TABLE OF CONTENTS
I. INTRODUCTION ............................................. 312
A. The Challenged Standards .............................. 312
B. The Parties ........................................... 312
C. Background ............................................ 313
D. Procedural History .................................... 314
II. THE VARIABLE PERCENTAGE REDUCTION OPTION ................. 316
A. EPA's Authority Under Section 111 to Issue
a Variable Standard ................................... 318
1. The Statutory Language ............................. 318
2. The Legislative History ............................ 319
B. The Reasonableness of EPA's Decision to
Issue a Variable Standard ............................. 322
1. Technical Background ............................... 323
2. EPA's Explanation for the Variable
Standard ........................................... 325
(a) The Factors Considered by EPA .................. 325
(b) EPA's Regulatory Analysis ...................... 326
(c) EPA's Stated Rationale for the Variable
Standard ....................................... 327
3. An Examination of EPA's Rationale for
the Variable Standard .............................. 328
(a) The Legitimacy of EPA's Regulatory
Analysis ....................................... 329
(1) EPA's Authority to Analyze
Long Term National and Regional
Impacts .................................... 329
(2) The Reliability of EPA's Econometric
Computer Model ............................. 332
(b) The Reasonableness of EPA's Conclusions ........ 336
(1) The Reasonableness of EPA's
Conclusion that Variable Control
Reflects a Better Balance
of the Section 111 Factors Than
Uniform Control ............................ 336
(2) The Reasonableness of EPA's
Conclusion that Variable Control
Promotes the Policies of
the Act .................................... 338
C. The Dry Scrubbing Controversy ......................... 340
1. The Role of Dry Scrubbing Technology
in EPA's Rationale for the Variable
Standard ........................................... 340
2. The Legitimacy of Considering Emerging
Technology in Setting Section 111
Standards .......................................... 346
3. The Adequacy of the Record for Dry
Scrubbing's Role in EPA's Rationale ................ 347
D. The Adequacy of Notice and the Opportunity
to Comment on the Rationale for the Variable
Standard .............................................. 352
III. THE 90 PERCENT REMOVAL STANDARD .......................... 356
A. Notice As to the Basis of the 90 Percent
Standard .............................................. 356
1. The Basis of the Final Standard .................... 356
2. Notice that the Basis of the Standard
Had Changed Since Proposal ......................... 358
B. The Achievability of the 90 Percent Standard .......... 360
1. The Support For EPA's Conclusions
About FGD Performance .............................. 360
(a) The Achievability of 92 Percent
Long Term Removal Efficiency ................... 361
(b) The Reasonableness of EPA's Assumption
About FGD Variability .......................... 364
2. The Support for EPA's Conclusion that
the 90 Percent Standard Was Achievable
by the Use of Coal Washing in
Conjunction with Scrubbing ......................... 367
(a) Description of the Coal Washing
Process ........................................ 368
(b) The Percentage Reduction Achievable
by Washing High Sulfur Coal .................... 369
IV. THE STANDARD FOR EMISSION OF PARTICULATE
MATTER ................................................... 374
A. Technical Background .................................. 374
1. ESP Control Technology ............................. 374
2. Baghouse Control Technology ........................ 375
B. The Evolution of the Particulate Standard ............. 376
C. The Achievability of the Standard ..................... 377
1. EPA's ESP Data ..................................... 377
2. EPA's Baghouse Data ................................ 380
V. THE 1.2 LBS./MBTU EMISSION CEILING ....................... 384
A. EPA's Rationale for the Emission Ceiling .............. 384
B. EDF's Procedural Attack ............................... 386
1. Late Comments ...................................... 387
2. Meetings ........................................... 387
C. Standard for Judicial Review of EPA Procedures ........ 391
D. Statutory Provisions Concerning Procedure ............. 392
E. Validity of EPA's Procedures During the
the Post-Comment Period ............................... 396
1. Written Comments Submitted During
the Post-Comment Period ............................ 397
2. Meetings Held with Individuals Outside
EPA ................................................ 400
(a) Intra-Executive Branch Meetings ................ 404
(b) Meetings Involving Alleged Congressional
Pressure ....................................... 408
VI. CONCLUSION ............................................... 410
APPENDIX ...................................................... 411
Opinion for the Court filed by Circuit Judge WALD.
Circuit Judge ROBB concurs in the result.
WALD, Circuit Judge:
This case concerns the extent to which new coal-fired steam generators that produce electricity must control their emissions of sulfur dioxide and particulate matter into the air. In June of 1979 EPA revised the regulations called "new source performance standards" ("NSPS" or "standards") governing emission control by coal burning power plants. On this appeal we consider challenges to the revised NSPS brought by environmental groups which contend that the standards are too lax and by electric utilities which contend that the standards are too rigorous. Together these petitioners present an array of statutory, substantive, and procedural grounds for overturning the challenged standards. For the reasons stated below, we hold that EPA did not exceed its statutory authority under the Clean Air Act1 in promulgating the NSPS, and we decline to set aside the standards.
A. The Challenged Standards
The Clean Air Act provides for direct federal regulation of emissions from new stationary sources of air pollution by authorizing EPA to set performance standards for significant sources of air pollution which may be reasonably anticipated to endanger public health or welfare.2 In June 1979 EPA promulgated the NSPS involved in this case.3 The new standards increase pollution controls for new coal-fired electric power plants4 by tightening restrictions on emissions of sulfur dioxide and particulate matter.5 Sulfur dioxide emissions are limited to a maximum of 1.2 lbs./MBtu6 (or 520 ng/j)7 and a 90 percent reduction of potential uncontrolled sulfur dioxide emissions is required except when emissions to the atmosphere are less than 0.60 lbs./MBtu (or 260 ng/j). When sulfur dioxide emissions are less than 0.60 lbs./MBtu potential emissions must be reduced by no less than 70 percent. In addition, emissions of particulate matter are limited to 0.03 lbs./MBtu (or 13 ng/j).
B. The Parties
Petitioners in this case are Sierra Club and the State of California Air Resources Board ("CARB"), which oppose the variable 70 to 90 percent reduction requirement of the NSPS; Appalachian Power Co. ("APCO"), et al., a group comprised of APCO, the Edison Electric Institute, the National Rural Electric Cooperative Association, and 86 individual utilities ("Electric Utilities"), which challenge both the maximum 90 percent reduction requirement and the 0.03 lbs./MBtu limit on emissions of particulate matter; and, the Environmental Defense Fund ("EDF"), which challenges the 1.2 lbs./MBtu ceiling imposed by the NSPS.
Intervenor-respondents filing briefs in these consolidated actions are the Electric Utilities and the Missouri Association of Municipal Utilities ("MAMU"), aligned in favor of both the variable percentage reduction standard and the 1.2 lbs./MBtu emissions ceiling; and the National Coal Association ("NCA"), which opposes EDF's claim that the 1.2 lbs./MBtu ceiling is invalid due to procedural impropriety.
Respondents are the United States Environmental Protection Agency ("EPA") and its Administrator, Douglas M. Costle.
The importance of the challenged standards arises not only from the magnitude of the environmental and health interests involved, but also from the critical implications the new pollution controls have for the economy — at the local and national levels. Further heightening the significance of this controversy is the crucial role coal burning power plants are expected to play in our nation's effort to cope with the problems associated with energy scarcity.8
Coal is the dominant fuel used for generating electricity in the United States.9 When coal is burned, it releases sulfur dioxide and particulate matter into the atmosphere. At the very least these pollutants are known to cause or contribute to respiratory illnesses.10 In 1975 alone electric power plants emitted 18.6 million tons of sulfur dioxide. If the former NSPS had not been changed the total annual national sulfur dioxide emissions could have exceeded 23 million tons by 1995: a 27 percent increase.11 The increase in emissions which could be expected if the former standards continued in effect would be more dramatic on a regional basis. For example, utility sulfur dioxide emissions could be expected to increase 1300 percent by 1995 in the West South Central region of the country (Texas, Oklahoma, Arkansas, and Louisiana).12 In 1976 power plant emissions accounted for 64 percent of the total estimated sulfur dioxide emissions and 24 percent of the total estimated particulate matter emissions in the entire country.13
EPA's revised NSPS are designed to curtail these emissions. EPA predicts that the new standards would reduce national sulfur dioxide emissions from new plants by 50 percent and national particulate matter emissions by 70 percent by 1995.14 The cost of the new controls, however, is substantial. EPA estimates that utilities will have to spend tens of billions of dollars by 1995 on pollution control under the new NSPS.15 Consumers will ultimately bear these costs, both directly in the form of residential utility bills, and indirectly in the form of higher consumer prices due to increased energy costs.16 Coinciding with these trends the utility industry is expected to have continued and significant growth. Under the new NSPS EPA projects that overall utility capacity should increase by about 50 percent with approximately 300 new fossil-fuel fired power plants to begin operation within the next ten years.17 And approximately 350 new plants (capable of generating 250 Gigawatts ("GW")) are expected to be constructed by 1995.18 Present levels of national coal production and consumption will triple by 1995.19 With oil scarce, the future of nuclear and solar energy uncertain, and hydro limited, "the nation's rich and cheap coal reserves call for exploitation."20 Not surprisingly, coal burning power plants' already preeminent share of electric power produced in the United States will grow over the remainder of this century.21
While the volume and technical complexity of the material necessary for our review is daunting,22 we have endeavored to consider thoroughly the claims and myriad arguments proffered by the parties. We will discuss the basis of our decision on the principal challenges of the parties. We will not attempt, however, to discuss each and every point briefed, nor do we feel compelled to adhere religiously to the analytic framework devised by the parties.
D. Procedural History
In 1970 Congress for the first time authorized the federal government to set performance standards limiting emissions from newly built or modified sources of air pollution.23 These sources to be controlled were those that EPA determined emitted pollution contributing substantially to the endangerment of the public health or welfare.24 EPA decided that large coal-fired generators fell within that category.25 In December 1971 EPA issued a NSPS for these sources.26 That first NSPS applied to units capable of firing more than 250 MBtu per hour, and limited sulfur dioxide emissions to 1.2 lbs./MBtu and particulate matter emissions to 0.10 lbs./MBtu.27 Under this standard it was possible to satisfy the emission limitations simply by burning coal with a low sulfur content.28
In 1976 the Sierra Club and the Oljato and Red Mesa Chapters of the Navajo Tribe petitioned EPA to revise the NSPS so as to require a 90 percent reduction in sulfur dioxide emissions.29 The petition claimed that advances in technology since 1971 justified a revision of the standard. In response to the petition EPA began an investigation of whether the standard should be changed.30
While EPA's decision was pending the Clean Air Act Amendments of 1977 were signed into law. Section 111 of the amendments, discussed more fully below, required EPA to revise the standards of performance for electric power plants within one year after the August 1977 enactment date.31 When it appeared that EPA would not meet this deadline, the Sierra Club filed a complaint in the District Court for the District of Columbia. The court approved a stipulation requiring the proposed regulations to issue in September 1978, and promulgation of final regulations within six months after the proposal. Eventually, after further delay, the final NSPS were promulgated in June 1979.32
Several parties petitioned EPA for reconsideration of the revised NSPS. In February of 1980 EPA denied all the petitions for reconsideration.33
The present appeal followed. Petitions for review of the NSPS were filed in this court by the Electric Utilities (No. 79-1719), Sierra Club (No. 79-1565), EDF (No. 79-1874), and CARB (No. 79-1867). In addition, petitions to review EPA's denial of the requests for reconsideration of the final NSPS were filed by the Electric Utilities (No. 80-1187), Sierra Club (No. 80-1201), EDF (No. 80-1213), and CARB (No. 80-1338). All of these cases have been consolidated.
II. THE VARIABLE PERCENTAGE REDUCTION OPTION
We have already noted that the final NSPS adopted by EPA include an optional variable percentage reduction standard. Under this optional standard a utility plant can permissibly reduce its sulfur dioxide emissions by less than 90 percent of potential uncontrolled emissions if the amount of sulfur dioxide emitted following the use of pollution control technology is less than 0.60 lbs./MBtu.34 In no instance, however, can a plant reduce emissions by less than 70 percent of potential uncontrolled emissions.35 As a result of this option, the NSPS requirements for percentage reduction of sulfur dioxide removal vary on a sliding scale ranging from a minimum of 70 percent to a maximum of 90 percent.36 There is no dispute that the 70 percent floor in the standard necessarily means that, given the present state of pollution control technology, utilities will have to employ some form of flue gas desulfurization ("FGD" or "scrubbing") technology.37
Sierra Club contests EPA's authority under section 111 of the Act to vary from a uniform national percentage reduction standard ("uniform standard" or "full control")38 and the reasonableness of EPA's justification for doing so in light of the administrative record.39 Additionally, Sierra Club argues that the variable standard is fatally flawed and must be set aside regardless of supporting evidence on the record because the rulemaking was procedurally defective.40 The procedural objections to the variable control component of the NSPS stem from Sierra Club's assertion that EPA did not give adequate notice of the basis for the variable standard or provide sufficient opportunity for adversarial comment on the agency's purported justification for the rule. We turn to the question of EPA's statutory authority first.
A. EPA's Authority Under Section 111 to Issue A Variable Standard
Sierra Club's challenge to variable control raises the fundamental issue of whether EPA violated section 111 of the Clean Air Act by establishing a sliding scale for the reduction of sulfur dioxide emissions based on the sulfur content of coal burned in new utility plants. We find that section 111 of the Act authorizes such a variable standard.41
1. The Statutory Language
To evaluate the competing interpretations of section 111 we turn first to its text.42 Initially we find that the language of section 111 neither imposes a single, nationally uniform, percentage reduction standard nor prohibits EPA from varying the standard. Rather, section 111 merely requires inter alia:
[T]he achievement of a percentage reduction in the emissions from such category of sources [like new coal burning utility plants] from the emissions which would have resulted from the use of fuels which are not subject to treatment prior to combustion....43
The absence of any express mandate in this language to adhere to a single percentage reduction standard critically undercuts Sierra Club's arguments that EPA could not vary the standard below the level which is technologically feasible.
In fact, EPA is expressly authorized by section 111 to "distinguish among classes, types and sizes within categories of new sources for the purpose of establishing ... standards."44 Thus, the statute provides on its face that EPA does not have to set a uniform percentage reduction requirement for an entire category of emission sources. On the basis of this language alone, it would seem presumptively reasonable for EPA to set different percentage reduction standards for utility plants that burn coal of varying sulfur content.45 Certainly the text of the statute nowhere forbids a distinction based on sulfur content.
Other provisions of section 111 also belie the notion that EPA lacks discretion to vary the percentage reduction requirement according to the sulfur content of coal. For example, section 111(a) explicitly instructs EPA to balance multiple concerns when promulgating a NSPS:
[A] standard of performance shall reflect the degree of emission limitation and the percentage reduction achievable through application of the best technological system of continuous emission reduction which (taking into consideration the cost of achieving such emission reduction, any nonair quality health and environmental impact and energy requirements) the Administrator determines has been adequately demonstrated.46
(Emphasis supplied.) Having given EPA this mandate, Congress surely could not have meant to bind the agency to issuance of a uniform standard even though the agency's balancing of cost, energy, and nonair quality health and environmental factors indicated that the percentage standard should vary according to the sulfur content of coal.
Furthermore, reading section 111 to permit a variable standard based on the sulfur content of coal comports with common sense which suggests that the amount of sulfur in coal is the most relevant factor in designing standards to reduce emissions of sulfur in the gaseous wastes of coal combustion. Quite obviously, the "best technological system," considering cost, energy, and nonair health and environmental factors may well vary depending on the sulfur content of the coal that is burned.
2. The Legislative History
Sierra Club relies on portions of the legislative history of the 1977 Amendments to the Clean Air Act to demonstrate that, no matter how logical it may seem to permit the percentage reduction standard to vary according to the sulfur content of coal burned by utilities, Congress nevertheless meant to forbid such variable control levels. But the statements in the legislative history Sierra Club cites to us, even assuming that they are entitled to substantial weight in the face of a relatively clear text, do not adequately support Sierra Club's interpretation of the Act.
We note initially that a specific percentage reduction requirement was added to the text of section 111 only by the Conference Committee after the bills had passed both Houses of Congress.47 The Conference Committee Report explaining the requirement clearly contemplated the adoption of a variable standard:
[T]he Conferees agreed that the Administrator may, in his discretion, set a range of pollutant reduction that reflects varying fuel characteristics. Any departure from the uniform national percentage reduction requirement, however, must be accompanied by a finding that such a departure does not undermine the basic purposes of the House provision and the other provisions of the act, such as maximizing the use of locally available fuel.48
(Emphasis supplied.) Subsequently, during the House consideration of the Report the Conference Committee submitted a "Clarifying Statement" which stated that it was only fuel characteristics that could justify a departure from uniform control:
While the Conferees agreed that the Administrator may set the percentage reduction requirement as a percentage range, the Conferees expect the Administrator to be exceedingly cautious if he should elect to do so. Any such range of percent reduction would be allowed only to reflect varying fuel characteristics, and must be based on a carefully and completely documented finding by the Administrator that such departure from the strict requirement does not undermine the basic purposes of the House provision as expressed on pages 183 through 195 of the House Report number 95-294.49
(Emphasis supplied.) Similarly, Senate consideration of the Report included a Clarifying Statement by Senator Muskie that recognized a variable range of percentage reduction was permissible:
EPA's Administrator is given the flexibility to set a range of pollutant removal based on varying fuel characteristics if he finds that the NSPS objectives [of the Act] are not undermined.50
Even in the face of this legislative history, however, Sierra Club contends that "it is crystal clear that sulfur content of coal was not to be one of the `varying fuel characteristics' which would justify a departure from a uniform standard."51 We are frankly at a loss to understand this statement. Neither the remarks of Representative Rogers, the House manager,52 nor Senator Muskie, the Senate manager,53 when introducing the Conference Committee Reports to their respective Houses, supports Sierra Club's narrow interpretation of the phrase "fuel characteristic" to exclude sulfur content. Sierra Club relies on Senator Domenici's "additional statement" submitted after the close of the floor debate on the Senate version of the bill (S. 252) — before the bill went to Conference — which reads in part:
The House amendments to the Clean Air Act (H.R. 6161) contain a provision — section 111 — that effectively requires all new coal-fired powerplants to meet the same percentage reduction of pollution removal on new powerplants regardless of the sulfur content of the coal burned. Any doubts on this matter are dispelled by the explicit report language.54
(Emphasis supplied.) When read in context, however, it is clear that Senator Domenici was arguing that the Senate should not acquiesce in Conference to any demand for a uniform percentage reduction applied to all coals which he believed was implied in the House bill. Sierra Club's reliance on this statement is misplaced because the statement was made before the Conference Committee met at a time when neither the House nor the Senate bill contained any express provision for a percentage reduction. In fact, Senator Domenici's views opposing a uniform standard based on sulfur content can be viewed as a motivating factor in the Conference Committee's adoption of a more flexible standard in the final bill that allowed such a variation.55 Nothing else in the legislative history comes close to a directive that sulfur content is not a relevant fuel characteristic for setting a variable standard.56
Sierra Club also argues that even if section 111 permits the standard to vary depending on the sulfur content of coal, its language was designed to permit a nonuniform standard only in the limited circumstance where a "best technological system" could not achieve the national percentage on certain types of coal. That is, EPA could vary the standard only to reflect the different maximum feasible percentage reductions achievable for different sulfur content coals. Under this view EPA could not relax the standard when a higher percentage reduction is technologically feasible. Thus, in this case because it is not disputed that wet scrubbing could achieve 90 percent reduction on low sulfur coal, EPA has no authority to vary the standard below the 90 percent level.
We do not believe that this interpretation of section 111 is warranted by a fair reading of the Act or the underlying legislative history. The text gives EPA broad discretion to weigh different factors in setting the standard. The legislative history indicates that EPA should be "exceedingly cautious" in allowing the standard to vary, but nevertheless, recognizes that such a determination is within the range of EPA's discretion.57 The required finding that must underlie a variable standard is much broader than a mere determination that uniformity is not achievable. Rather, EPA has the discretion to vary the standard upon finding "that such a departure [from uniform control] does not undermine the basic purposes of the Act."58 Here EPA has made such a finding. While the reasonableness of this finding is challenged on this appeal and will be reviewed below,59 here it cannot be said that in making the determination EPA acted beyond its statutory authority.60
In addition to its arguments about the proper interpretation of section 111, Sierra Club maintains that variable control violates the total statutory scheme of the Clean Air Act because it is irreconcilable with other important features of the 1977 Amendments to the Act.61 In particular, Sierra Club contends that variable control is inherently inconsistent with the provisions in the Act designed to prevent the deterioration of air quality62 and visibility63 in designated areas, primarily in the Southwest. EPA responds first that, in fact, the variable standard is consistent with the other programs established by the Clean Air Act. This argument is grounded in the extensive regulatory analysis performed by EPA which showed that variable control served the interests of air quality and visibility as well as any uniform standard. We save for later our review of the legitimacy of that regulatory analysis and the reasonableness of EPA's conclusion that the variable standard actually promotes improved air quality and visibility.64 At this point, we note that if its analysis is reasonable, then EPA is right and the variable standard does not conflict with other goals. Second, EPA argues convincingly that the variable standard does not infringe the other special programs of the Act, even if variable control would result in some decreases in air quality and visibility in certain parts of the country. This is because the NSPS authorized in section 111(a) are but one of many Clean Air Act requirements that might be applied by EPA or state agencies to new plants.65 To the extent that there are localized problems with air quality and visibility under national NSPS, the Act permits more stringent requirements to be applied in problem areas in addition to the NSPS. Thus, accepting arguendo Sierra Club's view of the facts — that the new NSPS would not foster air quality and visibility in certain specially protected areas — it still cannot be held as a matter of law that the standard must fail because it impedes the other statutory mechanisms for coping with these local problems.
We find, in sum, that EPA has the authority under section 111 of the Act to promulgate the variable standard and now turn to a consideration of whether EPA's decision to adopt the variable standard was reasonable and supported by the record.
B. The Reasonableness of EPA's Decision to Issue a Variable Standard
In reviewing the merits of EPA's variable standard, our function is to ensure that "the agency, given an essentially legislative task to perform, has carried it out in a manner calculated to negate the dangers of arbitrariness and irrationality in the formulation of rules ... for the future."66 If we find EPA's choice of variable control to be arbitrary and capricious then we will set the standard aside.67 We do not consider the policy issues de novo, substituting our judgment for that of the agency, but evaluate whether the agency has exercised reasoned discretion. This means that the agency must consider all of the relevant factors and demonstrate a reasonable connection between the facts on the record and the resulting policy choice.68
1. Technical Background
The controversy over EPA's justification for variable control centers on two processes for flue gas desulfurization ("FGD" or "scrubbing") referred to by the parties as wet scrubbing and dry scrubbing.69 Scrubbing, in contrast to other techniques for reduction of sulfur emission from coal combustion,70 involves the maintenance of a large scale chemical reaction to clean the smoke produced by coal combustion.71 Typically, as exhaust gases flow up a power plant smokestack, they are exposed to an absorbent medium that is sprayed in their path. Sulfur dioxide in the gas reacts with the chemical absorbent and takes a form which can be collected and removed from the exhaust.72
The type of wet scrubbing process relied on by EPA during this rulemaking was a "properly designed, installed, operated and maintained" lime or limestone FGD system.73 This wet scrubbing system requires that large quantities of lime or ground limestone be mixed with water to form a slurry. The slurry is sprayed into flue gas and absorbs sulfur dioxide which reacts with the slurry to form precipitates (predominantly calcium sulfite and calcium sulfate), which are in turn removed, dewatered, and disposed of in the form of sludge.74 A simplified flow diagram of a wet lime/limestone system is shown as Figure 1 in the appendix to this opinion.
Dry scrubbing is a newer and relatively less established technological alternative to wet lime/limestone scrubbing systems.75 Interest in developing dry scrubbing has been stimulated by perceived advantages over wet scrubbing.76 The dry scrubbing design which EPA focused on during the rulemaking77 removes sulfur dioxide in two stages which incorporate the use of a spray dryer and a baghouse.78 In this system a spray dryer (similar to a wet scrubber) is used with lime, soda ash, or other reagents to scrub sulfur dioxide from flue gases. Unlike wet scrubbing systems, since the flue gas leaving the spray dryer is "hot" (150-180° F) due to the minimal use of water in the spray dryer (by design), no additional reheating of the exhaust plume is required.79 Following the spray dryer, a baghouse is used to collect all particulate matter (including sulfur dioxide reactants).80 Simplified flow diagrams of typical dry scrubbers are shown as Figures 2 and 3 in the appendix to this opinion.
2. EPA's Explanation for the Variable Standard
(a) The Factors Considered By EPA
While the parties dispute the proper analytic method for balancing the relevant factors, they agree, with one exception, on the factors themselves which are relevant to EPA's decision to issue the variable standard. These factors are enumerated in section 111 of the Act and in the legislative history.
Section 111(a)(1), as revised in 1977, requires EPA to weigh cost, energy, and nonair quality health and environmental factors in setting a percentage reduction standard achievable by the best technological system of continuous emission reduction.81 During its passage through Congress the Conferees issued a clarifying statement that EPA may promulgate a variable percentage reduction standard so long as the agency determines that the standard does not undermine the essential purposes of the Act.82 The parties agree that these purposes are as follows:
1. The standards must not give a competitive advantage to one State over another in attracting industry.
2. The standards must maximize the potential for long-term economic growth by reducing emissions as much as practicable. This would increase the amount of industrial growth possible within the limits set by the air quality standards.
3. The standards must to the extent practical force the installation of all the control technology that will ever be necessary on new plants at the time of construction when it is cheaper to install, thereby minimizing the need for retrofit in the future when air quality standards begin to set limits to growth.
4 and 5. The standards to the extent practical must force new sources to burn high-sulfur fuel thus freeing low-sulfur fuel for use in existing sources where it is harder to control emissions and where low-sulfur fuel is needed for compliance. This will (1) allow old sources to operate longer and (2) expand environmentally acceptable energy supplies.
6. The standards should be stringent in order to force the development of improved technology.83
Sierra Club objects that EPA also took account of the impact of alternative standards on future national levels of sulfur dioxide emissions. Paradoxically, Sierra Club argues that "EPA may not consider total air emissions in deciding on a proper NSPS."84 Sierra Club reasons that by specifying only nonair quality health environmental considerations in section 111 Congress meant to exclude EPA's discretion to consider air quality effects of different standards.
We find this position untenable given that one of the agreed upon legislative purposes, set out above, requires that the standards must maximize the potential for long term economic growth "by reducing emissions as much as practicable."85 (Emphasis supplied.) In any event, we can think of no sensible interpretation of the statutory words "best technological system" which would not incorporate the amount of air pollution as a relevant factor to be weighed when determining the optimal standard for controlling sulfur dioxide emissions. Control technologies cannot be "best" if they create greater problems than they solve.86 In fact, we do not see how we could uphold a variable standard if EPA had not evaluated its effect on air emissions.
(b) EPA's Regulatory Analysis
EPA performed a regulatory analysis in three phases to evaluate alternative standards. Phase one of the analysis began before EPA published its proposed standards. Prior to framing alternative standards for consideration, EPA evaluated different control technologies in terms of performance, costs, energy requirements and environmental impacts. EPA also performed a preliminary study of these factors at the national, regional and plantsite levels and toward this end employed econometric computer models to forecast the nature of the utility industry in future years. The initial modeling effort was completed in April 1978 and revised in August 1978.
After this preliminary analysis EPA proposed a uniform 85 percent reduction standard while reserving judgment on whether the uniform percentage reduction standard was preferable to several alternative standards. EPA announced that the final decision would await additional analysis and public comment on the proposal.87 At that time the agency also announced that five methods of "wet" scrubbing were adequately demonstrated and that these technologies could all attain the proposed 85 percent limitation.88 The ensuing rulemaking thus focused not on which technology should be employed, but on the appropriate level of control.89
The regulatory analysis entered phase two following the September 1978 proposal. EPA conducted additional analyses of the impacts of alternative sulfur dioxide standards. The impacts analyzed included total air emissions, utility investment in new plant and pollution equipment, consumer costs, energy production and consumption, coal use, utility consumption of oil and natural gas, and the amount of western low sulfur coal shipped East.90 In addition, supplementary analyses were performed to assess the impact of alternative emission ceilings on specific regional coal reserves, to verify the performance characteristics of alternative scrubbing technologies, and to assess the sulfur reduction potential of coal preparation techniques. As part of the phase two analysis, a joint working group comprised of representatives from EPA, the Department of Energy, the Council of Economic Advisors, the Council on Wage and Price Stability, and others reviewed the underlying assumptions of the econometric model used in the August 1978 analysis and worked to develop new standards for testing in the computer model. As a result of the joint working group's efforts some assumptions were changed91 and a number of alternative standards were defined and considered.92 During phase two EPA also considered public comments on the proposal, identified critical parameters of uncertainty in the model, and revised the model so as to incorporate new considerations such as credits for coal washing which previously had not been properly accounted for in the model. The results of the phase two analysis were published93 and discussed at a public hearing in December 1978.94
Phase three of EPA's regulatory analysis occurred after the public hearing and after the close of the formal comment period on the proposed NSPS. This third stage featured for the first time EPA's formal consideration through its computer model of the impacts of dry scrubbing technology, a subject which will be discussed at length below.95 During phase three the model was run to forecast the impacts of each potential standard, first assuming the use of wet scrubbers only, and then assuming that utilities would use dry scrubbers in situations where it was economically and technologically feasible to do so.96 EPA obtained separate results under the alternative wet scrubbing and dry scrubbing assumptions.97 In brief, the phase three modeling analysis indicated that "[t]he variable control option produces emissions that are equal to or lower than the other options under both the wet scrubbing and dry scrubbing assumptions." Further, under the wet and dry assumptions, variable control as compared to uniform control was predicted to result in more coal capacity in newer and "cleaner" utility plants, to have a clear cost advantage, to use less oil, and to have an equivalent impact on coal production.98
(c) EPA's Stated Rationale for the Variable Standard
EPA's explanation for the adoption of the variable standard is contained in a long preamble accompanying the publication of the final NSPS.99 EPA stated that comments received from advocates of variable control supported a departure from uniform control principally on the basis that variable control "best satisfies the statutory language of Section 111 because it would achieve virtually the same emission reductions at a national level as a uniform approach but at substantially lower costs."100 "In addition [the commentators] argue that a variable control option would provide a better opportunity for the development of ... dry SO2 control systems which they felt held considerable promise for bringing about SO2 emission reductions at lower costs and in a more reliable manner."101 These comments, EPA explained, were the impetus for the phase three modeling analysis. EPA concluded that the results of this further analysis, as reported in the preamble and scrutinized below,102 demonstrated that the variable control option was best. EPA justified the variable standard in terms of the policies of the Act as follows:
The standard reflects a balance in environmental, economic, and energy consideration by being sufficiently stringent to bring about substantial reductions in SO2 emissions (3 million tons in 1995) yet does so at reasonable costs without significant energy penalties. When compared to a uniform 90 percent reduction, the standard achieves the same emission reductions at the national level. More importantly, by providing an opportunity for full development of dry SO2 technology the standard offers potential for further emission reductions (100 to 200 thousand tons per year), cost savings (over $1 billion per year), and a reduction in oil consumption (200 thousand barrels per day) when compared to a uniform standard. The standard through its balance and recognition of varying coal characteristics, serves to expand environmentally acceptable energy supplies without conveying a competitive advantage to any one coal producing region. The maintenance of the emission limitation at 520 ng/j (1.2 lb SO2 million Btu) will serve to encourage the use of locally available high-sulfur coals. By providing for a range of percent reductions, the standard offers flexibility in regard to burning of intermediate sulfur content coals. By placing a minimum requirement of 70 percent on low-sulfur coals, the final rule encourages the full development and application of dry SO2 control systems on a range of coals. At the same time, the minimum requirement is sufficiently stringent to reduce the amount of low-sulfur coal that moves eastward when compared to the current standard. Admittedly, a uniform 90 percent requirement would reduce such movements further, but in the Administrator's opinion, such gains would be of marginal value when compared to expected increases in high-sulfur coal production. By achieving a balanced coal demand within the utility sector and by promoting the development of less expensive SO2 control technology, the final standard will expand environmentally acceptable energy supplies to existing power plants and industrial sources.
By substantially reducing SO2 emissions, the standard will enhance the potential for long term economic growth at both the national and regional levels. While more restrictive requirements may have resulted in marginal air quality improvements locally, their higher costs may well have served to retard rather than promote air quality improvement nationally by delaying the retirement of older, poorly controlled plants.
The standard must also be viewed within the broad context of the Clean Air Act Amendments of 1977. It serves as a minimum requirement for both prevention of significant deterioration and non-attainment considerations. When warranted by local conditions, ample authority exists to impose more restrictive requirements through the case-by-case new source review process. When exercised in conjunction with the standard, these authorities will assure that our pristine areas and national parks are adequately protected. Similarly, in those areas where the attainment and maintenance of the ambient air quality standard is threatened, more restrictive requirements will be imposed.103
Sierra Club insists that EPA's conclusions about dry scrubbing technology are in effect the "cornerstone" of the variable standard. Therefore, Sierra Club argues, the standard must be judged solely on the basis of the validity of EPA's so-called dry scrubbing rationale. We do not agree and believe that it is appropriate to focus on the statutory, substantive, and procedural issues surrounding the dry scrubbing controversy in a later section of this opinion. For now we address Sierra Club's challenges to the standard which are independent of the questions raised about the role of dry scrubbing in the outcome of the final rule.
3. An Examination of EPA's Rationale for the Variable Standard
Sierra Club challenges both EPA's findings about the relative national and regional impacts of alternative standards and the conclusions the agency drew from these findings. First, Sierra Club raises a number of overlapping arguments in support of its view that the findings themselves are methodologically defective. Sierra Club's position, as we understand it, is that the findings cannot amount to substantial evidence because the agency's regulatory analysis which generated the findings was ill-conceived and impermissible under the Act. Specifically, Sierra Club objects to the conceptual framework by which EPA took account of the cost, energy, and environmental considerations mandated by section 111 of the Act. In addition, Sierra Club asserts that the econometric model employed by EPA was so speculative and otherwise unreliable that the modeling results are not substantial evidence. Finally, even accepting EPA's findings, Sierra Club contests EPA's judgment that the variable standard promotes the objectives of the Act. We address these questions seriatim.
(a) The Legitimacy of EPA's Regulatory Analysis
(1) EPA's Authority to Analyze Long Term National and Regional Impacts
Sierra Club argues that section 111 only allows EPA to weigh the cost, energy, and nonair quality health and environmental impacts specified in the statute in order to identify the "best technology." Once that technology is selected, according to Sierra Club, the standard must be set at whatever level is "achievable" by such technology, i. e., the maximum technologically feasible level of control. In this case, since EPA designated wet scrubbing as the technology of choice, the standard must be set at the maximum control level achievable by wet scrubbing. Sierra Club maintains that the kind of macrobalancing EPA has performed has already been done by Congress, and that Congress decided, as reflected in the language of section 111, to require the fullest degree of control that is achievable by the "best technological system" — in this case, a uniform reduction standard of at least 90 percent. The only comparison of aggregate impacts by EPA which Sierra Club believes would have been legitimate under section 111 is a comparison between the impacts of full scrubbing low sulfur coal and the impacts of full scrubbing high sulfur coal. Since there is no doubt that wet scrubbing low sulfur coal is more economical, uses less energy and has fewer other detrimental consequences than wet scrubbing high sulfur coal, Sierra Club argues that this comparison, if anything, demonstrates that the percentage reduction standard should be higher for low sulfur coal than for high sulfur coal. Sierra Club also objects that EPA improperly projected long range future impacts of alternative standards and "insists that the only factors which may be considered by EPA are the technological, economic, environmental, and energy impacts of presently existing technology."104
We reject Sierra Club's restrictive reading of the balancing exercise mandated by section 111 as too narrow to accomplish the purposes of the Act. This is so for several reasons.
1. Sierra Club's interpretation of section 111 is internally inconsistent. Sierra Club would permit EPA to consider the enumerated statutory factors only for the purpose of defining the best technology. This exercise, however, would necessarily involve evaluating the cost, energy, and environmental impacts of different technological systems — not in the abstract — but at some given level of operation. Thus, even the limited determination that Sierra Club would allow EPA to make would logically involve determining at what level a particular control system was "best" in terms of cost, environment, and energy. And so Sierra Club's argument that section 111 requires the standard to be set at 90 percent or higher in the case of wet scrubbing (if technologically achievable) would not follow if wet scrubbing at 85 percent removal efficiency was the "best" in terms of the statutory factors. In short, there is simply no way to set the standard at the level of maximum technological feasibility while simultaneously responding to the cost, energy, and environmental concerns that Congress wrote upon the face of the statute.
2. The language of section 111 not only authorizes variable control but also gives EPA authority when determining the best technological system to weigh cost, energy, and environmental impacts in the broadest sense — at the national and regional levels and over time as opposed to simply at the plant level in the immediate present.105 The pertinent portion of section 111 reads:
[A] standard of performance shall reflect the degree of emission limitation and the percentage reduction achievable through application of the best technological system of continuous emission reduction which (taking into consideration the cost of achieving such emission reduction, any nonair quality health and environmental impact and energy requirements) the Administrator determines has been adequately demonstrated.106
The extraordinary clumsiness of the phrasing of this section does not ease our task of interpretation. Nevertheless, we believe it is clear that this language is far different from the words Congress would have chosen to mandate that EPA set standards at the maximum degree of pollution control technologically achievable. Parsed, section 111 most reasonably seems to require that EPA identify the emission levels that are "achievable" with "adequately demonstrated technology." After EPA makes this determination, it must exercise its discretion to choose an achievable emission level which represents the best balance of economic, environmental, and energy considerations. It follows that to exercise this discretion EPA must examine the effects of technology on the grand scale in order to decide which level of control is best. For example, an efficient water intensive technology capable of 95 percent removal efficiency might be "best" in the East where water is plentiful, but environmentally disastrous in the water-scarce West where a different technology, capable of only 80 percent reduction efficiency might be "best." We cannot believe that Congress meant for EPA to ignore such future aggregate impacts of alternative standards. The standard is, after all, a national standard with long-term effects.
It seems likely that if Congress meant to require a monolithic standard and to curtail EPA's discretion to weigh various policy considerations it would have explicitly said so in section 111, as it did in other parts of the statute. For instance, in prescribing standards for areas that have not succeeded in meeting national ambient air quality standards ("nonattainment areas") Congress required that the standard be the "lowest achievable emission rate," and stressed that cost factors, while still cognizable, were not to play as important a role as they do in selecting the "best" system under section 111.107
3. In addition, when section 111 was amended in 1977 Congress did not narrow EPA's discretion to perform broad based analysis of potential impacts of a standard, even though Congress was aware that previous administrative and judicial interpretations of the former version of section 111 permitted the assessment of the long term costs to industry, consumers, and the environment.108 In Essex Chemical Corp. v. EPA,109 and in Portland Cement Co. v. EPA110 this court specifically endorsed a broad interpretation of "costs" when it held that section 111 did not require a National Environmental Policy Act impact statement. Essential to the court's reasoning was the understanding that "section 111 of the Clean Air Act, properly construed, requires the functional equivalent of a NEPA impact statement."111 The court explained:
The standard of the "best system" is comprehensive, and we cannot imagine that Congress intended that "best" could apply to a system which did more damage to water than it prevented to air.112
Largely as a result of Portland Cement, the "cost" considerations of former section 111 were specifically supplemented by Congress in the 1977 Amendments to cover considerations of nonair quality health and environmental impacts and energy requirements. In so doing, Congress made no attempt to cut back on EPA's ability to apply the new terms broadly nor did Congress reduce the range of discretion that had been read previously into the "cost" factor.
4. The legislative history clearly supports our reading of amended section 111 as authorizing EPA to balance long term national and regional impacts of alternative standards. The Conferees defined the best technology in terms of "long-term growth," "long-term cost savings," effects on the "coal market," including prices and utilization of coal reserves, and "incentives for improved technology."113 Indeed, the Reports from both Houses on the Senate and House bills illustrate very clearly that Congress itself was using a long-term lens with a broad focus on future costs, environmental and energy effects of different technological systems when it discussed section 111.114
5. Broad analysis of alternative standards promotes the several purposes of the Act articulated in the legislative history which Sierra Club recognizes are relevant to the decision to adopt the variable standard. In stating those purposes, Congress indicated that it wanted assurances from EPA that the new standards would not exacerbate existing problems, e. g., produce adverse effects on the coal market, impediments to long term growth, and inhibition of technological innovation. Congress could not have expected such complex interconnecting goals to be satisfied or even approximated without affording EPA a great deal of elbow room to react to and plan for changing events. If EPA was to do as Sierra Club suggests, and set the standard according to the single factor of technological achievability then EPA could not even begin to intelligently balance the stated purposes of the Act.
6. Finally, it is sensible for EPA to assess the likely impacts of the NSPS in 1995 as opposed to an earlier time. Since the standard is only applicable to "new sources" the standard will not have a measurable effect until there are a significant number of new plants in operation. EPA found that this will not occur until 1995 and explained:
Beyond 1995, our data show that many of the power plants on line today will be approaching retirement age. As utilization of older capacity declines, demand will be picked up by newer, better controlled plants. As this replacement occurs, national SO2 emissions will begin to decline. Based on this projection, the Administrator believes that the 1990-1995 time frame will represent the peak years for SO2 emissions and is, therefore, the relevant time frame for this analysis.115
For all of these reasons we conclude that EPA was justified in relying on long term analysis of national and regional cost, environmental, and energy impacts of alternative percentage reduction standards in order to select the "best technological system" upon which to base the NSPS.
(2) The Reliability of EPA's Econometric Computer Model
We are more sympathetic to Sierra Club's complaint about the reliability of EPA's econometric model. Such models, despite their complex design and aura of scientific validity, are at best imperfect and subject to manipulation as EPA forthrightly recognizes.116 The results ultimately are shaped by the assumptions adopted at the outset, and can change drastically for a given range of input data if key assumptions are adjusted even slightly.117 The accuracy of the model's predictions also hinges on whether the underlying assumptions reflect reality, which is no small feat in this volatile world.118
Still we cannot agree with Sierra Club that it was improper for EPA to employ an econometric computer model, or hold as a matter of law that EPA erred by relying on the model to forecast the future impacts of alternative standards fifteen years hence.119
Realistically, computer modeling is a useful and often essential tool for performing the Herculean labors Congress imposed on EPA in the Clean Air Act. In addition to the competing objectives that EPA must satisfy under section 111, the Act explicitly requires EPA to prepare an Economic Impact Statement before promulgating NSPS. The assessment of potential impacts in this statement must be as "extensive as practicable" and must determine the potential inflationary or recessionary effects of the standard, the effects on competition, consumer costs, and energy use.120 Given the complexity and magnitude of the analyses EPA must perform on economic impacts alone, computer modeling, for all its flaws, is invaluable.
Even absent such statutory requirements, we would deem it reasonable to use computer modeling, and to design the model to consider not just "present day" factors, but the consequences over time of the proposed agency action. In American Public Gas Ass'n v. FPC, a challenge was raised that the results of economic modeling, similar to that used here, did not rise to the level of "substantial evidence" necessary to support the agency's findings and conclusions.121 This court disagreed, stating that "[r]easoned decisionmaking can use an economic model to provide useful information about economic realities...."122 However, the agency must sufficiently explain the assumptions and methodology used in preparing the model; it must provide a "complete analytic defense of its model [and] respond to each objection with a reasoned presentation."123 The technical complexity of the analysis does not relieve the agency of the burden to consider all relevant factors and to identify the stepping stones to its final decision. There must be a rational connection between the factual inputs, modeling assumptions, modeling results and conclusions drawn from these results.124
In this case, the utility model itself and its key assumptions were discussed in the proposed rule and background documents.125 EPA invited public comments on the model and its assumptions, with the agency recognizing the sensitivity of the model to a "few key initial assumptions." The joint interagency working group reviewed results of model runs, revised assumptions, and required new runs of the model when it was deemed appropriate. The principal comments received by EPA on the model and the initial assumptions were discussed, together with the results of the three phases of the modeling and the major post-proposal changes to the model, in the preamble to the final NSPS.126 In reviewing this record on the use of the econometric model we have carefully examined, within the limits of our competence, EPA's explanation for the model's premises, the results, and the conclusions drawn therefrom to test them for internal consistency and reasonableness. Although EPA has the benefit of the presumption of good faith and regularity in agency action, we have attempted to ascertain whether the results have been improperly skewed by the modeling format. We conclude that EPA's reliance on its model did not exceed the bounds of its usefulness and that its conduct of the modeling exercise was proper in all respects. We are in fact reassured by EPA's own consciousness of the limits of its model,127 and its invitation and response to public comment on all aspects of the model.128 The safety valves in the use of such sophisticated methodology are the requirement of public exposure of the assumptions and data incorporated into the analysis and the acceptance and consideration of public comment,129 the admission of uncertainties where they exist,130 and the insistence that ultimate responsibility for the policy decision remains with the agency rather than the computer.131 With these precautions the tools of econometric computer analysis can intelligently broaden rather than constrain the policymaker's options and avoid the "artificial narrowing of options that [can be] arbitrary and capricious."132
Sierra Club has not only challenged the use of the model itself, but has also questioned here, as it has throughout the rulemaking, some of the specific assumptions built into the model. In some instances EPA actually adjusted the model to account for Sierra Club's objections and demonstrated that the outcome of the final rule would not have changed. EPA did this, for example, by responding to Sierra Club's recommendations for changes in estimated oil prices and nuclear capacity. EPA's findings and conclusions after adopting Sierra Club's assumptions for future oil prices and nuclear capacity are detailed below.133 The most critical assumptions Sierra Club still objects to are those concerning utility behavior which incorporate what it terms a "perverse hypothesis" that less stringent controls can result in lower total emissions. EPA answered that charge by explaining why variable control could promise equivalent or better reduction of emissions than the stricter full control option:
One finding that has been clearly demonstrated by the two years of analysis is that lower emission standards on new plants do not necessarily result in lower national SO2 emissions when total emissions from the entire utility system are considered. There are two reasons for this finding. First, the lowest emissions tend to result from strategies that encourage the construction of new coal capacity. This capacity, almost regardless of the alternative analyzed, will be less polluting than the existing coal- or oil-fired capacity that it replaces. Second, the higher cost of operating the new capacity (due to higher pollution costs) may cause the newer, cleaner plants to be utilized less than they would be under a less stringent alternative. These situations are demonstrated by the analyses presented here.134
The crucial assumption leading to these findings is that utilities are "cost minimizers."135 The cost minimization assumption implies that when faced with a decision the utility will choose the low-cost option, if risks between the options are neutral. Under the cost minimization model the higher the costs of pollution controls required by the NSPS, the more utilities will delay the retirement of older plants which do not have to comply with the NSPS, and the more utilities will be discouraged from building and operating new plants which must meet the NSPS. Since uniform control is costlier than variable control, uniform control is expected to result in greater reliance on old plants and less utilization of new plants than will variable control, which in turn leads to higher emissions.136 We see no basis for concluding that the adoption of this assumption about utility preferences constituted a clear error of judgment; indeed we are hard pressed to conjure up an alternative assumption about utility behavior that could be put into the computer model.137
(b) The Reasonableness of EPA's Conclusions
(1) The Reasonableness of EPA's Conclusion That Variable Control Reflects a Better Balance of the Section 111 Factors Than Uniform Control138
According to EPA, the variable standard strikes the proper balance between environmental, economic, and energy considerations, whether or not wet scrubbing or dry scrubbing technology is used. We agree because the findings of EPA's regulatory analysis made under the assumption that wet scrubbing technology would be utilized, listed in Tables 2 through 5 from the preamble to the final rule (shown as Figures 5 through 8 in the appendix to this opinion), show clear advantages for adoption of the variable control option over the full control option favored by Sierra Club — apart from any considerations as to the use or savings emanating from the emergent dry scrubbing technology.
Table 2 details EPA's projections for the national levels of sulfur dioxide emissions in 1995. The expected total national emissions for wet scrubbing under both variable control and full control are the same — approximately 20.6 million tons. However, Table 2 reveals that variable control achieves this level of emission with greater utility plant capacity burning more coal, while at the same time generating less sludge. The total projected capacity of plants burning coal in 1995 is 533 GW for the variable control option and 521 GW for the full control option. Wet scrubbing at levels permitted by variable control will result in only 50 million tons of sludge while full control is predicted to generate 55 million tons.
Table 3 details projections of 1995 regional sulfur dioxide emissions. A comparison of the figures listed in Table 3 for wet scrubbing at the variable control and full control levels indicates that there are trade-offs of the different regional impacts of the two standards. The figures show that the East is better off under variable control while the West is better off under full control. With variable control the East will have 300,000 tons less emissions than under full control, but variable control will allow 200,000 tons more emissions in the West than would be experienced with full control in the West. The other two regions — the Midwest and West South Central — can expect roughly the same levels of emissions under either level of control.
Table 4 illustrates the effect of the proposed standards on 1995 coal production, western coal shipped East, and oil consumption by utilities. With regard to these energy impacts the comparative performance of variable control and full control is mixed. National coal production is expected to triple regardless of the level of the percentage reduction standard, from 647 million tons in 1975 to almost 1800 million tons by 1995. Although more western coal will be shipped East under variable control than under full control, 71 million tons versus 59 million tons, both levels of control will result in much lower shipments of coal eastward than the 122 million tons expected under the former standards. In addition, the local production of coal in the East will be slightly greater under variable control than under full control, 470 million tons versus 463 million tons, despite the fact that variable control will result in greater shipments of western coal to the East. Finally, variable control is estimated to involve consumption of 200,000 fewer barrels per day of oil than will full control.
Table 5 shows the expected economic impacts in 1995. Since, as has been noted, the model estimates greater plant capacity under variable control than under full control, the cost of this extra capacity makes the cumulative utility capital expenditures 6 billion dollars higher under variable control than under full control.
Capital expenditures are, however, only part of the overall cost under alternative standards. Annualized cost, for example, includes capital charges, fuel costs, and operation and maintenance costs associated with utility equipment — including pollution control equipment. Table 5 lists annualized cost under different standards as increments over the cost which could be expected under the former standard. EPA projects that despite the greater capital costs, the annual cost of variable control will be half a billion dollars less than the annual cost of full control. In other words, variable control is expected to cost 3.6 billion dollars a year more than the former standard, while full control will cost 4.1 billion dollars a year more than the former standard.
Table 5's figures also demonstrate that variable control is more cost effective than full control. One measure of cost effectiveness is the incremental cost per ton of sulfur dioxide removal. The stated figures for the incremental cost per ton of sulfur dioxide removal are obtained by dividing the expected increase in annual cost by the expected decrease in emissions, as compared to the projected cost and emissions for the former standard. Since both variable control and full control should reduce emissions by 3.1 million tons, variable control is more cost effective than full control because the increased annual cost of variable control is 3.6 billion dollars while the increased annual cost of full control is 4.1 billion dollars, or half a billion dollars more. The incremental cost of sulfur dioxide removal, which is the ratio of the cost figures over the 3.1 million tons of increased emission reduction, is 1,161 dollars/ton for variable control and 1,323 dollars/ton for full control.
Finally, EPA's analysis of wet scrubbing predicted that consumers would incur both lower direct costs (in the form of monthly residential energy bills) and lower indirect costs (reflecting price increases due to higher energy costs) under variable control as opposed to full control.
Results of subsequent selective modeling conducted after the publication of the final NSPS during the pendency of the petitions for reconsideration do not contradict the earlier results discussed above (the results are listed in tables shown as Figures 9 through 11 in the appendix to this opinion). Sierra Club's petition for reconsideration criticized EPA's econometric model because the assumed future oil prices were too low and the assumed growth rate of nuclear power plants was too high. To evaluate the petition EPA reran the computer model, performing what the agency called two "sensitivity tests." These sensitivity tests first assumed higher oil prices and then assumed both higher oil prices and a lower nuclear growth rate, while holding the other modeling assumptions constant. Sierra Club claims "that when EPA ran the model ... with more realistic assumptions concerning oil prices and nuclear power growth, full control showed lower total emissions [than variable control]."139 (Emphasis in original.) But we find that the results of the rerun analysis are entirely consistent with the adoption of a variable standard. The first sensitivity test which assumed only higher oil prices produced no significant changes in the relative advantages of the alternative control options, and both sensitivity tests indicated that variable control was the most attractive option.140 Even though the sensitivity test which assumed both higher oil prices and lower nuclear capacity showed, for the first time during the modeling analysis, that full control would produce lower national emissions than variable control, the difference in total emissions between the two options was only 100,000 tons. "[T]he cost of this additional 100,000 tons of control was estimated at $1.8 billion, which represents more than a 40 percent increase in the ... cost."141 EPA further explained:
The principal environmental benefit of full control [assuming both higher oil prices and lower nuclear capacity] would be felt in the West and West South Central. Through case-by-case new source review ample authority exists [under the Clean Air Act] to require more stringent controls as necessary to protect our pristine areas and national parks.... As a result, the Administrator continues to believe that the flexibility offered by the variable standard will lead to the best balance of energy, environmental, and economic impacts....142
For these reasons we do not believe that the post-promulgation modeling analysis provides convincing evidence that full control is preferable to variable control.
In sum, the results of EPA's econometric modeling which forecast substantial benefits to be obtained by adopting the variable standard, provide adequate support for EPA's decision to select that course on the basis of the environmental, energy, and cost factors specified in section 111.
(2) The Reasonableness of EPA's Conclusion That Variable Control Promotes the Policies of the Act
The results of EPA's regulatory analysis also persuade us that the variable standard does indeed advance policies of the Act other than those specifically incorporated in section 111.
First, variable control is not — as Sierra Club alleges — inconsistent with the purposes underlying the Act's programs for the prevention of significant deterioration of air quality, and the provisions for eliminating the impairment of visibility in certain designated areas.143 Specifically, because EPA predicts that variable control will produce equivalent or lower total emissions of sulfur dioxide than any other control option, variable control protects air quality and visibility at least as well as any other standard.144 Regionally, all the control options produce about the same emissions in the Midwest and West South Central regions.145 Although variable control might result in higher annual emissions in the West than would be expected under full control, full control might yield even a greater increment of air pollution in the East.146 We agree that "it would not have been a reasonable exercise of discretion to impose additional costs of over a billion dollars per year merely in order to transfer several hundred thousand tons of sulfur dioxide annually from the West to the East."147 Nonetheless, we are not insensitive to possible regional hardships and do not mean to imply that they may be ignored. The NSPS, as EPA recognizes, are only a minimum national standard, and there are other mechanisms provided in the Clean Air Act which should be activated in appropriate circumstances to protect troubled areas.148
Second, the findings also support EPA's determination that variable control serves what the parties agree are the relevant purposes of section 111, which the legislative history says must be accommodated whenever EPA chooses to vary the percentage reduction standard. For example, the competitive advantage previously enjoyed by some states under former standards will be eased, since all new coal-fired sources are subject to the same emissions ceiling and all must apply some level of continuous emission reduction technology to control sulfur dioxide emissions. Consequently the standards assist in eliminating the advantage of using only low sulfur coal throughout the country or of relocating to areas where scrubbing was not previously required because low sulfur coal was available locally. The advantage of the lower percentage reduction requirement available to plants burning low sulfur coal is offset to a degree by countervailing considerations, such as the costs of mining, transportation, and relocation, competition for supply, and state regulations.
Other purposes of section 111 also appear to be well served. One highlighted in the Conference Committee Report was "maximizing the use of locally available fuels."149 EPA found that the 70 percent minimum floor was "sufficiently stringent to reduce the amount of low-sulfur coal that moves eastward when compared to the [former] standard. Admittedly, a uniform 90 percent requirement would reduce such movements further, but ... such gains would be of marginal value when compared to expected increases in high-sulfur coal production."150 EPA's figures show that 59 million tons of low sulfur coal would be shipped East under full control while 71 million tons would be shipped East under variable control. In light of the projected increase in coal production, from 647 million tons in 1975 to close to 1800 million tons in 1995151 we find no abuse of discretion in EPA's determination that the additional 12 million tons projected to be shipped East under variable control would not significantly interfere with the use of locally available coal. We are not especially troubled by Sierra Club's concern that under variable control, utilities in the Midwest "can choose between burning local high-sulfur or imported Western low-sulfur coal."152 Added flexibility is an attractive feature of variable control for many reasons, and among other things, will promote a more "balanced coal demand within the utility sector."153 In addition, we are informed by Intervenor MAMU that variable control will increase the propensity of some midwestern utilities to use local coal rather than to import low sulfur coal from the West.154
The combination of stringency and flexibility afforded by variable control satisfies still another stated purpose of section 111: freeing low sulfur coal for use in existing plants where it is harder to control emissions and where low sulfur fuel is needed to achieve compliance. The expected increase in the production of all types of coal, by itself, would probably provide sufficient supplies of low sulfur coal to existing "dirty" plants. Under variable control, however, these older plants should be retired at a more rapid rate than under full control. Thus, as low sulfur coal supplies grow, the demand from older plants should decline. Variable control should also reduce the number of long distance shipments of low sulfur coal to newer plants since some scrubbing will be required whatever kind of coal is used. It will become just as economical for some utilities, like the midwestern plants noted above, to burn higher sulfur content local coal, rather than to consume more distant low sulfur coal.
Overall, we must conclude that EPA's rationale for the variable standard — completely apart from dry scrubbing — is supported by the agency's findings, and that the agency's rationale demonstrates that the variable standard satisfies both the language of section 111 and the policies enumerated in its legislative history.
C. The Dry Scrubbing Controversy
Notwithstanding the reasons proffered by EPA for adopting the variable standard under the assumption that utilities would use wet scrubbing technology, Sierra Club argues that EPA's findings and conclusions about the emergent dry scrubbing technology are actually the cornerstone of the agency's rationale for the variable standard. For this reason, Sierra Club maintains, the variable standard must be set aside because: (1) EPA is not authorized by section 111 to consider the impact of the NSPS on the development of a new technology like dry scrubbing and (2) EPA's determinations about the efficiency of dry scrubbing technology, and the means of encouraging its development, are not supported by the record.
We address these arguments because EPA's published rationale for the variable standard attests to the important role dry scrubbing technology played in the evolution of the rule. As we have shown, EPA's justification for the rule in terms of statutory criteria does not depend on dry scrubbing technology. However, the agency's explanation of the evolution of the standard indicates that the impetus for EPA's examination of the 70 to 90 percent control option in the post-comment period phase three analysis stemmed from the agency's desire to reexamine the potential of dry scrubbing.155 According to EPA's rationale the development of dry scrubbing technology was perceived as a major bonus of promulgating the variable standard.156 We conclude that EPA's consideration of dry scrubbing as a reason for its selection of a nonuniform standard is consistent with the agency's authority under section 111, and that there is support in the record for doing so. In a final section we review the procedural objections that Sierra Club raises to EPA's consideration of dry scrubbing.
1. The Role of Dry Scrubbing Technology in EPA's Rationale for the Variable Standard
In the preamble accompanying the proposed NSPS, EPA referred only in passing to dry scrubbing. Dry scrubbing was described as one of several new and promising "emerging technologies" with considerable "potential" but not yet sufficiently demonstrated to be a basis for the proposed standard.157 The background documents to the proposal also gave limited attention to dry scrubbing.158 The proposed standard was clearly based on the level of emission control achievable by wet scrubbing systems, although EPA noted that "the use of other systems should not be discouraged. In this regard, a number of emerging technologies show promise."159
When the results of the phase two analysis were published in December 1978 EPA did not mention dry scrubbing at all. By that time EPA had received some comments from advocates of dry scrubbing and at the public hearing held that month testimony was given favoring a 70 percent floor on the percentage reduction standard so as not to preclude the use of dry scrubbers.160
Dry scrubbing was first incorporated into EPA's regulatory analysis in phase three — conducted after the close of the comment period. In the preamble to the final rule EPA explained that phase three modeling introduced:
the incorporation of dry SO2 scrubbing systems. Dry scrubbers were assumed to be available for new and retrofit applications. The costs of the systems were estimated ... based on pilot plant studies and contract prices for systems currently under construction. Based on economic analysis, the use of dry scrubbers was assumed for low-sulfur coal [less than 3 lbs./MBtu] applications in which the control requirement was 70 percent or less. For higher sulfur content coals wet scrubbers were assumed to be more economical.161
The results of the phase three analysis showed that variable control was preferable to uniform control under the wet scrubbing assumption, and to an even greater extent under the dry scrubbing assumption.162 (The results are listed in Figures 5-8 in the appendix to this opinion.)
These findings of the phase three regulatory analysis were published in June 1979 when the final rule was announced. The preamble to the final rule, in contrast to the agency's statement at the time of the proposal and after phase two, is replete with references to both wet scrubbing and dry scrubbing. The preamble reveals that both pollution control technologies played a role in EPA's decision. For example, EPA stated:
Although the standards are based on [wet] lime or limestone FGD systems, other commercially available [wet] FGD systems ... are also capable of achieving the final standard. In addition, when specifying the form of the final standards, the Administrator considered the potential of dry SO2 control systems....163
EPA's rationale for the final NSPS notes that the critical issue of variable control "was made more complex by the emergence of dry SO2 control systems."164 EPA claimed that "[a]s a result of public comments on the discussion of dry SO2 technology in the [September 1978] proposal, the EPA staff examined the potential of this technology in greater detail."165 EPA explained that it found dry scrubbing to be "progressing rapidly" as an attractive alternative to wet scrubbing but that the applicability of dry scrubbing was limited to low sulfur coal.166
EPA's rationale continued:
Faced with these findings, the Administrator had to determine what effect the structure of the final regulation would have on the continuing development and application of this technology. A thorough engineering review of the available data indicated that a requirement of 90 percent reduction in potential SO2 emissions would be likely to constrain the full development of this technology by limiting its potential applicability to high alkaline content, low-sulfur coals. For non-alkaline, low-sulphur coals, the certainty of economically achieving a 90 percent reduction level is markedly reduced. In the face of this finding, it would be unlikely that the technology would be vigorously pursued for these low alkaline fuels which comprise approximately one half of the Nation's low-sulfur coal reserves. In view of this, the Administrator sought a percentage reduction requirement that would provide an opportunity for dry SO2 technology to be developed for all low-sulfur coal reserves and yet would be sufficiently stringent to assure that the technology was developed to its fullest potential. The Administrator concluded that a variable control approach with a minimum requirement of 70 percent reduction potential in SO2 emissions (30-day rolling average) for low sulfur coals would fulfill this objective.
In addition to promoting the development of dry SO2 systems, a variable approach offers several other advantages often cited by the utility industry [advantages listed] .... While these points alone would not be sufficient to warrant adoption of a variable standard, they do serve to supplement the benefits associated with permitting the use of dry scrubbing.167
Sierra Club insists that this statement amounts to a concession that the agency's justification for variable control hinges on dry scrubbing and urges the following interpretation of the above-quoted language:
In the final rule-making the agency conceded that, based only on a consideration of wet scrubbing techniques, a deviation from uniform control should not have been adopted, and that any supposed advantages for partial controls "alone would not be sufficient to warrant adoption of a variable standard...."168
We do not believe, however, that a fair reading of the passage supports Sierra Club's interpretation. Rather the language indicates only that the incidental benefits of variable control, at the micro or plant level169 "alone would not be sufficient to warrant adoption of a variable standard" without regard to the macro or national level economic, environmental, and energy considerations required by section 111 of the Act. Thus read, it is by no means a concession that the regulatory analysis of the alternative wet scrubbing scenario does not support variable control, but only an enumeration of the "supplemental [plant level] benefits associated with permitting the use of dry scrubbing."170
In fact, EPA's rationale carefully distinguishes between the macro level considerations supporting variable control and the additional advantages of developing dry scrubbing:
By fashioning the SO2 standard in this manner, the Administrator believes he has satisfied both the statutory language of section 111 and the pertinent part of the Conference Report. The standard reflects a balance in environmental, economic, and energy considerations by being sufficiently stringent to bring about substantial reductions in SO2 emissions ... yet does so at reasonable costs without significant energy penalties. When compared to a uniform 90 percent reduction, the standard achieves the same emission reductions at the national level. More importantly, by providing an opportunity for full development of dry SO2 technology the standard offers potential for further emission reductions ..., cost savings ..., and a reduction in oil consumption ... when compared to a uniform standard.171
Our own reading of EPA's rationale convinces us that although dry scrubbing is not relied on as the exclusive or even primary justification for the variable standard, the agency's consideration of dry scrubbing did influence the course of the rulemaking by suggesting the 70 percent figure for a variable control standard that was tested and ultimately adopted. We have little doubt, based on a thorough reading of the record, that if EPA had never brought dry scrubbing to the forefront in the later stages of the rulemaking, the agency might still have selected and successfully defended the reasonableness of the variable standard. However, that is not what occurred here. Consequently our review of the standard turns to the question of whether it was legitimate under section 111 for EPA to take account of dry scrubbing in the way it did.
2. The Legitimacy of Considering Emerging Technology in Setting Section 111 Standards
Sierra Club strongly argues that EPA had no business worrying about emerging technology because section 111 does not explicitly include "technological innovation" as one of the factors EPA is to balance along with considerations of cost, energy and nonair health and environmental impacts. In fact, Sierra Club argues, Congress enacted a separate and exclusive provision, section 111(j),172 to deal with the subject of technological innovation. Section 111(j) authorizes EPA to encourage new technology through the grant of waivers from the NSPS to those sources of emissions planning to use "an innovative technological system or systems of continuous emission reduction."173
We do not believe that EPA is precluded from encouraging technological innovation through the NSPS either because section 111(a) prohibits the agency from doing so or because section 111(j) is meant to be the exclusive statutory mechanism for promoting new methods of pollution control.
Our interpretation of section 111(a) is that the mandated balancing of cost, energy, and nonair quality health and environmental factors embraces consideration of technological innovation as part of that balance. The statutory factors which EPA must weigh are broadly defined and include within their ambit subfactors such as technological innovation.
We have no reason to believe Congress meant to foreclose in section 111(a) any consideration by EPA of the stimulation of technologies that promise significant cost, energy, nonair health and environmental benefits. Our view is consistent with legislative history, which reveals Congress' concerns that the NSPS should not stymie technological innovation.174 So long as EPA considers innovative technologies in terms of their prospective economic, energy, nonair health and environmental impacts the agency is within the scope of its authorized analysis.175 This is not to say, however, that NSPS may be relaxed just to accommodate an uncertain and unproven technology. Rather, when balancing the enumerated factors to determine the basic standard it is appropriate to consider which level of required control will encourage or preclude development of a technology that promises significant advantages with respect to those concerns.
According to our reading, section 111(j) supplements rather than restricts EPA's discretion under section 111(a) to encourage innovative technology. Section 111(j) specifies the circumstances in which EPA may grant waivers from NSPS to individual plants using innovative technology which has not been adequately demonstrated and which will inter alia:
achieve greater continuous emission reduction than that required to be achieved under the standards of performance which would otherwise apply, or achieve at least an equivalent reduction at lower cost in terms of energy, economic, or nonair quality environmental impact....176
In our view, a symbiotic rather than mutually exclusive relationship exists between the NSPS of section 111(a) and the technology waivers of section 111(j). Section 111(j) waivers are limited to individual plant situations where the technology is capable of eventually achieving a level of emissions reduction which is equal to or greater than that specified in the section 111(a) NSPS. Thus, the ability to grant section 111(j) waivers would be unduly restricted or entirely foreclosed if the NSPS were set too high. If Congress intended any substantial use of 111(j) waivers, as we presume that it did, it is only plausible that Congress also meant for emerging technologies to be given consideration when EPA promulgated NSPS.
The Act does not state that section 111(j) is the exclusive provision for dealing with technological innovation. Elsewhere in the Act, however, Congress has explicitly indicated when EPA's discretion to work toward a particular end is limited to the confines of a specific provision.177 Absent such language in section 111(j), we hesitate to say that EPA is limited to the narrow terms of this section in dealing with the important area of new and developing technologies, particularly in situations where section 111(j) waivers to individual plants would not provide sufficient incentives to spread new technology throughout the industry.178
In short, we can find no statutory bar to EPA's consideration of how various NSPS options will affect the development of new technologies which have economic, energy, and environmental implications of their own.
3. The Adequacy of the Record for Dry Scrubbing's Role in EPA's Rationale
EPA has explained that its evaluation of the technical capabilities and commercial attractiveness of dry scrubbers precipitated the decision to test the 70 percent floor option, and its desire to encourage the development of dry scrubbing reinforced the agency's preference for the variable standard. We now proceed to examine whether there is a reasonable basis in the record for the agency's conclusions about the appropriate level of control in light of the potential of dry scrubbers.
We note that the nature of our inquiry here is different from that required had EPA relied on dry scrubbing as adequately demonstrated technology and found that the variable standard was the best level of control — in light of the relevant considerations — achievable by dry scrubbing. Here the variable standard was chosen as the best level of control achievable by wet scrubbing, admittedly an adequately demonstrated technology.
We also do not confront the situation where EPA has set a standard at a particular level believed necessary to promote a new technology, even though the agency's regulatory analysis demonstrated that another level of control was optimal, after considering all relevant factors apart from the new technology. In such a case our close scrutiny of the factual basis for the standard would involve a balance: we would have to weigh the magnitude of the standard's departure from the level of control otherwise indicated by the agency's regulatory analysis against the weight of the support in the record that substantial benefits would eventually accrue from adjusting the standard on the basis of the new technology. The greater the imprint of the new technology on the final rule, the more demanding our review of the evidence about the potential benefits and capabilities of new technology. By the very nature of its newness, it would be inevitably harder for EPA to acquire as precise and complete information about the emerging technology as would be available in the case of older, more established technologies. Thus the difficulty of justifying a standard that diverges from a level determined by weighing cost, energy, and environmental effects of the best presently demonstrated technology, solely to account for new technology, should provide built-in safeguards against overuse of such a justification and prevent circumvention of the primary statutory goals.
The present situation does not present such a risk because the control level EPA believed was necessary to promote dry scrubbing was found to be optimal even aside from dry scrubbing considerations. This circumstance obviates the need to determine whether EPA's conclusions about dry scrubbing can be relied upon as an independent basis for the variable standard. We do require, however, that the record substantiate the reasonableness of EPA's preliminary step of turning to a standard with a 70 percent floor for the purpose of comparing its potential impacts with other control options, and basing its subsequent adoption of that 70 percent standard, even partially, on the prospects of encouraging development of dry scrubbing.
Sierra Club's quarrel in this regard is not with EPA's judgment that 70 percent removal on low sulfur coal is technologically and economically feasible for dry scrubbers. On the contrary, as Sierra Club argues in advocating a higher standard, the record contains considerable evidence which indicates that dry scrubbing can achieve significantly better than the 70 percent floor adopted by EPA.179 EPA itself announced in the proposed NSPS that dry scrubbing appeared capable of eventually matching the performance of wet scrubbing.180
Sierra Club does object, however, to EPA's conclusion that the 70 percent standard is necessary to encourage utilities to use dry scrubbers. Much of the record evidence concerning whether it is appropriate to set the minimum standard at 70 percent in order to promote the development of dry scrubbing technology revolves around the assertion that dry scrubbing low alkaline coal is more expensive than dry scrubbing high alkaline coal because of "uncertainty in the stoichiometric ratio" required to achieve a 70 percent reduction for nonalkaline coals.181
Sierra Club's response is that the "alkaline" issue is a red herring.182 EPA's concern over "uncertainty in the stoichiometric ratio" means only — according to Sierra Club — that EPA does not know precisely how much more additional lime will have to be added to a dry scrubber system when the natural alkalinity of the coal is too low to obtain the necessary results in the sulfur dioxide removal process. However, the need to add reagent to low alkaline coal is not unique to dry scrubbing — lime or limestone reagents are frequently required in wet scrubbing systems burning low alkaline coal.183
The implications of the "uncertainty in the stoichiometric ratio" are thus principally financial. If the cost of the extra lime injected into the dry scrubber is too high, presumably it will make the system economically unattractive for removal efficiencies greater than 70 percent. We have examined the source materials containing the economic analysis offered in support of this conclusion, and find only limited substantiation for the perception that 70 percent removal efficiency is the limit of economical dry scrubber performance, or that the extra cost of scrubbing low alkali coal would tend to make dry scrubbing economically less attractive than wet scrubbing at a higher than 70 percent level.184
There is, however, other evidence in the record concerning EPA's prediction that utilities will refrain from dry scrubbing if the minimum standard is higher than 70 percent or conversely be enticed to use them if the standard drops to 70 percent. Witnesses at the December 1978 hearing voiced a preference for relaxing the percentage reduction standard on the basis that it would encourage dry scrubbing. Two comments received by EPA proposed 70 percent as the preferred level of control for that purpose.185 The Electric Utilities, as the spokesman of the industry, also submitted comments which emphasized that a standard higher than 70 percent would inhibit utilities from trying dry scrubbing since they would perceive no advantage in using an emerging and still risky technology over a demonstrated one like wet scrubbing.186 This view in turn must be considered in light of other evidence that several plants have in fact opted for dry scrubbing for performance levels greater than 70 percent,187 while still other utilities operating at the 70 percent level are choosing wet scrubbers.188 Overall EPA has projected that most new plants would, in fact, install wet scrubbers.189
In sum, the support in the record for selecting 70 percent as the magic percentage for encouragement of dry scrubbing is less than overwhelming. However, three things do stand out on the record. First, dry scrubbing has significant potential as a cheaper, energy conserving, and environmentally sound alternative to wet scrubbing. Second, despite the considerable potential advantages of dry scrubbing systems over wet scrubbing systems, dry scrubbing is not yet a proven or "adequately demonstrated" technology and its future is uncertain. Third, all of the testing of dry scrubbing performance has been done on high alkaline coal, while low alkaline coal makes up 50 percent of national coal reserves and 90 percent of the low sulfur coal reserves in the West.
On the basis of this evidence we find it was reasonable for EPA to seek to encourage dry scrubbing and to be concerned with the effect of the NSPS on the future of the new technology. Furthermore, on the basis of the entire record it was not unreasonable for EPA to believe that the leeway afforded by the variable standard from the requirements of the stricter 90 percent reduction requirements would promote dry scrubbing. Given the state of this record we would have been reluctant to uphold EPA's discretion to vary the standard solely on the basis of dry scrubbing, but we are satisfied that it was legitimate for EPA to find that variable control with a 70 percent floor was an option that would be worthwhile to further examine in terms of its national and regional impacts under alternative wet scrubbing and dry scrubbing assumptions.
To put our conclusion upholding the 70 percent standard in further perspective: the 70 percent minimum level of emission control is permissible only when the overall emission level is reduced from 1.2 lbs./MBtu to .60 lbs./MBtu. Therefore, it cannot be contended that the new NSPS are more lax than the former NSPS, but only that when a plant halves its total emissions, then the percentage reduction standard does not independently require a uniformly high percentage reduction at all times — regardless of the sulfur content of the coal burned or other policy considerations such as fostering the availability of cheaper, cleaner, energy saving technologies.
We now complete our review of the variable standard by considering whether the substantively acceptable standard was procedurally defective.
D. The Adequacy of Notice and the Opportunity to Comment on the Rationale for the Variable Standard
Sierra Club complains that EPA's focus on dry scrubbing and the determinative phase three regulatory analysis came only after the public comment period closed in January 1979; hence, interested parties in the rulemaking were not informed of these developments in time to make meaningful comments before the final rule issued in June 1979. Sierra Club argued in its petition for reconsideration that a new proposal should have been resubmitted for public comment at the point that EPA seriously began to consider the variable standard ultimately adopted.190
The procedural requirements for notice and comment applicable to this case are specified in section 307(d) of the Act,191 which is discussed below in our review of EDF's appeal of the 1.2 lbs./MBtu emission ceiling. For now it suffices to state that section 307(d) provides for published notice of a rulemaking to be accompanied by a statement of its basis and purpose which includes (a) the factual data on which it is based, (b) the methodology used in obtaining and analyzing the data, and (c) the major legal and policy considerations underlying the rule. All documents which become available after the proposed rule has been promulgated which are "of central relevance to the outcome of the rule" must be entered into the docket, and the agency must allow enough time for participants in the rulemaking to respond to those documents with comments.192 However, these provisions do not require EPA to select a final rule from among the precise proposals under consideration during the comment period. Rather, incremental changes are permissible so long as the final rule is a "logical outgrowth" of the proposals highlighted and discussed during the notice and comment period.193 It is entirely proper and often necessary for the agency to continue its deliberations and internal decisionmaking process after the close of public comment in order to assimilate those comments and arrive at a policy choice. Our decision here involves a question of degree: whether a sliding scale 70-90 percent reduction requirement is a logical outgrowth from the numerous alternative options discussed by EPA in its September 1978 proposal and in the December 1978 interim announcement. In addition we must decide if EPA's late consideration of dry scrubbing was such a detour from the course of the rulemaking that in fairness to the public and interested parties the agency should be required to retrace its steps, give notice of its new focus, and reopen the comment period.
EPA, of course, denies the existence of any procedural irregularity in the promulgation of the variable standard. When the agency denied Sierra Club's petition for reconsideration it stated that the notion that "the phase 3 analysis was a new venture" was "false."194 According to EPA, phase three featured no new modeling concepts or input assumptions, only "refinements" of earlier analysis. EPA insists that the 70 percent level of control which was introduced in order to consider the impact of dry scrubbing, raises basically the same policy issues as all the percentage reduction options previously analyzed.
While we find that Sierra Club's criticism of the procedural history of the variable standard is not unwarranted, we conclude that there was sufficient notice and comment on matters which are "of central relevance" to support EPA's rationale for the variable standard. It is indeed true that the 70 percent level of control did not surface as a serious contender until after the final bell sounded in the public arena, and many of the supporting documents related to dry scrubbing were inserted into the record in a last round flurry of agency activity. Yet it is important to evaluate EPA's ultimate selection of 70 percent as the floor for the standard in the context of the agency's decision to adopt some form of non-uniform sliding scale standard. There can be no doubt that the latter decision benefited from ample notice and comment. EPA stated at the time of the proposal that it might adopt a variable percentage reduction standard as opposed to the proposed 85 percent uniform standard, and emphasized that "[t]he principal issue associated with this proposal is whether electric steam generating units firing low-sulfur content coal should be required to achieve the same percentage reduction in potential SO2 emissions as those burning higher sulfur content coal."195 The preamble to the proposed rule continued:
Resolving this question of full versus partial control is difficult because of the significant environmental, energy, and economic implications associated with each alternative. The Administrator has not made a final decision on which of the alternatives should be adopted in the final standard and solicits additional data on these impacts before promulgating the final regulations.196
(Emphasis supplied.) Throughout the rulemaking period from December 1977197 until February 1980,198 extensive public comment centered on the issue of whether the percentage reduction should be uniformly applied.
The possibility of adopting a "sliding scale" standard varying with the sulfur content of coal was also discussed during and after the comment period. EPA considered several sliding scale options in terms of their economic, energy, and environmental impacts. These options included two sliding scales with maximum control at 85 percent, one of these with minimum control at the 20 percent level initially favored by the Electric Utilities199 and the other with minimum control at the 33 percent favored by the Department of Energy.200 In addition, EPA analyzed a 20 to 82 percent sliding scale and two options with a uniform 90 percent requirement in the West and sliding scales in the East (33 to 90 percent and 50 to 90 percent).201 After the close of the public comment period, EPA evaluated the 70 to 90 percent alternative and published the results under both wet and dry scrubbing assumptions. Finally, after promulgation of the final rule EPA considered the sliding scale option at the 50 percent minimum level urged by the Electric Utilities in their petition for reconsideration.202
The record also shows that the phase three analysis with a 70 percent minimum floor was a continuation of the type of regulatory analysis performed during phases one and two. Aside from adjustments in the model to account for the assumption that utilities would use dry scrubbers, there were no radical changes in the econometric model that tested the likely impacts of the 70 to 90 percent control option. The analytical format that EPA relied on in phase three was subject to broad review in phases one and two by both the public participants in the rulemaking and by the joint interagency working group.203 In fact, Sierra Club's objections to the phase three modeling are virtually identical to the criticisms of the phases one and two modeling that Sierra Club registered with the agency.204 EPA repeatedly solicited comments on the model, and on at least two occasions it responded to public comment by adjusting the analysis.205
The subject of dry scrubbing and the question of how the NSPS would affect emerging technology were not introduced for the first time in phase three. On the contrary, there was a discussion of these matters in December 1977 — well before EPA even proposed the NSPS — at a public hearing held by the National Air Pollution Control Techniques Advisory Committee (NAPCTAC).206 Both dry scrubbing and the concern for emergent technology were discussed (though not emphasized) by EPA in the preamble to the proposed rule and the background support and technical documents.207 The Electric Utilities filed initial comments in December 1978208 and filed supplemental and reply comments in January 1979209 which advocated varying the percentage reduction standard from the maximum feasible level of wet scrubbers in order to encourage new technologies like dry scrubbing. Individual utilities and local governmental agencies testified and filed comments which detailed the capabilities of dry scrubbers and the advantages of promoting the use of this technology.210 On the other hand, the comments filed jointly in January 1979 by EDF and the Natural Resources Defense Council ("NRDC") opposed EPA's statutory authority to adjust the NSPS to account for new technology like dry scrubbing.211 Sierra Club apparently did not comment on dry scrubbing or emerging technologies until after the final NSPS were promulgated.212
The record further shows that the parties had actual notice before the final rule was published in June 1979 of EPA's focus on the 70 to 90 percent variable standard and of its heightened interest in dry scrubbing.213 For example, on April 20, 1979, NRDC and EDF jointly wrote to EPA vigorously criticizing EPA's "new rationale" for allowing a sliding scale standard with a 70 percent minimum removal requirement in order "to avoid stifling the development of certain new technologies, such as dry scrubbing."214 The letter was docketed in the administrative record on May 2, 1980. The Electric Utilities' letter of April 23, 1979 also discussed the 70 to 90 percent standard and the importance of dry scrubbers to EPA's rationale for this alternative standard.215 Copies of this letter were sent to interested parties including Sierra Club, EDF, and NRDC.
Beginning in early 1979 and continuing until the publication of the final rule, EPA prepared and entered on the record several documents concerning dry scrubbing. These documents included EPA's economic comparison of dry scrubbing with wet scrubbing,216 a memorandum on the technical capability of dry scrubbers, and material on recent developments in dry scrubbing technology.217 The preamble to the final rule and the supporting background document incorporated large portions of these late entries into their text.
Viewing this record as a whole, we conclude that a variable sliding scale percentage reduction requirement was generally understood to be a serious possibility from the start of the rulemaking and that the issue of whether the standard should accommodate emerging technologies was raised by public comments. We also find that there was information on the potential of dry scrubbers in the record from the beginning of the rulemaking, although this was generously supplemented at the last stages of the proceeding.
Nonetheless, since the record contains the alternative "wet scrubbing" justification for the variable standard, and since the parties knew about EPA's growing interest in dry scrubbing, we cannot say that the failure to provide for more time to comment on the late entries into the dry scrubbing controversy is "so serious and related to matters of central relevance to the rule that there is a substantial likelihood that the rule would have been significantly changed if such errors had not been made."218 Otherwise, we find that the 70 percent minimum level of control came within the scope of alternative standards previously considered which ranged from a 20 percent minimum to a 90 percent maximum, and was not so significant a departure from prior sliding scale proposals as to require a new public comment period.
This rulemaking was by no means a neat and tidy proceeding, and it might well have been the wiser course if EPA had chosen to publish a new proposal for another round of comment,219 but we cannot say that the absence of new notice and comment is a fatal defect. For these reasons we decline to remand the variable standard to EPA on the procedural grounds argued by Sierra Club.
III. THE 90 PERCENT REMOVAL STANDARD
The Electric Utilities attack the NSPS insofar as they require a 90 percent reduction in the potential sulfur dioxide emissions from the combustion of high sulfur coal (at least 6 lbs./MBtu) measured on a thirty day rolling average.220 EPA's position is that the record demonstrates that the 90 percent standard is achievable even in the most difficult situations by the use of precombustion fuel preparation techniques such as physical coal cleaning ("PCC" or coal washing) followed by post-combustion wet scrubbing of flue gas ("FGD" or scrubbing). The Electric Utilities contend that when the 90 percent standard was promulgated it was based on the use of FGD alone and that it is procedurally improper for EPA to rely on a post hoc rationale for the standard which depends on the combined use of FGD and coal washing. The Electric Utilities also say that the 90 percent standard is not adequately demonstrated to be achievable even if FGD and coal washing or any other technology are used together. They point to several alleged deficiencies in the record and in EPA's statistical analysis to illustrate the lack of support for the agency's conclusions about the achievability of the standard.
A. Notice as to the Basis of the 90 Percent Standard
The Electric Utilities argue that both the proposed and the final NSPS were based on the level of emission reduction achievable by the application of FGD technology alone. Then, when the Electric Utilities showed in their petition for reconsideration that the 90 percent standard was not achievable by scrubbing alone, they say EPA, in denying the petition, impermissibly redefined the basis for the standard to include a combination of coal washing and scrubbing. Alternatively, the Electric Utilities maintain that even if the final rule was based on combined technology, EPA violated the procedural provisions of the Act which require that the promulgated rule must be accompanied by an explanation of any major changes in the final rule from the proposed rule.221
1. The Basis of the Final Standard
EPA steadfastly insists that the final standard was clearly based on the performance of scrubbing in conjunction with precombustion technologies like coal washing and admits that "[i]f the standard were based solely on FGD, it would probably not be achievable."222 The Electric Utilities, EPA claims, are "trying to distract the Court from EPA's analysis of the actual standard by erecting and destroying a straw man."223 We do not believe that the Electric Utilities' contention may be dismissed quite that easily.
It is undisputed that the proposed percentage reduction standard — a uniform 85 percent removal requirement measured on a 24 hour period — was based on the capability of "well designed, maintained and operated flue gas desulfurization systems" alone.224 The proposal specifically, albeit briefly, discussed and permitted a credit for fuel pretreatment such as coal washing to be counted toward meeting the 85 percent standard.225 EPA stated, however, that fuel pretreatment was not required to achieve the proposed standard.226
In the preamble to the final rule, the technological basis for the 90 percent standard was less explicit:
Under section 111(a) of the Act, a standard of performance for a fossil-fuel-fired stationary source must reflect the degree of emission limitation and percentage reduction achievable through the application of the best technological system of continuous emission reduction taking into consideration cost and any nonair quality health and environmental impacts and energy requirements. In addition, credit may be given for any cleaning of the fuel, or reduction in pollutant characteristics of the fuel, after mining and prior to combustion.227
At one point in the same preamble EPA conceded that for some high sulfur coals, the standard could only be achieved by combining scrubbing and coal washing:
Based on the public record and additional analyses performed, the Administrator concluded that a 90 percent reduction in potential SO2 emissions (30-day rolling average) has been adequately demonstrated for high-sulfur coals. This level can be achieved at the individual plant level even under the most demanding conditions through the application of flue gas desulfurization (FGD) systems together with sulfur reductions achieved by currently practiced coal preparation techniques. Reductions achieved in the fly ash and bottom ash are also applicable.228
But later in the same preamble EPA states:
SO2 Control Technology — The final SO2 standards are based on the performance of a properly designed, installed, operated and maintained FGD system. Although the standards are based on lime and limestone FGD systems, other commercially available FGD systems (e. g., Wellman-Lord, double alkali and magnesium oxide) are also capable of achieving the final standard.229
The preamble also contains a discussion of the potential of physical coal cleaning. EPA stated that while it did not consider coal washing alone to be adequately demonstrated to achieve the standard, coal washing offered the following benefits when used in conjunction with FGD:
(1) the SO2 concentrations entering the FGD system are lower and less variable than would occur without coal cleaning, (2) percent removal credit is allowed toward complying with the SO2 standard percent removal requirement, and (3) the SO2 emission limit can be achieved when using coal having a sulfur content above that which would be needed when coal cleaning is not practiced.230
We believe that these passages, when read together, can be construed reasonably to say that the standard did not envision achievement of the standard by scrubbing alone. Rather, at the time of promulgation EPA contemplated that in some cases where high sulfur coal would be burned it would be necessary to use pretreatment technologies as well. Thus, we cannot agree with the Electric Utilities that EPA changed the basis for the standard seven months after the final rule on denial of the petition for reconsideration.
2. Notice That the Basis of the Standard Had Changed Since Proposal
The Electric Utilities also argue that the switch from a proposed standard based on FGD with optional credit for fuel pretreatment, to a final rule that is achievable only through the combined use of FGD and fuel pretreatment is a "major change" that, under the Act, requires an expanded and forthright explanation. Nonetheless we do not think that EPA's ultimate dependence on fuel pretreatment to demonstrate the achievability of the standard in the "most demanding situations" came about without notice or that it actually prejudiced the Electric Utilities in any way that would require us to remand the standard for a new round of public comment.
From the beginning of the rulemaking EPA specified that credits for coal washing could be taken toward meeting the percentage reduction requirement.231 The technical and support documents for the proposed standard gave substantial attention to coal washing.232 For example, the record includes a February 1978 Interagency Report prepared for EPA by the Batelle Memorial Institute entitled "Physical Coal Cleaning for Utility Boiler SO2 Emission." This report focused on the potential of PCC in implementing the proposed NSPS and specifically examined the combined use of FGD and PCC. The study concluded that the combination of coal washing plus FGD would be useful in meeting stricter standards (e. g., 0.4 lbs./MBtu ceiling as compared to 1.2 lbs./MBtu standard ultimately adopted) because "large quantities of high-sulfur coals cannot be cleaned to this level with FGD alone."233 In fact, the Batelle Report concluded that "PCC may be useful in combination with other controls meeting this type of standard. PCC would allow the scrubber or other control system to operate at a lower efficiency since credit would be given for sulfur removal."234
EPA received several comments on the subject of sulfur removal by fuel pretreatment and on the combined use of coal washing and FGD.235 NRDC and EDF's joint comments noted that there were numerous advantages to using FGD after coal washing such as the reduction of coal variability and sludge, and the lowering of shipping and maintenance costs. In addition, NRDC specifically recommended a standard based on the combined scrubbing-washing technologies.236 The Electric Utilities submitted comments that supported EPA's position that coal washing alone could not be the basis for the NSPS but also objected to NRDC's recommendation that the NSPS be based on both coal washing and FGD.237
EPA targeted the importance of coal washing on at least two other occasions. With the agency's December 1978 announcement of the phase two modeling results EPA stated that it had reassessed its modeling assumptions and had added coal washing to its analysis because "the coal washing credit ... was found to have a significant effect" on the modeling results.238 On April 5, 1979 EPA held a meeting with the representatives of the coal industry, the Electric Utilities, and the environmental groups to discuss the state of the art of coal washing.239 Materials distributed at the meeting concerning coal washing were put into the record shortly thereafter.240
Post-comment period submissions by the Electric Utilities,241 NCA,242 and EDF243 strongly suggest that the parties understood that the final standard was to be based on both FGD and fuel pretreatment. For example, the Electric Utilities' March 2, 1979, comment criticized the "coal washing assumptions being relied upon to support the new more stringent standards purportedly being recommended" and stated that "the basis for the change [to a more stringent emissions ceiling and 90 percent reduction level] also appears to be the assumption that coal washing is universally applicable on high sulfur coals and that this will help resolve the additional difficulties inherent in scrubbing high sulfur coal."244 Electric Utilities voiced its concern about EPA's conclusion that "coal washing is an adequate basis on which to justify raising the maximum performance requirement from 85 percent monthly removal ... to 90 percent monthly removal."245 Again on April 23, the Electric Utilities criticized EPA's consideration of NSPS including a 90 percent requirement, because "the assumptions concerning demonstrated scrubber removal efficiency and coal washing underlying the staff proposal are not supported by the record...."246 Then on May 18, the Electric Utilities reiterated its opposition to EPA's "assumption that scrubbers can routinely achieve 90 percent removal in high sulfur coal either alone or in combination with coal washing."247 Having made this series of statements only a few months before promulgation of the standard the Electric Utilities cannot argue convincingly that they were unaware that the final rule might require FGD to be used with coal washing for high sulfur coals.248
On this record, if EPA had plainly stated that the basis of the final rule had changed since the time of its proposal, there would be no question that the change would have been permissible. Indeed, section 307 of the Act by its very terms appears to anticipate that changes will take place during the course of the rulemaking.249 Despite the lack of a formal announcement, we find that the public had actual notice and an adequate explanation of the changes in the basis of the final rule. We believe that the parties were aware of the potential significance of coal washing to the justification for the final standard and that they were not prejudiced by a lack of opportunity to contest the agency's conclusions about the capability of coal washing.
B. The Achievability of the 90 Percent Standard
EPA's rationale for the achievability of the 90 percent standard is based on three determinations: First, that the long term median250 sulfur dioxide removal of 92 percent is achievable by the use of FGD alone. Second, that the variability in the performance of a well operated FGD system will not exceed 0.36. EPA calculates that when the 92 percent median is adjusted to account for the 0.36 FGD variability (as well as a so-called "autocorrelation factor")251 FGD systems can achieve between 86 and 89 percent average efficiency. Third, EPA concluded that utilities with FGD systems operating at these levels would be able to comply with the 90 percent removal requirement by employing other technologies for sulfur removal such as coal washing and ash retention.252
The Electric Utilities challenge each element of EPA's rationale for the 90 percent standard. The evidence on both sides is extraordinarily technical and often confusing. We have studied the record and the briefs with an eye toward judging whether, given the agency's expertise in evaluating conflicting data and selecting among reasonable approaches for pursuing statutory goals, EPA has plotted a reasonable course through the evidentiary thicket and stated a logical rationale for the route it chose.
1. The Support For EPA's Conclusions About FGD Performance
We begin with EPA's finding that a 92 percent long term median reduction is necessary, along with a performance variability range of no greater than 0.36 standard deviation (at the 95th percent confidence level) to meet the 90 percent standard. The Electric Utilities are correct, and EPA concedes, that there is no data on the record showing that the 92 percent long term median figure is actually achieved on a continuous basis by any currently operating commercial lime or limestone system. However, EPA projects the 92 percent figure in part from: (1) data obtained from a very limited number of lime FGD systems that do perform for varying numbers of 24 hour periods (from 22 to 42) at medians of 88 and 89 percent, (2) other short term data from nonlime/limestone FGD systems that report efficiencies in excess of 90 percent, and (3) reference to "Japanese experience [which] shows that technology exists to obtain greater than 90 percent SO2 removal."253 Finally, EPA cites as a basis for its 90 percent standard the prediction that future plants can be designed and operated to improve on efficiency levels currently experienced throughout the industry.254
The Electric Utilities say, on the other hand, that short term performances of less than 90 percent cannot be projected into a long term 92 percent median. Further, the short term data on performances of more than 90 percent is irrelevant because it is not taken from lime/limestone scrubbers cleaning high sulfur coal exhaust. They discredit the Japanese data because the conditions under which it was conducted are undocumented, and are indicative only of short term scrubbing of low to medium, rather than high sulfur coal. Further, the Electric Utilities profess no knowledge of the improved design features or operational practices that EPA relies on to predict future scrubbing achievability, and argue alternatively that many of EPA's "improvements" have been already rejected by the industry as impractical or ineffective.
Even if the 92 percent figure were achievable, the Electric Utilities contend that EPA's 0.36 variability range is not representative since it is based on data from only one plant. In addition, they are troubled by the fact that EPA did not include the autocorrelation factor in its statistical analysis until the Electric Utilities, in their petition for reconsideration, submitted their own analysis demonstrating the relevance of autocorrelation when determining whether a long term median will achieve a particular given level of control.255
EPA responds that its most recent statistical analysis of scrubbing efficiency "is in close agreement with [the Electric Utilities'] analysis when the same process variation and amount of autocorrelation was assumed."256 Although EPA's analysis indicates that only 88 to 89 percent control (in a thirty day rolling average for a base loaded plant) and 86 to 87 percent control (in a thirty day rolling average for a peak loaded plant) are achievable, improvements in process variability can be expected to yield more than 89 percent minimum reduction by the use of FGD alone.257 These levels of performance, EPA argues, exceed the minimum FGD performance (85 percent) necessary to achieve total reduction of 90 percent when other precombustion technologies are used.
(a) The Achievability of 92 Percent Long Term Removal Efficiency
EPA states that "the most important measure of an FGD system ... is its median removal efficiency."258 EPA determined originally that a 92 percent long term median removal was necessary to comply with the proposed 85 percent uniform standard measured on a 24 hour basis. At the time the final rule was promulgated it was not clear whether the 92 percent median performance was necessary to achieve the 90 percent standard based on the performance of both FGD and fuel pretreatment. However, in denying the Electric Utilities' petition for reconsideration EPA explained that the 92 percent median removal by FGD was essential when combined with precombustion sulfur removal to achieve an overall 90 percent level of performance.259
The 92 percent figure is derived from data on existing plants and on EPA's projection regarding the capabilities of well designed and operated scrubbers in future plants. As EPA correctly points out, the NSPS apply only to new plants, and since most current FGD systems were not designed to operate at 90 percent removal efficiency, the NSPS should not be based solely on existing levels of removal efficiency. Nevertheless, to determine whether it was reasonable for EPA, even with expected improvements, to project a 92 percent median, we must first look at the agency's test results at currently operating plants.
EPA's test data show that long term removal efficiencies exceeding 92 percent occur on low sulfur coals both in this country and in Japan.260 Of course, this alone is insufficient to support the standard since the 90 percent standard applies to plants burning high sulfur coals and it is undisputed that high sulfur coals pose the greatest challenge for FGD processes.
EPA's data also show that FGD systems using scrubbing reagents that are more reactive than lime or limestone were achieving very high sulfur removal efficiencies on a short term basis. For example, the magnesium oxide FGD at Philadelphia Electric's Eddystone Station achieved a median sulfur dioxide removal of 96.8 percent in an 8 day test.261 The sodium based double-alkali FGD at Gulf Power Company's Scholz Station achieved as high as 99 percent short term removal (we cannot determine from the record what the median performance was).262 Tests at the Wellman-Lord FGD unit at Northern Indiana Public Service Company's Mitchell Station including a 41 day continuous period of operation, demonstrated 89.2 percent median removal for a total of fifty-six 24 hour periods.263
The Electric Utilities argue and we are inclined to agree that this data is not conclusive since EPA specified that the standards were based on lime or limestone systems, and not the more expensive and less available regenerative systems, or systems using reagents and additives more reactive than lime. Both the proposed and the final standards, and the background support documents conclude that the standard is achievable by many types of FGD, including lime/limestone systems, Wellman-Lord, magnesium oxide, and double-alkali processes. However, EPA's emphasis on the lime and limestone processes and the uncontested fact that the non-lime/limestone processes are not widely available, preclude EPA from demonstrating the achievability of the standard on the basis of data from non-lime and limestone scrubbers.
The data advanced by EPA from lime/limestone systems are sparse, primarily because few plants had any need to attempt to achieve removal efficiencies over 90 percent on high sulfur coal under the former standard.264 A lime system at Columbus and Southern Ohio Electric Company's Conesville No. 5 plant burning high sulfur coal (4.5 to 4.9 percent) and designed for 90 percent removal attained a median of 88.8 percent.265 The data underlying that median, however, represents only 15 percent of a six month test period; the rest of the time the Conesville unit was malfunctioning.266 The only other data is from a pilot plant utilizing a lime system burning high sulfur (3.0 percent) coal at the Tennessee Valley Authority's Shawnee Station. The Shawnee Station achieved an 88.6 median but the unit is only a 10 MW prototype plant.267
EPA suggests that a 92 percent long term median removal may be interpolated between the lower 88.6 to 88.8 percent medians of the two lime scrubber units and the higher removal efficiencies of the non-lime Japanese and American units burning low sulfur coal. Counsel for EPA states that 92 percent represents "an intermediate level among the [FGD] data."268 The Electric Utilities cast this reasoning in the worst, but not entirely unreasonable, light:
EPA counsel's logic is truly remarkable: since one lime/limestone system can get 89 percent for an average of 24 out of 162 days and one regenerable system can get 96 percent for an average of 8 out of 22 days, then all future systems (lime/limestone and non lime/limestone) can get 92 percent long term mean removals. On Sunday, September 21, George Brett of the Kansas City Royals had a .400 batting average. On that day, Eddie Murray of the Baltimore Orioles was batting .302. Given these facts, EPA counsel's logic would lead to the conclusion that, in the future, all baseball players can be reasonably expected to bat .350.269
(Emphasis in original, footnotes omitted.) We agree that splitting the difference between data for lime scrubbers treating high sulfur coal exhaust and the data for non-lime/limestone systems and systems treating low sulfur coal is an unacceptable method for demonstrating that a 92 percent median is achievable by a lime system burning high sulfur coal.
Thus we cannot accept EPA's 92 percent median solely on the basis of evidence that only one commercial scale plant and one small pilot unit can almost but not quite meet the standard. To uphold the standard we must rely on EPA's predictions that a 92 percent median can be achieved with some design and operational improvements in new scrubbers. Two reports in the record and the Supplemental Background Document for the NSPS elaborate in great detail the kind of changes in future scrubbers that could be expected to increase their performance.270 We have reviewed the materials and they appear to explain how the changes would improve FGD performance. The predictions in these documents are also supported to some extent by test results where use of a recommended improvement such as the addition of reactive agents to the lime slurry, actually increased removal efficiency above 92 percent over very short periods of time.271 In addition, we find it informative that the vendors of FGD equipment corroborate the achievability of the standard although their support for the standard, taken alone, would not be decisive.272 A survey of 12 vendors showed that nine vendors routinely guaranteed 90 percent removal, and five of these sometimes guaranteed 95 percent removal for short term "acceptance testing."273 EPA also indicates that several of the successful FGD systems in Japan which the Electric Utilities claim are inapplicable and unavailable in the United States are provided to the Japanese by American vendors.274 Finally, the comments of the Industrial Gas Cleaning Institute (IGCI) which is a trade association representing the manufacturers of over 80 percent of American industrial air pollution equipment stated that:
The control equipment industry has the present capacity to design, manufacture, and install ... FGD control systems which will meet the emission level requirements of the proposed [NSPS] [which required a 92 percent median]....275
Recognizing that the Clean Air Act is a technology-forcing statute,276 we believe EPA does have authority to hold the industry to a standard of improved design and operational advances, so long as there is substantial evidence that such improvements are feasible and will produce the improved performance necessary to meet the standard. We accept EPA's documentation on the potential for improved performance of scrubbers to achieve a long term median of 92 percent. As a result, we uphold EPA's judgment that the standard can be set at a level that is higher than has been actually demonstrated over the long term by currently operating lime scrubbers at plants burning high sulfur coal.
(b) The Reasonableness Of EPA's Assumption About FGD Variability
Having determined that a median removal efficiency of 92 percent is achievable, the next question is whether EPA has properly accounted for performance variability. Variability or geometric standard deviation is a term used to describe the statistical distribution of levels of performance around the median. The relevance of variability in determining the achievability of a final standard can be seen from a simplified example. A system with 92 percent median efficiency and low variability might always achieve between 90 and 94 percent removal and thus always comply with the 90 percent standard. A system with a 92 percent median efficiency and high variability might achieve between 85 and 99 percent removal, and therefore it would not always comply with a 90 percent standard.
EPA based its variability assumption on its conclusion that the maximum variability of 0.36 percent at the Cane Run plant represents the outer boundaries of FGD variability. EPA explains that "variability in the performance of a well-operated FGD system will not exceed the maximum variability of the most variable well-run lime or limestone plant that has been tested [Cane Run]."277 EPA states that Cane Run represents a "worst case assumption" in five different respects:
it is based on lime/limestone FGD, which is more variable than regenerable and sodium-based FGD; it is based on the most variable properly operated lime/limestone FGD tested; it is based on a peaking unit, which produces greater variability in flue gas volume than a base-load unit; it is based on maximum variability to be expected, rather than the average variability, at Cane Run; and it does not take account of techniques (automatic process controls and coal blending) that can reduce variability. The allowance for variability insures that the standard is achievable on a statistically consistent basis.278
Initially we were skeptical about the validity of assuming that the variability sample of one plant was representative of the entire industry. Upon review we are satisfied that EPA's case justifying its variability assumption is a reasonable one. The agency collected variability data from many plants including the seven plants in the agency's data base for FGD median efficiency.279 The data from three of these plants was defined as "representative" for standard setting purposes. Of these three, Cane Run was the most variable; the rest were rejected for valid reasons, such as, the plants burned low sulfur coal, they used non-lime/limestone scrubbers, or they experienced scrubber malfunctions and thus were not representative of well operated lime FGD systems.280 For this reason we are not persuaded by the Electric Utilities' argument that the aggregate data on variability, taken from all high sulfur burning plants in EPA's data base, showed a higher variability (about 0.42-0.43) than Cane Run (0.36).281 In addition a consultant's report on which EPA relied showed that variability could be reduced in future systems which incorporated specific design and operational improvements.282
The Electric Utilities also argued that the variability experienced at the Cane Run plant is not representative of worst case variability because the variability is greater at the Bruce Mansfield plant, which is one of the other two plants EPA considers to be a well run lime/limestone FGD system. The Electric Utilities' analysis of the Bruce Mansfield data concluded that FGD variability at such a plant could be at least as high as 0.38.283 Originally, EPA had not included in its variability analysis the Bruce Mansfield data relied on by the Electric Utilities.284 EPA argues that this additional data is unreliable for a number of reasons and should not be included in the variability data base.285 However, we are satisfied with EPA's demonstration that even if the additional data are considered, they support a variability figure of 0.33 for the Bruce Mansfield plant.286 Therefore, the Bruce Mansfield data are consistent with the reasonableness of EPA's conclusion of worst case FGD variability of 0.36.
In sum, the record supports the achievability of the level of FGD performance which is necessary to comply with the 90 percent standard by a long term median efficiency of 92 percent with a maximum variability of 0.36.
2. The Support for EPA's Conclusion That the 90 Percent Standard Was Achievable by the Use of Coal Washing in Conjunction With Scrubbing
EPA has sufficiently established that a minimum 86 percent removal efficiency (on a thirty day rolling average) — derived from a 92 percent median with 0.36 variability — is achievable on a continuous basis by the use of scrubbers alone. We now consider whether coal preparation techniques can enable utilities to increase the overall removal efficiency to meet the 90 percent standard. EPA claims that when physical coal cleaning or coal washing (the most prevalent method of coal preparation) is used together with "pulverization" and/or "ash retention techniques," then FGD systems capable of only 85 percent removal can meet the 90 percent standard. The Electric Utilities argue, not without cause, that the record underlying EPA's conclusions about the supplemental sulfur removal potential of pulverization and ash retention is deficient.287 They also claim that EPA has no basis for expecting that the removal efficiencies achieved by these individual technologies can be added together when the technologies are combined, since some of the sulfur each system is designed to remove may have been eliminated already through the prior use of another sulfur reduction technology. However, we do not find it necessary to reach the merits of EPA's case for an additional 10 percent reduction based on pulverization and ash retention because we are satisfied on the record that coal washing used in conjunction with FGD will achieve the standard.
(a) Description of the Coal Washing Process
Coal washing is a widely practiced and relatively inexpensive method of removing sulfur from coal currently used on approximately one-half of the coal produced in the United States including about 40 percent of the coal destined for electric utility plants.288 Sulfur reduction under current commercial coal washing practices varies with several factors.
Coal washing separates unwanted sulfur from coal by suspending crushed coal in a fluid, then separating the dense sulfur particles, which sink to the bottom, from the less dense particles of cleaned coal, which rise to the top.289 Commercially available coal washing processes remove inorganic sulfur, e. g., high density pyrite crystals, but not organic sulfur which is chemically bonded to the coal. Three factors determine the maximum percentage reduction of sulfur: (1) the relative proportions of organic (non-washable) and inorganic pyritic (washable) sulfur, (2) the particle size to which the raw coal is crushed before it is suspended in the separating fluid, and (3) the specific gravity of the fluid in which the coal is immersed.290
The proportions of organic and pyritic sulfur vary widely depending on the coal. While coal washing can remove from 35 to 70 percent of pyritic sulfur, the total percentage of removable sulfur is limited by the amount of nonwashable organic sulfur in the coal, which ranges from 30 to 70 percent of total sulfur content.291 High sulfur coal has a high proportion of pyritic sulfur compared to low sulfur coal. For this reason, it is easier to wash a given percentage of sulfur out of high sulfur coal than low sulfur coal.
As a general rule, when coal is crushed into smaller particles, more sulfur can be washed out. However, the smaller the coal particles, the more coal that is inadvertently washed out with the sulfur, and the more energy that is consumed in processing the coal. Both of these kinds of energy losses are measured in terms of the percentage of Btu's in raw coal that are recovered after washing. For example, standard coal washing results in an energy loss of from 5 to 20 percent, which translates to a Btu recovery ranging from 95 to 80 percent. In addition to energy loss, handling and processing costs increase with the amount of crushing. These economic constraints limit the maximum amount of coal washing that is feasible in any given situation.
The specific gravity at which the separation is performed is also a critical factor. The lower the specific gravity the greater the amount of pyritic sulfur that is eliminated. However, as lower specific gravities are used more coal is also lost and thus fewer Btu's are recovered.292
Lowering the sulfur content of coal by a given percentage through coal washing reduces the potential sulfur dioxide emitted from burning the coal by the same percentage since the sulfur in coal is converted into sulfur dioxide in a fixed proportion. The percentage of sulfur reduction achieved by a coal washing in turn reduces the level of efficiency of a scrubber required to achieve the overall 90 percent reduction standard. But there is significantly less than a percent for percent trade-off between coal washing and necessary scrubbing. Every percent removed by coal washing reduces the needed efficiency of FGD by much less than one percent. A brief example will illustrate the point: given a coal with uncontrolled emissions of 10 lbs./MBtu, the final standard would require a 90 percent overall reduction to 1.0 lbs./MBtu. If coal washing reduced potential emissions 25 percent to 7.5 lbs./MBtu, then the scrubbers would have to eliminate 6.5 lbs./MBtu to limit emissions to 1 lb./MBtu. This means that the scrubbers must operate at 87 percent reduction efficiency.293
(b) The Percentage Reduction Achievable By Washing High Sulfur Coal
EPA asserts that it is conservative to assume that a 27 percent reduction in sulfur content is achievable through washing high sulfur coals (greater than 6.0 lbs./MBtu). If the agency is correct, then a 90 percent total reduction can be achieved even in the "worst case" scenario involving high sulfur coals with a scrubber operating at 86-87 percent efficiency. Coals with sulfur content less than 6.0 lbs./MBtu are not subject to the 90 percent standard, thus a 27 percent reduction by coal washing is unnecessary on those coals. The Electric Utilities do not agree that 27 percent removal has been demonstrated for washing these high sulfur coals, or that the 90 percent standard can be achieved on them by a combination of coal washing and scrubbing.
Our inspection of the record reveals a wealth of technical data on the washability of every type and variety of coal in the United States. A Bureau of Mines ("BOM") Report, prepared in 1976 and included in the record of this rulemaking, examined the percentage of sulfur removal by different levels of washing in 455 raw coal channel samples in different regions, states, and coal beds.294 This study concluded that the national average washability was 27 percent with crushing to 1½ inch top size, specific gravity at 1.6 and the average sulfur content of all the coal sampled at 4.9 lbs./MBtu.295 The main regions producing high sulfur coal — Eastern Midwest and Northern Appalachia — averaged washability of 30 percent and the Western Midwest averaged just under 30 percent with crushing to 1½ inch top size and specific gravity at 1.6.296 The washability of specific coal could exceed the average if its sulfur content exceeded the average, if there was greater size reduction, or a higher specific gravity was employed. EPA concluded initially from the BOM data that a 35 percent potential sulfur reduction was reasonable for coals of 2.5 lbs./MBtu or more, but later assumed this level of reduction was applicable for coals of 5 lbs./MBtu or more,297 including, of course, the 6 lbs./MBtu sulfur coals for which washing is essential to meet the standard.
In the final rule, however, EPA appeared to back off from the 35 percent figure to the more conservative 27 percent298 which was, in effect, endorsed by the NCA and the Electric Utilities, who had objected that the 35 percent figure was "optimistic by 5 to 10 percentage points."299
In the denial of the Electric Utilities' petition for reconsideration EPA cited NCA's data and said:
On high-sulfur midwestern coals that would be subject to the 90 percent reduction requirement an average of 27 percent sulfur removal was achieved by conventional coal washing plants in 1978.... These data represent current industry practice and do not necessarily represent full application of state-of-the-art in coal cleaning technology.300
The Electric Utilities' own consultant wrote that "[t]he available docket information basically supports the 27 percent average coal washing credit cited by EPA."301
There is little doubt that the record supports an average 27 percent coal washing reduction potential on high sulfur coal. The remaining question is whether such an "average" figure can be relied upon to support the requirement of section 111 that the "achievability" of a standard must be "demonstrated." The Electric Utilities argue that the 27 percent does not take account of variability so that washing on the low ends of the "average" would not be sufficient to meet the standard. Both NCA's limited data and the Bureau of Mines' seam by seam data indicate deviations from the "average" sulfur removal attainable by washing the coal sampled. The data, however, also suggest to us that fluctuations in the removal potential of washing are closely tied to the varying characteristics of the coal itself (primarily its sulfur content) and not the reliability of the washing process.302 Thus, instances where coal washing might achieve below 27 percent would seem to involve treatment of coal whose sulfur content is considerably less than 6.0 lbs./MBtu and therefore not subject to the 90 percent reduction standard.
In any event, the best answer to the variability argument seems to lie in the mechanics of the compliance provisions of the rule that set out the methods of taking coal washing credits. A standard may be based on reliable data about the "average" removal from coal washing if the utilities who must comply with the standard are permitted the same leeway in "averaging" for compliance purposes.303 We believe that is so in this case for the following reasons. In order to take credit for coal washing the sulfur content of the raw coal and its sulfur content after washing must be determined in advance of feeding the coal into the boiler. Under EPA's regulations coal supplies can give utilities a certificate of credits with each lot of washed coal fuel delivered [or for all lots delivered for a calendar quarter].304 The certificates have to show (1) sulfur analysis of a coal input and output for the washing plant, (2) the quantity delivered, (3) the heat content, and (4) the calculation of pretreatment credit.305 After the utility plant receives the fuel and the pretreatment certificate, a summary of credits for all fuel received in a calendar quarter is prepared and averaged to obtain a quarterly pretreatment credit. This average credit is applied to 30 day periods in the following quarter. That is, for a utility to determine its present compliance level it combines the "average" coal washing credit from the previous quarter with the FGD control level in each 30 day measuring period in the current quarter.306 If an average of 27 percent removal by coal washing is achieved on a quarterly basis then the utility is credited with 27 percent toward compliance in the next quarter whether or not washing of certain lots varied above or below the 27 percent level. This allows a utility to monitor its coal supply and affords flexibility in meeting the standard. If incoming coal supplies threaten to increase the quarterly average to unacceptable levels, then buying for the rest of the quarter can be adjusted. EPA expected that utilities would enter into contract arrangements with their suppliers to obtain and guarantee that coal supplies meet the needed treatment criteria.307
If for some reason a utility obtains coal that has not been washed to the desired level the result is not an automatic violation of the standard, as the Electric Utilities imply. Rather, utilities have the option of: (1) not burning the coal, (2) enforcing their supply contracts, or (3) burning the coal realizing that when averaged with other quarterly supplies the credit needed to comply will be obtained.308
Thus the variability of individual lots of washed coal does not assume so critical a role as variability does in FGD systems which are under continuous monitoring. Since there is no doubt that a 27 percent "average" is achievable with the highest sulfur coals that must meet the 90 percent reduction, we believe that the standard can be upheld on the basis of the coal washing data in the record.
The Electric Utilities' final argument is that the record contains no data on the actual combined use of coal washing and FGD, and that EPA has not demonstrated that the standard is achievable simply by hypothesizing that coal washing and FGD can work together.309
We do not believe that, on the facts in this record, such an actual demonstration was necessary. The record is replete with documentation that lower sulfur coal is easier to scrub than higher sulfur coal, and the Electric Utilities have not cited nor have we found any evidence that washed coals create any special problems for scrubbers. The main difference between washed and unwashed coal of the same sulfur content is that the washed coal contains fewer impurities that may be expected to interfere with FGD performance. In fact, the record contains reports analyzing the potential reductions achievable by the combined technologies and evidence that washing can be expected to increase FGD performance.310 At an empirical level, since 40 percent of the coal now being used by utilities has been washed, the industry has certainly had the requisite experience to anticipate and document any potential problems.
For all of the foregoing reasons we find that EPA has adequately demonstrated the achievability of the 90 percent standard.
IV. THE STANDARD FOR EMISSION OF PARTICULATE MATTER
Most particulates emitted from electric utility sources come from coal-fired units.311 In 1976, for example, particulate emissions from coal-fired electric utilities accounted for 3.42 million out of 3.57 million tons coming from all electric utilities and 14.9 million tons from sources of all kinds.312 The former NSPS limited emissions of particulate matter to 0.10 lbs./MBtu.313 The revised standard limits emissions to 0.03 lbs./MBtu, and is based on the performance of either a well designed, operated and maintained electrostatic precipitator ("ESP") or a baghouse control system.314 The Electric Utilities, the only party to challenge the particulate standard, argue that the record does not support the conclusion that the promulgated standard is achievable by either ESP or baghouse technology.
A. Technical Background
1. ESP Control Technology
Since ESP's were introduced to the utility industry in the 1920's, they have become the most widely used means of controlling particulate emissions from coal-fired boilers.315 An ESP removes particles from a gas exhaust stream by electrically charging the particles and then attracting the charged particles to metal plates having the opposite charge.316 The size of the collecting plates and thus the size of the ESP are directly related to the capability of the ESP to control emissions. The larger the surface area of the ESP collecting plates the greater the ability of the ESP to remove particles from flue gas, assuming all other factors that affect ESP performance are held constant. Not surprisingly, installation and operating costs also increase with the area of the collecting plates. On the other hand, the volume of particulate-laden gas that needs to be treated by an ESP affects ESP effectiveness inversely. As the gas volume increases, the effectiveness of a given ESP decreases, again assuming all other factors are held constant. The ratio of the size of the collecting plates to the volume of gas treated, is known as the specific collecting area ("SCA"). SCA is measured in terms of square feet of plates per 1000 actual cubic feet per minute of gas ("Ft2/1000ACFM"). The greater the SCA the more effective the ESP. Increasing the area of the collecting plates increases the SCA and ESP effectiveness; in contrast, increasing gas volume decreases the SCA and ESP effectiveness.317
Two other important determinants of ESP effectiveness and cost are the characteristics of the fly ash produced by the burned coal and the temperature of the gas exhaust stream.318 The ash produced by the combustion of some low sulfur coals is highly resistant to electrical charging, which creates a problem for ESP collection. Generally, a larger SCA is needed for highly resistant ash than would be needed for cleaning the ash of most coals "whose resistivity is below the level where problems occur."319 For these high resistance coals, ash resistivity is at a maximum in the temperature range experienced after the flue gas exhausts from the air combustion preheater (downstream or "cold side" of the preheater). Conversely, the ash resistivity is much lower at the "hot side" (upstream) of the preheater.320 Consequently, high resistivity coal ash is easier to collect on the hot side than on the cold side of the preheater. However, this does not always mean that a smaller ESP can be applied on the hot side than on the cold side because gas volume is greater on the hot side. The advantages of hot side ESP are offset also by greater construction costs due to higher quality of materials, thicker insulation, and special design features to accommodate the expansion and warping potential of the collection plates.321 A typical ESP is shown as Figure 19 in the appendix to this opinion.
2. Baghouse Control Technology
Baghouse control of particulate emissions is an increasingly popular alternative to ESP's among electric utilities, largely because baghouses are not affected by particulate resistivity to electrical charge, and thus are less expensive to operate for power plants burning low sulfur coals with high resistance ash.322
In a baghouse system, particulate-laden gas is passed through a fabric filter so that particulates in the gas are retained on the upstream or dirty-gas side of the fabric. In a baghouse system the dirty gases flow into the housing, upward through the bags, and then out of the clean gas outlet.323 Typical baghouse systems are shown in Figures 20 and 21 in the appendix to this opinion.
A baghouse consists of numerous tubular fabric bags (each about one foot in diameter and thirty feet long) arranged in compartments or cells parallel to each other, standing like a stack of porous straws in a box. The flue gas from the boiler is divided so that a small fraction of the gas passes through the individual bags. Large baghouse systems can be designed as one large cell or module or many individual modules. With several modules, one module can be taken out of service for maintenance without affecting the operation of the other modules. If enough modules or isolated components are provided, even the largest pulverized coal-fired steam generators can be kept on line at full load while necessary maintenance is performed.324
Two determinants of baghouse performance are the "air-to-cloth" ratio and the "pressure drop."325 For a given flue gas stream, a greater fabric surface (larger baghouse) yields a lower air-to-cloth ratio. The pressure drop represents the pressure (measured in inches of water ("In.H2O")) required to force the flue gas through the fabric bags. For a given quantity of flue gas, the pressure drop increases as the fabric area decreases, since greater force is needed to push the flue gas through a smaller filter surface. Accordingly, at any boiler operating at full load, a larger air-to-cloth ratio (smaller baghouse) tends to be associated with a higher pressure drop, while a lower air-to-cloth ratio (larger baghouse) tends to be associated with a lower pressure drop.
There is a maximum pressure drop that can be achieved by a given boiler operating at full capacity. If the baghouse is undersized, that is, if the air-to-cloth ratio is too high, then the boiler will not be able to achieve full load. Thus, unlike ESP systems, where undersizing does not affect the capacity of the steam generator, but does affect the efficiency of the emission control system, the penalty for undersizing a baghouse is loss of boiler capacity.326
B. The Evolution of the Particulate Standard
In 1975, EPA began to investigate whether the 0.10 lbs./MBtu standard for particulates should be revised because the existing techniques for control of particulate emission appeared to be more effective than when the standard was originally promulgated in 1971.327 By 1976, EPA's research determined that there were, in fact, ESP systems installed at coal-fired power plants which were controlling emissions substantially below the 1971 standard. At this point EPA sought to acquire additional data on the achievability of a stricter standard. EPA attempted to locate plants which generated high resistivity ash and were equipped with ESP's with large SCA's.328 With the assistance of the Industrial Environmental Research Institute ("IERL") and the Industrial Gas Cleaning Institute ("IGCI"), EPA located six sources meeting these criteria. All of these sources were surveyed by plant visits: EPA found that "adequate emission test data were available" for three of the sources, two of the other three sources were tested by EPA, and the sixth source was tested by the utility company with an EPA observer present.329 Additional emission test data were gathered from IERL and State agencies to form a data base from tests of twenty-one different ESP systems.330 On the basis of this data base EPA concluded that ESP's are capable of achieving a limit of 0.03 lbs./MBtu even when the most difficult to control fly ash is generated. In the preamble to the final rule, EPA stated its judgment that utilities firing low sulfur coal (less than 1.0 percent) must have an ESP with a hot side SCA of 650 Ft2/1000ACFM or a cold side SCA of 1000Ft2/1000ACFM to achieve the standard. EPA suggested that firing medium sulfur coals (between 1.0 and 1.9 percent) would require a hot side SCA in excess of 452 Ft2/1000ACFM or a cold side SCA in excess of 435 Ft2/1000ACFM, while firing high sulfur coal would require a cold side SCA of at least 360 Ft2/1000ACFM. No hot side SCA figure was recommended for utilities firing high sulfur coal. EPA concluded that all of these SCA's were reasonable considering cost, energy, and nonair environmental impacts.331
In January 1977, while EPA was reassessing the capability of ESP control, the agency found that baghouse technology was developing well enough to warrant considering baghouses as an alternative to ESP's for even the largest utilities.332 EPA made a survey of sixteen baghouse installations333 and from this survey EPA concluded that the actual effectiveness of baghouses could prove equal, if not superior to ESP systems. Although the baghouses surveyed were installed at small plants, the agency found that the systems were composed of numerous individual cells and that even the largest steam generator could use baghouse control by increasing the number of cells as needed to handle the gas flow. Two plants were visited and subsequently tested by EPA in the spring and summer of 1977. This data base was supplemented by test results provided by IERL and by test data from the State of West Virginia to form a data base of tests from eight different baghouse systems. On the basis of this data base EPA concluded that baghouses with an air-to-cloth ratio of 2ACFM/Ft2 will achieve the standard at a pressure drop of less than 5 In.H2O. These specifications were considered reasonable in light of cost, energy, and nonair environmental factors.334
Although finding that either technology could be employed at a reasonable cost to achieve the standard, EPA predicted that utilities would rely on ESP's for high sulfur coal applications and baghouses for low sulfur coal applications because of cost advantages.335
C. The Achievability of the Standard
In order for EPA to demonstrate the achievability of the standard for particulate matter it must: (1) identify variable conditions that might contribute to the amount of expected emissions, and (2) establish that the test data relied on by the agency are representative of potential industry-wide performance, given the range of variables that affect the achievability of the standard. National Lime Association v. EPA.336
Citing these principles, the Electric Utilities argue that EPA's data base does not show that the particulate matter standard is achievable on a continuous basis, for either ESP or baghouse control. Initially, the Electric Utilities fault EPA's analysis of ESP and baghouse performance because it did not take into account whether the standard is achievable by utilities firing lignite coal, which they contend is a relevant variable. EPA counters that it did consider the achievability of the standard with lignite, and that it concluded that lignite was not a relevant factor because it was no more difficult to control lignite emissions than the emissions from other high resistivity coals. Most of the Electric Utilities' other objections are to the representativeness of the test data for both ESP's and baghouses. They say that the ESP and baghouse test results are incomplete, that the samples are too small and that the data reflect only short term performance while the standard requires long run continuous compliance. In addition, they claim that EPA has failed to reveal its methods of data collection and the conditions present during the tests. Finally, the Electric Utilities contend that the baghouse data are not representative because all of the data available at the time of promulgation of the standard were from small scale baghouse operations whereas the data obtained after promulgation from commercial size installations actually refute EPA's conclusions. These contentions require that we evaluate EPA's data in greater detail. As discussed below we uphold the particulate standard despite concern about EPA's ESP data because we find that the agency's baghouse data alone establish the achievability of the standard.
1. EPA's ESP Data
EPA satisfied its burden under the first prong of the National Lime test by describing the variable conditions that could be expected to affect ESP performance (e. g., coal ash characteristics, ESP size, gas volume and flow, maintenance practices, etc.).337 However, EPA has not met its burden for establishing that its ESP data are representative of industry-wide performance under the second prong of the National Lime test for achievability.338 We reach this conclusion even if EPA is correct that it devoted enough attention to another potential variable — the impact of firing lignite coals on the achievability of the standard.339
EPA collected emission data from twenty-one ESP-equipped steam generating units firing relatively low sulfur coal (0.4 to 1.9 percent). EPA explained that it "evaluated emission levels from units burning relatively low-sulfur coal because it is more difficult for an ESP to collect particulate matter emissions generated by the combustion of low-sulfur coal than high-sulfur coal."340 But the agency's own background document identifying the relevant performance factors does not list the sulfur content of coal as one of the variables and instead focuses on the resistivity of coal ash.341 While some low sulfur coals produce high resistivity ash, "sulfur content of the fuel alone is an extremely poor predictor of the resistivity for low sulfur fuels."342 Although EPA identified resistivity as an important variable, its reported data do not indicate the resistivity of the coal fired at the test units. In contrast, two nonagency reports on the record concerning ESP performance include resistivity data.343
The ESP test results are listed in a table in the Background Information Document for the Proposed Standard which is shown as Figure 22 in the appendix to this opinion.344 EPA reported that nine of the twenty-one units tested performed better than the 0.03 lbs./MBtu standard even though none of the twenty-one ESP units were designed to achieve this level of performance.345 Of the twelve units that exceeded the 0.03 lbs./MBtu level none experienced emissions greater than 0.05 lbs./MBtu.346 The former standard had been 0.10 lbs./MBtu or twice the worst test performance. Superficially, these performances appear to support the conclusion that ESP's can be designed to achieve the 0.03 lbs./MBtu standard. However, we have serious reservations about the completeness of the reported data for the twenty-one ESP-equipped units.
The table in which EPA's ESP data is compiled lists complete data for only twelve of the twenty-one units.347 Of this group only two of the units with complete data were reported as achieving the standard and only one of these two was firing low sulfur coal (less than 1.0 percent).348 The data that is missing on the table includes the SCA measurement and whether the ESP is hot side or cold side. EPA's background documents identify these two factors as critical.
In another table, shown as Figure 23 in the appendix to this opinion, EPA lists a summary of data on "difficult" ESP cases.349 This table reports data for seven ESP units firing coal less than 1.0 percent. The data of six of these units350 are repeated from the table of twenty-one units, but there were more than six units — eleven to be exact — in the previous table that fired less than 1.0 percent sulfur coal.351 EPA offers no explanation for this selectivity other than the statement that the seven difficult cases were selected based on the "reports of owners and ESP vendors that the low sulfur coals fired at the plant produced a difficult to collect ash."352 This statement calls attention to the fact that something more than just the sulfur content of the coal is relevant to the difficulty of ESP collection, and EPA has not shown that its data account for this factor — the resistivity of the ash. The seventh unit (unit 22) listed on the table of difficult cases was not included on the first table of units and it represents an additional unit that EPA does not appear to discuss or explain anywhere else on the record. While complete data for all seven of the so-called difficult cases are listed, only one of these cases limited particulate emissions below the 0.03 level.353 This one unit then is EPA's substantiation that the standard is achievable "even in the most difficult cases."354
We note that EPA's data are inadequate in other respects. First, the background document which contains the data reveals very little information about the duration of the tests and the conditions under which they were taken. We have been able to find some partial references in the joint appendix which disclose some of this information.355 We also are aware that the background document includes a citation to a test report for each unit tested.356 These reports are not included in the joint appendix, and no reference to the administrative docket is offered in most of EPA's citations. Presumably, some of the information that would be useful to our review is included in these documents. However, especially in a case of this magnitude and complexity, it is not reasonable for the agency to expect the court on its own to gather together all of the scattered pieces of information that are necessary to make a coherent whole.357
The Electric Utilities have also argued that the record does not reveal EPA's method of data collection. We believe, however, that throughout the rulemaking process EPA has given notice of the established test methods relied on; thus, the data base is not objectionable on this basis.358
Our concerns with the lack of documentation for EPA's data base are somewhat allayed by independent evidence on the record which tends to support the reasonableness of the agency's conclusions about ESP performance. High efficiency particulate removal has been practiced for decades on high and medium sulfur coal.359 Although EPA did not believe that utilities would use ESP's in high resistivity low sulfur coal applications, it found that ESP control of these cases was feasible so long as the ESP's were large enough — with a hot side SCA greater than 650 Ft2/1000 ACFM or a cold side SCA as large as 1000 Ft2/1000 ACFM. These figures are corroborated by an independent report prepared in 1974.360 Another report, prepared in 1978 by the Electric Power Research Institute ("EPRI") suggests that high efficiency removal by ESP's is feasible but questions whether it would be economical to drop much below the former NSPS of 0.10 lbs./MBtu.361 The EPRI report concluded that baghouses would become the economical choice for all coals with standards below the 0.10 lbs./MBtu level.362 Despite the presence of some evidence in the record that suggests that EPA's findings are reasonable, the agency's presentation of its own data on which it says it is relying does not provide the guarantee required by National Lime that the test results be representative of industry-wide performance.363
2. EPA's Baghouse Data
We can uphold the 0.03 lbs./MBtu standard for particulates because we are satisfied that there is substantial evidence on the record supporting the achievability of the standard by baghouse technology. The preamble to the final rule refers to this evidence, which includes extensive tests of small scale installations, limited data from one full scale commercial sized operation, in addition to industry practices that indicate that baghouse control technology is a viable method of complying with the standard.364 Evidence added to the record since promulgation and evaluated with the petitions for reconsideration confirms the reasonableness of EPA's conclusion that the standard is achievable with baghouses.
EPA identified, in accordance with National Lime, the variable conditions that affect baghouse performance (e. g., baghouse size, pressure drop, cleaning, maintenance, etc.).365 In accordance with the second prong of the National Lime test the agency accounted for these variables with the following data: EPA's data base from small scale baghouse operations was comprised of fifty emission test runs conducted at eight baghouse-equipped generating units. Although none of these baghouse units was designed to achieve the 0.03 lbs./MBtu level (again EPA does not report what level the baghouses were designed for), forty-eight of the test results achieved this level of emissions and only one test run at each of two units exceeded the 0.03 lbs./MBtu level. Neither of these two tests appears to have exceeded the 0.05 lbs./MBtu level. The data are summarized in a table found in the Background Information Document for the Proposed Standard and are shown as Figure 24 in the appendix to this opinion.366 Although complete data are not provided for all of the tested units, the omissions are not as critical as was the case with the agency's unsatisfactory ESP data and do not frustrate our ability to determine that the data supports the agency's stated rationale.
Complete data are listed for all four of the units tested by EPA methods.367 At the other four units,368 West Virginia test methods were used which EPA states are similar to its own test methods. In these cases data are not given for air-to-cloth ratios and pressure drops. However, all of the performances at the four EPA tested units, for which complete data are listed, except for one test run at Unit 8, achieved the 0.03 lbs./MBtu standard. At these installations, three of the four units that never exceeded the standard369 reported pressure drops less than or equal to 5 In. H2O (ranging from 2.1 to 5.0 In. H2O). The fourth unit (Unit 8) that failed to achieve the standard on one test listed pressure drops ranging from 8 to 10 In. H2O. Air-to-cloth ratios for all four of the EPA tested units ranged from 1.9 to 3.0 ACFM/Ft2. From this data it is possible to discern how EPA derived the baghouse specifications it judged were necessary to achieve the standard. We see that its judgment that an air-to-cloth ratio of 2 ACFM/Ft2 will achieve the standard at a pressure drop less than 5 in. H2O was, if anything, conservative.
All of the small scale units were less than 44 MW. Before announcing the final standard EPA was able to obtain test data from one new unit with 350 MW capacity.370 The baghouse control system for this facility was designed to achieve 0.01 lbs./MBtu, or three times less emissions than allowed by the standard.371 The baghouse system there had a pressure drop of 5 in. H2O and an air-to-cloth ratio of 3.32 ACFM/Ft2 which is undersize according to EPA's estimate that a 2.0 ACFM/Ft2 air-to-cloth ratio would be appropriate for a pressure drop of 5 in. H2O.372 In the preamble to the final rule EPA reported:
Although some operating problems have been encountered, the unit is being operated within its design emission limit and the level of the standard. During the testing the power plant operated in excess of 300 MW electrical output. Work is continuing on the control system to improve its performance.373
In addition to the operating experience from eight small scale facilities and one large scale facility, EPA noted that the trend in the utility industry reflected considerable confidence in full scale baghouse technology:
As of May 1979, at least 26 baghouse-equipped coal fired utility steam generators were operating, and an additional 28 utility units are planned to start operation by the end of 1982. About two-thirds of the 30 planned baghouse-controlled power generation systems will have an electrical output capacity greater than 150 MW....374
The Electric Utilities vigorously argue that EPA's data are not representative of the performance at full scale electric power plants and say that there are technological barriers to scaling baghouses up to commercial size facilities.375 These concerns are belied by the large numbers of utilities that are, in fact, moving to baghouse control systems. In any event, EPA has acknowledged since the standard was proposed that its performance data are based on small scale plants (which were the only size baghouse installation available at the time). We cannot fault EPA for failing to produce nonexisting data on baghouses at larger plants. We only can look carefully at the smaller plant data, as we have done, and at EPA's explanation for the representativeness of these data. EPA reasons that data obtained from small units were representative of the performance of large units because baghouses are designed and constructed as modules, and therefore a large unit basically involves more modules.376 This explanation is reasonable and borne out by the record. For example, the largest baghouse system in operation at the time of promulgation — a full size 350 MW power plant — had twenty-eight baghouse modules, each servicing 12.5 MW of generating capacity.377 While this new unit experienced some operation difficulties such as bag failure and high pressure drops,378 EPA explained that the failures were temporary and due to undersizing the baghouse by 40 percent from the recommended air-to-cloth ratio as well as to problems associated with working out bugs in a new system. EPA found that these problems were surmountable technological barriers to operating baghouses on a large scale.379 At another plant, four pulverized coal-fired boilers supplying a total of 112.5 MW have been retrofitted with baghouses.380 As the Electric Utilities state, this utility station employs four small baghouses serving less than 30 MW each.381 All of the baghouses in this commercial size station have consistently complied with the standard.382 At still another utility plant two units have been retrofitted with baghouses and are guaranteed by the manufacturer to achieve 0.03 lbs./MBtu, but test results on these plants are not available.383 This record demonstrates that EPA's test data are representative of full scale performance.
The Electric Utilities also object that the agency's baghouse data are unrepresentative because only two of the eight small scale units were utility boilers384 while the remaining six units were industrial boilers and most stoker-fired as opposed to pulverized coal-fired.385 But again EPA has offered a rational explanation for the representativeness of these data: "[e]ffective baghouses are designed on a multicell basis. The prime difference between an industrial and a utility baghouse is the number of cells. Consequently, data on the effectiveness and reliability of industrial boiler baghouse design is applicable to utility steam generators."386 The reasonableness of the explanation is verified by the large number of utilities adopting baghouses as control technology.
The Electric Utilities also rely on post-promulgation data acquired from two full size plants to refute EPA's conclusion that the 0.03 standard is achievable by baghouse control at reasonable cost.387 In their petition for reconsideration and on this appeal the Electric Utilities argue that the costs of operating baghouses at Southwestern Public Service's Harrington Station ("Harrington") and Texas Utility Generating Company's Monticello Station ("Monticello") are in excess of EPA's estimated costs for baghouse control. They contend that the bag replacement costs during the first year of operation, approximately $250,000 at the 350 MW Harrington unit and $321,000 for each of the two 305 MW Monticello units, surpassed EPA's estimates. To derive EPA's estimated figures for baghouse replacement it is necessary to refer to several different documents containing economic data which are scattered throughout the joint appendix.388 Then, because EPA's estimates are only listed for units of 200, 500, and 1000 MW, it is necessary to interpolate between the 200 and 500 MW cases to obtain the correct estimate for the Harrington 350 MW unit and Monticello 305 MW units.389 These calculations show that the estimated cost of bag replacement for a unit the size of Harrington is $440,000 and $390,000 for Monticello size units. These estimated allowances exceed the actual bag replacement costs for the two stations in question. Thus, we find that EPA was correct in stating in its denial of the petition for reconsideration that "cost estimates developed by EPA provide liberal allowance for start-up and continued maintenance."390
The Electric Utilities also maintained that the high pressure drop encountered at Harrington and Monticello would increase operating costs. EPA responds first that the Harrington plant is undersize, and that both plants experienced difficulties due to poor bag design and fabric selection.391 Because of these temporary problems, costs at Harrington and Monticello are not indicative of the costs at a properly operated baghouse installation. Furthermore, EPA found that even if the high pressure drops predicted by the Electric Utilities were maintained (11 In. H2O as compared with EPA's recommended 5 In. H2O) the result would be a cost increase of $191,000 per year.392 Since this increase amounted to a 4.5 percent increase in the total annualized cost projected for full size baghouse operation and less than 1 percent increase relative to utility operating costs, EPA concluded that the Electric Utilities' forecasted cost was not excessive and did not make the cost of compliance with the standard unreasonable. This is a judgment call with which we are not inclined to quarrel. Besides, EPA reported that the 11 In. H2O pressure drop predicted by the Electric Utilities had been reduced by corrective measures at Harrington to 8 In. H2O.393 This level of pressure drop would reduce the annual cost increment by about $90,000 from $191,000 at 11 In. H2O.394
In sum, EPA has sufficiently established that the standard for particulate emission is achievable by baghouse control by accounting for relevant variables and by demonstrating the representativeness of its data.395 The post-promulgation record does not require a contrary conclusion or indicate that the standard is unreasonably costly. For these reasons we affirm the particulate standard on the basis of the performance of baghouse technology.
V. THE 1.2 LBS./MBTU EMISSION CEILING
EPA proposed and ultimately adopted a 1.2 lbs./MBtu ceiling for total sulfur dioxide emissions which is applicable regardless of the percentage of sulfur dioxide reduction attained.396 The 1.2 lbs./MBtu standard is identical to the emission ceiling required by the former standard.397 The achievability of the standard is undisputed.
EDF398 challenges this part of the final NSPS on procedural grounds, contending that although there may be evidence supporting the 1.2 lbs./MBtu standard, EPA should have and would have adopted a stricter standard if it had not engaged in post-comment period irregularities and succumbed to political pressures.399 EDF raises its procedural objections in the context of its view that a more stringent emission ceiling would have been better than the 1.2 lbs./MBtu limit because it would decrease total emissions significantly without impeding the production or use of coal.400 Although the substantive validity of the 1.2 lbs./MBtu standard is not before the court, it is not possible to evaluate EDF's procedural argument without first examining the evolution of the standard and EPA's explanation for adopting the emission ceiling.
A. EPA's Rationale for the Emission Ceiling
EPA explained in the preamble to the proposed rule that two primary factors were considered in selecting the 1.2 lbs./MBtu ceiling: FGD performance, and the impact of the ceiling on high sulfur coal reserves.401 EPA further explained that it had considered whether to propose a 1.2 lbs./MBtu limitation with and without three exemptions per month.402 EPA's modeling analysis showed that under either option there would be no significant differences in total national production and there would be sufficient reserves available to satisfy national demand for coal. However, EPA predicted that on a regional basis a 1.2 lbs./MBtu ceiling without exemptions would adversely affect the Midwest.403 Consequently, EPA proposed that the emission limitation should have three exemptions but solicited comments on the level of the emission limit and the appropriateness of the 3 day exemption.404
Following the September 1978 proposal the joint interagency working group investigated options lower than the 1.2 lbs./MBtu ceiling, according to EPA, in order "to take full advantage of the cost effectiveness benefits of a joint coal washing/scrubbing strategy on high-sulfur coal."405 The joint working group reasoned that since coal washing is relatively inexpensive, an emission ceiling which would require 90 percent scrubbing in addition to coal washing "could substantially reduce emissions in the East and Midwest at a relatively low cost." Since coal washing is a widespread practice, it was thought that the 1.2 lbs./MBtu proposal would not have a seriously detrimental impact upon Eastern coal production.406 During phase two EPA analyzed 10 different full control and partial control options with its econometric model. These various options included emission ceilings at the 1.2 lbs./MBtu, 0.80 lbs./MBtu and the 0.55 lbs./MBtu levels.407 The modeling results, published before the close of the public comment period in December 1978, confirmed the joint working group's conclusion that the 1.2 lbs./MBtu standard should be lowered.408 The results of the phase two modeling exercise were cited by internal EPA memoranda in January409 and March410 1979 as a basis for lowering the 1.2 lbs./MBtu standard. After the phase two modeling, however, EPA undertook "a more detailed analysis of regional coal production impacts," using BOM seam by seam data on the sulfur content of the reserves and the coal washing potential for those reserves. This analysis identified the amount of reserves that would require more than 90 percent scrubbing of washed coal to meet alternative ceilings.411
As a result of concerns expressed on the record by NCA and others about the impacts of more rigorous emission ceilings, EPA called a meeting of principal participants in the rulemaking for April 5, 1979. At the meeting EPA presented its new analysis which showed that a 0.55 lbs./MBtu limit would require more than 90 percent scrubbing on 5 to 10 percent of Northern Appalachian reserves and 12 to 25 percent of Eastern Midwest reserves. A 0.80 ceiling would require more than 90 percent scrubbing on less than 5 percent of the reserves in each of these regions.412 NCA presented its own analysis on the sulfur content and washability of reserves held by its member companies, "a very small portion of the total reserves but including reserves which are planned to be developed in the near future."413 The NCA data confirmed the EPA analysis within 5 percentage points.414 At the same meeting the Administrator "reviewed his assessment of state of the art coal cleaning technology" and accepted NCA's recommendation that common practice, i. e., crushing to 1½ inch top size rather than to smaller sizes with separation at 1.6 specific gravity, be used as the standard when evaluating the impact of coal washing.415
After the April 5 meeting EPA also "concluded that the actual buying practices of utilities rather than the mere technical usability of coals should be considered."416 This exercise was expected to identify high sulfur coals that utilities would not use in order to avoid the risk of failing to satisfy the standard. EPA assumed that utilities would only purchase coal that would provide about a 10 percent margin below the emission limit in order to minimize risk, and, further that utilities would only purchase coal that would meet the emission limit (with the 10 percent margin) with no more than a 90 percent overall reduction in potential emissions.417 EPA claimed that these assumptions reflected utility preference for buying washed coal for which only 85 percent scrubbing is needed to meet both the percent reduction and the emission limitation, as compared to the previous assumption that utilities would perform 90 percent scrubbing on washed coal (resulting in more than 90 percent reduction in overall emissions). The agency's analysis, according to EPA, showed that up to 22 percent of high sulfur coal reserves in the Eastern Midwest and parts of the Northern Appalachian coal regions would require more than 90 percent reduction if emissions were held to a 1.0 lbs./MBtu standard. Thus, although acknowledging that stricter controls were technically feasible, EPA chose to retain the 1.2 lbs./MBtu standard because "conservatism in utility perceptions of scrubber performance could create a significant disincentive against the use of these coals and disrupt the coal markets in these regions."418 EPA concluded that "[a] more stringent emission limit would be counter to one of the basic purposes of the 1977 Amendments, that is, encouraging the use of higher sulfur coals."419
B. EDF's Procedural Attack
EDF alleges that as a result of an "ex parte blitz" by coal industry advocates conducted after the close of the comment period, EPA backed away from adopting the .55 lbs./MBtu limit, and instead adopted the higher 1.2 lbs./MBtu restriction. EDF asserts that even before the comment period had ended EPA had already narrowed its focus to include only options which provided for the .55 lbs./MBtu ceiling.420 EDF also claims that as of March 9, 1979, the three proposals which EPA had under active consideration all included the more stringent .55 lbs./MBtu ceiling, and the earlier 1.2 lbs./MBtu ceiling had been discarded.421 Whether or not EDF's scenario is credible, it is true that EPA did circulate a draft NSPS with an emissions ceiling below the 1.2 lbs./MBtu level for interagency comment during February, 1978.422 Following a "leak" of this proposal, EDF says, the so-called "ex parte blitz" began. "Scores" of pro-industry "ex parte" comments were received by EPA in the post-comment period, states EDF, and various meetings with coal industry advocates — including Senator Robert Byrd of West Virginia — took place during that period. These communications, EDF asserts, were unlawful and prejudicial to its position.
In order for this court to assess these claims, we must identify the particular actions and incidents which gave rise to EDF's complaints.423 Aside from a passing reference to a telephone call from an EPA official to the Chief Executive Officer of NCA,424 EDF's procedural objections stem from either (1) comments filed after the close of the official comment period, or (2) meetings between EPA officials and various government and private parties interested in the outcome of the final rule, all of which took place after the close of the comment period.
1. Late Comments
The comment period for the NSPS began on September 19, 1978, and closed on January 15, 1979.425 After January 15, EPA received almost 300 written submissions on the proposed rule from a broad range of interests. EPA accepted these comments and entered them all on its administrative docket. EPA did not, however, officially reopen the comment period, nor did it notify the public through the Federal Register or by other means that it had received and was entering the "late" comments. According to EDF, most of the approximately 300 late comments were received after the "leak" of the new .55 lbs./MBtu proposal. EDF claims that of the 138 late comments from non-government sources, at least 30 were from "representatives of the coal or utility industries,"426 and of the 53 comments from members of Congress, 22 were either forwarded by the Congressmen from industry interests, or else were prepared and submitted by Congressmen as advocates of those interests.427
EDF objects to nine different meetings.428 A chronological list and synopsis of the challenged meetings follows:
1. March 14, 1979 — This was a one and a half hour briefing at the White House for high-level officials from the Department of Energy (DOE), the Council of Economic Advisers (CEA), the White House staff, the Department of Interior, the Council on Environmental Quality (CEQ), the Office of Management and Budget (OMB), and the National Park Service.429 The meeting was reported in a May 9, 1979 memorandum from EPA to Senator Muskie's staff, responding to the Senator's request for a monthly report of contacts between EPA staff and other federal officials concerning the NSPS.430 A summary of the meeting and the materials distributed were docketed on May 30, 1979. EDF also obtained, after promulgation of the final rule, a copy of the memorandum to Senator Muskie in response to its Freedom of Information Act ("FOIA") request.431
2. April 5, 1979 — This is the meeting discussed at length above.432 The meeting was attended by representatives of EPA, DOE, NCA, EDF, Congressman Paul Simon's office, ICF, Inc. (who performed the microanalysis), and Hunton & Williams (who represented the Electric Utilities). The participants were notified in advance of the agenda for the meeting.433 Materials relating to EPA's and NCA's presentations during the meeting were distributed and copies were later put into the docket along with detailed minutes of the meeting.434 Followup calls and letters between NCA and EPA came on April 20, 23, and 29, commenting or elaborating upon the April 5 data. All of these followup contacts were recorded in the docket.435
3. April 23, 1979 — This was a 30-45 minute meeting held at then Senate Majority Leader Robert Byrd's request, in his office, attended by EPA Administrator Douglas Costle, Chief Presidential Assistant Stuart Eizenstat, and NCA officials.436 A summary of this meeting was put in the docket on May 1, 1979, and copies of the summary were sent to EDF and to other parties.437 In its denial of the petition for reconsideration, EPA was adamant that no new information was transmitted to EPA at this meeting.438
4. April 27, 1979 — This was a briefing on dry scrubbing technology conducted by EPA for representatives of the Office of Science and Technology Policy, the Council on Wage and Price Stability, DOE, the President's domestic policy staff, OMB, and various offices within EPA.439 A description of this briefing and copies of the material distributed were docketed on May 1, 1979.440
5. April 30, 1979 — At 10:00 a. m., a one hour White House briefing was held for the President, the White House staff, and high ranking members of the Executive Branch "concerning the issues and options presented by the rulemaking."441 This meeting was noted on an EPA official's personal calendar which EDF obtained after promulgation in response to its FOIA request,442 but was never noted in the rulemaking docket.
6. April 30, 1979 — At 2:30 p. m., a technical briefing on dry scrubbing technology at the White House was conducted by EPA for the White House staff. A short memorandum describing this briefing was docketed on May 30, 1979.443
7. May 1, 1979 — Another White House briefing was held on the subject of FGD technology.444 A description of the meeting and materials distributed were docketed on May 30, 1979.445
8. May 1, 1979 — EPA conducted a one hour briefing of staff members of the Senate Committee on Environmental and Public Works concerning EPA's analysis of the effect of alternative emission ceilings on coal reserves. The briefing was "substantially the same as the briefing given to Senator Byrd on May 2, 1980."446 No persons other than Committee staff members and EPA officials attended the briefing. This meeting, like the one at 10:00 a. m. on April 30, was never entered on the rulemaking docket but was listed on an EPA official's calendar obtained by EDF in response to its FOIA request. This EPA official has since stated that it was an oversight not to have a memorandum of this briefing prepared for the docket.447
9. May 2, 1979 — This was a brief meeting between Senator Byrd, EPA, DOE and NCA officials held ostensibly for Senator Byrd to hear EPA's comments on the NCA data.448 A 49 word, not very informative, memorandum describing the meeting was entered on the docket on June 1, 1979.449
On June 16, 1980, responding to motions filed by EDF,450 this court ordered EPA to file affidavits providing additional information regarding five of these nine meetings (March 14, April 23, April 27, April 30, and May 2, 1979).451 After EPA complied with the order, EDF argued that the other meetings held on April 30 and May 1 were still undocumented,452 whereupon EPA voluntarily filed an affidavit describing them.453
EDF believes that the communications just outlined, when taken as a whole, were so extensive and had such a serious impact on the NSPS rulemaking, that they violated EDF's rights to due process in the proceeding, and that these "ex parte" contacts were procedural errors of such magnitude that this court must reverse. EDF does not specify which particular features in each of the above-enumerated communications violated due process or constituted errors under the statute; indeed, EDF nowhere lists the communications in a form designed to clarify why any particular communication was unlawful. Instead, EDF labels all post-comment communications with EPA — from whatever source and in whatever form — as "ex parte," and claims that "this court has repeatedly stated that ex parte contacts of substance violate due process."454
At the outset, we decline to begin our task of reviewing EPA's procedures by labeling all post-comment communications with the agency as "ex parte." Such an approach essentially begs the question whether these particular communications in an informal rulemaking proceeding were unlawful.455 Instead of beginning with a conclusion that these communications were "ex parte," we must evaluate the various communications in terms of their timing, source, mode, content, and the extent of their disclosure on the docket, in order to discover whether any of them violated the procedural requirements of the Clean Air Act, or of due process.
C. Standard for Judicial Review of EPA Procedures
This court's scope of review is delimited by the special procedural provisions of the Clean Air Act,456 which declare that we may reverse the Administrator's decision for procedural error only if (i) his failure to observe procedural requirements was arbitrary and capricious, (ii) an objection was raised during the comment period, or the grounds for such objection arose only after the comment period and the objection is "of central relevance to the outcome of the rule," and (iii) "the errors were so serious and related to matters of such central relevance to the rule that there is a substantial likelihood that the rule would have been significantly changed if such errors had not been made."457 The essential message of so rigorous a standard is that Congress was concerned that EPA's rulemaking not be casually overturned for procedural reasons, and we of course must respect that judgment.
Our authority to reverse informal administrative rulemaking for procedural reasons is also informed by Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council, Inc.458 In its unanimous opinion, the Supreme Court unambiguously cautioned this court against imposing its own notions of proper procedures upon an administrative agency entrusted with substantive functions by Congress. The Court declared that so long as an agency abided by the minimum procedural requirements laid down by statute, this court was not free to impose additional procedural rights if the agency did not choose to grant them.459 Except in "extremely rare" circumstances, the Court stated, there is no justification for a reviewing court to overturn agency action because of the failure to employ procedures beyond those required by Congress.460
[W]hen there is a contemporaneous explanation of the agency decision, the validity of that action must "stand or fall on the propriety of that finding, judged, of course, by the appropriate standard of review. If that finding is not sustainable on the administrative record made, then the ... decision must be vacated and the matter remanded ... for further consideration." .... The [reviewing] court should engage in this kind of review and not stray beyond the judicial province to explore the procedural format or to impose upon the agency its own notion of which procedures are "best" and most likely to further some vague, undefined public good.461
Bearing this caveat in mind, we now set out the procedural requirements which Congress mandated for this rulemaking. Since EDF does not question — nor do we doubt — the constitutional sufficiency of the procedures mandated by the Clean Air Act, we shall reverse only (1) if the statutory requirements, or the procedures reasonably inferable from them or from basic notions of constitutional due process,462 were breached by EPA, and (2) where such breaches under the statute were "so serious and related to matters of such central relevance to the rule that there is a substantial likelihood that the rule would have been significantly changed if such errors had not been made."463
D. Statutory Provisions Concerning Procedure
The procedural provisions of the Clean Air Act specifying the creation and content of the administrative rulemaking record are contained in section 307.464 Responding in part to criticism that there was no formalized record which courts could rely upon when reviewing EPA rules,465 Congress enacted new procedures which represented "[b]y and large ... a legislative adoption of the suggestions for a rulemaking record set forth in a law review article dealing with EPA. (Pedersen, `Formal Records and Informal Rulemaking,' 85 Yale L.J. 38 (1975))."466 The Pedersen article argued for rules which would provide a reviewing court with a "procedural record" which it could rely upon when passing on agency rules. A "procedural record" is, according to Pedersen, a record defined by formal norms explicitly governing the inclusion and exclusion of data and which becomes a record "without further action." "No inquiry [is] necessary [as] to whether the `agency' actually had `considered' the documents in the record, or whether they passed some test of relevance, anymore than is generally done in the course of appellate review of trial court decisions."467 In this sense the Pedersen proposal for a "procedural record" in informal rulemaking resembles the records assembled in trial courts and agency formal adjudications, where record material is defined simply and exclusively as "material which has been accepted under a given obligatory set of procedures,"468 without regard to whether it was actually considered by the decisionmaker.
Pedersen distinguished his "procedural record" idea from what he called a "historical record" approach, in which a reviewing court requires the rulemaking agency to provide it with a post hoc assemblage of materials consisting of the data actually considered (as a matter of "historical" fact) by the agency decisionmaker, regardless of the procedures, if any, which governed the materials' inclusion in any formal record. Pedersen criticized the "historical record" approach469 insofar as it affected the judicial review function, because (1) it confronted the courts with an "openended and disorganized" mass of evidence where "no one knows exactly what [the record should] consist of until judicial review is well underway,"470 and (2) it eroded the bar against probing the actual mental processes of the decisionmaker found in United States v. Morgan,471 since the "historical record" approach requires the court to decide what evidence "really" influenced the agency's thinking. Pedersen's reform proposal thus represented an explicit rejection of the "historical record" approach and was designed "to move rulemaking records as far as possible toward the `procedural' end of the spectrum without reimposing the adjudicatory hearing requirements that have done so much damage in the past."472
Following Pedersen's recommendations, the 1977 Amendments required the agency to establish a "rulemaking docket" for each proposed rule which would form the basis of the record for judicial review.473 The docket must contain, inter alia,474 (1) "notice of the proposed rulemaking ... accompanied by a statement of its basis and purpose," and a specification of the public comment period; (2) "all written comments and documentary information on the proposed rule received from any person ... during the comment period[;] [t]he transcript of public hearings, if any[;] and [a]ll documents ... which become available after the proposed rule has been published and which the Administrator determines are of central relevance to the rulemaking...."; (3) drafts of proposed rules submitted for interagency review, and all documents accompanying them and responding to them; and (4) the promulgated rule and the various accompanying agency documents which explain and justify it.
In contrast to other recent statutes,475 there is no mention of any restrictions upon "ex parte" contacts. However, the statute apparently did envision that participants would normally submit comments, documentary material, and oral presentations during a prescribed comment period. Only two provisions in the statute touch upon the post-comment period, one of which, as noted immediately supra, states that "[a]ll documents which become available after the proposed rule has been published and which the Administrator determines are of central relevance to the rulemaking shall be placed in the docket as soon as possible after their availability."476 But since all the post-comment period written submissions which EDF complains of were in fact entered upon the docket,477 EDF cannot complain that this provision has been violated.478
The only other provision treating post-comment period procedures states that
Only an objection to a rule or procedure which was raised with reasonable specificity during the period for public comment (including any public hearing) may be raised during judicial review. If the person raising an objection can demonstrate to the Administrator that it was impracticable to raise such objection within such time or if the grounds for such objection arose after the period for public comment (but within the time specified for judicial review) and if such objection is of central relevance to the outcome of the rule, the Administrator shall convene a proceeding for reconsideration of the rule and provide the same procedural rights as would have been afforded had the information been available at the time the rule was proposed. If the Administrator refuses to convene such a proceeding, such person may seek review of such refusal in the United States court of appeals for the appropriate circuit....479
In this case EPA refused to convene a reconsideration proceeding, stating
The Administrator does not believe that the procedures cited by EDF were improper. Moreover, as discussed below, any arguable errors were not of central relevance to the outcome of the rule, and therefore do not constitute grounds for granting EDF's petition to reconsider.480
Of course, if this assessment by EPA of EDF's petition for reconsideration were wrong, it would constitute reversible error. But since this court can reverse an agency on procedural grounds only if it finds a failure to observe procedures "required by law,"481 we must first decide whether the procedures followed by EPA between January 15 and June 1, 1979 were unlawful. Only if we so find would we then face the second issue whether the unlawful errors were "of such central relevance to the rule that there is a substantial likelihood that the rule would have been significantly changed if such errors had not been made."482 We now hold that EPA's procedures during the post-comment period were lawful, and therefore do not face the issue whether any alleged errors were of "central relevance" to the outcome.
E. Validity of EPA's Procedures During the Post-Comment Period
The post-comment period communications about which EDF complains vary widely in their content and mode; some are written documents or letters, others are oral conversations and briefings, while still others are meetings where alleged political arm-twisting took place. For analytical purposes we have grouped the communications into categories and shall discuss each of them separately. As a general matter, however, we note at the outset that nothing in the statute prohibits EPA from admitting all post-comment communications into the record; nothing expressly requires it, either.483 Most likely the drafters envisioned promulgation of a rule soon after the close of the public comment period, and did not envision a months-long hiatus where continued outside communications with the agency would continue unabated. We must therefore attempt to glean the law for this case by inference from the procedural framework provided in the statute.
1. Written Comments Submitted During Post-Comment Period
Although no express authority to admit post-comment documents exists, the statute does provide that:
All documents which become available after the proposed rule has been published and which the Administrator determines are of central relevance to the rulemaking shall be placed in the docket as soon as possible after their availability.484
This provision, in contrast to others in the same subparagraph, is not limited to the comment period. Apparently it allows EPA not only to put documents into the record after the comment period is over,485 but also to define which documents are "of central relevance" so as to require that they be placed in the docket. The principal purpose of the drafters was to define in advance, for the benefit of reviewing courts, the record upon which EPA would rely in defending the rule it finally adopted; it was not their purpose to guarantee that every piece of paper or phone call related to the rule which was received by EPA during the post-comment period be included in the docket. EPA thus has authority to place post-comment documents into the docket, but it need not do so in all instances.
Such a reading of the statute accords well with the realities of Washington administrative policymaking, where rumors, leaks, and overreactions by concerned groups abound, particularly as the time for promulgation draws near. In a proceeding such as this, one of vital concern to so many interests — industry, environmental groups, as well as Congress and the Administration — it would be unrealistic to think there would not naturally be attempts on all sides to stay in contact with EPA right up to the moment the final rule is promulgated.486 The drafters of the 1977 Amendments were practical people, well versed in such activity, and we decline now to infer from their silence that they intended to prohibit the lodging of documents with the agency at any time prior to promulgation. Common sense, after all, must play a part in our interpretation of these statutory procedures.
EPA of course could have extended, or reopened, the comment period after January 15 in order formally to accommodate the flood of new documents; it has done so in other cases.487 But under the circumstances of this case, we do not find that it was necessary for EPA to reopen the formal comment period. In the first place, the comment period lasted over four months, and although the length of the comment period was not specified in the 1977 Amendments, the statute did put a premium on speedy decisionmaking by setting a one year deadline from the Amendments' enactment to the rules' promulgation.488 EPA failed to meet that deadline, and subsequently entered into a consent decree489 where it promised to adopt the final rules by March 19, 1979, over seven months late. EPA also failed to meet that deadline, and it was once more extended until June 1, 1979 upon agreement of the parties pursuant to court order.490 Reopening the formal comment period in the late spring of 1979 would have confronted the agency with a possible violation of the court order, and would further have frustrated the Congressional intent that these rules be promulgated expeditiously.
If, however, documents of central importance upon which EPA intended to rely had been entered on the docket too late for any meaningful public comment prior to promulgation, then both the structure and spirit of section 307 would have been violated. The Congressional drafters, after all, intended to provide "thorough and careful procedural safeguards ... [to] insure an effective opportunity for public participation in the rulemaking process."491 Indeed the Administrator is obligated by the statute to convene a proceeding to reconsider the rule where an objection of central importance to it is proffered, and the basis of the objection arose after the comment period had closed.492 Thus we do not hold that there are no circumstances in which reopening the comment period would ever be required.
The case before us, however, does not present an instance where documents vital to EPA's support for its rule were submitted so late as to preclude any effective public comment. The vast majority of the written comments referred to earlier at text accompanying nn.425-27, supra, were submitted in ample time to afford an opportunity for response. Regarding those documents submitted closer to the promulgation date, our review does not reveal that they played any significant role in the agency's support for the rule.493 The decisive point, however, is that EDF itself has failed to show us any particular document or documents to which it lacked an opportunity to respond, and which also were vital to EPA's support for the rule.
EDF makes only one particularized allegation concerning its inability to respond adequately to documents submitted during the post-comment period. It argues that at the April 5 meeting called by EPA, representatives of NCA produced new data494 purporting to show a significant impact upon available coal reserves of more restrictive emissions ceilings. EDF alleges that additional documents supporting a higher ceiling were thereafter forwarded by NCA to EPA following the April 5 meeting. We find, however, that EDF was not denied an adequate opportunity to respond to this material. EDF was provided with advance notice of the April 5 meeting's time, place, and agenda. At the meeting EDF proceeded to question the assumptions used in the coal industry's studies.495 After the meeting, on April 19, 1979, it sent a detailed memorandum to EPA asserting that NCA's new claims were "false" and "unsupported by the sheafs of new data the Coal Association has hastened to submit...."496 Of course EDF would have preferred "additional time" to be able to furnish a "more complete evaluation" of the studies, but we do not find that EDF's preference for more time constitutes substantial evidence of an inability to respond. EDF had many weeks between the meeting and the promulgation of the rule to submit additional material, and its rebuttal material has all been entered on the docket and considered by EPA.497 The mere wishes of a participant in an informal rulemaking for more time to respond to documents in the record to which it already had opportunity to respond cannot force an agency to delay rulemaking simply because some new rebuttal evidence may be forthcoming; this is particularly so when the statute mandates speedy promulgation of the rule. Were it otherwise, participants could delay promulgation indefinitely since new information continually comes to light on the subject of many proposed rules. Finality, after all, has a place in administrative rulemaking, just as it does in judicial decisionmaking.498
We therefore conclude that it was not improper in this case for EPA to docket and consider the post-comment documents submitted to it. Nor was it improper for EPA to decline to reopen the formal comment period and delay promulgation, since there was no proof that evidence crucial to the rule's validity was entered too late for any effective public comment.
2. Meetings Held With Individuals Outside EPA
The statute does not explicitly treat the issue of post-comment period meetings with individuals outside EPA. Oral face-to-face discussions are not prohibited anywhere, anytime, in the Act. The absence of such prohibition may have arisen from the nature of the informal rulemaking procedures Congress had in mind. Where agency action resembles judicial action, where it involves formal rulemaking, adjudication, or quasi-adjudication among "conflicting private claims to a valuable privilege,"499 the insulation of the decisionmaker from ex parte contacts is justified by basic notions of due process to the parties involved.500 But where agency action involves informal rulemaking of a policymaking sort, the concept of ex parte contacts is of more questionable utility.501
Under our system of government,502 the very legitimacy of general policymaking performed by unelected administrators depends in no small part upon the openness, accessibility, and amenability of these officials to the needs and ideas of the public from whom their ultimate authority derives, and upon whom their commands must fall. As judges we are insulated from these pressures because of the nature of the judicial process in which we participate; but we must refrain from the easy temptation to look askance at all face-to-face lobbying efforts, regardless of the forum in which they occur, merely because we see them as inappropriate in the judicial context.503 Furthermore, the importance to effective regulation of continuing contact with a regulated industry, other affected groups, and the public cannot be underestimated. Informal contacts may enable the agency to win needed support for its program, reduce future enforcement requirements by helping those regulated to anticipate and shape their plans for the future, and spur the provision of information which the agency needs.504 The possibility of course exists that in permitting ex parte communications with rulemakers we create the danger of "one administrative record for the public and this court and another for the Commission."505 Under the Clean Air Act procedures, however, "[t]he promulgated rule may not be based (in part or whole) on any information or data which has not been placed in the docket...."506 Thus EPA must justify its rulemaking solely on the basis of the record it compiles and makes public.
Regardless of this court's views on the need to restrict all post-comment contacts in the informal rulemaking context, however, it is clear to us that Congress has decided not to do so in the statute which controls this case. As we have previously noted:
Where Congress wanted to prohibit ex parte contacts it clearly did so. Thus APA § 5(c) forbids ex parte contacts when an "adjudication" is underway, but even that prohibition does not apply to "the agency or a member or members of the body comprising the agency." 5 U.S.C. § 554(d)(C) (1970).... If Congress wanted to forbid or limit ex parte contact in every case of informal rulemaking, it certainly had a perfect opportunity of doing so when it enacted the Government in the Sunshine Act, Pub.L. No. 94-409, 90 Stat. 1241 (Sept. 13, 1976).... That it did not extend the ex parte contact provisions of the amended section 557 to section 553 — even though such an extension was urged upon it during the hearing — is a sound indication that Congress still does not favor a per se prohibition or even a "logging" requirement in all such proceedings.507
Lacking a statutory basis for its position, EDF would have us extend our decision in Home Box Office, Inc. v. FCC508 to cover all meetings with individuals outside EPA during the post-comment period. Later decisions of this court, however, have declined to apply Home Box Office to informal rulemaking of the general policymaking sort involved here,509 and there is no precedent for applying it to the procedures found in the Clean Air Act Amendments of 1977.
It still can be argued, however, that if oral communications510 are to be freely permitted after the close of the comment period, then at least some adequate summary of them must be made in order to preserve the integrity of the rulemaking docket, which under the statute must be the sole repository of material upon which EPA intends to rely.511 The statute does not require the docketing of all post-comment period conversations and meetings,512 but we believe that a fair inference can be drawn that in some instances such docketing may be needed in order to give practical effect to section 307(d)(4)(B)(i), which provides that all documents "of central relevance to the rulemaking" shall be placed in the docket as soon as possible after their availability. This is so because unless oral communications of central relevance to the rulemaking are also docketed in some fashion or other, information central to the justification of the rule could be obtained without ever appearing on the docket, simply by communicating it by voice rather than by pen, thereby frustrating the command of section 307 that the final rule not be "based (in part or whole) on any information or data which has not been placed in the docket...."513
EDF is understandably wary of a rule which permits the agency to decide for itself when oral communications are of such central relevance that a docket entry for them is required. Yet the statute itself vests EPA with discretion to decide whether "documents" are of central relevance and therefore must be placed in the docket; surely EPA can be given no less discretion in docketing oral communications, concerning which the statute has no explicit requirements whatsoever. Furthermore, this court has already recognized that the relative significance of various communications to the outcome of the rule is a factor in determining whether their disclosure is required.514 A judicially imposed blanket requirement that all post-comment period oral communications be docketed would, on the other hand, contravene our limited powers of review,515 would stifle desirable experimentation in the area by Congress and the agencies,516 and is unnecessary for achieving the goal of an established, procedure-defined docket, viz., to enable reviewing courts to fully evaluate the stated justification given by the agency for its final rule.517
Turning to the particular oral communications in this case, we find that only two of the nine contested meetings were undocketed by EPA.518 The agency has maintained that, as to the May 1 meeting where Senate staff people were briefed on EPA's analysis concerning the impact of alternative emissions ceilings upon coal reserves, its failure to place a summary of the briefing in the docket was an oversight. We find no evidence that this oversight was anything but an honest inadvertence; furthermore, a briefing of this sort by EPA which simply provides background information about an upcoming rule is not the type of oral communication which would require a docket entry under the statute.
The other undocketed meeting occurred at the White House and involved the President and his White House staff. Because this meeting involves considerations unique to intra-executive meetings, it is discussed in the section immediately infra.
(a) Intra-Executive Branch Meetings
We have already held that a blanket prohibition against meetings during the post-comment period with individuals outside EPA is unwarranted, and this perforce applies to meetings with White House officials. We have not yet addressed, however, the issue whether such oral communications with White House staff, or the President himself, must be docketed on the rulemaking record, and we now turn to that issue. The facts, as noted earlier, present us with a single undocketed meeting held on April 30, 1979, at 10:00 a. m., attended by the President, White House staff, other high ranking members of the Executive Branch, as well as EPA officials, and which concerned the issues and options presented by the rulemaking.
We note initially that section 307 makes specific provision for including in the rulemaking docket the "written comments" of other executive agencies along with accompanying documents on any proposed draft rules circulated in advance of the rulemaking proceeding. Drafts of the final rule submitted to an executive review process prior to promulgation, as well as all "written comments," "documents," and "written responses" resulting from such interagency review process, are also to be put in the docket prior to promulgation.519 This specific requirement does not mention informal meetings or conversations concerning the rule which are not part of the initial or final review processes, nor does it refer to oral comments of any sort. Yet it is hard to believe Congress was unaware that intra-executive meetings and oral comments would occur throughout the rulemaking process. We assume, therefore, that unless expressly forbidden by Congress, such intra-executive contacts520 may take place, both during and after the public comment period; the only real issue is whether they must be noted and summarized in the docket.
The court recognizes the basic need of the President and his White House staff to monitor the consistency of executive agency regulations with Administration policy. He and his White House advisers surely must be briefed fully and frequently about rules in the making, and their contributions to policymaking considered. The executive power under our Constitution, after all, is not shared — it rests exclusively with the President. The idea of a "plural executive," or a President with a council of state, was considered and rejected by the Constitutional Convention.521 Instead the Founders chose to risk the potential for tyranny inherent in placing power in one person, in order to gain the advantages of accountability fixed on a single source. To ensure the President's control and supervision over the Executive Branch, the Constitution — and its judicial gloss — vests him with the powers of appointment and removal, the power to demand written opinions from executive officers, and the right to invoke executive privilege to protect consultative privacy.522 In the particular case of EPA, Presidential authority is clear since it has never been considered an "independent agency," but always part of the Executive Branch.523
The authority of the President to control and supervise executive policymaking is derived from the Constitution;524 the desirability of such control is demonstrable from the practical realities of administrative rulemaking.525 Regulations such as those involved here demand a careful weighing of cost, environmental, and energy considerations.526 They also have broad implications for national economic policy. Our form of government simply could not function effectively or rationally if key executive policymakers were isolated from each other and from the Chief Executive. Single mission agencies do not always have the answers to complex regulatory problems. An overworked administrator exposed on a 24-hour basis to a dedicated but zealous staff needs to know the arguments and ideas of policymakers in other agencies as well as in the White House.
We recognize, however, that there may be instances where the docketing of conversations between the President or his staff and other Executive Branch officers or rulemakers may be necessary to ensure due process. This may be true, for example, where such conversations directly concern the outcome of adjudications or quasi-adjudicatory proceedings; there is no inherent executive power to control the rights of individuals in such settings.527 Docketing may also be necessary in some circumstances where a statute like this one specifically requires that essential "information or data" upon which a rule is based be docketed.528 But in the absence of any further Congressional requirements, we hold that it was not unlawful in this case for EPA not to docket a face-to-face policy session involving the President and EPA officials during the post-comment period, since EPA makes no effort to base the rule on any "information or data" arising from that meeting.529 Where the President himself is directly involved in oral communications with Executive Branch officials, Article II considerations — combined with the strictures of Vermont Yankee — require that courts tread with extraordinary caution in mandating disclosure beyond that already required by statute.
The purposes of full-record review which underlie the need for disclosing ex parte conversations in some settings do not require that courts know the details of every White House contact, including a Presidential one, in this informal rulemaking setting. After all, any rule issued here with or without White House assistance must have the requisite factual support in the rulemaking record, and under this particular statute the Administrator may not base the rule in whole or in part on any "information or data"530 which is not in the record, no matter what the source. The courts will monitor all this, but they need not be omniscient to perform their role effectively. Of course, it is always possible that undisclosed Presidential prodding may direct an outcome that is factually based on the record, but different from the outcome that would have obtained in the absence of Presidential involvement. In such a case, it would be true that the political process did affect the outcome in a way the courts could not police. But we do not believe that Congress intended that the courts convert informal rulemaking into a rarified technocratic process, unaffected by political considerations or the presence of Presidential power.531 In sum, we find that the existence of intra-Executive Branch meetings during the post-comment period, and the failure to docket one such meeting involving the President, violated neither the procedures mandated by the Clean Air Act nor due process.
(b) Meetings Involving Alleged Congressional Pressure
Finally, EDF challenges the rulemaking on the basis of alleged Congressional pressure, citing principally two meetings with Senator Byrd.532 EDF asserts that under the controlling case law the political interference demonstrated in this case represents a separate and independent ground for invalidating this rulemaking. But among the cases EDF cites in support of its position,533 only D. C. Federation of Civil Associations v. Volpe534 seems relevant to the facts here.
In D. C. Federation the Secretary of Transportation, pursuant to applicable federal statutes, made certain safety and environmental findings in designating a proposed bridge as part of the interstate highway system. Civic associations sought to have these determinations set aside for their failure to meet certain statutory standards, and because of possible tainting by reason of improper Congressional influence. Such influence chiefly included public statements by the Chairman of the House Subcommittee on the District of Columbia, Representative Natcher, indicating in no uncertain terms that money earmarked for the construction of the District of Columbia's subway system would be withheld unless the Secretary approved the bridge. While a majority of this court could not decide whether Representative Natcher's extraneous pressure had in fact influenced the Secretary's decision, a majority did agree on the controlling principle of law: "that the decision [of the Secretary] would be invalid if based in whole or in part on the pressures emanating from Representative Natcher."535 In remanding to the Secretary for new determinations concerning the bridge, however, the court went out of its way to "emphasize that we have not found — nor, for that matter, have we sought — any suggestion of impropriety or illegality in the actions of Representative Natcher and others who strongly advocate the bridge."536 The court remanded simply so that the Secretary could make this decision strictly and solely on the basis of considerations made relevant by Congress in the applicable statute.537
D. C. Federation thus requires that two conditions be met before an administrative rulemaking may be overturned simply on the grounds of Congressional pressure. First, the content of the pressure upon the Secretary is designed to force him to decide upon factors not made relevant by Congress in the applicable statute. Representative Natcher's threats were of precisely that character, since deciding to approve the bridge in order to free the "hostage" mass transit appropriation was not among the decisionmaking factors Congress had in mind when it enacted the highway approval provisions of Title 23 of the United States Code. Second, the Secretary's determination must be affected by those extraneous considerations.538
In the case before us, there is no persuasive evidence that either criterion is satisfied. Senator Byrd requested a meeting in order to express "strongly" his already well-known views that the SO2 standards' impact on coal reserves was a matter of concern to him. EPA initiated a second responsive meeting to report its reaction to the reserve data submitted by the NCA. In neither meeting is there any allegation that EPA made any commitments to Senator Byrd. The meetings did underscore Senator Byrd's deep concerns for EPA, but there is no evidence he attempted actively to use "extraneous" pressures to further his position. Americans rightly expect their elected representatives to voice their grievances and preferences concerning the administration of our laws. We believe it entirely proper for Congressional representatives vigorously to represent the interests of their constituents before administrative agencies engaged in informal, general policy rulemaking, so long as individual Congressmen do not frustrate the intent of Congress as a whole as expressed in statute, nor undermine applicable rules of procedure. Where Congressmen keep their comments focused on the substance of the proposed rule — and we have no substantial evidence to cause us to believe Senator Byrd did not do so here539 — administrative agencies are expected to balance Congressional pressure with the pressures emanating from all other sources. To hold otherwise would deprive the agencies of legitimate sources of information and call into question the validity of nearly every controversial rulemaking.
* * * * * *
In sum, we conclude that EPA's adoption of the 1.2 lbs./MBtu emissions ceiling was free from procedural error. The post-comment period contacts here violated neither the statute nor the integrity of the proceeding. We also hold that it was not improper for the agency to docket and consider documents submitted to it during the post-comment period, since no document vital to EPA's support for the rule was submitted so late as to preclude any effective public comment. Hence we find no reason under section 307 to overturn the 1.2 lbs./MBtu standard. The field of course is open for Congress or the agency to formulate further procedural rules in this area.
Since the issues in this proceeding were joined in 1973 when the Navajo Indians first complained about sulfur dioxide fumes over their Southwest homes, we have had several lawsuits, almost four years of substantive and procedural maneuvering before the EPA, and now this extended court challenge. In the interim, Congress has amended the Clean Air Act once and may be ready to do so again. The standard we uphold has already been in effect for almost two years, and could be revised within another two years.
We reach our decision after interminable record searching (and considerable soul searching). We have read the record with as hard a look as mortal judges can probably give its thousands of pages.540 We have adopted a simple and straight-forward standard of review, probed the agency's rationale, studied its references (and those of appellants), endeavored to understand them where they were intelligible (parts were simply impenetrable), and on close questions given the agency the benefit of the doubt out of deference for the terrible complexity of its job. We are not engineers, computer modelers, economists or statisticians, although many of the documents in this record require such expertise — and more.
Cases like this highlight the critical responsibilities Congress has entrusted to the courts in proceedings of such length, complexity and disorder. Conflicting interests play fiercely for enormous stakes, advocates are prolific and agile, obfuscation runs high, common sense correspondingly low, the public interest is often obscured.
We cannot redo the agency's job; Congress has told us, at least in proceedings under this Act, that it will not brook reversal for small procedural errors; Vermont Yankee reinforces the admonition. So in the end we can only make our best effort to understand, to see if the result makes sense, and to assure that nothing unlawful or irrational has taken place. In this case, we have taken a long while to come to a short conclusion: the rule is reasonable.
ROBB, Circuit Judge, concurs in the result.
1 Typical Wet Scrubber
2 Typical Dry Scrubber
3 Typical Dry Scrubber
4 EPA Phase Three Modeling Analysis, Table 1
5 EPA Phase Three Modeling Analysis, Table 2
6 EPA Phase Three Modeling Analysis, Table 3
7 EPA Phase Three Modeling Analysis, Table 4
8 EPA Phase Three Modeling Analysis, Table 5
9 EPA Reconsideration Analysis, Table 1
10 EPA Reconsideration Analysis, Table 2
11 EPA Reconsideration Analysis, Table 3
12 EPA Economic Analysis of Dry Scrubbing
13 EPA Economic Analysis of Dry Scrubbing
14 EPA Economic Analysis of Dry Scrubbing
15 EPA Economic Analysis of Dry Scrubbing
16 EPA Economic Analysis of Dry Scrubbing
17 EPA Economic Analysis of Dry Scrubbing
18 EPA Economic Analysis of Dry Scrubbing
19 Typical ESP
20 Typical Baghouse (2 Cell)
21 Typical Baghouse (Multicell)
22 EPA's ESP Data
23 EPA's ESP Data for Difficult Cases
24 EPA's Baghouse Data
Table 1. — Key Modeling Assumptions
Growth rates ........................1975-1985; 4.8%/yr.
Nuclear capacity ....................1985: 97 GW.
Oil prices ($ 1975) .................1985: $12.90/bbl.
Coal transportation .................1% per year real increase.
Coal mining labor costs .............U.M.W. settlement and 1% real increase
Capital charge rate .................12.5% for pollution control expenditures.
Coal reporting basis ................1978 dollars.
FGD costs ...........................No change from phase 2 analysis except
for the addition of dry scrubbing
systems for certain applications.
Coal cleaning credit ................5%-35% SO2 reduction assumed for
high sulfur bituminous coals only.
Bottom ash and fly ash content ......No credit assumed.
44 Fed. Reg. at 33608
Table 2. — National 1995 SO2 Emissions From Utility Boilersa
Level of controlb
1975 Current Full control Partial Variable
Plant category actual standards control control
Wetd Drye Wet Dry Wet Dry Wet Dry
SIP/NSPS Plantsc ............ 15.5 15.8 16.0 16.2 15.9 16.2 16.0 16.1
New Plantsf ................. 7.1 7.0 3.1 3.1 3.6 3.4 3.3 3.1
Oil Plants .................. 1.0 1.0 1.4 1.4 1.3 1.2 1.3 1.2
Emissions .......... 18.6 23.7 23.8 20.6 20.7 20.8 20.9 20.6 20.5
Capacity (GW) ...... 205 552 554 521 520 534 537 533 537
Sludge generated (million
tons dry) ................. 23 27 55 58 43 39 50 41
a Results of joint EPA/DOE analyses completed in May 1979 based on oil
prices of $12.90, $16.40, and $21.00/bbl in the years 1985, 1990, and 1995,
b With 520 ng/J maximum emission limit.
c Plants subject to existing State regulations or the current NSPS of 1.2 lb
d Based on wet SO2 scrubbing costs.
e Based on dry SO2 scrubbing costs where applicable.
f Plants subject to the revised standards.
44 Fed. Reg. at 33608.
Table 3. — Regional 1995 SO2 Emissions From Utility Boilersa
Level of controlb
1975 Current Full control Partial Variable
actual standards control control
Wetc Dryd Wet Dry Wet Dry Wet Dry
Emissions .......... 18.6 23.7 23.8 20.6 20.7 20.8 20.9 20.6 20.5
Easte ..................... 11.2 11.2 10.1 10.1 9.8 9.8 9.8 9.7
Midwestf .................. 8.1 8.3 7.9 7.9 7.9 8.0 7.9 8.0
West South Centralg ....... 2.6 2.6 1.7 1.7 1.8 1.8 1.8 1.7
Westh ..................... 1.7 1.7 0.9 0.9 1.2 1.2 1.1 1.1
Capacity (GW) ...... 205 552 554 521 520 534 537 533 537
a Results of joint EPA/DOE analyses completed in May 1979 based on oil
prices of $12.90, $16.40, and $21.00/bbl in the years 1985, 1990, and 1995,
b With 520 ng/J maximum emission limit.
c Based on wet SO2 scrubbing costs.
d Based on dry SO2 scrubbing costs where applicable.
e New England, Middle Atlantic, South Atlantic, and East South Central
f East North Central and West North Central Census Region.
g West South Central Census Region.
h Mountain and Pacific Census Regions.
44 Fed. Reg. at 33608.
Table 4. — Impacts on Fuels in 1995a
Level of controlb
1975 Current standards Full control Partial control Variable control
actual 33% minimum 70% minimum
Wetc Dryd Wet Dry Wet Dry Wet Dry
U.S. Coal Production (million
Appalachia .................... 396 489 524 463 465 475 486 470 484
Midwest ....................... 151 404 391 487 488 456 452 465 450
Northern Great Plains ......... 54 655 630 633 628 622 576 632 602
West .......................... 46 230 222 182 180 212 228 203 217
Total ........................ 647 1,778 1,767 1,765 1,761 1,765 1,742 1,770 1,752
Western Coal Shipped East
(million tons) .................. 21 122 99 59 55 68 59 71 70
Oil Consumption by Power
Plants (million bbl/day):
Power Plants .................. ... 1.2 1.2 1.6 1.6 1.4 1.4 1.4 1.4
Coal Transportation ........... ... 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2
Total ....................... 3.1 1.4 1.4 1.8 1.8 1.6 1.6 1.6 1.6
a Results of EPA analyses completed in May 1979 based on oil prices of $12.90, $16.40, and $21.00/bbl in the years
1985, 1990, and 1995, respectively.
b With 520 ng/J maximum emission limit.
c Based on wet SO2 scrubbing costs.
d Based on dry SO2 scrubbing where applicable.
44 Fed. Reg. at 33609.
Table 5. — 1995 Economic Impactsa
Level of controlb
Current standards Full control Partial control Variable control
33% minimum 70% minimum
Wetc Dryd Wet Dry Wet Dry Wet Dry
Average Monthly Residential Bills ($/
month) .................................. $53.00 $52.85 $54.50 $54.45 $54.15 $53.95 $54.30 $54.05
Indirect Consumer Impacts ($/month) ....... ...... ...... 1.50 1.60 1.15 1.10 1.30 1.20
Incremental Utility Capital Expenditures,
Cumulative 1976-1995 ($ billions) ....... ...... ...... 4 5 6 - 3 10 - 1
Incremental Annualized Cost ($ billions) .. ...... ...... 4.1 4.4 3.2 3.0 3.6 3.3
Present Value of Incremental Utility
Revenue Requirements ($ billions) ....... ...... ...... 41 45 32 31 37 33
Incremental Cost of SO2 Reduction ($/ton) . ...... ...... 1,322 1,428 1,094 1,012 1,163 1,036
a Results of EPA analyses completed in May 1979 based on oil prices of $12.90, $16.40, and $21.00/bbl in the years
1985, 1990, and 1995, respectively.
b With 520 ng/J maximum emission limit.
c Based on wet SO2 scrubbing costs.
d Based on dry SO2 scrubbing costs where applicable.
44 Fed. Reg. at 33609.
Table 1. — Summary of 1995 Impacts With Phase 3 Assumptions1
Level of control with 520 ng/J maximum
Current 50 pct 70 pct Full
standards minimum minimum control
National SO2 Emissions (million tons) .. 23.8 20.6 20.5 20.7
East2 .............................. 11.2 9.7 9.7 10.1
Midwest ............................ 8.3 8.0 8.0 7.9
West South Central ................. 2.6 1.8 1.7 1.7
West ............................... 1.7 1.1 1.1 0.9
Incremental Annualized Cost
(billions 1978 $) .................... 2.9 3.3 4.4
Incremental Cost of SO2 Reduction
(1978 $/ton) .............. .......... 914 1,036 1,428
Oil Consumption (million bbl/day) ...... 1.4 1.6 1.6 1.8
Coal Production (million tons) ......... 1,767 1,745 1,752 1,751
Total Coal Capacity (GW) ............... 554 537 537 520
1 With wet and dry scrubbing and the following energy assumptions:
Oil prices Nuclear
($ 1975) Capacity
1985 .... $12.90 97
1990 .... 16.40 165
1995 .... 21.00 228
2 See 44 FR 33608 for designation of census regions.
Assumed Oil Prices
[Dollars per Barrel]
Sensitivity Phase 3
1985 ........ 25 16
1990 ........ 30 20
1995 ........ 38 26
45 Fed. Reg. at 8218.
Table 2. — Summary of 1995 Impacts With Higher Oil Prices1
Level of control with 520 ng/J maximum
Current Variable Variable Full
standards control, control, control
50 pct 70 pct
National SO2 Emissions (million tons) .. 23.2 19.6 19.6 19.7
East2 ............................. 10.9 9.1 9.1 9.5
Midwest ............................ 8.2 7.9 7.8 7.8
West South Central ................. 2.6 1.7 1.6 1.5
West ............................... 1.6 1.1 1.0 0.9
Incremental Annualized Cost
(billions 1978 $) .................... ..... 3.3 3.6 5.0
Incremental Cost of SO2 Reduction
(1978 $/ton) ......................... ..... 967 977 1,049
Oil Consumption (million bbl/day) ...... 0.9 0.9 0.9 0.9
Coal Production (million tons) ......... 1,800 1,797 1,802 1,632
Total Coal Capacity (GW) ............... 588 587 587 587
1 With wet and dry scrubbing and the following energy assumptions:
Oil prices Nuclear
($ 1975) Capacity
1985 .... $20.20 97
1990 .... 24.20 165
1995 .... 30.70 228
2 See 44 FR 33608 for designation of census regions.
45 Fed. Reg. at 8219.
Table 3. — Summary of 1995 Impacts With Higher Oil Prices and Less
Level of control with 520 ng/J maximum
Current Variable Variable Full
standards control, control, control
50 pct 70 pct
National SO2 Emissions (million tons) .. 25.0 20.9 20.6 20.5
East2 ............................... 12.0 9.8 9.7 10.1
Midwest ................................ 8.6 8.2 8.1 8.0
West South Central ..................... 8.6 1.6 1.7 1.6
West ................................... 1.7 1.2 1.1 0.9
Incremental Annualized Cost
(billions 1978 $) .................... ..... 3.8 4.1 5.9
Incremental Cost of SO2 Reduction
(1978 $/ton) ......................... ..... 883 914 1,259
Oil Consumption (million bbl/day) ...... 0.9 0.9 0.9 0.9
Coal Production (million tons) ......... 1,940 1,943 1,948 1,964
Total Coal Capacity (GW) ............... 644 644 644 643
1 With wet and dry scrubbing and the following energy assumptions:
Oil prices Nuclear
($ 1975) Capacity
1985 .... $20.20 92
1990 .... 24.20 141
1995 .... 30.70 173
2 See 44 FR 33608 to designation of census regions.
Assumed Nuclear Capacity
Sensitivity Phase 3
1985 ........ 92 GW 97 GW
1990 ........ 141 GW 165 GW
1995 ........ 173 GW 228 GW
45 Fed. Reg. at 8219.
Table 3-7 — Economics of Dry SO2 Control
Coal Sulfur Content
1.68 2.3 4.0
Total Capital 30.30 30.30 30.30
Investment $ KW
Lime 2,678,000 3,666,000 6,376,000
Labor 454,000 454,000 454,000
Replacements 500,000 500,000 500,000
Power Costs 213,000 213,000 213,000
Disposal 525,000 718,000 1,250,000
TOTAL 4,370,000 5,551,000 8,793,000
O&M Mills/KWH 1.5 1.9 3.1
Based Upon 500 MWe Plant, 70% Removal
65% Load Factor
$40/Ton of Lime
$5/Ton Waste Disposal Cost
Non-Alkaline Ash (1977 Dollars)
Ad. Doc. No. V-B-1, supra note 75, at 3-67, J.A. at 2671.
Table 3-8 — Representative Coals Used in EPA's Economic Model
Pounds of SO2/ Coal Sulfur
Designation Million BTU Content, % Coal Rank
A 0.8 0.4 Subbituminous
B 1.2 0.6 Subbituminous
D 1.7 0.68-0.8 Subbituminous, bituminous
F 3.3 1.32-1.65 Subbituminous, bituminous
G 5.0 2.5 Bituminous
H greater than 5.0 greater than 2.5 Bituminous
Ad. Doc. No. V-B-1, supra note 75, at 3-69, J.A. at 2673.
Table 4-2 — Summary of Data on Pulverized Coal-Fired Steam Generator
High Efficiency Electrostatic Precipitator Systems
Specific Collection Area (SCA)
Unit & Reference Square Metres Per Actual Cubic Control Effectiveness
Identification Unit Size ESP Type Metres per Second Nanograms per Joule
Number Megawatts (% Sulfur) (Ft2/1000 ACFM) (lb/106 Btu)
19 500 Cold (0.6) 96.1 18-20a
210 800 Hot (0.5) 60.4 12-18a
310 800 Hot (0.5) 60.4 14.16a
411 570 Cold (1.9) 57.3-59.1 14-17
512 570 Cold (1.9) 52.8-54.3 10-18
613 500 Cold (0.5) 174 3.0-8.6a
714 657 NG (1.1) NG 16-17a
815 46 NG (1.4) NG 9.0-15a
915 46 NG (1.4) NG 6.4-14a
1016 1300 Cold (0.9) 65.9 17-20
1117 69 Hot (1.4) NG 8.2-15a
1218 250 Cold (0.7) 158 19-20
1318 250 Cold (0.7) 158 20-21
1419 190 Hot (1.4) 53.0 4.3-5.6a
1520 350 Hot (0.4) NG 18.9a,b
1621 700 Hot (1.4) 53.7 15-17a,b
1722 141 NG (0.8) NG 11-13a
1822 187 NG (0.8) NG 6.0-7.3a
1923 411 NG (1.1) NG 5.6-10a
2024 74 NG (1.4) NG 10.13a
2125 680 Cold (0.5) 151.0 12-17a,b
a EPA Method 5
b EPA Sponsored Test
NG = Data not given
Ad. Doc. No. III-B-1, supra note 311, at 4-34, J.A. at 1394.
Table 4-3 — Summary of Data on Difficult Electrostatic Precipitator Control
Cases for Pulverized Coal-Fired Steam Generators
Specific Collection Area (SCA) Average
Unit & Reference Square Metres Per Actual Cubic Control Effectiveness
Identification Unit Size ESP Type Metres per Second Nanograms per Joule
Number Megawatts (% Sulfur) (Ft2/1000 ACFM) (lb/106 Btu)
210 800 Hot (0.5) 60.4 13.7a
310 800 Hot (0.5) 60.4 15.5a
2226 330 Hot (0.9) 57.5 17.2a
1218 250 Cold (0.7) 158 19.8
1318 250 Cold (0.7) 158 20.6
613 500 Cold (0.5) 174 4.0a
2125 680 Cold (0.5) 151 13.6
a = EPA Method 5 b = EPA Sponsored Test
Ad. Doc. No. III-B-1, supra note 311, at 4-38, J.A. at 1398.
Table 4-5 — Summary of Data on Baghouses Applied to Coal-Fired Steam Generator Combustion Gases
Air to Cloth Ratio
Coal Sulfur Actual Cubic Metres Effectiveness
Source Content and Type Per Minute per Pressure Drop Nanograms
Identification Source Source Size Type Cleaning Square Metre Kilopascals per Joule
Number Type Megawatts Filtration Method (ACFM/Ft2) (inches H2O) (lb/106 Btu)
127 S (2.0) 10* Outside-in Reverse air 0.61-0.76 0.52-0.70 4.7a,b
(2.0-2.5) (2.1-2.8) (0.011)
127 S (2.0) 10* Outside-in Pulse jet 0.76-0.91 0.55-1.1 9.9a,b
(2.5-3.0) (2.2-4.3) (0.023)
228 PC (2.0) 44** Inside-out Reverse air 0.58 0.62 1.2-43a,b
(1.9) (2.5) (0.0028-0.0100)
329 S (0.7) 13** Inside-out Shaking and 0.85 1.25 0.60-7.7a,b
reverse air (2.8) (5.0) (0.0014-0.0180)
430 S 18* NG NG NG NG 11.2c
530 S 12.5* NG NG NG NG 3.9c
630 S 24* NG NG NG NG 18c
730 S 6.4* NG NG NG NG 6.4c
831 PC (0.5) 25* Inside-out Reverse air 0.69 2-2.5 15.5a,b
(2.26) (8-10) (0.036)
NG = Data not given
S = Stoker
PC = Pulverized Coal
* = Industrial Boiler
a = EPA Method 5
b = EPA Sponsored Test
c = Non-EPA or No Data
** = Utility Boiler
Ad. Doc. No. III-B-1, supra note 311, at 4-41, J.A. at 1401.