Angus Macbeth was on the brief for petitioner Chemical Waste Management, Inc., in No. 87-1459.

William R. Weissman and Douglas H. Green (for Edison Electric Institute, et al.) and G. William Frick and Thomas S. Llewellyn, Washington, D.C., (for American Petroleum Institute) were on the joint brief for intervenors.

Jacqueline M. Warren entered an appearance for petitioner Natural Resources Defense Council, Inc., in No. 86-1677.

Robert F. Van Voorhees and Mark B. Halverson, Washington, D.C., entered appearances for petitioner Chemetco, Inc., in No. 87-1057.

James K. Jackson, Washington, D.C., and Ralph J. Colleli, Jr., entered appearances for intervenor American Petroleum Institute in No. 86-1677. Arnold S. Block, Philadelphia, Pa., entered an appearance for intervenor American Petroleum Institute in Nos. 87-1016 and 87-1459.

Before WALD, Chief Judge, SILBERMAN and D.H. GINSBURG, Circuit Judges.

Opinion PER CURIAM.

Opinion concurring in part and concurring in the result filed by Circuit Judge SILBERMAN.

In 1984, Congress amended the Resource Conservation and Recovery Act ("RCRA"), 42 U.S.C. Sec. 6921-6991 (1982 & Supp. IV 1986), to prohibit land disposal of certain hazardous solvents and wastes containing dioxins except in narrow circumstances to be defined by Environmental Protection Agency ("EPA") regulations. See Hazardous and Solid Waste Amendments, Sec. 201(a), 42 U.S.C. Sec. 6924(e) (Supp. IV 1986). In these consolidated cases, petitioners seek review of EPA's final "solvents and dioxins" rule published pursuant to Congress' 1984 mandate. We conclude that the rule under review is consistent with RCRA, but remand one aspect of the rulemaking to the agency for further explanation.

A. Statutory Scheme.

The Hazardous and Solid Waste Amendments of 1984 ("HSWA"), Pub.L. No. 98-616, 98 Stat. 3221 (1984), inter alia, substantially strengthened EPA's control over the land disposal of hazardous wastes regulated under RCRA's "cradle to grave" statutory scheme. In preambular language to the HSWA, Congress, believing that "land disposal facilities were not capable of assuring long-term containment of certain hazardous wastes," expressed the policy that "reliance on land disposal should be minimized or eliminated." 42 U.S.C. Sec. 6901(b)(7). In order to effectuate this policy, HSWA amended section 3004 of RCRA to prohibit land disposal of hazardous waste unless the waste is "pretreated" in a manner that minimizes "short-term and long-term threats to human health and the environment," id. Sec. 6924(m), or unless EPA can determine that the waste is to be disposed of in such a fashion as to ensure that "there will be no migration of hazardous constituents from the disposal [facility]...." Id. Sec. 6924(d)(1), (e)(1), & (g)(5).

As amended, RCRA requires EPA to implement the land disposal prohibition in three phases, addressing the most hazardous "listed" wastes first. See id. Sec. 6924(g).1 In accordance with strict statutory deadlines, the Administrator is obligated to specify those methods of land disposal of each listed hazardous waste which "will be protective of human health and the environment." Id. In addition, "[s]imultaneously with the promulgation of regulations ... prohibiting ... land disposal of a particular hazardous waste, the Administrator" is required to

promulgate regulations specifying those levels or methods of treatment, if any, which substantially diminish the toxicity of the waste or substantially reduce the likelihood of migration of hazardous constitutents from the waste so that short-term and long-term threats to human health and the environment are minimized.

Id. Sec. 6924(m).

Respecting two categories of hazardous wastes, including the solvents and dioxins at issue here2 Congress, however, declined to wait for phased EPA implementation of the land disposal prohibition. For these wastes, Congress imposed earlier restrictions, prohibiting land disposal after dates specified in the HSWA except in accordance with pretreatment standards or pursuant to regulations specifying "protective" methods of disposal. Id. Sec. 6924(e)(1). These prohibitions, as applied to the solvents and dioxins listed in the HSWA, were to take effect November 8, 1986. Id.

In order to further RCRA's basic purpose of mandating treatment of hazardous wastes in lieu of land disposal, Congress further provided that storage of wastes falling within the land disposal prohibition would be "prohibited unless such storage is solely for the purpose of the accumulation of such quantities of hazardous waste as are necessary to facilitate proper recovery, treatment or disposal." Id. Sec. 6924(j). Congress believed that permitting storage of large quantities of waste as a means of forestalling required treatment would involve health threats equally serious to those posed by land disposal, and therefore opted in large part for a "treat as you go" regulatory regime.

B. The Rulemaking Under Review.

In January 1986, EPA issued a notice of proposed rule-making announcing its draft implementation of the land disposal prohibition for solvents and dioxins. See 51 Fed.Reg. 1602 (1986) (hereinafter "Proposed Rule"). Approximately ten months later, after receiving extensive public commentary on the draft blueprint, EPA published a final solvents and dioxins rule differing in some respects from its draft approach. See 51 Fed.Reg. 40,572 (1986) (hereinafter "Final Rule"). These differences were especially striking in EPA's implementation of section 3004(j) and section 3004(m) of RCRA, governing the storage prohibition and treatment standards, respectively, for solvents and dioxins. These portions of the rule, together with other discrete portions of the rulemaking faulted by petitioners, are summarized below.

1. Section 3004(m) Treatment Standards.

In the Proposed Rule, EPA announced its tentative support for a treatment regime embodying both risk-based and technology-based standards. The technology-based standards would be founded upon what EPA determined to be the Best Demonstrated Available Technology ("BDAT"); parallel risk-based or "screening" levels were to reflect "the maximum concentration [of a hazardous constituent] below which the Agency believes there is no regulatory concern for the land disposal program and which is protective of human health and the environment." Proposed Rule at 1611. The Proposed Rule provided that these two sets of standards would be melded in the following manner:

First, if BDAT standards were more rigorous than the relevant health-screening levels, the latter would be used to "cap the reductions in toxicity and/or mobility that otherwise would result from the application of BDAT treatment[.]" Id. Thus, "treatment for treatment's sake" would be avoided. Second, if BDAT standards were less rigorous than health-screening levels, BDAT standards would govern and the screening level would be used as "a goal for future changes to the treatment standards as new and more efficient treatment technologies become available." Id. at 1612. Finally, when EPA determined that the use of BDAT would pose a greater risk to human health and the environment than land disposal, or would provide insufficient safeguards against the threats produced by land disposal, the screening level would actually become the 3004(m) treatment standard. Id.

EPA invited public comment on alternative approaches as well. The first alternative identified in the Proposed Rule (and the one ultimately selected by EPA) was based purely on the capabilities of the "best demonstrated available technology." Id. at 1613. Capping treatment levels to avoid treatment for treatment's sake, according to EPA, could be accomplished under this technology-based scheme by "the petition process":

Under this approach, if a prescribed level or method of treatment under section 3004(m) resulted in concentration levels that an owner/operator believed to be overly protective, the owner/operator could petition the Agency to allow the use of an alternative treatment level or method or no treatment at all by demonstrating that less treatment would still meet the petition standard of protecting human health and environment.

Id. at 1613. And the function served by health-screening levels of providing a default standard when the application of BDAT technology would itself pose a threat to human health and the environment could likewise be fulfilled by the petition process: "an owner operator could [ ] petition the Agency ... to allow continued land disposal of the waste upon a demonstration that land disposal of the waste would not result in harm to human health and the environment." Id.

The Agency received comments supporting both approaches, but ultimately settled on the pure-technology alternative. Of particular importance to EPA's decision were the comments filed by eleven members of Congress, all of whom served as conferees on the 1984 RCRA amendments. As EPA recorded in the preamble to the Final Rule:

[these] members of Congress argued strongly that [the health screening] approach did not fulfill the intent of the law. They asserted that because of the scientific uncertainty inherent in risk-based decisions, Congress expressly directed the Agency to set treatment standards based on the capabilities of existing technology.

The Agency believes that the technology-based approach adopted in [the] final rule, although not the only approach allowable under the law, best responds to the above stated comments.

Final Rule at 40,578.

EPA also relied on passages in the legislative history supporting an approach under which owners and operator of hazardous waste facilities would be required to use " 'the best [technology] that has been demonstrated to be achievable.' " Id. (quoting 103 CONG.REC. S9178 (daily ed. July 25, 1984) (statement of Senator Chaffee). And the agency reiterated that the chief advantage offered by the health-screening approach--avoiding "treatment for treatment's sake"--could "be better addressed through changes in other aspects of its regulatory program." Id. As an example of what parts of the program might be altered, EPA announced that it was "considering the use of its risk-based methodologies to characterize wastes as hazardous pursuant to section 3001 [of RCRA]." Id.; see 42 U.S.C. Sec. 6921 (1982 & Supp. IV 1986).3

Petitioner CMA challenges this aspect of the rule as an unreasonable construction of section 3004(m)'s mandate to ensure that "short-term and long-term threats to human health and the environment are minimized." 42 U.S.C. Sec. 6924(m) (1982 & Supp. IV 1986). In the alternative, CMA argues that EPA has failed to explain the basis--in terms of relevant human health and environmental considerations--for its BDAT regime, which allegedly requires treatment in some circumstances to levels far below the standards for human exposure under other statutes administered by EPA. Thus, CMA claims that EPA's action in promulgating a technology-based rule is arbitrary and capricious.

2. Section 3004(j) Storage Prohibition.

Section 3004(j) of RCRA, as noted above, prohibits the storage of wastes falling within a land disposal prohibition "unless such storage is solely for the purpose of the accumulation of such quantities of hazardous waste as are necessary to facilitate proper recovery, treatment, or disposal," 42 U.S.C. Sec. 6924(j) (1982 & Supp. IV 1986). In the Proposed Rule, EPA tentatively implemented this provision to allow generators to accumulate hazardous wastes on-site for up to 90 days, no questions asked. EPA selected this period in the belief "that it would allow a reasonable period for accumulation prior to further management without interfering with a generator's production process[.]" Proposed Rule at 1709. It observed that as a matter of prevailing industrial practice "most wastes were removed from the site of generation within 90 days." Id. Nevertheless, out of concern that "a longer time may, in some cases, be necessary to accumulate sufficient quantities to facilitate proper recovery, treatment, or disposal," id., the agency solicited comments on alternative storage periods that might be appropriate.

The comments received by the Agency ranged far and wide, but all found the 90-day period inadequate. A majority of the commentors favored a one-year storage period in order to accommodate small-quantity generators and others whose waste streams "accumulate[ ] more slowly than others." Final Rule at 40,582. On the basis of these remarks, EPA agreed that 90 days was an insufficient period for the adequate accumulation of wastes to facilitate recovery, treatment or disposal.

EPA ultimately settled on a one-year storage period, but the implementing regulation differed significantly in character from the 90-day proposal. The Final Rule provides:

An owner/operator of a treatment facility may store [ ] wastes for up to one year unless the Agency can demonstrate that such storage was not solely for the purpose of accumulation of such quantities of hazardous waste as are necessary to facilitate proper recovery, treatment, or disposal.

Id. at 40,643 (emphasis added) (codified at 40 C.F.R. Sec. 268.50(b) (1988)). A companion provision requires owners and operators to bear the burden of proving that storage for over a one-year period was for proper purposes under RCRA. See id. (codified at 40 C.F.R. Sec. 268.50(c) (1988)).

Characterizing the final storage rule, in effect, as a "shifting of the statutory burden of proof" which "effectively allows a one year override of the statutory prohibition" against storage, petitioners Hazardous Waste Treatment Council ("HWTC") and the Natural Resources Defense Council ("NRDC") challenge the rule as inconsistent with section 3004(j) of RCRA.

3. Responsibility for Testing Wastes Prior to Disposal.

A determination as to whether and to what degree treatment of a waste is required prior to land disposal depends upon the concentration of hazardous constituents in the waste. To facilitate these determinations and to ensure compliance with the land disposal prohibitions and applicable treatment standards, EPA proposed to implement requirements for mandatory testing in some circumstances. Proposed Rule at 1691.

The Agency was immediately confronted with the question of who, among generators, treatment facilities and land disposal facilities, should shoulder the responsibility of testing the waste prior to disposal. While several alternatives were available, EPA initially proposed that the land disposal facility alone be responsible for such testing. Proposed Rule at 1692.

Under this approach, the disposal facility must either conduct an analysis of the waste or obtain an analysis of the waste from the generator or treater. Similarly, the owner or operator of a land disposal facility could arrange for the generator or treatment facility to supply all or part of the required testing data. However, if the generator or treater did not supply the testing data and the land disposal facility owner or operator chose to accept the waste, the owner or operator would be responsible for conducting the required testing.

Id. at 1691. The agency cautioned that this approach did not leave the generator without responsibility altogether. The generator was still obliged to determine "whether he must treat his waste prior to disposal." Id. "[R]ather than specifically requiring the generator to conduct testing, [however], the Agency [proposed to] allow determination of whether wastes meet the regulatory thresholds to be based on either testing or knowledge of the characteristics of the waste." Id. The Agency found this proposal desirable because "[i]t is flexible, does not require redundant testing, fits into the current regulatory scheme for the waste analysis plan and requires the testing to take place where the liability for disposal exists--at the land disposal facility." Id. at 1691.

The Final Rule bears substantial resemblance to that initially proposed by EPA, with one principal exception. As the agency explained, "[b]ecause the [treatment] approach promulgated [in the Final Rule] does not cap BDAT with screening levels, more wastes will require treatment to meet the specified treatment standards." Final Rule at 40,597. Given this expanded role for the treatment industry, EPA decided in the Final Rule to impose testing requirements on both treatment facilities and land disposal facilities. But, the agency followed the proposed rule insofar as it did not require testing by generators. See id. Thus, when sending waste to either a treatment facility for pretreatment or directly to a land disposal facility, the Final Rule permits generators to base their determinations as to the concentration of hazardous constituents in the waste on "waste analysis data, knowledge of the waste, or both." Id. No matter what the basis for their determinations, generators forwarding wastes directly to land disposal facilities must certify their conclusions to the facilities' operators. False certifications, under the Final Rule, may result in criminal penalties. See 40 C.F.R. Sec. 268.7 (1988); see also 42 U.S.C. Sec. 6928(d)(3) (Supp. IV 1986).

Petitioners HWTC and NRDC contend that it is arbitrary and capricious for EPA to require operators of treatment and land disposal facilities, but not generators, to test wastes within the land disposal prohibition.

CMA challenges EPA's adoption of BDAT treatment standards in preference to the approach it proposed initially primarily on the ground that the regulation is not a reasonable interpretation of the statute. CMA obliquely, and Intervenors Edison Electric and the American Petroleum Institute explicitly, argues in the alternative that the agency did not adequately explain its decision to take the course that it did. We conclude, as to CMA's primary challenge, that EPA's decision to reject the use of screening levels is a reasonable interpretation of the statute. We also find, however, that EPA's justification of its choice is so fatally flawed that we cannot, in conscience, affirm it. We therefore grant the petitions for review to the extent of remanding this issue to the agency for a fuller explanation.

A. The Consistency of EPA's Interpretation with RCRA.

Our role in evaluating an agency's interpretation of its enabling statute is as strictly circumscribed as it is simply stated: We first examine the statute to ascertain whether it clearly forecloses the course that the agency has taken; if it is ambiguous with respect to that question, we go on to determine whether the agency's interpretation is a reasonable resolution of the ambiguity. Chevron v. Natural Resources Defense Council, 467 U.S. 837, 842-45, 104 S. Ct. 2778, 2781-83, 81 L. Ed. 2d 694 (1984).

1. Chevron Step I: Is the Statute Clear?

We repeat the mandate of Sec. 3004(m)(1): the Administrator is required to promulgate "regulations specifying those levels or methods of treatment, if any, which substantially diminish the toxicity of the waste or substantially reduce the likelihood of migration of hazardous constitutents from the waste so that short-term and long-term threats to human health and the environment are minimized." 42 U.S.C. Sec. 6924(m)(1).

CMA reads the statute as requiring EPA to determine the levels of concentration in waste at which the various solvents here at issue are "safe" and to use those "screening levels" as floors below which treatment would not be required. CMA supports its interpretation with the observation that the statute directs EPA to set standards only to the extent that "threats to human health and the environment are minimized." We are unpersuaded, however, that Congress intended to compel EPA to rely upon screening levels in preference to the levels achievable by BDAT.

The statute directs EPA to set treatment standards based upon either "levels or methods" of treatment. Such a mandate makes clear that the choice whether to use "levels" (screening levels) or "methods" (BDAT) lies within the informed discretion of the agency, as long as the result is "that short-term and long-term threats to human health and the environment are minimized." To "minimize" something is, to quote the Oxford English Dictionary, to "reduce [it] to the smallest possible amount, extent, or degree." But Congress recognized, in the very amendments here at issue, that there are "long-term uncertainties associated with land disposal," 42 U.S.C. Sec. 6924(d)(1)(A). In the face of such uncertainties, it cannot be said that a statute that requires that threats be minimized unambigously requires EPA to set levels at which it is conclusively presumed that no threat to health or the environment exists.

Nor are we at all persuaded by CMA's interpretation of NRDC v. EPA, 824 F.2d 1146, 1163 (D.C.Cir.1987) (en banc ), in which we held that EPA was not permitted to "substitute[ ] technological feasibility for health as the primary consideration under Section 112 [of the Clean Air Act]." That provision requires the Administrator to set air pollution standards "at the level which in his judgment provides an ample margin of safety to protect the public health." 42 U.S.C. Sec. 7412(b)(1)(B). EPA had set emission standards for vinyl chloride, however, "based solely on the level attainable by the best available control technology," 824 F.2d at 1149, despite its finding that such levels would create health risks. It had neither stated that the risks it found were insignificant, nor explained how the risks it accepted were consistent with its statutory duty to provide "an ample margin of safety." Id. This court held that EPA had erred in failing to consider whether the best available technology was sufficient to provide the statutorily mandated margin of safety. Id. at 1164-66.

Contrary to CMA's implication, however, the court did not hold, or even imply, the converse--that EPA could not require generators to use technologies that would reduce emissions to a point below that which would provide an "ample margin of safety." Indeed, the court noted that "Congress ... recognized in section 112 that the determination of what is 'safe' will always be marked by scientific uncertainty and thus exhorted the Administration to set ... standards that will provide an 'ample margin' of safety," id. at 1165; we then concluded that "[o]nce 'safety' is assured, the Administrator should be free to diminish as much of the statistically determined risk as possible by setting the standard at the lowest feasible level." Id.

This is not to say that EPA is free, under Sec. 3004(m), to require generators to treat their waste beyond the point at which there is no "threat" to human health or to the environment. That Congress's concern in adopting Sec. 3004(m) was with health and the environment would necessarily make it unreasonable for EPA to promulgate treatment standards wholly without regard to whether there might be a threat to man or nature. That concern is better dealt with, however, at Chevron 's second step; for, having concluded that the statute does not unambiguously and in all circumstances foreclose EPA from adopting treatment levels based upon the levels achievable by BDAT, we must now explore whether the particular levels established by the regulations supply a reasonable resolution of the statutory ambiguity.

2. Chevron Step II: Is EPA's Interpretation Reasonable?

The screening levels that EPA initially proposed were not those at which the wastes were thought to be entirely safe. Rather, EPA set the levels to reduce risks from the solvents to an "acceptable" level, and it explored, at great length, the manifest (and manifold) uncertainties inherent in any attempt to specify "safe" concentration levels. The agency discussed, for example, the lack of any safe level of exposure to carcinogenic solvents, 51 Fed.Reg. at 1,628; the extent to which reference dose levels (from which it derived its screening levels) understate the dangers that hazardous solvents pose to particularly sensitive members of the population, id. at 1,627; the necessarily artificial assumptions that accompany any attempt to model the migration of hazardous wastes from a disposal site, id. at 1,642-53; and the lack of dependable data on the effects that solvents have on the liners that bound disposal facilities for the purpose of ensuring that the wastes disposed in a facility stay there, id. at 1,714-15. Indeed, several parties made voluminous comments on the Proposed Rule to the effect that EPA's estimates of the various probabilities were far more problematic than even EPA recognized. See, e.g., Comments of Natural Resources Defense Council, Record at 29,000-62.

CMA suggests, despite these uncertainties, that the adoption of a BDAT treatment regime would result in treatment to "below established levels of hazard." It relies for this proposition almost entirely upon a chart in which it contrasts the BDAT levels with (1) levels EPA has defined as "Maximum Contaminant Levels" (MCLs) under the Safe Drinking Water Act; (2) EPA's proposed "Organic Toxicity Characteristics," threshold levels below which EPA will not list a waste as hazardous by reason of its having in it a particular toxin; and (3) levels at which EPA has recently granted petitions by waste generators to "delist" a particular waste, that is, to remove it from the list of wastes that are deemed hazardous. CMA points out that the BDAT standards would require treatment to levels that are, in many cases, significantly below these "established levels of hazard."

If indeed EPA had determined that wastes at any of the three levels pointed to by CMA posed no threat to human health or the environment, we would have little hesitation in concluding that it was unreasonable for EPA to mandate treatment to substantially lower levels. In fact, however, none of the levels to which CMA compares the BDAT standards purports to establish a level at which safety is assured or "threats to human health and the environment are minimized." Each is a level established for a different purpose and under a different set of statutory criteria than concern us here; each is therefore irrelevant to the inquiry we undertake today.

The drinking water levels, for example, are established under a scheme requiring EPA to set "goals" at a level at which "no known or anticipated adverse effects on the health of persons occur." 42 U.S.C. Sec. 300g-1(b)(4). EPA is then to set MCLs as close to its goals as "feasible," taking into account, among other things, treatment costs. 42 U.S.C. Secs. 300g-1(b)(4), (5). Since SDWA goals are set only to deal with "known or anticipated" adverse health effects, a mere "threat" to human health is not enough in that context. Moreover, SDWA levels are set without reference to threats to the environment. Finally, EPA must consider costs in setting its MCLs; there is no similar limitation in Sec. 3004 of RCRA.

Similarly, in promulgating the OTC levels, EPA made clear that, "[i]n establishing a scientifically justifiable approach for arriving at [OTC levels], EPA wanted to assure a high degree of confidence that a waste which releases toxicants at concentrations above the [OTC level] would pose a hazard to human health." EPA Hazardous Waste Management System; Identification and Listing of Hazardous Waste ..., Proposed Rule, 51 Fed.Reg. 21,648, 21,649 (1986) (emphases added). Thus it is clear that wastes with toxicant levels below the OTC thresholds may still pose "threats to human health [or] the environment." Id. at 21,648 (emphases added).

Finally, CMA points to the "delisting levels" as appropriate points of comparison. The term is a bit misleading, however. EPA delists particular wastes in response to individual petitions, see, e.g., 42 U.S.C. Sec. 6921(f)(1), and it has not adopted formal, or even de facto, levels below which any waste will be delisted. That EPA has delisted, in particular circumstances, wastes containing concentrations of solvents higher than those called for by the BDAT standards adds nothing to CMA's argument. The treatment standards establish a generic approach, requiring that all wastes deemed to be hazardous be treated to a set level in order to minimize threats to health and to the environment. If a waste is listed as hazardous, and an individual generator wants to dispose of it without meeting the BDAT standards, it may petition to have its particular waste delisted. If the agency grants the delisting petition, only the petitioner is affected; the generally required level of treatment remains the same. Hence, there is no inconsistency between a "delisting level," accepted in particular circumstances, that permits a higher level of a particular contaminant then the BDAT level otherwise generally applicable.

In sum, EPA's catalog of the uncertainties inherent in the alternative approach using screening levels supports the reasonableness of its reliance upon BDAT instead. Accordingly, finding no merit in CMA's contention that EPA has required treatment to "below established levels of hazard," we find that EPA's interpretation of Sec. 3004(m) is reasonable.

Our concurring colleague suggests that our discussion of the reasonableness of the BDAT standard is unnecessary, if not "perhaps analytically impossible." Con.Op. at 371. Contrary to the impression given in his separate opinion, however, the basis upon which we find EPA's interpretation reasonable here is not one that we have supplied, but the one EPA itself put forth. In its Initial Rule document discussing BDAT as well as screening levels, and in its briefs to this court, EPA has presented precisely the arguments we find persuasive here. While, as we shall see, those arguments are inadequate to justify the choice made, in the Final Rule, in favor of BDAT as against screening levels--which also seem to present a reasonable approach--they do demonstrate that the BDAT approach is reasonable.

B. Was EPA's Explanation Adequate?

The Supreme Court has made it abundantly clear that a reviewing court is not to supplement an agency's reasons for proceeding as it did, nor to paper over its plainly defective rationale: "The reviewing court should not attempt itself to make up for such deficiencies [in the agency's explanation]; we may not supply a reasoned basis for the agency's action that the agency itself has not given." Motor Vehicles Manufacturers Ass'n v. State Farm Mut. Auto Ins. Co., 463 U.S. 29, 43, 103 S. Ct. 2856, 2866-67, 77 L. Ed. 2d 443 (1983) (citing SEC v. Chenery Corp., 332 U.S. 194, 196, 67 S. Ct. 1575, 1577, 91 L. Ed. 1995 (1947)). "We will, however, 'uphold a decision of less than ideal clarity if the agency's path may reasonably be discerned.' " Id. (quoting 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)). Accordingly, in order to determine whether we can affirm EPA's action here, we must parse the language of the Final Rule to see whether it can be interpreted to make a sensible argument for the approach EPA adopted. We find that it cannot.

As we have said, EPA, in its Proposed Rule, expressed a tentative preference for an approach that combined screening levels and BDAT. It indicated that it thought either that approach or BDAT alone was consistent with the statute, and recognized that there were myriad uncertainties inherent in any attempt to model the health and environmental effects of the land disposal of hazardous wastes. It initially concluded, however, that despite those uncertainties, the better approach was to adopt the combination of screening levels and BDAT. Nevertheless, in the Final Rule, it rejected its earlier approach, and adopted a regime of treatment levels defined by BDAT alone.

In order fully to convey the inadequacy of EPA's explanation, we quote the relevant portion of the Final Rule at length:

Although a number of comments on the proposed rule favored the first approach; that is, the use of screening levels to "cap" treatment that can be achieved under BDAT, several commenters, including eleven members of Congress, argued strongly that this approach did not fulfill the intent of the law. They asserted that because of the scientific uncertainty inherent in risk-based decisions, Congress expressly directed the Agency to set treatment standards based on the capabilities of existing technology.

The Agency believes that the technology-based approach adopted in today's final rule, although not the only approach allowable under the law, best responds to the above-stated comments. Accordingly, the final rule establishes treatment standards under RCRA section 3004(m) based exclusively on levels achievable by BDAT. The Agency believes that the treatment standards will generally be protective of human health and the environment. Levels less stringent than BDAT may also be protective.

The plain language of the statute does not compel the Agency to set treatment standards based exclusively on the capabilities of existing technology.... By calling for standards that minimize threats to human health and the environment, the statute clearly allows for the kind of risk-based standard originally proposed by the Agency. However, the plain language of the statute does not preclude a technology-based approach. This is made clear by the legislative history accompanying the introduction of the final section 3004(m) language. The legislative history provides that "[T]he requisite levels of [sic] methods of treatment established by the Agency should be the best that has been demonstrated to be achievable" and that "[T]he intent here is to require utilization of available technology in lieu of continued land disposal without prior treatment." (Vol. 130, Cong.Rec. 9178, (daily ed., July 25, 1984)). Thus, EPA is acting within the authority vested by the statute in selecting [sic] to promulgate a final regulation using its proposed alternative approach of setting treatment standards based on BDAT.

The Agency believes that its major purpose in adopting the risk-based approach of the proposal (i.e., to allow different standards for relatively low-risk, low-hazard wastes) may be better addressed through changes in other aspects of its regulatory program. For example, EPA is considering the use of its risk-based methodologies to characterize wastes as hazardous pursuant to section 3001.

51 Fed.Reg. at 40,578.

To summarize: after EPA issued the Proposed Rule, some commenters, including eleven members of Congress, chastised the agency on the ground that the use of screening levels was inconsistent with the intent of the statute. They stated that because of the uncertainties involved, Congress had mandated that BDAT alone be used to set treatment standards. EPA determined that the "best respon[se]" to those comments was to adopt a BDAT standard. It emphasized, however, that either course was consistent with the statute (and that it was therefore not required to use BDAT alone). Finally, it asserted, without explanation, that its major purpose in initially proposing screening levels "may be better addressed through changes in other aspects of its regulatory program," and gave an example of one such aspect that might be changed.

This explanation is inadequate. It should go without saying that members of Congress have no power, once a statute has been passed, to alter its interpretation by post-hoc "explanations" of what it means; there may be societies where "history" belongs to those in power, but ours is not among them. In our scheme of things, we consider legislative history because it is just that: history. It forms the background against which Congress adopted the relevant statute. Post-enactment statements are a different matter, and they are not to be considered by an agency or by a court as legislative history. An agency has an obligation to consider the comments of legislators, of course, but on the same footing as it would those of other commenters; such comments may have, as Justice Frankfurter said in a different context, "power to persuade, if lacking power to control." Skidmore v. Swift & Co., 323 U.S. 134, 140, 65 S. Ct. 161, 164, 89 L. Ed. 124 (1944).

It is unclear whether EPA recognized this fundamental point. On the one hand, it suggested that the adoption of a BDAT-only regime "best-respond[ed]" to the comments suggesting that the statute required such a rule. On the other hand, EPA went on at some length to establish that the comments were in error, in that screening levels are permissible under the statute. EPA's "rationale," in other words, is that several members of Congress (among others) urged upon it the claim that Proposition X ("Congress mandated BDAT") requires Result A ("EPA adopts BDAT"), and that although Proposition X is inaccurate, the best response to the commenters is to adopt Result A.

Nor is anything added by EPA's bald assertion that its reason for initially preferring Result B (screening levels) "may be" better served by other changes in the statutory scheme. In its Proposed Rule, EPA had, after extensive analysis of the various alternatives, come to the opposite conclusion. It is insufficient, in that context, for EPA to proceed in a different direction simply on the basis of an unexplained and unelaborated statement that it might have been wrong when it earlier concluded otherwise.

In the entire relevant text of the Final Rule, EPA neither invokes nor discusses the uncertainties inherent in the land disposal process in support of its determination to use BDAT. EPA's only mention of the concept is in its description of the commenters' argument that, because of such uncertainties, Congress mandated BDAT--an argument that EPA rejected. While it may be that EPA intended that reference to act as an incorporation of all the uncertainties it outlined in its Proposed Rule, or all the many challenges to its assumptions that commenters submitted in response to the Proposed Rule, that intent, if indeed it exists, is so shrouded in mist that for this court to say that we could discern its outlines would be as illogical as the agency's explanation in the Final Rule itself.

Accordingly, we grant the petitions for review in this respect.

HWTC and NRDC contend that the Administrator's regulation allowing generators to store wastes on-site for periods of up to one year unless EPA "can demonstrate that such storage was not solely for the purpose of accumulati[ng]" quantities of waste suitable for treatment, 40 C.F.R. Sec. 268.50(b) (1988), violates Congress' "plain intent" in enacting section 3004(j) of RCRA. According to these petitioners, this provision's flat prohibition against storage of wastes "unless such storage is solely for [proper purposes]," 42 U.S.C. Sec. 6924(j) (Supp. V 1987), requires generators--and not EPA--to bear the burden of proving that their motives in storing prohibited wastes are consistent with section 3004(j), no matter what the circumstances. "By shifting the statutory burden of proof of EPA ...", we are told, "the rule effectively allows a one year override of the statutory prohibition."

It is unclear to what petitioners refer when they speak of the "statutory burden of proof." At common law, the "burden of proof" concept bore two somewhat dissimilar meanings. In classical applications, the party carrying the burden of proof bore the risk of nonpersuasion of the factfinder; in other words, the "burden of proof" customarily implied the "burden of persuasion." 9 WIGMORE, EVIDENCE Sec. 2486 (Chadbourn rev. 1981). A secondary meaning developed, however, in service of the role of the trial judge in the common-law tribunal. To enable the judge "to keep the jury within the bounds of reasonable action," the party bearing the burden of proof had a threshold responsibility of satisfying the judge that sufficient evidence had been advanced "to form a reasonable basis for the verdict." Id. Sec. 2487, at 293. The discharge of this so-called "burden of production" was a prerequisite, at common law, to getting one's case to the jury at all.

In administrative proceedings, the APA provides a default rule for allocating proof burdens when regulatory statutes do not set forth separate rules. Specifically, "the proponent of a rule or order," usually the agency in proceedings charging statutory violations, "has the burden of proof." 5 U.S.C. Sec. 556(d) (1982). We have held, though, that the APA uses the term in its secondary application; "the 'burden of proof' it casts upon the 'proponent' is the burden of coming forward with proof, and not the ultimate burden of persuasion." Environmental Defense Fund, Inc. v. EPA, 548 F.2d 998, 1013 (D.C.Cir.1976) cert. denied, 431 U.S. 925, 97 S. Ct. 2199, 53 L. Ed. 2d 239 (1977); accord, Old Ben Coal Corp. v. Interior Bd. of Mine Operations Appeals, 523 F.2d 25, 30 (7th Cir.1975). As the House Report accompanying the APA explains, "section [556(d) ] means that every proponent of a rule or order or the denial thereof has the burden of coming forward with sufficient evidence therefor." H. REP. No. 1980, 79th CONG., 2D SESS. 34 (1946), U.S.Code Cong. & Admin.News 1946, p. 1195. While the locus of the ultimate burden of persuasion may be unclear--indeed, it may rest on the opponent of an agency order, see 3 K. DAVIS, ADMIN. L. TREATISE Sec. 16.9 at 258 (2d ed. 1980)--it is beyond doubt that the initial burden of going forward with a prima facie case of unlawful conduct rests on the agency charging the statutory violation, unless the regulatory statute provides otherwise.

Insofar as petitioners allege that EPA has shifted this threshold burden of going forward from the regulated industry to itself in contravention of RCRA, their contention plainly runs aground on the APA.4 For under the APA's guiding provisions, " '[no] agency is entitled to presume that the conduct of any person or status of any enterprise is unlawful or improper" unless the agency's organic statute provides otherwise. Environmental Defense Fund, Inc., 548 F.2d at 1014-15 (quoting S.REP. NO. 752, 79th Cong., 1st Sess. 22 (1945)); see also Industrial Union Dept., AFL-CIO v. American Petroleum Inst., 448 U.S. 607, 653 & n. 61, 100 S. Ct. 2844, 2869 & n. 61, 65 L. Ed. 2d 1010 (1980) (opinion of Stevens, J.). Petitioners point to nothing in RCRA that purports explicitly or implicitly to alter the APA's background rule placing this threshold burden of going forward on the agency. The language of section 3004(j) is silent as to allocations of production burdens (or, for that matter, the ultimate burden of persuasion) between the agency (the charging party) and the regulated respondent. The legislative history of RCRA's storage prohibition, as cited to us by HWTC and NRDC, is equally unilluminating; it merely repeats the proscription of section 3004(j). And petitioners identify no other provision in the statute that so much as even addresses procedural burdens in administrative proceedings under RCRA. See generally 42 U.S.C. Sec. 6928 (1982 & Supp. IV 1986) (governing enforcement proceedings). The Administrator's interpretation of section 3004(j) to comport with the procedural tradition of the APA can hardly be termed unreasonable in this setting. See Chevron U.S.A. Inc. v. Natural Resources Defense Council, 467 U.S. 837, 843, 104 S. Ct. 2778, 2782, 81 L. Ed. 2d 694 (1984) ("if the statute is silent or ambiguous with respect to [a] specific issue, the question for the court is whether the agency's answer is based on a permissible construction of the statute").

Were the Administrator, thus, to assess noncompliance penalties against a generator in the belief that the generator had accumulated prohibited wastes for improper purposes, the generator would be entitled to air the agency's charges in a public factfinding hearing. See 42 U.S.C. Sec. 6928(b) (Supp. V 1987). And at any such hearing, the Administrator, as the "proponent" of the agency compliance order, would be required to come forward at least with a prima facie case suggesting improper storage. 5 U.S.C. Sec. 556(d) (1982); see Environmental Defense Fund, Inc., 548 F.2d at 1014-15; Old Ben Coal Corp., 523 F.2d at 30. As we read the challenged regulation, as such, it simply sets forth in RCRA terminology the APA's default rule regarding the burden of production; to wit, "unless the Agency can demonstrate that [the generator's] storage [of prohibited wastes for less than one year] was not solely for the purpose of accumulation of such quantities of hazardous waste as are necessary to facilitate proper recovery, treatment, or disposal," the storage will not be penalized under RCRA. The regulation shifts nothing; it places no evidentiary burden on the agency that it would not otherwise be required to bear under section 556(d).

The Administrator, to be sure, has published companion regulations providing, with respect to storage of prohibited wastes for periods greater than one year, that "the owner/operator bears the burden of proving that such storage was solely for the purpose of accumulation of such quantities of waste as are necessary [for treatment]." 40 C.F.R. Sec. 268.50(c) (1988). But this in no way suggests, as petitioners argue, that the Administrator has "reverse[d] the statutory presumption against storage" for periods of one year or less. Rather, the Administrator has simply determined (reasonably, we think) that producing evidence that a generator has stored prohibited wastes for a period greater than one year fulfills the agency's section 556(d) obligation to come forward in administrative proceedings with a prima facie case of unlawful storage. Once the Administrator makes such a prima facie showing, of course, the burden may permissibly shift to the generator to demonstrate that such lengthy storage was in fact motivated by legitimate considerations under RCRA. See First Nat'l Bank of Bellaire v. Comp. of Currency, 697 F.2d 674, 683 (5th Cir.1983); Environmental Defense Fund, Inc., 548 F.2d at 1014-15; Old Ben Coal Corp., 523 F.2d at 30.

At bottom, what petitioners quarrel with is the precise point at which the Administrator can satisfy his initial burden of production in proceedings charging a section 3004(j) violation solely by introducing evidence of the duration of the generator's storage. Indeed, petitioners conceded at oral argument that they in all likelihood would not have challenged the Administrator's proposed rule--providing a 90-day storage window--had it been carried forward. But petitioners offer no basis to question the Administrator's professional judgment on this score. The record amply supports the Administrator's conclusion that aggregation of wastes for proper treatment may require accumulation for periods of up to one year. It was eminently reasonable, under these circumstances, for the Administrator to determine that he would have to come forward with more than the mere duration of storage for less than one year to make out a prima facie case under section 3004(j). Accordingly, we hold that 40 C.F.R. Sec. 268.50(b) is reasonable and consistent with RCRA.

As part of its implementation of the Hazardous and Solid Waste Amendments of 1984 ("HSWA"), Pub.L. No. 98-616, 98 Stat. 3221, the EPA developed an enforcement plan to assure that wastes that are prohibited from land disposal will not make their way into the ground. Under the EPA's scheme, restricted wastes will follow one of two paths. First, if the generator of the waste determines that he is managing a restricted waste and the waste does not meet the applicable treatment standards, he must notify the treatment facility of the appropriate treatment standards, see 40 C.F.R. Sec. 268.7(a)(1); the treatment facility is then required, pursuant to 40 C.F.R. Sec. 268.7(b), to test the treatment residue to assure that the waste, once treated, meets those standards before forwarding the waste to a land disposal facility,5 which is also required to test the waste, 40 C.F.R. Sec. 268.7(c). Alternatively, if a generator determines that he is managing a restricted waste, but that the waste can be land disposed without further treatment, he may ship the waste directly to landfill operators, the final handlers of the waste who, under the EPA scheme, bear ultimate responsibility for testing and determining that land disposed wastes meet the applicable treatment standards. See 51 Fed.Reg. 40,597 (November 7, 1986).

Although earlier handlers of wastes--both waste generators and treatment facilities--are also required by the regulations to certify that waste leaving their control and marked for land disposal meets the appropriate treatment standards, only the latter are expressly required to test the waste in order to certify compliance. See 40 C.F.R. Sec. 268.7(b). Generators of waste are "recommend[ed]" to conduct "a comprehensive analysis of each waste stream ... at least annually," 51 Fed.Reg. at 40,598, but in the end the agency's regulations leave generators the option of certifying that their wastes comply with treatment standards on the basis of, inter alia, their "knowledge" of the waste:

If a generator determines that he is managing a restricted waste under this part, and determines that the waste can be land disposed without further treatment, with each shipment of waste he must submit, to the land disposal facility, a notice and a certification stating that the waste meets the applicable treatment standards....

(ii) The certification must be signed by an authorized representative and must state the following:

I certify under penalty of law that I personally have examined and am familiar with the waste through analysis and testing or through knowledge of the waste to support this certification that the waste complies with the treatment standards specified in 40 C.F.R. Part 268 Subpart D and all applicable prohibitions set forth in 40 C.F.R. 268.32 or RCRA section 3004(d). I believe that the information I submitted is true, accurate and complete. I am aware that there are significant penalties for submitting a false certification, including the possibility of a fine and imprisonment.

40 C.F.R. Sec. 268.7(a)(2) (emphasis added).

HWTC and NRDC challenge the agency's decision to allow generators to rely on their knowledge to certify that wastes are within treatment standards. Petitioners note that wastes requiring treatment must be tested before being sent to land disposal facilities, and they therefore argue that it is arbitrary and capricious for the agency to fail to require generators of waste to test their waste streams in order to certify that admittedly restricted wastes conform to the applicable treatment standards. They charge that since the applicable treatment standards are stated in terms of specific and minute concentrations of hazardous constituents, without actual test data, "generators cannot possibly determine whether their wastes are generated meet these treatment standards and can be land disposed." Brief for Petitioners HWTC and NRDC at 14.6 They urge this court to replace the agency's rule with a requirement of their own: "[W]astes, which a generator has determined (by whatever means) to be: 1) hazardous and 2) subject to a land disposal restriction (e.g., they are a listed solvent or dioxin waste), must be tested by the generator if the generator is to certify that the wastes meet treatment standards and can be transported directly to a land disposal facility." HWTC/NDRC Reply Br. at 12 (emphasis in original). This is a requirement we are unwilling to impose.

First, unlike petitioners we find it neither nonsensical nor absurd to expect that generators may to some extent "know their waste" without testing each batch produced. Indeed, waste generators who apply the same methods to the same inputs in the same manner as part of the same production process every day are, after a while, likely to be in a very good position to know the hazardous contents of their waste. As we read the EPA's rules and statements during the rulemaking process, the agency's scheme does not allow generators to make guesses about the hazardous nature of their wastes without empirical or analytical foundation. Rather, waste generators are allowed to rely on actual "knowledge" they have acquired only if such knowledge enables them to certify that their waste complies with applicable treatment standards. Generators are required to keep records of all data that goes into their certifications, see 40 C.F.R. Sec. 268.7(a)(4), and they are subject to penalties for erroneous certifications. Thus, contrary to petitioners' assertion that "nothing in the rule itself ... requires generators shipping wastes directly to a landfill to test the waste to determine compliance with the treatment standards," Letter from HWTC (March 28, 1989) at 2, the EPA's scheme will necessarily require at least some initial testing of generators' waste stream in order to comply with the rules' plain directives.7 If down the road the generators' familiarity with their wastes does indeed render them capable of certifying the wastes' contents without conducting more frequent testing, then we see no reason to compel the EPA to require such unnecessary testing.

Furthermore, we do not find the EPA's decision to require treatment facilities to conduct testing but to allow generators to rely on their knowledge to be "arbitrary." The rulemaking record adequately reflects the EPA's sense that while generators can be expected to have reasonable knowledge of familiar wastes, off-site treatment facilities do not always have similar familiarity with the waste they handle. Moreover, it is the treatment facility's job to transform waste otherwise deemed too dangerous to permit into landfills into acceptable form. It is therefore not irrational for the EPA to introduce a backup, arguably "redundant" testing stage for these wastes requiring treatment, and even to consider this a "critical" stage in the process. See 51 Fed.Reg. at 40,597.

Although the agency's certification system may be somewhat imprecise with regard to generators of waste, this imprecision is not fatal. Rather, the EPA has explicitly stated that the crucial stage in the process, upon which the agency has placed its most heavy reliance, is the point at which the waste reaches the land disposal facility: at this juncture, just prior to land disposal, waste must be rigorously tested to confirm that it is what others have represented it to be and that it may permissibly be land disposed. Given the agency's reliance on testing by landfill owners and operators to intercept erroneously identified waste, we cannot say that the EPA acted arbitrarily or capriciously in deciding not to require elaborate and even redundant testing8 by generators presumably able to identify in a large number of cases the hazardous components of the waste they generate.

HWTC and NRDC further argue that the testing required of disposal facilities will be inadequate to assure that only wastes that are permitted to be land disposed will actually enter landfills.9 In particular, they complain that "substantial percentages of individual waste shipments received by a landfill operator are not required to be tested for compliance with the treatment standards" by actual testing. Letter from HWTC (March 28, 1989) at 3. Despite these concerns, however, the regulations are structured to assure that the frequency of testing is sufficient to identify wastes that do not comply with treatment standards.For the purposes of compliance with the land disposal restrictions rule, a waste analysis plan for an off-site disposal facility must address the procedures for screening incoming shipments of waste to ensure that wastes received conform to the certification made by the generator or treatment facility. That is, the waste analysis plan must address the procedures necessary for determining whether an extract of the waste or treated waste meets the treatment standards.

51 Fed.Reg. at 40,598.10 In a sense, then, petitioners' concerns are premature: while the EPA's scheme is designed to assure adequate testing, which includes case-by-case determinations of the frequency with which actual testing will need to be conducted on waste shipments, petitioners anticipate that the EPA will authorize testing schedules that are inadequate. We prefer to anticipate that the agency will faithfully execute its responsibilities under the statute, and will impose testing requirements that will guarantee that Congress' purposes in enacting the statute are implemented. If the agency does not live up to this expectation, there will be time and opportunity for petitioners' challenge.

At its base, the challenge of HWTC and NRDC is undergirded by a peculiar set of epistemological assumptions. In brief, these petitioners appear to argue that only much more frequent testing of waste at every stage of its handling would ever allow us to "know" whether any given batch of waste (itself an arguably "arbitrary" dividing line) conforms to the EPA's treatment standards. Absent continuous testing at the point of generation, they argue, generators cannot certify what levels of hazardous constituents their waste contains; and if landfill operators are not required to test each individual waste shipment for compliance with the treatment standards by the stringent "Toxicity Characteristic Leaching Procedure" testing method, then the testing requirement at the land disposal stage cannot be relied upon to catch waste that generators have erroneously certified as falling naturally within these standards.

While we have no desire to enter a metaphysical debate over the source and nature of all knowledge, common sense compels recognition of the fact that much of what we think of as "knowledge" in the practical world is nothing more than extrapolation from a more limited set of experiences. As relevant to the present case, we cannot say that the statute requires testing beyond what is practically necessary to assure with a high degree of confidence that prohibited wastes are not being land disposed. We therefore hold that the EPA's decision to allow generators to rely in appropriate circumstances on their knowledge of their restricted waste to certify that it naturally meets treatment standards is reasonable.

We conclude that the solvents and dioxins rule is not arbitrary, capricious, or contrary to RCRA in any of the respects argued by petitioners, but remand the matter for the EPA to clarify its reasons for adopting the Final Rule in preference to the Proposed Rule. In order to avoid disrupting EPA's regulatory program, we will withhold issuance of our mandate for 90 days, during which the agency may either withdraw the Final Rule or publish an adequate statement of basis and purpose.

Judgment Accordingly.

SILBERMAN, Circuit Judge, concurring in part and concurring in the result:

I concur in all of the majority's per curiam opinion but its purported resolution of the Chevron "Step II" question concerning the reasonableness of BDAT treatment standards as a construction or application of RCRA. While CMA's "Step I" challenge to EPA's construction of RCRA section 3004(m)--i.e., whether the statute "clearly forecloses" the approach charted by the agency, Maj. op. 361-62 --was available for final judicial review, I do not believe it proper for the court to have reached the Step II question as to whether the selection of BDAT treatment levels was "a reasonable policy choice for the agency to make." Chevron U.S.A. Inc. v. Natural Res. Defense Council, 467 U.S. 837, 845, 104 S. Ct. 2778, 2783, 81 L. Ed. 2d 694 (1984). In the absence of a valid agency explanation as to how it has attempted to accommodate the competing interests Congress has committed to its care via RCRA, it is in my view inappropriate (perhaps analytically impossible) even to address, much less resolve, CMA's challenge to the reasonableness of EPA's treatment regime under the statute. Because the court today remands for further EPA explanation of its adoption of BDAT standards, the majority's Chevron Step II discussion should be considered dicta.

I agree with the majority's conclusion that Congress did not have a "specific intention" that technology-based treatment standards not be employed in implementation of section 3004(m), see Chevron, 467 U.S. at 845, 104 S.Ct. at 2783, and to that extent I further agree that the questions of statutory interpretation presented by CMA's petition are appropriately resolved under Step II of Chevron. My reading of the critical statutory language requiring EPA to set treatment standards so that "threats to human health and the environment are minimized" suggests a threshold ambiguity as to whether Congress intended the agency, insofar as it was technologically possible, to eliminate any statistically discernible risk to human health and the environment, or whether Congress intended there to be some sort of balancing. The dictionary definition of the word "minimize," see Maj. op. at 361, provides no ready answer to this question; a command that the agency "reduce" a threat "to the smallest possible degree" leaves open the factors that the agency can account for in determining what is, in fact, possible (or feasible?) under the circumstances.

We are also in agreement over the significance that EPA must attach to actual or reasonably perceived "threats to human health and the environment" in the course of fleshing out section 3004(m)'s meaning. As the majority notes, EPA is not

free ... to require generators to treat their waste beyond the point at which there is no 'threat' to human health or the environment. That Congress's concern in adopting Sec. 3004(m) was with health and the environment would necessarily make it unreasonable for EPA to promulgate treatment standards wholly without regard to whether there might be a threat to man or nature.

Maj. op. at 362. EPA is instead obliged to explain how its selection of BDAT treatment standards--which, as CMA notes, will require generators to treat certain wastes to levels of purity beyond those EPA requires for drinking water--is guided by RCRA's concern with health and environmental threats. But the majority persuasively demonstrates that EPA's explanation in the Final Rule falls woefully shy of this mark as a matter of administrative law, leaving the court without any hint whatsoever as to EPA's theory of the compatibility of the Final Rule with RCRA's purposes. Under these circumstances, resolution of the statutory questions confronting the court, at least those belonging to Step II of Chevron, is an improper exercise of judicial creativity.

In order to conclude--as the majority does--that the accommodation of competing RCRA policies reflected in the agency's treatment regulation is "one that Congress would have sanctioned," United States v. Shimer, 367 U.S. 374, 383, 81 S. Ct. 1554, 1560, 6 L. Ed. 2d 908 (1961) (quoted in Chevron, 467 U.S. at 845, 104 S.Ct. at 2783), the court necessarily must determine that the approach is "rational and consistent with the statute." NLRB v. United Food & Comm'l Workers U., 484 U.S. 112, 108 S. Ct. 413, 421, 98 L. Ed. 2d 429 (1987). This requires a determination that the agency has fashioned its approach in reliance on considerations made relevant by Congress under the substantive statute, which in turn requires an examination of the agency's stated reasons for adopting the challenged course. See, e.g., AFL-CIO v. Brock, 835 F.2d 912, 917 (D.C.Cir.1987) (equating Chevron 's second step with arbitrary and capricious review of agency policies); NRDC v. EPA, 824 F.2d 1146, 1163 (D.C.Cir.1987) (striking down agency statutory interpretation based on unreasonable "application of [relevant statutory] factors"); see also Chevron, 467 U.S. at 843, 104 S.Ct. at 2782 (requiring affirmance of administrative statutory construction if interpretation is a "reasonable policy choice for the agency to make"). The critical inquiry, as such, in the court's Chevron Step II inquiry is "whether the agency has advanced what the Chevron Court called 'a reasonable explanation for its conclusion that the regulations serve the ... objectives in question.' " Continental Air Lines v. Dep't of Transp., 843 F.2d 1444, 1452 (D.C.Cir.1988) (quoting Chevron, 467 U.S. at 863, 104 S.Ct. at 2791) (emphasis added).

EPA's explanation in the instant case, however, is utterly devoid of any rationale whatsoever for the agency's statutory construction or its policy choice. As the court observes, EPA's intentions in promulgating treatment standards are "so shrouded in mist that for this court to say that we could discern its outlines would be as illogical as the agency's explanation in the Final Rule itself." Maj. op. at 366. I would go further: I think it doubtful that EPA attempted at all to explain its presumptive view that the employment of BDAT treatment standards across the board would reasonably serve congressional intent. For after observing in the Final Rule that the plain language and legislative history of RCRA do not squarely preclude a technology-based approach, EPA failed to indicate what statutory policies tipped the balance in favor of that approach as opposed to the initial health-screening levels announced in the Proposed Rule. Before the court can determine that "the agency's answer is based on a permissible construction of the statute," Chevron, 467 U.S. at 843, 104 S.Ct. at 2782, the agency must explain how it has translated RCRA into its treatment standards regulation.

The majority's Chevron Step II analysis itself convincingly illustrates how important a role the agency's explanation of its policy accommodation plays in post-Chevron federal judicial review of agency statutory interpretation. Nowhere in its discussion does the majority address the agency's view of the way in which the BDAT regime serves RCRA's purposes; because the agency offered no such view, this should not be surprising. This part of the majority's opinion is instead devoted exclusively to the petitioner's objection that, with respect to certain solvents and dioxins, BDAT treatment levels will result in treatment to "below established levels of hazard." See Maj. op. at 362. The majority responds to CMA's argument by pointing out that none of the "established levels" to which CMA refers was developed under a statutory standard requiring minimization of "threats to human health and the environment." But EPA did not say that. We have before us no indication, as the majority later observes, that EPA was driven away from "established" health-screening levels on the basis of the asserted incomparability of the statutory standards under which those levels were determined. Nor, as the majority later notes as well, see Maj. op. at 365-66, is there any indication that EPA was impelled toward technology-based levels because of the "long-term uncertainties associated with land disposal" Congress identified in RCRA. 42 U.S.C. Sec. 6924(d)(1)(A). The majority's treatment of CMA's--as opposed to EPA's--analysis at best suggests that CMA's approach is not compelled by the legislative text, a proposition relevant only to Chevron 's first step.

Indeed, the majority's discussion of the reasonableness of EPA's interpretation of the statute necessarily proceeds without reference to an agency interpretation because no such construction exists.1 One is left to wonder how the majority can give deference to a statutory construction (or an explanation as to how an agency iniative is consistent with the statute) that nowhere appears in the Final Rule. The majority's Chevron Step II analysis, in my opinion, is nothing more than an advisory opinion to the effect that were the court presented with a Final Rule that echoed the majority's discussion of an appropriate balance to strike among RCRA's purposes, the court would sustain the agency's view of the statute. As a consequence, the majority ends up deferring not to an agency statutory construction, but rather simply to a result. Assuming this judicial approach ever were permissible, surely after Chevron it no longer is.

Given the complexity of the subject matter and the fundamental ambiguity in Congress' direction, a complete agency explication of its view of the statute would be especially helpful in this case. For instance, the question of how (and why) Congress would have intended EPA to require generators to treat the wastewaters they intend to pour into the ground to levels more pure than Congress requires for drinking water drawn out of the ground would surely benefit from the views of those to whom Congress entrusted regulatory responsibility. More fundamentally, it is incumbent upon EPA to identify the incremental "threats to human health and the environment" that it hopes to address by opting uniformly for more stringent technology-based standards in lieu of health-based standards of whatever origin. Cf. Small Refined Lead Phase-Down Task Force v. EPA, 705 F.2d 506, 523 (D.C.Cir.1983) ("adverse health effects," in and of themselves, do not permit EPA "to justify any ... standard at all, without explaining why it chose the level it did"); United Steelworkers of America v. Marshall, 647 F.2d 1189, 1207 (D.C.Cir.1980) (agency must "explain the logic and the policies underlying any legislative choice"), cert. denied, 453 U.S. 913, 101 S. Ct. 3148, 69 L. Ed. 2d 997 (1981); Lead Indus. Ass'n v. EPA, 647 F.2d 1130, 1162 (D.C.Cir.) (choice between two policy approaches must be explained), cert. denied, 449 U.S. 1042, 101 S. Ct. 621, 66 L. Ed. 2d 503 (1980); Industrial U. Dep't, AFL-CIO v. Hodgson, 499 F.2d 467, 476 (D.C.Cir.1974) ("when [an administrator] is obliged to make policy judgments ..., he should so state and go on to identify the considerations he found persuasive"). Only then can the court legitimately defer to the agency's construction of RCRA, for proper judicial deference to an agency interpretation requires an understanding of the agency's objectives that can only be gleaned from the agency's presentation of its rule. With all respect, the majority's analysis, however appealing, see Maj. op. at 362-64, cannot substitute for this obligatory agency explanation.

I do not mean to ignore the conceptual distinction between review of an agency's statutory construction and of an agency's actions under the arbitrary and capricious standard. We have in the past said "[i]t would be inappropriate ... to import wholesale [arbitrary and capricious review principles] and apply [them] in [the] conceptually distinct arena" of statutory construction. Continental Air Lines, Inc. v. Dep't of Transp., 843 F.2d 1444, 1452 (D.C.Cir.1988). But at the same time we have often recognized that Chevron 's second step and review of an agency's action under the arbitrary and capricious standard, although starting from different legal premises, often converge and sometimes overlap. See, e.g., General Am. Transp. Corp. v. ICC, 872 F.2d 1048, 1053 (D.C.Cir.1989); AFL-CIO v. Brock, 835 F.2d at 917; Natural Res. Defense Council v. EPA, 824 F.2d at 1163; Rettig v. Pension Benefit Guaranty Corp., 744 F.2d 133, 152 (D.C.Cir.1984). One thing, in any event, is quite clear: these "distinct" judicial review functions proceed from a common foundation--the agency's expressed view. Thus, if the agency has offered an inadequate explanation as to how its chosen policy is consistent with Congress' mandate, the court's Chevron Step II analysis is necessarily hypothetical.

It would appear that EPA faced with formidable political forces opposing its Proposed Rule, simply acquiesced in the approach desired by those forces, but was unwilling to offer as its own a statutory/policy rationale to justify its acquiescence. In the Final Rule, EPA in effect stated that it recognized, and subordinated itself to, the senators and congressmen who protested against EPA's Proposed Rule without in any way affirming the legal (or policy) superiority of the legislators' position. My colleagues acknowledge EPA's behavior is intolerable as a matter of administrative law, see Meredith Corp. v. FCC, 809 F.2d 863, 872-73 (D.C.Cir.1987) (holding that FCC was obliged to address constitutional challenge to fairness doctrine notwithstanding "non-legislative expressions of congressional concern" that the question be reserved for Congress); Sierra Club v. Costle, 657 F.2d 298, 404-10 (D.C.Cir.1981) (holding that EPA's ex parte exchanges with congressional leaders and White House officials did not render informal rule procedurally infirm since EPA set forth its own independent rationale for the rule selected), but nevertheless "rescue" EPA from its predicament by supplying the statutory/policy analysis which, if it had been adopted by EPA, would have obviated the need for a remand. Under the circumstances, I do not know why the court's remand is other than an empty gesture, one which conforms to principles of judicial review of agency policymaking only in form.