I concur in Parts I and II of Judge Bork’s opinion and in the result reached by Part III. I write separately, however, to emphasize my understanding of the showing necessary before a court may supplement the record by examining transcripts of a closed commission hearing. I cannot accept the view, supra at 44-45, that a petitioner asking a court to review transcripts of this nature must always make a prior and independent showing of agency wrongdoing before the court will examine the transcripts. Like a bank policy of offering loans only to borrowers who do not need loans, this view suggests that the transcripts can serve only as cumulative evidence to support a claim already independently established. The plurality claims support for this notion from the Supreme Court’s observation that a strong showing of bad faith or improper behavior must precede examination of the administrative decisionmaker’s mental processes. Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 91 S.Ct. 814, 28 L.Ed.2d 136 (1971). Overton Park based this rule, however, on the availability of administrative findings “made at the same time as the decision.” Id. at 420, 91 S.Ct. at 825. The Court then immediately added, “But here there are no such formal findings and it may be that the only way there can be effective judicial review is by examining the decisionmakers themselves.” Id.
The Overton Park rule protects the administrative decisionmaker's mental processes from routine judicial scrutiny. In other appropriate cases, a court may examine agency transcripts without entering this sensitive terrain of the decisionmaker’s thought patterns. The most fruitful yield from agency transcripts may well be information about the tangible ingredients that entered the agency’s decision, not inferences about the decisionmaker’s biases, motivations, and human weaknesses.
In reconciling the concerns of Overton Park with the potential uses of agency transcripts, my view is that the showing demanded of a petitioner is an allegation, strongly supported by the record, affidavits, and specific references to the transcripts, that the agency has acted in bad faith or with improper purpose. If the court decides that review of the transcripts is called for, it can devise in camera procedures to ensure the sanctity of the administrative process. But the plurality’s attempt to safeguard agency deliberations by an absolute judicial refusal to inspect transcripts at the threshold of inquiry sweeps too broadly. In practical effect, the plurality’s rule may deprive petitioner of the only available, and perhaps the most complete, evidence of agency wrongdoing. Moreover, it creates incentives for concealment from the public and reviewing courts by announcing to the agency that any improp*214er actions taken during their proceedings will be unreviewable so long as no tangible evidence of these improper actions escapes from the meeting room. Only by leaving open the possibility of at least in camera judical inspection of transcripts can the legitimate interests of petitioners and the public in judicial review of agencies and their procedures be vindicated. By stating-that a petitioner must produce a “smoking gun” before the court will even look at closed agency proceedings, the plurality insulates the agency’s deliberations far beyond the protection contemplated by Over-ton Park or any other controlling precedents.
In this case, I agree with the plurality that petitioners’ request to supplement the record with the transcripts must be denied. Petitioners have not made nearly substantial enough a showing to warrant inclusion of the transcripts in the record. Neither petitioners’ allegations nor the portion of the transcripts disclosed to us suggests that the agency acted in bad faith in excluding earthquakes from its consideration of emergency planning at Diablo Canyon. For this reason, I concur in the result.
WALD, Circuit Judge, with whom SPOTTSWOOD W. ROBINSON, III Chief Judge, and J. SKELLY WRIGHT and GINSBURG, Circuit Judges, join, dissenting:Today a majority of the en banc court upholds the Nuclear Regulatory Commission’s conclusion that its emergency planning regulations neither require nor permit consideration of earthquakes as events which could initiate or complicate an accident at a nuclear power plant located three miles from an active fault.1 The court reaches this conclusion by narrowly limiting its focus to Commission decisions dealing with earthquakes and emergency planning and by relying on a deceptive set of calculations based on the numerical probabilities of a radiological emergency and an earthquake occurring simultaneously at the Diablo Canyon plant. This case must, however, be viewed in the context of the broader purposes of emergency planning and against the backdrop of the Commission’s other emergency planning decisions. The majority’s opinion upholding the Commission’s rationale in toto effectively nullifies emergency planning as an effort to predict what actions would be needed should an accident occur. Instead, under the majority’s view, the Commission can avoid the public commitment it made in its regulations simply by declaring that the probability of any accident occurring is too low to bother with. Had the majority surveyed the wider picture, it should have realized that the Commission's decision is woefully inadequate as a piece of logical reasoning, lacking in record support, and patently inconsistent with other interpretations of the emergency planning regulations as well as the underlying purposes of those regulations.2
I. Scope of Review
Petitioners argue that the effect of earthquakes on emergency planning at Dia*215blo Canyon is a “material safety issue" requiring a hearing under § 189(a) of the Atomic Energy Act as interpreted in Union of Concerned Scientists v. NRC, 735 F.2d 1437 (D.C.Cir.1984), cert. denied, — U.S. -, 105 S.Ct. 815, 83 L.Ed.2d 808 (1985). Because the materiality of a safety issue is in large part a function of the Commission’s regulations,3 this claim reduces to a charge that the NRC’s interpretation and application of its emergency planning regulations is arbitrary and capricious. Cf. GUARD v. NRC, 753 F.2d 1144, 1150 (D.C.Cir.1985). As the majority rightly points out, this court must normally defer to an agency’s interpretation of its own regulation “ ‘unless it is plainly erroneous or inconsistent with the regulation.’ ” United States v. Larionoff 431 U.S. 864, 872, 97 S.Ct. 2150, 2155, 53 L.Ed.2d 48 (1977) (citation omitted). This rule does not, however, mean that the court abdicates its responsibility to assess the arbitrariness and capriciousness of agency action. Instead, it serves to focus the court’s inquiry on several discrete tasks of which one, but only one, is to assure that the interpretation is consistent with the language of the regulation.
A court also needs to assure itself that the regulations are “consistent with the statute under which they are promulgated.” Id. at 873, 97 S.Ct. at 2156. This statutory inquiry is difficult to conduct for NRC regulations, however, because the Atomic Energy Act creates “a regulatory scheme that ... is ‘virtually unique in the degree to which broad responsibility is reposed in the administrative agency, free of close prescription in its charter as to how it shall proceed in achieving the statutory objectives.’ ” Carstens v. NRC, 742 F.2d 1546, 1551 (D.C.Cir.1984) (citation omitted), cert. denied, — U.S.-, 105 S.Ct. 2675, 86 L.Ed.2d 694 (1985). The Act does not, of course, specifically mention emergency planning, let alone earthquakes.
A third task for a court reviewing the NRC’s regulatory interpretations is to inquire into their consistency with both the stated purposes of the regulations and other interpretations of the same regulations. The Supreme Court has held that the Commission’s interpretation of its regulations must be accepted only if it “sensibly conforms to the purpose and wording of the regulations” and is “consistent with prior agency decisions.” Northern Indiana Public Service Co. v. Porter County Chapter of the Izaak Walton League of America, 423 U.S. 12, 14-15, 96 S.Ct. 172, 173-74, 46 L.Ed.2d 156 (1975) (assessing Atomic Energy Commission regulations). A reviewing court thus must “determine if [the regulatory interpretation] is consistent with ... the purpose which the regulation is intended to serve.” Cheshire Hospital v. New Hampshire-Vermont Hospitalization Service, Inc., 689 F.2d 1112, 1117 (1st Cir.1982). It must also ensure that the agency has treated like cases similarly or provided a reasoned explanation for any variations. Airmark Corp. v. FAA, 758 F.2d 685, 691-92 (D.C.Cir.1985).
Finally, a reviewing court must insist that the agency provide a clear explanation of the factual and policy bases for its regulatory interpretation. Errors rendering an action arbitrary and capricious cannot even be spotted unless the agency has “articulate[d] a satisfactory explanation for its action including a ‘rational connection between the facts found and the choice made.’ ” Motor Vehicle Manufacturers Association v. State Farm Mutual Automobile Insurance Co., 463 U.S. 29, 43, 103 S.Ct. 2856, 2866, 77 L.Ed.2d 443 (1983) (citation omitted). As this court regularly points out, “[a]lthough our judicial duties demand great deference to agency expertise, we cannot defer, indeed we cannot *216even engage in meaningful review, unless we are told which factual distinctions separate arguably [similar situations], and why those distinctions are important.” Public Media Center v. FCC, 587 F.2d 1322, 1331 (D.C.Cir.1978). An agency cannot fulfill the requirement of an adequate explanation merely by insisting that its conclusions are rational and supported by the record. Instead, it must give the court “the rationale underlying the importance of factual distinctions as well as the factual distinctions themselves.” Id. at 1332.
The NRC’s order excluding earthquakes from consideration in the emergency planning for Diablo Canyon cannot stand up to this kind of scrutiny. The NRC has totally defaulted in providing any satisfactory explanation of why it interpreted the emergency planning regulations to exclude all consideration of earthquake complications in the emergency planning for a facility whose design proceedings centered on the plant’s proximity to an active earthquake fault. In particular, many of the rationales the NRC has provided are inconsistent with the stated purpose of the emergency planning regulations and with earlier and later applications of those regulations.
II. Ignoring the Purposes of Emergency Planning
The majority never faces up to petitioners’ most telling argument: that the Commission’s interpretation of its emergency planning regulations is inconsistent with the fundamental purposes of those regulations. Petitioners’ Br. at 16-17; see Diablo Canyon, 20 N.R.C. at 262 (Commissioner Asselstine, dissenting) (Commission’s decision is inconsistent with the regulations’ “judgment that adequate emergency planning is an essential element in protecting the public health and safety independent of the Commission’s other regulations and safety reviews focusing on the design of the plant itself’ ”). The requirement that a regulatory interpretation be consistent with the original reasons for promulgating the regulations must be strictly enforced in order to prevent agencies from making major changes in regulatory direction sub silentio — as the Commission appears to be doing here.
A. The Purposes of the Regulations
When the emergency planning regulations were issued in 1980, they reflected a major shift from the Commission’s conception of emergency planning prior to the accident at the Three Mile Island (“TMI”) nuclear power plant. As the NRC explained when it proposed the regulations,
The proposed rule is predicated on the Commission’s considered judgment in the aftermath of the accident at Three Mile Island that safe siting and design-engineered features alone do not optimize protection of the public health and safety---- Emergency planning was conceived as a secondary but additional measure to be exercised in the unlikely event that an accident would happen. The Commission’s perspective was severely altered by the unexpected sequence of events that occurred at Three Mile Island. The accident showed clearly that the protection provided by siting and engineered safety features must be bolstered by the ability to take protective measures during the course of an accident.
44 Fed.Reg. 75,167, 75,169 (1979). These views were echoed in the preamble accompanying the final regulations. 45 Fed.Reg. 55,402, 55,403 (1980).
In adopting these emergency planning regulations, the NRC espoused the “fundamental philosophy” of emergency planning first proposed by the Kemeny Commission in the aftermath of the TMI accident. This philosophy requires all relevant actors to “do everything possible to prevent accidents of this seriousness, but at the same time [to] assume that such an accident may occur and be prepared for response to the resulting emergency.” Report of the President’s Commission on the Accident at Three Mile Island 17 (1979), reprinted in Record Volume 69, Joint Intervenors’ (J.I.) *217Exhibit 114A (for identification only).4 A previous Commission interpretation of the regulations similarily noted that “[t]he underlying assumption of the NRC’s emergency planning regulations in 10 CFR § 50.47 is that, despite application of stringent safety measures, a serious nuclear accident may occur.” Southern California Edison Co. (San Onofre), 17 N.R.C. 528, 533 (1983), vacated on other grounds sub nom. GUARD v. NRC, 753 F.2d 1144 (D.C.Cir.1985). Basing the regulations on the assumption that an accident can occur despite other safeguards highlights the importance of “emergency planning as equivalent to, rather than secondary to, siting and design in public protection.” 44 Fed. Reg. at 75,169. These contemporaneous statements of the purpose of the regulations, and not the NRC’s later representation to this court that “emergency planning [is] a backstop rather than a front-line defense,” NRC Br. at 26, must guide the interpretation of the emergency planning regulations.
B. Inconsistencies with the Regulation’s Purposes
Two of the most important reasons for the exclusion of earthquake effects from emergency planning cited by the majority and the Commission are inconsistent with the stated purposes of the regulations. The first is the Commission’s conclusion that earthquakes smaller than the Safe Shutdown Earthquake (“SSE”) need not be considered as initiators of accidents because “the seismic design of the plant was reviewed to render extremely small the probability that such an earthquake would result in a radiologic release.” 20 N.R.C. at 251. The second notion relied upon primarily, by the majority, is that the likelihood of a radiological accident occuring at all is a relevant concern in deciding the scope of emergency planning, even though such planning begins with the assumption that an accident — unthinkable as it may be — mil in fact occur.
Nuclear power plants are designed to withstand a variety of severe natural phenomena, including earthquakes. See 10 C.F.R. pt. 50, app. A, criterion 2 (1985). If the operation of these plants always lived up to their designers’ hopes, no emergency planning would ever be necessary. The entire thrust of the change in philosophy reflected in the 1980 regulations, however, was to build emergency planning around the assumption “that, despite application of stringent safety measures, a serious nuclear accident may occur.” Southern California Edison Co., 17 N.R.C. at 533. The regulations are designed to ensure that “the protection provided by siting and engineered design features [will] be bolstered by the ability to take protective measures during the course of an accident." 45 Fed. Reg. 55,402, 55,403 (1980) (preamble to final regulations). The NRC cannot now, consistently with the stated purpose of its regulations, interpret those regulations to exclude altogether consideration of accidents initiated by earthquakes solely on the ground that Diablo Canyon’s design makes such accidents highly improbable. Cf. GUARD v. NRC, 753 F.2d 1144, 1149-50 (D.C.Cir.1985) (finding interpretation to be irrational in part because it was based on “an assumption [not] properly indulged in an emergency preparedness regulation”). But see infra at 51 & n.7 (NRC may exclude earthquakes as initiators because the probability of occurrence is small). In so doing, the Commission is engaging in circular reasoning, since the very purpose of the exercise is to plan for the unthinkable eventuality that the design safeguards will not prevent an accident.
The majority and the Commission, see infra at 52 & n.10, repeat this mistake when they factor in the probability of a radiological accident occurring at all when evaluating whether earthquakes occur frequently enough to merit consideration as complicating factors. Here again, emergency planning starts from the assumption that an accident has already occurred. Ob*218viously the probability of an adverse event, such as heavy rain or an earthquake, complicating emergency planning after an accident must be taken into account when deciding whether to plan for such an eventuality. But the pertinent probability is that of the complicating, event alone, not multiplied by the probability of a radiological accident. The majority eventually does compare the independent probabilities of earthquakes and other allegedly frequently occurring phenomena, maj. op. at 42, but only after downplaying the frequency of earthquakes by multiplying their probability by the probability of a radiological accident, maj. op. at 39-40. This calculation, stressed in the majority’s opening paragraph, produces deceptively low figures in large part because the low statistical probability of a radiological accident ever occurring insures that the simultaneous occurrence of an accident and any complicating factor is concededly an extremely unlikely event. See infra at 55. Thus, “[t]he probability arguments used by the Commission are really arguments that we do not need any emergency planning, rather than that we need not consider earthquakes in emergency planning.” 20 N.R.C. at 262 (Commissioner Asselstine, dissenting).
By endorsing the NRC’s consideration of design adequacy and radiological accident probabilities in emergency planning, the majority invites the continuing erosion of the emergency planning standards. Little is left for the NRC to plan for once it eliminates from consideration not only initiating events the plant has been designed to withstand but any complicating event with a low probability of occurrence after it has been multiplied by a factor of 0.00001 to reflect the 1 in 100,000 chance of a radiological accident occurring in the first place.5 These exceptions will easily swallow the whole of the plan. The NRC is, of course, free to change its regulations to make “emergency planning ... a backstop rather than a front-line defense.” NRC Br. at 26. As of now, however, emergency planning is supposed to be “equivalent to, rather than ... secondary to, siting and design in public protection.” 44 Fed.Reg. at 75,169. The regulations must be interpreted accordingly.6
III. The “Arbitrary and Capricious” Nature of the Commission’s Rationales
The Diablo Canyon decision is arbitrary and capricious solely because of its fundamental inconsistency with the putative purposes of the emergency planning regulations. In addition, the Commission’s three stated reasons for excluding earthquakes from emergency planning at Diablo Canyon, and the majority’s acquiescence in them, provide further evidence of the decision’s arbitrariness.
A. Earthquake-Initiated, Radiological Emergency
The Commission’s first rationale for interpreting the emergency planning regulations to exclude earthquake complications is that the probability of an earthquake causing a radiological release is too small to be of concern. This conclusion is dependent on two separate findings about different-sized earthquakes:
For earthquakes up to and including the Safe Shutdown Earthquake (SSE), the seismic design of the plant was reviewed to render extremely small the probability that such an earthquake would result in *219a radiologic release. While a radiologic release might result from an earthquake greater than the SSE, the probability of occurrence of such an earthquake is extremely low.
20 N.R.C. at 251 (footnotes omitted). I agree that by definition earthquakes greater than the SSE occur too infrequently to warrant consideration, since the SSE is the strongest earthquake that could ever be expected to hit the Diablo Canyon site.7 On the other hand, I believe that the Commission erred by excluding smaller earthquakes from consideration as accident initiators.
One of the two arguments supporting the Commission’s decision to ignore earthquakes smaller than the SSE has already been addressed. The Commission’s conclusion that such earthquakes will not initiate a release given the plant’s design is inconsistent with “[t]he underlying assumption of the NRC’s emergency planning regulations ... that, despite application of stringent safety measures, a serious nuclear accident may occur.” Southern California Edison Co. (San Onofre), 17 N.R.C. 528, 533 (1983); see supra at 48-49. In the emergency planning context, design alone cannot justify barring consideration of a natural hazard which may initiate a radiological accident.8
The majority also argues that no natural hazard need be considered as an accident initiator because the Commission’s guidance document on emergency planning specifies that “[n]o single specific accident sequence should be isolated as the one for which to plan because each accident could have different consequences, both in nature and degree.” NRC & FEMA, Criteria for Preparation and Evaluation of Radiological Emergency Response Plans and Preparedness in Support of Nuclear Power Plants 6 (Revision 1 1980) [hereinafter referred to as NUREG-0654]; see maj. op. at 32. This argument was not, however, relied upon by the NRC in the Diablo Canyon decision, 20 N.R.C. at 251-52, and was only alluded to in the San Onofre decision, 14 N.R.C. at 1092. In any event, the argument is based on a misreading of the guidance document, NUREG0654. The admonition not to plan based solely on a single accident sequence does not preclude planning based on an assessment of the differing consequences of a range of accidents triggered by differing initiating events. The previous sentence in NUREG-0654 states that the objective of emergency response plans is to provide protection against a spectrum of accidents, NUREG-0654 at 6, and different types of accidents obviously require different types of emergency planning.9
*220Indeed, the Commission itself recently acknowledged that “the capability of the surrounding population to respond to an accident initiated by a severe external event, such as an earthquake or hurricane, would differ significantly from the capability to respond to other accidents.” Consolidated Edison Co. (Indian Point), 21 N.R.C. 1043, 1058 (1985). This refreshing dose of common sense from the Commission suggests that the Diablo Canyon stand on earthquakes may indeed be an aberration dictated by frustration at the ten year delay in bringing the plant on-line. Be that as it may, the NRC’s reasons for ignoring any consideration of smaller-than-SSE earthquakes as initiators of radiological accidents makes no sense given the purposes of emergency planning and the requirements of the regulations as explained in NUREG-0654.
B. Simultaneous Occurrence of an Earthquake and an Independently Caused Radiological Emergency
The Commission and the majority have also dismissed the need to consider the simultaneous occurrence of a radiologic release and an unrelated earthquake. The NRC decision distinguished such an occurrence from other off-site complications which are routinely considered in emergency response planning as follows:
NUREG-0654 does call for some consideration of site-specific adverse or emergency conditions on emergency response. In prior cases, such frequently occurring natural phenomena as snow, heavy rain, and fog have been considered. With one exception, the focus has always been on frequently occurring natural phenomena. The Commission believes, based on the information provided by the parties, that earthquakes of sufficient size to disrupt emergency response at Diablo Canyon would be so infrequent that their specific consideration is not warranted.
20 N.R.C. at 252 (footnote omitted) (emphasis added). The majority has expanded upon this rationale, quantifying the Commission’s qualitative conclusion that earthquakes of a size likely to affect emergency response occur too infrequently to warrant consideration and, in the process, factoring in the probability of a radiological accident occurring in the first place.10
None of the reasons given by the Commission — or the majority — for excluding consideration of earthquakes which coincide with a radiological accident can stand up even under the deferential scrutiny of “arbitrary and capricious” review. I have already explained why the majority’s focus on the probability of a nuclear accident is irrelevant in deciding how to respond to such an accident. See supra at 8-9. The major failing of this portion of the NRC’s decision, however, is that the Commission has totally failed to present any coherent standard for determining which natural phenomena meet its “frequently occurring” standard or to substantiate the application of any such standard to Diablo Canyon. The Commission’s “frequently occurring” standard for deciding which natural phenomena merit attention *221as complicating factors in emergency planning is inadequately explained and justified, inconsistent with the way the Commission has applied emergency planning regulations to other offsite natural phenomena in the Diablo Canyon proceeding and in other licensing proceedings, and impossible to apply to Diablo Canyon based on the record in this proceeding.
1. The Inadequate Definition of “Frequently Occurring”
The Commission says that earthquakes are not the type of “frequently occurring” natural hazards to which the emergency planning regulations are addressed. The majority goes even further and purports to have difficulty in finding any references to natural hazards at all in the planning regulations or guidance document. Maj. op. at 11-15. It is noteworthy that neither the Commission’s final order in this case nor the relevant staff memoranda11 ever questioned the proposition that the emergency planning regulations, as explained in NU-REG-0654, require consideration of some offsite natural hazards which may initiate or complicate emergency planning. Diablo Canyon, 20 N.R.C. at 252; Pacific Gas & Electric Co. (Diablo Canyon), 19 N.R.C. 937, 941-44 (1984). The critical question is which such phenomena require consideration. To illuminate that question, I will briefly trace the regulations’ requirement that some natural hazards be considered in emergency planning.
The overall emergency planning regulation is indeed broadly worded, providing only that “no operating license for a nuclear power reactor will be issued unless a finding is made by NRC that there is reasonable assurance that adequate protective measures can and will be taken in the event of a radiological emergency.” 10 C.F.R. § 50.47(a)(1)(1985). Its scope is, however, refined by the further requirement that response plans comply with sixteen listed standards. 10 C.F.R. § 50.47(b). The relevant standard here requires the utility to develop, “consistent with Federal guidance,” a range of protective actions “appropriate to the locale” to be taken in the event of an emergency and guidelines for choosing among these alternative actions. 10 C.F.R. § 50.47(b)(10). The federal guidance referred to is NUREG-0654, which provides in several places for consideration of natural hazards and other offsite phenomena.12 The Commission and staff have themselves focussed on natural hazards by requiring that evacuation time estimates consider site-specific adverse weather characteristics, 20 N.R.C. at 252; 19 N.R.C. at 943-44; NUREG-0654, app. 4 at 4-6, which the staff defines as those “which might reasonably be expected to occur during the plant lifetime at a particular site and be severe enough to affect the time estimates for a particular event,” 19 N.R.C. at 944 (emphasis added).
The demands of NUREG-0654, as acknowledged by the Commission, are that site-specific natural hazards be considered in emergency planning. The majority must therefore rely on its more extreme argument that even if NUREG-0654 requires consideration of natural hazards, that document is only guidance and is not binding on the agency. Maj. op. at 33. The standard at issue here, however, specifically requires plans to comply with NUREG0654’s “guidance.” 10 C.F.R. § 50.-47(b)(10). And in its recently proposed rule on earthquakes and emergency planning, *222the Commission described NUREG-0654 as a “document, developed jointly by the NRC & [the Federal Emergency Management Agency (“FEMA”), which] forms the basis for both NRC and FEMA regulations on emergency planning.” 49 Fed.Reg. 49,640, 49,640 (1984). Further, NUREG-0654 itself notes that “FEMA and NRC regard all of the planning standards identified and contained herein as essential for an adequate radiological emergency plan.” NU-REG-0654 at 5. Finally, NRC licensing boards accord NUREG-0654 “considerable weight” in evaluating emergency plans because it was written by a joint FEMA/NRC committee, considered during the rulemaking process, and specifically referenced in the emergency planning rules. Long Island Lighting Co. (Shoreham), 21 N.R.C. 644, 652-53 (A.S.L.B.1985); Public Service Co. (Seabrook), 17 N.R.C. 1170, 1177 n. 5 (A.S.L.B.1983). In view of NUREG-0654’s history and the stated reliance on this document by the NRC and its licensing boards, this court should accord NUREG-0654 “considerable weight” in interpreting the emergency planning regulations, rather than altogether denying its pertinence. Cf. Community for Creative Non-Violence v. Watt, 670 F.2d 1213, 1216 (D.C.Cir.1982) (court may rely on agency’s contemporaneously issued policy statement in interpreting agency regulations).
The majority, finally, claims that reading the regulations to require consideration of any natural hazards requires reading them to require consideration of all such hazards. Maj. op. at 31 n. 2, 32. Yet the Commission has established that the regulations require consideration of some but not all contingencies caused by offsite natural phenomena. NUREG-0654 and the Commission’s order in the Diablo Canyon proceeding both acknowledge that some such hazards must be assessed. 20 N.R.C. at 252. Allowing the Commission to draw some line between the hazards it will and will not consider is not necessarily inconsistent with the purpose of the emergency planning regulations, which is to identify “prudent risk reduction measures.” Southern California Edison Co. (San Onofre), 17 N.R.C. 528, 533 (1983) (emphasis in original). Thus, the Commission may consider the probability of a natural hazard occurring and exclude from consideration those which occur so infrequently as to not countenance prudent risk reduction measures. I agree with the majority, maj. op. at 12 n. 2, that the only legitimate issue in this case is whether the Commission’s line-drawing was rational; I conclude, however, that petitioners have demonstrated that the Commission’s choice was not adequately supported.
The Commission has not adequately explained its decision to limit consideration of natural hazards to “frequently occurring” ones. The only explanation of the “frequently occurring” standard given by the Commission was to list several examples which it said had been considered in “prior cases.” By failing to define what constitutes a frequent rate of occurrence — or at least to describe where the cut-off point between frequently and infrequently occurring phenomena might lie — the Commission has made it impossible to apply the standard to hazards other than those specifically listed. And the reference to “prior cases,” as I demonstrate in the next section, adds nothing to the bare “frequently occurring” standard.
The majority is ultimately driven to uphold the Commission’s “frequently occurring” rationale by supplementing it with arithmetical calculations designed to illustrate the absurdly small probabilities of the simultaneous occurrence of a nuclear plant accident and an earthquake of any size. There are two general problems with the majority’s approach. First, as I have stressed repeatedly, factoring in the probability that an accident will occur conflicts with the fundamental principles of emergency planning, which must proceed on the assumption that a radiological accident has in fact already occurred. See supra at 5-9. Second, since, under the majority’s analysis, the probability of any natural hazard under consideration must always be multiplied by the probability of a radiological accident, this exercise is of no use whatso*223ever in drawing the line between frequently and infrequently occurring phenomena.
By automatically multiplying the 1 in 100,000 chance of a nuclear accident by the likelihood of any natural hazard occurring, maj. op. at 39-40, the majority reduces all simultaneous occurrences to a “never-never land” beyond rational planning. For example, there is only a one in a million chance that a nuclear accident would coincide with a severe blizzard if such a storm independently has a 10% chance of occurring in any given year, and a one in ten million chance an accident would coincide with a 100 year flood. Yet the Commission has considered both severe blizzards and 100 year floods in emergency planning. See infra at 56-57. How then can the Commission, under the majority’s approach, rationalize these as “frequently occurring” phenomena? It is not surprising that the majority quickly switches its approach when it discusses the probabilities of concededly “frequently occurring” natural ■ hazards such as fog and heavy rain, and analyzes their occurrence in absolute terms rather than in terms of the probabilities of their coinciding with a nuclear accident. Maj. op. at 42. ' This differential treatment does not, however, make for a convincing rationale as to what natural hazards the Commission should or should not consider.
Thus the majority’s quantitative focus cannot disguise the plain truth that the Commission has totally failed to provide any sensible working definition of its “frequently occurring” standard. While the Commission says that the emergency planning regulations only contemplate consideration of “such frequently occurring natural phenomena as snow, heavy rain, and fog,” 20 N.R.C. at 252, it has failed to explain when conditions other than those specifically listed rise to the level of “frequently occurring natural phenomena.” And, as I discuss next, this failure is not remedied by looking to those natural hazards considered in prior cases or in the Diablo Canyon proceeding. The Commission’s failure to define and apply the “frequently occurring” standard in any comprehensible way renders its interpretation of the emergency planning regulations as excluding consideration of earthquakes in this case an idiosyncratic one, without roots in any rational criteria. Cf. Railway Labor Executives’ Association v. United States Railroad Retirement Board, 749 F.2d 856, 862 (D.C. Cir.1984) (vacating agency’s statutory interpretation for failure to both articulate and apply a standard).
2. Inconsistency with Other Applications of the Emergency Planning Regulations
The Commission might have rescued its unfounded "frequently occurring” standard if, as the Diablo Canyon opinion claimed, “prior cases” had defined and applied such a standard. 20 N.R.C. at 252. Indeed, consistency of application in pri- or — and subsequent — cases is a highly important element in assessing whether a regulatory interpretation merits deference. See supra at 47-48. The Commission’s backhand reference to “prior cases,” however, does nothing to identify its criteria or a cutoff point for a “frequently occurring” natural hazard.
The majority accepts too quickly the Commission’s bald assertion that prior cases followed a “frequently occurring” standard; it adds that it can find no cases inconsistent with that standard. The majority’s lens, however, is a narrow one; it confines its inquiry to the only two Commission opinions which specifically addressed the issue of earthquakes and emergency planning. Of course, natural hazards other than earthquakes have been considered in many cases under the applicable regulation, 10 C.F.R. § 50.47(b)(10), and those cases shed considerable light on the prevailing interpretation of that regulation. The majority refuses to look at these cases because, it says, petitioners did not provide specific citations. Maj. op. at 40-41. At the same time, however, the majority unquestioningly accepts the Commission’s equally undocumented assertion that it has considered only “frequently occurring” natural phenomena in prior cases. Id. at 33. Fairness would seem to dictate that both *224parties cite cases to support their opposing claims before the court accepts either. Here, unfortunately, neither party did so. As a result, I examined all of the available emergency planning decisions to see whether the Commission has in fact consistently interpreted the regulations to require consideration of only “frequently occurring natural phenomena.” 20 N.R.C. at 252. My examination concludes it has not.
Several licensing board and Commission decisions have taken into consideration natural hazards which coiild complicate emergency planning, although these opinions acknowledged that such events occurred only infrequently. The NRC and its licensing boards regularly consider the complicating effects of very severe winter storms on evacuation. Consolidated Edison Co. (Indian Point), 21 N.R.C. 1043, 1059 (1985); Philadelphia Electric Co. (Limerick), 21 N.R.C. 1219, 1358-60 (A.S.L.B.1985); Long Island Lighting Co: (Shoreham), 21 N.R.C. 644, 815 (A.S.L.B.1985).13 Several licensing boards, including the one for Diablo Canyon, have considered the simultaneous occurrence of peak summer beach crowds and heavy rains “even though it is doubtful that a peak vacation period would coincide with heavy rains.” Rec. Vol. 102, App.Ex. 78 at 95 (evacuation time study); see Public Service Co. (Seabrook), 17 N.R.C. 1170, 1176-80 (A.S.L.B.1983) (allowing litigation as to evacuation times for busy summer weekend with adverse weather). Finally, as the Commission noted, the emergency plan for the Trojan plant considered the complicating effects of a volcanic eruption at Mt. St. Helens. 20 N.R.C. at 252 & n. 4.14
Neither does the Diablo Canyon proceeding itself provide much aid in discerning which phenomena occur frequently enough to warrant consideration. An evacuation time study admitted into the record evaluated the effects of such low probability events as the simultaneous occurrence of heavy rainstorms and peak summer crowds (an event far less likely to happen than heavy rains alone) and of flood levels projected to occur once every 100 years. Rec. Vol. 102, App.Ex. 78 at 67-69, 95. If a once in 100 years flood is frequent enough to warrant attention, why is not the once in 275 years recurrence of the Operating Basic Earthquake? Both probabilities are of the same order of magnitude.
These “prior cases” appear to follow the staffs position15 that adverse weather conditions should be considered as complicating factors in emergency response if they *225“might reasonably be expected to occur during the plant lifetime at a particular site and be severe enough to affect the time estimates for a particular event.” 19 N.R.C. at 944. They certainly indicate that the Commission’s rigid position in Diablo Canyon that earthquakes may not be considered in emergency planning differs from prior and subsequent interpretations of the same regulation with regard to other infrequently occurring phenomena. Without any explanation of these differences, the NRC’s interpretation of the emergency planning regulations as encompassing only “frequently occurring” natural hazards does not merit deference.
3. Misapplication of the Standard
Even if the Commission had defined and previously applied a “frequently occurring” standard for determining which off-site phenomena merit attention in emergency planning, the Commission lacked substantial evidence in this record to support application of any such standard. In order to apply any version of this standard, the NRC would have had to compare the probability of occurrence of “frequently occurring natural hazards” with that of an earthquake “of sufficient size to disrupt emergency response at Diablo Canyon.” 20 N.R.C. at 252. This calculation was not and could not have been performed by the NRC based on the existing record. The NRC never discussed — and indeed the record contains no evidence on — what size earthquake would disrupt emergency response and how frequently such earthquakes occur in the Diablo Canyon area.
Because the severity and frequency of earthquakes are related, the Commission cannot say that earthquakes are or are not frequently occurring natural phenomena. Rather it must conclude that, in the Diablo Canyon area, earthquakes of a given size occur only infrequently. A relevant discussion would have to include information as to what size earthquake would disrupt off-site response and how frequently such earthquakes occur near Diablo Canyon.
Here the relevant size earthquake is that “sufficient ... to disrupt emergency response at Diablo Canyon.” 20 N.R.C. at 252. But the NRC did not and apparently could not explain what size earthquake would have disruptive offsite effects. The record does not contain any information on the offsite consequences of different-sized earthquakes because all evidence on offsite effects of earthquakes of any size was specifically rejected as inadmissible. See infra at 60 n. 19. Indeed, the only record evidence on offsite consequences of earthquakes was contained in two staff memoranda, and that evidence indicated that earthquakes smaller than the SSE could disrupt offsite emergency response at Diablo Canyon.16 The majority’s citation to a *226public document, not in the record, on the effects of several earthquakes that have occurred in the Diablo Canyon area, maj. op. at 39-40, cannot override the NRC’s conscious decision to exclude all evidence on what' size earthquakes produce what types of offsite effects.
The NRC staff, PG & E, and the Commission (in the “special circumstances” 17 portion of its opinion) implicitly assume that the appropriate size earthquake to consider is the Operating Basis Earthquake (“OBE”). They argue that such an earthquake has a low occurrence rate in the Diablo Canyon area, about 1 in 275 years. Pacific Gas & Electric Co. (Diablo Canyon), 13 N.R.C. 903, 992 (A.S.L.A.B.1981). The NRC has not, however, provided any reason to believe that the OBE is the smallest earthquake that could disrupt offsite emergency response. The OBE is simply “the strongest seismic event considered likely to occur during the operating lifetime of a nuclear power plant.” Id. at 989; see 10 C.F.R. pt. 100, app. A, § III(d) (1985); cf. maj. op. at 38-39 (less severe earthquakes won’t affect the plant site). The Commission and the majority seem to have focused on this size earthquake not because of its capacity to cause offsite disruptions, but because there is some record evidence on its frequency. The record is absolutely bare, however, as to whether earthquakes smaller than the OBE could cause significant offsite disruptions.
There is, on the other hand, some evidence in the record that smaller earthquakes occur much more frequently than larger ones in the Diablo Canyon area. Rec. Vol. 47, Board Ex. 2J at Tables I & II & Fig. 2 (earthquakes of magnitude 5.0 are expected to occur 45 times in the next fifty years, those of magnitude 5.5, 16 times, and those of magnitude 6.0, 6 times); see Rec. Vol. 47, Board Ex. 2F at Table II. This record evidence flatly contradicts the majority’s conclusion, again based on extra-record evidence, that the probability of any size earthquake occurring in the Diablo Canyon area in any given year is about one in fifty. Maj. op. at 39-40.18
*227In sum, the Commission has failed to either define or apply the “frequently occurring” standard in any rational way as it affected earthquakes around Diablo Canyon. As a rationale for excluding the complicating effects of all earthquakes from emergency planning, the standard flops badly.
C. Flexibility of the Emergency Plan
The Commission’s last attempt at rationalizing the exclusion of earthquake planning from the Diablo Canyon licensing proceedings was billed as not an independent ground for its decision but rather a “consideration” which “bolstered” its conclusion. The Commission explained that
[s]pecific consideration has been given in this case to the effects of other relatively frequent natural phenomena____ In the extreme, these phenomena are capable of resulting in area-wide disruptions similar to some of the disruptions which may result from an earthquake____ Thus, while no explicit consideration has been given to disruptions caused by earthquakes, the emergency plans do have considerable flexibility to handle the disruptions caused by various natural phenomena which occur with far greater frequency than do damaging earthquakes, and this implicitly includes some flexibility to handle disruptions by earthquakes as well.
20 N.R.C. at 252-53.
The NRC is correct in saying that it may — perhaps must — assess the flexibility of an emergency response plan to meet different kinds of exigencies. The Commission has previously explained that “there should be core planning with sufficient planning flexibility to develop a reasonable ad hoc response to those very serious low probability accidents which could affect the general public.” Southern California Edison Co. (San Onofre), 17 N.R.C. 528, 533 (1983). Although the sufficient flexibility rationale is an acceptable one in general, however, it fails to save the Commission’s decision in this case because its application lacks substantia] evidence in the record.
The Commission’s sanguinity about the Diablo Canyon emergency plan’s flexibility is grounded in its assumption that the disruption which would be caused by an earthquake is comparable to the disruption which would be caused by other natural phenomena — such as fog, severe storms, and heavy rain — which were considered in developing the emergency plan. 20 N.R.C. at 252. To make such a finding, however, the Commission needed to compare the effects of earthquakes and the effects of the other natural phenomena. The majority is wrong in asserting that “common sense” alone demonstrates the similarity of the effects of earthquakes and these other natural phenomena on emergency responses to a nuclear accident. Maj. op. at 38. Common sense rather tells us that a factual record is needed to draw such a conclusion. For example, the Commission cites only one piece of evidence on the effects of heavy fog — that it increases evacuation time to ten hours. 20 N.R.C. at 252. Even if that isolated datum constituted sufficient evidence on the effects of natural phenomena other than earthquakes, it is meaningless by itself because the Commission has no record evidence about the effects of an earthquake on evacuation time to compare it with. The problem, in a nutshell, is that the record lacks any evidence on the offsite consequences of an earthquake because the licensing board concluded that all such evidence was inadmissible.
Ironically, the Commission’s sufficient flexibility rationale assumes what the Commission goes to great pains to deny — that earthquake effects should be considered in emergency planning. All that petitioners seek is the opportunity to litigate the issue of whether the Diablo Canyon plan is flexible enough to accommodate complications caused by earthquakes. The Commission cannot assume that flexibility without any record evidence and parade it as an excuse for not allowing relevant evidence about the disruptive effects of earthquakes into the record. Cf. GUARD v. NRC, 753 F.2d 1144, 1149 (D.C.Cir.1985) (a court will not *228consider even record evidence when the NRC’s interpretation of an emergency planning regulation had excluded consideration of that evidence).
III. Conclusion
The NRC’s absolute refusal to consider any evidence of complications caused by earthquakes which might cause or occur simultaneously with a radiologic release at Diablo Canyon is inexplicable in legal, logical, or common sense terms. The Commission’s decision is inconsistent with the terms of its own regulation and guidance document and with other interpretations of the same regulations. Parts of the decision contradict the purposes of emergency planning.
I am wholly at a loss to understand why the Commission has worked so strenuously to exclude all consideration of earthquakes from these licensing proceedings, when earthquake complications could easily have been explored on the basis of previously prepared exhibits and cross-examination of witnesses already testifying in the proceedings.19 I can only surmise that the Commission’s members painted themselves into a corner from which they refused to retreat. It defies common sense to exclude evidence about the complicating effects of earthquakes from a proceeding dealing with how to respond to a nuclear accident at a plant located three miles from an active fault, a plant in which seismic concerns dominated the design and construction proceedings for well over a decade. The majority’s preoccupation with probability calculations simply does not justify the Commission’s stubborn refusal to do the obvious. The majority has allowed the Commission to interpret its regulations in a manner which undermines the basic purpose of emergency planning and singles out earthquakes for different treatment from other offsite natural phenomena, without giving any good reason for its neglect.
The Emperor has no clothes — earthquakes should have been considered in the emergency planning for a radiological accident at Diablo Canyon. The county government knows this and has factored them into its emergency plans; PG & E commissioned a study on earthquakes at the NRC staff’s request and then was told there was no need to litigate or implement *229it. After more than ten years of public alarm, only a divided Commission and this divided court persist in pretending that earthquakes are not material to emergency planning for a nuclear plant located only three miles from an active geological fault. If that judgment is at fault, history will allow no rehearing.
I respectfully dissent.
. The original San Onofre decision on earthquakes and emergency planning held that the regulations do not require consideration of earthquakes and indicated further that such consideration would not be permitted barring amendment of the regulations. Southern California Edison Co. (San Onofre), 14 N.R.C. 1091, 1091-92 (1981) [hereinafter cited as San Onofre]. Similarly, in the Diablo Canyon order the Commission held both that the regulations did not require consideration of earthquakes and that it would not permit consideration in this case. Pacific Gas & Elec. Co. (Diablo Canyon), 20 N.R.C. 249, 250, 253-54 (1984) [hereinafter cited as Diablo Canyon ]; see infra at 58 n. 17 & 60 n. 19.
. At the same time this court granted, in part, petitioners’ suggestion for rehearing en banc, we granted petitioners’ motion for leave to file supplementary exhibits consisting largely of transcripts of closed Commission meetings in which the Commissioners discussed the decision not to allow earthquakes to be considered in the emergency planning for Diablo Canyon. San Luis Obispo Mothers for Peace v. NRC, 760 F.2d 1320, 1321 (D.C.Cir.1985) (en banc). Because, however, I would find the Commission’s order arbitrary and capricious on its face, and in view of the lack of evidence supporting it, I do not reach the second question posed in our unpublished order dated August 12, 1985, on the degree of consideration the court may give to those transcripts.
. The majority mischaracterizes the holding of Union of Concerned Scientists when it suggests that only issues defined as material in the Commission’s regulations merit a hearing. Maj. op. at 29-30. The court in Union of Concerned Scientists clearly stated that a court may always review, under an arbitrary and capricious standard, an agency’s determination that an issue is not material to safety. 735 F.2d at 1448 n. 20. Such review would be appropriate when a party alleges that existing regulations erroneously fail to require consideration of a material.safety issue.
. AH citations to the record will hereinafter refer to the volume of the record ("Rec. Vol.”) and, when possible, to either the exhibit number ("Ex.”) or transcript page ("Tr.”).
. The majority, the NRC staff, and PG & E all rely on an estimated accident probability of 10-5 or 1 in 100,000. Maj. op. at 38; NRC Staffs Memorandum Regarding Consideration of Effects of Earthquakes on Emergency Planning (CLI-84-4) at 4 n. 3 (May 3, 1984), in Rec.Vol. 168; PG & E Br. at 13-14 & nn. 14-15.
. The majority errs in finding that it requires "a strange reading of the statute to say that it permits no emergency planning at all (the situation for over thirty years), but that, once an emergency planning regulation is promulgated, it must mandate consideration of earthquakes.” Maj. op. at 36. All petitioners are saying is that once the NRC interpreted the Atomic Energy Act to require emergency planning regulations in order to protect public health and safety, it must interpret those regulations consistently with that purpose. Any subsequent change in statutory interpretation by the agency would, of course, require a reasoned explanation.
.Pacific Gas & Electric Co. (Diablo Canyon), 10 N.R.C. 453, 490 (A.S.L.B.1979). The NRC adequately explained why the regulations do not require consideration of the complicating effects of earthquakes greater than the SSE. As I explain later, the Commission can consider the likelihood of an initiating or complicating event occurring and exclude from consideration those with such low probabilities that they would not warrant prudent risk reduction methods. See infra at 54. Here the Commission noted that an earthquake greater than the SSE was extremely unlikely to occur and would cause so much damage that emergency response would have only marginal benefits. 20 N.R.C. at 251-52. While drawing the line between probabilities will sometimes prove difficult, see infra at 53-57, the Commission clearly does not have to consider an event as unlikely as an earthquake greater than the 7.5 magnitude SSE for Diablo Canyon. 10 N.R.C. at 489 (7.5 magnitude earthquake will occur once every 100,000 years).
. The panel opinion’s unanimous finding "that the likelihood that an earthquake will trigger a nuclear accident at the facility is so small as to be rated zero," 751 F.2d at 1304, is thus inapposite here. That conclusion was based on an evaluation of design precautions. But emergency planning standards operate on different assumptions than design standards: in the emergency planning context, the NRC assumes that an accident can occur despite the plant’s engineering and siting safeguards. Thus, in interpreting the emergency planning regulations the Commission cannot assume that design safeguards will be totally effective and eliminate the possibility that an earthquake smaller than the SSE will cause a radiologic release.
. The Diablo Canyon emergency plan accordingly contains onsite procedures for plant operators to follow when any of several initiating events, including earthquakes, triggers one of four emergency action levels. Rec.Vol. 98, App.Ex. 73 at Table 4.1-1 (emergency action *220levels); Rec.Vol. 100, App. Ex. 75 at EP M-4 (procedures for earthquakes).
. The Commission never quantified the probabilities involved in the manner suggested by the majority, the NRC staff, and PG & E. Maj. op. at 37-40; NRC Staffs Memorandum Regarding Consideration of Effects of Earthquakes on Emergency Planning (CLI-84-4) at 4 n. 3 (May 3, 1984) in Rec.Vol. 168; PG & E Br. at 13-14 & nn. 14-15. At one point in the. opinion the majority assumes that the Commission adopted the whole of the staffs analysis, maj. op. at 36, but the staff analysis to which the Commission referred, 20 N.R.C. at 251 & n. 1, was a January, 1984 memorandum which contained no such quantitative analysis, see 19 N.R.C. at 946-52. Indeed, it seems odd to infer that the Commission performed such a quantitative, probabilistic analysis when its discussion of the occurrence rates of natural hazards was wholly qualitative and never considered the probability of an accident occurring at all. 20 N.R.C. at 251-52. To the extent the NRC did not in fact rely on this quantitative rationale, the majority advances without warrant a post-hoc rationalization for the agency’s decision. Motor Vehicle Mfrs. Ass’n v. State Farm Mutual Automobile Insurance Co., 463 U.S. 29, 43, 103 S.Ct. 2856, 2866, 77 L.Ed.2d 443 (1983).
. While the majority correctly points out that the staffs prior position on interpretation of the regulations normally cannot bind the Commission, the second memorandum — dated January 3, 1984 — seems to have been adopted by the Commission in Diablo Canyon. See 20 N.R.C. at 251 & n. 1 (explaining that the NRC "agrees with the staffs analysis in this case” as advanced in the 1984 memorandum).
. See NUREG-0654 at 42 (procedures that provide for emergency actions to be taken should "tak[ej into account local offsite conditions that exist at the time of the emergency"); id. at 63 (plan for implementing protective measures should identify and provide means to deal with "potential impediments (e.g., seasonal impassability of roads) to use of evacuation routes”); id app. 4 at 4-6 (evacuation time estimates must consider site-specific adverse weather conditions).
. Indeed, the Shoreham licensing board has apparently considered offsite phenomena with only a "remote” probability of occurrence on more than one occasion:
Once again we are called on to predictively resolve an issue generated by the postulated simultaneous occurrence of independent events: in this case, snowstorms of varying intensity occurring simultaneously with a serious radiological emergency at Shoreham. No law of nature prevents the occurrence; the record is silent on its probability (although we think it remote) ...
21 N.R.C. at 815. The board concluded that such "remote" situations must be considered but only require the formulation of general response plans. Id. As we have noted, the appropriate probability to assess is that of the offsite phenomenon, rather than the simultaneous occurrence of the phenomenon and a radiological accident. See supra at 49-50.
. The Commission claims that a volcanic eruption was considered only because of "the expectation that another explosion [was] imminent at Mt. St. Helens.” 20 N.R.C. at 542 n. 4. There is no indication of such an expectation in the Trojan opinion itself, however; the only discussion of probabilities notes that "if an accident occurred in combination with transportation difficulties due to severe volcanic ashfall, effective protecting measures can still be implemented, albeit with greater difficulty. The probability of these two events occurring simultaneously is, however, extremely low." Portland General Electric Co. (Trojan), 12 N.R.C. 241, 243 (Off. of Nuclear Reactor Reg. 1980).
.While the majority correctly points out that the staffs prior position on interpretation of the regulations cannot bind the Commission, the staffs working definition of 10 C.F.R. § 50.-47(b)(10) is relevant in determining what interpretation has been applied in other cases. NU-REG-0654 defines § 50.47(b)(10) to require evacuation plans to consider site-specific adverse weather conditions. NUREG-0654, app. 4 at 4-6. The definition quoted in the text from the 1982 staff memorandum indicates which conditions the staff believes must be considered to meet that requirement.
. While the staff memoranda cannot create a legal interpretation binding on the Commission, see maj. op. at 33, they do indicate the state of the factual record on the offsite consequences of earthquakes in California. The staff never veered from its position that the offsite consequences of earthquakes at Diablo Canyon warranted consideration. Earthquakes were a major issue throughout the Diablo Canyon licensing proceedings because the plant is located only three miles from the Hosgri Fault, a fact unknown to the utility when it selected the site. Pacific Gas & Electric Co. (Diablo Canyon), 16 N.R.C. 756, 760 (A.S.L.B.1982). Seismic issues pervaded the design portion of the licensing proceedings. See generally Diablo Canyon, 13 N.R.C. 903. Not surprisingly, before the San Onofre decision the NRC staff working on Diablo Canyon had requested applicant PG & E to evaluate the potential complicating effects of earthquakes on emergency planning, specifically asking about disruption of offsite communication networks and transportation routes because “[i]n California, such occurrences appear to be frequent enough to warrant consideration in your emergency plans." Rev.Vol. 69, J.I. Ex. 117.
In the meantime, the NRC staff was considering earthquake effects on emergency planning because the San Onofre decision had said that the NRC would "consider on a generic basis whether regulations should be changed to address the potential impacts of a severe earthquake on emergency planning.” 14 N.R.C. at 1092. The staff’s resulting memorandum to the Commission, dated June 22, 1982, concluded that ”[p]lanning for earthquakes which might have implications for response actions ... in areas where the seismic risk of earthquakes to offsite structures is relatively high may be appropriate (e.g., for California sites and other areas of relatively high seismic hazard in the *226Western U.S.)." 19 N.R.C. at 941 (emphasis added). A second memorandum again noted that ‘‘[ojffsite damage generated by earthquakes can significantly affect nuclear emergency response,” especially on the West Coast where ground motion levels capable of causing severe offsite damage may be lower than the plant’s Safe Shutdown Earthquake. Id. at 947.
. In an earlier order to the parties in this proceeding, the Commission asked whether the regulations required consideration of earthquakes and, if not, whether such consideration should be permitted for Diablo Canyon because of “special circumstances.” 19 N.R.C. at 938-39. The Commission was referring to its regulation providing for waivers of and exceptions to regulations when "special circumstances with respect to the subject matter of the particular proceeding are such that application of the rule or regulation (or provision thereof) would not serve the purposes for which the rule or regulation was adopted." 10 C.F.R. § 2.758(b) (1985). This waiver provision, however, seems to be totally inapposite to the Diablo Canyon proceeding. The "special circumstances” regulation is used to consider site-specific attributes which affect the application of generic regulations. The emergency planning regulation at issue here already takes site-specific factors into account. See supra at 33. Obviously if an earthquake is not a site-specific phenomenon worthy of consideration it is also not a special circumstance requiring waiver of the emergency planning regulations.
Neither does the Commission’s concession that special circumstances might sometimes permit consideration of earthquakes change the fact that the Commission’s decision interprets the emergency planning regulations to neither require nor permit consideration of earthquake complications. See infra at 46 n. 1. The "special circumstances" regulation only allows the NRC not to apply a regulation. Thus, the only way earthquakes can be considered under the Commission's interpretation of its emergency planning regulations is for the regulations not to be applied.
. Indeed, if we are to rely on extra-record evidence on the frequency of earthquakes I would note that the last month has been marked by heavy earthquake activity in California. Three earthquakes struck Northern California between March 29 and March 31, with the last measuring 5.3-5.6, the strongest earthquake to hit the area since a 6.2 earthquake in April, 1984. The March 31 earthquake was felt in San Luis Obispo. N.Y. Times, April 1, 1986, at A18, col. 6; Washington Post, April 1, 1986 at A4, col. 4; Time, April 14, 1986, at 33. In addition, three mild earthquakes occurred in Southern California in quick succession in early April. Washington Post, April 6, 1986, at A21, col. 2.
. The earthquake issue could easily have been resolved at an early stage in the proceedings. PG & E responded to the staff’s initial request for information on earthquakes, see supra at 25 n. 16, by hiring a consultant, the TERA Corporation, to prepare a report on the complicating effects of earthquakes on emergency planning. See Rec.Vols. 102-03, Applicant’s Exs. 79, 79(A) & 79(B) (for identification only) (TERA report). The San Onofre decision was handed down one week before the pre-hearing conference in the Diablo Canyon full power operating license proceeding. At that conference, the Atomic Safety and Licensing Board concluded that San Onofre barred any consideration of the complicating effects of earthquakes on emergency planning, Rec.Voi. 88, Tr. at 11,445-51, a holding swiftly incorporated into an unpublished order, Memorandum & Order, Docket Nos. 50-275 OL, 50-323 OL, slip op. at 2 (A.S.L.B. Dec. 23, 1981), in Rec.Voi. 88. Although at the pre-hearing conference the Board had noted that there was sufficient time to appeal before the hearings began if its ruling was in error, Rec.Vol. 88, Tr. at 11,450, the Board later denied a request to certify an appeal because a "decision in regular course by the Commission in response to an appeal from the Board’s final initial opinion" would suffice, Memorandum & Order, Docket Nos. 50-275 OL, 50-323 OL, slip op. at 2 (A.S. L.B. Jan. 11, 1982), in Rec.Vol. 89.
Accordingly, when emergency planning issues were discussed during the hearings, the licensing board excluded all evidence on earthquakes. The applicant, PG & E, unsuccessfully attempted to introduce its TERA Report into the record, arguing that “just because we aren’t to litigate the effects of earthquakes ... doesn’t mean that it’s not accepted into evidence if it’s part of somebody’s plan.” Rec.Vol. 90, Tr. at 11,759. Judge Wolf bluntly rejected that reasoning, stating that "[w]e will not permit any evidence regarding earthquakes in this hearing." Id. at 11,760. One portion of the TERA Report, on estimation of evacuation times, was later admitted as Applicant’s Exhibit 84, but only after all references to earthquakes were blacked out and only after Judge Wolf reaffirmed that he would permit no questioning on the earthquake-related portions of the report. Id. at 12,111-14, 12,186-90. PG & E was also barred from placing in evidence -Exhibit 80(A), a portion of the revisions to the County's emergency plan for Diablo Canyon which addressed earthquake complications. Id. at 11,766-68. Thus, the record could easily have contained all of the information necessary to litigate the earthquake issue.