Chemical Manufacturers Association v. U.S. Environmental Protection Agency, Nos. 87-4849

ON PETITIONS FOR REHEARING AND SUGGESTION FOR REHEARING EN BANC Before, RUBIN, GARZA, and KING, Circuit Judges. ALVIN B. RUBIN and KING, Circuit Judges: At the conclusion of our original opinion 1 we encouraged the parties to file petitions for rehearing in order to ensure that we had given each issue in this complex case its due. 2 After considering the parties’ arguments, we now clarify certain aspects of our earlier opinion and grant rehearing and remand two parts of the regulations to the Environmental Protection Agency (EPA) for further rulemaking proceedings. As in our original opinion, we have divided responsibility for drafting this opinion: Judge King prepared the sections concerning the regulations based on the best practicable control technology (BPT), and Judge Rubin prepared the sections concerning regulations based on the best available technology (BAT). No issues have been raised on rehearing with respect to the parts of the original opinion authored by Judge Garza. I. Best Practicable Technology (BPT) Issues A. Fundamentally-Different-Factors (FDF) Variance Issues Several of the petitioners have requested that we clarify that our opinion does not preclude them from seeking an FDF variance. Although we had attempted to make clear that nothing in our opinion should be deemed prejudicial to the petitioners’ ability to seek FDF variances, we will address the specific points raised by the individual petitioners in their requests for rehearing. 1. DUPONT DuPont objects to the language in our opinion which states that “[a]n FDF variance would not exempt petitioners from the limitations; it would merely subject them to less stringent limits using the best technology.” 3 DuPont asserts that this is not correct — an FDF variance might allow an applicant to leave its present technology in place or to install technology less costly than the model technology. Because this language was merely dicta, we direct that the final paragraph of part II. F.5. 4 in our original opinion be disregarded. The fact remains, however, that this court may not grant the relief petitioner requests. Under Sections 509(b)(1) 5 and 505(a)(2) 6 of the CWA, we do not have jurisdiction in this proceeding to require the EPA to process petitioners’ applications in a more expeditious manner, 7 nor may we stay the regulations pending the EPA’s actions. As we noted in our original opinion, the Act provides explicitly that an application for an FDF variance “ ‘shall not stay the applicant’s obligation to comply with the ef *257 fluent limitation guideline or categorical pretreatment standard which is the subject of the application.’” 8 Therefore, we adhere to our refusal to order the EPA to process FDF-variance applications. 2. UNION CARBIDE Union Carbide asserts that we may have prejudiced its ability to obtain an FDF variance for its Taft, Louisiana, plant. The discussion of the Taft plant issue in our original opinion may not be sufficiently clear because a number of petitioners had raised plant-specific claims as objections to the EPA’s categorical rulemaking and that was one of a number of such claims. At any rate, we affirm that such objections must be raised in an FDF variance proceeding, and not in a proceeding of review of the national rules. 9 Our discussion of the Taft plant issue was intended only to hold that the EPA did not err in refusing to address this plant-specific claim in the context of the national rulemaking. Nothing in our discussion should be deemed prejudicial to Union Carbide’s efforts to obtain an FDF variance. 3. PONDS PETITIONERS Finally, the “ponds petitioners” assert that our original opinion 10 mischarac-terized their argument — petitioners did not seek an exemption from the BPT limitations, but sought to have the regulations declared unlawful. Our original opinion made it clear that the objections raised by the ponds petitioners regarding the EPA’s consideration of the cost of compliance would not suffice to invalidate the regulations as a whole. 11 We also held that petitioners had failed to establish that the EPA erred in refusing to create a subcategory for plants employing waste stabilization ponds. 12 In so holding, we stated that “even if Texas Eastman, DuPont, and Air Products are required to install new activated-sludge systems in order to comply with BPT, the costs of these systems would be within the range generally estimated for the industry as a whole,” and concluded that this did not provide a basis for “exempting” petitioners from the limits that apply to the rest of the industry. 13 The word “exemption” was intended to refer to petitioners’ claim that they should have been subject, as a separate subcategory, to less stringent guidelines because the cost of compliance would be higher for petitioners than for the industry as a whole. Petitioners now assert that they intended only to question the overall validity of EPA’s costing methodology. They observe that there is nothing in the record to support the conclusion that the actual costs to petitioners of installing entirely new treatment technology would not be “fundamentally different” from costs to other plants in the industry, and argue that the language in the original opinion is misleading on this point and may be prejudicial to petitioners’ efforts to obtain FDF variances. Because we may have misconstrued petitioners’ argument on the cost issue, we will alter the final paragraph of part II.E.3. 14 to read as follows: The EPA concedes that petitioners may be required to install entirely new treatment units consisting of activated sludge and secondary clarification. However, the EPA estimated the costs of such steps for more than a quarter of the plants in the industry which would require upgrades to comply with BPT. 1 We therefore agree with the EPA that the EPA’s failure to consider the actual cost of installing entirely new treatment systems at three of petitioners’ plants does not distort the *258 EPA’s overall cost assessments sufficiently to undermine the validity of the BPT limits as a whole. We adhere to the remainder of our earlier discussion. We express no opinion whether costs, or any other consideration raised by petitioners, would provide a basis for an FDF variance. B. Phenol PSES Notice & Comment Issue Our original opinion failed to address one issue raised by petitioners Union Carbide, Allied, and others: whether EPA violated the Administrative Procedure Act by failing to afford adequate notice and an opportunity to comment on the inclusion of Phenol in PSES. 15 Petitioners now rely in part on our resolution of the BAT-subcategory notice and comment issue in favor of NRDC. 16 We found that the EPA had not afforded the parties sufficient opportunity for comment on the Agency’s final rule establishing subcategories for BAT because the EPA had indicated in its notice that it would not establish any subcategories for BAT. 17 Our conclusion was based on the following language from the record: EPA considered whether the industry should be subcategorized for BAT purposes by evaluating the same subcate-gorization factors which were considered for BPT. EPA has decided to promulgate a single set of BAT limitations which would be applicable to all OCPSF facilities.... The available data for BAT show that plants ... can achieve similar low toxic pollutant effluent concentrations by installing the best available treatment components_ Therefore, the Agency believes that BAT subcategories do not appear to be necessary for effective, equitable regulation. However, EPA will continue to explore the possibility of subcategorizing the industry for BAT purposes and invites comments and supporting date on appropriate approaches. 18 We held that “[t]he last sentence notwithstanding, this notice was not sufficient to fairly apprise interested parties that BAT subcategorization was still a live issue.” 19 Petitioners contend that until the 1986 proposal — the last proposal document published before the final rule — the EPA had never mentioned phenol in the PSES rule-making. Petitioners claim, moreover, that the 1986 proposal indicated that the EPA did not intend to include phenol in the PSES limits. Petitioners cite the following language we found inadequate with respect to the BAT-subcategory issue: When a pollutant is reduced to similar effluent levels and the pass-through analysis compares average industry to average POTW percent removals, ... the calculation of percent removal for the POTW and industrial treatment systems becomes a function of influent concentration only for these pollutants rather than a function of actual removal efficien-cy_ [T]he Agency is considering comparing percent removals only for comparable ranges of influent concentrations. Using this approach ... phenol ... would not be regulated. 20 If this were in fact an accurate representation of the record, we might find petitioners’ claim to have merit. The record, however, belies petitioners’ argument. First, the EPA’s 1986 proposal indicated that the Agency was considering five dif *259 ferent options for PSES — one of which, PSES V(a), would exclude phenol, anthra-cene, and phenanthrene. 21 Second, in language which petitioners have artfully el-lipsed from the quotation above, the EPA stated that it was “considering an alternate modification of the pass-through analysis for certain non-volatile organic priority pollutants such as phenol.” 22 The EPA then explained how this “alternate” approach would differ from the straight BAT-comparison otherwise employed to determine pass through. The EPA also qualified its statement: “if sufficient data exist, the Agency is considering comparing percent removals only for comparable ranges of influent concentrations. Using this approach, three non-volatile organic pollutants — phenol, phenanthrene, and anthra- cene — which previously required PSES/PSNS standards would not be regulated.” 23 Far from indicating that regulation of phenol was no longer a “live issue,” the record demonstrates that excluding phenol from PSES was one option under consideration and would depend on the outcome of further study. The quoted language makes clear that phenol would otherwise be regulated. This conclusion is consistent with the supporting documents for the EPA’s 1985 proposal, which reported the results of the EPA’s pass-through analysis. Phenol was included in the table and BAT percent removal for phenol was found to be .8% greater than the POTW percent removal. Under the five or ten percent differential approach that the EPA had considered earlier, phenol would not be regulated. The EPA made clear in its 1986 proposal, however, that it was abandoning the percent differential approach for determining pass through in favor of a straight BAT-comparison approach. This notice should have been sufficient to apprise interested parties that PSES would regulate any pollutant that the EPA had found to be removed less effectively by POTWs than by direct dis-chargers complying with BAT. Because the documents supporting the EPA’s earlier proposal supplied the information necessary to conclude that phenol would be regulated under the BAT-comparison approach, and because the 1986 proposal itself made clear that phenol would be exempted from regulation only if the EPA chose an “alternate modification of the pass-through analysis,” we conclude that petitioners were fairly apprised that phenol might be subject to PSES. C. Publicly-Owned Treatment Works (POTW) Issue Petitioners Gulf Coast Waste Disposal Authority (“Gulf Coast”) and its industrial users request that we alter the portion of our original opinion holding that the EPA properly refused to create a separate subcategory for PSES for users of allegedly exemplary POTWs. 24 Petitioners contend that we erroneously equated sub-categorization with the award of removal credits. They assert that removal credits occupy a role distinct from subcategorization. Petitioners argue that PSES depends upon a determination that a given pollutant passes through POTWs without adequate treatment and that subcategorization must therefore be available for the users of POTWs when pass through of certain regulated pollutants does not occur. In contrast, petitioners argue, removal credits may be granted even when POTW treatment is less effective than BAT — that is, even when pass through occurs. Petitioners conclude that we should therefore construe the statute to allow subcategorization based upon the performance of individual POTWs without rendering the removal credits provision meaningless — for removal credits would still apply to the users of less effective POTWs. In our original opinion, we rejected the contention that the EPA must account for the performance of individual POTWs in its national rulemaking and agreed instead with the EPA’s interpretation of the stat *260 ute. 25 The EPA has interpreted the subca-tegorization provision to apply to characteristics of the petitioners’ own facilities, and not to the characteristics of a POTW used by petitioners. The EPA concluded that the performance of individual POTWs may be considered only in the context of the removal credits provision. Petitioners’ arguments in their request for rehearing do not persuade us that this interpretation of the statute is unreasonable. Specifically, we do not find convincing petitioners’ argument that subcategorization and removal credits necessarily play distinct roles. Rather, the statute indicates that removal credits are intended to be the mechanism for relieving indirect dischar-gers from PSES when the POTW and its users are capable, together, of meeting the relevant standards with reduced pretreatment by the indirect discharger. The statute provides that removal credits will be available only when “the discharge from such works does not violate that effluent limitation or standard which would be applicable to such toxic pollutant if it were discharged by such source other than through a publicly owned treatment works.” 26 Removal credits simply reallocate between the POTW and the indirect dischargers the responsibility for removing the total amount of pollutant necessary to achieve the applicable limit. Together, the POTW and indirect dischargers must still meet the limits. 27 In other words, if no pollutants “pass through” a given POTW, the indirect dischargers who use that POTW may seek removal credits to maintain the status quo. If removal credits are awarded, the allocation of responsibility between the two parties would remain the same as it was before the promulgation of PSES. The difference between subcategorization and the award of removal credits is that if a subcategory is created for indirect dischargers using an exemplary POTW, indirect dischargers would not be required first to comply with pretreatment standards and then to seek removal credits. That is, the status quo would never be altered. As we explained in our original opinion, however, Congress clearly intended the PSES to be promulgated on a national, categorical level and intended the removal credits provision, like the FDF variance provision, to provide a “safety valve” for the resolution of individualized claims. 28 This conclusion is further supported by our concern that the creation of subcategories for users of exemplary POTWS would contravene Congress’ intent in prohibiting the award of removal credits in the absence of sludge regulations. The consequence of either subcategorization or the award of removal credits would be to make the POTW responsible for removing a greater proportion of toxic pollutants than it would have been required to remove if its users complied with PSES. Thus, pollutants that would not otherwise be present in the POTW’s influent might contaminate the sludge that the POTW must in turn dispose of. 29 We therefore conclude that we cannot require PSES to be relaxed by subcategorization until such time as the EPA promulgates regulations addressing the presence of toxic pollutants in sludge. 30 Finally, even if we were to construe the statute to permit the EPA to create *261 subcategories based on the performance of POTWs, we would uphold the EPA’s conclusion that the data submitted by petitioners did not provide an adequate basis for establishing a subcategory for the users of the Gulf Coast POTWs. The EPA has adequately explained its reasons for concluding that petitioners’ data was insufficient to establish that toxic pollutants consistently do not pass through Gulf Coast’s facilities. 31 The variation in the data regarding individual pollutants and Gulf Coast’s individual POTWs further supports the view that such claims would be addressed more appropriately under the removal credits provision. II. Best Available Technology (BAT) Issues A. The EPA’s Use of Weighted Averaging in Deriving the Long-Term Averages In order to derive the BAT limitations the EPA calculated the long-term average performance of the plants in its data base that treated a particular pollutant by weighting the sum of the average of detect and non-detect values reported for each plant. 32 In its initial brief CMA contended that the EPA “fabricated” data by assigning non-detect values to plants that did not report any, thus lowering the long-term averages and the resulting limitations. In our original opinion we stated that assigning non-detect values to plants that did not report them resulted from the EPA’s use of the recognized statistical technique of weighted averaging, and upheld the EPA’s method as within its discretion. 33 CMA now contends that the EPA misrepresented its technique as weighted averaging. CMA points out that, in calculating the long-term averages, the EPA assigned the same average proportion of non-detect values to each plant in the data base even if a particular plant did not report any non-detect values. The EPA assumed, in its own words, “that nondetected values should be weighted in accordance with the frequency with which nondetected values for the pollutant generally were found in the daily-data plants.” 34 This assumption, CMA argues, is not supported by the record. CMA correctly points out that the EPA’s assumption initially appears implausible because plants with efficient treatment systems seem likely to report a higher proportion of non-detect values than poorly performing plants. CMA may also be correct that the EPA’s statistical procedure is not “weighted averaging” in the usual sense. Nevertheless, it is too late to make these objections. When CMA had an opportunity to comment on the challenged procedure in 1985, it stated that it found the procedure “acceptable:” EPA uses a weighting procedure in calculating plant-pollutant long-term averages to account for concentrations in the data base which are below the assigned detection limit. This procedure may be acceptable for this use. It makes the statistical calculations for the variability factor more tractable and, more importantly, comes closer to representing an analytically identifiable concentration. It should be emphasized, however, that ' this procedure overestimates the mass of priority pollutants being discharged by a substantial amount. 35 CMA now seems to have concluded that the EPA’s methodology underestimates rather than overestimates the mass of pollutant *262 discharged. Nevertheless, having failed to object to the EPA’s methodology when it could easily have been changed, CMA cannot object now. 36 Even apart from considerations of the timing of CMA’s objections, we note that the EPA has presented an elaborate justification of its methodology. The EPA argues that the method it employed is statistically reasonable, and that because it is merely an averaging technique — whether “weighted” or not — its use did not prejudice CMA. We see no reason to alter our conclusion that the EPA’s methodology was within its broad discretion in the choice of statistical techniques. 37 B. The Effects of Dilution on the Measurement of Plant Performance In its original Petition for Review, CMA argued that the influent streams of five of the forty plants in the BAT data base contained water that did not derive from OCPSF processes, and that this resulted in unduly low effluent-concentration measurements and BAT limitations. 38 We rejected this argument, deferring to the EPA’s determination that the dilution identified by CMA had no significant effect on the measurement of treatment performance because it occurred before influent sampling. 39 CMA now argues that this court misconceived the importance of dilution, and recasts its argument as follows: Suppose, for example, that one liter of organic chemical wastewater influent to a treatment plant (Stream A) contains ten grams of pollutant X, and the plant achieves 80 percent removal. The resulting effluent will contain two grams per liter. Now if the original 10 grams/liter wastewater were diluted prior to treatment by another liter of water (Stream B) (not containing pollutant X), then the two liters of combined influent will have a measured concentration of 5 grams per liter.... After treatment at the same 80 percent removal rate, the combined effluent will now measure one gram per liter. The dilution has had no effect on treatment performance as measured by percent removal. But dilution has had the effect of making the treated effluent concentration in Stream A appear to be much lower than it really is — in this case by 50 percent. Yet in this example, it is the lower one gram/liter measurement that EPA used to develop the guideline limit. Although CMA’s argument initially seems plausible, it is based on the premise that BAT achieves the same percentage removal (80% in CMA’s example) regardless of the influent concentration, and that the concentration in an undiluted influent cannot be reduced to the same concentration at the end of the stream as the concentration in a diluted influent. The record does not establish this premise. It is probably true that the EPA’s decision not to modify its sampling data to take account of dilution resulted in BAT limitations somewhat lower than if dilution had been taken into account. The EPA measured the influent of a plant, however, at a point downstream of the place where dilution, if any, occurred. As a result, modifying its data to take account of dilution as CMA proposes would have forced the EPA to treat data from identical processing systems operating upon identical influent streams differently depending upon whether the influent streams were “natural” or “diluted.” We see no reason to conclude that the EPA was rationally required to treat a wastestream that had been diluted before it was measured differently from a wastestream that came directly from a processing plant without being diluted so long as the streams, when measured, measured *263 the same. The EPA’s refusal to adopt this course, we conclude, was rational and within its discretion. In addition, the record shows that, despite the dilution complained of by CMA, the influent pollutant concentrations at most of the five plants it identifies 40 appear to be well above the industry average for the relevant pollutants. 41 C. The Calculation of Variability Factors From Less Than 100% of the Data To develop the BAT limits the EPA multiplied the long-term average performance of the plants in the data base by a variability factor. The EPA calculated the variability factors by excluding daily data representing the 1% highest discharges and monthly data representing the 5% highest discharges. 42 Petitioners PPG and Dow urge, as they argued in their earlier briefs, that the BAT limitations are not achievable because these data were excluded. In our original opinion, we held that the EPA had reasonably excluded data points exceeding the 99th and 95th percentiles because the EPA could reasonably assume that these points were isolated and extreme departures from average performance that were due to quality-control problems or upsets. 43 PPG and Dow argue, however, that the EPA had already excluded data representing quality-control problems and upsets from the data base. The remaining 1% and 5% instances of deviation, in their view, therefore must occur during normal operations, and should have been used in the calculation of the variability factors. The record does not support the petitioners’ contention that all individual data points (as distinguished from individual data sets) that might have resulted from upsets and quality-control problems were deleted from the data base. The record states: In general, the Agency’s BAT toxic pollutant data base editing criteria were as follows: ... Data not representative of BAT technology performance were eliminated from the data base. Examples of reasons for not being representative of BAT technology performance include process spills; treatment system upsets; equipment malfunctions; performance not up to design specifications; past historical performance; or performance exhibited by other plants in the data base with BAT technology in place. 44 The EPA also asserted that “[t]he data were reviewed in detail and edited to assure that only data representing BAT-level design and operation were retained for purposes of developing limitations.” 45 The EPA edited from the data base the entire data sets from plants that generally failed to conform to BAT standards; however, it did not use its editing criteria to exclude individual data points. 46 Indeed, because the detailed day-to-day operating conditions of the data-base plants were known only to the plant operators and not to the EPA, 47 it would have been impossible in most instances for the EPA to determine the reasons for an unusually high discharge. The EPA’s editing criteria were designed to select a representative group of well-operated plants utilizing BAT technology, but even well-operated plants occasionally will experience quality-control problems. The EPA therefore reasonably could assume that the individual data points representing the most extreme departures from normal operation were caused by quality-control problems that either were unlikely to recur or might be overcome by more efficient operation, and, in its discretion, the EPA could exclude them in calculating the varia *264 bility factors. This conclusion is reinforced, moreover, by the fact that the EPA identified specific quality-control measures that could be used to reduce the extreme variability reported by some of the data base plants. 48 D. The Achievability of the BAT Limitations for Volatile Pollutants Based on Steam-Stripper Technology PPG and Dow object to both of our holdings rejecting their challenge to the BAT limitations for volatile pollutants based on steam stripper technology. 49 First, they complain that we wrongly held that the EPA reasonably excluded the 85 ppb highest discharge from plant 415 from its data base in developing the BAT limitations for trichlorethylene (TCE). The exclusion of this discharge, they contend, makes the EPA’s BAT .daily, limit of 69 ppb for TCE unachievable. Second, Dow and PPG object to our holding that the EPA can determine the “best” plant for the development of BAT limitations on a pollutant-by-pollutant basis without demonstrating that any single plant meets the limits for all of the pollutants in its effluent. With respect to the first issue, we conclude, as we stated in our previous opinion, that the EPA reasonably could infer that the isolated 85 ppg discharge was due either to an upset or to a quality-control problem. Because the data concerning the cause of an unusually high discharge often will be uniquely within a plant operator’s control, and because the EPA has considerable expertise in interpreting sampling data, we do not believe that the EPA’s exclusion of the isolated and extremely high discharge was unreasonable. We also reaffirm our holding that the EPA may determine the “best” plant upon which to base BAT limitations on a pollutant-by-pollutant basis. The. fragments of legislative history that the petitioners quote to us are out of context and do not address the point at issue. The Fourth Circuit’s decision in Tanners’ Council of America v. Train 50 does hold that EPA limitations had not been shown to be achievable when the plants in the data base had met “the limitations for some, but not all of the pollution parameters.” 51 That court’s rationale was unclear, however, and we have declined to follow its decision. Association of Pacific Fisheries v. EPA 52 and CPC International, Inc. v. Train, 53 also relied on by PPG and Dow, are not contrary to this court’s ruling. In CPC International the EPA’s regulations were remanded because they were not supported by data from any plant, 54 and in Pacific Fisheries, a regulation was remanded because the study upon which the EPA relied did not demonstrate that the sole data-base plant could comply with the regulations. 55 Here, in contrast, at least one plant can meet every BAT limitation, and, as we held originally, the fact that no plant has been shown to be able to meet all of the limitations does not demonstrate that all the limitations are not achievable. E.. The Achievability of the BAT Limitations for Priority Pollutants for Which In-Plant Biological Treatment is the Model Technology The EPA designated in-plant biological treatment as the model technology for the treatment of twenty priority pollutants, To establish the BAT2 limitations for these pollutants, however, the EPA relied on a data base consisting solely of three end-of-pipe biological treatment plants. 56 As a result, CMA argues, the EPA has failed to demonstrate that the BAT2 limits for the priority pollutants are achievable. *265 We initially rejected CMA’s argument, stating: [t]he petitioners have failed to demonstrate that end-of-pipe biological treatment systems are sufficiently different from in-plant systems to make the EPA’s reliance on end-of-pipe data irrational. For all we can tell from the parts of the record that have been cited, the only difference between the two systems is that they are installed at different positions in the production process. 57 CMA has now clarified its position. End-of-pipe systems, CMA points out, typically employ much longer detention times than in-plant systems. Indeed, the record shows that the three end-of-pipe plants used to develop the BAT2 limitations for the priority pollutants — Plants 1293T, 948F, and 2536T — had detention times of 17.2, 3.5, and 1.6 days respectively. In contrast, the EPA used a maximum detention time of 2.1 days to estimate the costs of in-plant treatment systems, a period substantially shorter than the detention times at two of the three end-of-pipe plants. Because detention time is a key variable determining the effectiveness of biological treatment, CMA asserts that the achievability of the BAT2 limitations for the priority pollutants, which the CWA requires the EPA to demonstrate, 58 has not been established. The EPA justifies its use of data from the three end-of-pipe plants on the grounds that end-of-pipe and in-plant systems utilize the same biological processes and the three plants treated wastestreams comparable to those that would enter an in-plant treatment system. 59 It does not follow, however, that the differing detention times can rationally be ignored; if anything, these similarities suggest that the differences in detention times are unlikely to be compensated for by an offsetting variable. The EPA further maintains that CMA’s focus on detention time is. inappropriate because detention time is only one of many factors that affect the efficacy of a biological treatment system. Other factors include the organic loading of biodegradable material in the influent, the concentration of biodegrading organisms in the aeration basin (MLVSS), and the length of time that these organisms remain in the aeration basin. The EPA adds that some plants treat priority pollutants successfully even with detention times less than thirty hours, and that increasing MLVSS concentrations can shorten- the time necessary for successful treatment. . Nevertheless, the record contains no performance data for in-plant treatment of the twenty priority pollutants at issue, and the EPA concedes that detention time does affect the efficacy of a treatment system. Although the record supports the EPA’s assertion that higher MLVSS concentrations decrease the detention time necessary for a given level of treatment, it does not make clear exactly what level of pollution would result from any given combination of shorter detention time and increased MLVSS. The EPA’s claim .that the BAT2 limitations can be met because MLVSS concentrations can be increased consequently appears to be no more than an educated guess. The. EPA bears the burden of producing a reasonable basis on the record for its regulations. 60 The EPA has failed, however, to demonstrate a reasonable basis for its. conclusion that in-plant treatment can eliminate pollutants as effectively as the end-of-pipe systems of Plants 1293T and 948F. The limitations based on these plants therefore are arbitrary and capricious, and must be remanded to the EPA for further rulemaking proceedings. The *266 portions of our previous opinion to the contrary are withdrawn. F. The EPA’s Erroneous Inclusion of Three Complexed Metals in the Limits for Uncomplexed Metals Appendix A of the effluent limitations establishes limits for the discharge of toxic uncomplexed metals. Appendix B lists “complexed metals,” that is metals bonded with an organic molecule, and provides that limits for such pollutants will be established on a case-by-case basis by the individual NPDES permit writer. 61 In its initial brief DuPont contended that the EPA erroneously included three complexed metals, including tetraethyl lead, tetramethyl lead, and anti-knock fuel additives, in Appendix A. DuPont contended that these compounds should have been listed in Appendix B with other complexed metals. In its rehearing petition DuPont notes that this Court failed to address this issue. By notice filed May 12, 1989, the EPA conceded that these three compounds are complexed metals and it erred by including them in Appendix A. This Court therefore grants the petition for review, strikes these three compounds from Appendix A, and remands the issue to the Administrator for further rulemaking proceedings. Conclusion The petitioners’ request for a rehearing is denied in all respects except that 1) our previous opinion is clarified as explained above; 2) the limitations for priority pollutants for which in-plant biological treatment is the model technology are remanded to the EPA for further rulemaking proceedings insofar as they are based on Plants 1293T and 948F; and 3) the three complexed metals erroneously included in Appendix A are ordered stricken from that Appendix, and the issue is remanded to the Administrator for further rulemaking proceedings. Footnotes [1] 870 F.2d 177 (5th Cir.1989). [2] See id. at 266. [3] Id. at 226. [4] Id. [5] 33 U.S.C. § 1369(b)(1) (1982 & Supp. V 1987). [6] 33 U.S.C. § 1365(a)(2) (1982 & Supp. V 1987). [7] 870 F.2d at 226. [8] Id. (quoting 33 U.S.C. § 1311(n)(6) (1982 & Supp. V 1987)). [9] Id. at 221-22. [10] See id. at 217-21. [11] Id. at 220. [12] Id. at 219. [13] Id. at 220. [14] Id. at 219-20. [1] Petitioners assert on rehearing that the record does not in fact support the EPA’s contention that it estimated the costs of activated sludge systems for "nearly half of the plants in the industry requiring treatment improvements.” See 870 F.2d at 219-20. Petitioners' own calculation of 20% was based, however, on all of the plants in the industry, rather than only those plants requiring treatment improvements. Upon reexamining the record, we note that the EPA costed activated sludge systems for 59 plants — out of 206 requiring system upgrades (other than contract hauling) and 297 plants altogether. [15] 5 U.S.C. § 553(b)(3), (c) (1988). [16] 870 F.2d at 235-36. [17] Id. at 236. [18] 50 Fed.Reg. 29,079; see 870 F.2d at 236. [19] 870 F.2d at 236. [20] 51 Fed.Reg 44,090. [21] Id. at 44,084. [22] Id. at 44,090. [23] Id. (emphasis added). [24] 870 F.2d at 257-61. [25] Id. at 244-46; see also Cerro Copper Products Co. v. Ruckelshaus, 766 F.2d 1060, 1068 (7th Cir.1985). [26] 33 U.S.C. § 1317(b)(1) (1982). [27] The formula for awarding removal credits is y = x/l-r where x = the discharge limit of a pollutant according to the EPA’s pretreatment standards, r = the POTW's consistent removal rate for that pollutant, and y = the revised discharge limit. Cerro Copper, 766 F.2d at 1063 n. 1. [28] 870 F.2d at 258-59; see also Cerro Copper, 766 F.2d at 1068. [29] See Armco, Inc. v. EPA, 869 F.2d 975, 980 (6th Cir.1989). [30] We do not find Congress’ refusal to establish separate regulations for pollutants which may contaminate sludge to be to the contrary. Congress simply concluded that this concern would be addressed by the regulation of sludge under Section 405(d) of the CWA. See 870 F.2d at 248. [31] In particular, the EPA found it difficult to determine the actual level of pollutants in in-fluent and therefore could not determine the effects of dilution on the POTWs’ "removal" rates. As we explained in our original opinion, one of the EPA's central concerns in promulgating PSES was to ensure that dilution of pollutants by POTWs did not substitute for effective treatment. Id. at 245-46. [32] See id. at 227. [33] See id. at 227-28. [34] 50 Fed.Reg 29,080. [35] CMA’s December 23, 1985 Comments on EPA’s July 17, 1985 and October 11, 1985 Notices of Availability of New Information for the Organic Chemicals, Plastics and Synthetic Fibers Category at V-70, reprinted in Joint App. at 1078. [36] See, e.g., Campos-Guardado v. INS, 809 F.2d 285, 291 (5th Cir.), cert. denied, 484 U.S. 826, 108 S.Ct. 92, 98 L.Ed.2d 53 (1987); Brotherhood of Railway, Airline, & Steamship Clerks v. St. Louis Southwestern Ry., 676 F.2d 132, 136 (5th Cir.1982); Myron v. Martin, 670 F.2d 49, 51 (5th Cir.1982); Reynolds Metals Co. v. EPA, 760 F.2d 549, 563 (4th Cir.1985); Association of Pacific Fisheries v. EPA, 615 F.2d 794, 817 (9th Cir.1980). [37] See 870 F.2d at 228. [38] See id. at 233. [39] Id. at 234. [40] Plants 2313, 725, 2394, 1293, and 1494. [41] See Dev. Doc. V-105-07, reprinted in Joint App. 3646-48; Admin.R. 115331-764, reprinted in Joint App. 5219-424. [42] See 870 F.2d at 228-29. [43] See id. at 230. [44] Dev.Doc. VII-183, 185, reprinted in Joint App. 3883, 3885. [45] 52 Fed.Reg. 42,540. [46] See Dev.Doc. VII-189-90, reprinted in Joint App. 3889-90. [47] Admin.R. 103216 n. *, reprinted in Joint App. at 2987 n.*. [48] See 52 Fed.Reg. 42,564. [49] See 870 F.2d at 238-39. [50] 540 F.2d 1188 (4th Cir.1976). [51] Id. at 1192-94. [52] 615 F.2d 794 (9th Cir.1980). [53] 540 F.2d 1329 (8th Cir.1976). [54] See id. at 1338-40. [55] See 615 F.2d at 819. [56] See 870 F.2d at 240. [57] Id. at 240. [58] See 33 U.S.C. § 1314(b)(2) (1982). [59] Dev.Doc. VII-49, reprinted in Joint App. at 3749. [60] See Pacific Fisheries, 615 F.2d at 819; E.I. du Pont de Nemours & Co. v. Train, 541 F.2d 1018, 1037-38 (4th Cir.1976), rev’d in part on other grounds, 430 U.S. 112, 97 S.Ct. 965, 51 L.Ed.2d 204 (1977); CPC Int'l, 540 F.2d at 1338-40; FMC Corp. v. Train, 539 F.2d 973, 981-82 (4th Cir.1976); American Iron & Steel Inst. v. EPA, 526 F.2d 1027, 1062-63, 1065 (3d Cir.1975). [61] 52 Fed.Reg. 42,542-43.