Full opinion text
OPINION OF THE COURT TABLE OF CONTENTS PAGE I. Introduction 722 II The Statutory Framework 724 A. The Atomic Energy Act and the Licensing Process 724 B. The National Environmental Policy Act 725 III. Procedural History 728 IV. The Appeal of Limerick Ecology Action 728 A. Does a Finding of Adequate Protection Under the AEA Preclude Consideration Under NEPA? 729 B. Severe Accident Mitigation Design Alternatives 731 1. The Commission’s Decision 731 2. The Propriety of General Exclusion by Final Policy Statement 733 a. The NRC’s Pronouncement: Policy Statement or Rule? 733 b. The NRC's Failure to Give Careful Consideration to' SAMDAs 736 3. Is the Risk Remote and Speculative? 739 4. Summary 741 C. Sabotage Risk 741 1. The Commission’s Decision 741 2. Discussion 742 a. Whether Worst Case Analysis Is Required 743 b. Whether the NRC’s Determination that LEA’s Sabotage Contention Was Unsupported Is Arbitrary and Capricious 743 D. Industrial or Economic Damages After One Year 745 V. The Appeal of the Graterford Inmates 747 A. The Regulatory Framework and Procedural History 748 1. The Regulatory Framework 748 2. Procedural History 748 B. Contentions Not Considered by the NRC 749 1. The Need for Union Approval of the Plan 750 2. Panic Potential 750 3. The Civilian Evacuation Training Claim 750 C. The Accuracy of Estimate of Time of Evacuation Claim 751 D. The Manpower Mobilization by Telephone Claim 752 VI. Conclusion 754 BECKER, Circuit Judge. I. INTRODUCTION This opinion addresses several petitions for review of orders of the Nuclear Regulatory Commission (“NRC”) granting a full power license to the Philadelphia Electric Co. (“PECO”) for operation of Unit I of the Limerick Nuclear Power Generating Station (“Limerick”) in Limerick, Pennsylvania. The Limerick plant is, along with the Indian Point plant near New York City, and the Zion plant near Chicago, one of three operating nuclear plants in the country located within 50 miles of a major metropolitan area. The Limerick plant is twenty-five miles from Philadelphia and approximately eight miles from the State Correctional Institution at Graterford (“Grater-ford”), the largest maximum security prison in Pennsylvania. Two parties challenge the grant of a full power license. First, Limerick Ecology Action, Inc. (“LEA”), a citizens group formed in opposition to the plant, contends that, in granting the full power license, the NRC violated the National Environmental Policy Act of 1969, 42 U.S.C. §§ 4321 to 4361 (1982) (“NEPA”), by failing adequately to consider: (1) severe accident mitigation design alternatives (“SAMDAs”); (2) the threat of reactor sabotage; and (3) the possible industrial and economic effects that might arise more than one year after a severe accident. The NRC has described severe accidents as “those in which substantial damage is done to the reactor core whether or not there are serious offsite consequences.” Policy Statement on Severe Reactor Accidents Regarding Future Designs and Existing Plants, 50 Fed.Reg. 32,138, 32,138 (1985) (“Final Policy Statement”). Second, Thomas Martin, a Grater-ford inmate, challenges the adequacy of the Commonwealth of Pennsylvania’s plans to protect or evacuate the inmates in the event of a nuclear accident. We are confronted at the outset by the NRC’s contention that by making decisions under the Atomic Energy Act, 42 U.S.C. §§ 2011 to 2282 (1982) (“AEA”), it has precluded the need for consideration of environmental implications under NEPA. Because we conclude that consideration under NEPA should not be precluded by the AEA, we must address LEA’s three specific contentions. LEA’s first contention requires us to address the question whether, by excluding consideration of the environmental impact of SAMDAs through the use of a policy statement instead of a rulemaking, the NRC violated the first of NEPA’s twin aims: consideration of “every significant aspect” of the environmental consequences of government actions. The Supreme Court has stated that “[t]he role of the courts [under NEPA] is simply to ensure that the agency has adequately considered and disclosed the environmental impact of its actions and that its decision is not arbitrary and capricious.” Baltimore Gas & Electric Co. v. Natural Resources Defense Council, Inc., 462 U.S. 87, 97-98, 103 S.Ct. 2246, 2252, 76 L.Ed.2d 437 (1983). Although NEPA requires the Commission to undertake “careful consideration,” Baltimore Gas, 462 U.S. at 98, 103 S.Ct. at 2253, of environmental consequences, under Baltimore Gas it may issue a rulemak-ing to address and evaluate environmental impacts that are “generic”, i.e., not plant-specific. We find in this case that (1) the SAMDAs were addressed through a policy statement, not a rulemaking, and that the policy statement does not represent the requisite careful consideration of the environmental consequences; and (2) the Commission did not find that such risks are remote and speculative and failed to give the requisite careful consideration to SAM-DAs. Because the level of consideration given was legally inadequate, we will grant LEA’s petition for review as to its first contention and remand the case to the NRC for consideration of severe accident mitigation design alternatives. LEA’s second contention requires us to consider whether the NRC’s refusal to consider specifically and separately the risk of sabotage in the Final Environmental Statement, NUREG-0974 (1984) (“FES”) or in the licensing proceedings, on the ground that estimation of the risk is beyond the state of the art of risk assessment, violated NEPA. Because we conclude that LEA did not produce sufficient evidence to support its claim, and therefore that the NRC did not act. arbitrarily or capriciously in denying specific and separate consideration of the risk of sabotage, we will deny LEA’s petition on this issue. LEA’s third and final contention requires us to decide whether the NRC impermissi-bly excluded consideration in the FES of the amount of any industrial or economic damages arising more than one year after a nuclear accident. Again, we conclude that the NRC did not act arbitrarily or capriciously in refusing to consider such damages. We then turn to the AEA based challenge of petitioner Thomas Martin, a Gra-terford inmate, to the prison evacuation plan. Martin challenges the adequacy of the nuclear accident emergency response evacuation plan prepared by the Commonwealth’s Bureau of Corrections, which provides for evacuation of maximum security inmates through two important steps: (1) telephone calls to off-duty guards to notify them of the need to assist in the evacuation; and (2) transportation of inmates from the prison on commercial buses driven by civilian bus drivers. Martin claims that the NRC erred in approving civilian evacuation training, evacuation time, and manpower mobilization aspects of the emergency response plan. Martin also contends that the NRC erred in excluding from consideration three of his other objections to the response plan: lack of participation in the planning by the guards’ union; the potential for panic by inmates and guards; and the need for actual training (as opposed to the mere offer of training) of civilian drivers. Finally, Martin contends that the Atomic Safety and Licensing Board (“Licensing Board”) of the NRC violated his procedural due process rights in its treatment of his various contentions. We hold that substantial evidence exists to support the Commission’s acceptance of the estimated time for evacuation and the off-duty staff notification system and therefore that the NRC did not abuse its discretion with respect to these issues. We also hold that the NRC’s exclusion of the issues of the union’s awareness of the plan and of the inmates’ claimed potential for panic was within the bounds of the agency’s discretion. The issue of whether the civilian drivers must actually receive training was, however, properly raised by the inmates below, and we hold that the NRC abused its discretion in excluding consideration of that issue. We therefore remand to the Licensing Board for initial consideration of that contention. Finally, we conclude that there was no violation of Martin’s due process rights. Proper understanding of the issues requires that we first review the statutory framework and next the facts and procedural history of the case. We will then turn to a resolution of the legal issues raised by the petitions. II. THE STATUTORY FRAMEWORK A. The Atomic Energy Act and the Licensing Process The Atomic Energy Act provides for a two-stage approval process for consideration of the public health and safety aspects of nuclear power plant licensing: (1) consideration of whether the applicant should be able to construct a plant, see 42 U.S.C. § 2235; and (2) consideration of whether a facility should be licensed to generate electricity. Id. The NRC technical staff reviews an applicant’s safety and environmental reports, conducts technical studies, and employs consultants to review safety and environmental concerns. The staff’s conclusions are then set forth in a Safety Evaluation Report, an Environmental Impact Statement, and supplements to both documents. Section 189(a) of the AEA, 42 U.S.C. § 2239(a) (1982), the Administrative Procedure Act 5 U.S.C. §§ 551 to 706 (1982) (“APA”), and Commission regulations, 10 C.F.R. Part 2 (1988), establish a hearing framework for examining issues. First, a three-member Licensing Board is formed, generally composed of one lawyer and two technical members. 10 C.F.R. Part 2, App. A. For operating license applications, a hearing must be held on material issues that are specifically and timely raised upon request of an interested person. Id. at §§ 2.105, 2.714; see generally id. at Part 2, App. A. The findings and conclusions of the Licensing Board constitute the agency’s initial decision. Licensing Board decisions may be appealed to a three-member Appeal Board within forty days. Id. at § 2.786. However, the ultimate determination as to whether a facility’s operation “will be in accord with the common defense and security and will provide adequate protection to the health and safety of the public” rests with the Commission itself. 42 U.S.C. § 2232(a) (1982). B. The National Environmental Policy Act The National Environmental Policy Act of 1969, 42 U.S.C. §§ 4321 to 4347 (1982) (“NEPA”), requires that federal agencies prepare a “detailed statement,” known as an environmental impact statement, for every major federal action “significantly affecting the quality of the human environment.” 42 U.S.C. § 4332(2)(C). The statement must address “any adverse environmental effects” of the project and “alternatives to the proposed action.” Id. We have observed that “[t]he purpose of NEPA is to focus national policymaking on the interdependence between human beings and the environment.” Dunn v. United States, 842 F.2d 1420, 1426 (3d Cir.1988). In Baltimore Gas & Electric Co. v. Natural Resources Defense Counsel, Inc., 462 U.S. 87, 103 S.Ct. 2246, 76 L.Ed.2d 437 (1983), the Supreme Court identified NEPA’s “twin aims.” “First, it ‘places upon an agency the obligation to consider every significant aspect of the environmental impact of a proposed action.’ ” Baltimore Gas, 462 U.S. at 97, 103 S.Ct. at 2252 (quoting Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Counsel, Inc., 435 U.S. 519, 553, 98 S.Ct. 1197, 1216, 55 L.Ed.2d 460 (1978)). “Second, it ensures that the agency will inform the public that it has indeed considered environmental concerns in its decisionmak-ing process.” Id. However, in Vermont Yankee, 435 U.S. at 558, 98 S.Ct. at 1219, the Supreme Court made clear that NEPA imposes duties upon agencies that are “essentially procedural.” See also Strycker’s Bay Neighborhood Council, Inc. v. Karlen, 444 U.S. 223, 228, 100 S.Ct. 497, 500, 62 L.Ed.2d 433 (1980) (per curiam) (holding that NEPA permits a court to ensure that an agency has considered environmental consequences but not to review the weight given to the consequences). The Council on Environmental Quality (“CEQ”) issues general regulations for NEPA, while each agency issues its own implementing regulations. We have determined that the CEQ guidelines are not binding on an agency that has not expressly adopted them. Township of Lower Alloways Creek v. Public Service Electric & Gas Co., 687 F.2d 732, 740 n. 16 (1982). The NRC has acknowledged its obligation to comply with NEPA, however, by issuing regulations governing the consideration of the environmental impact of the licensing and regulatory actions of the agency. See 10 C.F.R. §§ 51.1 (1988); Susquehanna Valley Alliance v. Three Mile Island Nuclear Reactor, 619 F.2d 231, 239 (3d Cir.1980), cert. denied, 449 U.S. 1096, 101 S.Ct. 893, 66 L.Ed.2d 824 (1981). These regulations require the Commission to consider “the alternatives available for reducing or avoiding adverse environmental and other effects.” 10 C.F.R. § 51.71(d) (incorporated by reference in 10 C.F.R. § 51.90). The Commission has also issued a NEPA policy statement mandating consideration of alternative plant designs, discussed infra At 730. In 1971, the Atomic Energy Commission (“AEC”), the predecessor of the NRC, published proposed regulations to implement NEPA which created an accident classification scale, rating classes of accidents according to risk. See Consideration of Accidents in Implementation of the National Environmental Policy Act of 1969, 36 Fed. Reg. 22,851 (1971). Although never promulgated in final form, the NRC followed these proposed regulations throughout the 1970s. In classes one through eight, the AEC placed so-called “design-basis accidents,” those accidents hypothesized on the basis of certain specific failures that the design attempted to protect against. Apart from class-one accidents, which were classified as “trivial,” the AEC required that NEPA reviews for each plant consider the risks from that plant’s design of all of these design-basis accidents. In class nine, the AEC placed severe potential accidents that involved more complex sequences of successive failures, such as a partial reactor core melt accompanied by the production of large quantities of steam and gas or the rupture of the containment vessel. The AEC proposed regulations assumed that a severe accident’s probability was so low that no consideration of its risks or consequences was necessary in the review of individual plants. Early NRC environmental impact statements did not consider the risk of severe accidents, because these proposed NRC regulations classified severe accidents as too unlikely to occur to necessitate consideration of consequences. Courts upheld this policy. See, e.g., Carolina Environmental Study Group v. United States, 510 F.2d 796, 798-800 (D.C.Cir.1975). However, beginning in 1975, new and controversial studies such as the Reactor Safety Study indicated that the risk of severe core accidents was greater than previously thought. See Reactor Safety Study: An Assessment of Accident Risks in U.S. Commercial Nuclear Power Plants, NU-REG-75/014 at 135 (1975). In addition, a new risk assessment method was developed, called Probabilistic Risk Assessment (“PRA”), which involved modeling potential accident sequences and estimating their likelihood. A 1977 study could not determine whether these probability estimates were high or low, but it did determine that enormous possibilities for error existed in the estimates because of the difficulties of the PRA approach. See Risk Assessment Review Group Report to the U.S. Nuclear Regulatory Commission, NUREG/CR-0400 at vi-x, 4-5 (1978). In combination with the March 28, 1979, Three Mile Island Unit 2 accident (“TMI”), which involved a partial core melt, these studies induced the NRC in 1980 to issue a policy statement revoking the earlier classification scheme and requiring consideration of severe accidents in future NEPA reviews, see Statement of Interim Policy, Nuclear Power Plant Accident Considerations Under the National Environmental Policy Act of 1969, 45 Fed.Reg. 40,101 (1980) (“NEPA Interim Policy Statement”). The studies (coupled with TMI) also induced the NRC to expand the scope of emergency planning requirements, see 10 C.F.R. Part 50, App. E (1988), and to initiate a research program into severe accident risks and mitigation alternatives, including a review of Limerick and other facilities located near major population centers. “Filtered-vented containment systems,” which LEA asserts should have been considered in the Limerick FES, are one of the mitigation alternatives studied, having received widespread attention and some acceptance in Europe. See infra at note 11. Following the 1980 actions, the NRC took several actions, ostensibly under the AEA’s required safety procedures that it has subsequently applied to NEPA environmental reviews. First, the NRC set qualitative and quantitative goals for the risk of severe accidents in the Safety Goal Development Program, 48 Fed.Reg. 10,772 (1983) (“Safety Goals Policy Statement”). This Policy Statement indicated that plant designs should ensure that the “likelihood” of an accident resulting in a large-scale core melt is “less than one in 10,000” per year of operation. Id. at 10,775. The NRC refused, however, to adopt quantitative limits on the risk of severe accidents through regulations, because it concluded that PRA techniques “have substantial associated uncertainties” that “raise[] a serious question whether ... [a] quantitative risk goal can be verified with a sufficient degree of confidence.” Id. Instead, the NRC retained its regulatory “defense-in-depth approach ... in order to prevent accidents from happening and to mitigate their consequences” through “[s]iting in less populated areas” and “[ejmergency response capabilities.” Id. Second, the NRC issued a release in which it continued to eschew numerical licensing goals. See Proposed Commission Policy Statement on Severe Accidents and Related Views on Nuclear Reactor Regulation, 48 Fed.Reg. 16,014, 16,016 (1983) (“Proposed Policy Statement”). Because the NRC proposed to find the risk of accident for already designed plants acceptable, it proposed to conclude that the capability of current designs or alternatives “to control or mitigate severe accidents should not be addressed in case-related safety hearings.” Id. at 16,018. However, the NRC did instruct that applications for new construction permits for boiling water reactors should provide for “filtered-vented containment systems, or a variation of such systems ... if these yield a cost-effective reduction in risk.” Id. at 16,019. Commissioners Asselstine and Gilinsky dissented on the ground that the Commission had not received any briefings from the staff regarding the studies initiated in 1980 and hence could not possibly reach any conclusions based on those studies. Id. at 16,023. Subsequently, in its Policy Statement on Severe Reactor Accidents Regarding Future Designs and Existing Plants, 50 Fed. Reg. 32,138 (1985) (“Final Policy Statement”), the Commission concluded that no changes in design regulations were needed, apparently because PRA’s indicated “that existing plants pose no undue risk to public health and safety.” Id. at 32,138. Most important for this case is the fact that, although the Commission stated that “[a] fundamental objective” was “to take all reasonable steps” not only to reduce the chances of a severe accident but also “to mitigate the consequences of such an accident should one occur,” id. at 32,139, the Commission excluded consideration of severe accident mitigation design alternatives from individual licensing proceedings. It concluded that “[individual licensing proceedings are not appropriate forums for a broad examination of the Commission’s regulatory policies relating to evaluation, control and mitigation of accidents more severe than the design basis.” Id. at 32,-144. The Commission did not explain its reasoning, but simply stated that “plant-specific review of severe accident vulnerabilities ... is not considered to be necessary to determine adequate safety or compliance with NRC safety regulations under the Atomic Energy Act.” Id. at 32,138. Commissioner Asselstine again dissented, arguing that the unreliability of PRA analysis precluded confidence in the NRC’s judgment about reactor safety. Although the NRC issued the Final Policy Statement under the Atomic Energy Act, the NRC specifically applied the Statement to exclude environmental considerations under NEPA in the case sub judice. In addition to applying the Final Policy Statement to NEPA reviews, the Commission amended its NEPA regulations in 1984. See Environmental Protective Regulations for Domestic Licensing and Related Regulatory Functions and Related Conforming Amendments, 49 Fed.Reg. 9352 (1984). The amended regulations adopt the CEQ guidelines with a few significant exceptions. The CEQ guidelines, 40 C.F.R. § 1502.22(b) (1985) (superseded by 40 C.F. R. § 1502.22(b) (1987)), state that where an agency acts despite not having access to information regarding the potential adverse impact of its action, the agency “shall include a worst case analysis and an indication of the probability ... of its occurrence.” In its notice amending its NEPA regulations, however, the Commission concluded that because NEPA has been held to be procedural only, and because the Commission viewed the worst-case analysis requirement as substantive, it was not bound by the provision. The NRC also indicated that the 1980 interim NEPA policy on severe accident risks would remain in effect pending further study and experience. See 49 Fed.Reg. at 9356-57. In sum, the NRC originally thought severe accidents too unlikely to justify consideration of their likelihood in reviewing and determining the safety of nuclear plants. It retreated from that viewpoint following the TMI accident and subsequently set safety goals with respect to severe accidents. However, it refused to set quantitative limits; it provided that severe accident mitigation design alternatives should not be studied on a case-by-case basis; and it excluded environmental considerations under NEPA in the case sub judice. III. PROCEDURAL HISTORY The Limerick Generating Station is a nuclear electric generating plant with two light water reactors: Limerick Unit 1 is licensed and running; Limerick Unit 2 is still under construction. PECO received a permit to construct Units 1 and 2 in 1974, and applied for a license to operate Unit 1 in 1981. See 46 Fed.Reg. 42,557 (1981) (notice published by NRC informing the public that it could request a hearing on PECO’s application). In response, the NRC convened a Licensing Board, which conducted hearings lasting ninety-five days and generating a 20,000-page transcript. Much of the activity concerned (1) challenges to the FES prepared by the staff to comply with NEPA and the Radiologic Emergency Response Plan (“evacuation plan”) prepared in conjunction with the Federal Emergency Management Agency and (2) whether the Commonwealth of Pennsylvania had complied with NRC emergency planning regulations. See generally 40 C.F.R. Part 50, App. E (1988). In addition to holding numerous conferences and public hearings during this period, the Licensing Board issued four partial initial decisions (“PIDs”) under 10 C.F.R. § 2.714 regarding (1) supplementary cooling water systems, 17 N.R.C. 413 (1983); (2) on-site emergency planning, environmental and safety contentions, 20 N.R.C. 446 (1984); (3) off-site emergency planning issues, 21 N.R.C. 1219 (1985); and (4) Gra-terford prison emergency planning, 22 N.R. C. 101 (1985), supplemented at 24 N.R.C. 731 (1986). The licensing process culminated in the issuance of an operating license for Limerick I on June 19, 1987. In 1982, LEA applied for and was admitted as an intervenor in the licensing proceedings. LEA filed petitions to review the second PID and the final licensing decision. Thomas Martin, a prisoner at Graterford, filed petitions to review the fourth PID and the final licensing decision. We consider LEA’s petitions for review in Part III and Martin’s petitions in Part IV. Our standard of review of an order granting a nuclear power operating license under section 10(e) of the APA, 5 U.S.C. § 706, is “deferential; that order may not be overturned unless it is found to be ‘arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.’ ” In re Three Mile Island Alert, Inc., 771 F.2d 720, 727 (3d Cir.1985), cert. denied, 475 U.S. 1082, 106 S.Ct. 1460, 89 L.Ed.2d 717 (1986). IV. THE APPEAL OF LIMERICK ECOLOGY ACTION LEA appeals from the Commission’s rejection of three contentions decided in the second PID. The Appeal Board affirmed the second PID, Philadelphia Electric Company (Limerick Generating Station, Units 1 and 2), 22 N.R.C. 681 (1985), as did the NRC itself, by refusing review, 23 N.R.C. 125 (1986). A. Does a Finding of Adequate Protection Under the AEA Preclude Consideration Under NEPA? In rejecting consideration of design alternatives, the Appeal Board concluded that “NEPA could not logically require more than the safety provisions of the Atomic Energy Act.” 22 NRC at 696 n. 10. Before turning to the three contentions addressed in the second PID, we must consider the NRC’s argument that a finding of adequate protection of public health and safety under section 182(a) of the AEA precludes the need for further consideration under NEPA. The NRC seeks to extend that decision here to encompass all three rejected contentions. LEA responds that NEPA imposes additional requirements over and above the AEA, and that the Appeal Board erred in excluding such considerations. The language of NEPA indicates that Congress did not intend that it be precluded by the AEA. Section 102 of NEPA requires agencies to comply “to the fullest extent possible.” 42 U.S.C. § 4332. Although NEPA imposes responsibilities that are purely procedural, see Vermont Yankee, 435 U.S. at 558, 98 S.Ct. at 1219, there is no language in NEPA itself that would permit its procedural requirements to be limited by the AEA. Moreover, there is no language in AEA that would indicate AEA precludes NEPA. The legislative history of the phrase “to the fullest extent possible” indicates that Congress intended that NEPA not be limited by other statutes by implication. The proposed language, which was replaced by “to the fullest extent possible” in the current statute, stated that “nothing in this Act shall increase, decrease or change any responsibility or authority of any Federal official or agency created by any other provision of law.” Conf.Rep. No. 765, 91st Cong., 1st Sess. 9-10, reprinted in 1969 U.S.Code Cong. & Admin.News 2751, 2767, 2770. The Conference Report stated that “[t]he purpose of the new language is to make it clear that each agency of the Federal Government shall comply with the directives [of section 102] unless the existing law applicable to such agency’s operations expressly prohibits or makes full compliance with one of the directives impossible.” Id. The Report concluded that “it is the intent of the conferees that the provision ‘to the fullest extent possible’ shall not be used by any Federal agency as a means of avoiding compliance with the directives set out in section 102.” Id. Hence, the legislative history unequivocally supports LEA’s contention that the AEA cannot preclude application of NEPA by implication. The Commission in the case sub judice does not maintain that the AEA contains express provisions prohibiting compliance with NEPA, nor does it argue that compliance is impossible. Third, courts have repeatedly held that, as suggested by the legislative history, compliance with NEPA is required unless specifically excluded by statute or existing law makes compliance impossible. See, e.g., Public Service Co. of New Hampshire v. NRC, 582 F.2d 77, 81 (1st Cir.) (“The directive to agencies to minimize all unnecessary adverse environmental impact obtains except when specifically excluded by statute or when existing law makes compliance with NEPA impossible.”), cert. denied, 439 U.S. 1046, 99 S.Ct. 721, 58 L.Ed. 2d 705 (1978). Accordingly, “[u]nless there are specific statutory provisions which necessarily collide with NEPA, the Commission was under a duty to consider and, to the extent within its authority, minimize environmental damage____” Public Service, 582 F.2d at 81 (footnote omitted). On the basis, therefore, of the language of NEPA and AEA, the legislative history of NEPA, and the existing case law, we find no intent by Congress that the AEA preclude application of NEPA. Furthermore, the NRC points to no cases indicating that exclusion of consideration of an issue under the AEA requires exclusion of the same issue from consideration under NEPA. As we discussed supra note 7, the case law suggests that where review of environmental implications occurs under a “functional equivalent” of NEPA, the review need not be repeated. But these cases do not suggest that NEPA can never require consideration of additional alternatives simply because there is some overlap in the considerations required by both statutes. Nor do the cases indicate that an issue not considered under the AEA need not be considered under NEPA. In Citizens for Safe Power, Inc. v. NRC, 524 F.2d 1291, 1299 (D.C.Cir.1975), the court indicated that where the concerns under the AEA and NEPA are the same, conclusions reached on the basis of evidence received in “environmental” hearings conducted under NEPA may be applied to “health and safety” considerations under the AEA. As the court stated, to hold otherwise would amount to “stultifying formalism.” Id. The court did not indicate, however, that when issues are excluded from consideration under the AEA they must also be excluded under NEPA. In contrast, the court noted, albeit in dictum, that it is “unreasonable to suppose that [environmental] risks are automatically acceptable, and may be imposed upon the public by virtue of the AEA, merely because operation of a facility will conform to the Commission’s basic health and safety standards.” Id. It is this automatic exclusion which the NRC seeks here and which we refuse to adopt. In its affirmance of the Appeal Board’s decision, the NRC stated that the Commission’s Final Policy Statement “was intended to address both NEPA and Atomic Energy Act reviews____ Insofar as this type of accident mitigation is concerned, NEPA and the Atomic Energy Act reviews are both directed at cost-effective measures to reduce the risk from accidental discharges of radioactive materials, and it would make no sense for the Commission to implement different review policies under the two statutes.” Philadelphia Electric Co., 23 N.R.C. 125, 127 (1986). However, the NRC’s regulations implementing NEPA require it to consider “the alternatives available for reducing or avoiding adverse environmental and other effects.” 10 C.F.R. § 51.71(d) (incorporated by reference in 10 C.F.R. § 51.90). And the Commission’s most recent NEPA policy statement makes clear that among the alternatives to be considered are alternative plant designs. 49 Fed.Reg. 9352, 9354 (1984). Thus, whether or not there are two review policies, the NRC was required to address SAMDAs and cannot now look to sufficiency under the AEA to avoid that obligation. In sum, by whatever route the NRC claims to have determined the environmental impact of Limerick, it has not succeeded, or attempted to succeed, in convincing this Court that the procedural requirements of NEPA have been met. B. Severe Accident Mitigation Damage Alternatives 1. The Commission’s Decision In its second PID, the Licensing Board resolved numerous technical, environmental, and on-site emergency planning issues in favor of PECO, and authorized issuance of a low-power license for Limerick Unit 1. LEA contends that generic staff studies of SAMDAs indicate that such alternatives should be considered in the licensing process. Severe accident mitigation design alternatives are, as the name suggests, possible plant design modifications that are intended not to prevent an accident, but to lessen the severity of the impact of an accident should one occur. The NRC staff prepared a FES for Limerick that sought to comply with NEPA and particularly with the NRC’s 1980 Interim Policy Statement on Nuclear Power Plant Accident Considerations, 45 Fed.Reg. 40,101. The FES stated that, consistent with the interim policy statement, it would include “a reasoned consideration of the environmental risks (impacts) attributable to accidents at the particular facility or facilities within the scope of [the] statement.” FES at 1-3. The FES contained an extensive review of the risk of severe accidents at Limerick, including the results of a PRA performed for Limerick. Nevertheless, although the FES reviewed existing containment mechanisms, see FES at 5-69, it neither considered nor specifically rejected SAMDAs. Instead, the FES concluded: Accident risks from Limerick are expected to be a small fraction of the risks the general public incurs from other sources. Further, the best estimate calculations show that the risks of potential reactor accidents at Limerick are within the range of such risks from other nuclear power plants. Based on the foregoing considerations of environmental impacts of accidents, which have not been found to be significant, the staff has concluded that there are no special or unique circumstances about the Limerick site and environs that would warrant consideration of alternatives for Limerick Units 1 and 2. FES at 5-126. LEA challenged this failure to consider design alternatives. The Licensing Board excluded consideration of mitigation design alternatives, however, on the ground that because the contention was too general, it failed the Commission’s “basis and specificity” requirement under 10 C.F.R. § 2.714(b). See 22 N.R.C. at 693. According to the Licensing Board, LEA failed because it did not identify a “particular, cost-effective design alternative for a particular accident sequence.” Id. The Appeal Board concluded that LEA’s contentions met the basis and specificity requirement because they were based on the studies sponsored by the NRC, which included preliminary conclusions regarding design mitigation alternatives at Limerick itself. Philadelphia Electric Co., 22 N.R.C. at 693-94. In fact, the NRC has studied, designed and assessed the costs of mitigation systems. The Appeal Board upheld the exclusion of design mitigation alternatives, however, holding that the contention was precluded by the Final Policy Statement. Id. at 693-96. The Appeal Board explicitly relied on the Final Policy Statement’s directive that “severe accident mitigation measures, beyond already existing Commission requirements, ‘should not be addressed in case-related safety hearings.’ ” Id. at 696-97 (quoting Final Policy Statement, 50 Fed.Reg. at 32,-145). The Board noted that because the Final Policy Statement found that existing plants posed no undue risk to the public health and safety and that research was ongoing, the policy statement precluded review of design alternatives. As we have explained above, the Board also noted that there is “Commission precedent holding that NEPA could not logically require more than the safety provisions of the Atomic Energy Act, and court precedent recognizing the inherent interrelationship of these statutes and thus issues raised under each.” 22 N.R.C. at 696 n. 10 (citing Public Service & Electric Co., 9 N.R.C. 14, 39 (1979); Citizens for Safe Power, Inc. v. NRC, 524 F.2d 1291, 1299-1300 (D.C.Cir.1975)). Additionally, the Appeal Board reasoned that although SAMDAs were being excluded from consideration in the Limerick licensing proceedings, they were not being ignored. Although the Board did not state or even intimate that it would, without the Final Policy Statement, reject the design alternatives contention, it pointed to several additional factors supporting its decision. The Board noted that extensive research into design alternatives was ongoing. In addition, the Board pointed out that although the FES did not consider design alternatives for Limerick, it did consider severe accidents, concluding that “there are no special or unique circumstances about the Limerick site,” 22 N.R.C. at 697 (citing FES at 5-126), and maintaining that because the consideration of a severe accident was done at the Commission’s discretion (and was not required), “[a] fortiori, consideration of possible design alternatives to mitigate a severe accident is not required either.” Id. In sum, the Appeal Board upheld the decision not to consider design alternatives on the grounds that the Policy Statement’s conclusion that existing designs were sufficiently safe to exclude consideration of alternatives controlled its decision; that NEPA “could not logically require more than the safety provisions of the Atomic Energy Act”; that ongoing studies were still considering design alternatives; and that the FES’s consideration of severe accidents (although not of design alternatives) was sufficient. The NRC refused review, thus affirming the decision of the Appeal Board. In the order declining review, the Commission briefly stated that the Final Policy Statement was intended to address both NEPA and AEA reviews. The Commission stated that “[ijnsofar as this type of accident mitigation is concerned, NEPA and Atomic Energy Act reviews are both directed at cost-effective measures to reduce the risk from accidental discharges of radioactive materials, and it would make no sense for the Commission to implement different review policies under the two statutes.” 23 N.R. C. at 127. Commissioner Asselstine dissented. Id. at 128. He noted that the order extended the Final Policy Statement’s requirement of exclusion of design alternatives under the AEA to exclusion also under NEPA. He also noted that the NRC’s estimate of the chance of an accident at least as severe as TMI occurring in the next 20 years is “50-50.” According to Commissioner Asselstine, “a 50-50 chance within the next twenty years is [not] an acceptable level of risk,” particularly for plants such as Limerick and Indian Point, which are located in densely populated areas. Id. at 129. 2. The Propriety of General Exclusion by the Final Policy Statement The parties do not dispute that the Appeal Board excluded consideration of design alternatives on the basis of the Final Policy Statement, and that the NRC’s opinion affirmed on this ground. See 23 N.R.C. at 126. On appeal, LEA and the Commonwealth of Pennsylvania in its amicus brief acknowledge that after Baltimore Gas, the NRC may preclude consideration of generic issues by rulemaking. They contend, however, that because the Final Policy Statement was a “policy statement” as opposed to a “rule” and because it seeks to apply to all plants yet does not concern a generic issue, it does not have the force of a rule-making and should not be permitted to exclude consideration of mitigation alternatives. The distinction is important: courts have repeatedly held that if an agency action is merely a policy statement, “[w]hen the agency applies the policy in a particular situation, it must be prepared to support the policy just as if the policy statement had never been issued.” Pacific Gas & Electric Co. v. Federal Power Commission, 506 F.2d 33, 37 (D.C.Cir.1974). a. The NRC’s Pronouncement: Policy Statement or Rule? We begin by briefly reviewing the process by which the NRC arrived at its Final Policy Statement. On October 2, 1980, the NRC published The Advanced Notice of Proposed Rulemaking, 45 Fed.Reg. 65,474 (1980), in which it indicated that a long-term rulemaking effort was being initiated that would establish policy goals and requirements relating to core-melt accidents greater than the “design basis.” The NRC retreated from this formal rulemaking proceeding, however, in its Proposed Policy Statement, 48 Fed.Reg. 16,014 (1983). In the Proposed Policy Statement, the Commission proposed to replace long-term generic rulemaking in favor of future “rule-makings on specific standard plant designs and regulatory decisions on other classes of existing or future plants which may, or may not, include rulemaking.” Id. at 16,-014. Instead of a severe accident rulemak-ing, the NRC stated that it would promulgate rules for future plants only, and that for operating plants or those under construction it would exclude consideration of design alternatives as a matter of Commission policy while research into design alternatives was ongoing. The proposed exclusion of design alternatives was unequivocal, although it was characterized as a policy statement, not a rule. The Final Policy Statement revised the Proposed Policy Statement, but on the issue of design alternatives for existing plants or those under construction, the NRC reiterated its refusal to consider such alternatives in individual licensing proceedings. Again, the NRC asserted that it was excluding such considerations on the basis of its “policy statement,” and stated that it “sees no present basis for immediate action on generic rulemaking [such as a rule requiring certain mitigation design alternatives] or other regulatory changes for these plants because of severe accident risk.” 50 Fed.Reg. at 32,138. Because the NRC has repeatedly characterized its exclusion of design alternatives as a policy statement, not a rulemaking, we must determine the impact of that determination on the validity of the exclusion in the case sub judice. The agency’s label of an agency action, although one factor to be considered, does not control whether the action is in fact a rulemaking. Cerro Metal Products v. Marshall, 620 F.2d 964, 981-82 (3d Cir.1980). Instead, “it is the substance of what the [agency] has purported to do and has done which is decisive.” Columbia Broadcasting System v. United States, 316 U.S. 407, 416, 62 S.Ct. 1194, 1200, 86 L.Ed. 1563 (1942); see also Lewis-Mota v. Secretary of Labor, 469 F.2d 478, 481-82 (2d Cir.1972) (“[T]he label that the particular agency puts upon its given exercise of administrative power is not, for our purposes, conclusive; rather it is what the agency does in fact.”). In Pacific Gas & Electric, 506 F.2d at 37, the District of Columbia Circuit recognized that the distinction between substantive rules and policy statements is not altogether clear, but noted several distinctions between the two. First, the court noted that a policy statement differs, because “[i]t is not finally determinative of the issues or rights to which it is addressed.” Id. at 38. Here, however, the Final Policy statement purports finally to preclude consideration of design alternatives in individual licensing proceedings. See American Business Assoc. v. United States, 627 F.2d 525, 531 (D.C.Cir.1980) (where a statement does not allow an official to exercise discretion, it is a rule rather than an interpretation); see also J. O’Reilly, Administrative Rulemaking § 3.08, at 59 (1983) (“Agency discretion and its absence is one of the measures of a rule’s substantive character.”). Second, the Pacific Gas court determined that an action is a rule where “not subject to challenge in particular cases.” 506 F.2d at 39. Cf. Ryder Truck Lines, Inc. v. United States, 716 F.2d 1369, 1377-78 (11th Cir.1983) (where an action sets forth a rebuttable presumption which may be challenged in individual proceedings, it is not a binding norm), cert. denied, 466 U.S. 927, 104 S.Ct. 1407, 1408, 80 L.Ed.2d 181 (1984). Obviously, the intent of the Final Policy Statement at issue here is to exclude challenges in particular licensing decisions. Third, as the Pacific Gas court noted, policy statements are not subject to the notice and comment requirements of the APA. The Final Policy Statement, however, did undergo notice and comment. Thus, although described as a policy statement, the NRC appears to have intended the Final Policy Statement to have the effect of a substantive rule, i.e., it appears that it was intended to be: (1) finally determinative as to the issue of design alternatives for Limerick; (2) not subject to challenge in the individual licensing proceeding; and (3) subjected to notice and comment procedures. Although the Final Policy Statement more closely approximates a substantive rule than a policy statement, the NRC asserts before this Court that it did not promulgate a binding rule: Because the policy statement was not promulgated by the Commission as a binding rule, petitioners were free (and arguably obliged) to challenge it when the Appeal Board applied the policy statement to affirm the Licensing Board’s exclusion of their contention. Because the Commission does not argue that the policy statement was immune to challenge in individual licensing proceedings, as a rule would have been, the petitioners’ various arguments that turn on the fact that the policy statement was not a rule are simply irrelevant. NRC Br. at 33 n. 16. The NRC cannot have it both ways, however. The Appeal Board decision and the subsequent order of the full Commission unequivocally indicated that the issue of mitigation design alternatives could not be challenged in the Limerick licensing proceeding. It is plain that, notwithstanding the Commission’s protestation to the contrary in its brief, the Commission did, in fact, rely on the prior statement itself without examining the substantive arguments for considering design alternatives. We conclude that the Final Policy Statement here should not be accorded the stature of a rule. First, it was described as a policy statement by the NRC in the proposed and final Federal Register notices. An informed public interest group such as LEA, aware of the notice, and also (presumably) aware of the substantial precedent that policy statements cannot preclude consideration in individual licensing proceedings, might reasonably have been led to concentrate scarce resources on a challenge to a decision in a specific licensing proceeding such as this one rather than to a generic policy statement which, by definition, can have no binding effect. The NRC’s assertion that the “policy statement was not promulgated ... as a binding rule” indicates in no uncertain terms the stature that the Commission accords the Final Policy Statement before this court. We cannot hold the public to a higher standard of divining the actual function of the Statement. Moreover, it is uncertain whether, even if LEA had sought judicial review of the Final Policy Statement at the time of its promulgation, review would have been permitted. General policy statements, because they are ineffective except as applied and defended in specific proceedings, are often insulated from judicial review at the time of issuance. See, e.g., Regular Common Carrier Conference v. United States, 628 F.2d 248, 252 (D.C.Cir.1980). Although it is uncertain whether, because the Final Policy Statement sought to foreclose consideration of design alternatives in all proceedings, the statement could have been challenged in court at the time of its publication, a reviewing court might reasonably have concluded that because the NRC chose to proceed under the “policy statement” rubric, it intended to defend the policy where challenged in licensing proceedings and hence a generic challenge would be “premature.” National Association of Insurance Agents, Inc. v. Board of Governors, 489 F.2d 1268, 1271 (D.C.Cir.1974). At all events, we need not decide whether the Final Policy Statement could have been challenged at the time of issuance because it is contrary to the intent of the APA to force the public to divine the obfuscated intention of the NRC. We conclude that the NRC’s Final Policy Statement is entitled to no greater deference than any other policy statement, i.e., none. b. The NRC’s Failure to Give Careful Consideration to SAMDAs In Baltimore Gas & Electric Co. v. Natural Resources Defense Council, 462 U.S. 87, 103 S.Ct. 2246, 76 L.Ed.2d 437 (1983), the Supreme Court upheld an NRC rule establishing that licensing boards should assume for the purposes of individual licensing decisions that the permanent storage of radioactive wastes would have no significant environmental impact and could not affect the decision whether to license a particular nuclear plant. First adopted after “extensive informal rulemaking proceedings” and later finalized after formal rulemaking proceedings, see 44 Fed.Reg. 45,362 (1979), the NRC's “zero-release” assumption, as indicated in a table compiling the environmental effects of nuclear fuel storage, dictated that “ ‘[n]o further discussion of such environmental effects shall be required.’” Baltimore Gas, 462 U.S. at 93, 103 S.Ct. at 2250 (quoting 42 Fed.Reg. 13,806 (1977)). The Court of Appeals for the District of Columbia Circuit had held that the rules were arbitrary and capricious and not in accordance with NEPA, because the NRC had failed to include the uncertainties surrounding the zero-release assumption into the licensing process in a way that the uncertainties could affect the outcomes of any particular licensing decision. The D.C. Circuit invalidated the rule adopting the zero-release assumption in part because the rule prevented the uncertainties in the assumption from affecting any individual licensing decision, but did not find the environmental effects to be insignificant as required by NEPA. The Supreme Court’s reversal was predicated upon the fact that the zero-release assumption was within the bounds of reasoned decisionmaking under the APA. Although the Commission did not allow the uncertainties involved in making the zero-release assumption to affect any particular licensing, the Court determined that through the extensive rulemaking it “made the careful consideration and disclosure required by NEPA.” Id. at 98, 103 S.Ct. at 2253. The Court held that exclusion by a generic rule was appropriate because, since the rule was predicated upon storage of wastes at a “common long-term repository” (bedded-salt repositories), “[t]he environmental effects of much of the fuel cycle are not plant specific.” Id. at 101, 103 S.Ct. at 2254. According to the Court, the generic nature of the problem therefore could be regulated efficiently and consistently by a generic proceeding. The Court then concluded that because the zero-release assumption, when read in context with the other assumptions made in the rule, represented a reasonable statement of risks and because the predictions of environmental effects were scientific judgments within the Commission’s “special expertise,” the zero-release assumption was not arbitrary and capricious. Id. at 103, 103 S.Ct. at 2255. Baltimore Gas is readily distinguishable from the NRC’s decision not to consider SAMDAs, however. First, in Baltimore Gas the NRC proceeded on the basis of a formal rulemaking, whereas in this case, as the discussion supra explains, the NRC promulgated the decision not to consider mitigation alternatives as a policy statement. To the extent we hold the NRC to its unequivocal language in the Final Policy Statement and on review, the NRC cannot simply reject consideration of design alternatives, but must fully defend its reasons for doing so. This the Licensing Board, Appeal Board, and full Commission explicitly declined to do. ASLB Order (unpublished), slip op. at 1, 3 (April 20, 1984); 22 N.R.C. at 693-96; NRC Order, slip op. at 2 (March 2, 1986). As we discussed above, numerous courts have held that policy statements must be supported just as if they did not exist. See Pacific Gas & Electric Co., 506 F.2d at 38. Even the rulemaking in Baltimore Gas, 462 U.S. at 98-99, 103 S.Ct. at 2252-53, which, as a rulemaking, was not as vulnerable to challenge, reflected “the careful consideration and disclosure required by NEPA.” We therefore review whether the FES or the Final Policy Statement itself provides support for the exclusion of SAM-DAs. First, to qualify, the FES must contain sufficient discussion of the relevant issues and opposing viewpoints to enable the decisionmaker to take a “hard look” at the environmental factors and to make a reasoned decision. Kleppe v. Sierra Club, 427 U.S. 390, 410 n. 21, 96 S.Ct. 2718, 2730 n. 21, 49 L.Ed.2d 576 (1976). The impact statement must be sufficient to enable those who did not have a part in its compilation to understand and consider meaningfully the factors involved. Environmental Defense Fund, Inc. v. Corps of Engineers, 492 F.2d 1123, 11367 (5th Cir.1974). Cf. Dunlop v. Bachowski, 421 U.S. 560, 572, 95 S.Ct. 1851, 1860, 44 L.Ed.2d 377 (1975) (noting that a statement by an agency of the reasons for its determination is crucial to effective judicial review). Here, as we discussed supra At 733, the FES neither considered nor specifically rejected SAMDAs. Second, preclusion from consideration in individual licensings must be based on a judgment that the issue could not affect the decisions therein. However, the policy statement at issue excludes consideration of design alternatives without making any conclusions about the effectiveness or cost of any particular alternative. Nor does the Final Policy Statement indicate that the NRC has determined that current plants are so safe that no design mitigation could be worthwhile. Rather, the Commission merely states its belief that SAMDAs “should not be addressed in case-related safety hearings.” 50 Fed.Reg. at 32,145. Yet, this conclusion is undercut by the Commission’s statement earlier in the Final Policy Statement that the NRC “intends to take all reasonable steps ... to mitigate the consequences of [a severe] acci-dent____” Id. at 32,139. The NRC further pointed out that “the occurrence of a severe accident is more likely at some plants than at others. At each plant there will be systems, components or procedures that are the most significant contributors to severe accident risk,” id., implying that SAM-DAs should most appropriately be considered on a case-by-case basis. It would seem, even on the Commission’s own terms, that a failure to consider SAMDAs in the Limerick proceeding could affect the final decision and, therefore, that preclusion from consideration was an abuse of discretion. Third, it is axiomatic that the generic approach of Baltimore Gas will not suffice where the underlying issues are not generic. In Baltimore Gas, the rulemaking concerned an unquestionably generic issue: the environmental effect of the long-term storage of nuclear waste from all plants, assuming that those wastes are stored not at individual plants, but in a “common long-term repository.” Baltimore Gas, 462 U.S. at 101, 103 S.Ct. at 2254. Given that the NRC’s long-term storage plans did not provide for permanent storage at each individual site, the common storage of the wastes in bedded-salt repositories provided a generic basis for the regulation because the effect of long-term storage could be expected to arise from the situs of the waste, rather than from the particular characteristics of the plants at which the waste was generated. The Commission’s assertion as to the generic nature of severe accident risks would be analogous to Baltimore Gas only if the “zero-release” assumption in Baltimore Gas assumed that nuclear fuel would be stored at each individual plant permanently. However, to assume that in such a situation the NRC could permissibly promulgate a generic rule, as the NRC would have us conclude, would render meaningless the discussion in Baltimore Gas regarding the importance of common storage to the promulgation of a generic rule. As the Commission itself has noted, the impact of SAMDAs on the environment will differ with the particular plant’s design, construction and location. Moreover, as a logical proposition, because risk equals the likelihood of an occurrence times the severity of the consequences, see Baltimore Gas, 462 U.S. at 104-05, 103 S.Ct. at 2255-56 (quoting NU-REG-0116 at 2-11), even assuming that all plants are of exactly equal design and construction, which they obviously are not, the risk will vary with the potential consequences. Because the potential consequences will largely be the product of the location of the plant, the risk will vary tremendously across all plants. As the NRC itself has noted, “the population distribution in the vicinity of the site affects the magnitude and location of potential consequences from radiation releases.” 48 Fed.Reg. at 16,020. This is particularly true for plants such as Limerick which were built near densely populated areas. The FES noted that “the [Limerick] site is substantially higher than average in terms of population density,” FES at 1-4, and that approximately seven million persons are expected to live within fifty miles of Limerick in the year 2000. Id. at 5-98. In other words, the same probability of the same accident in a plant such as Limerick will produce a higher risk than that produced by the same accident at a plant not located within twenty-five miles of a major metropolitan area. Therefore, it is unlikely that severe accident mitigation can be treated as a generic issue. In fact, in its conclusion that severe accidents pose “no undue risk,” the Final Policy Statement did not take into account the added risks of a plant located in a densely populated area. Since the Commission never addressed this added risk problem in either the licensing proceeding or the Final Policy Statement, the record does not support the NRC’s assertion that severe accident potential may be treated as a generic issue. To summarize, the policy statement was not a rulemaking and therefore did not absolve the NRC of the required consideration of environmental effects. We conclude that the FES failed adequately to consider SAMDAs and, therefore, the deci-sionmaker did not take the requisite “hard look” at SAMDAs. We further conclude that a decision with respect to SAMDAs could affect the final decision and therefore preclusion of consideration of SAMDAs was inappropriate. Finally, on the record, the underlying issue of SAMDAs may not be treated as a generic issue and therefore summary treatment of SAMDAs was inappropriate. 3. Is the Risk Remote and Speculative? The NRC asserts that, even if we conclude that the exclusion from consideration under the Final Policy Statement was otherwise improper, we should uphold the Appeal Board’s decision because the risks of severe accidents are “remote and speculative.” It is undisputed that NEPA does not require consideration of remote and speculative risks. See, e.g., San Luis Obispo Mothers for Peace v. NRC, 751 F.2d 1287, 1300-01 (D.C.Cir.1984), rehearing en banc granted on other grounds, 760 F.2d 1320 (D.C.Cir.1985), aff'd on rehearing en banc, 789 F.2d 26 (D.C.Cir.), cert. denied, 479 U.S. 923, 107 S.Ct. 330, 93 L.Ed.2d 302 (1986); Carolina Environmental Study Group v. United States, 510 F.2d 796, 799 (D.C.Cir.1975). Neither the Appeal Board nor the Commission based its decision on the belief that the risk of severe accidents is remote and speculative, however, and “argument by counsel cannot take the place of an agency’s statement of reasons or findings.” WAIT Radio v. FCC, 418 F.2d 1153, 1158 (D.C.Cir.1969). Accord Sierra Club v. EPA, 719 F.2d 436, 466 (D.C.Cir.1983), cert. denied, 468 U.S. 1204, 104 S.Ct. 3571, 82 L.Ed.2d 870 (1984). See also SEC v. Chenery, 332 U.S. 194, 196, 67 S.Ct. 1575, 1577, 91 L.Ed. 1995 (1947) (agency action must be judged solely on grounds invoked by the agency). The Appeal Board determined that design alternatives could be excluded under the AEA and that NEPA could require no more. After concluding that the AEA satisfied NEPA concerns, the Appeal Board added that further studies were ongoing which might indi