TONE, Circuit Judge.
This is a review of effluent limitations promulgated by the Administrator of the Environmental Protection Agency under the Federal Water Pollution Control Act Amendments of 1972, 33 U.S.C. § 1251 et seq., 86 Stat. 816 et seq. (hereinafter “the Act”).1 Petitioner is the American Meat Institute (“AMI”), whose members operate slaughterhouses and meat-packing plants throughout the country. The regulations under review limit the quantities of various pollutants which these plants can discharge into waterways. Our jurisdiction is invoked under § 509(b) of the Act.
The Statute
The objective of the Act “is to restore and maintain the chemical, physical, and biological integrity of the Nation’s waters.” Section 101(a). The Act sets as national goals the elimination by 1985 of all “discharge[s] of pollutants into the navigable waters,” and the achievement by 1983, “wherever attainable” of a water quality adequate to maintain aquatic life and allow recreational use. Id.
As intermediate steps to the 1985 goal, § 301(b) of the Act requires2 the achievement
(1) by July 1, 1977 of “effluent limitations for point sources3 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”; and
(2) by July 1, 1983 of “effluent limitations for categories and classes of point sources . . . which (i) shall require application of the best available technology economically achievable [445]*445for 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 304(b)(2) of this Act4
For convenience, we shall refer to the technology which must be used by 1977 as the 1977 technology, and to that which must be used by 1983 as the 1983 technology.
The 1977 and 1983 technologies are to be defined by the Administrator under § 304. Subsection (b) of that section provides that “[f]or the purpose of adopting or revising effluent limitations under this Act,” the Administrator is to publish “regulations, providing guidelines for effluent limitations.” These guidelines are to be promulgated within one year after enactment of the Act,5 “after consultation with appropriate Federal and State agencies and other interested persons,” and they are to be revised at least annually, if appropriate. The guidelines are to identify, in terms of specific pollutants, “the degree of effluent reduction attainable through the application of” the 1977 and 1983 technologies. Thus, subdivision (1) of § 304(b), referring to the 1977 criterion, requires identification of “the degree of effluent reduction attainable through the application of the best practicable control technology currently available for classes and categories of point sources.”6 Subdivision (2), referring to the 1983 criterion, requires identification of “the degree of effluent reduction attainable through the application of the best control measures and practices achievable including treatment techniques, process and procedure innovations, operating methods, and other alternatives for classes and categories of point sources . . . .”
In connection with both the 1977 and 1983 criteria, the guidelines are to specify “factors to be taken into account” in determining the applicable technology. These factors are to include, for the 1977 technology, “consideration of the total cost of application of technology in relation to the effluent reduction benefits to be achieved from such application,” and, for the 1983 technology, “the cost of achieving such effluent reduction.” For both the 1977 and 1983 technologies the factors are to include “the age of equipment and facilities involved, the process employed, the engineering aspects of the application of various types of control techniques [and], process changes,” as well as “non-water quality environmental impact (including energy requirements), and such other factors as the [446]*446Administrator deems appropriate . . . § 304(b)(1)(B) and (2)(B). Finally, the guidelines are to “identify control measures and practices available to eliminate the discharge of pollutants from categories and classes of point sources, taking into account the cost of achieving such elimination of the discharge of pollutants.” § 304(b)(3).
To complement §§ 301 and 304, which govern existing sources, § 306 requires the Administrator to promulgate “regulations establishing Federal standards of performance for new sources” within certain categories of sources. These regulations are to cover only plants on which construction began after publication of proposed new-source regulations for that category.
Section 402 adds to the regulatory scheme a permit system for discharges which replaces the permit system formerly administered by the Army Corps of Engineers under the Act of 1899, 30 Stat. 1152, 33 U.S.C. § 407. Permits may be granted by the Administrator provided the discharger complies with all the requirements of the Act, including those of §§ 301, 302, and 306. The Administrator may delegate his permit-granting authority to the states, if they provide sufficient assurances that they will enforce these requirements.7
Background of the Regulations
The regulations before us cover the “Red Meat Processing Segment of the Meat Products Point Source Category.” The common characteristic of the plants in this segment of the meat industry is that they all slaughter animals (but not poultry) and produce fresh meat, which may be sold as whole, half, or quarter carcasses, or as smaller meat cuts. Plants that produce only fresh meat are called slaughterhouses; those that also produce cured, smoked, canned, or other prepared meat products are called packinghouses. Both types of plants usually perform some by-product processing, such as rendering (separation of fats and water from tissue), blood processing, and hide processing.
EPA8 employed North Star Research Institute to study the industrial processes used by slaughterhouses and packinghouses, the wastes generated, and the treatment technologies in use or available to these plants, and to recommend, inter alia, effluent limitations under § 301(b). North Star proceeded to study relevant literature and information on the meat industry it had previously gathered for EPA. In conjunction with AMI, it prepared questionnaires which were distributed to slaughterhouses and packinghouses. From the responses to the questionnaires and information acquired from various other sources, North Star classified the plants into four subcategories and attempted to identify those in each subcategory having the most effluent control. To verify the questionnaire responses, selected plants from these groups were inspected and monitored to a very limited extent. In June 1973, North Star submitted to EPA a report in which the information North Star had gathered was collected and summarized, and analyses and recommendations were presented.
After reviewing the North Star report, distributing copies to industry representatives, and receiving their comments, EPA revised the report and published the revision as a Draft Development Document in October 1973. The standards recommended in this document were then incorporated into proposed regulations, which the agency published [447]*447the same month. Proposed EPA Reg. 40 C.F.R., part 432, 38 Fed.Reg. 29858 (Oct. 29, 1973).
After publication of the proposed regulations, EPA received further comments. On February 28, 1974, it promulgated the final regulations which are the subject of this review proceeding. 40 C.F.R., part 432, 39 Fed.Reg. 7894. In addition, a revised version of the October 1973 Draft Development Document was published under date of February 1974 as the Final Development Document (hereinafter sometimes cited as FDD).
The Regulations
The regulations classify slaughterhouses and packinghouses into the following four subcategories: 9
(1) simple slaughterhouses, which slaughter animals and perform a limited number, usually no more than two, by-product processing operations (subpart A, §§ 432.10 — 432.16);
(2) complex slaughterhouses, which slaughter animals and perform several, usually three or more, by-product processing operations (§§ 432.20 — 432.36); and
(3) low-processing packinghouses, which not only slaughter animals but process meat from animals killed at that plant into cured, smoked, canned, and other prepared meat products, normally processing less than the total kill (§§ 432.30-432.36); and
(4) high-processing packinghouses, which not only slaughter animals but process meat from both animals killed at the plant and animals killed elsewhere (§§ 432.40 — 432.46).
For existing sources in each subcategory, the regulations set forth “[e]ffluent limitations guidelines” for 1977, which are apparently intended to constitute both guidelines under § 304(b) and effluent limitations under § 301(b), 40 C.F.R. §§ 432.12, 432.22, 432.32, 432.42. The same is true of the 1983 standards. 40 C.F.R. §§ 432.13, 432.23, 432.33, 432.-43.10
The regulations limit the discharge of “BOD5,” “TSS,” and ammonia, in addition to other pollutants not involved in this proceeding. Two of these terms require explanation:
BOD5. The initials “BOD” stand for “biochemical oxygen demand” and describe pollutants which, when they decompose, deplete oxygen necessary to support aquatic life. BOD5 is BOD measured over a five-day period.
TSS. The initials “TSS” stand for “total suspended solids,” which are particles of organic and inorganic matter suspended in the water or floating on its surface.
The regulations permit the discharge of certain amounts of BOD5 and TSS per 1,000 pounds (or per 1,000 kilograms) of live weight killed (“LWK”). The 1983 ammonia standard is set in terms of milligrams of ammonia per liter of effluent (mg/1), which shows the concentration of ammonia in the effluent. The regulations challenged in this case are the existing source limitations for 1977 and 1983 relating to BOD5 and TSS, and those for 1983 relating to ammonia.
[448]*448These limitations are set out in the following table:
A. Simple B0D5
slaughter- TSS houses Ammonia
B. Complex B0D5
slaughter- TSS houses Ammonia
C. Low B0D5
processing TSS packinghouses Ammonia
D. High B0D5
processing TSS packinghouses Ammonia
1977 1983
Maximum Maximum Dally Daily
Average Average
for 30 for 30
Consecutive Consecutive Days Days
.12 .03
.20 ’ .05
— 4.00
.21 .04
.25 .07
— 4.00
.17 .04
.24 .06
— 4.00
.24 .08
.31 .10
— 4.00
The maximum discharge for any individual day is twice the maximum daily average for any 30 consecutive days.
I.
Jurisdiction and EPA’s Authority To Promulgate Effluent Limitations Under § 301
At the threshold, we are met with a challenge to our jurisdiction. AMI’s petition for direct review of the existing source regulations is grounded on § 509(b)(1), which provides in pertinent part as follows:
“Review of the Administrator’s action . (E) in approving or promulgating any effluent limitation or other limitation under section 301, 302, or 306 . . . may be had by any interested person in the Circuit Court of Appeals of the United States for the Federal judicial district in which such person resides or transacts such business . . . .”11
We thus have jurisdiction to review the existing-source regulations before us if they are “effluent limitation[s] . under section 301.”
The regulations unquestionably fall within the statutory definition of effluent limitations. Section 502(11) defines “effluent limitation” as “any restriction established by a State or the Administrator on quantities, rates, and concentrations” of discharges from point sources. Furthermore, the preambles to both the proposed and final regulations state that the regulations are promulgated pursuant to § 301 and § 304(b). 39 Fed.Reg. 7894 (1974) (final regulations); 38 Fed.Reg. 29858 (1973) (proposed regulations). Thus, on the surface, there would appear to be no question that the regulations are “effluent limitations” and “promulgated under § 301.”
Nevertheless, and although the parties agree that the Administrator had authority to establish these regulations under § 301, his authority has been challenged by amici curiae,
Amici assert that individual effluent limitations must be established for each existing point source through the permit-issuing process of § 402, using regulations promulgated under § 304(b) as guidelines. Thus, effluent limitations would be set on a case-by-case basis, rather than being prescribed by regulations covering entire subcategories. Under this view, the Administrator lacked the authority to establish across-the-board effluent limitations by regulation, so the regulations were not properly issued as § 301 limitations and are therefore not reviewable here. In essence, this was the position adopted by the Eighth Circuit in CPC International Inc. v. Train, 515 F.2d 1032, 1037 (8th Cir. 1975). In that case, the court held that it lacked jurisdiction to review similar regulations promulgated for a different point source category. The Third Circuit, in American Iron and Steel Institute v. EPA, 526 F.2d 1027, No. 74-1640 (Nov. 7, 1975), has reached an opposite result, as have several district courts. E. I. DuPont de Nemours & Co. v. Train, 383 F.Supp. 1244, 1253 (W.D.Va.1974), appeal pending, No. 74--2237 (4th Cir.); American Paper Inst. v. Train, 381 F.Supp. 553, 554 (D.D.C.1973), appeal pending, No. 74-1544 (D.C.Cir.); American Petroleum Inst. v. Train, No. 74-F-8 at 6 (D.Col., April 8, 1975).
In EPA’s view, the Act calls for the setting of across-the-board effluent limitations pursuant to § 301(b), based on guidelines prescribed pursuant to § 304(b).14 The permit-issuing process, according to EPA, is a mechanism for verifying compliance by each plant and individualizing the effluent limitations to the extent required by the peculiarities of individual point sources.15
In choosing between these conflicting views, we are guided by the teaching of the Supreme Court in Train v. Natural Resources Defense Council, Inc., 421 U.S. 60, 95 S.Ct. 1470, 43 L.Ed.2d 731 (1975), which arose under the Clean Air Amendments of 1970, 42 U.S.C. § 1857a et seq. The courts of appeals had given varying interpretations of that Act, all of which differed [450]*450from the one adopted by the agency. Noting that “[t]he disparity among the courts of appeals rather strongly indicates that the question does not admit of an easy answer,” the Court said that while the agency’s construction was not “the only one it permissibly could have adopted, ... it was at the very least sufficiently reasonable that it should have been accepted by the reviewing courts,” 421 U.S. at 75, 95 S.Ct. at 1480, and, further, “sufficiently reasonable to preclude the Court of Appeals [in that case] from substituting its judgment for that of the Agency.” Id. at 87, 95 S.Ct. at 1485. Our inquiry then is not whether the agency’s interpretation of § 301 is the only permissible one, but rather whether it is sufficiently reasonable to preclude us from substituting our judgment for that of the agency.16. See also, McLaren v. Fleischer, 256 U.S. 477, 480-481, 41 S.Ct. 577, 65 L.Ed. 1052 (1921).
The heart of the controversy is the interpretation of §§ 301, 304, and 509(b). As we have already noted, in providing for direct review in the courts of appeals of “the Administrator’s action . in approving or promulgating any effluent limitation . . . under section 301,” § 509(b)(1) appears to contemplate that the Administrator will adopt effluent limitations under § 301. While § 301 itself does not expressly direct the Administrator to promulgate effluent limitations, subsection (a) of § 301 provides that “[e]xcept as in compliance with this section . . the discharge of any pollutant by any person shall be unlawful”; subsection (b) requires the achievement of certain “effluent limitations for point sources” by 1977 and other “effluent limitations for categories and classes of point sources” by 1983 — language difficult to reconcile with the view that individual effluent limitations are to be set when each permit is issued; and subsection (e) adds that “[e]ffluent limitations established pursuant to this section or section 302 of this Act shall be applied to all point sources . .” Finally, the first sentence in § 304(b) provides that “[f]or the purpose of adopting or revising effluent limitations under this Act the Administrator shall . . . publish . . . guidelines for effluent limitations . . . .”
In addition to these provisions, the language of several other sections of the Act relating to effluent limitations supports EPA. Section 302(a) allows, under certain circumstances, stricter effluent limitations than the “effluent limitations required under section 301(b)(2).” See also § 302(c). Section 303(d)(1)(A) requires each state to “identify those waters within its boundaries for which the effluent limitations required by section 301(b)(1) . . . are not stringent enough to implement any water quality standard applicable to such waters.” And § 309(a)(3), (c), and (d) prohibit violations of “section 301, 302 ., or . . of any permit condition or limitation implementing any of such sections in a permit issued under section 402 of this Act . . . .” See also § 316(b) and (c).17 The reference to limitations in these sections, while not specifying how or by whom they are to be established, is nonetheless “further support for the position that Congress intended the section 301(b) limitations to have an independent existence” apart from the permit process. American Iron [451]*451and Steel Institute v. EPA, supra, 526 F.2d at 1039.
Under § 401(a)(1), applicants for any federal license must obtain state certification that they comply with § 301 or that “there is not an applicable effluent limitation . . . under sections 301(b) and 302 . . .” We find this language especially significant because it cannot be construed as referring to §§ 301(a), (c) or (f), the explanation the Eighth Circuit gave for other references to “effluent limitations under § 301.” CPC International Inc. v. Train, supra, 515 F.2d at 1042-1043. In addition, § 505(f), which defines “effluent standard or limitation under this Act” for purposes of § 505 (the citizen suit provision), includes in the definition, “(2) an effluent limitation or other limitation under section 301 or 302 of this Act,” and “(6) a permit or condition thereof . .” We agree with the courts in American Iron and Steel Institute v. EPA, supra, 526 F.2d at 1038, and E. I. DuPont de Nemours & Co. v. Train, 383 F.Supp. 1244, 1251 (W.D.Va.1974), appeal pending, No. 74-2237 (4th Cir.), that under the interpretation of the Act urged by amici here subsections (2) and (6) of § 505(f) would be redundant, and disagree with the Eight Circuit (CPC International Inc. v. Train, supra, 515 F.2d at 1043) that the reference to § 301 in § 505(f)(2) is to § 301(f). In summary, the most natural reading of the language of the Act is that § 301 is a source of authority to promulgate effluent limitations, independent of the § 402 permit procedure.
The legislative history also contains support for the EPA position. Senator Bentsen, a member of the Public Works Committee that reported out the original version of the Act, stated during the Senate debate:
“In phase I, for point sources of pollutants, effluent limits shall be established not later than January 1, 1976 [now July 1, 1977], which comply with specifically defined levels of effluent control and treatment. As defined in section 301(b)(1) of the bill, and as elaborated in the regulations which we anticipate the Administrator shall issue pursuant to sections 301 and 304, these . . . goals shall be at least the ‘best practicable control technology currently available’ for [industrial] point sources . . . .” Quoted in Congressional Research Service, A Legislative History of Water Pollution Control Act Amendments of 1972, at 1283 (1973) (emphasis added) (hereinafter “Leg.Hist.”).
The Senate Report stated specifically that, “pursuant to subsection 301(b)(1) (A) and section 304(b)” the Administrator is to interpret “best practicable” as a “basis for specifying clear and precise effluent limitations.” Leg.Hist. 1468. Also, during Senate consideration of the conference committee report, Senator Muskie, the principal author of the Act, explained:
“[T]he conference agreement provides that each poluter within a category or class of industrial sources will be required to achieve nationally uniform effluent limitations based on ‘best practicable’ technology no later than July 1, 1977. This does not mean that the Administrator cannot require compliance by an earlier date; it means that these limitations must be achieved no later than July 1, 1977, that they must be uniform, and that they will be final upon the issuance of a permit under section 402 of the bill.” Leg.Hist. at 162 (emphasis added).
In a written explanation prepared by Senator Muskie and submitted by him to the Senate during the debate on the conference report, he indicated that “practicability” and “availability” are not to be determined on a plant-by-plant basis.18 Rather, he explained, the conferees intended that
“the factors described in section 304(b) [cost, age of equipment, type of manufacturing process, engineering aspects of pollution control techniques] be con[452]*452sidered only within classes or categories of point sources and that such factors not be considered at the time of the application of an effluent limitation to an individual point source within such a category or class.” Leg.Hist. 172.
Cf. Leg.Hist. 169-170, 254-255. It is unnecessary to multiply these examples. Those we have given are- sufficient to show the reasonableness of the EPA position.
Much of the remaining legislative history, including some cited by the Eighth Circuit, CPC International Inc. v. Train, supra, 515 F.2d at 1039-1040, is ambiguous. The ambiguity derives from § 502(11), which defines “effluent limitation” to mean any restriction on discharges established by the Administrator or a state, “including schedules of compliance.” Consequently, statements in the legislative history about the role of the states in setting “effluent limitations” can be read to refer either to target limitations or to schedules of compliance.
The EPA position represents a reasonable accommodation of the policies embodied in the Act. On the one hand, nátionwide effluent limits will insure the uniformity Senator Muskie and others stressed. See Leg.Hist. 132, 162, 170, 309, 466-467, 517, 577, 711, 1219, 1405. On the other hand, the states will retain a major role in regulating water pollution, in accord with § 101(b) of the Act. In administering the permit system, the states will have to specify schedules of compliance and determine in some cases whether a variance is justified or whether stricter discharge controls are needed to achieve water quality standards. Thus, the EPA position gives weight to both the policy of uniformity and that of federalism.
EPA’s interpretation also avoids anomalies that would result from acceptance of the Eight Circuit’s interpretation of the Act in the CPC case. Under the CPC interpretation, individual EPA permits (§ 402) based on the nationally-uniform guidelines would be directly reviewed in the court of appeals, § 509(b)(1)(F); yet the nationwide guidelines themselves would be reviewed in the first instance by the district court. CPC International Inc. v. Train, supra, 515 F.2d at 1038. Similarly, variances from the 1983 effluent limitations (§ 301(c)) would be directly reviewable in the court of appeals, 515 F.2d at 1043; but the effluent limitations themselves, which apply to “categories and classes” of point sources, would be reviewed initially by the district court. These results conflict with the congressional purpose of using direct review in the courts of appeals to insure expeditious and consistent application of effluent guidelines. See E. I. DuPont de Nemours & Co. v. Train, supra, 383 F.Supp. at 1253-1254. Cf. Leg.Hist. 1503.
We conclude that the position chosen by the EPA “was ‘correct,’ to the extent that it can be said with complete assurance that any particular interpretation of a complex statute such as this is the ‘correct’ one.” Train v. Natural Resources Defense Council, Inc., supra, 421 U.S. at 87, 95 S.Ct. 1470, 1485, 43 L.Ed.2d 731. We therefore sustain EPA’s interpretation of the statute, and find that it had the authority to issue effluent limitations under § 301 and that we have the authority to review the regulations under § 509(b)(1).*
II.
The Standard of Review
We begin our discussion of the merits by noting the relevant standard of review. Under § 10(e) of the Administrative Procedure Act, 5 U.S.C. § 706(2)(A), agency action in an informal rulemaking proceeding is to be sustained unless “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” See also Camp v. Pitts, 411 U.S. 138, 141-142, 93 S.Ct. 1241, 36 [453]*453L.Ed.2d 106 (1973). This standard requires us to determine whether “the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment.” Citizens To Preserve Overton Park v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 824, 28 L.Ed.2d 136 (1971). We are not to set the effluent limitations ourselves or substitute our judgment for the agency’s. Id.; Portland Cement Association v. Ruckelshaus, 158 U.S.App.D.C. 308, 486 F.2d 375, 402 (1973), cert. denied, 417 U.S. 921, 94 S.Ct. 2628, 41 L.Ed.2d 226 (1974). Rather, we are to determine whether the limitations set by the agency are “the result of reasoned decision-making.” Essex Chemical Corp. v. Ruckelshaus, 158 U.S.App.D.C. 360, 486 F.2d 427, 434 (1973), cert. denied sub nom., Appalachian Power Company v. EPA, 416 U.S. 969, 94 S.Ct. 1991, 40 L.Ed.2d 558 (1974). If the basis stated by the agency for its decision is insufficient, we may not supply another that the agency itself has not chosen to rely on. SEC v. Chenery Corp., 318 U.S. 80, 63 S.Ct. 454, 87 L.Ed. 494 (1943), 332 U.S. 194, 196, 67 S.Ct. 1575, 91 L.Ed. 1995 (1947); FPC v. Texaco Inc., 417 U.S. 380, 395-396, 94 S.Ct. 2315, 41 L.Ed.2d 141 (1974). We must, however, “uphold a decision of less than ideal clarity if the agency’s path may reasonably be discerned.” Bowman Transportation, Inc. v. Arkansas-Best Freight Systems, Inc., 419 U.S. 281, 286, 95 S.Ct. 438, 442, 42 L.Ed.2d 447 (1974).
III.
The 1977 BOD5 and TSS Limitations
AMI’s first challenge is directed at the 1977 effluent limitations, which require application of “the best practicable control technology currently available.” For guidance in interpreting that term, EPA looks to Senator Muskie’s written explanation to the Senate, referred to at note 18, supra, in which he stated as follows:
“In defining ‘best practicable’ for any given industrial category, the Committee expects the Administrator to take a number of factors into account. These factors should include the age of the plants, their size, the unit processes involved, and the cost of applying such controls.
“The Administrator should establish the range of ‘best practicable’ levels based upon the average of the best existing performance by plants of various sizes, ages, and unit processes within each industrial category. In those industrial categories where present practices are uniformly inadequate, the Administrator should 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 practicably applied.
“ ‘Best practicable’ can be interpreted as the equivalent of secondary treatment for industry, but this interpretation should not be construed to limit the authority of the Administrator.” Leg.Hist. 169-170.
This, we think, is a reasonable view of the Administrator’s responsibility. The “best practicable technology” will normally be defined based on the average performance of the best existing plants. If, however, the Administrator concludes that present practices in an industrial category are uniformly inadequate, he may require levels of control based on technology not presently in use in the category (or, it would seem, technology in use only by a single plant), if he determines, by applying the criteria listed in § 304(b)(1)(B), that this technology can be practicably applied throughout the category. One of these criteria is the cost of applying the proposed technology in relation to the resulting effluent reduction. With these principles in mind, we turn to AMI’s challenges to the 1977 standards.
A.
Technology Relied on by EPA for Achievement of 1977 BOD5 and TSS Standards
It appears from the EPA comments introducing the final regulation, 39 Fed. Reg. 7896 (February 28, 1974), that the [454]*4541977 effluent limitations are based primarily on the technology of biological treatment through a three-lagoon system. This is considered “secondary” treatment, that is, treatment which takes place after the waste water has passed through “primary,” in-plant treatment systems. In a three-lagoon system, waste water from the plant flows first into the anaerobic lagoon, where organic matter in the effluent is partially consumed by anaerobic bacteria (bacteria that do not require free oxygen). To increase oxygen levels in the waste water, it is then mechanically aerated in the aerated lagoon. The water then flows to the aerobic lagoon, where most of the remaining organic matter is consumed by aerobic bacteria (bacteria that do need oxygen). After being held there for a relatively long period, the waste water is discharged.
EPA estimates, based on a sampling survey, that 63% of the slaughterhouses and packinghouses discharging into waterways19 already have lagoon systems. FDD 125. Apparently, many of these systems do not include mechanical aeration. The cost-benefit analysis made by EPA is based on the cost of adding mechanical aeration to those lagoon systems. 39 Fed.Reg. 7896. Other treatment systems mentioned in the comments were not subjected to the cost-benefit analysis required by § 304(b). Because only the three-lagoon system has been cost-justified, it is the only technology that can be considered “practicable” as that term is defined in § 304(b).
B.
The Effect of Climate and Temperature on the Efficiency of Aerobic and Anaerobic Lagoon Systems
AMI’s first argument, aimed at the 1977 effluent limitations for all four subcategories of plants, is that, while the proposed lagoon system qualifies as practicable, it cannot achieve the limitations on a year-round basis because of seasonal and climatic effects. Winter conditions, according to AMI, impair the efficiency of both anaerobic and aerobic lagoons, while algae growth in the summer increases BOD5 and TSS.
(1) The Effect of Winter Temperatures on the Anaerobic Lagoon
The optimum temperature for an anaerobic lagoon is approximately 90 °F. Cold temperatures cause it to function less efficiently by slowing bacterial activity. The issue is the magnitude of this effect. AMI relies on an authority which says that removals are reduced to 70%.20 It conceded in its submittal to the agency, however, that the effect of winter temperatures on the anaerobic lagoon is small.21
EPA argues that winter temperatures are counteracted by the heat of incoming waste water (80-100 °F.) and by the insulating grease cover that forms over the pool. An article concerning the Wilson plant at Cherokee, Iowa, reports that the grease cover on the anaerobic pool, after taking some time to build up, insulated the effluent and maintained satisfactory temperatures.22 The anaerobic [455]*455pond at that plant operated at a 92% level of efficiency in February 197023 EPA’s conclusion is also supported by data on other plants supplied by the State of Iowa, which show, for example, that at one plant the anaerobic temperature on two dates in January 1972 was 77-78°F.
AMI has not satisfactorily refuted this evidence. Nor has it otherwise demonstrated that whatever decrease in anaerobic removals does take place in winter will affect the efficiency of the system as a whole sufficiently to prevent attainment of the standards. AMI’s own graph, submitted for the purpose of showing seasonal effects at American Beef’s plant at Oakland, Iowa, indicates that over a two-year period the plant met the BOD5 limitation in all but one of the coldest winter months (December through February).24 EPA also points to data from other plants which show that the 1977 limitations were met for extended periods that included winter months. These data are especially significant because they do not show the direct correlation AMI suggests between removal efficiency and cold weather. Finally, even assuming that cold weather reduces the efficiency of the anaerobic lagoon, overall system efficiency would be impaired only slightly, because a lagoon system as a whole responds much less to changed conditions than any of its parts.25
We conclude that there is firm record support for EPA’s conclusion on the effect of cold weather on the efficiency of the anaerobic lagoon.
(2) The Effect of Winter Temperatures on the Aerobic Lagoon
Like anaerobic lagoons, aerobic lagoons operate less efficiently in winter. Cold temperatures inhibit aerobic microorganisms, and ice and snow covers reduce the oxygen content of water. EPA argues that these difficulties can be ameliorated by increasing detention time, thereby giving the microorganisms more time to work, by using additional aerobic ponds, or by using submerged aerators. We agree with AMI that EPA’s argument as to these countermeasures is inadequately supported by the record.26
The record does suggest, however, that winter conditions do not make compliance with the 1977 standards impossible, since some plants have succeeded in complying with the BOD5 standards in winter. One such plant was the Wilson plant at Cherokee, Iowa, which maintained a 45% level of aerobic removal of BOD5 in February. The American Beef plant at Oakland, Iowa, also met the BOD5 limitations during the winter months, as did several other plants. AMI’s argument that some of these plants should be disregarded because they did not discharge in some winter [456]*456months is unsound. As counsel for EPA pointed out during oral argument, a plant which does not discharge during a given period may be continuing its operations while storing its effluent. Our examination of the record confirms that the plants in question continued operations during periods when they did not discharge. The time of release is unimportant, so long as the effluent is successfully treated before release.
Neither party has submitted information on TSS removal in winter. The data on BOD5 removal indicates, however, the incorrectness of AMI’s basic postulate that aerobic activity is sharply reduced during cold weather. Moreover, AMI’s own submittal indicates that TSS concentrations in winter are comparable to those in summer, and we conclude below that the TSS limitations are attainable in summer.
Thus, although the record support for EPA’s position with respect to the effect of cold weather on aerobic ponds is less conclusive than the evidence concerning anaerobic ponds, we think it is still sufficient, bearing in mind the restricted scope of review under the “arbitrary and capricious” standard.
(3) The Effect of Summer Weather on Aerobic Lagoons
Warm weather promotes the growth of algae. On the basis of comments in the record by industry representatives, state pollution authorities, and others about the effect of algae on aerobic lagoons, AMI argues that algae growth increases TSS and BOD5 counts. Two of these comments refer in general terms to problems at individual plants without giving detailed supporting data; another comment is heavily qualified and inconclusive; 27 and the others are purely conclusory.
EPA states that the Illini Beef plant at Genesco, Illinois, and the Swift plant at Glenwood, Iowa, were able to meet the standards during summer months, as was the Routh plant at Sandusky, Ohio. AMI does not respond directly to these assertions, and, from our examination of the record we conclude that data from these plants fail to show the correlation between summer weather and TSS predicted by AMI.
With respect to BOD5, EPA cites data from five plants that complied with the 1977 effluent limitations during the summer. AMI does not contest the figures regarding summer performance for two of these plants (Wilson, Cherokee, and Swift, Glenwood) but argues that Illini Beef and another plant should be disregarded because EPA itself, as shown in the Final Development Document, excluded them from consideration. The fifth plant was not mentioned or relied on by EPA in the Final Development Document. AMI also points out that a sixth plant, American Beef, failed to meet the standards during the summer. Nevertheless, the ability of even two plants using the proposed technology to meet the BOD5 and TSS standards in summer demonstrates that the standards are attainable in warm weather and is sufficient to overcome AMI’s weakly supported position.
In summary, we find sufficient basis in the record for the Administrator’s conclusion that temperature changes do not render the 1977 effluent limitations unattainable by the 1977 technology he designated.
C.
Record Support for 1977 BOD5 and TSS Limitations
We next consider AMI’s argument that the 1977 effluent limitations are arbitrary and capricious because they are not supported by the record. Limitations for three of the four subcategories were based on the performance of exem[457]*457plary plants in each subcategory.28 AMI accepts this method but contests the reliability and significance of the figures used. Limitations for the fourth subcategory, high-processing packinghouses, were based on statistical techniques. AMI concedes that this approach is “not necessarily wrong,” but urges that EPA erred in implementing it.
(1) Simple Slaughterhouses BOD5.
EPA set B0D5 limitations for simple slaughterhouses as the average of the performance of four plants. Data for one of these, Cornwell at Pureellville, Virginia, must be disregarded because the plant did not use the 1977 technology-
Data for the second plant, Collins Packing at Greenfield, Ohio, was derived from questionnaire responses. While the questionnaire responses for most other plants contain data collected at regular intervals over long periods, those submitted by Collins consisted of the results of only two tests. AMI argues that the results of one of these tests were so low as to raise a question about its reliability, and that the other test showed the plant’s effluent level to be over the standard. EPA’s figures show that if the second test alone were used, the plant would be only .0034 lb. BOD5/1000 lbs. LWK over the limitation, and AMI does not dispute this figure in its reply brief. The determination of whether the first test should be rejected because the result is unexpectedly low is peculiarly within EPA’s expertise, and we shall not substitute our judgment for that of the agency. Hence, we accept Collins as support for the standard.
The third plant was Iowa Beef at La-Mars, Iowa. AMI argues that data from this plant should have been disregarded because
“[t]he [aerobic] lagoon was only treating the portion of the waste that did not leak out and could accordingly provide the remaining waste a better degree of treatment. This plant cannot be considered representative of plants that must treat full strength full volume wastes.”
This argument rests on two unarticulated assumptions: first, that wastes leak out faster than water, so that the concentration of waste is decreased by leakage; and second, that aerobic lagoons treat large volumes of waste water less effectively than small volumes. We find no support for these assumptions in the record, and conclude that EPA was justified in utilizing the data from LaMars. This data showed that the plant was .01 lb. BOD5/1000 lbs. LWK over the standard, but since EPA used an averaging process to set the standard for this subcategory, at least one of the numbers averaged must be greater than the average.
Data for the fourth plant, Swift & Company at Glenwood, Iowa, came both from the North Star tests and the State of Iowa. Although AMI originally argued that the North Star tests did not show compliance with the 1977 limitation, it did not refute in its reply brief EPA’s showing to the contrary. The parties also disagree about whether the state data show the plant to be in compliance during some of the months it discharged. Neither side reveals its calculations, but our own29 show that the [458]*458plant was in compliance for 20 consecutive months, as EPA contends. AMI also argues that Swift’s performance should be discounted because the lagoon system was “relatively new,” but the record contains no basis for the implication that efficiency decreases with age. We conclude that the 1977 BOD5 limitation for simple slaughterhouses, even considering the performance of only the three plants which qualify is adequately supported by the record.
TSS.
The TSS limitations for simple slaughterhouses are based in part on treatment systems at three of the four plants on which the BOD5 limitations for this subcategory were based. The Final Development Document states that the three plants operated at 97% efficiency and that assuming an average raw waste load, this would be sufficient to meet the TSS limitations. FDD 142 — 143. We agree with AMI that the record does not support the 97% figure for these three plants.30
EPA does not, however, rely solely on this figure. The Final Development Document places primary emphasis on the actual discharge levels at the three plants. While Swift, Glenwood violated the 1977 limitations during the North Star tests, Iowa State data show that plant to be in compliance for four of the five months in 1972 for which complete data are available. EPA says that Collins was' also well below the 1977 limitations, but the supporting data for this statement is not in the record. (Two other plants were able to attain the 1977 limitations, but EPA did not' rely on these plants as support for the standards because their operations are “very unusual.”)
Meager as the foregoing evidence may seem, AMI itself, in its final submission during the rule-making process, recommended the TSS limitation EPA ultimately adopted. It did so “[o]n the basis of results obtained from several exemplary well operated lagoon systems in various parts of the United States.” We think EPA could reasonably conclude that this limitation was attainable.31
(2) Complex Slaughterhouses BOD5.
EPA based its effluent limitations for complex slaughterhouses on five plants. We are unable to determine from the record whether one of these, Tama Beef at Tama, Iowa, complied with the limitations. The record shows that none of the remaining plants used full 1977 technology. Assuming, as appears to be correct, that mechanical aeration will produce the 40% efficiency claimed by EPA,32 two of the plants, Missouri Beef at Rockport, Missouri, and Iowa Beef at West Point, Iowa, could apparently meet the standards easily by adding the aeration step called for by the 1977 technology. The other two plants also support the attainability of the limitation. AMI admits that one of these, American Beef, met the standards for 15 months out of a two-year period,33 and that the other, [459]*459Armour at Sterling, Illinois, did so for over three-quarters of a two-year period. We conclude that there was sufficient basis for the B0D5 standard for complex slaughterhouses.
EPA contends that one complex slaughterhouse studied, American Beef, achieved the 1977 TSS limitation of .25 kg/1000 kg LWK. See FDD 143. According to AMI, however, North Star tests showed violations at this plant on two of the three test days. One violation was admittedly slight (.01 lb./1000 lbs. LWK over the limit). As to the second alleged violation, AMI points to a test result of .86, greatly in excess of the limit. EPA counters that the latter test result was actually .086, and attributes AMI’s figure to a misreading of the record due to illegibility. EPA maintains that this figure is consistent with other data in the record and results in an average discharge for the three test days of .19, well under the .25 limit.
In its reply, AMI shifted its attack,34 arguing that EPA improperly relied on the North Star data rather than data obtained from the questionnaires. EPA responded at oral argument that the questionnaire data was “suspect” because “concentrations of suspended solids were in the range of a 100 milligrams per liter.” This explanation was unsupported by reference to the record or expert authority, and we have found no support for it in the record.
EPA’s rejection of questionnaire data on this occasion is inconsistent with its preference for data of that kind in other instances in which that data and test data were at odds. In these other instances EPA chose to rely on data from questionnaires and state tests rather than on conflicting North Star tests, on the theory that data collected frequently over extended periods were more reliable than the isolated tests conducted by North Star. We find this approach reasonable and within EPA’s discretion to adopt, but once adopted it should have been applied consistently, unless EPA could provide a reasoned and record-supported explanation for not doing so. EPA has failed to explain satisfactorily why it took a contrary approach on this single occasion.
EPA also attempts to support the TSS limitation with data from five simple slaughterhouses, which it contends met the complex slaughterhouse limitations and had comparable raw waste loads (i. e., volumes of effluent after in-plant treatment).35 Of these five plants, Corn-well must be disregarded because it did not use the 1977 technology. The raw waste load for two other plants, Collins and Iowa Beef, was less than half the average waste load for complex slaughterhouses, and that for both Illini Beef and Swift, Glenwood was 19% below the average.36 We thus have trouble finding [460]*460that any of these four plants furnishes a basis for comparison.
In sum, when we read EPA’s asserted rationale for the 1977 TSS limitation for complex slaughterhouses against the record, we must conclude that EPA has not supplied a reasoned basis for that limitation. We therefore remand the limitation to EPA for further consideration.
(3) Low-Processing Packinghouses BOD5.
EPA refers to five plants in support of the BOD5 standard for low-processing packinghouses. Two of the plants do not use lagoon systems. One of these must be disregarded completely, and the other is relevant only to demonstrate best practicable technology for plants currently using anaerobic contact.37 On the other hand, the Wilson, Cherokee plant fully complies with the 1977 limitations, as does the Routh Packing plant at San-dusky, Ohio, though its operations are in some respects atypical.38 The Iowa Beef plant could also be brought within the standards by the addition of mechanical aeration, in view of the fact that the plant qualifies for an adjusted limitation because it uses imported hides.39 We conclude that EPA’s effluent limitation is sufficiently supported.
We are unpersuaded by AMI’s argument that in setting the standards for low-processing packinghouses below those for complex slaughterhouses, EPA contravened “the definitions and the inherent nature of these categories.” Because packinghouses perform the functions of slaughterhouses, together with additional processes that add to the raw waste load, AMI says the limitation for packinghouses should be higher rather than lower.
Since meat packing operations in low-processing packinghouses contribute relatively little to the waste load, they may be ignored for present purposes. The dispute therefore focuses on the amount and nature of by-product processing done by plants in the two subcategories. EPA contends that the by-product processing done by low-processing packinghouses is “less extensive” than that done by complex slaughterhouses. AMI argues that the contrary is true, and we find the record to be inconclusive. Nor does the record contain information on whether plants in one subcategory tend to use processes that produce a higher waste load than those used by plants in the other subcategory.40
What does appear clearly from the questionnaire data in the record is that, whatever the reason, the average raw waste load is much lower for low-processing packinghouses than for complex slaughterhouses. AMI attacks this data as inaccurate, in part because it conflicts with North Star test data (which was available for only two plants in these [461]*461subcategories). As we said earlier, EPA could properly rely on the questionnaire data even when North Star samplings were inconsistent with that data. We see no reason to believe that industry members would submit inaccurate data, especially when doing so might cause stricter limitations.41
Because treatment systems for other plants in this subcategory performed “rather poorly,” EPA relies solely on the Routh plant as support for its TSS limitation for low-processing packinghouses. We interpret this as a finding that, with the exception of this plant, technology in this subcategory was uniformly inadequate — a finding which EPA was entitled to make. See p. 453, supra. EPA maintains that its “determination to establish the effluent limitation at a level that is double the level being achieved by a major plant is reasonable.” Relying on questionnaire data,42 EPA says the Routh plant’s TSS emissions were less than half the 1977 limitation.
AMI makes two arguments concerning Routh: first, that Routh did not use the 1977 technology; and second, that it was an atypical plant. The first argument was not raised by AMI until its reply brief, where it stated in a footnote that Routh “does not actually employ ‘best practicable control technology’ since it has a series of dissolved air floatation units rather than an anaerobic lagoon.” We would not find this persuasive, even if the argument were timely. (See note 44, infra.) The record demonstrates that Routh’s dissolved air floatation units, combined with other primary treatment, reduced TSS levels from 2.87 to .79, a reduction of 72%. This is a lesser reduction than is ordinarily achieved by an anaerobic lagoon, which, according to the Final Development Document, can remove up to 95% of TSS. Our own inspection of the record confirms that anaerobic lagoons are able to achieve removal rates well over 75%. For instance, North Star tests at Ulini Beef showed a 96.4% removal rate for TSS. Even the American Beef plant, one of the poorer performers with regard to TSS removal, obtained 78% anaerobic removal. Thus, Routh replaced the anaerobic component of the 1977 technology with an apparently less efficient component, but was still able to meet the standards.
AMI also attacks the Routh plant as atypical. The Final Development Document states that
“[t]he degree of secondary processing conducted at any packinghouse is somewhat variable, although a large number of by-product recovery operations are typically practiced.” FDD at 26.
Most low-processing packinghouses apparently do blood or hide processing and inedible rendering, but Routh does none of these. Also, Routh’s raw waste load was below average for a low-processing packinghouse, though waste load varied greatly among such plants. See FDD 43. Despite these atypical features, EPA could reasonably find that a limitation twice that attained by Routh could be attained by other low-processing packinghouses, especially since Routh used a technology less efficient than that recommended for 1977.43
[462]*462(4) High-Processing Packinghouses BOD5.
High-processing packinghouses vary greatly in the quantity of off-site kill processed. For this reason, the method used by the Administrator to set effluent limitations for this subcategory was different from that used for other subeategories. He applied “the exemplary treatment technology proven in use by plants in the other three subcategories to the average raw waste values” for high-processing packinghouses. FDD 144. Using a statistical equation relating raw waste BOD5 to LWK and processed product production, he determined the raw waste BOD5 for plants having a .55 ratio of processed products to on-site kill. He then assumed a removal efficiency of 98.5% based on the performance of various secondary treatment systems, and calculated BOD5 limits for these plants based on the 98.5% figure. Other high-processing packinghouses were given an additional allowance over this base figure.
AMI contends that the 98.5% figure is incorrect because the standards for the other categories require only 98.0% removal. EPA answers that “98.5 percent was found to be a reasonable intermediate point between average and exemplary performance of several biological systems, including anaerobic-aerated-aerobic lagoons . . . .” While this statement is vague, there is support for the 98.5% figure in the record. A' table in the Final Development Document of removal values for various biological systems shows average values of 95.4% for two-lagoon systems and 98.3% for three-lagoon systems, with the best plants reaching 98.9% and 99.5% respectively. FDD 94A. EPA was entitled to base its effluent limitations on the average of the best plants, rather than on the average of all plants, in the subcategories from which the exemplary technology was borrowed. Accordingly, selecting the 98.5% figure, which is lower than the average of the best plants, was not arbitrary or capricious.
AMI made no complaint of the 1977 TSS limitation for high-processing packinghouses in its opening brief, aside from of its general attack concerning the effect of temperatures on lagoon systems, which we disposed of above. EPA’s answering brief noted this omission and made no argument with reference to the TSS limitation. In its reply brief, however, AMI challenged the TSS limitation as “purportedly based on suspended solids removals achieved in the three other categories,” and stated that “if those limitations must be reconsidered, as we contend, the high-processing packinghouse limitations should be reconsidered also.” The challenge not only comes too late,44 but is also too general to permit serious evaluation. Accordingly, we sustain the 1977 TSS limitation for high-processing packinghouses.
IV.
The 1983 Standards
The distinction between the “best practicable” standard governing the 1977 technology and the “best available . . . economically achievable” standard governing the 1983 technology, according to Senator Muskie, “is intended to reflect the need to press toward increasingly higher levels of control in six-year stages.” Leg.Hist. 170. While some factors, are relevant to setting both standards (see § 304(b)(1)(B), (2)(B)), the 1983 effluent limitations are to be based on “a broader range of technological alternatives,” including techniques “which exist in operation or which can be applied as a result of public and private research efforts.” Leg.Hist. 170. No formal cost-benefit analysis is required [463]*463in determining the “best available” technology, though the Administrator is to take cost into consideration. In addition, “rather than establishing the range of levels in reference to the average of the best performers in an industrial category [which is the norm for the 1977 technology, see p. 453, supra], the range should, at a minimum, . be established with reference to the best performer in any industrial category.” Id. In light of the stringent effluent limitations contemplated by the Act for 1983 and the declared national policy of eliminating the discharge of all pollutants by 1985 (§ 101(a)), we believe that the EPA must be upheld if it can show the existence of some technology which, if implemented, may reasonably be expected to achieve the 1983 standards.
AMI’s position is that EPA’s standards for BOD5, TSS, and ammonia concentration are unattainable by the 1983 technology designated by EPA, and therefore cannot stand. We turn now to these arguments.
Technology Relied Upon by EPA for Achievement of the 1983 BOD5 and TSS Standards
The 1983 standards contemplate the implementation of a number of additions and improvements to “secondary” systems, which typically will be the three-lagoon systems on which the 1977 standards are based. Three of these “tertiary” or advanced treatment techniques are the subject of AMI’s criticism: land disposal, sand filtering, and microstraining.
Land disposal is an alternative to discharge into waterways or public treatment systems. It involves disposing of waste water by distributing it through irrigation systems over relatively flat land, surrounded by dikes, upon which a cover crop of grass or hay may be grown. Barring underground seepage, land disposal totally eliminates the discharge of pollutants into public waters and is therefore highly preferable to other treatment systems. As AMI points out, however, in some locales land shortages or soil conditions may prevent use of land disposal, and in others freezing and lack of ground cover may prevent use of that technique in winter. AMI also contends that the EPA underestimated the amount of land necessary for effective disposal.
A sand filter is a specially-prepared bed of sand or similar filter medium through which waste water passes and from which BOD5 and TSS are removed, either by raking the upper layers, through an under-drainage system, or by backwashing. AMI maintains that this technique is ineffective in removing algae, and that, aside from this problem, it is incapable of removing a sufficient proportion45 of the BOD5 and TSS allowed by the 1977 standards to meet the 1983 standards.46
Microstrainers are partially submerged rotating drums which remove BOD5 and TSS through use of a fine mesh fabric. AMI contends that they too are incapable of extracting algae, that they are a recent innovation on which performance data is sparse, and that no record estimate of their efficiency is based on a reliable source.
The points raised by AMI are not without substance. EPA, in fact, has recognized that land disposal will not be universally available and has suggested its use only as a practical and efficient alternative. Also, while the EPA now [464]*464appears ready to defend the efficiency of microstrainers, the Final Development Document notes the paucity of information on their reliability in full-scale operation. FDD 100. The defect in AMI’s argument, however, is that it fails to consider additions to the 1977 technology other than the three just examined.
In particular, in-plant or “primary” techniques may be used to reduce the level of effluent in the water stream at an early stage, thereby reducing the efficiency at which tertiary systems such as sand filters must operate in order to attain the standards. One of these primary techniques is dissolved air floatation, which involves releasing a mixture of compressed air and liquid into the waste water stream to force small particles to the surface. When used with an alkalinity control and chemical flocculation, which causes the particles to aggregate, dissolved air floatation is capable of removing 90% of BOD5 and 98% of TSS, according to unconverted EPA statistics. Various other in-plant controls and modifications, such as improved handling of viscera, paunch, and blood wastes, all demonstrated as technically feasible, are also described in the Final Development Document.47
As to the problem of algae, AMI has submitted no data to show that algae growth in aerobic lagoons will result in excessive TSS in plants utilizing the in-plant controls and modifications and the extensive primary treatment EPA requires. On the other hand, EPA has suggested one method of reducing the algae content of the effluent discharged from the lagoons, namely, by locating the intakes for discharge pipes about a foot below the lagoon’s surface, since algae tend to float on the surface. Also, it-notes that the authority relied upon by AMI in its analysis of the effectiveness of sand filters actually indicates that they are capable of removing 33 to 45% of algae content. We conclude that EPA’s conclusions and the resulting 1983 limitations are based on a reasoned analysis of the record, with one exception. Since the 1983 TSS limitations are based in part on the 1977 limitation,48 the 1983 limitation for complex slaughterhouses should be reconsidered in light of our holding with respect to the 1977 TSS limitation for this subcategory.
1983 Ammonia Standards
A final pollutant which EPA seeks to control by 1983 is ammonia. According to the Final Development Document, the concentration of ammonia in plant waste water ranges from 7 to 50 mg/1 before treatment, rising to a level of 100 mg/1 or more after treatment in an anaerobic lagoon.
In the preamble to the effluent limitations, EPA states as follows:
“The ammonia limitation was derived from engineering judgment as to the reliable capability of the air stripping method of ammonia control. It would appear that the limitation is a reasonable current estimate of the capability of ammonia stripping techniques for controlling this parameter.”
In its brief in this court, however, EPA abandons sub silentio its reliance upon ammonia stripping and relies upon nitrification as the technology by which the effluent limitations can be met, adding that “use of ammonia towers [stripping] is not necessary to meet the standard. Nevertheless, ammonia towers are effective under proper conditions and may be utilized by the industry to meet the standard.”
[465]*465We first examine ammonia stripping, the technology on which EPA originally said it based the limitations. Ammonia stripping removes ammonia from liquid waste by exposing it to an ammonia-free gas such as air. After adjusting the alkalinity of the waste water, it is pumped to the top of a stripping tower and allowed to flow downward, while a counter-current of air introduced at the bottom of the tower flows upward. The method is capable of removing up to 98% of ammonia content. Reaching this efficiency, however, requires a level of air flow that is extremely costly to maintain. Moreover, removal efficiencies begin to drop once the temperature falls below 68° F., according to at least one record reference. Housing the tower and heating the air prior to introducing it into the tower are mentioned in the Final Development Document as means of dealing with the temperature problem but are characterized elsewhere in the record as impractical and extremely expensive. AMI notes finally that stripping towers are subject to scaling — the buildup of a calcium carbonate scale on the tower baffles, eventually reducing the tower’s efficiency— which is very difficult or impossible to remove in hard water areas. Thus, while some of the problems connected with ammonia stripping were recognized in the Final Development Document, neither it nor EPA’s brief suggests techniques which are likely to solve those problems. And, though counsel for EPA stated at oral argument that the agency has not abandoned this technique, we think that the difficulties are of a sufficient magnitude to preclude EPA from relying on this technique alone to meet the standards, at least on this record.49
Since EPA based the ammonia limitations on technology not shown to be feasible, remand is necessary. We cannot sustain the regulation on a basis which counsel now asserts but which the agency did not rely upon in formulating the regulation. Burlington Truck Lines, Inc. v. United States, 371 U.S. 156, 168, 83 S.Ct. 239, 9 L.Ed.2d 207 (1962). It is nevertheless appropriate to go on to express, for EPA’s consideration on remand, our views on the shortcomings of the present record with respect to nitrification, the method EPA relied upon in its brief before this court. Nitrification is the process of decomposing ammonia molecules by oxidizing their nitrogen into nitrites and nitrates, which are then usually converted to nitrogen and nitrogen oxide by a denitrification process. As noted above, the concentration of ammonia in waste water after it passes the anaerobic lagoon stage is 100 mg/1 or more. Hence, in order for nitrification to meet the 1983 limitation of 4.0 mg/1, the process must be capable of reducing the amount of ammonia content by at least 96%. The three studies in the record50 on which EPA relies to support the capability of the nitrification technique to reach 1983 standards show reduction of final ammonia concentrations below 4.0 mg/1, but the ammonia level of the influent (or liquid flowing into the processing unit) in each case was significantly less than 100 mg/1, yielding removal efficiencies of 10 to 90%. Counsel for EPA stated at oral argument that the problem with ammonia removal was [466]*466not in percentage reduction but rather in achieving the base level, implying that a treatment process capable of achieving, for example, a 2 mg/1 effluent could do so regardless of the ammonia concentration of the influent. Perhaps this is so, and if it is, perhaps the point would be obvious to an expert. Since we, however, lack the expertise necessary to make such a judgment and the record is barren on the point, we are unable to uphold EPA, particularly in light of record evidence that the best available ammonia-removing technology applied to influents with 100 to 150 mg/1 ammonia yields a final effluent of 10 to 15 mg/1, substantially in excess of the 1983 standards.51 We therefore remand the 1983 ammonia standards for reconsideration and development of data adequately supporting whatever standards are ultimately adopted.
V.
Effluent Limitations in Issue Compared With Subsequently Issued Meat Processing Standards
Approximately six months after promulgating the regulations under review here for slaughterhouses and packinghouses, EPA issued proposed standards for meat-packing plants, i. e., plants that process meat but do not do their own slaughtering. 39 Fed.Reg. 31486 (August 28, 1974).52 Because few meat processors have their own waste treatment systems, EPA- used data from its study of slaughterhouses and packinghouses in establishing the meat processing standards. AMI complains that “EPA set meat processing standards, especially 1983 meat processing standards, at levels which are significantly more lenient than those appealed by AMI in this proceeding,” and that “it is arbitrary and capricious for EPA to reach two different conclusions as to achievable limitations based upon the same data.”
We do not have before us the record on which the regulations for the meat processing industry were based, and an intelligent comparison of the two industries and the regulations for each is impossible in the absence of full records for both industries. Moreover, as the court said in Portland Cement Association v. Ruckelshaus, supra, 486 F.2d at 389,
“Inter-industry comparisons of this kind are not generally required, or even productive; and they were not contemplated by Congress in this Act [the Clean Air Act], ... It would be unmanageable if, in reviewing the cement standards, the court should have to consider whether or not there was a mistake in the incinerator standard, with all the differences in parties, practice, industry procedures, and record for decision.”
We have decided in parts III and IV of this opinion that the challenged 1977 and 1983 standards for BOD5 and TSS for the Meat Products Point Source Category (with the exception of TSS for complex slaughterhouses) have adequate record support. The fact that less stringent limitations have been set for another category does not require further justification for the present ones.
VI.
Conclusion
To summarize, we have decided that we have jurisdiction to review directly the existing source regulations challenged here. We have also found that temperature and climatic effects do not render the 1977 limitations unattainable, and that, with the exception of the TSS limitation for complex slaughterhouses, the 1977 limitations are based on a reasoned analysis supported by the record. Finally, we have found that the 1983 limitations for BOD5 and TSS are adequately supported (with the same excep[467]*467tion), but that the limitation for ammonia is not.
The only remaining problem is that of fashioning a remedy. The statutory deadline for promulgating existing source standards has passed, and the deadline for compliance with the 1977 standards is fast approaching. In order to expedite the case, we remand it to the EPA with the instructions set forth below, and retain jurisdiction pending the remand. CPC International Inc., supra, 515 F.2d at 1050 (citing cases).
EPA may conclude, on reconsideration, that evidence in the existing record on which it did not previously rely is adequate to support the limitations. If so, it may issue a clarifying explanation and reissue the limitations without further hearings, notice, or opportunity for comment. See South Terminal Corp. v. EPA, 504 F.2d 646, 665 (1st Cir. 1974), and cases cited therein. On the other hand, it may decide to seek new data or to reconsider the limitations themselves. If so, it must follow the procedural requirements of the Act. In either event, if the petitioner is dissatisfied with the agency’s final action with respect to any effluent limitation, it shall have 21 days to file any objection to the effluent limitation in this court. The objection will then be heard by this panel of the court on an accelerated briefing schedule.
We remand for expeditious consideration by the EPA, pursuant to these instructions, of the 1983 ammonia limitations and the 1977 and 1983 TSS limitations for complex slaughterhouses. With these exceptions, we uphold the challenged effluent limitations.
Affirmed in part, remanded in part.
This opinion has been circulated among all judges of this court in regular active service, in view of the conflict between our holding in Part I and that of the Eighth Circuit in CPC International Inc. v. Train, supra, 515 F.2d 1032, and no judge has requested that the issue decided in Part I be reheard in banc.