COFFIN, Chief Judge.
These consolidated petitions have been brought by eleven manufacturers of pesticides
In 1974 the Agency hired an outside contractor, Roy F. Weston, Inc. (Weston), to analyze the industry. Weston submitted its final report in December of 1975. In early 1976 EPA hired a second contractor, Environmental Science and Engineering, Inc. (ESE), to evaluate Weston's work. ESE determined that Weston's work needed improvement and undertook its own study of the industry. In late 1976 EPA published interim final regulations that were immediately effective, but on which EPA invited public comment. 41 Fed.Reg. 48087 (1976). An interim development document and an economic analysis explaining the derivation of the interim regulations were also released.
After the final regulations issued, one of the petitioners filed a motion for reconsideration alleging, among other things, that analytical techniques were not available to detect many pesticides at the levels stated in the regulations. EPA reexamined the record and discovered that some measurement methods that EPA thought were available might not be reliable. Accordingly, EPA amended the regulations so that the pesticide content of process waste water would be limited for the producers of only 49 out of several hundred pesticides. 43 Fed.Reg. 44845, 44856 (1978).
The first petition for review challenged the interim final regulations. When the final regulations were published we granted permission to amend so as to include review of the finals. BASF Wyandotte Corp. v. Costle, 582 F.2d 108 (1st Cir. 1978). Subsequently the petitions for review filed in other circuits were transferred to this circuit. The consolidated petitions assert several procedural and substantive errors in the regulations and their promulgation.
I. Organic Pesticide Manufacturing
A. Administrative Procedure Act Compliance
Petitioners' first complaint is that EPA failed to comply with the requirements of the Administrative Procedure Act in that the final regulations were so different from the interim final regulations that the interims were not notice of "either the terms or substance of the proposed rule or a description of the subjects and issues involved." 5 U.S.C. § 553(b)(3). This requirement is a critical one because it supports the assumption we make with regard to EPA's substantive decisions that those decisions are in fact the product of informed, expert reasoning tested by exposure to diverse public comment. Though our review of an agency's final decision is relatively narrow, we must be strict in reviewing an agency's compliance with procedural rules. See Weyerhaeuser Co. v. Costle, 590 F.2d 1011, 1027-1028 (D.C. Cir. 1978).
In this case EPA issued interim final regulations and sought comments on them. Industry representatives, government agencies, and others submitted voluminous comments on many aspects of the interim regulations. It is clear that EPA gave careful consideration to these comments. The Agency summarized the public comment, together with the Agency responses in the prologue to the final regulations. 43 Fed. Reg. 17781-85 (1978). The Agency accepted several suggestions made in comments critical of the interim regulations. For instance, EPA deleted some parameters by which the interim regulations controlled
The procedural rules were meant to ensure meaningful public participation in agency proceedings, not to be a straitjacket for agencies. An agency's promulgation of proposed rules is not a guarantee that those rules will be changed only in the ways the targets of the rules suggest. "The requirement of submission of a proposed rule for comment does not automatically generate a new opportunity for comment merely because the rule promulgated by the agency differs from the rule it proposed, partly at least in response to submissions." International Harvester Co. v. Ruckelshaus, 155 U.S.App.D.C. 411, 428, 478 F.2d 615, 632 (1973); Weyerhaeuser Co. v. Costle, supra, at 1031; American Frozen Food Institute v. Train, 176 U.S.App.D.C. 105, 132, 539 F.2d 107, 134 (1976). Even substantial changes in the original plan may be made so long as they are "in character with the original scheme" and "a logical outgrowth" of the notice and comment already given. South Terminal Corp. v. EPA, 504 F.2d 646, 658, 659 (1st Cir. 1974).
The essential inquiry is whether the commenters have had a fair opportunity to present their views on the contents of the final plan.
The first and principal change complained of is the Agency's decision to merge the first three interim subcategories
The industry comments were almost unanimous in condemning the three original subcategories both for being internally inconsistent and insufficiently differentiated from the other categories.
Neither suggestion is grounds for remand. It should be clear to commenters when they criticize a regulatory scheme that if the agency accepts those criticisms, a new scheme will be substituted. The commenters cannot claim they had no notice to propose and discuss alternatives. And in fact they did so, suggesting a number of new approaches based on different ways to subcategorize the industry. In fact, at least one commenter recognized that if the existing scheme was not defensible, one alternative would be to abandon subcategories. This commenter, also a petitioner, Monsanto, wrote:
Clearly Monsanto realized that the Agency would have to consider alternatives and was not finally committed to subcategories. Moreover, Monsanto's response indicates that the Agency had not misled petitioners into thinking that they need only state views on the desirability of more subcategories rather than the undesirability of fewer.
Another repeated call of the industry commenters was that the subcategories should be more equitably treated because the record did not support the significantly different guideline limits assigned to the different subcategories.
For the same reasons, we do not think EPA was obligated to provide further opportunity to comment on "its decision to limit the regulations to those pesticides for which there are reliable analytical methods" or "its determination that all pesticides can be treated to a single level regardless of differences in treatability of particular pesticides". To the extent the latter differs from the decision to consolidate categories, it was clearly signalled in the interim regulations. As the commenters pointed out, the interim subcategories were based on physical structure and each included chemicals of diverse treatability. The industry's comments plainly suggested that treatability ought to be taken into account.
The last point we must address is a different and more difficult one. One of EPA's justifications for consolidating the subcategories and for reducing the discharge limits for many pesticides was the conclusion, based in part on "additional raw waste load and treatment data, and additional pilot plant and laboratory data" that "the waste waters of all organic pesticide chemicals can be treated or controlled to the levels documented in the Agency's data base." 43 Fed.Reg. at 17777. The BASF petitioners argue that since the EPA's conclusions were premised on "new data on the applicability and effectiveness of treatment by activated carbon and by hydrolysis" EPA should have entertained a new round of comments.
Certainly nothing is more important than the bottom line numbers which determine whether individual plants are or are not in compliance with the regulations. The data that an agency has used to set proposed limits obviously should be subject to public comment if possible. Indeed, as another circuit court has noted, factual matters "are especially subject to verification through the notice and comment process and less acceptably removed therefrom." Weyerhaeuser, slip op. at 1030 n. 26.
This does not mean, however, that any new numbers gathered after publication of proposed regulations must be submitted for comment. It is perfectly predictable that new data will come in during the comment period, either submitted by the public with comments or collected by the agency in a continuing effort to give the regulations a more accurate foundation. The agency should be encouraged to use such information in its final calculations without thereby
We conclude that EPA was within the law. The final regulations set the pesticide discharge limit on the basis of data from nine plants with recommended pesticide removal technology. Seven of these nine plants were mentioned in the interim development document as forming the basis for or supporting the limits set by the interim regulations.
More importantly, the interim development document disclosed the method by which EPA calculated the limits. This methodology was substantially the same for the interim and final regulations. EPA afforded petitioners the chance to make the most meaningful possible contribution to the regulatory process because the petitioners could use the information to collect data at their own plants, analyze it, and submit it to EPA. Not only would such input have guaranteed that EPA would be acting on the most complete data base possible, but it would have guaranteed that the special problems faced by each petitioner would have been taken account of by EPA and contributed to the resulting limitations.
As to these seven plants, petitioners knew all they needed to know in order to make a meaningful contribution to the regulatory process and a meaningful comment on the regulations proposed. All they did not know was what the actual raw numbers would be. While these numbers were certainly crucial in setting the limits and of obviously great interest to the petitioners, we are not convinced that knowing them would have significantly improved petitioners' opportunity to comment. They would have had a different numerical target for their many complaints about the limits set, but that would not have greatly advanced their ability to make positive contributions to the process by criticizing from their own knowledge the treatment technologies recommended, the success they could anticipate by implementing those technologies, and the analytical and statistical methodology EPA used to turn the raw numbers into effluent limits. In short, in this case, we think it was far more important for EPA to solicit comments on how it intended to collect and use data than on the data itself.
As for the two plants that did not contribute to the generation of the interim limits, the same arguments apply in part. All they added to the process were more numbers to be used in the same way as the other numbers. Their treatment technologies are not markedly different from the others and conform to EPA's fully disclosed model technology.
B. EPA's Scientific Methodology
The BASF petitioners have tenaciously attacked the methodology and data used by ESE in calculating the final effluent limitations for organic pesticides. If ESE's work is not reliable, then the final limitations cannot stand. Partly because of a belated turnover of some 2000 pages of laboratory documents — all of those used by ESE for this rulemaking proceeding — we have been faced with three rounds of briefing memoranda, dealing with targets that were both moving and increasingly particularistic. If we are not to lose our bearings in this extensive cross-fire of chemical expertise, it is important for us to gain and hold a perspective both as to facts and law.
ESE was retained in 1976, after EPA became dissatisfied with the work done over the prior two years by another contractor. ESE thereafter collected data from pesticide manufacturers and did its own experimentation on samples of waste water, the manufacturers furnishing 95 per cent of the data, and ESE laboratory work accounting for 5 per cent. The methodology and resulting data used by ESE to identify and quantify organic pesticides relate to two techniques, gas chromatography (GC) and thin layer chromatography (TLC). The method attracting most of the controversy in this appeal is gas chromatography. It involves vaporizing a sample of effluent, passing it through a column filled with various materials, which impeded the flow of each compound at a characteristic rate. Results are recorded on a continuous chart which shows a "peak" when each compound exits the column, thus identifying it in terms of the time it appears, and which indicates concentration by the height and area of the peak. The method is neither simplistic nor self-executing. It involves the use of several columns, the testing and calibrating of each with samples containing known quantities of known pesticides, continual checking, and constant alertness for the presence of unforeseen substances which can distort the recording on the strip chart. The method succeeds only in careful and expert hands.
The lore of gas chromatography has been recognized for some time, being the subject of a handbook compiled in 1972 by EPA's Analytical Quality Control Laboratory (and the source of ESE's practices in this rulemaking procedure), the "Analysis of Pesticide Residues . . ." in 1974 and "Analytical Procedures for Pesticides Approved by EPA per 40 C.F.R. Part 136", App.1997-2090. These compilations dealt with means
The factors underlying the present controversy as to the methodology of identifying and measuring organic pesticides are threefold: the complexity of the process, the high degree of elimination of pollutants required by the regulations, and the penalties for non-compliance. 33 U.S.C. § 1319. Petitioners point with dismay to the analytical obstacles, pitfalls, and limitations which EPA acknowledges and warns against,
The considerations governing our task of review in such a case as this have been recently illumined in Weyerhaeuser Co. v. Costle, at 1025, where Judge McGowan, writing for the court, said:
After quoting from the court's earlier opinion in Industrial Union Dep't v. Hodgson, 162 U.S.App.D.C. 331, 338-39 n. 18, 499 F.2d 467, 474-75 n. 18 (1974) ("Where existing methodology or research in a new area of regulation is deficient, the agency necessarily enjoys a broad discretion to attempt to formulate a solution to the best of its ability on the basis of available information."), Judge McGowan summarized the court's view of its mission:
With these precepts in mind, we consider petitioners' challenges.
The first is to the reliability and adequacy of the two basic methods endorsed by EPA, gas chromatography and thin layer chromatography. Petitioner's general attack is based, as we have noted in note 17, on the acknowledged uncertainties attending the successful use of the methods in dealing with complex mixtures. It seems to us that EPA's cautions are to be preferred by far to overclaiming or overselling. It also seems to us that action of this nature, although not without risk to manufacturers, is preferable to no action at all. We are mindful of the fact that Congress contemplated a certain amount of "technology forcing", Weyerhaeuser, at 1057, expecting EPA "to press sometimes beyond the most advanced technology currently being used". Id. at 1061. Here, there is no claim that the technology has not been developed but merely that EPA has overstated its public acceptance. If the issue were whether the American Society for Testing and Materials (ASTM) and the American Public Health Association (APHA) had approved GC methods for testing all three classes of organic pesticides for which EPA claims GC to be appropriate, EPA might lose. But it seems clear to us, from the literature, including the series of manuals and handbooks dealing with both GC and TLC going back at least to 1972, and from the fact that several EPA test methods for organic pesticides have almost completed running the gamut of ASTM and APHA approval processes, that any a priori attack on the methods must fail.
This leads us to more specific criticisms. The most sweeping of these is petitioners' claim that ESE, so far as the record discloses, pursued no systematic and reliable quality control procedures in running its tests on samples of effluent. In its initial form the argument was that the many steps
ESE came on the scene after a prior contractor had been found inadequate. It obviously had a mandate to produce scientifically acceptable work; EPA had demonstrated its reaction to inadequacy. Between late 1976 and the fall of 1977 ESE invested 16,500 man hours in this rulemaking process. (App. 1491) As we have noted, all of its working papers, totalling 2000 pages have been turned over to petitioners. The specific criticisms we shall later discuss are the total harvest of hindsight. Most, we think, are answered satisfactorily by EPA; some are ignored. We find ourselves, as a reviewing court, dealing with adversaries who deal with the issues on different levels. Petitioners have combed the record to select what they deem to be errors or examples of sloppy work. EPA, on the other hand, seems to have adopted a rather bland approach, selecting only what it views as important challenges, and giving them a minimum response.
The issue of ESE's quality control standards offers an example of these diverse approaches. Petitioners decry the absence of a description of methods, quality control charts, the failure to prepare a "reproducible standard curve", the failure to run an adequate number of standards, an inadequate number of spiked and "method blank" samples, and duplicate analyses insufficient in quantity and quality. Their citations to the record are few. EPA's response to this range of criticism is largely set forth in the following footnote language in its memorandum responding to petitioners' reply brief: "Complaints of inadequate numbers of standard and spiked samples are refuted by laboratory documents supplied petitioners pursuant to the Court's request at argument."
We have therefore conducted our own review. It was restricted to 238 pages of ESE documents (App. 3352-3590). It was necessarily superficial, since we are not laboratory technicians, since the function of the various forms and entries was often not self evident, and since many of the forms were illegible, some were apparently duplicates (App. 3469, 3470; 3413, 3556) or triplicates (App. 3363, 3474, 3582), and a sizeable number were upside down. We found frequent evidence that ESE testing had involved the running of duplicates,
We cannot be sure that these references are an adequate response to criticisms levied at ESE's general control procedures. But we cannot play the role of Superchemist.
Our feeling is strengthened by our review of some of the specific instances of alleged error in results charged by petitioners. They have pointed to the wide divergence of the analyses conducted principally by manufacturers Rhom and Haas and Eli Lilly and those performed by ESE on "split" samples of the same waste water. ESE consistently found much higher quantities of pesticide in the samples. As we have indicated above, we have not found sufficient reason to reject ESE's analytical procedures or results. And, while the methods generally used by the companies were sufficiently reliable,
Petitioners also focused on what first appeared to them as a "patent error" in ESE's calculations of the amount of waste water used in a test, amounting to more than a 55 gallon drum. But this turned out to be correct, EPA saying that the sample was so polluted that it had to be diluted 250,000 times. Petitioners noted two strip charts for sample 9010, speculating that a "shoulder" indicated interference and that a peak which ran off the paper could not be only 3.6 centimeters high. But EPA counters by saying the first peak was so pesticide-laden that it did indeed run off the paper, but that the rerun was indeed 3.6 centimeters high. The supposed interference denoted by the shoulder was the occasion for the technician's comment we have noted concerning the jiggling of the recorder pen. App. 3412. Other charges were that two strip charts supposedly representing duplicate analyses were not "remotely comparable" — the response being that they were successive analyses with the differences due to the large amount of pesticide present in the early extractions; that ESE improperly used higher influent numbers ascertained by a Perkin-Elmer instrument with no explanation for rejecting the lower numbers obtained by a Varian instrument — the reply being that each instrument is more effective in different ranges of numbers. Although petitioners have a retort in each such instance, we must say that EPA clearly has the better of these arguments.
There are other even more minor issues where we confess that we are not sure of the winner — whether as to a particular analysis, a standard strip chart should have been verified, whether ESE's injections were too little in volume, whether the lack of confirmatory tests on alternate columns of different polarity was significant. As to these and other issues, EPA either did not respond or did not do so at sufficient length to be understandable. But we have said enough to indicate that residual doubt on
Petitioners have raised one additional issue of a substantive technical nature in challenging EPA's choice of COD as a parameter. The contention is that it serves no purpose where wastes possessing large amounts of salts are to be dealt with. Diamond Shamrock's data are said to be of doubtful validity. But without such data, leaving only that developed by Monsanto's Muscatine and Anniston plants, the standard would be even more stringent. Wholly apart from absence of prejudice, EPA points to the wide use of COD in industry as a measure of long term biochemical oxygen demand and a comment from Olin that "[I]t is a mystery why the Agency feels compelled to establish a limitation on BOD when COD measurements are more feasible." C.App. 212.
C. EPA's Determination of Best Practicable Control Technology Currently Available
Even accepting the work of ESE as accurate, petitioners argue that the results do not support EPA's conclusion that carbon adsorption and hydrolysis are effective technologies for pre-treating pesticide waste streams to remove organic pesticide chemicals. Closely related to this challenge are the arguments that the numbers with which EPA calculated the final effluent limits were inaccurate, thus invalidating those limits, and that EPA could not reasonably combine all manufacturers of organic pesticides into one category required to meet the same limits. EPA's decision that by using the model technology, including either carbon adsorption or hydrolysis, all plants achieved "similar" pesticide content in their effluent led EPA to conclude "that the waste waters of all organic pesticide chemicals can be treated or controlled to the levels documented in the Agency's data base." 43 Fed.Reg. 17777 (1978).
In carbon adsorption, the organic molecules reach the surface of the carbon where they diffuse into the carbon's porous structure and bind with the carbon. The molecular structure and solubility of the pesticide influence its adsorption characteristics. Periodically the carbon becomes saturated and requires regeneration which can be accomplished by incineration. The resulting flue gases are quenched, scrubbed, and discharged to the atmosphere. The carbon can then be reused, although its adsorptive capacity will be somewhat reduced.
Hydrolysis is a chemical process in which the pesticide compounds are broken down by adding a caustic (or an acid) to the waste water setting off a reaction with the water and the organic compounds. The hydroxyl or hydrogen ions attach to some part of the pesticide chemical molecule, either displacing part of the group or breaking a bond so that two or more new compounds form. The success of hydrolysis is a function of the temperature and pH (relative acidity or alkalinity) of the solution and the properties of the chemical. Some compounds hydrolyze relatively easily compared to others. The ability of a compound to be hydrolyzed can be measured in terms of the half-life of the reaction, the time it takes at a given temperature and pH to hydrolyze half the chemical present. The shorter the half-life, the easier it is to destroy the pesticide through hydrolysis.
The regulations do not require the use of any particular treatment technology so long as the effluent limitations are met. One of the premises of the regulations, however, is that by using either carbon adsorption or hydrolysis, together with equalization and biological treatment, any plant could meet the regulations. Petitioners argue that there is no basis for this conclusion and that, therefore, we should remand the regulations as to the organic pesticides category. They suggest that the data collected at
We emphasize again that our review of agency rule-making is very limited, especially where the Agency must overcome technological and scientific uncertainty in making its delegated discretionary decisions. See Weyerhaeuser, at 1025. We will not remand so long as the Agency has explained the facts and policies on which it relied; the facts have some basis in the record; and a reasonable person could make the judgment the Agency made. Id. at 1026. Thus, the petitioners carry an extremely heavy burden when petitioning for review in such a case as this.
EPA identified eight full-scale carbon treatment systems used to reduce pesticides and five full-scale hydrolysis systems. The Agency sought information on all these facilities. One of the plants using carbon adsorption (one of the petitioners before us) disclosed neither the operating conditions of the system nor individual analyses of its effectiveness. Information on the remaining 12 systems is presented in great detail in section VII of the Development Document. Five of the seven carbon systems achieve removal rates in excess of 99 per cent of the influent pesticide as do two of the three hydrolysis systems for which data is available as to the pesticide content of both the influent and effluent of the treatment system. The two hydrolysis systems, for which only effluent data is available reduced the pesticide level to less than 1 mg/1 (one part per million) and to below the detection point respectively. EPA has concluded that the two substandard carbon systems and one sub-standard hydrolysis system could all achieve much better results by adjusting the operating conditions of the facilities (replacing the carbon more often or holding the waste water in the treatment system for a longer period).
Petitioners attack the use of data from the carbon systems because the systems were not designed to remove pesticides. This attack is trivial. EPA faces a severe problem in regulating the pesticide industry because a great deal is not known about treatment of pesticide waste waters and because the industry is very reticent about revealing what it does (or could) know. Given this admitted information shortage EPA must make use of the information it has, recognizing the limits of the information; EPA cannot refuse to carry out its mandate, waiting for the day when it might possess perfect information.
The same answer applies to petitioners' claim that the data on carbon treatment systems were too sparse to support
Petitioners suggest that even the limited data available at three of the carbon treatment plants are unrepresentative because the data were collected too soon after replacement of the carbon columns or because some wastes bypassed the treatment system being studied. The sampling at Eli Lilly covered the first 4 days of a 7 day carbon cycle. The record does not reveal why the carbon was changed just before sampling began or whether either ESE or Lilly timed the change purposely. The data, however, covered more than half the cycle, and the record reveals that the results obtained during the ESE sampling period were, if anything, unrepresentative on the high side since they showed a higher concentration of pesticide in the treatment effluent than the company found in its own studies. C.App. 217. The sampling at Hardwicke was done early in that company's normal 30 day carbon cycle, but, as EPA noted, the carbon should be changed far more frequently and the sampling was fairly representative of what would have been a reasonable carbon cycle for the plant.
Petitioners also argue that Lilly and Olin generated waste streams containing pesticides that did not pass through the treatment systems. Their citations with regard to Lilly are unconvincing. The direct communication from Lilly says nothing about a pesticide-bearing waste stream by-passing carbon treatment. The MITRE study does not support the claim. It says that the only two waste sources containing the pesticide may be combined prior to treatment. C.App. 206. We find no indication that the "Floor drains, Cooling waters, and Miscellaneous Waste Streams" contained any pesticide. C.App. 207.
The problem with respect to Olin is not so easily disposed of. On March 15, 1977, Olin submitted its comments on the interim final regulations. It stated that EPA's data omitted "supplementary waste streams containing some process waste waters." C.App. 213A. An attachment to these comments described the supplementary waste streams as including "among other things, spills, wash-downs, vent scrubber effluents, and surface runoff. In this wastestream, the following waste by-product streams are not included: 1. PCNB Plant Hydrochloric Acid; 2. PCNB Plant Spent Sulfuric Acid; 3. TCAN Plant Hydrocholoric Acid." C.App. 213A. Finally, this letter stated as "basically an intelligent guess" that 1.3 pounds of pesticide were in the supplementary waste stream (not including "TCAN, HCL, PCNB HCL and PCNB spent nitration acid."). EPA visited Olin's plant on August 17, 1977. According to a memorandum of that visit, "[p]lant personnel were unable to provide back-up information supporting the revised waste load estimates reported to EPA in a letter from Olin dated March 15, 1977. They agreed to provide the basis of these revised waste loads upon receipt of a letter from EPA." C.App. 304. On August 18, 1977, EPA sent a letter to Olin seeking, inter alia, more information about the supplementary waste streams; confirmation that "these supplementary waste streams have not been, and are not currently being treated by activated carbon, but rather are neutralized and discharged"; and confirmation that the existing "waste
EPA's response to petitioners' argument is that "[t]he truth as acknowledged by Olin, is that there are no [unaccounted for] streams." EPA brief at 62. EPA's brief-writers misread the exchange quoted above. They interpreted Olin's acknowledgement that the three acid by-product streams were not being discharged as meaning that no supplementary waste streams were being discharged. A clear distinction was made at all times in the exchange of correspondence between the acid "by-product waste streams" and the pesticide-bearing supplementary waste stream. The latter is clearly the one of concern, and Olin clearly stated that it was being discharged with no treatment other than neutralization.
The implication is that Olin was discharging more pounds of pesticide per thousand pounds of production than the figure EPA used in calculating the limits, and, therefore, the limits were miscalculated and should be corrected.
Similar arguments are directed against plants EPA used to illustrate hydrolysis treatment. Petitioners object to reliance on two plants claiming that in fact neither has an hydrolysis system. In fact the record reveals that both plants hydrolyze the wastes, albeit not in separate hydrolysis basins. Kerr-McGee adds a caustic to elevate the pH above 11 and elevates the temperature to facilitate the reaction. App. 991. Monsanto's Anniston plant similarly sets off a reaction with addition of caustic to raise the pH, App. 2459, though the temperature is not raised. App. 126-27. In effect, hydrolysis and biological treatment occur simultaneously. Although there may be some problem in attributing the destruction of the pesticide to one or the other cause, both treatments are recommended, and the success of these systems is relevant to whether companies can meet the guideline limits
Having decided to use a single category, EPA had to set the limits for it. Its decision to use only data from full-scale operating systems and to ignore zero-discharge plants was well within its discretion. Thus it needed a way to express the limitation more precisely than a requirement such as that all plants "do as well as the plants we have surveyed". It needed to express the results of the survey data as a number typical of the model group — that is, an average. Averaging the effluent data from the model plants is very different from averaging apples and oranges or bananas and marshmallows. If the plants performed identically, no average would be necessary. Though the plants performed differently, even very differently, we cannot see why averaging is unacceptable. The same parameter was measured in the same way and in the same units. The set of observations thus derived made up the population of which an average was desired. The problem, as we see it, was not whether an average could be taken, but how to derive that average.
EPA chose to calculate a weighted average that gave equal influence to each observation rather than to each plant. Thus the plant with 4 observations had far less influence on the final average than the plant with 450 observations, but each observation at the two plants had exactly the same influence. We are sure there were many alternatives to this system, but we cannot say that EPA went beyond its discretion in picking this one. We agree with the Fourth Circuit that "the choice of statistical methods is a matter best left to the sound discretion of the Administrator." FMC Corp. v. Train, 539 F.2d 973, 986 (4th Cir. 1976); American Petroleum Institute v. EPA, 540 F.2d 1023, 1035 (10th Cir. 1976). The choice of any given method may mean that an alternative method would yield different results. The necessary corollary, however, is that any other system chosen would be open to the same criticism. We will not leave the Agency so vulnerable.
Even if the plants studied demonstrate that carbon or hydrolysis treatment is effective for the pesticides they produce, petitioners point out that they produce only a small minority of the 49 pesticides whose discharge is controlled. This is true, but it hardly means that the regulations must be limited to those pesticides for which EPA has collected information on operating treatment systems or even to the larger group of pesticides currently being treated
D. EPA's Consideration of Cost
The Act requires EPA's assessment of best practicable control technology currently available to "include consideration of the total cost of application of technology in relation to the effluent reduction benefits to be achieved." 33 U.S.C. § 1314(b)(1)(B). Cost, however, is not a paramount consideration. Congress "self-consciously made the legislative determination that the health and safety gains that achievement of the Act's aspirations would bring to future generations will in some cases outweigh the economic dislocation it causes to the present generation." Weyerhaeuser, at 1037. The obligation the Act imposes on EPA is only to perform a limited cost-benefit balancing to make sure that costs are not "wholly out of proportion" to the benefits achieved. A Legislative History of the Water Pollution Control Act Amendments of 1972 170 (1973) (statement of Senator Muskie); Weyerhaeuser, at 1045 n. 52. Thus, the balancing is a relatively subsidiary task and need not be precise.
The Agency based these estimates on model treatment techniques assuming a variety of plant sizes and treatment difficulties. The petitioners criticize the estimates and their application to plants on a variety of grounds. Though many of the criticisms may be valid, suggesting that the estimates are not perfectly accurate, none of them convince us that EPA's evaluation of costs must be rejected. There is no way the cost analysis could be more than an estimate, especially given the reticence of the industry to supply information, and EPA needed to develop no more than a rough idea of the costs the industry would incur. On the basis of its estimates, EPA decided to promulgate these regulations. We cannot say that the Agency has failed to consider costs in relation to benefits, and we find no basis on which to overrule EPA and decide that costs are wholly out of proportion to benefits.
II. Metallo-Organic Pesticide Manufacturing
Subpart B of the regulations is "applicable to discharges resulting from the manufacture of metallo-organic active ingredients containing mercury, cadmium, arsenic, or copper." 40 C.F.R. § 455.30; 43 Fed.Reg. 17786-87 (1978).
Diamond Shamrock has twice in the past reported that its arsenic manufacturing process had no discharge of polluted waste water. In August, 1975, it reported to EPA that its "process has a negative water balance [and is] able to utilize essentially all, including rainwater from the process area, in the product." A 1972 report stated, "The arsenical unit has no discharge — its design is such as to recycle all waters into the final product." Other reports, however, are either
Though the record is far from clear, it indicates to us that, while Diamond Shamrock may once have been a non-discharger, at least as of the time these regulations were promulgated the plant was discharging pollutants and had so advised EPA. This does not, however, end the inquiry. The fact remains that the other eight manufacturers in the metallo-organic subpart, including the two other producers of arsenic pesticides, achieve zero discharge and have not joined the petition for review,
Diamond Shamrock's attempts to refute EPA's evidence concerning the other plants are unpersuasive. The evidence consists of notes of phone conversations between ESE and officials at the other two arsenic plants, Vineland and Ansul. According to the notes, Vineland's chief chemist claimed all waste water is recycled to the process, and the Ansul official said, "All wastewater from the pesticide production is reused in the process; the only discharge is of non-contaminated cooling water." Though transcribed notes of phone calls are not the best possible evidence, it seems that EPA confronted stern industry resistance in its attempts to put together a more reliable data base. Industry cannot be allowed to profit from the success of its resistance to regulatory inspections. Moreover, in proceedings such as these EPA has to be able to rely on industry-supplied information, especially where dishonesty would be counter-productive for the informants.
It appears that the industry can and does produce arsenic-based pesticides without discharging polluted waste water. The statute, though, requires more; EPA must consider specific factors including "the total cost of application of technology in relation to the effluent reduction benefits to be achieved ..., and ... the process employed, the engineering aspects of the application of various types of control techniques, process changes, [and] non-water quality environmental impact ..." 33 U.S.C. § 1314(b)(1)(B). EPA's excuse for not considering these factors is that Diamond Shamrock does not now, or at least did not in the past, discharge polluted waste
The record as we construe it does not support EPA's reasoning. Diamond Shamrock is now a discharger. It will have to change something, whether in its production process or its treatment process, in order to meet the regulation. EPA must give some consideration to the cost and impact, if any, of implementing these changes and compare the cost to the reduction benefits. Until it has performed that task, EPA has not discharged its statutory responsibilities. Having failed to consider the relevant factors, EPA abused its discretion, and we must remand.
As indicated in our discussion of the required cost analysis above, the consideration may be general, and EPA has considerable discretion to determine the scope of the investigation. We also emphasize that our remand on this issue is a limited one to give EPA an opportunity to perform the statutorily mandated assignment. If EPA discovers that in fact Diamond Shamrock will not incur costs or that the benefits of the regulation are not wholly out of proportion to the costs to the industry, then we will be prepared to uphold the regulations as written.
III. Formulators and Packagers Subcategory
Subpart C of the regulations apply to formulators and packagers of pesticides. 40 C.F.R. §§ 455.40-455.42. This segment of the industry, encompassing approximately 5300 companies, mixes technical grade pesticide chemicals with inert ingredients by dry-based, solvent-based, or liquid-based processes to produce usable commercial products. These companies are required to release "no discharge of process waste water pollutants to navigable waters". 40 C.F.R. § 455.42. EPA expects them to meet this limit by a combination of in-process controls and collection and evaporation of whatever waste water cannot be avoided. NACA, on behalf of the formulators and packagers, has intervened in these petitions for review to challenge the no discharge limitation on a number of grounds. NACA argues that EPA's data base is so inconsistent, inaccurate, and incomplete that it cannot support any conclusion and that even taking the data base at face value it does not support the conclusion EPA has reached. Further, NACA argues that EPA has failed to meet the statutory mandate to consider certain factors. 33 U.S.C. § 1314(b)(1)(B).
We agree with NACA that EPA's treatment of this industry segment is not a model of administrative regulatory methodology, but, nonetheless, we find sufficient reliable data in the record to support EPA's conclusion that zero discharge is the best practicable control technology currently available for this category. We rest this conclusion on two sources of data with respect to the reliability and value of which EPA and NACA nearly agree.
These results are confirmed by the second source, a report compiled by a contractor hired by NACA and submitted to EPA by NACA. The report's conclusions included the following:
This study contacted 105 companies, getting 91 usable responses. Of these 91, 66 discharged no waste water. Only 5 discharged directly to navigable waters, and another 20 discharged to private treatment facilities or practiced deep well injection or ocean dumping. Thus 73 per cent of the companies have no discharge at all, and 95 per cent comply with these regulations. The results are virtually identical to ESE's results. Whatever the failings of these studies, the numbers clearly show that the great majority of formulators already meet the limitations, and it follows that meeting the regulations is practicable for the industry looked at as a whole.
NACA does not seriously dispute that conclusion. Rather NACA suggests that we should not rely on the numbers at all and that the category should not be looked at as a whole. NACA attacks the methodology of the ESE survey because two different survey forms were used, those were not completely filled out, the results have been inconsistently reported, and the survey failed to generate any information on the effectiveness of any pre-discharge waste water treatment systems. Neither we nor EPA rely on this survey for anything beyond the information that was collected — the company's formulating process and whether or not it generated or discharged process waste water. That other questions were not asked or answered is for this purpose irrelevant. Similarly, that the survey results have been inconsistently reported is of no matter so long as the final tabulation is accurate. NACA challenges the survey it submitted because that survey was not designed to serve the purpose for which EPA used it, was hastily done, and left many questions unanswered. Again, our use of it, like EPA's, is only for the results it did collect, not for whatever it may have failed to determine.
The ESE survey collected information on the type of process used as well as waste generation and discharge.
Given EPA's supported finding on the basis of the survey that the industry in large part achieved zero discharge, EPA did not need to go through the further steps of designating some number of these plants "exemplary" and averaging the zero discharge reports of the exemplary plants to set a zero discharge limit for the subcategory. To be sure, it might have been well for EPA to visit some of the facilities reporting zero discharge in order to verify the reports, but at some point EPA must be able to rely on information provided by industry, and here the companies had no incentive to report discharges below actual levels, thereby inviting regulations they could not meet.
We turn now to the factors the statute mandates that EPA consider. In particular, NACA argues that EPA has failed adequately to consider "non-water quality environmental impact" and "the total cost of application of technology in relation to the effluent reduction benefits to be achieved." 33 U.S.C. § 1314(b)(1)(B).
As we have said, the cost/benefit comparison mandated by Congress is "intended to limit the application of technology only where the additional degree of effluent reduction is wholly out of proportion to the costs of achieving such marginal level of reduction for any class or category of sources." Legislative History, supra, at 170; Weyerhaeuser, at 1045 & n. 52. Congress did not intend cost to be an unwieldy barrier to attainment of pollution control goals. We have already explained the nature of our review.
EPA devoted significant effort to discovering the economic impact of these regulations on the formulating and packaging segment of the industry. Not only did EPA attempt to find out the number of companies that would have to incur new costs to comply with the regulations, but it also prepared estimates of the cost of compliance on the assumption that small, medium, and large plants would have to build evaporation treatment systems from scratch. NACA attacks these cost estimates on two fronts. The first is that EPA failed to consider adequately or accurately the cost of hauling waste water or sludge remaining after evaporation to a landfill. EPA's estimates are only for evaporation systems, not the alternative acceptable technology of hauling all waste water to a landfill. But evaporation is the more expensive of the two alternatives.
NACA's second attack is that EPA used unrealistic estimates of evaporation rates. EPA used the median evaporation rate for the country in developing its estimates. NACA recognizes this by saying that the figure used "overestimates the annual evaporation rate for approximately half the nation" and "over half the pesticide formulation plants". EPA is entitled to look at costs on an industry-wide basis as opposed to plant-by-plant. To do so EPA must be able to make some assumption about the conditions facing the industry. Certainly, using an appropriately chosen average figure is a legitimate statistical technique, especially given that EPA is doing no more than developing rough estimates to help it determine whether the cost "is wholly out of proportion" to the benefit. Moreover, EPA notes that the evaporation tank volumes allotted are more than ample to accommodate the very slow build up of waste water that might accumulate where evaporation is significantly less than the national average. This build up can be hauled periodically to landfills.
Given EPA's determination that the cost of installing and operating an evaporation system would be acceptable even for a plant that starts with no system at all and that most formulators already comply with the regulations, EPA gave ample consideration to the comparison of cost with benefit. To be sure, the record contained only scanty data on the quantity of pollutants in formulators' waste water.
The Agency must also "take into account ... non-water quality environmental impact". The District of Columbia Circuit has held that this means the Agency must inform itself of the magnitude of potential problems and reach an "express and considered conclusion about their bearing". Weyerhaeuser, at 1045. The Agency has done so, though in a most cursory way. 43 Fed.Reg. 17776, 17780 (1978). The Agency noted that the major problem would be sludge and concentrated waste disposal. Though there is no specific finding of the magnitude of the problem and no express statement of conclusions, it is clear that EPA considered the problems and decided they were not of sufficient magnitude to prevent promulgation of these regulations. The Agency noted that it has published guidelines on solid waste disposal, 40 C.F.R. Part 241, and that it is in the process of developing regulations under the Resource
Finally, NACA challenges EPA's decision that plants which both manufacture and formulate pesticides must meet the limits for manufacturing plants with no credit for any discharge from the formulation process. NACA suggests that this will force combination plants to separate their waste streams because if combined the manufacturer will have to treat the waste water to levels better than those attained by BPT. As EPA points out though, most formulation processes do not generate or discharge waste water, including many at plants that also manufacture pesticides. Also, waste flows from formulation, where they exist, will be very small compared to the flow from manufacturing. As a consequence any burden on combined plants choosing to treat wastes together should be minimal. If that assumption proves wrong, of course, the plant retains the option of keeping the formulation waste separate and evaporating it or hauling it to a landfill. It seems eminently reasonable to us that if formulation plants are required to attain zero discharge, they should not be able to discharge extra pollutants simply because they happen also to manufacture pesticides.
Because of the minor problems noted above, the regulations must be remanded. We have no reason to believe that either of the reasons for this remand — the misunderstanding concerning one company's reporting of the waste content of its discharges and the Agency's failure to consider the possible costs of implementing the regulations in the metallo-organic subcategory — will pose a serious stumbling block to final approval of the regulations. Accordingly we will retain jurisdiction over these consolidated petitions for review so that we can bring a speedy conclusion to this case once the Agency has set the record straight and made whatever corrections it deems necessary.
EPA now advises us that there is not any on-going investigation.