In 1971, pursuant to the Clean Air Amendments of 1970, Pub.L.No. 91-604, 84 Stat. 1679, the Administrator of the Environmental Protection Agency (EPA) promulgated a series of national ambient air quality standards, known as the NAAQS, for sulfur oxides. The standards established maximum acceptable average concentrations for one-year periods, 24-hour periods, and 3-hour periods. Under the Act and the EPA regulations, if the outdoor air of a locality exceeds either the 24-hour or 3-hour standard more than once during a year, construction of new pollution sources would be restricted and existing sources could be restricted in the technology they may employ.

Prior to 1979, compliance with the 24-hour NAAQS for sulfur oxides, 40 C.F.R. § 50.4(b), was generally determined by analyzing daily average concentrations based on air samples collected during midnight-to-midnight periods. In 1979, however, pursuant to section 110 of the Clean Air Act, as amended, 42 U.S.C. § 7410(a)(2)(C) (Supp. III 1979), the Administrator promulgated a new rule which implemented a national air quality monitoring system.1 In promulgating the rule, EPA purported to require the new monitoring system to measure sulfur oxide concentrations continuously and to calculate hourly running averages for 24-hour periods, rather than only midnight-to-midnight averages. The use of running averages would reveal whether concentrations of sulfur oxides exceed the NAAQS level for any 24-hour period during the year, whereas midnight-to-midnight measurement would only reveal exceedances for 365 specific 24-hour periods. Depending on the times of the day at which sulfur oxide concentrations reach their high and low points, a locality with concentrations that exceed the level set by the standard for many 24-hour periods would be detected by the running average technique but would escape detection under the midnight-to-midnight sampling method.

These appeals challenge the 1979 rule that requires the running average technique for measuring sulfur oxide concentrations. Petitioners operate major stationary sources of sulfur dioxide. They would face significant economic burdens if the air monitoring statistics showed their localities to be in violation of the NAAQS for sulfur oxides. Petitioners charge that the challenged rule, in effect, revises the 24-hour NAAQS for sulfur oxides, and that it therefore should have been promulgated according to the exacting procedures set forth in section 307(d) of the Clean Air Act, 42 U.S.C. § 7607(d) (Supp. III 1979). Respondent argues that the rule does not constitute a revision of the NAAQS, but merely establishes closer and more uniform air quality monitoring, as mandated by section 319, 42 U.S.C. § 7619 (Supp. III 1979). Therefore, respondent contends, the agency properly proceeded by informal rulemaking, 5 U.S.C. § 553 (1976).

The scope of review, as well as the lawfulness of EPA procedures, depends upon the characterization of the agency's action. If the rule effects a revision of the NAAQS, then its adoption would require the procedures set out in § 7607(d), which are more demanding than those of the Administrative Procedure Act (APA). At the same time, however, § 7607(d) would preclude our invalidation of the rule on procedural grounds without a much stronger showing of prejudicial error than is required in cases under the APA.2 If, on the other hand, the rule does not effect a revision of the NAAQS, then it would have to be adopted only according to the APA procedures.

To decide whether the agency's action amounts to a revision of the 24-hour NAAQS for sulfur oxides, this court must resolve a fundamental ambiguity in the meaning of the NAAQS itself. The question, explained more fully below, is whether an exceedance of the standard's "(m)aximum 24-hour concentration" can be computed on the basis of all periods of twenty-four consecutive hours ("running averages"), or whether it must be computed on the basis of midnight-to-midnight periods only. Unfortunately, EPA has complied no administrative record directed toward that issue. Petitioners were apparently caught off-guard because of the obscurity of the notice given. They claim they were unaware that this issue was presented in the challenged rulemaking, and so offered no comments at all. We thus find ourselves in a "catch-22": the agency's procedures have produced no visible basis upon which to review its interpretation of the 24-hour NAAQS; yet, without reviewing that interpretation, we cannot determine the adequacy of its procedures. We, therefore, must examine the 24-hour sulfur oxide NAAQS directly, but only to the extent necessary to determine whether EPA's interpretation was not wholly unreasonable.

We conclude that the 24-hour NAAQS does not require a midnight-to-midnight sampling period, and that the new rule, therefore, does not revise that standard. Therefore, EPA's rulemaking proceeding under the APA, rather than the Clean Air Act, was not facially unlawful. We also find, however, that EPA failed to publish a critical provision of the new rule in the Federal Register and thereby violated the publication requirement of the APA. Belated publication cannot cure this omission. Because the proposed rule did not notify the public that the use of running averages was being proposed and because that proposal was an important aspect of the rule, EPA failed to comply with the notice requirement of the APA in its attempt to mandate the use of running averages. We therefore remand the record to the EPA to consider that part of the rule, after proper notice, by whatever procedures are adequate to develop a proper record for decision. Further, we admonish the parties that an important interpretation of an agency standard must, in the first instance and in clear fashion, be carried out by the agency itself, with review by the court following only upon an adequate record.

Congress has been concerned with air pollution for over a quarter of a century. See Act of July 14, 1955, Pub.L.No.84-159, 69 Stat. 322. In 1970, it enacted sweeping changes in the Clean Air Act.3 The amendments directed EPA's Administrator to promulgate NAAQS and provided for the adoption of state plans to achieve and maintain those standards. The "primary" NAAQS prescribe maximum acceptable concentrations of various pollutants in the outdoor air which, "allowing an adequate margin of safety, are requisite to protect the public health." 42 U.S.C. § 7409(b)(1). The "secondary" NAAQS prescribe levels "requisite to protect the public welfare from any known or anticipated adverse effects." Id. § 7409(b)(2). These standards function as benchmarks for the state implementation plans. The statute provides that both the primary and secondary NAAQS be based on "air quality criteria" for each pollutant, which

accurately reflect the latest scientific knowledge useful in indicating the kind and extent of all identifiable effects on public health or welfare which may be expected from the presence of such pollutant in the ambient air, in varying quantities.

Id. § 7408(a)(2).

The NAAQS at issue here concerns sulfur oxides, measured as sulfur dioxide. Sulfur dioxide contributes importantly to the scope and severity of the nation's air pollution problem. It is emitted in large quantities when industrial or public utility power plants burn high sulfur fossil fuels, especially coal.⁴ In the atmosphere, it is converted to sulfur trioxide, sulfuric acid, and salts of sulfuric acid. J.A. 459. Acute episodes of sulfur dioxide pollution, especially in connection with high levels of particulate matter, result in increased mortality and morbidity. In the world's most disastrous incident, about 4000 excess deaths were attributed to the smog that struck London in December 1952. Typically the elderly and persons with preexisting pulmonary and cardiac disease are the most susceptible. J.A. 566-67. Significant health effects are also produced by long-term exposure to sulfur dioxide pollution:

Acute respiratory infections in children, chronic respiratory diseases in adults, and decreased levels of ventilatory lung function in both children and adults have been found to be related to concentrations of SO2 and particulates ....

Rall, Review of the Health Effects of Sulfur Oxides, 8 Env'tal Health Perspectives 97, 99 (1974), quoted in Republic Steel Corp. v. Costle, 621 F.2d 797, 800 (6th Cir. 1980). Finally, atmospheres polluted with oxides of sulfur are believed to attack and damage a wide range of materials and property, and to cause a variety of harms to vegetation. J.A. 506, 518.

The Administrator of EPA promulgated primary and secondary sulfur oxides NAAQS on April 30, 1971, 36 Fed.Reg. 8186. The current version is set out in full in the margin.5 These standards were based on the Criteria issued in 1969, supra note 4. The primary NAAQS set maximum average concentrations for long-term (one year) and short-term (24-hour) periods; the secondary NAAQS sets a maximum 3-hour average concentration.6 The 24-hour and 3-hour concentrations are "not to be exceeded more than once per year." All concentrations are to be measured by "the reference method," described in Appendix A to 40 C.F.R. Part 50 (hereinafter Appendix A), or by an "equivalent method." A separate set of procedures is set out in 40 C.F.R. 53.30-53.32 to test a "candidate method" against the reference method in order to certify it as "equivalent."

The reference method is not automated: air is sampled through a "gas bubbler" for a given length of time, then removed and chemically analyzed. The average concentration of sulfur dioxide for that period is determined from the contents of the bubbler. Appendix A provides for several sampling periods in order to accommodate a variety of data-gathering tasks.7 Until about 1977, compliance with the 24-hour NAAQS was determined by analyzing daily average concentrations, based on 24-hour samples.8 The reference method specifies that 24-hour samples be collected "from midnight to midnight."9 Thus, from a given monitoring station, not more than 365 24-hour samples could be tested by the reference method each year. If the "second highest" concentration among these samples exceeded the prescribed level for the 24-hour NAAQS, a non-attainment designation would be made.

It is possible, however, for a locality to experience 24-hour concentrations of sulfur dioxide that are above the NAAQS concentration levels and yet go undetected by a monitoring station using the 24-hour sampling technique of the reference method. Not only does the use of 24-hour samples under the reference method allow sampling for only 365 out of the infinite number of twenty-four hour periods during the year, it requires that each sampling period begin and end at the same times every day. As a result, sulfur dioxide concentrations that exceed the NAAQS level during a period other than midnight to midnight noon to noon, for example would not be detected unless the average concentration also exceeds the standard during a contiguous midnight-to-midnight period. The graph below illustrates the importance of the sampling period in identifying periods of concentration above the standard:NOTE: OPINION CONTAINS TABLE OR OTHER DATA THAT IS NOT VIEWABLE

If the average concentration of sulfur dioxide in a locality follows the path charted either at discrete times during the year or regularly the 24-hour sampling technique of the reference method would not detect concentrations higher than the NAAQS levels. In this example, continuous monitoring and calculation of running averages, in contrast, would indicate an exceedance of that level for the 24-hour periods beginning close to noon and ending at the same time the next day.

Over the years, a number of newer methods have been certified as "equivalent," and have gradually replaced the reference method at many air-monitoring facilities. These newer methods permit instantaneous and continuous monitoring of sulfur dioxide concentrations. Running averages of concentrations can be calculated for continuous 24-hour periods. Using such methods, it becomes easier to detect exceedances of the NAAQS level for any period of 24 consecutive hours, whether it runs from midnight to midnight or not.10

In February, 1977, EPA prepared and distributed to the state monitoring offices a revised "Guidelines for the Interpretation of Air Quality Standards" (hereinafter Guidelines). J.A. 424. The Guidelines took advantage of the newer techniques to request data based on running averages. EPA recommended that, wherever continuous monitors were available, running 24-hour average concentrations should be calculated, with each hour beginning a new 24-hour period. Id., J.A. at 435. The Guidelines also provided a methodology for computing the statistics used in determining compliance with the 24-hour NAAQS, Id., J.A. at 437-38. Any 24-hour period with an average concentration greater than the standard would be deemed an exceedance. If a locality experienced two such periods that did not overlap in one year, the locality would be deemed out of compliance. Id. at 437-38. The Guidelines were not published in the Federal Register.

In August, 1977, Congress again amended the Clean Air Act.11 It added section 307(d), 42 U.S.C. § 7607(d), which made the promulgation or revision of any NAAQS subject to demanding procedures.12 It also added a new section 319, directing the EPA Administrator to "promulgate regulations establishing an air quality monitoring system throughout the United States which ... utilizes uniform air quality monitoring criteria and methodology." 42 U.S.C. § 7619. The section 319 regulations were not subject to the requirements of section 307(d), but were to follow "notice and opportunity for public hearing."

One year later, EPA proposed rules to implement section 319. 43 Fed.Reg. 34892 (1978). The notice expressly stated that the APA's informal rulemaking procedures would be followed. Id. at 34896. The proposed rules required each state to maintain a network of "state and local air monitoring stations" (SLAMS), some of which would be designated "national air monitoring stations" (NAMS). The NAMS were to be equipped with certified monitoring devices capable of continuous measurement. Id. at 34918. The proposed rules also contained about 30 pages of technical appendices designed to assure the quality and uniformity of data collected by these networks. One of these appendices Appendix F required each station in the SLAMS network to calculate and report annually, among other statistics, the "(h)ighest and second highest 24-hour averages" for sulfur dioxide, presumably to determine compliance with the 24-hour NAAQS. Id., at 34928. A footnote to that requirement (hereinafter Footnote 3) specified that the averages should be "(b)ased on nonoverlapping values computed according to procedures described in reference (1)." "Reference (1)" was identified at the end of the appendix as the Revised Guidelines document, discussed supra. Thus, the proposed rule attempted to adopt the use of running averages as a means of determining compliance with the 24-hour NAAQS for sulfur oxides. Previously, EPA had only recommended that method.

The proposed rules were adopted with little change in May, 1979. None of the comments indicated a contradiction between the proposed monitoring and reporting method and the 24-hour NAAQS. See 44 Fed.Reg. 27567 (1979). The only discussion of reporting methods came from state environmental agencies, which had earlier received the Guidelines. J.A. 668-690. We are informed by counsel that petitioners did not comment because they were unaware that the rule would affect the stringency with which the NAAQS was applied. Although running averages were apparently being reported by some monitoring stations at that time, petitioners did not learn of this until the county in which they do business was designated a nonattainment area. That occurred in October, 1978, at the very end of the comment period for the challenged rulemaking.13

Now petitioners ask this court to set the rule aside and to instruct EPA that "the midnight-to-midnight reference method remains in effect." Petitioners' Brief at 48. They claim, first, that the reference method of Appendix A is integral to the NAAQS for sulfur oxides, and can only be changed through the procedures set forth in 42 U.S.C. § 7607(d). They imply, further, that the reference method requires 24-hour samples to be used in evaluating the 24-hour NAAQS. Since these samples must be collected from midnight to midnight, petitioners believe that the NAAQS must be applied from midnight to midnight. Any method which departs from this midnight-to-midnight practice, therefore, is said to be inconsistent with the 24-hour NAAQS.14

Petitioners argue that a running average requirement creates a significantly more stringent 24-hour NAAQS. This was done, they assert, without substantive support and in violation of the procedures provided for such changes by the Clean Air Act, 42 U.S.C. § 7607(d)(1)(A). In addition, because EPA adopted a portion of the rule by means of an obscure reference, never incorporated into the Federal Register cited in a footnote to an appendix to a 46-page rule on air quality reporting petitioners argue that the agency violated the publication and notice requirements of the APA.

The rulemaking record presents this court with a dilemma. It discloses no hint that EPA considered petitioners' arguments, and thus provides no basis for resolving the issues they raise. If petitioners are correct that the NAAQS was revised sub silentio, then any inadequacy in the record might well have been due to a failure to follow Clean Air Act procedures. If, on the other hand, the 24-hour NAAQS does not incorporate the reference method's instruction on midnight-to-midnight sampling, then no revision has occurred, and the agency was correct to follow APA procedures. The court is being asked to choose an interpretation of the NAAQS before the agency has even spoken.

This court is charged with determining "the meaning or applicability of the terms of an agency action" pursuant to the APA, 5 U.S.C. § 706 (1976). But such determinations should be made on an adequate record, after the agency itself has addressed the issue. Whether the challenged rule involves a change in the NAAQS depends on factual and policy judgments which EPA is best equipped to make. Direct resolution of that issue by the court would bypass a carefully designed administrative process. If incorrect, the resolution could itself effect a change in the NAAQS. As we shall describe, an initial interpretation of the 24-hour NAAQS by the court is necessary in this case: the agency must be given some assurance that its reconsideration will proceed under the correct statute. But whatever procedures are chosen on remand, we trust that the record will enable the court to perform its traditional reviewing function. This court is prepared to reconsider its interpretation of the 24-hour sulfur oxide standard if and when EPA has first addressed that issue.

To perform our reviewing function, we must reach at least a preliminary interpretation of the 24-hour NAAQS. Whether any procedural errors were sufficiently prejudicial to warrant further notice and comment, for example, may depend on whether the APA or the more demanding test of the Clean Air Act, 42 U.S.C. § 7607(d), is applied.15 In any event, our opinion should inform EPA as to the procedures that are statutorily required on remand. On the basis of our own review of the record before us, we believe that the reference method of Appendix A does not confine the 24-hour sulfur dioxide NAAQS to midnight-to-midnight sampling periods. Accordingly, we believe EPA acted within its authority when it attempted to change the reporting requirements so as to mandate running averages through informal APA rulemaking. Our conclusion, however, is not meant to prevent the agency from reaching a different interpretation of the NAAQS on remand.

Ordinarily, we would accord the greatest deference to EPA's construction of its own standard, as well as to its choice of procedures. In this case, however, the agency's construction can only be inferred from its actions; we have no direct and contemporaneous statement of its views on the meaning of the NAAQS. We therefore look principally to the language and structure of the regulatory scheme, and to the policies that underlie it. Our interpretation is based on three grounds. First, the reference method itself does not clearly mandate midnight-to-midnight samples. Second, a change in method does not necessarily entail a change in the NAAQS. Third, running averages are consistent with the apparent purpose of the 24-hour NAAQS.

The reference method provides detailed instructions for measuring sulfur dioxide concentrations.15 Appendix A describes three separate sampling periods: 1/2 hour, 1 hour, and 24 hours. Petitioners argue that the 24-hour sample instructions are to be used in enforcing the 24-hour NAAQS, and therefore control its meaning. Appendix A's instruction to collect 24-hour samples from midnight to midnight is thus said to preclude the use of 24-hour running averages.

We do not agree. Nothing in Appendix A or anywhere else states that 24-hour samples must be used for 24-hour NAAQS. The other two sampling periods described in Appendix A 1/2 hour and 1 hour do not correspond to the averaging times of the other two NAAQS i. e., 3 hours and one year. There is thus no support for the notion that each NAAQS must use a corresponding sampling period.16 Indeed, the 24-hour NAAQS, by its terms, could be measured from any of the sampling periods specified in Appendix A, including the 30-minute and 1-hour periods. Although it would surely be inconvenient to collect and measure 24 1-hour samples manually, that practice would permit the computation of hourly running averages using the reference method.17 And as Appendix A expressly provides: "(o)ne can select different combinations of sampling rate and time to meet special needs." 40 C.F.R. Part 50, App. A (7.1).

The instruction to sample from midnight to midnight thus appears directed at assuring uniformity and reliability in determining and comparing sulfur dioxide concentrations. There is no indication that it was intended to restrict or clarify the meaning of the 24-hour NAAQS.

Petitioners argue that a change in Appendix A must follow the procedures specified for revising a NAAQS, 42 U.S.C. § 7607(d)(1)(A). Whereas Congress was explicit in the procedures required for revising a NAAQS, it provided much greater flexibility to the Administrator in the areas of measurement and evaluation. A nonattainment area, for example, is to be designated through monitored data, air quality modeling, or "other methods determined by the Administrator to be reliable." 42 U.S.C. § 7501(2). The technical details of measurement methodology have typically developed through less formal procedures. Even if Appendix A permitted only 24-hour samples, therefore, the agency might nevertheless institute continues monitoring and running averages without revising the 24-hour NAAQS.

In the case before us, "equivalent methods," see note 14 supra, have altered the practice under the reference method. Nothing in the equivalency regulations specifies a midnight-to-midnight sampling period. Petitioners argue that these methods should be limited to such a period, otherwise they cannot be equivalent. But that begs the question of how central the midnight-to-midnight instruction really is. It is equally plausible to argue that the instruction is not carried into the equivalency regulations because it is not essential that methods are equivalent if and when they show identical concentration levels for the same time periods. Monitoring stations today must compare concentrations over varying time periods; they evaluate air quality based on modeling techniques which are not time-specific.18 In these circumstances, a simpler notion of equivalency may better serve to keep pace with technological change and the needs of flexible data-gathering.

It is also possible to revise a reference method directly without revising a NAAQS. See 40 C.F.R. §§ 53.14, 53.16. Although respondent does not claim to have made such a revision here, we note that the Administrator has broad discretion to act, following procedures distinct from those of 42 U.S.C. § 7607(d).

Finally, the interpretive questions at issue in this case have historically been resolved in guideline documents, rather than in reporting rules or NAAQS proceedings. See J.A. 435, 945. The "fixed interval" approach, supported by petitioner, has been justified in terms of computational advantages. The running average approach has been justified as providing a more accurate indication of underlying concentrations, as well as more protection. The agency has never acted as though these decisions revised either the NAAQS or the reference method.

As important as measurement methods are, they occupy a somewhat peripheral place in the structure of these regulations. The focus is clearly on the NAAQS: simple, inherently controversial, and perhaps therefore procedurally difficult both to adopt and to change. The measurement methods, by contrast are complex, technical, flexible, and, in theory, noncontroversial. It is unlikely that Congress intended EPA to use the complex procedural machinery applicable to the NAAQS to make technical adjustments in the methods of measurement.19

Beyond the language and structure of these regulations, there is a policy reason for refusing to confine the 24-hour period to a midnight starting time. The existence of both a short- and a long-term primary standard strongly implies that the former is aimed at acute threats to health. If so, it should not matter whether high pollutant concentrations run from midnight to midnight or noon to noon: one episode presents as great a threat as the other. Petitioners' interpretation would permit many of these acute threats to go undetected, whereas the standard declares that one such period per year is enough.20

Taking more than 365 measurements per year, which continuous measurement and running averages allow, thus becomes a matter of stricter enforcement. It detects high pollution levels, concentrated in time, just as placing more monitors in an area would detect pockets of pollution concentrated in space. It is not unreasonable to assume that such concentrations are unhealthful whenever they occur. That they escaped detection ten years ago does not mean they should be permitted to continue today. Indeed, the provision for certifying equivalent methods may well have been intended to allow the employment of improved enforcement technology as it develops.

Petitioners claim that the air quality criteria studies, which underlie the standards, were based on midnight-to-midnight measurements. We note that the parties disagree on this question, and the answer is by no means clear. In any event, we decline to resolve it. It is of course possible that the health effects described in these early studies were caused by pollutant concentrations higher than those recorded, which went undetected because they occurred over parts of two midnight-to-midnight periods. If that is true, new studies using running averages might undermine some of the support for the existing 24-hour standard.21 It is not our role to decide or speculate about that issue. The interpretation of the standard, however, should remain unaffected by new studies. The Clean Air Act expressly provides procedures for challenging a standard on the basis of new information. See 42 U.S.C. § 7607(d)(7)(B). Challenging the standard through the back door of reinterpretation and bypassing the agency altogether is distinctly not favored.

We thus agree with respondent that the new rule did not revise the NAAQS for sulfur oxides, and that the conduct of the rulemaking and the scope of our review are governed by the APA, not the Clean Air Act. We proceed now to consider whether the APA was violated in any respect.

A central policy issue raised by the new rule is whether the 24-hour NAAQS for sulfur oxides should only apply to midnight-to-midnight periods. The most obvious reason for EPA's silence on that issue is that no apposite comments were offered at the rulemaking. While an absence of comments in no way implies inadequate notice, it does alert us to that possibility. Petitioners charge that essential provisions in both the proposed and final rules cannot be understood without the Guidelines document,22 which was neither published nor properly incorporated in the Federal Register. We agree.

In most respects, the agency's notice met the APA's notice requirement. It included "the terms ... of the proposed rule," 5 U.S.C. § 553(b)(3) (1976), and was fully adequate to apprise the public "of the 'subjects and issues involved' so that (interested parties) may offer informed criticism and comments." Ethyl Corp. v. EPA, 541 F.2d 1, 48 (D.C.Cir.), cert. denied, 426 U.S. 941, 96 S. Ct. 2662, 49 L. Ed. 2d 394 (1976). The notice, however, did not fully meet the APA's notice requirement. In one particular, the rule itself, without the Guidelines, is extremely unclear. By basing certain critical compliance data on "nonoverlapping values," Footnote 3 appears simply to demand that reported exceedances not overlap. 40 C.F.R. Part 58, App. F at 2.1.2 & fn.3. The provision does not in itself effect a change in the longstanding practice of accepting midnight-to-midnight measurements. Unless the Guidelines' definition of "nonoverlapping values" is read into the rule, in other words, we cannot say that Footnote 3 requires running averages to be reported. Nor can we find that the same language in the proposed rule notified the public that such a requirement was being considered.

Respondent argues that the text of the rule informed the public of the reporting procedures to be followed and that the Guidelines did not impose any additional requirements. Brief at 35. But the plain meaning of "nonoverlapping values" does not necessarily imply elimination of the midnight-to-midnight measurement period and adoption of the running average technique. Use of midnight-to-midnight averages would not conflict with the requirement that "nonoverlapping values" be used. If running averages are imposed by the rule, it can only be through Footnote 3's subsequent direction: "computed according to procedures described in reference (1)." That is, a special definition of "nonoverlapping values" or the method described in reference (1) would be necessary to impose the additional requirement of running averages. As we see it, therefore, EPA intended the Guidelines' procedure to be made mandatory by means of the rule.

If a required definition or procedure is part of a rule, it must be published or incorporated by reference in the Federal Register, 5 U.S.C. § 552(a)(1)(D) (1976). Respondent does not deny that the Guidelines were neither published nor incorporated, or claim that petitioners had actual and timely notice of that document. Under these circumstances, petitioners "may not in any manner be required to resort to, or be adversely affected by" such unpublished materials. 5 U.S.C. § 552(a)(1). In short, the Guidelines referred to in Footnote 3 are without legal effect, and the rule does not require running averages.

If the omission had been minor or merely technical, it might perhaps be cured by subsequent publication, without resorting to further agency procedures. See 1 C.F.R. § 51.1(e). In the circumstances of this case, however, we believe that the failure to publish goes to the adequacy of the agency's notice. By EPA's own account, midnight-to-midnight sampling can indicate sulfur dioxide concentrations 30 to 40 percent lower than those that would be measured using running averages. Guidelines, J.A. 435. This, even if it is due to the use of more precise measurement methods, is hardly a "logical outgrowth" of a rulemaking to improve the quality and uniformity of air monitoring data. See Sierra Club v. Costle, supra, 657 F.2d at 352-353; Weyerhaeuser Co. v. Costle, 590 F.2d 1011, 1030-31 (D.C.Cir.1978). The notice and comment process of that rulemaking was simply not directed at the effect the new measurement technique would have on the stringency of the standards as applied.

The importance of the issue was belied by the obscurity of the footnote intended to give notice, by the lack of meaningful comment, and by the complete absence of explanation in the agency's statement of basis and purpose. While belated publication or incorporation of the Guidelines might thus satisfy the duty to publish a rule, 5 U.S.C. § 552(a)(1) (1976), it will not here serve the more fundamental purposes of the APA's publication requirements: it will not provide petitioners the opportunity to comment on a significant part of the agency's decisionmaking process as required by section 553. See Weyerhaeuser v. Costle, supra, 590 F.2d at 1030-31; American Frozen Food Inst. v. Train, 539 F.2d 107, 135 (D.C.Cir.1976).

If on remand EPA wishes to require the reporting of running averages, it should conduct notice and comment proceedings to consider and explain its decision properly. We decline, however, to take the further step of declaring the method unlawful, as petitioners urge. To do so would change the meaning of the standard and circumvent the very notice and comment procedures that petitioners claim were absent here. Petitioners will have ample opportunity to challenge EPA's interpretation of the sulfur oxides standard when the rule is taken up before the agency. Alternatively, a petition may be filed directly with EPA to interpret or amend the standard, to withdraw the Guidelines, or to specify midnight-to-midnight reporting procedures. See 42 U.S.C. § 7607(d)(7) (B); 5 U.S.C. § 553(e) (1976). Either route would provide a reviewing court with a contemporaneous record of the agency's consideration of this issue, rather than with the "post hoc rationalizations of counsel." See Oljato Chapter of the Navajo Tribe et al. v. Train, 515 F.2d 654, 665-68 (D.C.Cir.1975).IV. CONCLUSION

For the reasons set forth, the record is remanded so that respondent may consider the issues raised by petitioners in a properly noticed rulemaking. The agency is free to conduct that rulemaking pursuant to the APA, and to continue using its Guidelines to encourage but not to require, the reporting of running averages. Certain materials lodged with this court belong, if anywhere, in the administrative record; petitioners' motion to supplement the record on this appeal is accordingly denied. As stated above, this court is prepared to reconsider its interpretation of the 24-hour sulfur oxides standard if and when EPA has first addressed that issue.

So ordered.