors that it considers worthy of study. So long as EPA pays some attention to the congressionally specified factors, the section on its face lets EPA relate the various factors as it deems necessary. The legislative history reveals that clear congressional policies support the section’s facial structure. The original House and Senate versions of the section differed significantly. A major point of contention between the two versions involved the House’s stronger concern over the economic effects of imposing stringent effluent limitations. Ultimately a compromise was ■reached at Conference and accepted by both houses. It provided, first, that the Agency must use “limited” cost-benefit balancing in deriving 1977 standards, but not in arriving at the 1983 standards. Legislative History, at 170 (statement of Sen. Muskie). A “mid-course” evaluation was then to occur in the mid- to late-1970’s, after the Act had been in effect for several years but before full implementation of the 1983 standards. Id. at 175. The latter step was designed to allow Congress to rewrite- the 1983 standards in order to continue the cost-benefit balancing during that period as well, if Congress found after early experience that such a course was necessary. Section 315 of the Act, 33 U.S.C. § 1325 (establishing a National Study Commission to consider mid-course changes). Thus the fact that Congress indicated its greater concern for cost-benefit calculation in the short run by making cost and benefit “comparison factors” for 1977, but only “consideration factors” for 1983, demonstrates the more relaxed view it took of EPA’s treatment of consideration factors relevant to both the 1977 and 1983 standards. Indeed, after studying the need for mid-course correction, Congress decided to retain the 1972 arrangement, which, except for limited modifications not germane here, gives the Agency more complete discretion in considering the relevant factors in the future. 53 Judicial decisions have carefully observed that the cost and benefit factors require more rigorous EPA consideration of cost versus benefit in the 1977 standards than in the 1983 standards. American Paper Inst., supra, 177 U.S.App.D.C. at 191, 543 F.2d at 338 (cost-benefit balancing in 1977, not in 1983); accord American Frozen Food Inst., supra, 176 U.S.App.D.C. at 117, 539 F.2d at 119; American Meat Inst., supra, 526 F.2d at 462-63; see W. Rodgers, supra at 466. But see Appalachian Power Co., supra, 545 F.2d at 1361, criticized in La Pierre, supra, 62 Iowa L.Rev. at 819-20. Since the consideration factors specified in the 1977 provision track the language of all of the factors — including cost and benefit — in the 1983 provision, these cases support a lower level of administrative rigor and judicial review with respect to the 1977 considera *1047 tion factors than with respect to the costs and benefit factors relevant to that year’s limitations. We are fortified in this view by its coincidence with the Agency’s interpretation. As the Supreme Court held last year, the EPA’s construction of this Act is entitled to some deference. duPont, supra, 430 U.S. at 134-35, 97 S.Ct. 965. Consequently, we must review the comparison factors to determine if EPA weighed them through the “limited” balancing test as intended by Congress. On the other hand, we may review the consideration factors only to determine if EPA was fully aware of them and reached its own express conclusions about them. Since the two types of factors are separate, we divide our discussion accordingly. a. Petitioners do not challenge the cost-benefit analysis for the whole industry. They do, however, challenge the analysis for the sulfite sector, contending that EPA used an “overall” instead of an “incremental” method of balancing, and that its figures on the cost of BPCTCA for the dissolving sulfite subcategory were underestimates. We uphold EPA’s determination against both contentions. EPA’s approach was similar to the one we upheld in American Paper Inst., supra, 177 U.S.App.D.C. at 191-92, 543 F.2d at 338-39. The Agency assessed the costs of internal and external effluent treatment measures, not only for the industry, but also for each subcategory. This included a separate cost assessment for the sulfite subcategories. App. 2380-83. An economic analysis was prepared to determine the impact of the costs on the industry. App. 1211-448. It found that the industry as a whole would readily absorb the cost of compliance with the 1977 standards, estimated at $1.6 billion. Out of 270 mills employing 120,000 people, eight mills would likely be closed and 1800 people laid off. The Agency noted that the impact on the three heavily polluting sulfite subcategories would be the greatest. Of less than 30 sulfite mills, three would probably close, resulting in 550 people being laid off. Against these costs, EPA balanced the main effluent reduction benefit: overall 5,000 fewer tons per day of BOD discharged into the nation’s waters. 54 EPA refined this balance by calculating the cost per pound of BOD removed for each subcategory. App. 2478-79. Although sulfite mills must make large investments in waste treatment facilities, the cost-benefit balance is favorable for the limitations on these mills, because of the large volume of waste they produce and thus the greater treatment efficiency. See note 8 supra. Petitioners’ first contention is that EPA not only should have calculated the overall cost-benefit balance, but also should have made an “incremental” calculation of that balance. More precisely, they contend that EPA must undertake to measure the costs and benefits of each additional increment of waste treatment control, from bare minimum up to complete pollution removal. In support of this contention, they point to Senator Muskie’s description of cost-benefit balancing,- which suggests a focus on the “additional degree” or “marginal” amount of effluent reduction. See note 52 supra. Petitioners concede that we accepted EPA’s calculation of the overall cost-benefit balancé, without any further marginal or incremental analysis, in American Paper Inst., supra, 177 U.S.App.D.C. at 191, 543 F.2d at 338. Nonetheless, they suggest that the present case can be distinguished, because in these proceedings, unlike in American Paper Inst., industry representatives submitted an incremental breakdown of costs and benefits to the Agency. 55 *1048 The failure of American Paper Inst, to require EPA to perform its own incremental analysis is justified for a number of reasons beyond some oversight on the part of paper industry petitioners in that case. While EPA has no discretion to avoid cost-benefit balancing for its 1977 standards, it does have some discretion to decide how it will perform the cost-benefit balancing task. “[E]ven with th[e] 1977 standard, the cost of compliance was not a factor to be given primary importance,” American Iron & Steel Inst., supra, 526 F.2d at 1051, and, as such, cost need not be balanced against benefits with pinpoint precision. A requirement that EPA perform the elaborate task of calculating incremental balances would bog the Agency down in burdensome proceedings on a relatively subsidiary task. Hence, the Agency need not on its own undertake more than a net cost-benefit balancing to fulfill its obligation under section 304. However, when an incremental analysis has been performed by industry and submitted to EPA, it is worthy of scrutiny by the Agency, for it may “avoid the risk of hidden imbalances between cost and benefit.” Id. at 1076 n. 19 (Adams, J., concurring). If such a “hidden imbalance” were revealed here, and if the Agency had ignored it, we might remand for further consideration. But in this case the incremental analysis proffered by industry showed that the last and most expensive increment of BOD treated in sulfite mills cost less than $.15 per pound of BOD removed, which is below the average cost of treatment in most of the industry’s subcategories. App. 3545-51B. See note 8 supra. We would be reluctant to find that EPA had ignored a “hidden imbalance” when the most unfavorable incremental cost-benefit balance that is challenged falls well within the range of averages for the industry as a whole. 56 Petitioners’ next contention is that EPA underestimated the cost of BPCTCA for the dissolving sulfite subcategory, by misfiguring the cost of SSL recovery. 57 We have examined petitioners’ contention, bearing in *1049 mind that we do not review EPA’s cost figuring de novo, but accord EPA discretion to arrive at a cost figure within a broad zone of reasonable estimate. See Permian Basin Area Rate Cases, 390 U.S. 747, 88 S.Ct. 1344, 20 L.Ed.2d 312 (1968). Petitioners’ complaint revolves around the Agency’s refusal to include the capital cost of installing SSL recovery in some of its cost calculations. 58 As EPA notes, five out of six dissolving sulfite mills have already installed SSL recovery. Where, as here, most of a cost element has been incurred already, or will be incurred for a purpose other than meeting federal effluent limitations, EPA may exclude that element from the cost of BPCTCA. 59 Accordingly, we uphold EPA’s estimate of the cost of BPCTCA. b. “[N]on-water quality environmental impact^] (including energy requirements)” are among the “consideration factors” listed in section 304, and are the sole factors of that kind on which petitioners premise a challenge to the limitations. We have already seen that the Act does not specify a particular structure for EPA’s treatment of the consideration factors but instead leaves the Agency with discretion in deciding how they will be “taken into account.” In exercising that discretion, it is clear that EPA devoted considerable attention to assessing environmental impact and adequately set forth its conclusions with respect thereto. App. 2360-467 (Development Document). Most crucially, in view of the Act’s emphasis in listing “energy requirements” as part of environmental impacts, EPA developed estimates of the new energy demands for the industry as a whole — about 2.4% of the industry’s total energy use — and for each industry subcategory. For the sulfite subcategories, with their higher waste loads requiring greater waste treatment, the figure was an 18% increase in energy demand. 60 EPA also developed estimates of the sludge disposal problem, which is the reverse side of effluent reduction benefit, *1050 because the waste that is removed from effluent must be disposed of as sludge. 61 We are consequently convinced that EPA took adequately into account the environmental impacts of its regulations. 62 Petitioners assert, however, that we must impose on EPA a further and special requirement to engage in environmental balancing. They cite allegedly dramatic examples of negative environmental impacts from the air pollution and sludge disposal incident to waste treatment, and contend that EPA failed to give these enough “weight” in the balance. As we have discussed, we believe Congress entrusted the manner of deliberation about all of the “consideration factors” to EPA’s discretion, and we are prepared to uphold EPA on that basis alone. Nonetheless, the special policies in the Act with respect to environmental protection warrant some additional comments. Our consideration begins with the history of the Act, which shows that Congress developed a willingness to entrust EPA with more latitude than other agencies in carrying out programs that affect the environment. In the years immediately prior to the passage of the Act, the federal government had mounted a campaign to control water pollution by a permit program under the Rivers and Harbors Act of 1899, 33 U.S.C. § 407. During that period, Congress also had enacted the National Environmental Policy Act (NEPA), 42 U.S.C. § 4321 et seq., which was designed to force federal decisionmakers to consider the environmental effects of their decisions. Because the water pollution permit program under the Rivers and Harbors Act was aimed at enhancing the environment, some of NEPA’s legislative history suggested that the permit program officials might be exempt from the duty under NEPA to prepare environmental impact statements (EIS’s). Nevertheless, in December 1971, a court required EIS’s for water permits under the Rivers and Harbors Act, Kalur v. Resor, 335 F.Supp. 1 (D.D.C.1971), and in so doing severely impeded the permit program’s administrability. It is not surprising, therefore, that Congress became aware of this threat to the regulation of water pollution and that the legislators took it into account the following year in formulating the successor to that permit program — the 1972 Act. See, e. g., Legislative History, at 260 (remarks of Rep. Dingell). Accordingly, section 511(d) of the House version of the Act, H.R. 11896, 93d Cong., 1st Sess., excused EPA from complying with NEPA in issuance of water permits. Nonetheless, the House imposed environmental balancing duties on EPA in a different way. Section 101(g) of H.R. 11896 provided generally that “[i]n the implementation of this Act, agencies responsible therefor shall consider all potential impacts relating to the water, land, and- air to insure that other significant environmental degradations and damage to the health and welfare of man does not result.” 63 The House also included “environmental impact” in the particular section listing the many factors to be “tak[en] into account” in setting ef *1051 fluent limitations. Id. § 304. The Senate version of the Act had no corresponding provisions. After lengthy deliberations, 64 the Conference Committee knitted the two versions of the Act in a combination that minimized, without entirely omitting, EPA’s obligation thereunder explicitly to consider the non-water environment. Like the House bill, the Conference version excused EPA from preparing EIS’s. 65 Unlike the House bill, the Conference version completely deleted section 101(g), with its general duty to avoid all significantly adverse environmental impacts. Only section 304 was retained, in a modified and limited form, see note 68 infra, which simply made “non-water quality environmental impacts” one of the many listed factors for EPA to “take into account,” along with such other factors as engineering aspects, age, and process changes. Congress’ intent in passing this legislation was obviously not to minimize the importance of protecting the environment. The late 1960’s and early 1970’s saw the passage of a number of statutes aimed at dramatically increasing the protection given the environment, and the Act involved herein is among the most important. Rather, Congress was resolved to rely on EPA’s own internal structure and personnel attitudes to ensure that the net result of all of its programs would be a substantially enhanced natural environment. In essence, Congress was convinced that EPA’s internal dynamics and procedures were the “functional equivalent” of the NEPA duties imposed on other agencies. 66 As Senator Muskie declared, “[t]he whole concept of EPA is that environmental considerations *1052 are to be determined in one place by an agency whose sole mission is protection of the environment.” 67 Based on this conviction about EPA’s commitment to environmental goals, the legislators apparently felt that a judicially reviewable environmental balancing duty would serve no purpose except to establish a NEPA duty “as a tactic of litigation and delay” and thus make the EPA program unworkable. Portland Cement Ass’n v. Ruckelshaus, 158 U.S.App.D.C. 308, 317, 486 F.2d 375, 384 (1973), cert. denied, 417 U.S. 921, 94 S.Ct. 2628, 41 L.Ed.2d 226 (1974). That Section 304 requires EPA to “take into account” non-water quality environmental impacts, therefore, reflects several concerns apart from a fear that the Agency will have an inadequate commitment to protection of the air and land. Perhaps most important, if these factors were not listed, EPA arguably would have no authority to temper its effluent regulations when its own conclusion was that such tempering was needed to protect the land and air. 68 While committing the Agency to lead a comprehensive attack on water pollution, Congress did not intend that attack to prevent the Agency from realizing its other environmental goals. Furthermore, there are some environmental considerations, such as wildlife conservation, that are not *1053 within EPA’s expertise, but that Congress wanted considered in EPA’s deliberations on water pollution. See Legislative History, at 883-84 (letter from Reps. Reuss, Pelly, et a 1.). Thus, it included a reference to all non-water quality environmental impacts to assure the proper broadening of EPA’s environmental impact deliberations. Cf. Sun Enterprises, Ltd. v. Train, 532 F.2d 285, 289-90 (2d Cir. 1976) (EPA required under the Fish and Wildlife Coordination Act to consult with the Interior Department about relevant actions taken under the 1972 Act). Finally, the requirement of some advertence on the Agency’s part to non-water environmental impacts provided a mechanism for assuring that EPA’s internal structure and procedures will evolve as expected. The Act’s listing of environmental impacts as a factor encourages the Agency, if more incentive is necessary, to seek information from relevant sources outside the Agency and from personnel in sections of the Agency devoteá to non-water matters. Once that communication is assured, the likelihood that the expected inter- and intraagency sensitivity to environmental benefits and impacts will not occur is slight indeed. 69 Thus, since Congress intended EPA’s internal structure to protect the non-water environment, the judicial function is completed when we have assured ourselves that EPA expressly considered the probable environmental impacts of its regulations. As we have noted, EPA fully investigated the environmental impacts, and thereby fulfilled this aspect of its statutorily mandated duty. VI. THE AGENCY’S EXERCISE OF DISCRETION Petitioners’ remaining challenges to the effluent limitations involve attacks solely on the rationality of the Agency’s discretionary decisionmaking in various areas. Although we have examined all of petitioners’ attacks on the Agency’s exercise of discretion and found them wanting, several of them deserve further discussion. Those challenges fall into two major categories: whether EPA has responded adequately to the diversity of the industry in its subcategorization and limit-setting decisions, and whether it has identified “practicable” technology capable of achieving those limitations. A. In setting effluent limitations, EPA had to take into account the great diversity among the almost 300 mills subject to Phase II regulation. It did so by two methods: subdivision and averaging. EPA began by subcategorizing the industry on the basis of the major differences in industrial processes and other factors. Further distinctions were made within many subcategories. Industry commented on the earlier subdivision proposals, and EPA accommodated many of industry’s concerns, eventually identifying 16 subcategories, divided into 66 subdivisions. On average, that is, EPA tailored one set of limitations to every five mills. This extensive subdivision safeguarded against overzealous standards, increased the confidence that can be placed in the practicability of the regulations, and diminished the need to handle variation through the variance process. Of course, differences still exist among mills within the same subcategories and subdivisions. EPA took these differences into account by basing its limitations on BPCTCA, i. e., the pollution control achieved at the “average of the best mills.” For its averages, EPA drew on the best *1054 available data base. 70 That the limitations still impose greater burdens on some mills than others, i. e., on those with below average treatment facilities, is an unavoidable consequence of Congress’ goal of uniformity in effluent limitations rather than in treatment expenditures. Petitioners’ specific challenges to EPA’s treatment of variability focus on four sources of variability: cold climate, 71 the *1055 profitability of SSL recovery, “upset” conditions, and hydraulic flow. At the outset, we note that petitioners have á heavy burden in challenging EPA’s treatment of variability. That treatment involves a host of administrative, technical, and statistical considerations as to which we lack the mandate to second-guess the Agency. For the reasons discussed below, we uphold EPA’s limitations against all four challenges. 1. Petitioners argue that EPA inadequately accounted for temperature differences in setting effluent limitations. As all parties agree, cold climates affect biological waste treatment. Bacteria break waste down more slowly at low temperatures, and different populations of bacteria predominate at different temperatures, 72 so that mills in cold regions have greater difficulty in using biological treatment. See Tanners’ Council v. Train, 540 F.2d 1188, 1194 (4th Cir. 1976); American Meat Inst., supra, 526 F.2d at 455. EPA considered the climate problem and concluded that northern paper mills could practicably use activated sludge systems as their method of biological treatment. Under this method, waste is treated before it can cool significantly even in cold climates. 73 Although petitioners counter that activated sludge has many drawbacks, including a susceptibility to “shocks” and higher cost, 74 EPA took those drawbacks into account in its deliberations. Thus, it included equalization basins in activated sludge BPCTCA to insulate against “shocks” and, more importantly, it determined that the effluent benefits were worth the extra cost. In light of this latter determination, subcategorization on the basis of climate would be a fruitless act, since the subcategory incurring the higher cost (mills in colder areas) could justifiably be held to the same effluent limitations as the one incurring the lower cost (those in warmer climes). Nor is disapproval of the regulations mandated by prior cases that have discussed temperature subcategorization. The critical issue in those cases has been whether EPA compiled a full record on which it based its decision, and here we find adequate record support for EPA’s determination. 75 *1056 2. Petitioners next contend that EPA acted arbitrarily in refusing to create a separate subcategory for sulfite mills lacking a profitable process for “SSL recovery.” As previously noted, the “cooking” of wood with chemicals in the sulfite process leaves behind a waste-laden solution called spent sulfite liquor (SSL). See pp.---of 191 U.S.App.D.C., pp. 1022-1024 of 590 F.2d supra. SSL can be recovered by collecting, evaporating, and/or burning it. When sodium and magnesium are used in the cooking solution, SSL recovery is profitable because those expensive chemicals may be economically recovered for reuse rather than discharged. When two less expensive chemicals, ammonia and calcium, are used, SSL recovery is less profitable or unprofitable, and SSL recovery is simply useful as a waste treatment method for disposing of SSL without water pollution. See note 10 supra. Petitioners assert that the difference between mills at which SSL is profitable, and those at which it is not, is a more basic process difference than some other such differences that were used as the basis for EPA subcategorization. As such, they assert, the Agency acted arbitrarily in refusing to establish separate subcategories in this instance. We believe that EPA acted within its broad discretion in regulating all sulfite mills without regard to the profitability of SSL recovery. Process is only one of several factors relevant to subcategorization. Other factors may be influential or decisive. Here, subcategorization would have had a negative effect on the cost and effluent benefit balance, because, then, SSL recovery — a highly cost-efficient pollution-control technique — could have been foregone. 76 By contrast, the other process differences pointed to by petitioners as less drastic but nonetheless the basis for EPA subcategorization do not require that techniques as cost-efficient as SSL recovery be abandoned as BPCTCA. 77 3. Waste treatment facilities occasionally release excess pollutants due to such unusual events as plant start-up and shutdown, equipment failures, human mistakes, and natural disasters.' EPA accounted for this type of performance variability by using a careful statistical approach. In setting daily and monthly effluent limitations based on performance data from many plants, it made allowance for more than 99% of all the variability in performance. Petitioners contend that this degree of accuracy is insufficient in that the statute’s civil and criminal penalties are based on an absolute liability standard. They note that the 99% figure shows that, in the past, fully adequate mill treatment systems have occasionally exceeded the proposed effluent lim *1057 itations, and they ask the Agency to promulgate an “excursion” or “upset” regulation specifying circumstances in which exceeding the limitations will be excused. The excursion issue has troubled EPA and the courts. 78 On the one hand, denying such provisions seemingly asks plants to do the impossible — to handle the most unusual upsets in plant conditions. On the other hand, there are strong policy reasons for denying such provisions and leaving the handling of upsets to EPA’s enforcement discretion. Last year, EPA reviewed and elaborated its position on the matter. NPDES Decision of the General Counsel No. 57 (March 16, 1977). Henceforth, EPA will identify “upset conditions” and allow excursion provisions in some industries. In other industries, it will not promulgate such provisions. Instead, it will take variability into account by setting generous daily and monthly effluent limitations and by exercising its discretion not to enforce when appropriate. The proceeding involved in the present case was one of the latter kind. EPA refused to promulgate an excursion provision, stating that it “expects that the performance of the worst cases [i. e., plants that exceed limitations] will be improved by proper controls and that the effluent limitations can be achieved by properly operated and maintained plants.” 42 Fed.Reg. 1420 (1977) (final regulations). We believe that EPA’s position in Decision No. 57 is sound, and we uphold its denial of excursion provisions in this case. A contrary decision might hamper the Agency’s ability to “force technology” and hence could impede enforcement. Congress intended effluent limitations to compel plants to improve their performance, even when innovation was required. See pp. ---of 191 U.S.App.D.C., pp. 1061-1062 of 590 F.2d infra. In this “technology forcing” context, denial of an excursion provision justifiably compels plants to develop monitoring, repair, and back-up capabilities to avoid excessive discharges. 79 Moreover, depriving EPA of discretion to refuse excursions would interfere with the congressional goal of “swift and direct” enforcement. 80 Excursion provisions cannot be framed in simple numerical terms, as, say, an allowance of four excessive discharges per year, without giving mills a license to dump wastes at will on several occasions annually. American Petroleum Inst. v. EPA, 540 F.2d 1023, 1036 (10th Cir. 1976) (denying excursions). Thus, such provisions must be stated in complex terms, e. g., by specifying plant incidents justifying excessive discharges. 81 As the Supreme Court has recognized recently, however, there is a major difference in pollution regulation enforcement between simple numerical standards and *1058 complex requirements. 82 Once excursion provisions are promulgated, an enforcement case no longer turns on the sharply defined issue of whether the plant discharged more pollutant than it was allowed to, but instead depends on murky determinations concerning the sequence of events in the plant, whether those events would have been avoidable if other equipment had been installed, and whether the discharge was within the intent of the excursion provision. Consequently, what Congress planned as a simple proceeding suitable for summary judgments would become a form of inquest into the nature of system malfunction. We consequently reject the argument that EPA must promulgate excursion provisions so that the effluent limitations will reflect BPCTCA technology that is effective 100 percent of the time. But see Marathon Oil Co., supra, 564 F.2d at 1273-74. In the nature of things, no general limit, individual permit, or even any upset provision can anticipate all upset situations. After a certain point, the transgression of regulatory limits caused by “uncontrollable acts of third parties,” such as strikes, sabotage, operator intoxication or insanity, and a variety of other eventualities, must be a matter for the administrative exercise of case-by-case enforcement discretion, not for specification in advance by regulation. See CPC Int’l, Inc., supra, 540 F.2d at 1338 (denying excursions). A line must be drawn as to what is treated in the general rule and what is handled case-by-case, and we believe it appropriate to defer when EPA, in the interest of preserving an enforcement system based on straightforward numbers, establishes a general rule with better than 99 percent accuracy and leaves the rest to prosecutorial discretion. We are likewise unpersuaded by an analogy to excursions under the Clean Air Act, 83 and by the threat of citizen suits, 84 and/or of a lack of good faith on the part of government prosecutors. Accordingly, we uphold EPA’s decision to deny excursion provisions. 85 *1059 4. The Agency typically expresses its effluent limitations in terms of pounds of pollutants per ton of goods produced. That figure, in turn, is arrived at by multiplying water flow times pollution concentration— that is to say, by deriving the product of (1) the gallons of water used per ton of goods and (2) the pounds of pollutant per gallon of water: Pollution ratio * = Flow ** X Concentration *** This approach is designed with the commendable purpose of avoiding limitations expressed in terms, such as pounds of pollution per gallon of water, that would allow mills to avoid the regulatory impact by diluting their effluent. See Marathon Oil Co., supra, 564 F.2d at 1269. Nonetheless, it places the Agency partially in the position of regulating water flow, which tends to vary from mill to mill. That is, in promulgating its limitations, the Agency first reached a conclusion about the normal flow in given segments of the industry and then directed its technology search at finding means to extract the likely pollution from that amount of water. It is true that if a mill has a higher than average flow, it will probably have a lower than average concentration of pollutants. Nonetheless, to treat that higher amount of water, despite its lower concentration of pollutants, might take more facilities or more time, and hence might cost more. As such, those mills that utilize higher levels of flows than those deemed normal by EPA may be at some disadvantage in achieving the effluent limitations. It is important to note, however, that EPA does not require operators to achieve any specific level of flow, but only that the pollution content in that flow as a whole be below a constant level set relative to tons of product. Hence, EPA does leave industry the option of utilizing a somewhat higher flow but building pollution control technology capable of accommodating it. In light of the nonmandatory nature of the flow determinations made by EPA, we may easily dispose of one of petitioners’ contentions. The industry appears to argue that by setting limitations and identifying practicable technology with a given flow in mind, EPA has failed adequately to identify the cost of the required pollution control (or, stated differently, to identify practicable technology) for those plants with higher than average flows. 86 The simple answer to this argument is that EPA studied mills with varying levels of flow and set limitations that it deemed within reach of all of those mills, so long as they installed “practicable” (i. e., in part, not overly costly) technology. Having found none of the limitations themselves — or the practicable technology determinations — unreasonable, therefore, the fact that some mills use flows different from those considered normal for the purposes of some of EPA’s calculations seems irrelevant. Petitioners further claim, however, that even if reasonable under the normal analysis, the limitations are unsatisfactory because they force any plant with higher than average flows to add internal controls to reduce water usage. In support of their position, petitioners point to the legislative history suggesting that Congress expected most pollution controls to require end-of-pipe, or external, technology, and they repeat the well established rule derived therefrom that internal controls that are not in common use in the regulated industry may not be required by EPA in the 1977 standards. See American Paper Inst., supra, 177 U.S.App.D.C. at 194, 543 F.2d at 341 (citing cases); American Iron & Steel Inst., supra, 526 F.2d at 1060-61. Although this rule requires the Agency to make a “common use,” rather than “practicability” finding whenever the control technology it identifies is internal, we review its decision *1060 for abuse of discretion exactly as we review its other findings. In this instance, as in most others, EPA has fully explained its methodology so that the bases for its decision on average flow for each subcategory are visible. App. 2073-177, 2478-81 (Development Document). The Agency achieved this series of figures by a careful process. Initially many of its subcategorization divisions were based on characteristic flows within segments of the industry, so that the mills to be compared with each other were fairly uniform to begin with. Next, the Agency measured the average flow within each subcategory — discarding unusually high and, more often unusually low ones. Notably, the Agency based its decision on the average of all mills rather than on the “average of the best,” which is its usual — and unchallenged — method of determining practicability. Hence, EPA made a conscious effort to distinguish internal flow controls, which require a “common usage” finding, from other technological requirements, which may be less industrially prevalent so long as they are “practicable.” Moreover, beyond these measurements, the Agency also undertook a detailed study of flow control methods and the cost thereof for each subcategory and found them available and used to one degree or another in all subcategories. App. 2195-242, 2364-97 (Development Document). In fact, it showed, as in American Paper Inst., supra, 177 U.S.App.D.C. 194, 543 F.2d at 341, that some of all internal pollution control measures identified as useful by the Agency are used by no less than 75% of the mills studied within any given subcategory. App. 2214 (Development Document). And, although not all of these controls are aimed at reducing flow, the Agency determined that, overall, more than 50% of all of the mills studied achieved the subcategory flow rate utilized by EPA in setting the limits, which we find sufficiently indicative of the normality of internal flow controls. 87 Accordingly, the Agency’s determination of common flow is within the authorized bounds of discretion. B. 1. Some of the petitioners have strongly challenged EPA’s choice of BPCTCA for the dissolving sulfite subcategory. As discussed above, the intense “cooking” process used in the dissolved sulfite process dissolves most of the wood’s cellulose in waste solutions. See pp.-- - of 191 U.S.App.D.C., pp. 1022-1024 of 590 F.2d supra. During “biological” treatment this dissolved matter is taken up by bacteria. Disposing of the watery mass of these bacteria, i. e., sludge, is a serious problem, for their tough cell walls make it difficult to drain the water from, i. e., “dewater,” them. Moreover, without draining, sludge is difficult to incinerate or use as landfill. 88 *1061 It has long been recognized that the dewatering problem is difficult and that all the available approaches have certain drawbacks. See App. 1515-30 (1969 study). EPA began its consideration of the problem in 1973 with a “state of the art” review of treatment technology, which devoted much attention to alternative dewatering approaches. App. 3693-711. It has since amassed considerable other material on the problem, including the industry’s own data and criticism of EPA’s dewatering proposals, culminating in discussions in its final Development Document of dewatering technology and the estimated cost and environmental impact thereof. App. 2262-64, 2346-59, 2382-83, 2406-08. Recognizing the difficulty of the problem, EPA has not promoted a single dewatering technology, but instead has offered- diverse approaches — some sequential, some alternative — indicating that each particular dissolving sulfite mill might choose a different approach. In its model, the bacterial solution is chemically prethickened and then mixed with nonbacterial waste, which is easier to dewater and further thickens the mass. The resulting mixture is chemically conditioned and subjected to vacuum filtration, a process in which rotating filters suck a soggy “cake” out of the prethickened waste. This cake, which may still be 20% or less solid, can be used for landfill, either at this point or later after pressing to remove even more water. Alternatively, the sludge can be dried using special boilers, and then burned in those same boilers, with the heat of incineration used to dry more sludge. Industry pointed out problems and drawbacks with all these proposed steps and alternatives. 89 The Agency responded by pointing to its record support, including the fact that one out of the six dissolving sulfite mills covered by the regulations has already implemented dewatering technology. 90 Although perhaps a closer question than the others reviewed herein, we do not find EPA’s choice of BPCTCA in this instance to be an abuse of its broad discretion. EPA has clearly devoted much attention to the problem and has explained its conclusions. In major part it seems to have been motivated by a congressionally mandated policy concern: the determination to push pollution control technology rapidly and decisively forward to the limits of practicability after 1977 and to the limits of achievability after 1983. See Legislative History, at 1460 (Senate Report) (Administrator given 1977 “mandate to press technology and economics to achieve those levels of effluent reduction which he believes to be practicable . . ..”). Indeed, even in the earlier set of regulations, Congress expected EPA to press sometimes beyond the most advanced technology currently being used: “In those industrial categories where present practices are uniformly inadequate,” Congress has admonished the Administrator to interpret “ ‘best practicable’ to require higher levels of control than any currently in place if he determines that the technology to achieve those higher levels can be practically applied.” Id. at 169-70 (statement of Sen. Muskie). To force industry implementation of pollution control technology, EPA has used a number of approaches upheld by the courts. EPA has based its 1977 standards on exemplary facilities, including foreign ones, American Frozen Food Inst., supra, 176 *1062 U.S.App.D.C. at 138, 539 F.2d at 140, and has utilized “transfer technology.” See note 70 supra. In this case, EPA has used yet another approach — specification of a diverse technological array as BPCTCA 91 —which we uphold in light of the lack of effluent treatment that has traditionally characterized sulfite mills. Under this diversified approach, the Agency has identified a number of alternative dewatering techniques with the expectation that, if several techniques are available, one or another will work at all of the regulated mills. Even if each technique has some drawbacks, together all of them constitute a “practicable technology” suitable to BPCTCA. In sum, because we find that the dewatering methods identified by EPA have had some success in the past and because the judgment as to whether an array of diverse approaches is sufficient in this instance “is best left in the hands of the Agency,” Tanners' Council, supra, 540 F.2d at 1192, we accept the Agency’s conclusions.- 2. Petitioners also make a broad challenge to all the TSS limitations for the industry. They contend that there are certain “nonsettleable solids” that will not be removed by BPCTCA, and the EPA failed to rely on actual treatment performance data in setting TSS limitations that require removing those solids. We considered and rejected a similar challenge in American Paper Inst., supra, 177 U.S.App.D.C. at 198, 543 F.2d at 345, and we are inclined to do likewise here. Petitioners admit that in the particular subcategory it chose as an example, the bleached kraft subcategory, 10 out of the 32 mills already meet all of EPA’s effluent limitations, and even more meet just the TSS limitations. Nonetheless, they complain that these conforming mills, as well as others relied on by EPA, utilize advanced internal and external control technologies that were installed solely to meet stringent state and local water quality standards for particular receiving waters. App. 764. This observation is largely correct but it does not present a cognizable grievance. The mills on which EPA has modeled BPCTCA are without question capable of meeting EPA’s effluent limitations, and the Agency accordingly is justified in insisting that all other mills come up to this undoubtedly “practicable” standard. For all of the foregoing reasons, the 1977 effluent limitations for the bleached segment of the American paper industry are upheld, and the petitions denied, except that the BOD limitation for acetate grade dissolving sulfite mills is remanded to the Agency for further proceedings consistent herewith. It is so ordered. Footnotes [1] In relevant part, i. e., with reference to the 1977-83 set of limitations for private industry, § 301(b) provides: In order to carry out the objective of this Act, there shall be achieved— (1)(A) not later than July 1, 1977, effluent limitations for point sources, other than publicly owned treatment works, (i) which shall require the application of the best practicable control technology currently available as defined by the Administrator pursuant to section 304(b) of this Act . Section 42 of the Clean Water Act of 1977, 33 U.S.C.A. § 1311(b) (1978), replaced the 1972 Act’s framework for post-1977 standards (1983 standards) with a new one. The 1972 Act’s framework for post-1977 standards is well understood, and so we have used it to provide perspective. To avoid discussing issues not presented in this case, we have deliberately not attempted to describe the new Act’s post-1977 framework except where necessary. [2] Again as relevant here, see note 1 supra, § 304(b) provides: For the purpose of adopting or revising effluent limitations under this chapter the Administrator shall, after consultation with appropriate Federal and State agencies and other interested persons, publish within one year of October 18, 1972, regulations, providing guidelines for effluent limitations and, at least annually thereafter, revise, if appropriate, such regulations. Such regulations shall— (1)(A) identify, in terms of amounts of constituents and chemical, physical, and biological characteristics of pollutants, the degree of effluent reduction attainable through the application of the best practicable control technology currently available for classes and categories of point sources (other than publicly owned treatment works); and (B) specify factors to be taken into account in determining the control measures and practices to be applicable to point sources (other than publicly owned treatment works) within such categories or classes. Factors relating to the assessment of best practicable control technology currently available to *1020 comply with subsection (b)(1) of section 301 of this Act shall include consideration of 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 (including energy requirements), and such other factors as the Administrator deems appropriate. Although, as written, § 304(b) required EPA to act with respect to all industries within a year of the statute’s passage and, in particular, § (1)(A) thereunder required the agency within that period to have taken many of the steps preliminarily necessary to the promulgation of the limitations under § 301(b), see note 1 supra, the Agency found its task too overwhelming to complete in so short a time. See E. I. duPont de Nemours & Co. v. Train, 430 U.S. 112, 131-32 & n.22, 97 S.Ct. 965, 51 L.Ed.2d 204 (1977). Consequently, EPA has generally chosen to telescope into one proceeding per industry the identification of the attainable effluent reductions and the factors relevant thereto — under § 304(b) — and the actual establishment of the various industry-wide limitations — under § 301(b). It has accomplished this goal by using traditional, if usually more exhaustive, notice-and-comment rulemaking proceedings, some of which are still in progress. See duPont, supra, 430 U.S. at 131-32 n.22, 97 S.Ct. 965. The courts have validated this exercise of EPA discretion in carrying out the Act upon a showing that the unified procedure has effectively achieved the ends of the statute’s bifurcated and apparently impractical plans. E. g., American Frozen Food Inst. v. Train, 176 U.S.App.D.C. 105, 128-29, 539 F.2d 107, 130-31 (1976). [3] Although unexplained in the record, we may ' surmise that the delay stemmed largely from the greater diversity of the Phase II segment of the industry, compare 39 Fed.Reg. 18742 (1974) (Phase I regulations, identifying only five discrete subcategories of the affected industry) with 42 Fed.Reg. 1398 (1977) (Phase II regulations, identifying 16 such subcategories), the prudent decision of the EPA to solicit comments at several stages of an already complicated promulgation process, see Hunciker & Pagano, Federal Environmental Litigation in 1976; The Federal Water Pollution Control Act, 1 Harv.E.L.Rev. 59, 87 (1977) (usual “standard-setting procedure requir[es] twenty-seven distinct steps”), and, more generally, from the overwhelming technical and litigative burden shouldered by the Agency under the Act, see Note, Effective National Regulation of Point Sources Under the 1972 Federal Water Pollution Control Act, 10 Ga.L.Rev. 983, 1024 n.216 (1976) (as of 1975, the Agency was defending 250 court challenges to effluent regulations). [4] Rather than using the Federal Register to solicit comment at this preliminary stage, EPA circulated the draft to 450 government, industrial, and private citizen organizations. See generally 38 Fed.Reg. 21202-06 (1973). [5] Zinc is also a pollution parameter measure for industry subcategories that use it in their processes. Zinc effluent limitations have not been challenged. The term “parameter” has been used to describe BOD, TSS, pH, and similar measures because of their function. In mathematics, a parameter is defined as an “arbitrary constant” —a variable that keeps a constant role in a formula as it takes on different (arbitrary) numerical values. For example, in the formula for a parabola (y = ax 2 + b), the horizontal or “x” variable is a parameter because it keeps a constant role in the formula while different numerical values are plugged in to calculate different vertical or “y” levels. For all EPA regulations, BOD and the like keep a constant role — what they are, and how they are measured, stay the same — even as they take on different numerical values for the acceptable level in each industrial category depending on the available technology for that category. [6] As the Second Circuit recently noted, “Biochemical Oxygen Demand is not strictly speaking a pollutant at all.” C & H Sugar Co. v. EPA, 553 F.2d 280, 282 n. 7 (2d Cir. 1977). BOD is a measure of how much oxygen is used by organisms in water that break down biodegradable pollutants. As such, it indicates how much dissolved oxygen in the water will be used in fermenting the wastes. It is important because that fermentation process is capable of *1023 depleting so much of the water’s oxygen supply that fish and other life in the water asphyxiate. The BOD level is also important because it correlates with the level of harmful organic chemicals. Accordingly, effluent with high BOD can cause damage even when dumped into waters that have more than ample amounts of dissolved oxygen. The same is true of pH: effluent high in toxic acids may cause damage even when dumped into sea water, whose salts buffer its pH level. It is well recognized that EPA can use pollution parameters that are not harmful in themselves, but act as indicators of harm. See American Paper Inst. v. Train, 177 U.S.App.D.C. 181, 202, 543 F.2d 328, 349, cert. dismissed, 429 U.S. 967, 97 S.Ct. 398, 50 L.Ed.2d 335 (1976) (color can be used as a pollution parameter). [7] Separate challenges were brought by mills using nonsulfite processes, but they raised issues common to the whole industry. Of almost 300 mills affected by these regulations, about 30 are sulfite mills. [8] Although the sulfite subcategories must invest heavily, EPA has not given them a disproportionate duty; due to the large volume of waste to be treated, investment in sulfite waste control is highly cost-efficient. Under the 1977 regulations, sulfite mills will pay $.10 to $.19 for each pound of BOD they remove from effluent. Of the 13 other industry subcategories, only one will pay less than $.10 per pound of BOD removed. Of the rest, five will have comparable, and seven will have higher, costs. App. 2479. [9] Council on Economic Priorities, Paper Profits: Pollution in the Pulp and Paper Industry 16-17 (1971). This carefully documented study analyzes air and water pollution generated by 24 paper companies at 131 locations. It discusses the paper industry’s unhappy record of liquid waste dumping and failure to install even minimal antipollution measures before the 1972. Act. The study determined that “in 1969, it is estimated that the industry discharged 15% of the total industrial effluent of the United States. ... In 1967, the paper and allied products industry used nearly 2.1 trillion gallons and discharged 1.9 trillion gallons. Only 34% of this discharge water received any treatment to alleviate its chemical and physical adulteration.” Id. at 3, 7. [10] When the wood is “cooked,” the solution uses a mixture of sulphurous acid and one of four chemicals: sodium, magnesium, ammonia, or calcium. The economics are such that it is profitable to perform SSL recovery to recover sodium and magnesium for reuse and not profitable to recover calcium. Evidence concerning the profitability of recovering ammonia is conflicting. Out of 23 sulfite mills, the six that use sodium and magnesium all perform SSL recovery, as do 14 out of the 17 that use ammonia or calcium. The latter have done so in part under the compulsion of past water pollution regulations. [11] We might add parenthetically that the degree of “concise[ness]” called for by the APA with respect to the Agency’s explanation of its actions — and, indeed, the degree of conciseness tolerated by the courts with respect both to that explanation as well as to the opportunity for public participation — may vary according the nature of the regulations. In the case of highly technical regulations, and especially regulations calling for refined scientific judgments, the reviewing court should be careful to insure that the procedures used as a vehicle for promulgation are ample enough to support their substantive cargo. E. g., Kennecott Copper Corp. v. EPA, 149 U.S.App.D.C. 231, 462 F.2d 846 (1972); Environmental Defense Fund v. Hardin, 138 U.S.App.D.C. 391, 398, 428 F.2d 1093, 1100 (1970). Cf. Vermont Yankee Nuclear Power Corp. v. National Resources Defense Council, Inc., 98 S.Ct. 1197, 1212-13 (1978). Moreover, the use in § 553 of the dyad “basis *1025 and purpose” suggests that the explanatory statement should discuss both the factual premises, if any, and the policy considerations underlying the administrative action. This requirement not only aids the reviewing court in assuring adherence to considerations appropriate under the statute, but also permits the court to decide if the agency’s conclusion is “sustainable on the administrative record made.” Id. at 1214, quoting Camp v. Pitts, 411 U.S. 138, 143, 93 S.Ct. 1241, 36 L.Ed.2d 106 (1973). [12] In this regard, we note that EPA has often been compelled to devote as much as half of its rulemaking staff under the Act to in-court defense of its already-issued regulations. See Hunciker & Pagano, supra note 3, at 61. [13] One of the petitioners in this case unwittingly illustrated this point. Its brief characterized one of the many claims made by it as involving “so basic and glaring [an error] that it is not necessary to have scientific and technical training in order to appreciate [it].” Supplemental Brief of Petitioners in No. 76-1675, at 42. Presumably our “scientific and technical training” enables us to cope with the others. [14] We use the term “exponent” in both of its nonmathematical senses. That is, the courts have not only traditionally served as “expounder[s], explainer[s and] interpreters” of due process, but also, through their own procedures as developed over the centuries, as “one[s] who . . . exemplify] or represent[ ].” Webster’s Third Dictionary of the English Language 802 (3d Ed. 1963). [15] It is important, however, for potential challengers of regulations to realize that the success of open and participatory procedures in achieving substantively sound administrative law depends as much upon them as upon the Agency. It is our responsibility, therefore, in attempting to enforce procedural fairness, not only to hold the Agency to adequate procedures but also to insist that petitioners have made substantial and good faith use thereof. To the extent that technical, factual, or policy ' concerns, that could have been raised earlier, are in fact raised by petitioners for the first time on appeal, the notice-comment-and-re sponse procedures will have been deprived of much of their vitality, and the party responsible therefor will accordingly be given less latitude in complaining about the results. Cf. ASARCO, Inc. v. EPA, 188 U.S.App.D.C. at 320-79 n.1, 578 F.2d at 320-21, n.1 (1978) (sufficient that issue have been raised before Agency by someone, even if not by petitioner; greater latitude to raise legal issues for the first time on appeal of regulations). [16] In general, the next three parts of the opinion will review the paper industry effluent regulations first for their procedural fairness, then for their adherence to the statute, and finally for any substantive abuse of discretion. Nonetheless, we will deal with any abuse-of-discretion challenges related to a discrete aspect of the regulations at the point where we discuss the procedural or statutory challenges thereto. [17] For example, in light of industry comments received on the Interim Final Limitations, the total suspended solids (TSS) daily limit for the dissolving sulfite subcategory was increased by 38 percent. Compare 41 Fed.Reg. 7677 (1976) with 42 Fed.Reg. 1406 (1977). [18] “Raw waste load” is the total amount of waste generated by the mill before treatment. After this waste goes through primary treatment (j. e., after being pumped into a sedimentation tank where some solid waste settles out), the remaining pollution, called “secondary waste load,” is subjected to secondary treatment (breakdown by bacteria) in order to cut down BOD. Accordingly, the amount of secondary waste load determines the amount and cost of secondary treatment sufficient to meet the BOD limitation. [19] The Agency, based the figure of 487 lbs/ton on data from one mill, Mill 403. [20] App. 2124. The Agency based the figure of 404 lbs/ton on new data for Mill 403, see note 19 supra, and on data from another mill, Mill 402. [21] On the one hand, Mill 402, see note 20 supra, utilized one type of waste treatment that mills using BPCTCA do not use, SSL neutralization, so for that reason its secondary water load would likely fall below a representative level. On the other hand, Mills 402 and 403, see notes 19 & 20 supra, did not use another type of waste treatment, SSL recovery, as fully as mills with BPCTCA, so for that reason they could be expected to have higher secondary waste loads than at perfectly representative mills. EPA counsel inform us that during rule-making the Agency staff estimated that these two factors would essentially cancel out, and decided not to reflect them in the record. It is now argued by the Agency that, if the two factors are fully taken into account, there should be a net adjustment downward of the data received from the two plants. [22] App. 1795-96. EPA assumed that Mills 402 and 403, see notes 19 & 20 supra, lacked primary treatment. If so, their secondary waste loads would be higher than at a mill using BPCTCA, which has primary treatment, and their secondary waste load would be above a representative figure. Accordingly, in making its cost calculations EPA adjusted their figures downwards by 22% to arrive at a representative figure. In fact, the mills use primary treatment, and the adjustment should not have been made. [23] Petitioners argued first that the 404 lbs/ton figure derived from the new data undervalued secondary waste load in mills using BPCTCA because the new data was facially incomplete and unrepresentative of mills using BPCTCA. See generally note 70 infra. Further, they pointed out the erroneous assumption behind the 22% downward adjustment to the 404 lbs/ton figure that EPA made in computing the cost of achieving the established effluent limitations. See note 22 supra. [24] Petitioners’ new data show that Mill 402’s SSL recovery is even less full than EPA assumed. See note 21 supra. This new data indicated that Mill 402’s secondary waste load figure was even further above a representative one than EPA assumed during rulemaking, and that a larger adjustment downward to obtain a representative figure would be appropriate. [25] EPA’s revised analysis may be summarized as follows: If EPA had taken into account that the mills actually did have primary treatment, it would not have made the downward adjustment discussed in note 22 supra. And if it had adjusted the figures because of Mill 402’s atypical use of SSL neutralization, it would have raised the secondary waste load figure obtained in order to make it more representative. See note 21 supra. On the other hand, had EPA taken into account that both mills lacked full SSL recovery, and had it used the new data on how far below capacity the SSL recovery was operating, it would have adjusted their secondary waste load figure significantly downwards. See notes 21 & 24 supra. EPA counsel contended that when the adjustments upwards and downwards were aggregated, the net secondary waste load figure would fall to 292 lbs/ton, i. e., slightly below the figure of 314 lbs/ton used in EPA’s calculation of the cost of BPCTCA. [26] These first two aspects of the EPA’s decisionmaking process that were not subject to prepromulgation public scrutiny are factual matters. As such, they are especially subject to verification through the notice and comment *1031 process and less acceptably removed therefrom. Moreover, although both parties herein accept the validity of the data referred to in the second factor mentioned, the fact that it nowhere appears in the record so divorces it from the notice and comment procedure as to make it suspect. [27] It might be argued that whether the data and recomputations argued for by petitioners or by EPA are used, the resulting secondary waste load figure, and thus the cost of treating the same to achieve the specified BOD limitation, would not be so high as to invalidate the limitation itself as an abuse of discretion. Hence, the argument might continue, the alleged underestimation of waste load is harmless, because only that limitation is subject to review. Yet this argument misconceives our review function. Even though we might find such a limitation within the realm of discretion accorded the Agency by Congress, it is our place only to review actions actually and properly taken by the Agency. If that action is improper, and if we cannot be sure that under the correct procedures the Agency would have reached the same conclusion, we cannot characterize the defect as harmless. [28] We interpret the Fourth Circuit’s and Supreme Court’s actions in duPont has having proceeded under the normal assumption in cases where no other evidence is available that the Agency will apply substantive parts of its regulations in a meaningful and lawful manner. The passage quoted in text from the Fourth Circuit’s opinion indicates, however, that as soon as a court is faced with a case in which the variance’s application in fact is no longer “a matter of speculation,” the court should abandon that assumption and proceed with its review as the circumstances of the case dictate. [29] We note that EPA recently rejected a variance authorization that a California agency had included in a permit for a mill operated by one of the petitioners in this case, Crown Simpson Pulp Co. See In re Louisiana-Pacific Corp., 10 E.R.C. 1841 (1977) (Decision of the Administrator). That decision is currently under review in the Ninth Circuit, and our decision here implies no opinion on the merits of that particular petition in light of the facts relevant solely thereto. Moreover, the government conceded in oral argument that the Ninth Circuit has jurisdiction not only to decide the merits of that particular variance request under the Agency’s applicable variance regulation, but also to decide the general validity of the variance provision itself under the criteria set out in the Act as interpreted in duPont. Cf. § 5.09(b)(1)(E) of the Act, 33 U.S.C. § 1369(b)(1)(E) (review petition concerning “any effluent limitation or other limitation under section 301 of the Act” must be brought within 90 days of promulgation of that limitation). Since our review seeks only to decide whether, in light of the evidence we now have concerning the Agency’s interpretation of its variance provision, it is capable of having the degree of flexibility required by duPont, our discussion does not foreclose any future explorations of whether, as interpreted hereafter and, particularly, as applied in a particular case, the variance provision has the requisite flexibility. [30] This conclusion does not require our full acceptance of the analysis in Appalachian Power Co. v. Train, 545 F.2d 1351, 1359-61, modified, 545 F.2d 1380 (4th Cir. 1976). There, the Fourth Circuit postulated that because the 1977 limitations were to be less stringent than those in 1983, the variance allowance for the former should be at least as flexible as the one for the latter, which is found in § 301(c). Our holding is somewhat different. Noting that the Act reveals quite clearly how the factors relevant to . granting a 1983 variance relate to those relevant to drawing up the general 1983 limitations, we insist only that the EPA’s application of the 1977 variance allowance bear a similar relationship to the factors the statute deems crucial to the development of the general 1977 limitations. [31] This provision, in broad outline, coincides with the requirements typical to variance clauses in most local zoning ordinances governing real property. Those clauses generally require “a showing . . . that (1) enforcement of the statute as written would cause to the applicant substantial hardship typical to his land; and (2) the proposed use will not disrupt the general zoning scheme.” 6 R. Powell, Real Property fl 870 (1977). [32] Section 301(b), 33 U.S.C. § 1311(b) states that “there shall be achieved— (2)(A) [not later than July 1, 1983,] effluent limitations for categories and classes of point sources, other than publicly owned treatment works, which (i) shall require application of the best available technology, economically achievable for such category or class, which will result in reasonable further progress toward the national goal of eliminating the discharge of all pollutants, as determined in accordance with regulations issued by the Administrator pursuant to section [301 of the Act,] which such effluent limitations shall require the elimination of discharges of all pollutants if the Administrator finds, on the basis of information available to him (including information developed pursuant to section [315 of the Act]), that such elimination is *1035 technologically and economically achievable for a category or class of point sources as determined in accordance with regulations issued by the Administrator pursuant to section [304(b)(2) of the Act,] or (ii) in the case of the introduction of a pollutant into a publicly owned treatment works which meets the requirements of subparagraph (B) of this paragraph, shall require compliance with any applicable pretreatment requirements and any other requirement under section [307 of the Act.]” [33] Section 304(b)(2)(B) requires the Agency to specify factors to be taken into account in determining the best measures and practices available to comply with subsection (b)(2) of section 301 of the Act to be applicable to any point source (other than publicly owned treatment works) within such categories or classes. Factors relating to the assessment of best available technology shall 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!.] [34] We stress that the variance provision only need apply when the entire impact of a limitation on an individual mill exceeds that which the EPA is authorized to place on the industry. Merely proving that one aspect of the impact, say fuel requirements, is extreme will not necessarily entitle an operator to a variance. He must show instead that when each aspect of the impact on his mill is aggregated, it exceeds the aggregation of impacts that the Agency may place on the industry as a whole. Moreover, we are not insisting upon a variance requirement that hands out exceptions simply because the impact on the individual is greater than that actually imposed by EPA on the industry. Instead, the proper referent is the impact that the Agency could have imposed on the industry under the Act. [35] We turn to the legislative history for the definition of “total cost” under § 304(b)(1)(B). It is meant to include those internal, or plant, costs sustained by the owner or operator and those external costs such as potential unemployment, dislocation, and rural area economic development sustained by the community area, or region. Legislative History, at 231 (remarks of Rep. Jones explaining Conference Committee Report on the floor of the House). Under this definition, certain economic factors must be considered but they need not be decisive if associated with commensurate pollution-ending gains, and they do not, without more, include the fact that the operator is experiencing difficulty in, or is unable to, absorb the costs. Before duPont, this question of the necessary relevance of economic factors in administering a 1977 variance provision under the Act was the subject of debate among the circuits. In Appalachian Power Co. v. Train, 545 F.2d 1351, 1359-60, modified, 545 F.2d 1380 (4th Cir. 1976), the Fourth Circuit held that all economic factors must be relevant, including, apparently, the “economic capacity of the owner” to absorb costs. Id. at 1359. This conclusion may be somewhat broader than ours and was reached by a different analysis inasmuch as it antedated duPont. See note 30 supra. Under that analysis, Appalachian Power rejected a variance provision in regulations limiting electric power company effluents because the provision totally precluded consideration of cost-related factors. On the other hand, the Tenth Circuit in American Petroleum Inst. v. EPA, 540 F.2d 1023, 1033 (10th Cir. 1976), cert. denied, 430 U.S. 922, 97 S.Ct. 1340, 51 L.Ed.2d 601 (1977), upheld a similar variance provision, concluding that EPA could draw the 1977 variance as narrowly as it wanted (or, presumably, dispense with it altogether) because “Congress did not provide for any 1977 variance.” Whatever the validity of this latter view before duPont, the Supreme Court has now made clear that in order for EPA to retain its desired power under the Act to set binding 1977 limitations on an industry-by-industry basis, its regulations must include a meaningful variance provision. Accordingly, we find the Fourth Circuit’s holding on this matter more persuasive. [36] This same rationale could suggest that Congress rejected the idea of 1977 variances altogether. This interpretation, however, which was first rejected in Natural Resources Defense Council, Inc. v. EPA, 537 F.2d 642, 645-47 (2d Cir. 1976) (EPA has discretion to establish variance provision even though not required to do so by the Act), has now firmly been laid to rest in duPont, 430 U.S. at 128, 97 S.Ct. 965 (EPA must establish variance provision to meet requirements of Act). A somewhat more subtle argument would suggest that even if a variance procedure is a necessity, it need not take any account of economic factors. In addition to the legislative history just cited, this argument could draw support from other congressional statements indicating that the Agency need not undertake a plant-by-plant analysis of the “economic impact of [each] effluent limitation,” but only that it must do so “on the basis of classes and categories of point sources.” Legislative History at 304 (Conference Report). Accord, id. at 254 (remarks of Rep. Dingell); id. at 170 (remarks of Sen. Muskie), quoted in note 52 infra. Read in context, however, none of these statements rules out a variance procedure that allows consideration of individual mill costs associated with pollution control. Instead, they simply relieve the Agency, in issuing the general limitations, from the burden of completing a mill-by-mill study of economic effects. In the case of variances, of course, the mill operator bears the burden of making the requisite showing so that, on the issue of economic impact, as on all other issues, the Agency is not forced to make any showing. Thus, the Conference Report, after expressly excusing EPA from a “plant by plant determination” of the “economic impact of an effluent limitation,” continues: However, after July 1, 1977, the owner or operator of a plant may seek relief from the requirement to achieve [the 1983] effluent limitations . . .. The burden [to meet the requirements of § 301(c)] is on the owner or operator. Id. at 304. Although this passage speaks in terms of the 1983 variance procedure in section 301(c), it clearly reveals that while Congress did not intend for EPA, in setting general limitations, to undertake a mill-by-mill economic impact study, it did expect the Agency to consider those undertaken by individual variance applicants. Under duPont, this same approach is also appropriate to the 1977 variance. [37] An identical variance provision is appended to each set of limitations for each of the 16 subcategories of the bleached paper industry. That provision, see, e. g., 42 Fed.Reg. 1402 (1977), provides as follows: In establishing the limitations set forth in this section, EPA took into account all information it was able to collect, develop and solicit with respect to factors (such as age and size of plant, raw materials, manufacturing processes, products produced, treatment technology available, energy requirements and costs) which can affect the industry sub-categorization and effluent levels established. It is, however, possible that data which would affect these limitations have not been available and, as a result, these limitations should be adjusted for certain plants in this industry. An individual discharger or other interested person may submit evidence to the [permit-granting agency] that factors relating to the equipment or facilities involved, the process applied, or other such factors related to such discharger are fundamentally different from the factors considered in the establishment of the guidelines. ... If such fundamentally different factors are found to exist, the [permit-granting agency] should establish for the discharger effluent limitations in the . permit either more or less stringent than the limitations established herein, to the extent dictated by such fundamentally different factors. Such limitations must be approved by the Administrator of the Environmental Protection Agency. [38] The catalyst for the change is not explained, although we might surmise that the successive decisions in Appalachian Power and duPont impressed the Agency with the importance the courts attach to a meaningful variance provision. See In re Louisiana-Pacific, 10 E.R.C. 1841, 1852 n. 27 (1977), citing and harmonizing Appalachian Power Co. v. Train, 545 F.2d 1351, 1360 n. 23, modified, 545 F.2d 1380 (4th Cir. 1976). In any case, on oral argument of this case, counsel for the Agency basically adhered to the Administrator’s analysis in Louisiana-Pacifíc, stressing only that, in his understanding, EPA did not consider relevant an individual operator’s ability to absorb the cost of pollution controls. [39] With respect to cost, the Administrator stressed that the variance analysis would not be based on a “detailed cost-benefit analysis at the permit granting stage” but instead turns on “cost differentials of the particular point source” in relation to EPA’s projected industry-wide costs. 10 E.R.C. at 1852 n. 27, quoting Appalachian Power Co. v. Train, 545 F.2d 1351, 1360, n. 23, modified, 545 F.2d 1380 (4th Cir. 1976). It is worthy of note that while EPA’s brief in this case largely toes the Agency’s earlier line, see pp. 48-49 supra, at one point it states that “[t]he variance clause was inserted in the [1977] regulations to allow for limited reconsideration of the section 304(b) factors for plants in unique circumstances.” Brief for Respondents, at 46 — 47. This interpretation of the clause adheres to the Administrator’s in Louisiana-Pacific and fortifies our view that, as presently interpreted, the variance has the flexibility necessary to allow EPA promulgation of generally applicable 1977 regulations. [40] See note 6 supra. Some petitioners contend that EPA should have taken receiving water capacity into account in setting pollutant parameters. They argue that the Agency should not have considered BOD as such a parameter because the ocean has so much dissolved oxygen that wastes with high BOD have negligible impact and that it should not have considered pH as a parameter because ocean salts buffer waste acidity. Other petitioners contend that EPA should have taken receiving water capacity into account in subcategorization. They argue that ocean-discharging plants are in a wholly different stance from other plants and should be in a separate subcategory. It is also urged that EPA should be forced to give variances for ocean-discharging plants. Finally, the argument for taking receiving water capacity into account is sometimes couched in terms of an environmental balancing test: since there is no environmental “credit” for preventing discharges that the oceans could treat, and there is an environmental “debit” for the air pollution and sludge disposal problems incident to treatment, the balance should tilt in favor of ocean-discharging plants. We regard all the contentions as equivalent in their crucial component — that Congress intended to let EPA take receiving water capacity into account. [41] Apart from this simple “common sense” version of the argument, there is a more sophisticated economic version called the “optimal pollution” theory. This economic theory contends that there is a level or type of pollution that, while technologically capable of being controlled, is uneconomic to treat because the benefit from treatment is small and the cost of treatment is large. See generally W. Baxter, People or Penguins: The Case for Optimal Pollution (1974); B. Ackerman, S. Rose-Ackerman, J. Sawyer & D. Henderson, The Uncertain Search for Environmental Quality (1974). These economic theories are premised on a view that we have both adequate information about the effects of pollution to set an optimal test, and adequate political and administrative flexibility to keep polluters at that level once we allow any pollution to go untreated. As discussed in this section, it appears that Congress doubted these premises. [42] See, e. g., Legislative History, at 493-94 (Great Lakes); id. at 133 (Atlantic Ocean), id. at 500 (Atlantic Ocean); id. at 718-23 (Puget Sound). [43] See note 6 supra. [44] See In re Louisiana-Pacific Corp., 10 E.R.C. 1841 (1977). This opinion exhaustively discusses the Act’s legislative history. Rather than repeat that discussion here, we have taken a step back and looked at the issue in its broad context. [45] Only one violator of the water quality regulatory acts was brought to court in the first 22 years of regulation. Note, supra note 3, at 985. This period of regulation, and the causes of its failure have been canvassed thoroughly. See D. Zwick, Water Wasteland (1971); Parentau & Tauman, The Efñuent Limitations Controversy, 6 Ecology L.Q. 1, 8-12 (1976); Note, supra note 3, at 984-87. [46] “[T]he greatest political barrier to effective pollution control is the threat by industrial polluters to move their factories out of any State that seriously tries to protect its environment.” Legislative History, at 577 (remarks of Rep. Reuss). “Varying local revenue capabilities, economic pressures, and citizen interest have often stagnated community and State initiative.” Id. at 156 (remarks of EPA Administrator Ruckelshaus). See American Frozen Food Inst. v. Train, 176 U.S.App.D.C. 105, 127, 539 F.2d 107, 129 (1976). [47] Section 502(6), (12), (19) of the Act, 33 U.S.C. § 1362(6), (12), (19). Under these definitions, each pollutant parameter is generic for an industry rather than specific to a site of pollution. Furthermore, EPA need only be satisfied that the alleged pollutant affects the “integrity” of water, not that it harms it. FMC Corp. v. Train, 539 F.2d 973, 983 (4th Cir. 1976). [48] The 1977 amendment for publicly-owned treatment plants actually emphasizes Congress’ reluctance to allow discharge into the ocean as a substitute for treatment. It spells out eight distinct factors that a publicly owned treatment facility must demonstrate to receive a permit that raises its allowable pollution levels based on water quality. Congress felt that publicly owned facilities deserved special treatment not accorded industrial facilities because of the practical difficulties facing the former. While 90% of the major industrial dischargers will meet the 1977 requirements of the Act, S.Rep. No. 370, 95th Cong., 1st Sess. 7, reprinted in [1977] U.S.Code Cong. & Admin.News, pp. 4326, 4333, only 50% of municipal dischargers will do so. It would be highly anomalous for Congress to recognize the special needs of these public polluters by giving them this care *1044 fully narrowed right to relaxed limitations based on receiving water capacity and on an 8-factor test, if all other polluters had an unconditional right to laxer treatment based on receiving water capacity. See also H.Rep. No. 830, 95th Cong., 1st Sess. 113, reprinted in [1977] U.S.Code Cong. & Admin.News, pp. 4424, 4487 (EPA directed to study whether “rum distillers might safely dispose of certain natural wastes untreated in the marine environment”). [49] Further support for this interpretation may be found in section 304 of the Act. This provision requires EPA in setting limitations to balance cost and. “effluent reduction benefits.” Although the section specifies other factors to be considered as well, none relates to the effect of regulation on the treated or receiving waters. The phrase “effluent reduction benefits” avoids any suggestion that receiving water quality is an issue. Effluent reduction occurs whenever less effluent is discharged, i. e., whenever a plant treats its wastes before discharge, and the same degree of reduction occurs whether the discharge is into a small stream or the Pacific Ocean. [50] See Legislative History, supra at 1255 (remarks of Sen. Muskie) (criticizing “useless two million dollar studies proving that sulfite waste liquor harms oysters”); Council on Economic Priorities, supra note 9, at 40. Under the pre1970’s legislation, federal pollution regulation relied principally but unsuccessfully on the “conference” with polluters. Of the 50 federal water pollution abatement conferences held by 1971, 15 directly involved pulp and paper mills. Twenty-eight of the 50 conferences, including several involving the paper industry, failed to produce satisfactory improvement. Id. [51] Although its meaning is clear, this section’s diction is confusing. In substance, it states that “[f]actors” relating to BPTCA (a) “shall include consideration” of cost and benefit and (b) shall “take into account” other specified and unspecified “factors” (emphasis added). [52] Senator Muskie described the “limited” balancing test: The modification of subsection 304(b)(1) is intended to clarify what is meant by the term “practicable”. 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. Legislative History, at 170 (emphasis added). [53] When the mid-course study by the National Study Commission was completed and congressional hearings were held, Congress concluded that “[l]ittle contained in the study of the Commission could be construed as justifying major change in the direction established in 1972. . With respect to the industrial program [which includes the regulations at issue in this case], the committee recognizes and applauds the significant success that most of the Nation’s major industries have attained.” S.Rep. No. 370, 95th Cong., 1st Sess. 1, reprinted in [1977] U.S.Code Cong. & Admin.News pp. 4326, 4327. Accordingly, the 1977 amendments did not change the factors involved in § 304(b)(2), 33 U.S.C.A. § 1314(b)(2) (1977). [54] Besides the main benefit of reduced BOD, there were also benefits from reduced TSS and more neutral pH. Petitioners do not challenge EPA’s manner of balancing cost versus these other benefits. [55] Petitioners provided EPA with a graph showing the cost per pound of BOD removed for each increment of waste treatment. The starting point was the cost of a minimal level of treatment: primary treatment (sedimentation) *1048 plus SSL recovery. See pp.---of 191 U.S.App.D.C., pp. 1022-1024 of 590 F.2d supra. Successive increments showed the cost of secondary or biological treatment of each of a series of increasingly dilute waste solutions: hot caustic extract, evaporator condensate, unbleached white water and fines, and the remaining waste solutions. As one would expect from the law of diminishing returns, each successive increment of waste treatment was less cost efficient. The first steps treated a pound of BOD most cheaply; each successive step removed a pound of BOD in a more costly manner than the previous step. [56] One petitioner urges that we remand on the basis of a different incremental analysis for one mill (the Scott Everett mill). The analysis shows that the last increment of treatment at that mill costs $.65 per pound of BOD removed. The record support for this assertion consists of extracts from testimony at a State of Washington effluent permit hearing on the mill. EPA notes in its brief, without contradiction from petitioners, that the Washington testimony “which occupies almost four thousand pages in the record (Rec. at 41609-45596), was submitted to EPA with a brief cover letter explaining that Scott Paper Company believed it relevant to the rulemaking. Until cited in their brief, the particular items of information now relied upon so heavily by petitioners were never specifically brought to the Agency’s attention.” Brief for Respondents at 72 n. 125. Generally, we are reluctant to consider matter presented to EPA in bulk without meaningful explanation of its significance. Such presentation certainly does not aid, and may well impede, the process of notice-and-comment rule-making aimed at ensuring informed agency action. See note 15 supra. Were we inclined to consider the matter, we doubt that the last incremental cost of $.65, which is only four times the average of the subcategory and not much above the averages for some other subcategories, would be considered “wholly out of proportion” to the benefits derived — the test set forth by Senator Muskie. See note 52 supra. Even if it were the case that a mill’s last incremental cost was “wholly out of proportion” to benefits such mill-specific matters are more appropriately considered in permit-granting proceedings. [57] Concerning SSL recovery, see pp.-- - of 191 U.S.App.D.C., pp. 1020-1024 of 590 F.2d supra, petitioners also claim that EPA underestimated BPCTCA costs for the dissolving sulfite category by underestimating (a) the size of the waste load to be treated and (b) the cost of sludge disposal. The former argument was discussed earlier in light of the procedural problem involved therein. Inasmuch as the latter claim is merely a different version of petitioners’ challenge concerning sludge dewatering technology, it is treated infra in our discussion of that challenge. [58] Petitioners’ reply brief reshapes this same argument into a fusillade of charges about EPA’s accounting, intending to show that SSL recovery cost should have been figured at $17.26/ton of product. EPA has argued that SSL incineration pays for itself once the equipment is installed because energy and chemicals are recovered. Petitioners’ own study from which the $17.26 figure is derived clearly confirms EPA’s contention: the $17.26 figure results from subtracting profits of operation of SSL recovery from the capitalized cost of installation. App. 3549. The issue thus becomes the one discussed in text, i. e., whether to count the cost of SSL installation as a cost of BPCTCA. [59] Petitioners also assert that EPA’s economic impact analysis was invalid because it used lower cost figures than did EPA’s BPCTCA cost analysis. There were two reasons for this difference. BPCTCA cost treatment may or may not include costs that have already been incurred, while the economic impact analysis definitely has no reason to include such costs. Also, the length of the rulemaking necessarily meant that although some costs were not yet incurred when some calculations were made, they had been incurred and were subject to consideration by the time of other calculations. We accordingly find EPA’s accounting practices within the range of reasonableness. [60] App. 2408. The parties dispute the method of calculating energy use in the sulfite subcategory. The sulfite waste treatment process involves taking spent sulfite liquor (SSL), and evaporating and burning it. Many steps in that process require energy, see App. 238, 717, although, as EPA points out, the burning of SSL also recovers some usable energy. App. 2418. Petitioners do not seriously contest EPA’s assessment of the increased energy requirement for the sulfite subcategories. They attempt only to show by dramatic example how much energy is involved. For example, they emphasize that at one mill the demand for electricity will rise 47%. As the Agency notes, this method of statement exaggerates the effect on “energy requirements,” the statutory term, since electricity is only part of the energy used by mills. Much additional energy is generated at the mills by burning coal for steam. App. 2418. Similarly, petitioners emphasize that the amount of energy, although relatively small in percentage terms compared to the energy used for non-waste treatment purposes, is large in absolute terms. As they detail, many homes and schools could run on that energy. Nonetheless, although pollution treatment for immense paper mills could not be expected to run on small energy outlays, Congress has insisted upon such treatment. Hence, we are not persuaded by the industry’s presentation of the figures but instead agree with EPA that a percentage comparison is a meaningful way to present and analyze energy use. [61] Petitioners vividly describe the land area that will be covered by the sludge to be disposed of. Admittedly, the amounts of sludge involved are impressive. Yet, the same waste would be discharged into the water if not dealt with as sludge, and Congress has made the choice between effluent and land disposal. [62] Petitioners also contend that EPA should have assessed various vaguely specified “off-site” environmental impacts, e. g., those involved in transporting waste treatment equipment and supplies. They have presented no hard data showing that any environmental impacts EPA failed to assess are comparable in magnitude to those impacts that EPA did assess. [63] Taken at face value, the provision appears highly restrictive, prohibiting water pollution programs that have significant air or land pollution side-effects even if those were outweighed by benefits to the water. The legislative history indicates, however, that more of a balancing test was intended. The House committee reporting on the provision explained that it “believes that there is little gained in stopping water pollution if the preventive actions cause more environmental damage than it eliminates. Legislative History, at 766. [64] For a discussion of the unusual length and difficulty of the conference on this Act, and the notable contributions it made, see Note, The Federal Water Pollution Control Act Amendments of 1972: Ambiguity as a Control Device, 10 Harv.J.Legis. 565, 569-70 (1976). [65] Section 511(c)(1) of the Act, 33 U.S.C. § 1371(c)(1). The exemption from NEPA was the focus of sharp debate in both houses of Congress, and the ultimate version of the exemption is ambiguous in one respect. Besides requiring agencies to carry out the procedural duty of preparing EIS’s, NEPA apparently requires agencies to carry out the substantive duty of balancing environmental costs against benefits. See Calvert Cliffs’ Coordinating Comm., Inc. v. AEC, 146 U.S.App.D.C. 33, 37, 449 F.2d 1109, 1113 (1971), quoted in Alaska v. Andrus, 188 U.S.App.D.C. 202, 211, 580 F.2d 465, 474 (1978). It is unclear whether § 511(c)(1) exempted EPA from engaging in this substantive NEPA duty. The House version of the provision had broadly exempted EPA from all NEPA duties, procedural and substantive, and it survived a vote on the House floor attempting to delete it. Legislative History, at 536-42. The conference version changed the wording of the House provision considerably, addressing itself to “major federal actions,” a phrase that refers simply to NEPA’s procedural duties. See 42 U.S.C. § 4332(2)(C). This change led to debate about the provision on the floor of the Senate. Senator Muskie maintained that § 511(c) eliminated all NEPA duties, procedural and substantive. Legislative History, at 179-83, 206. Other Congressmen contended that § 511(c) was phrased to eliminate only procedural, not substantive, NEPA duties. Legislative History, at 105-08 (Rep. Dingell), id. at 202-06 (Sen. Jackson), id. at 210-11 (Sen. Hart). A number of difficulties surround any attempt to interpret EPA’s NEPA duties. The law on substantive NEPA duties is still developing, so that an attempt to construe § 511(c) must wrestle with a possible exemption at a time when the general rule is not firmly fixed. Moreover, section 511(c)’s interpretation has been rendered increasingly complex and significant by the passage of a parallel provision for EPA’s Clean Air Act duties. Energy Supply and Environmental Coordination Act of 1974, § 7(c)(1), 15 U.S.C. § 793(c)(1). Accordingly, we continue to deem it appropriate to refrain from deciding the scope of § 511(c) until a case, unlike the present one which is based entirely on § 304, squarely presents the issue. Portland Cement Ass’n v. Ruckelshaus, 158 U.S.App.D.C. 308, 317, 486 F.2d 375, 384 (1973), cert. denied, 417 U.S. 921, 94 S.Ct. 2628, 41 L.Ed.2d 226 (1974). It is enough for our purposes in interpreting § 304 to note that EPA’s duties under NEPA are far more attenuated than any other agency’s, based on stronger congressional willingness to trust EPA in such matters. [66] The “functional equivalence” doctrine was first elaborated by this Court in Portland Cement Ass’n v. Ruckelshaus, 158 U.S.App.D.C. 308, 486 F.2d 375 (1973), cert. denied, 417 U.S. 921, 94 S.Ct. 2628, 41 L.Ed.2d 226 (1974). As courts have developed confidence in EPA’s commitment to the environment, that doctrine has become widely accepted. See Maryland v. Train, 415 F.Supp. 116, 121-22 (D.Md.1976) (collecting cases). [67] Legislative History, at 198. Senator Muskie was “the principal author of the [1972 Act], [the] sponsor and floor manager in the Senate, and the Senate leader at conference negotiations.” Save the Bay, Inc. v. EPA, 556 F.2d 1282, 1286 (5th Cir. 1977). Petitioners have relied upon comments on the floor of the House by Congressman Jones: The managers believe that it would be foolhardy to credit one environmental account and debit another by the same action. Their intent is that the assessment of “best practicable control technology currently available” shall be such that the net effect on water and other environmental needs will be positive and benefíciai, and that other impacts of water quality environmental efforts would not negate the overall benefit of the achievement of higher water quality. Legislative History, at 26 (emphasis added). In light of the structure of the Act and the rest of its legislative history, we are not persuaded that this statement indicates a congressional intent to impose mandatory environmental balancing duties on EPA. Representative Jones’ statement was not paralleled in the Conference Report, nor, extended to its fullest implications, would it comport with the language and structure of the Act that the Conference Committee produced. Were it taken to mean that EPA had to balance environmental impacts, it would indicate that the House’s original views had prevailed despite the Conference’s action in deleting section 101(g) and in simply including environmental impact as one apparently equal factor in a long list of concerns relevant to EPA’s regulatory mission. We believe, therefore, that this statement expressed a hope as to how entrusting EPA with its mission would ultimately work out, rather than a mandate as to EPA’s specific means of carrying out that mission. [68] The House version of § 304 used the bare term “environmental impacts.” The conference version’s shift to “non-water quality environmental impacts (including energy requirements)” reflected a carefully crafted alternative to two unacceptable formulations. On the one hand, as we have discussed, the Act intended to exclude consideration of receiving water quality completely. See pp.---of 191 U.S.App.D.C., pp. 1041-1044 of 590 F.2d supra. If the House provision had been retained, it would have been natural to construe “environmental impacts” as including impacts on water quality. That construction would have suggested that receiving water quality was relevant, since environmental impact on water varies with the nature of the receiving water. By specifying “non-water quality environmental impacts,” Congress kept receiving water quality considerations from being brought into play. On the other hand, if the House’s “environmental impacts” language had been excised entirely, it would have appeared that non-water impacts, as well as water impacts, were off limits as a relevant consideration. This possibility was especially serious at the time because the Act was drafted during a period when a parallel uncertainty was troubling the interpretation of the Clean Air Act, a statute that “does not obviously authorize EPA consideration of the adverse environmental consequences of emission controls.” Stewart, The Development of Administrative and Quasi-Constitutional Law in Judicial Review of Environmental Decision-making: Lessons from the Clean Air Act, 62 Iowa L.Rev. 713, 735 (1977) (footnote omitted). Not until the year after the Water Act’s passage was it held that the Clean Air Act implicitly gave EPA authority to consider environmental impacts in tempering its regulations. Portland Cement Ass’n v. Ruckelshaus, 158 U.S.App.D.C. 308, 486 F.2d 375 (D.C. Cir. 1973), cert. denied, 417 U.S. 921, 94 S.Ct. 2628, 41 L.Ed.2d 226 (1974). Here again, Congress employed careful language to avoid an unfortunate ambiguity in the Act. [69] In the course of EPA’s internal review of its effluent limitations, personnel with non-water environmental responsibilities presumably have a part in assessing the limitations and their impacts. It will assist review if these personnel are identified in the record, and if they indicate their reasons, in light of their own concerns and expertise, for accepting the non-water impacts of the limitations. Such indications will reassure Congress, the courts, industry, and the public that the internal EPA structure that is relied upon to ensure proper environmental consideration is functioning as expected. See generally Gaines, Decisionmaking Procedures at the Environmental Protection Agency, 62 Iowa L.Rev. 839 (1977). [70] The papergrade sulfite petitioners also argue that the BOD and TSS limitations set for them are too stringent because the data used by EPA in determining those levels were unrepresentative of papergrade sulfite mills. The upshot of petitioners’ argument is that because of the absence of extensive pollution control data from the actual mills to be regulated, or at least from some mills exactly like them, there can be no proof that “practicable” technology exists to achieve those goals. Hence, so it is argued, EPA may not impose any 1977 regulations and must rely completely on the 1983 regulations to achieve the 1985 goal of ending pollution. See, e. g., Brief for Petitioners in No. 76-1675, at 44. We begin our analysis of this claim by noting that while the sulfite subcategories produce the highest level of pollution in the industry, see Council on Economic Priorities, supra note 9, at 16-17, they have generally lagged behind other subcategories in putting pollution controls in place. See 42 Fed.Reg. 1419 (1977) (final regulations); App. 2493 (Development Document). Consequently, the Agency embarks upon the urgent task of regulating this segment of the industry with the least amount of directly relevant information to guide it. Congress and the courts have recognized that data-gathering problems faced by the Agency, and have allowed it to make a practicability finding based on “transfer technology,” that is, technology used solely in other industries but reasonably found to be transferable to the industry in question. See, e. g., Legislative History, at • 169-70 (statement of Sen. Muskie); C & H Sugar Co. v. EPA, 553 F.2d 280, 286 (2d Cir. 1977). Adopting this “transfer technology” analysis, we will be satisfied if the EPA’s exercise of discretion in making use of imperfectly representative data passes muster under the general abuse of discretion standard identified earlier. In this instance, that standard is easily met. The Agency readily acknowledged the thinness of its data bases, and explained the reasons— and its efforts to compensate — therefor. E. g., 42 Fed.Reg. 1419-20 (1977) (final regulations); App. 2307-15, 2493-96 (Development Document). Thus, there is no claim that EPA failed either to explain its conclusions or to base that explanation on facts before it. Cf. Tanners’ Council v. Train, 540 F.2d 1188, 1193 (4th Cir. 1976); FMC Corp. v. Train, 539 F.2d 973, 985 (4th Cir. 1976); CPC Int’l, Inc. v. Train, 515 F.2d 1032, 1048-50 (8th Cir. 1975). Petitioners simply take issue with the explanation, and under the circumstances, we feel that EPA could reasonably reach the conclusion it did based on the information it cited. See C & H Sugar Co. v. EPA, supra, 553 F.2d at 287-89. With respect to BOD, for example, the use of dissolving sulfite mill data to establish typical effluent levels for papergrade sulfite mills probably redounded to the benefit of papergrade mill operators because they can expect to have lower effluent levels than their dissolving sulfite counterparts. See 30 Fed.Reg. 1419 (1977). Similarly, the Agency has demonstrated to our satisfaction that the pilot plants .that produced data were designed to reflect full-scale results, and that the small amount of data (11 % of data from 4 of 24 mills used) derived from mixed-grade plants that were not then using the papergrade sulfite process did not significantly undermine the validity of the data. This conclusion is especially supportable in that an Agency comparison demonstrated that the limitations derived from the challenged data required papergrade sulfite mills to achieve a lesser degree of treatment than was demanded on average from nonsulfite mills. App. 2312 (Development Document). Finally, the record reveals that two papergrade sulfite mills can and have achieved the BOD limitations established by EPA, demonstrating even more directly that the technology identified as BPCTCA by the Agency is practicable. See App. 84-106, 2405. As to the TSS limit, petitioners’ complaint about the representativeness of the data stems from a technological dispute over the positive or negative effect that one pollution control mechanism (secondary treatment) aimed at controlling one pollutant parameter (BOD) will have on the level of another type of pollutant (TSS) when both mechanisms are in place at a mill, as they were not at the data-producing mills. EPA has offered evidence tending to support its side of this dispute, and we find the evidence sufficient to convince a reasonable person. See Brief for Respondents, Appendix C. [71] Although petitioners portray the Agency’s treatment of climate as indicative of its under-consideration of several other process-related factors, they only seriously rely on the climate challenge. In fact, this challenge was made twice: once as a general complaint about a lack of subcategorization based on climate and once as a cost argument that the sludge treatment specified for colder climates is unduly expensive. Petitioners also make a brief claim that EPA failed to consider adequately certain factors peculiar to Alaskan paper mills. Brief of Petitioners in No. 76-1690, at 2 n. 1. The effluent limitations were based on a record that included data from two Alaskan mills, and we *1055 do not find their application to Alaskan mills arbitrary or capricious. [72] Bacterial breakdown of paper waste is not very different from bacterial fermentation of foodstuff. Cold climates interfere with biological treatment the same way that refrigeration slows food decay. [73] The activated sludge method recycles and aerates the active biological culture continually to keep biological activity high, allowing treatment to be completed within 6-8 hours. The alternative treatment method is the aerated sludge basin (ASB) method. ASB does not recycle concentrated bacterial cultures, and therefore takes longer, i. e., around 10 days per batch of waste. This latter method is cheaper and more reliable, but takes more time and space. Both methods have been extensively used in various industries. [74] Petitioners also argue that other federal and state laws often force mills in northern climates to discharge effluent at a reduced temperature to prevent destruction of the ice-covering on the receiving water and to protect coldwater fish, and that activated sludge systems, unlike the ASB method, cannot accommodate these temperature requirements. Thus, some mills in northern climates will have to install costly cooling towers or ponds. See App. 779. Although petitioners’ record citations in support of this argument do not show that they apprised EPA about the permit problem during rulemaking, see note 15, supra, we would uphold activated sludge treatment as BPCTCA even had the issue been raised properly. Activated sludge treatment does not create the hot waste problem; it merely leaves that pre-existing problem intact. EPA may properly choose as BPCTCA a method that does not solve other pollution problems besides those regulated by the effluent limitation statute. Cf. United States Steel Corp. v. Train, 556 F.2d 822, 846-47 (7th Cir. 1977) (state pollution requirements do not affect the duty to meet federal effluent limitations). [75] Compare Tanners’ Council v. Train, 540 F.2d 1188, 1194 (4th Cir. 1976) (inadequate record; remanded) with American Meat Inst. v. EPA, 526 F.2d 442, 454-56 (7th Cir. 1975) (record support; affirmed). Petitioners contend the record is inadequate because while EPA listed five plants that met the effluent limitations year round, it listed none in the coldest regions of the country that used BPCTCA. Here again, EPA has been hampered by a lack of data. See note 70 supra. As in many industries, the most *1056 advanced plants in the paper industry have been built in warmer climes, so that much of the Agency’s data also come from those areas. It suffices, however, that EPA established the eight-hour operability of activated sludge systems, and demonstrated that such short detention times make the system relatively immune to cold. App. 2261. [76] Petitioners claim that they are not asking EPA to exempt mills that have not installed SSL recovery facilities from doing so but only to give such mills their own subcategory. The distinction is illusory. Subcategorization is decisive in determining whether mills will have to meet the high standards of superior plants, such as those using SSL recovery, or will only have to meet their own much lower standards that are achievable without that control technique. [77] Petitioners point to EPA’s subcategorization based on the process difference between two methods of SSL recovery: blow pits (the technology installed at older mills, which only recovers 85% of SSL) and vacuum drums (the newer technology, which recovers 95% of SSL). This process choice only makes a difference of 10% of SSL in the waste streams, while the one urged by petitioners would involve a difference of 85%. Put most simply, EPA need not accept a total lack of SSL recovery just • because it accepted a slightly less efficient method of SSL recovery. Looked at from another angle, EPA viewed SSL recovery as an internal control measure that could be included in BPCTCA because it is in common use in the industry. Of the 23 papergrade sulfite mills, only three or four lack SSL recovery. See pp.---of 191 U.S.App.D.C., pp. 1059-1061 of 590 F.2d infra. [78] Three courts reviewing effluent limitations have considered industry demands for excursions. One decision, the first, overruled EPA and granted an excursion provision, but the others have not followed this approach. Compare FMC Corp. v. Train, 539 F.2d 973, 986 (4th Cir. 1976) (granting excursions), with CPC Int'l, Inc. v. Train, 540 F.2d 1329, 1336-38 (8th Cir. 1976) (denying excursions), and American Petroleum Inst. v. EPA, 540 F.2d 1023, 1035-36 (10th Cir. 1976), cert. denied, 430 U.S. 922, 97 S.Ct. 1340, 51 L.Ed.2d 601 (1977) (same). Two courts reviewing individual mill permits have also considered industry demands for excursions. One court, reviewing permits for offshore drilling platforms in the infamously variable climate of Cook Inlet, Alaska, concluded that upset provisions had to be allowed. Marathon Oil Co. v. EPA, 564 F.2d 1253, 1272-74 (9th Cir. 1977). The other denied excursions. United States Steel Corp. v. Train, 556 F.2d 822, 841-42 (7th Cir. 1977). [79] See United States Steel Corp. v. Train, 556 F.2d 822, 842 (7th Cir. 1977). See also Legislative History, at 431: “[Manufacturers of control equipment do not generally produce new equipment until there is a market demand for it, and industrial polluters do not install improved equipment until forced to do so. . Therefore, the success of the new strategy depends on success in developing and marketing the necessary technology.” [80] “[I]f the timetables established throughout the Act are to be met, the threat of sanction must be real, and enforcement provisions must be swift and direct.” Legislative History, at 1483 (Senate Report). [81] This scheme was proposed in Marathon Oil Co. v. EPA, 564 F.2d 1253, 1273-74 (9th Cir. 1977). [82] Adamo Wrecking Co. v. United States, 434 U.S. 275, 98 S.Ct. 566, 573, 54 L.Ed.2d 538 (1978). In Adamo, the Supreme Court refused to permit criminal enforcement of a complex nonnumerical regulation under the Clean Air Act that specified proper “work techniques” for asbestos handling. [83] EPA has included excursions in the ambient air standard regulations it promulgated under the Clean Air Act. See 40 C.F.R. §§ 50.4-.10 (1977), discussed in Marathon Oil Co. v. EPA, 564 F.2d 1253, 1273 (9th Cir. 1977); FMC Corp. v. Train, 539 F.2d 973, 986 (4th Cir. 1976). However, the section of the Clean Air Act under which those standards are issued is not parallel to the effluent limitations section of the 1972 Water Act. Ambient air standards are based on public health considerations, and they apply to entire air quality regions with large numbers of polluters who together produce the air pollution health threat in a region. See 42 U.S.C. § 1857c-A (standards). Excursions are allowed because the public health can stand occasional brief exposures, and because the statistical circumstances are such that large numbers of air polluters cannot be readily orchestrated to avoid relatively frequent and severe collective upsets. In contrast, effluent limitations are based on technology considerations, and apply to individual water polluters. Whether the public health (or even receiving water quality) can take occasional exposures is irrelevant under the Act, and there are no collective circumstances to orchestrate because each polluter is individually responsible for adhering to the limitation. See also Essex Chem. Corp. v. Ruckelshaus, 158 U.S.App.D.C. 360, 486 F.2d 427 (1973), cert. denied, 416 U.S. 969, 94 S.Ct. 1991, 40 L.Ed.2d 558 (1974) (if, unlike here, EPA ignores possibility of upsets in setting clean air standards, it must take them into account separately). [84] Under § 505, 33 U.S.C. § 1365, citizens can sue polluters for excessive discharges. Petitioners contend at oral argument that even if EPA exercised good faith prosecutorial discretion in upset situations, it could not bind citizens to do so. We find this argument speculative at this time especially in light of the rarity of citizen enforcement. See 8 BNA Envir.Rep. 1649-50 (Aug. 24, 1977) (remarks of Assistant Attorney General). If and when a pattern of harassing citizen suits develops, EPA can consider requests to revise its regulations. [85] In this case, Weyerhaeuser Co. argued that it was unreasonable to require plants that shift the grade of product they produce to adhere to the effluent limitations for one particular grade. It was assured by EPA that the effluent limitations did not foreclose the issue: “In the situation posed by Weyerhaeuser, the precise form the effluent limitations take in the mill’s NPDES permit is not specified in the regulations; that is left for the permit-issuing authority to decide.” Brief for Respondent, at 107-08. [*] (lbs pollution/ton product) [**] (gals, water/ton product) [***] (lbs pollution/gals. water) [86] Because petitioners’ primary complaint seems to be that EPA’s limits may force them to utilize costly or impracticable internal controls to cut down on flow, there seems to be no claim that those mills with below average flow would be disadvantaged by EPA’s methodology- [87] See American Iron & Steel Inst. v. EPA, 526 F.2d 1027, 1060-61 (3d Cir. 1975). “Common use” does not mean universal use. We accordingly reject petitioners’ argument that a “common use” finding may not coexist even with a small degree of cross-subcategory variability in flow. See Joint Brief Addressing Common Issues at 36. Nor do we read the dicta in FMC Corp. v. Train, 539 F.2d 973, 980 (4th Cir. 1976), to require “uniform water usage per unit of product” before flow may be taken into account. In FMC, effluent limitations were remanded on the basis of variability in flow. Nonetheless, that case is clearly distinguishable on its facts, because there EPA’s own experts had found cross-subcategory variability in flow so “drastic” that they had not even attempted to. identify and use flows characteristic of plants using common internal controls. That those flows were indeed drastic is indicated by the fact that in a subcategory example discussed in FMC, the flow figure used by the Agency in establishing the limits was four times lower than the maximum flow among the plants in that subcategory. Id. By contrast, the relevant figures in the present case show that for 13 of the 16 subcategories the maximum flow was no more than one and one-half times the average flow used by EPA, and that in the most variable subcategory the differential factor was still only slightly greater than two. [88] All secondary waste treatment systems have the problem of disposing of biological waste. The problem is acute for dissolving sulfite mills because, compared to other mills, they have less primary (nonbacterial) waste — which is *1061 easier to drain — to mix with and thus help dewater the larger amount of bacterial waste. [89] For example, petitioners noted correctly that landfill of material with solids content of 20% or less may not be legal in some states since such moist landfill generates pernicious leachates. App. 2410. They assert further that even if bacterial wastes are prethickened and filtered to 20% solid, they still cannot be pressed successfully, so that the wastes inevitably will make poor landfill. Moreover, although some dissolving sulfite mills currently bum their primary, nonbacterial wastes, App. 26-27, 108-09, petitioners contended that the prethickening chemicals, and infeasibility of pressing, would prevent secondary sludge from being burned in existing sludge incineration equipment. App. 1519. [90] App. 1638-706. Petitioners apparently did not object during rulemaking to the Agency’s reliance on this mill, and, despite its greater progress toward effective effluent treatment, we do not find it fatally unrepresentative of dissolving sulfite mills. [91] Such reliance on a diverse array of promising approaches is common in industry, which often achieves goals by having independent research teams test the available alternatives to find the best one among them.