In our original opinion, 350 F.3d 1263 (D.C.Cir.2003), we considered a challenge to an EPA rule that exempted certain recycled zinc fertilizer products from regulation under the Resource Conservation and Recovery Act ("RCRA"), 42 U.S.C. § 6901. The new rule, in relevant part, exempted such products from RCRA so long as their contaminant levels fell below specified limits set by EPA. EPA claimed that recycled products meeting these regulations would have environmental impacts substantially similar to those of analogous products made from virgin materials, and that therefore EPA could lawfully classify them as not being "solid waste" for RCRA purposes despite their being hazardous materials within the meaning of the statute. We upheld this so-called "identity principle" — together with market valuation and EPA-required management practices — as a valid standard for distinguishing waste from non-waste, and we further upheld the principle's application to EPA's chosen limits for lead, arsenic, cadmium, and mercury in exempted products.

Petitioners sought rehearing on several grounds, one of which requires discussion, and, indeed, a limited remand. That ground is the argument that our conclusions with regard to the metals in question were based on a study submitted by The Fertilizer Institute ("TFI Study") that was not part of the record. A related aspect of this claim is the argument that even if the TFI Study or its relevant conclusions were part of the record, EPA did not rely on the study in promulgating its rule, and therefore we could not properly do so in upholding it. See SEC v. Chenery Corp., 318 U.S. 80, 87, 63 S. Ct. 454, 87 L. Ed. 626 (1943). We asked for and received a response from the EPA.

After considering the parties' submissions, we reach the following conclusions: Petitioners are correct that we relied on the TFI Study — in the limited sense of relying on its conclusions. Those conclusions, however, were in the record, so that the precise claim raised by petitioners is incorrect. Nonetheless, in our reliance on the TFI Study we may have gone farther than any express EPA language justified in equating it with an EPA study that was in the record and was expressly relied on by EPA, but which we as lay judges found ourselves unqualified to interpret. Thus our original opinion made certain connections that ought to have been made — assuming they can properly be made — by the agency. We therefore remand to EPA for a more detailed explanation of the relationship between its risk assessment study and the conclusions of the TFI Study. In all other respects the petition for rehearing is denied.

* * *

Petitioners appear to be correct that the TFI Study was not in the rulemaking record, though the conclusions of that study — i.e., the risk thresholds cited by EPA in its final rule and by this court in our original opinion — clearly were. But EPA's position that virgin fertilizers and recycled fertilizers meeting EPA's proposed contaminant limits are "identical" in their environmental impact did not rest directly on the TFI Study. In its Notice of Proposed Rulemaking ("NPRM"), for instance, EPA articulated its rough equation of the risk threats (of fertilizers made of virgin materials and of ones made with qualifying recycled materials) without supporting citation, saying that while "contaminant levels in non-hazardous feedstocks are slightly lower than those in hazardous feedstocks... any potential risks posed by hazardous and non-hazardous zinc feedstock materials would be substantially similar...." 65 Fed. Reg. 70,959 (emphasis added). EPA also said that its proposed contaminant limits were based "on contaminant levels that can be routinely and reliably achieved in ZSM [zinc sulfate monohydrate] fertilizer products," a finding which was based on samples from representative virgin products, and EPA further claimed that lower limits would not "result in any significant gain in environmental protection." 65 Fed.Reg. 70,969 (emphasis added). As EPA did not have the TFI Study before it at that point, its conclusion about the environmental impacts of recycled and virgin fertilizers was presumably based on EPA's own studies.

EPA's non-reliance on the TFI Study — and its affirmative reliance on its own studies — is made even more clear by the explanation that accompanied the final rule. Responding to commenters who called for more stringent technology-based limits (based on the alleged capability of fertilizer producers to achieve such limits), EPA pointed out that "[t]he Agency's fertilizer risk assessment indicates that the proposed limits are considerably below levels that we estimate (albeit roughly) to be safe for humans and ecosystems. Thus, the actual environmental benefit to be gained from more stringent limits would likely be negligible." 65 Fed. Reg. 48,405/2. Similarly, in responding to comments on the NPRM, EPA noted that "[w]hile there are uncertainties in EPA's study of fertilizer contaminant risks, we are confident in its basic conclusions, particularly since the limits are well below EPA's thresholds for acceptable risks to human health." EPA's Proposed Regulations for Zinc Fertilizers Made from Recycled Hazardous Secondary Materials: Response to Comments, Docket No. 8 (undated) at 23. Thus, EPA's conclusion that its identity principle was satisfied rested not on the TFI Study, but on EPA's own assessment, announced in the NPRM and apparently unchallenged, that recycled materials meeting EPA's limits posed no meaningful extra risk beyond that of fertilizers from virgin materials.

We note here that the validity of EPA's own risk assessment was potentially in play in the rulemaking — though not on precisely the issue of EPA's application of its identity principle. The reason that its pertinence before the agency was not on precisely that question is because that question was never posed by petitioners or any other party. But EPA invoked its risk assessment in responding to claims that it could and should have adopted more stringent technology-based limits, and it is precisely those claims to which petitioners have pointed before us in responding to EPA's argument, see EPA Br. at 44 & n.23, that the application of the identity principle had never been attacked in the rulemaking, see Petitioners' Reply Br. at 7 n.4. In treating petitioners' challenge to the identity principle as properly before us, we (implicitly) extrapolated from general arguments during the comment period that the limits should have been stricter. We similarly extrapolated from EPA's response to these comments, 67 Fed. Reg. 48,405/2, an appropriate defense of its application of the identity principle; since petitioners didn't frame their agency-level challenge in that way, EPA's discussion was of course not stated as a response to such a challenge.

In our original opinion the TFI Study became relevant primarily because the EPA risk assessment, and the other studies in the record upon which EPA relied, are difficult for non-expert judges to interpret. See Safe Food, 350 F.3d at 1271. Had the TFI Study never been submitted or discussed in the explanation of the final rule, we would presumably either have had to take EPA at its word that the differences in contaminant levels between virgin and recycled fertilizers are trivial when viewed in the perspective of real risks to health and the environment, or have had to remand to EPA with instructions to further explain how the results of the studies in the record could be translated into risk thresholds that we could use to verify EPA's claims regarding identical health and environmental impacts.

Our original opinion seized on a third option that made use of the TFI Study, the conclusions of which had been submitted by industry commenters who thought that EPA should adopt as its contaminant limits the much higher, risk-based levels that the TFI Study proposed as thresholds. EPA declined to do so, noting that such a decision would allow contaminant levels to increase dramatically and that this would not be a desirable environmental result, especially considering the uncertainty inherent in such risk estimates. Despite this uncertainty, however, EPA observed that "the general findings of EPA's risk assessment did not differ dramatically from those of the TFI-sponsored study." 67 Fed. Reg. 48,405/1. Given this claim of rough equation between the two studies — which petitioners to this day appear not seriously to have contested in their submissions to this court — we used the TFI Study risk estimates as a benchmark for determining whether EPA could reasonably find that the differences between EPA's contaminant limits and the contaminant levels found in virgin products were insignificant. We found that, for lead, mercury, arsenic, and cadmium, the TFI Study risk thresholds ranged from 20 to 372 times higher than EPA's proposed contaminant limit. Safe Food, 350 F.3d at 1270. Having thus used the TFI Study to "translate" EPA's own risk estimates into terms comparable with EPA's proposed limits, we concluded that EPA was justified in its initial finding — again, a finding never challenged either in the rulemaking or before us — that the risks from virgin and recycled fertilizers were, for all practical purposes, identical. But, although the EPA risk assessment and the TFI Study considered chromium, the summary of the TFI Study results for some reason did not include a proposed chromium risk threshold, and the findings of the EPA assessment were not comprehensible to us; accordingly we remanded for further explanation.

We now recognize that our original opinion was insufficiently clear about the use to which we put the TFI Study, in that the opinion suggests that the conclusion rested solely or primarily on the TFI Study itself. That is not the case. Rather, the use of the TFI Study numbers hinged on EPA's uncontested claim that the TFI Study estimates were roughly comparable to EPA's own risk assessment. Given that assurance of rough comparability, and notwithstanding our recognition that such estimates were subject to considerable uncertainty, we could see no error in EPA's application of its identity principle.

Nonetheless, we are mindful of the Chenery rule that we can uphold an agency decision only on the basis of arguments and evidence provided by the agency during the rulemaking proceedings. And we recognize that, as petitioners point out, we put the results of the TFI Study to a use that EPA appears not to have considered in its explanation of the rule. The record contains all the necessary pieces, but we put them together in a way that the agency had not. In particular, we accepted EPA's assertion in one section of the record that the TFI Study results "did not differ dramatically" from the EPA's own risk assessment, 67 Fed. Reg. at 48,405/1, to verify EPA's claim that the EPA risk assessment "indicates that the proposed limits are considerably below levels that we estimate (albeit roughly) to be safe for humans and ecosystems." Id. at 48,405/2. We conclude that we erred insofar as we simply accepted EPA's claim that its risk assessment, on which the agency could properly rely, generated results sufficiently comparable to the TFI Study, on which the agency did not rely, for us to use the latter study to evaluate whether EPA's claim regarding identity was plausible.

We therefore remand with instructions that EPA explain why the risk threshold estimates in the TFI Study are consistent with the EPA's own risk assessment, or, in the alternative, to do its own "translation" of its study, and the other studies properly in the record, into terms that a reviewing court could use to assess whether the EPA reasonably applied its identity principle. We stress the narrowness of this remand. Petitioners already had an opportunity and incentive to challenge the methodology of the EPA risk assessment and the other studies in the record, and to submit their own studies challenging EPA's consistent assertion that the differences in the contaminant levels found in virgin and recycled products are too small to be of moment from an environmental standpoint. Petitioners may not at this point have another bite at the apple. Nor may they at this stage challenge the methodology of the TFI Study if EPA chooses to approach the remand by showing how the TFI methodology and its own methodology are sufficiently comparable, as this would amount to a back-door challenge to EPA's methodology. Petitioners are limited at this stage to challenging the EPA's claim that its study and the TFI Study generate similar results.

So ordered.