In these consolidated cases, petitioners Reynolds Metals Company, Miller Brewing Company, and United States Brewer's Association, ask us to set aside as invalid effluent limitations promulgated for the canmaking industry1 by the Environmental Protection Agency (Agency) under the Clean Water Act2 (Act). The canmaking industry discharges in its effluent3 conventional, toxic and nonconventional pollutants. Standards for canmaking were promulgated to control all three types of pollutants, conventional, toxic and nonconventional, but it is that portion of the standards relating to the removal of the total toxic organics (TTO) and toxic metals that generate petitioners' objections. As in other "clean water" regulations, the Agency devised limitations for pollution from canmaking by first determining the best ways to remove pollutants (the model technology), then testing wastewaters to determine the effectiveness of the technology, and prohibiting pollutant discharges in excess of limits determined to be achievable by reference to the model technology.

Petitioners contend that the standards for effluent control are invalid in that the Agency erroneously concluded that because the removal of oil and grease had effectively removed total toxic organics in the aluminum forming and coil coating industries, it would achieve similar results in the canmaking industry. They also contend that the Agency arbitrarily refused to subcategorize the canmaking industry, erred in its pass-through analyses, overstated the presence of chromium, zinc, and copper, and failed to observe its statutory duty to "weigh" costs relating to one standard and to "consider" costs for another. We disagree with all of these contentions and affirm.

Canmaking encompasses all of the manufacturing processes employed in the production of various shaped metal containers used to package and store foods, beverages, and other products. The two major types are two-piece (seamless) and three-piece (seamed) cans. A vegetable or soup can is an example of a three-piece can and an aluminum soda can is an example of a two-piece can. It is only the seamless, or two-piece, cans that are subject to the regulation which is involved in this appeal. The EPA excluded from regulation manufacturers of three-piece cans, can ends, can tops and seamless cans which are not washed because these processes do not generate wastewater.

In the manufacturing of a two-piece can a coiled metal sheet is coated with an oil lubricant and straightened. A machine called a cupper then cuts a circular blank from the metal sheet and forms the blank into a cup that is drawn into the required height and diameter by a process known as ironing. This ironing process is performed by a machine called a body maker. The can bodies are then cleaned, the metal is treated, and coatings and decorations are applied. Finally, the open end of the can body is flanged to receive the can top.

The forming process employs oil lubricants at virtually all stages. In order to remove the lubricants from the can bodies, the process utilizes a can washer, which usually consists of six spray processing stations. After leaving the body maker, the cans are conveyed through the canwasher on a continuous metal belt. The six canwashing stages include (1) prewash, to remove the layer of lubricant remaining on the can from the body maker; (2) acid wash, to further clean and etch the surface of the can; (3) rinse, to further remove contaminants; (4) surface treatment, to prepare the can for decorating by the application of either chromium- or zirconium phosphate-based coatings; (5) rinse, to remove contaminants remaining from surface treatment; and (6) DI rinse, using de-ionized water to rinse off the last remnants of the processing solutions.

Nationally, eighty-six plants generate wastewater from the manufacture of two-piece cans. Of these, eighty are indirect dischargers and three are direct dischargers.4 The remaining three dispose of wastewater by land application. Pollutants found in two-piece canmaking wastewaters5 include (1) conventional pollutants, (2) toxic organics, (3) toxic metals, and (4) nonconventional pollutants.

The Clean Water Act directs the EPA to issue nationally-applicable effluent limitations guidelines and standards for classes or categories of point sources.6 E.I. duPont deNemours & Co. v. Train, 430 U.S. 112, 97 S. Ct. 965, 51 L. Ed. 2d 204 (1977). The standards normally are to apply uniformly.7 After standards and guidelines are established by regulations, facilities may achieve the specified effluent discharge allowance through the use of the technology described in the regulation or in any other manner. The Agency's actions in regulating industrial water pollution have been so frequently the subject of appellate review that detailed references to the statutory scheme mandating regulations seems redundant. An overview, however, is necessary to frame the issues presented by petitioners' objections to the removal technology recommended by the Agency for canmaking as well as the issues relating to the treatability of toxic metals and the final issue of whether the Agency properly considered costs of the removal technology.

In passing the Act, which amended the Federal Water Pollution Control Act of 1972, Congress set as a national goal the elimination, by 1985, of the discharge of pollutants into navigable waters. 33 U.S.C. Sec. 1251(a)(1) (1982). To reach that goal, the Act directed the Administrator of the Agency to promulgate regulations setting limits on the pollutants that can be discharged by "point sources." 33 U.S.C. Sec. 1362(14) (1982).

First, the Act required the Administrator to establish effluent limitations for point sources which discharge pollutants directly into navigable waters i.e. "direct dischargers". The Administrator had to define effluent limitations for categories or classes of point sources that would require existing direct dischargers to employ by 1977 the "best practicable control technology currently available" (BPT), 33 U.S.C. Secs. 1311(b)(1)(A), 1314(b)(1)(A), and to comply by 1984 with limitations based on the "best available technology economically achievable" (BAT). 33 U.S.C. Secs. 1311(b)(2)(A), 1314(b)(2)(B).8 For newly-constructed dischargers, the Administrator had to establish new source performance standards (NSPS) requiring the application of the "best available demonstrated control technology" to remove all types of pollutants. 33 U.S.C. Sec. 1316. The Administrator's BPT, BAT, and NSPS limitations were to be based upon a consideration of the factors specified in sections 304(b) and 306(b) of the Act. 33 U.S.C. Secs. 1314(b)(1)(B), 1316(b)(1)(B).

Second, the EPA is required to establish effluent limitations for point sources that expel wastewater into publicly owned treatment works (POTWs), which treat the wastewater prior to its introduction into public waterways, by requiring such indirect dischargers to pretreat wastewater before allowing it to flow into a POTW. Under unregulated conditions, indirect dischargers ultimately would introduce fewer pollutants into waterstreams than direct dischargers because indirect discharges flow through sewers into POTWs where much pollution is removed before it is, in turn, discharged into a national stream of water. In requiring standards for indirect dischargers, however, Congress realized that a POTW normally would not remove the same amount of pollutants from industrial wastewater as direct dischargers are now required to remove. Additionally, a POTW is unable to successfully operate on some pollutants--specific pollutants might interfere with or be incompatible with its operation. Because of these factors, the Agency is required to establish standards for pretreatment of wastewater before it enters a POTW "to prevent the discharge of any pollutant through [POTWs] which interferes with, passes through, or otherwise is incompatible with such works." 33 U.S.C. Sec. 1317(b)(1). The legislative history indicates that pretreatment standards are analogous to the standards for direct dischargers, i.e. the combined treatment of wastewater by an indirect discharger and the POTW should achieve the same level of pollution removal as would be realized if the industrial source were treating wastewater and then directly discharging it. See H.R.Conf.Rep. No. 830, 55th Cong., 1st Sess. 87, reprinted in 1977 U.S.Code Cong. & Ad.News 4326, 4424, 4462. The EPA accordingly has imposed pretreatment standards both for existing sources (Pretreatment Standards for Existing Sources or PSES) and for newly-constructed facilities (Pretreatment Standards for New Sources or PSNS). 33 U.S.C. Sec. 1317(c).

Third, though not relevant to this appeal, the Act requires that the Administrator set effluent limitations for POTWs that treat municipal sewage and industrial waste. 33 U.S.C. Secs. 1311(b)(1)(B), 1314(d)(1).

In setting standards, the EPA is directed to consider five factors: the age of equipment and facilities, the process employed, engineering aspects of the application of various types of control techniques, process changes, and nonwater quality environmental impacts (including energy demands). 33 U.S.C. Sec. 1314(b). A sixth factor involves cost, and in this regard the Agency is required, for setting BPT limitations, to refer to "total cost of application of technology in relation to the effluent reduction benefits to be achieved by such application." 33 U.S.C. Sec. 1314(b)(1)(B). For BAT, the Act mandates consideration of "the cost of achieving such effluent reduction." 33 U.S.C. Sec. 1314(b)(2)(B).

The EPA, in 1978, began collecting information later used to formulate effluent standards for the canmaking industry. Data was gathered from EPA studies, published literature, trade associations, and can manufacturers. Additionally, meetings were held between the Agency and industry representatives.

The Agency sent a data collection portfolio in 1978 to each company known or believed to be engaged in aluminum forming.9 The portfolios requested specific information concerning production, wastewater management and treatment, cost information, and other pollutant information based on 1977 data. Followup portfolios directed specifically at can manufacturers were mailed and returned in 1982 with similar information based on 1981 production records.10 The 1978 portfolios requested that each company indicate which of a list of 129 TTO pollutants were believed to be present, believed to be absent, known to be present, or known to be absent. The 1982 portfolios added toxic metals and cyanide to this list. Three toxic metals--chromium, copper, and zinc--were often identified in the 1982 responses as believed to be present or known to be present.

The EPA conducted engineering and sampling visits in 1978 and 1979 based on the responses to the first data collection portfolios. Prior to sampling, all available data, including plant and wastewater pretreatment facility layouts and diagrams, were gathered and reviewed. From this information, a detailed sampling plan was generated identifying the points at which samples would be collected. Engineering visits were conducted at seven canmaking plants and five plants were chosen for sampling--four manufacturing two-piece aluminum can bodies and one producing two-piece steel cans.

In conducting the sampling, the EPA took samples from each operation which discharged or used water, including rinses. Both influent and final effluent were analyzed for pollutants. When streams were treated and discharged separately, all of the effluents were measured. The samples were collected and analyzed in accordance with Sampling and Analysis Procedures for Screening of Industrial Effluents for Priority Pollutants, U.S. EPA, March 1977, revised April 1977. With respect to total toxic organics, this sampling revealed seven specific compounds at concentrations greater than 0.01 mg/L.11 Other pollutants detected included conventional pollutants (oil and grease, suspended solids, and pH), toxic metals (chromium, copper, nickel, and zinc), and nonconventional pollutants (aluminum, fluoride, manganese, and phosphorus).

After the data had been analyzed, the EPA, on February 10, 1983, published a proposed regulation in the Federal Register. 48 Fed.Reg. 6267 (Feb. 10, 1983). It outlined various technologies considered in reaching proposed effluent limitations for BAT, BPT, NSPS, PSNS, and PSES and explained its research methods. In setting limitations, the Agency considered various factors, including the cost of applying technology in relation to effluent reduction benefits, the age of the involved facilities and equipment, the processes employed, and additional environmental impacts. The Agency based its proposed limitations on a model technology consisting of a combination of oil and grease removal, chromium reduction and cyanide precipitation, and precipitation and sedimentation methods in conjunction with techniques aimed at reducing the flow of water through the canwashers. It invited comments, however, on more exacting technologies of possible use in meeting BAT, NSPS, PSNS and PSES limitations. The Agency proposed to regulate TTO under PSES and PSNS, but gave to the industry the alternative of monitoring only for oil and grease limits. The Agency reasoned that efficient removal of oil and grease eliminated 97% of the TTO, so that the costly monitoring for toxic organics was unnecessary and that compliance would be assumed upon a showing that the oil and grease standards were satisfied.

The Agency also explained that its proposed setting of limitations for certain pollutants was based on data gathered in regulating other categories of point sources. This included borrowing values for aluminum from aluminum-forming and coil coating data; for fluoride and phosphorus from values achieved in the electric and electronic component manufacturing industries; and for oil and grease from data derived from the coil coating, aluminum-forming, and copper-forming industries. The Agency referred to these industries from which it transferred data as the combined metals data base (CMDB). The Agency also referred to the CMDB in determining to regulate suspended solids, chromium and zinc.

The Agency, in the preamble to the proposed regulation, referred to the CMDB in explaining aspects of its proposed model technology. With respect to the oil removal component, i.e. skimming, dissolved air flotation, and chemical emulsion breaking, the Agency reasoned that because canmaking generated amounts of oil and grease comparable to that from coil coating and aluminum forming, this technology could be employed in canmaking as well. Although recognizing that canmaking wastestreams contained different pollutants than those appearing in coil coating and aluminum forming effluent, due to the greater number and variety of forming lubricants and cleaning formulations employed in canmaking, it concluded that "by controlling the most prevalent toxic metals, some conventional and nonconventional pollutants, and total toxic organics (TTO) with oil removal and lime and settle technology, pollutants present as a result of these variations will also be controlled." The incorporation into the proposed model technology of filtration and of hydroxide precipitation and sedimentation was also based on results achieved by similar technologies in the CMDB.

The proposed regulation solicited comments on all aspects of the regulation, including data on steel canmakers, the use of filtration, the effectiveness of oil skimming technologies and precipitation and sedimentation systems, the use of the CMDB, and the reasonableness and achievability of the Agency's cost analysis. The Agency also requested the submission of additional data from canmaking plants employing properly-operated model technologies.

Following the publication of the proposed rule, the Agency provided the development document and the economic impact analysis supporting the proposed rule to industry, government agencies, and the public. A public hearing was held in Washington, D.C. on April 27, 1983 at which one person presented testimony. Additionally, fourteen commenters submitted a total of approximately 330 individual comments on the proposed regulation.

Comments addressed (1) perceived inaccurate data, (2) difficulty in achieving water flow reduction, (3) transferability of technology or data from CMDB, (4) perceived inaccuracies in evaluating pollutants, (5) regulation of aluminum for indirect dischargers, (6) alleged erroneous finding that certain pollutants appeared at treatable levels in canmaking water streams, (7) alleged failures to consider use of synthetic lubricants, (8) use of mass-based standards rather than those based on concentration, (9) alleged miscalculation of compliance costs, (10) economic impacts, and (11) the effect of suggested deposit legislation on future demand for two-piece cans. Many of the comments were generated by a self-sampling program of fourteen aluminum canmaking plants initiated by industry trade associations, the Can Manufacturing Institute (CMI) and the United States Brewer's Association (USBA) after promulgation of the proposed regulation. The Agency accepted some suggestions contained in the comments but rejected most. It responded to all of them.

After the comments were submitted, and in response to the CMI and USBA sampling data, the Agency conducted post-proposal sampling for metals at seven plants and for TTO at five plants. The samples taken during this period were "grab" samples, i.e. short term samples which were not conducted in the same manner as the pre-proposal samplings. These grab samples consisted of process wastewater before treatment (seven plants), treated wastewater (six plants), and untreated individual process streams (two plants). This sampling revealed the presence of seven additional toxic organic pollutants appearing at treatable levels in canmaking wastestreams.12 The EPA published a notice in the Federal Register on September 22, 1983 describing the post-proposal sampling and the Agency's preliminary analysis of data submitted by the various commenters. 48 Fed.Reg. 43195 (Sept. 22, 1983). Six commenters submitted about fifty comments on the data and issues raised in the September 22, 1983 notice. The Agency, in turn, responded to the additional comments and made certain modifications based on industry submissions.

The Agency published the final regulation for the canmaking industry in November 1983. In promulgating the regulation, the Agency identified a model technology consisting of an "end of pipe treatment" in conjunction with flow reduction techniques. The "end of pipe treatment" includes the removal of oil and grease from wastewater through the use of oil skimming, chemical emulsion breaking, dissolved air flotation, or a combination of these processes. The removal of metals, fluoride, and phosphorus is accomplished by lime precipitation and chemical precipitation in which process alkaline compounds are used to cause metals such as chromium, copper and zinc to precipitate. Solids, as well as the metal ions precipitated as a result of the previous process, are eliminated by sedimentation. Additionally, pH is adjusted through the use of sodium hydroxide or lime plus sodium hydroxide. Finally, chromium reduction is realized by employing reducing agents which reduce hexavalent chromium to its trivalent form. Then chemical precipitation is employed to eliminate the resulting trivalent chromium. Using this model technology, the Agency established standards for the best practicable control technology currently available (BPT), the best available technology economically achievable (BAT), and established new source performance standards (NSPS) as well as pretreatment standards for both existing sources (PSES) and new sources (PSNS).

In setting BPT limitations, the Agency employed the model treatment, including flow reduction to reduce the flow of water through the canwasher. Specific effluent values were established for chromium, zinc, aluminum, phosphorus, fluoride, oil and grease, total suspended solids, and pH.

In setting BAT limitations, the Agency employed the model treatment, but included further flow reduction. Two other options proposed in the notice of proposed rulemaking were rejected.13 Effluent limitations were selected for pollutants including chromium, zinc, aluminum, fluoride, and phosphorus. Canmakers were required to limit the discharge of these pollutants to specified quantities expressed in terms of maximum monthly and daily discharges. TTO was not regulated under BAT because the Agency felt that it would be removed by the oil and grease removal systems mandated by BPT. Copper, lead, nickel, and manganese were not regulated because the Agency believed that these metals would be removed by the model technology when it was operated with sufficient efficiency to remove the pollution parameters chosen.

Effluent limitations for new sources were also instituted on the basis of application of the model technology. However, flow reduction was further increased. Effluent limitations were established for oil and grease, total suspended solids, chromium, zinc, aluminum, fluoride, phosphorus, and pH. The oil and grease limitation was used in order to control TTO, the Agency explained, because of these pollutants' high solubility in oil, i.e. removal of oil and grease would also remove acceptable amounts of TTO present in the wastestreams. Nickel, copper, and lead were not regulated under NSPS because the Agency believed that these pollutants would be reduced incidentally by the model treatment technology.

Although the regulation controls the discharge of a number of pollutants as indicated earlier, it is the regulation of total toxic organics and toxic metals that form the principal issues on this appeal. TTO is specifically controlled only by pretreatment standards--that is, under PSES and PSNS. As has been indicated, TTO is not specifically controlled for direct dischargers (BPT, BAT or NSPS) but only under PSES and PSNS for indirect dischargers. Many of petitioners' objections to this regulation then relate not to the data collected for BPT, BAT and NSPS technologies but only to the TTO data collected for control of indirect dischargers under PSES and PSNS.

The model technology selected by the Agency for setting PSES standards is the same as for BAT, while that selected for PSNS is identical to that for NSPS. The Agency also explained that "pass-through" existed with regards to TTO. It reasoned that while a POTW would remove 70% of these pollutants from untreated wastewaters, treating the wastewater by oil and grease removal, as demonstrated in the aluminum forming category, would achieve a 97% reduction. Pass-through having been established, the Agency promulgated effluent limitation standards for TTO. However, because the Agency recognized that monitoring for TTO was a costly and time consuming process, oil and grease standards were established as an alternative monitoring parameter, i.e. a canmaking facility could meet the effluent limitation for TTO by satisfying the standards set for the removal of oil and grease.

Although petitioners allege error in the Agency's regulation of toxic metals for direct dischargers (considered infra ), their disagreement with pretreatment standards or regulation of indirect dischargers of toxic metals under PSES and PSNS relate to the Agency's findings that toxic metals "pass-through" a POTW.

With respect to chromium, zinc, copper, and manganese, the Agency reasoned that a well-operated POTW would remove 50%-60% of these pollutants while the model technology would remove 92%. Accordingly, "pass-through" was demonstrated and pretreatment standards were established for these pollutants. No standard was established for aluminum because alum, an aluminum sulfate, is often added to wastewater at a POTW. Manganese and copper were chosen because these substances are employed as alloying constituents along with aluminum in canmaking and it was believed that removal of manganese and copper would result in acceptable removal of aluminum. The treatment effectiveness for copper and manganese was drawn from the CMDB.

Again, the errors argued by petitioners on this appeal are: that the effluent limitations for total toxic organics were so marred by erroneous data collection and selection that we must view the Agency's actions as arbitrary and capricious and its conclusions as resulting from unreasoned judgments; that the Agency erred in not subcategorizing the canmaking industry into point sources that use chromium as a can coating and those that use other coating material; that it erroneously applied the "pass-through" criteria in formulating the PSES and PSNS limitations on chromium, copper and zinc in the wastewater of indirect dischargers; and that it failed to exercise its statutory duty to consider the costs imposed by the regulation. With respect to petitioners' challenge regarding TTO limitations, we note that no argument has been advanced that the oil and grease limitations established by the Agency as an alternative monitoring parameter cannot be met.

The scope of our review of the Agency's action in this case is governed by Sec. 10(e)(2)(A) of the Administrative Procedure Act, 5 U.S.C. Sec. 706(2)(A). That standard provides that we may set aside the Agency's action only if it is found to be "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law." Id. Under this standard, we presume the validity of Agency action, Ethyl Corp. v. EPA, 541 F.2d 1, 34 (D.C.Cir.) (en banc), cert. denied, 426 U.S. 941, 96 S. Ct. 2662, 49 L. Ed. 2d 394 (1976), and our function is to scrutinize the Agency's activity to discern whether the record reveals that a rational basis exists for the Agency's decision. Id.; Bowman Transportation Inc. v. Arkansas-Best Freight System, Inc., 419 U.S. 281, 286, 95 S. Ct. 438, 442, 42 L. Ed. 2d 447 (1974).

The scope of our review is further colored by the policy of the Clean Water Act and the sophisticated data evaluations mandated by that lengthy and complicated statute. The Act expresses a congressional insistence to eliminate water pollution within a short time-span through the use of uniform effluent limitations imposed on an industry-wide basis. This need for quick action and cross-industry application demands that we exercise our review of these regulations with considerable circumspection. Consolidation Coal Co. v. Costle, 604 F.2d 239, 243 (4th Cir.1979); Weyerhaeuser Co. v. Costle, 590 F.2d 1011, 1025 (D.C.Cir.1978). Further, technological and scientific issues, such as those presented in this case, are by their very nature difficult to resolve by traditional principles of judicial decisionmaking. For this reason, "[w]e must look at the decision not as the chemist, biologist or statistician that we are qualified neither by training nor experience to be, but as a reviewing court exercising our narrowly defined duty of holding agencies to certain minimal standards of rationality." Ethyl Corp., 541 F.2d at 36. More specifically, we note that an agency's data selection and choice of statistical methods are entitled to great deference, FMC Corp. v. Train, 539 F.2d 973, 986 (4th Cir.1976); American Meat Institute v. EPA, 526 F.2d 442, 457 (7th Cir.1975), and its conclusions with respect to data and analysis need only fall within a "zone of reasonableness." Hercules, Inc. v. EPA, 598 F.2d 91, 107 (D.C.Cir.1978). This standard, however, does not compel us to abdicate our judicial function, and we are mindful that the Agency "must fully explicate its course of inquiry, its analysis, and its reasoning." Tanner's Council of America, Inc. v. Train, 540 F.2d 1188, 1191 (4th Cir.1976).

With these principles in mind, we review the Agency's action during rulemaking to determine if it abused its discretion in promulgating the regulation.

We consider first petitioners' several arguments relating to their contention that the Agency committed reversible error in setting pretreatment standards limiting effluence of total toxic organics. The Agency, in setting pretreatment standards, reasoned that oil and grease removal would result in a 97% removal of TTO. It reached that conclusion based on the testing of wastewater in rulemaking for the aluminum forming industry. Although the aluminum forming wastewater contained a much higher concentration of TTO (25.7 mg/L) than did canmaking wastewater, the Agency concluded that the percentage of TTO removable by the model technology (97%) would be similar. In establishing canmaking standards, the Agency relied on data indicating that wastewater from canmaking facilities contained an average of 2.727 milligrams of TTO in each liter of wastewater (2.727 mg/L). Deducting 97% of that, the Agency arrived at a mean expected TTO concentration of 0.08 mg/L and, factoring in variables in the model technology, the Agency established .32 mg/L as an allowable one-day maximum TTO concentration.14

The oil and grease removal technology recommended for the canmaking industry, therefore, was conceived in the regulation of another source category--the aluminum forming industry. The Agency's regulation of the aluminum forming industry is, of course, not a subject of this appeal, but the Agency transferred15 oil and grease removal technology from that category to the canmaking category. The petitioners initially contend that the regulation was not valid even as it applied to the aluminum forming industry due to the flawed method by which the Agency collected data.

(a)

The petitioners' primary complaints concerning the transfer of aluminum forming removal efficiency challenge the way in which the Agency sampled and tested the aluminum forming wastewater. They state that the Agency compared one day's influent concentration to another day's effluent concentration, or compared one day's influent or effluent concentration to the average concentration of several days effluent or influent. The petitioners urge that this error in testing was compounded by errors of taking wastewater samples from improper locations in the water flow systems. The Agency tested three plants in five days, say the petitioners, and the efficiency was unknown for one day and varied between 97% and 99% for the other four days. If the samples had been accurately taken, they contend that the TTO removal efficiency would have varied between 76% and 99%.

The Agency responds that use of even the comparisons and sampling points suggested by the petitioners reveals a constant removal efficiency approximating 97%. Additionally, the Agency points out that most of these objections were not raised in the rulemaking procedure. The parties tabulate and chart much of the same raw data and sometimes utilize the same tables and charts, but arrive at their different conclusions concerning the removal efficiency percentage. As frequently has been written, we do not sit as a scientific body minutely comparing competing research methods and results. See BASF Wyandotte Corp. v. Costle, 598 F.2d 637, 649-50 (1st Cir.1979). As a reviewing court, we have delved deeply enough into this essentially scientific disagreement to understand it for our purposes--judicial review of the administrative agency's actions under the standards of review we have previously discussed. The Agency here chose five representative canmaking plants where it gathered water samples. It had previously, as explained in its rulemaking, designed procedures and protocols for sampling and analysis to protect their scientific integrity. The petitioners argue with the choice of testing sites within the water systems and with the Agency's methods of comparing samples. The Agency explained its methods during rulemaking and insisted there and here on appeal that their methods were scientifically correct. We do not judge that; we view their actions under a judicial glass and readily discern that they have acted reasonably, given the regulated industry's representatives more than adequate opportunity to comment, considered the comments, and explained their rejection. This is indeed reasoned administrative decisionmaking, and we have no occasion to interfere in that process.

(b)

The petitioners next contend that, even if the 97% removal efficiency was correctly calculated in aluminum forming, it is not a valid assumption when applied in canmaking. First, they contend that removal of oil and grease removes a greater percentage of TTO from wastewater highly concentrated with TTO than wastewater with lower concentrations.

The petitioners argue that at concentrations of less than 2.12 mg/L, removal of oil and grease only removes 76% of the TTO. Demonstrating by graph a 97% removal rate for TTO at a concentration of 25.7 mg/L (as found in the aluminum forming category) and a 76% removal rate at concentration of 2.12 mg/L, they argue that the removal efficiency percentage for a 2.727 mg/L concentration, as found in the canmaking industry, is only 80%. Petitioners reach this conclusion by charting the removal rates that they contend would have been demonstrated had the sampling been conducted properly, i.e. removal efficiencies fluctuating between 76% and 99.0% depending on the amount of TTO present.

Additionally, petitioners cite the report of Murray P. Strier, that they contend was used along with others by the Agency, and which demonstrates that only trace amounts of eight of the fourteen regulated TTO pollutants found in canmaking wastewater could be removed by the model technology. Finally, in this part of their argument, petitioners list the tested achievable treatment levels for the fourteen toxic pollutants (ranging from 0 mg/L to .10 mg/L), total them, and arrive at an achievable level of TTO in the wastewater after treatment by the model technology of 0.413 mg/L.16 They, therefore, assert that the TTO discharge level permitted by the regulation (.32 mg/L) is significantly lower than what is actually achievable.

The Agency characterizes the petitioners' reasoning as seriously flawed. It is not the concentration of organics that determines the percentage that can be removed, argues the Agency; removal efficiency depends upon the octanol/water partition coefficient and the concentration of oil.17 If the organics have a high partition coefficient and there is sufficient oil in the wastewater, virtually all of the organics will be absorbed by the oil and removed by effective oil removal technology. Since all fourteen organics found in canmaking wastewaters are highly soluble in oil and there are high levels of oil present,18 scientific analysis reinforces the conclusion based on the sampling data collected from aluminum forming wastewater.

Moreover, the Agency disputes the petitioners' concentration estimates. The Agency points out that the data chosen by the petitioners to calculate the amount of toxic organics in canmaking wastewater represents the condition of the wastewater before flow reduction mandated by the model technology is applied. That part of the required technology, unchallenged by the petitioners, reduces the amount of water and obviously increases the concentration of TTO in the water. An Agency table shows that TTO concentration will increase several times after application of flow reduction required for meeting both PSNS and PSES. The Agency dismisses the "Strier report" as based on a different technology than that developed by the Agency and avers that although it possessed the Strier material, it was not used in their determinations. The Agency thoroughly considered removal efficiency in the canmaking context, and we find no abuse of its discretion.

(c)

The petitioners next contend that the Agency erred in arriving at its mean concentration value of 2.727 mg/L in canmaking wastewater because it did not include in its average the data from sampling points where it was indicated that a particular pollutant was not present. The Agency replies that it purported to show an average of only the pollutants that were present and subject to removal and that it would have been senseless to devise a standard which included toxic organics that did not have to be removed because they were not present in the first place. We agree with the Agency that it acted well within its assigned role in selecting the method to tabulate the data needed to reflect the pollutant composition of the wastewater under examination.

The petitioners, both in complaining about the efficacy of TTO sampling and in attacking the Agency's costs considerations (considered infra), advance another asserted error in sampling. They contend that in addition to the samples taken in 1977 and 1978, the 1983 "grab" samples should have been added to the equations and, if included, would have produced for them a favorable result. That argument overlooks the explanation offered by the Agency in the preamble to the regulation and reiterated on appeal--that the 1983 sampling was not designed for scientifically accurate computation but was designed to obtain approximate values and to respond to Agency conclusions called into question by the CMI and USBA sampling conducted after promulgation of the proposed regulation. This has been explained and reiterated administratively, yet the petitioners insist that we weigh judicially these Agency actions, which are without the sphere of judicial review. We decline.

(d)

The petitioners urge that the technology designed in aluminum forming is not transferable to canmaking. Here, they repeat many of their arguments that the Agency's calculation of the percentage efficiency of pollutant removal was faulty. They argue that the technology developed in the aluminum forming category is, for the same reasons, not legally transferable to the canmaking subcategory.

In Tanner's Council of America, Inc. v. Train, 540 F.2d 1188 (4th Cir.1976), we considered the propriety of transferring the results of pollution technology from one industry to another as the basis for Clean Water Act standards. We stated that "[t]his transfer of technology is permissible only 'if he (the Administrator) determines the technology to achieve those higher levels can be practicably applied.' " Id. at 1192, quoting S.Rep. No. 414, 92d Cong., 1st Sess. (1971), A Legislative History of the Water Pollution Control Act Amendments of 1972, 93d Cong., 1st Sess. 1468. We quoted with approval the criteria developed by the Eighth Circuit19 to determine if a technology can be practicably applied: the Agency must "(1) show that the transfer technology is available outside the industry; (2) determine that the technology is transferable to the industry; (3) make a reasonable prediction that the technology if used in the industry will be capable of removing the increment required by the effluent standards." Tanner's Council, 540 F.2d at 1192 (footnote omitted).

The Agency demonstrated in the aluminum forming category that the removal of oil and grease by the technology would result in acceptable reduction of TTO. The Agency also explained that in light of the similar amounts of oil and grease present in canmaking and aluminum forming wastestreams, the removal technology could be applied in both industries. Finally, as we have explained, the Agency offered a reasoned justification for the transfer, based on data demonstrating that the TTO found in canmaking effluent was highly soluble in oil and thus the technology could be expected to remove an adequate amount of these pollutants. We note further that this issue was fully aired during the rulemaking procedure and that the Agency has consistently held to its position and exhaustively explicated its reasons for the transfer of technology. We find no abuse of the Agency's discretion in this regard and hold that the transfer of technology was amply justified.

(e)

The petitioners also assert that the Agency abused its discretion in even regulating TTO under PSES and PSNS. The petitioners point to the Agency's decision not to regulate the coil coating category under PSES and PSNS because the TTO concentration in wastewater from that category was only approximately 1.47 mg/L and argue that, according to their calculations, TTO concentration in canmaking wastewater is only 1.145 mg/L.

The statutory criteria for determining to impose pretreatment standards is whether pollutants generated by a facility would interfere with, pass-through, or otherwise be incompatible with the POTW. See 33 U.S.C. Sec. 1317(b). The petitioners, again relying on their calculations rather than the Agency's, insist that direct dischargers, by removing oil and grease from canmaking wastewater, can remove only .413 mg/L. Using the Agency's assumption that a well-run POTW can remove 70% of TTO and the petitioners' TTO concentration calculations of 1.145 mg/L, they attempt to demonstrate that a POTW receiving wastewater without pretreatment by removal of oil and grease would eliminate all but .344 mg/L TTO (i.e. 30% x 1.145). Since the .344 residual TTO after POTW treatment is less than .413 mg/L, which petitioners argue is the maximum level of TTO achievable by a direct discharge through oil and grease removal, the pass-through criteria, they argue, is not met because the POTW can remove a greater percentage of TTO than a direct discharger.

Assuming the correctness of the petitioners' calculations, we would be greatly concerned with this argument. However, the 1.145 mg/L TTO concentration in canmaking is at odds with the Agency's tabulated 2.727 mg/L concentration. We would be hard put to accept that the 1.145 concentration had been proved by a preponderance of the evidence--yet, of course, that is not the test. Additionally, the .413 mg/L that the petitioners propound as the achievable level of TTO removal is considerably greater than the .32 mg/L limit imposed by the regulation. We are convinced that the Agency properly exercised its administrative role in reaching the conclusion that there is a 2.727 mg/L concentration of TTO in canmaking wastewater and that pass-through has been demonstrated--we need go no further.

Although not determinative, we note again that a number of the objections petitioners now level at Agency data were either not raised or not fully explained to the Agency during rulemaking. To raise such material for the first time on appeal is unfortunate from both an administrative and appellate standpoint. The Agency in this case has complied strictly with the notice and comment procedures required by the Administrative Procedure Act, 5 U.S.C. Sec. 553, and petitioners do not attack the regulation on the ground of procedural irregularity or infirmity. The Agency has not cloaked its consideration in secrecy--adequate notice was given in the proposed regulation and the Agency has exhibited an admirable willingness to consider matters brought up by comments submitted by petitioners and others in the industry. An enormous amount of explanatory and technical data has been generated, including development documents, comments and responses, economic analyses, and scientific data. Despite this adequate opportunity to comment and the clear explanation of the Agency's intent, many of the arguments relating to the Agency's conclusions regarding the removal of TTO in the aluminum forming category and the transfer of technology to the canmaking subcategory were not presented to the Agency during the rulemaking procedure. Under such circumstances "the notice-comment-and-response procedures will have been deprived of much of their validity, and the party responsible therefor will accordingly be given less latitude in complaining about the results." Weyerhaeuser, 590 F.2d at 1028 n. 15 (emphasis in original); see also National Association of Metal Finishers v. EPA, 719 F.2d 624, 638 (3d Cir.1983).

The petitioners' attack on the Agency's regulation of the discharge of chromium, zinc, and copper by indirect dischargers in one respect differs from their attack on the regulation of toxic organics and in another respect parallels their objection to toxic organic regulation.

(a)

The first objection is that the presence of chromium is overstated because the Agency abused its discretion by considering canmaking as a single category of a source of water pollution rather than creating a subcategory for plants that use a chromium-based manufacturing process. As was pointed out in our previous discussion, there are only a few canmaking plants that now use chromium as a coating. This is significant because not only do the plants using the chromium coating process discharge wastewaters with a higher percentage of chromium, but hexavalent chromium is more prevalent than trivalent chromium. Hexavalent chromium is many times more toxic than trivalent chromium. The Agency recognizes that levels of chromium pollutants from plants using the chromium process20 are much higher than levels from those using nonchromium processes, but it insists that some chromium is present in the wastewaters from all plants and that, regardless, the chromium effluent limitation which it set can be readily achieved by all plants.

The Act requires the Agency to establish effluent limitations "for categories and classes of point sources." 33 U.S.C. Sec. 1311(b)(2)(A). It must also "designate the category or categories of sources" to which pretreatment standards apply. 33 U.S.C. Sec. 1317(b)(3). Here, it is the fixing of a single pretreatment standard that precipitates the petitioners' complaint. This is another area of judicial review, however, where we will not reverse the Agency's determination unless it abused its discretion, and the Agency need not account for all possible differences among plants. American Iron and Steel Institute v. EPA, 568 F.2d 284, 297-99 (3rd Cir.1977).

In the development document, the Agency discussed thirteen factors it considered in deciding whether to subcategorize further the canmaking subcategory. One of these factors was the manufacturing process employed. In discussing this factor, the Agency made no mention of differing surface treatments; petitioners contend that these differing surface treatments constitute a difference in "manufacturing processes" and that the Agency abused its discretion in failing to subcategorize on this basis. Even if this were error, we do not feel that it is of sufficient magnitude to require reversal of the Agency's decision.

In the first place, the Agency on appeal stresses that while most of the plants now use nonchromium coating processes, they are constructed so that they can use either chromium- or nonchromium-based treatments. Consequently, although chromium surface treatments may be out of favor at this time, the manufacturing process itself remains capable of using chromium in the future. The regulation of a pollutant now in use in some plants and capable of being employed in others does not appear to us to be unreasonable. Secondly, the Agency points out that there are chromium pollutants, in some quantity, discharged from all plants.21 The Agency's task was to establish numerical standards limiting effluent pollution and it concentrated on grouping plants that could meet the same limitations. That this is a legitimate consideration, there can be no doubt. See Vol. 1, Legislative History of the Federal Water Pollution Control Act Amendments of 1972, at 172. The Agency urges that even if the canmaking industry was further subcategorized, the effluent standards would probably be the same. The petitioners have not shown that it would be otherwise, but even if they could we do not think the Agency has abused its discretion in creating a single canmaking category. Before making that decision, it considered all relevant factors and provided reasoned explanations for its actions for which there was a substantial basis in the record.22

(b)

The petitioners' second objection to the regulation of toxic metals is that chromium, zinc, and copper do not meet the pass-through criteria for regulation under PSES. In other words, they contend that a well operated POTW would remove more chromium, zinc, and copper from wastewater discharged into it without pretreatment than would be removed by direct dischargers employing the model technology.

As in its attack on the TTO standards, however, the petitioners use different data for their demonstration.23 Importantly, they overlook the water flow reduction which is part of the model technology--with the water flow reduced the concentrations of chromium, zinc, and copper are much greater and pass-through criteria easily met. Apart from that, petitioners show at most a disagreement with the Agency without a showing that the Agency was guilty of serious technological errors in testing, calculating, and applying the results of the tests so as to achieve their basic goal--a uniform achievable standard which would prevent an optimum amount of toxic metals from reaching the nation's waters.

The petitioners' final major assault on the regulation attacks the Agency's analysis and "consideration" of the cost effectiveness of treatment options. The Act requires the Agency, in identifying BPT, to consider "the total cost of application of technology in relation to the effluent reduction benefits to be achieved from such application." 33 U.S.C. Sec. 1314(b)(1)(B). In identifying BAT technology and promulgating NSPS, PSES and PSNS standards, however, the Agency is not required to consider "effluent reduction benefits," but must "take into account" direct and indirect costs. 33 U.S.C. Sec. 1314(b)(2)(B). The petitioners contend the Agency failed to fulfill these statutory duties.

For BPT there must be a "limited balancing" of costs against benefits, but as regards BAT, NSPS, PSES and PSNS no balancing is required--only that costs be considered along with the other factors discussed previously. 33 U.S.C. Secs. 1314(b)(1)(B), (b)(2)(B). National Ass'n Metal Finishers v. EPA, 719 F.2d 624, 662-663 (3rd Cir.1983); Weyerhaeuser v. Costle, 590 F.2d 1011 at 1046.

The petitioners concede that the Agency "considered" costs but contend that its analysis was so faulty that promulgating the regulation in face of what the actual cost and actual cost/benefits results should have been amounts to an abuse of discretion. The Agency calculated the cost of BPT for direct dischargers at $50/lb. The petitioners contend this cost should be $17,710/lb, or a cost 350 times that calculated by the Agency. Similarly, it is contended that the Agency grossly understated the cost of pretreatment. The petitioners argue that these alleged gross inaccuracies resulted from three principal factors--the Agency's failure to (1) establish a separate subcategory for plants using the chromium-based manufacturing process; (2) include its 1983 wastewater sampling data in its calculations; and (3) correctly differentiate between the amounts of high toxic chromium (hexavalent chromium) and low toxic chromium (trivalent) which were present in wastewater. They also contend that the Agency erred in calculating costs at incremental levels of the technology rather than the overall benefit for each treatment level.

The first two objections to the cost analysis consideration repeat the arguments we have rejected in part IV. There is no reason to reconsider them. An agency has a broad discretion in its selection of data and in the method of calculation, particularly when it involves highly scientific or technical considerations. Hercules, Inc. v. EPA, 598 F.2d 91, 108 (D.C.Cir.1978); American Petroleum Institute v. EPA, 540 F.2d 1023, 1035-36 (10th Cir.1976).

Similarly, we think that the Agency's action in considering costs at incremental levels to be properly within its discretion. The Agency explained in its Cost Effectiveness Analysis that cost-effectiveness was defined as "the incremental annualized cost of a pollution control option in an industry or industry subcategory per incremental pound equivalent of pollution removed by that control option." We find no abuse of discretion of its decision to analyze costs on this basis and hold that this was a reasonable effort by the Agency which must be upheld. FMC Corp. v. Train, 539 F.2d 973, 979 (4th Cir.1976).

The Agency's estimate of the hexavalent/trivalent mix in canmaking wastewater, however, gives us some pause. We are of the opinion that the Agency's reasoning in this one particular was far from faultless, but we are reluctant to remand the regulation because of this one error when a corrected result would not affect the regulation. Determining that it would not, we decline to reverse on this ground.

The "cost effectiveness" of a technology is defined as an annualized capital cost of the technology per "pound equivalent" of pollutant removed by such technology. "Pound equivalent" is a term used to express the varying degrees of toxicity of different pollutants, wherein toxicity is standardized by reference to the toxicity of copper. The "pound equivalent" of a particular pollutant is the number of pounds of copper that are equivalent in toxicity to one pound of a given pollutant. The toxic weight of hexavalent chromium is 19.3 and that for trivalent chromium is 0.127.

The true mix of hexavalent to trivalent chromium was a contested issue during the rulemaking. Commenters contended that of the chromium present in wastewater, virtually all of it was in trivalent form, although the Agency argues that no commenter submitted any data to support that claim. The Agency responded that "[b]ased on the data available ... chromium is present in the wastewaters of almost all canmaking plants, .... [I]n the absence of specific steps to reduce chromium, chromium in canmaking wastewaters can be expected to appear in hexavalent form." Nonetheless, the Agency attempted to compensate for its failure to distinguish between hexavalent and trivalent chromium in the final Cost Effectiveness Analysis which was issued contemporaneously with the final regulation:

Two key estimates were made with regard to chromium pollutant loadings. Since these values are reported for total chromium only, the precise mix between hexavalent and trivalent chromium (which have toxic weights of 19.3 and .127, respectively) is not known. To calculate CE values, it was estimated that the chromium mix is 50% hexavalent and 50% trivalent before treatment, and 24% hexavalent and 76% trivalent after lime and settle treatment.24

We do not agree with the Agency's argument that the petitioners had a primary duty to demonstrate the percentage of hexavalent chromium present in its wastewaters. We think, however, that the record demonstrates that the Agency satisfied the statutory requirement that it consider costs.

As we stated in FMC Corp. v. Train, 539 F.2d 973 (4th Cir.1976):

The Act's overriding objective of eliminating ... the discharge of pollution into the waters of our Nation indicates that Congress, in its legislative wisdom, has determined that the many intangible benefits of clean water justify vesting the Administrator with broad discretion, just short of being arbitrary or capricious, in his consideration of the cost of pollution abatement.

Id. at 978-79. See also American Iron and Steel Institute v. EPA, 526 F.2d 1027, 1031 (3rd Cir.1975) (cost of compliance "not a factor to be given primary importance").

In promulgating this regulation, the Agency, through a subcontractor,25 conducted exhaustive economic impact analyses and cost effectiveness analyses. The cost effectiveness analyses26 examined the cost effectiveness of regulatory alternatives with respect to indirect and direct dischargers as well as each type of controlled pollutant. Additionally, the subcontractor analyzed the impact of the regulation on such diverse aspects as plant-level profitability, capital requirements, plant closures, new plant construction, small businesses, and plant characteristics. Further the development documents contain estimated compliance costs which were the subject of numerous comments and responses.

Although we do not condone the Agency's treatment of the issue concerning the hexavalent/trivalent chromium mix, the record indicates that it carefully considered all other cost factors and, in this one particular, made an estimate of the differing quantities of hexavalent and trivalent chromium which has support in the administrative record. Importantly, it also concluded that even if its estimates were completely erroneous, it would not have affected the regulation. In sum, we believe that the record demonstrates that the Agency made a reasonable effort in analyzing costs and on that basis the regulation must be upheld. See FMC Corp. v. Train, 539 F.2d 973, 979 (4th Cir.1976).

For all of the foregoing reasons, the effluent limitations for the canmaking industry are upheld, and the petitions denied.

DENIED.