These consolidated cases present a variety of challenges to actions of the Environmental Protection Agency (EPA) in setting standards to govern emissions of particulate matter and oxides of nitrogen from diesel vehicles. The Natural Resources Defense Council (NRDC) argues that the agency's actions do not adequately protect the public health; General Motors Corporation (GM) and Intervenors Mercedes-Benz of North America, Inc., and Volkswagen of America, Inc., assert that the EPA did not give adequate consideration to safety factors, and that, in a variety of ways, the standards are too strict. Finding that the agency has stated adequate reasons for its decisions, and that its actions are consistent with statute, we uphold the challenged regulations in their entirety.

The EPA is authorized by the Clean Air Act (Act)1 to regulate emissions of harmful pollutants from motor vehicles. The Act itself specifies the quantity of acceptable emissions from light-duty vehicles for three classes of pollutants: carbon monoxide, hydrocarbons, and oxides of nitrogen. Act § 202(b)(1). Section 202(a)(1) of the Act confers on the EPA Administrator the general power to prescribe by regulation "standards applicable to the emission of any air pollutant from any class or classes of new motor vehicles or new motor vehicle engines, which in his judgment cause, or contribute to, air pollution which may reasonably be anticipated to endanger public health or welfare." These provisions are supplemented and qualified by various specific provisions relating to particular classes of vehicles or pollutants. E.g., Act §§ 202(a)(3)(A)(i), 202(a)(3)(F), 202(b)(6)(A).

The statutory standard for hydrocarbon emissions from light-duty vehicles is an absolute one. For models manufactured from 1977 to 1979, hydrocarbon emissions may not exceed 1.5 grams per vehicle mile; for those manufactured from 1980 on, the standards must require a reduction of at least ninety percent from the emission standards applying in 1970. Act § 202(b)(1)(A). The statutory standards for carbon monoxide and oxides of nitrogen are also absolute, but they are subject to a variety of waivers for certain manufacturers who lack the technological capacity to comply. See Act §§ 202(b) (1)(B), 202(b)(5), 202(b)(6).

The emission standards set by the EPA under its general regulatory power, in contrast, are "technology-based" the levels chosen must be premised on a finding of technological feasibility. Section 202(a)(2) of the Act provides that standards promulgated under section 202(a)(1) shall not take effect until "after such period as the Administrator finds necessary to permit the development and application of the requisite technology."

The requirement that emission standards be technologically achievable highlights the need for the EPA's power to divide the broad spectrum of motor vehicles into classes or categories. See Act §§ 202(a)(1), 202(a)(3)(A)(iv). Manufacturers produce a wide variety of motor vehicles of different sizes, some using different engine technologies resulting in unusual emission characteristics. In particular, diesel engines use a different fuel, emit exhaust at a lower temperature, and produce a different distribution of pollutants than traditional gasoline engines. For example, diesel carbon monoxide levels are typically lower than those from gasoline vehicles, see 45 Fed.Reg. 5480, 5493 n.192 (1980), but diesel vehicles produce particulate emissions at thirty to seventy times the rate of gasoline vehicles, see 45 Fed.Reg. 14,496 (1980), and also produce higher levels of the unregulated pollutants sulfur dioxide and benzo(a)pyrene, see 45 Fed.Reg. 5480, 5489 (1980).

The present challenges concern the EPA's promulgation of standards governing particulate2 emissions from light-duty diesel vehicles and light-duty diesel trucks,3 and the EPA's waiver of the statutory standard for oxides of nitrogen4 for light-duty vehicles.5 The EPA's particulate standard and NOx decisions are appropriately linked in the present proceeding because current technology creates an unfortunate trade-off between particulate control and control of oxides of nitrogen. The primary technique used today for reducing NOx emissions is exhaust gas recirculation (EGR). While lowering the NOx content of the exhaust, EGR increases the particulate content, and "the greater the EGR rate, the greater the increase in particulate emissions." Environmental Protection Agency, Regulatory Analysis (of) Light-Duty Diesel Particulate Regulations 33 (1980) (hereinafter cited as Regulatory Analysis), Joint Appendix (J.A.) 510. Thus the stringency of a technology-based particulate standard depends on the level of the NOx standard concurrently applied. We consider the EPA's actions and the NRDC and industry challenges in turn.6

The EPA announced its intention to promulgate standards for particulate emissions from light-duty diesels on February 1, 1979. See 44 Fed.Reg. 6650 (1979). The proposed standards would have limited diesel particulates to 0.60 grams per vehicle mile (gpm) in model year 1981, and to 0.20 gpm in model year 1983. The agency concluded that a single standard, governing all light-duty vehicles, was the preferable regulatory strategy, although 1979 certification data indicated that diesel particulate performance among those vehicles ranged from the 0.23 gpm achieved by the Volkswagen Rabbit to the 0.84 gpm emitted by the Oldsmobile 350. Id. at 6651. Furthermore, these restrictions would have applied equally to light-duty vehicles and light-duty trucks. Id. at 6654-55.

After analyzing the comments elicited by its notice of proposed rulemaking, the EPA promulgated as final standards a modification of the rules originally announced. See 45 Fed.Reg. 14,496 (1980). The limit of 0.60 gpm was retained, but its effective date was postponed to model year 1982, because the rulemaking process had absorbed so much time that testing and certification of 1981 models was no longer feasible. Id. at 14,497. The agency concluded that the technology necessary to make the 0.20 gpm standard feasible would probably not be developed in time for implementation in 1983 model vehicles; 1984 was a more likely goal, but the effective date was postponed to model year 1985 to give sufficient margin for error. Id. at 14,498. Finally, the EPA believed that light-duty trucks would not be able to perform as well as light-duty vehicles, and the 1985 standard for light-duty trucks was therefore adjusted to 0.26 gpm. Id. at 14,497.

The auto industry petitioners do not challenge the 1982 standard of 0.60 gpm, but they vigorously deny the likelihood that technology will be available to meet the lower standards in 1985.7 In setting the 1985 standards, the EPA predicted that a currently experimental particulate control device, known as a "trap-oxidizer," would be perfected early enough to allow its mass production and installation in 1985 model diesel vehicles. The manufacturers argue that this prediction lacked a sufficient evidentiary basis, and that the agency's action must therefore be invalidated as failing to meet the requirement of reasoned decisionmaking. They also argue that the EPA gave inadequate consideration to the safety risks involved in trap-oxidizer technology.

NRDC insists that the EPA's entire regulatory strategy is an inadequate response to the agency's statutory mandate to protect the public health. The EPA deliberately set a single standard for all light-duty diesel vehicles, predicting that even the worst performing diesel could meet it. NRDC argues that that regulatory choice is inconsistent with the EPA's statutory responsibilities; it urges a variable standard, imposing more rigorous requirements on better performing vehicles. NRDC also urges that the agency failed to consider the risks posed by diesel particulate as a carcinogen, and that in giving "appropriate consideration" to cost as a factor in standard-setting, it should have tried to discourage purchase of polluting vehicles through economic disincentives. Finally, NRDC attacks the postponement of the 0.20 gpm standard from 1984 to 1985 as unnecessary and irresponsible.

In order to evaluate these claims, we first determine the applicable statutory directives. We then turn to light-duty vehicles and, after discussing the proper standard of review in this case, assess the manufacturers' attack on the EPA's prediction of technological feasibility. Next, we examine the NRDC's claims and, finally, turn to light-duty trucks.

Before turning to the merits of the EPA's rulemaking activities, we must consider a threshold question that has attracted attention sporadically throughout the course of the proceedings. Although none of the parties contests the EPA's authority to regulate diesel particulate emissions, there is some disagreement as to the precise statutory provision conferring that authority. GM insists that the EPA's particulate rulemaking for non-heavy-duty vehicles8 may be viewed only as an exercise of its broad general power to prescribe motor vehicle emission standards under section 202(a)(1) of the Act;9 the EPA argues that an explicit provision of the Act specifically directs it to promulgate particulate standards.

In proposing the challenged regulations, the EPA cited as its statutory basis section 202(a)(3)(A)(iii) of the Act, added by the 1977 Clean Air Act Amendments, Pub.L.No.95-95, § 224, 91 Stat. 685. That provision, set out more fully in the margin,10 requires that "(t)he Administrator shall prescribe regulations under (section 202(a)(1)) applicable to emissions of particulate matter from classes or categories of vehicles manufactured during and after model year 1981 (or during any earlier model year, if practicable)." GM maintains that this provision empowers the agency only to set standards for heavy-duty vehicles, while the EPA reads it as applying to any "classes or categories of vehicles," including all light-duty diesels. The apparent meaning of the language of section 202(a)(3)(A)(iii), read in isolation, comports with the EPA's view, but we conclude that that language should not be read in isolation, see, e. g., Brown v. Duchesne, 60 U.S. (19 How.) 183, 194, 15 L. Ed. 595 (1857). The structure of section 202 casts doubt on the validity of a literal reading, and after a closer examination of the section's legislative history, we conclude that GM's position is correct.

The EPA's power under section 202(a)(1) antedates the 1977 amendments such authorization, to the agency and its predecessors, has been present in varying forms since the Clean Air Act Amendments of 1965, Pub.L.No.89-272, § 101, 79 Stat. 992. The legislative history of the 1970 amendments11 demonstrates that Congress "expected that section 202(a) authority would be used for regulation of particulate emissions." S.Rep.No.1196, 91st Cong., 2d Sess. 24 (1970), reprinted in 1 Legislative History of the Clean Air Act Amendments 424 (1974) (hereinafter cited as Legislative History). Congress did not write in a statutory particulate standard at that time, as it did for vehicular emissions of carbon monoxide, hydrocarbons, and oxides of nitrogen,12 because of technological obstacles:

No present measurement techniques exist to evaluate or establish standards for such particulate emissions. Such standards cannot be established on the basis of 1970 vehicles as required by subsection (b) because measurement techniques will not exist until 1972. At such time as measurement methods are developed the Secretary would be expected to establish standards for particulate emissions under 202(a) authority.

Id.

By 1977, however, Congress had become impatient with the EPA's failure to promulgate a particulate standard.

The need for a standard for the pollutant particulate matter for motor vehicles was identified in the Senate report in 1970. None has yet been proposed, except for a smoke standard for heavy duty diesels.

EPA should promulgate a standard requiring best available control technology for smoke from heavy duty diesels and other particulate emissions from heavy duty vehicles and motorcycles by model year 1978, if possible, or by model year 1979.

S.Rep.No.127, 95th Cong., 1st Sess. 67 (1977), reprinted in 3 Legislative History 1441. Congress therefore adopted the mandatory language of section 202(a)(3)(A)(iii).

The 1977 Senate report illustrates that congressional concern over vehicular particulate emissions focused on heavy-duty vehicles. They are the major source of the problem; as the EPA observed in the present rulemaking, light-duty "gasoline-powered vehicles with three-way catalysts emit very low levels of particulate." Regulatory Analysis at 31, J.A. 510. Not surprisingly, then, the origin of section 202(a)(3)(A)(iii) can be traced to a provision in the 1977 Senate bill "applicable to emissions of carbon monoxide, hydrocarbons, particulates, and oxides of nitrogen from heavy duty trucks, buses, and motorcycles and engines thereof," S. 252, 95th Cong., 1st Sess. § 19 (1977), see S.Rep.No.127, 95th Cong., 1st Sess. 193 (1977), reprinted in 3 Legislative History 1567.13

The Senate provision was compromised in conference with a less stringent House provision, and the contents of the Senate's bulky paragraph were distributed among subparagraphs of paragraph 202(a)(3)(A) under circumstances suggesting that the omission of the words "heavy duty" from subparagraph 202(a)(3)(A) (iii) was inadvertent. The section of the act that emerged from conference bore the title, "Emission Standards for Heavy Duty Vehicles or Engines and Certain Other Vehicles or Engines," Pub.L.No.95-95, § 224, 91 Stat. 685; the substance of that section indicates that the "certain other vehicles" are motorcycles. Subparagraphs 202(a)(3)(A)(i) and 202(a)(3)(A)(ii) still retain the limitation to heavy-duty vehicles. Subparagraph 202(a)(3)(A)(iv), which empowers the Administrator to establish "classes or categories of vehicles or engines for purposes of regulations under this paragraph," is not so restricted, but the conference report is written as if it were: "The Administrator is specifically authorized to subdivide heavy duty vehicles into classes or categories for purposes of this section." H.R.Rep.No.564, 95th Cong., 1st Sess. 164 (1977), reprinted in (1977) U.S.Code Cong. & Ad.News 1502, 1544, 3 Legislative History 544.

This accumulation of detail convinces us that Congress did not intend, in adding section 202(a)(3)(A)(iii), to mandate adoption of particulate standards for light-duty vehicles. Rather, Congress directed the EPA to give priority to establishing particulate emission standards for heavy-duty vehicles, and left the agency free to exercise its power under section 202(a)(1) to regulate light-duty automobiles, whether diesel-powered or otherwise.14

Our conclusion that the EPA cited the wrong section of the statute15 as authority for its rulemaking does not necessitate a remand, even though paragraph 202(a)(1) requires, as subparagraph 202(a)(3)(A)(iii) does not, a threshold finding that the regulated pollutant "may be anticipated to endanger public health or welfare." Particulates have long been recognized as one of the major targets of the Clean Air Act. See, e. g., U.S. Dep't of Health, Education, and Welfare, Air Quality Criteria for Particulate Matter (National Air Pollution Control Administration Publication No. AP-49) (1969). And the EPA, in issuing the present regulations, made specific findings on the environmental impact of diesel particulates:

Section 202(a)(3)(A)(iii) directs EPA to control particulate emissions; it does not require that the Agency first conduct an environmental impact assessment. Nevertheless, EPA has carefully examined the environmental impact of this rulemaking.

....

... Small particles, which are much more likely to be deposited in the alveolar region and which require much longer periods of time to be cleared from the respiratory tract, have a greater potential to affect human health than larger particles. Thus, control of diesel particulate, of which 100 percent is less than 15 micrometers in diameter and approximately 97 percent of which is less than 2.5 micrometers in diameter, is especially important with respect to human health.

45 Fed.Reg. 14,496, 14,498-99 (1980). Indeed, GM expressly concedes in its brief that "the Administrator considered and discussed the health risks of particulate emissions during rulemaking (and that his) determination that particulate emissions require priority control to protect human health would entitle him to prescribe particulate standards under § 202(a)(1) for light-duty vehicles." Reply Brief for Petitioner General Motors Corp. at 6 (hereinafter GM Brief).16

The EPA's choice of the 0.20 gpm standard for light-duty diesels in 1985 was the result of adjusting current diesel particulate emission data by the percentage of reduction expected from certain technological improvements, most notably the trap-oxidizer. The manufacturers' attack on the standard focuses on the EPA's prediction concerning the probable pace of development of trap-oxidizer technology. Before examining the details of the agency's reasoning and the industry challenges, however, we find it useful to discuss the legal standard that governs our inquiry.

The standard of review in this case is the traditional one for judicial scrutiny of agency rulemaking: we are to set aside any action found to be "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law." Act § 307(d)(9)(A). As nonscientists, we must recall that "(o)ur 'expertise' is not in setting standards for emission control but in determining if the standards as set are the result of reasoned decisionmaking." Essex Chemical Corp. v. Ruckelshaus, 486 F.2d 427, 434 (D.C.Cir.1973), cert. denied, 416 U.S. 969, 94 S. Ct. 1991, 40 L. Ed. 2d 558 (1974). Despite this limited role, our examination of the record must be searching, for

the necessity to review agency decisions, if it is to be more than a meaningless exercise, requires enough steeping in technical matters to determine whether the agency "has exercised a reasoned discretion." ... We cannot substitute our own judgment for that of the agency, but it is our duty to consider whether "the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment."

Portland Cement Ass'n v. Ruckelshaus, 486 F.2d 375, 402 (D.C.Cir.1973), cert. denied, 417 U.S. 921, 94 S. Ct. 2628, 41 L. Ed. 2d 226 (1974).

In the present case, GM attacks the EPA's estimation of the period of time "necessary to permit the development and application of the requisite technology" to achieve compliance with the 1985 particulate standards, see Act § 202(a)(2). The agency has determined that the technology will be available in time, and now seeks to defend its conclusion as a product of reasoned decisionmaking. Such predictions inherently involve a greater degree of uncertainty than estimations of the effectiveness of current technology. If we judge the EPA's action by the standard of certainty appropriate to current technology, the agency will be unable to set pollutant levels until the necessary technology is already available.

The legislative history of both the 1970 and the 1977 amendments demonstrates that Congress intended the agency to project future advances in pollution control capability. It was "expected to press for the development and application of improved technology rather than be limited by that which exists today." S.Rep.No.1196, 91st Cong., 2d Sess. 24 (1970), reprinted in 1 Legislative History 424; H.R.Rep.No.294, 95th Cong., 1st Sess. 273 (1977), reprinted in (1977) U.S.Code Cong. & Ad.News 1077, 1352, 4 Legislative History 2740. In designing the particulate standard, the EPA recognized the uncertainty necessarily accompanying its duty to predict:

When projecting a near-term standard when little time exists for technological advances, it is relatively simple for a regulatory agency to predict what the best available control technology will be, and to set a standard based on its application. It is more difficult to regulate on this basis in the long-term because of the uncertainty that inevitably surrounds expected technological improvements. Nevertheless, ... EPA has concluded that it is absolutely necessary to issue standards which motivate the private sector to maximize its efforts in reducing particulate emissions from light-duty vehicles.

Regulatory Analysis at 32, J.A. 511.

This court has upheld the agency's power to make such projections, while recognizing that it is "subject to the restraints of reasonableness, and does not open the door to ' "crystal ball" inquiry.' " International Harvester Co. v. Ruckelshaus, 478 F.2d 615, 629 (D.C.Cir.1973). The Clean Air Act requires the EPA to look to the future in setting standards, but the agency must also provide a reasoned explanation of its basis for believing that its projection is reliable. This includes a defense of its methodology for arriving at numerical estimates. Id.

The thoroughness and persuasiveness of the explanation we can expect from the agency will, of course, vary with the nature of the prediction undertaken. "Where existing methodology or research in a new area of regulation is deficient, the agency necessarily enjoys broad discretion to attempt to formulate a solution to the best of its ability on the basis of available information." Industrial Union Dep't v. Hodgson, 499 F.2d 467, 474 n.18 (D.C.Cir.1974). At one extreme, this court has recognized that the EPA's decision to regulate potentially harmful pollutants involves a large element of policy choice that cannot be demonstrably "correct," although it must have a genuine scientific basis.

The Administrator may apply his expertise to draw conclusions from suspected, not completely substantiated, relationships between facts, from trends among facts, from theoretical projections from imperfect data, from probative preliminary data not yet certifiable as "fact," and the like. We believe that a conclusion so drawn a risk assessment may, if rational, form the basis for health-related regulations under the "will endanger" language of Section 211.

Ethyl Corp. v. EPA, 541 F.2d 1, 28 (D.C.Cir.1976) cert. denied, 426 U.S. 941, 96 S. Ct. 2663, 49 L. Ed. 2d 394 (1976) (footnote omitted); see Environmental Defense Fund v. EPA, 598 F.2d 62, 83-85 (D.C.Cir.1978). We have also acknowledged the necessarily speculative nature of agency predictions in the social sciences, including judgments of the competitive impact of regulatory decisions. United States v. FCC, 652 F.2d 72, 100 (D.C.Cir.1980); National Small Shipments Traffic Conference v. CAB, 618 F.2d 819, 829-30 (D.C.Cir.1980). At the other extreme, this court's inquiry into agency methodology in the physical sciences has been far more exacting "where the facts pertinent to (a) standard's feasibility are available and easily discoverable by conventional technical means." National Lime Ass'n v. EPA, 627 F.2d 416, 454 (D.C.Cir.1980).

The present case lies between those two extremes. It does not involve questions at the frontier of physiological knowledge, but it does require a determination by the EPA of the likely sequence of further technological development. There is no known scientific technique for calculating when an as yet unsolved design problem will be ironed out. Thus, unlike the short-term feasibility assessments scrutinized in National Lime Association, the present determination presents the court with "the question how much deference is owed a judgment predicated on limited evidence when additional evidence cannot be adduced or adduced in the near future," id. at 454.

The time element in the EPA's prediction affects our reviewing task in three distinct ways. First, it introduces uncertainties in the agency's judgment that render that judgment vulnerable to attack. At the same time, however, the time element gives the EPA greater scope for confidence that theoretical solutions will be translated successfully into mechanical realizations, for "the question of availability is partially dependent on 'lead time', the time in which the technology will have to be available." Portland Cement Ass'n v. Ruckelshaus, 486 F.2d 375, 391 (D.C.Cir.1973), cert. denied, 417 U.S. 921, 94 S. Ct. 2628, 41 L. Ed. 2d 226 (1974). Finally, the presence of substantial lead time for development before manufacturers will have to commit themselves to mass production of a chosen prototype gives the agency greater leeway to modify its standards if the actual future course of technology diverges from expectation.

The relevance of lead time, and of the ability to modify standards in light of future developments, to the degree of justification the agency must offer may be seen in this court's opinion in International Harvester Co. v. Ruckelshaus, 478 F.2d 615 (D.C.Cir.1973). That case, despite numerous dissimilarities to the present one, provides a useful point of reference, and all the parties seek to claim it as their own. In International Harvester, the court reversed the EPA's refusal to suspend for one year strict new 1975 model year emission standards that had been set by Congress in the 1970 amendments. This court, reviewing in early 1973 an EPA decision of May 1972, stressed the harm that would result from "a relaxation of standards, and promulgation of an interim standard, at a later hour after the base hour for 'lead time' has been passed, and the production sequence set in motion." Id. at 637. Too late a relaxation would penalize technologically advanced firms, like Ford, which would already have begun manufacture of vehicles that achieved better emission control at the expense of road performance. For this and other reasons,17 the hardship resulting if a suspension were mistakenly denied outweighed the risks from a suspension needlessly granted. Because of that balance of hardships, the court probed deeply into the reliability of the EPA's methodology.18 The present case is quite different; the "base hour" for commencement of production is relatively distant, and until that time the probable effect of a relaxation of the standard would be to mitigate the consequences of any excessive strictness in the initial rule, not to create new hardships.

The significance of the time factor in International Harvester was increased by the fact that the EPA was not predicting future technological advances, but rather was imposing an interpretation on current industry data. That data uniformly indicated that the standards were not being met, yet the EPA claimed that "adjustments" of the data demonstrated the likelihood of compliance. But the court concluded that the agency had failed to demonstrate the reliability of its methodology sufficiently to defend its reinterpretation of apparently adverse data.19

International Harvester has been cited frequently in cases involving presently-available-technology standards, as well as in other cases in which the agency's "central argument is that the standard is achievable because it has been achieved," National Lime Association, 627 F.2d at 432-33 (emphasis in original); Bunker Hill Co. v. EPA, 572 F.2d 1286 (9th Cir. 1977); Duquesne Light Co. v. EPA, 522 F.2d 1186 (3d Cir. 1975), vacated, 427 U.S. 902, 96 S. Ct. 3185, 49 L. Ed. 2d 1196 (1976); CPC International, Inc. v. Train, 515 F.2d 1032 (8th Cir. 1975); Portland Cement Ass'n v. Ruckelshaus, 486 F.2d 375 (D.C.Cir.1973), cert. denied, 417 U.S. 921, 94 S. Ct. 2628, 41 L. Ed. 2d 226 (1974). The defense of a projection methodology in such cases has required "that variables be accounted for, that the representativeness of test conditions b(e) ascertained, that the validity of tests be assured and the statistical significance of results determined." National Lime Association, 627 F.2d at 452-53 (footnotes omitted). But statistically-based techniques for reviewing the methodology of contemporary projections do not translate well into rules for reviewing predictions of future progress. If the agency is to predict more than the results of merely assembling preexisting components, it must have some leeway to deduce results that are not represented by present data.

The EPA has generally been granted "considerable latitude in extrapolating from today's technology" when it predicts future technological developments for the purposes of the Clean Water Act, 33 U.S.C. §§ 1251-1376 (1976 & Supp. II 1978). See California & Hawaiian Sugar Co. v. EPA, 553 F.2d 280, 288 (2d Cir. 1977). The courts have had numerous occasions to review EPA determinations that a given control technique constitutes the "best available technology economically achievable" in the 1980s.20 Most of the opinions, including our own American Paper Institute v. Train, 543 F.2d 328, 352-53 (D.C.Cir.1976), cert. dismissed, 429 U.S. 967, 97 S. Ct. 398, 50 L. Ed. 2d 335 (1976), steer close by the shores of their factual contexts and yield little in the way of explicit doctrine. But their essential requirement is that the agency provide "a reasonable basis for belief that a new technology will be available and economically achievable."21 Hooker Chemicals & Plastics Corp. v. Train, 537 F.2d 620, 635 (2d Cir. 1976). When a technology is already in use in other industries, the court often expects more solid evidence that the technology can be transferred to the industry in question, or at least that relevant dissimilarities have been considered. American Meat Institute v. EPA, 526 F.2d 442, 465 (7th Cir. 1975).

To apply these general considerations to our task of review in the present case, we must examine the nature of the EPA's determination. The agency has predicted that the manufacturers will be able to develop a satisfactory version of the trap-oxidizer in the time remaining. This device was designed specifically for the purpose for which EPA intends it, and prototypes have achieved partial success. GM itself has characterized trap-oxidizers as "(t) he most promising particulate traps," and has admitted that "current program status (would) indicate a possibility of 1985 model year production."22 General Motors Response to EPA Notice of Proposed Rulemaking 132, 175 (1979) (hereinafter GM Response), J.A. 279, 284. The EPA's decision must be judged in terms of record evidence available in early 1980, allowing a "time frame of 2-21/2 years for completion of the design development phase (and) 2-21/2 years of production lead time."23 45 Fed.Reg. 48,133, 48,139 (1980).

Given this time frame, we feel that there is substantial room for deference to the EPA's expertise in projecting the likely course of development. The essential question in this case is the pace of that development, and absent a revolution in the study of industry, defense of such a projection can never possess the inescapable logic of a mathematical deduction. We think that the EPA will have demonstrated the reasonableness of its basis for prediction if it answers any theoretical objections to the trap-oxidizer method, identifies the major steps necessary in refinement of the device, and offers plausible reasons for believing that each of those steps can be completed in the time available. If the agency can make this showing, then we cannot say that its determination was the result of crystal ball inquiry, or that it neglected its duty of reasoned decisionmaking.

2. The Time "Necessary to Permit the Development and Application of the Requisite Technology"

Applying the standard described in the preceding section to the challenged particulate regulations, we can determine whether the EPA has presented an adequate exposition of its reasons for believing that the necessary technology will be available for 1985 model year light-duty diesels to comply with the standard. The EPA bases its prediction that the 1985 standard will be achieved on two factors: modifications decreasing the particulate output of diesel engines,24 and development of "aftertreatment" technology, that is, means by which the vehicle will remove particulate matter from its own exhaust. The larger proportion of the expected reduction in particulate emissions depends on aftertreatment, and it is the availability of that technology that provokes the major controversy in this case.

The EPA has identified a number of strategies for extracting particulates from diesel exhaust,25 but the 1985 standard was set in reliance on one preferred method and must stand or fall with the agency's prediction that that method will be available in time. This favored device is the trap-oxidizer, a mechanism that filters out particulates and then periodically incinerates its catch in order to maintain the trapping capacity of the filter.

The trap-oxidizer is essentially a compromise between two other particulate reduction strategies. At one extreme, the vehicle could rely on a mechanical filter alone but unless that filter were somehow able to maintain its trapping efficiency indefinitely, it would periodically need either replacement or cleaning. At the other extreme, particulates could be continuously incinerated in a catalytic converter but difficult engineering problems accompany the resulting need to maintain sufficient temperature in the converter (c. 1000o F.) and to keep the particulate matter inside the converter long enough to be burned. Citing the technical barriers to continuous incineration and the behavioral barriers to periodic restoration of a filter by car owners, the EPA recognized the trap-oxidizer as "the preferred method" of particulate control. Regulatory Analysis at 47-50, J.A. 526-29. The trap-oxidizer combines the short-term technical superiority of a filter with the long-term usefulness of a converter.

The EPA has predicted that trap-oxidizers will be available for use in model year 1985 vehicles. As the agency has repeatedly observed, the trap-oxidizer is familiar and unobjectionable as a concept. It is not only theoretically sound experimental data demonstrate that periodic incineration can maintain efficiency for over 10,000 miles. But to date, no filter material has been found that can withstand periodic incineration of the accumulated particulates throughout the 50,000-mile useful life of the vehicle26 while maintaining a high level of trapping efficiency. The agency noted that

the best durability of a trap reported to EPA was a metal mesh trap on an Opel vehicle, run on a modified AMA driving schedule with no hard accelerations, hills, or speeds above 45 mph. The trap survived 12,800 miles and at that time had a collection efficiency similar to its zero-mile efficiency of 55 percent.

Regulatory Analysis at 51, J.A. 530. Understandably, the EPA has concluded that further research is needed before devices with the appropriate characteristics will be available for use:

Clearly, more basic research still needs to be done in the areas of regeneration initiation and control, and trap durability. Enough progress has been achieved to convince EPA that a successful trap-oxidizer can be developed, but as of this time, no design has proven to have the required collection efficiency over the desired length of time.

Nevertheless, the agency concludes that it is merely a question of time before the trap-oxidizer is perfected. "The improvements that are necessary are engineering problems, and are more a function of the resources allocated to the problem than any scientific or technical breakthrough." 45 Fed.Reg. 14,496, 14,498 (1980). Based on the routine nature of most of the remaining problems, the rapid pace of progress in the field since 1978, and the industry's own forecasts of 1985 as a potential completion date, the agency has determined that the lead time remaining is sufficient for application of the requisite technology.

GM dismisses the agency's conclusion as baseless speculation and charges the EPA with naive optimism about the solution of myriad uncertainties, ranging from the development of a durable filter material to the proper location of the trap on the vehicle itself, see GM Brief at 28. GM regards the gaps in present knowledge as vitiating the entire standard-setting endeavor:

Until further experimental knowledge on these major development needs is obtained, it is totally impossible to specify when a successful system will be developed for passenger cars and light-duty trucks. Thus, any particulate standard which contemplates use of a regenerative trap-oxidizer must be judged premature and not technologically feasible.

GM Petition for Reconsideration of Standard at 8 (hereinafter GM Petition) J.A. 812. Thus, GM believes that no standard can be promulgated, regardless of its effective date, on the current record.

Before analyzing GM's technical objections, we must reiterate the standard of review that governs this case. The EPA is not obliged to provide detailed solutions to every engineering problem posed in the perfection of the trap-oxidizer. In the absence of theoretical objections to the technology, the agency need only identify the major steps necessary for development of the device, and give plausible reasons for its belief that the industry will be able to solve those problems in the time remaining. The EPA is not required to rebut all speculation that unspecified factors may hinder "real world" emission control.

The EPA has identified as the necessary remaining steps in development of trap-oxidizer technology the choice of a durable, efficient filter material, the selection of an incineration method, and the refinement of a control mechanism to bring about automatic initiation of the regeneration process. GM agrees with the agency's specification of these aspects of the trap-oxidizer as the ones requiring further research. GM Brief at 22 ("Three critical issues concerning trap oxidizer feasibility remain unresolved: trap durability, regeneration, and collection efficiency.").

a. Development of a Durable Filter

The most vigorously controverted issue in this case concerns durability, which the EPA has recognized as the key remaining problem. "Collection efficiencies and regeneration techniques have progressed to the point where the most critical issue is whether the efficiency and regeneration mechanism can be maintained over the useful life of the vehicle." 45 Fed.Reg. 14,496, 14,498 (1980). Both parties focus on the GM Opel test, in which a metal mesh trap survived over 12,000 miles; the EPA views this test as a stirring demonstration of how far research had progressed in the brief springtime of trap-oxidizer research, while GM sees the 12,000-mile mark as insubstantial compared to the 50,000- to 100,000-mile goal.

The EPA has predicted that the necessary work can be accomplished in time for 1985 model year production. The agency points to the wide variety of materials that have demonstrated appropriate initial efficiencies; several of these are hybrids,28 suggesting that new combinations of present candidates, rather than hitherto untested substances, may provide the answer. The EPA noted the rapid pace of achievement since attention turned to trap-oxidizers:

Considerable progress has occurred in the last 11/2 years. Initial aftertreatment research concentrated on the use of simple traps and gasoline engine catalytic converters, with basic problems of efficiency and backpressure. In the last 11/2 years we have seen marked improvements in efficiency and backpressure, and more importantly, a general consensus that the trap-oxidizer can periodically (and possibly even continually) incinerate the particulate matter. Methods for regeneration initiation and control have been investigated and repetitive incineration has been demonstrated for several trap-oxidizers.

45 Fed.Reg. 14,496, 14,498 (1980). The small number of successful tests reflected the recent focus on trap-oxidizers, not a persistent record of failure: "At this time, EPA has limited trap-oxidizer durability data, as researchers have been reluctant to fund durability testing until other, more basic questions were solved." Id.29 Finally, the manufacturers themselves had projected the possibility of introducing trap-oxidizers in 1985 models. See Summary and Analysis of Comments on the Notice of Proposed Rulemaking at 40-41 (1979) (hereinafter cited as Analysis of Comments), J.A. 358-59. The agency fully recognized that no durable, efficient filter material was presently available, but concluded that, based on current progress and past achievements, there were sufficient grounds for believing that one would be developed in time for the 1985 standard.

We conclude that these are plausible reasons for a determination that the industry is capable of solving the durability problem in the allotted time. The EPA could reasonably refuse to be discouraged by the limited initial success, as the project is relatively young. The rapidity of recent progress is a factor that the agency may consider in making a prediction of future capabilities. See Society of the Plastics Industry, Inc. v. OSHA, 509 F.2d 1301, 1309 (2d Cir. 1975), cert. denied, 421 U.S. 992, 95 S. Ct. 1998, 44 L. Ed. 2d 482 (1975). And the industry's own predictions, while not determinative, support the view that success in this kind of research can realistically be expected within the proposed time frame. We conclude that the EPA's durability prediction, though uncertain, is no more uncertain than such estimates inherently must be, and that the EPA has met the requirement of "reasoned decisionmaking."

b. Efficiency

We agree with GM's assertion that the very concept of efficiency "is virtually meaningless in the absence of acceptable durability and regenerative capacity." GM Reply Brief at 17. Rather than viewing efficiency as an independent factor that the EPA has failed to analyze adequately, we believe that the efficiency question must be seen as an aspect of the durability problem just considered. The variety of filter elements with sufficient initial efficiency is apparent from the record, and played an important role in the agency's prediction that a filter combining both durability and efficiency would be developed in the next few years.

We also observe, however, that efficiency differs from durability and regenerability in that the EPA's ability to make minor adjustments to the efficiency factor affords a safety valve for the standard. If a durable, regenerable trap-oxidizer is developed, but is found only to achieve an average efficiency of fifty-five percent, then most diesels will be in compliance with the standard, and only a small reduction in its stringency would be necessary to permit even the worst polluters to comply. If the research falls short on durability or regenerability, in contrast, the standard would have to be radically changed for all but the smallest diesels.

We find no fatal defect in the EPA's prediction that a trap-oxidizer with the requisite efficiency will be developed, given our resolution of the durability issue.

c. Regeneration Initiation and Control

The EPA has pointed to two viable means of incinerating accumulated particulates: throttling, which increases the heat of the engine exhaust itself by reducing air intake, and ignition from a separate heating element. Regulatory Analysis at 50, J.A. 529. In either case, use of a noble metal catalyst facilitates burning at lower temperatures, lessening the additional increment of heating required. Id. For systems relying on an independent heat element, dividing the trap into two chambers permits a further reduction in the amount of power required. Id. Throttling has been demonstrated as an effective method for incinerating particulates, and the EPA saw most promise in a system relying on throttling and a catalyst, plus insulation of the exhaust system to reduce heat loss. Id. at 51, J.A. 530. GM achieved good results with throttling alone, at least at speeds over thirty-five mph. GM Response at 110-13, J.A. 262-65.

GM complains of the excessive electric demand created by a heating element; as the EPA observed, that demand is considerably reduced by using a dual path trap, insulating the exhaust system, and introducing a noble metal catalyst. Regulatory Analysis at 50, J.A. 529. GM next seeks to characterize air intake throttling as unsafe and unreliable, despite the success of GM's own tests. GM emphasizes the failures in an EPA test conducted in 1979: the EPA investigators created temperatures in excess of 1100o F., and eventually discovered that they had blown a hole through their trap. J.A. 729-35. This anecdote does not demonstrate that proper control of the throttling process is unfeasible, particularly if a catalyst is used to lower the necessary temperature.

For both throttling and ignition approaches, some automatic method of initiating incineration is required. A simple mechanical device tied to the odometer is conceivable, but more subtle control can be achieved with microelectronics.

GM insists that the agency has no basis for believing that a control mechanism for initiating and regulating the incineration process can be developed. This argument is without merit. GM's own prototype throttling vehicle "utilize(d) a microprocessor controller to set the position of the throttle in the air intake as a function of engine speed and rack angle." GM Response at 110, J.A. 262. GM states no theoretical objection to automated control of the incineration process and does not deny that the necessary computer technology exists. GM merely points to the EPA's admission that "(i)t is impossible at this time to delineate the exact design" of the control unit, Regulatory Analysis at 51, J.A. 530, and insists that "the relevant parameters cannot be delineated unless and until a final design is selected." GM Reply Brief at 17 (emphasis in original). Nothing in the record suggests that adaptation of preexisting computer technology to the specific design of a GM trap-oxidizer will be anything other than routine, see Denial of Petition for Reconsideration, 45 Fed.Reg. 48,133, 48,137 (1980), and the EPA is not obliged to establish that no unknown parameters will later prove relevant to proper control.30

d. Conclusion

In summary, we find sufficient support for the EPA's necessarily predictive judgment and therefore uphold the EPA's particulate standard. The agency has given the manufacturers substantial lead time, and there is room for interim adjustments to the standard without significant hardship. Under those circumstances, the applicable standard of review allows the EPA considerable latitude to exercise its expertise through reasoned projections. We find that the agency has given an adequate explanation of its reasons for believing that the necessary steps in improving trap-oxidizer technology can be completed in the time remaining.31

C. NRDC's Objections to the Particulate Standards

The NRDC, in contrast to GM, thinks that the EPA's standards for diesel particulates are too lax, and attacks a number of the decisions in their regulatory genesis. We find no merit in the NRDC's objections.

The NRDC's first attack is on the regulatory date, relying on the EPA's suggestion that "there is a strong likelihood that trap-oxidizers will be feasible for vehicle application by 1984," Regulatory Analysis at 53, J.A. 532. In view of the uncertainty remaining as to whether even the 1985 date is realistic, we defer to the EPA's determination that a 1984 date would leave too little margin for error.

The second objection is to the EPA's consideration of cost in standard setting. The NRDC notes the EPA's estimate that "(e)ven with the particulate regulations, the overall cost of pollution control from diesels should be less than that from gasoline engines." Id. at 120-22, J.A. 599-601. The NRDC argues that the carcinogenic character of diesel particulates makes diesel vehicles so dangerous that "appropriate consideration" to cost would require the EPA to discourage purchase of diesels by making them more expensive. Emission standards should be tightened until the cost of compliance by diesels exceeds the cost of compliance by gasoline vehicles. But the correlation between scientific achievement and research investment is not so mechanical that the EPA could set a dollar figure and then compute the level of emission control it would buy. The agency has projected the probable course of future advances in particulate reduction to the best of its ability, assuming a reasonable range of research investment. The statute requires no more.

The NRDC also charges that the EPA "ignored" the risk of cancer posed by particulates. Even a glance at the record refutes the suggestion that the EPA was unaware of the risk, or irresponsibly discounted it.32 A more temperate rephrasing of this claim is that the regulatory approach would have been stricter if the EPA had focused more strongly on the carcinogenic potential of diesel emissions. In this form, the claim becomes another facet of the NRDC's attack on the choice of worst-performing diesel as the measure of technological feasibility.

Although we have held that the proper statutory basis of the EPA's power to issue particulate standards for light-duty vehicles is the general authority of section 202(a)(1), see Part II(A) supra, the EPA viewed itself as bound by the explicit mandate of section 202(a)(3)(A)(iii) to require "the greatest degree of emission reduction achievable." See Regulatory Analysis at 31, J.A. 510. Both provisions call for a determination that the technology needed for compliance will be available when the standards take effect. Assuming arguendo that appropriate consideration of the carcinogenic potential of diesel particulates would have required standards reflecting the greatest degree of reduction achievable, we turn to the NRDC's claim that the worst-performing-diesel standard does not comply with the directive of section 202(a)(3)(A)(iii).

The EPA defends its decision to promulgate a single particulate standard applicable to all light-duty diesel vehicles, rather than a number of standards applicable to various classes of diesels. One reason given is the absence of any apparent parameter on which to base such a classification.33 Another concern is the paradoxical result that graduated standards, by imposing stricter and more costly requirements on low-polluting vehicles, would skew competition in favor of the more loosely regulated, highly polluting models. Finally, Congress itself has taught this method by example, providing unitary emission standards for all passenger vehicles in both the 1970 and 1977 amendments. We conclude that the adoption of a single particulate standard for light-duty diesel vehicles was within the EPA's regulatory discretion.

Even if a single standard is proper, the NRDC asserts, it ought not to be one that every diesel can meet. The NRDC cites a favorable passage in International Harvester, in which this court agreed with the Administrator that the Act's technological feasibility provisions required only that the standards permit basic auto demand to be met, "even though this might occasion fewer models and a more limited choice of engine types." 478 F.2d at 640. In this case, however, the EPA viewed the legislative history of the 1977 amendments as justifying a more tolerant attitude toward diesel development in the near future. See text at note 36 infra. The availability of waivers of the 1.0 gpm oxides of nitrogen standard for "any class or category" of diesel vehicles that requires them, Act § 202(b)(6)(B) (emphasis added), certainly suggests an interest in exploring the full range of potential benefits of diesel technology. We do not hold that the statute mandates such an approach, but it was well within the EPA's regulatory discretion to impose "standards which provide significant particulate reductions, but which do not force any diesel models out of production." Regulatory Analysis at 32, J.A. 511.

As we have had occasion to observe, some light-duty trucks come within the "heavy-duty vehicle" category for which particulate emissions standards are authorized by section 202(a)(3)(A)(iii), while others remain in a residual category, neither heavy-duty vehicles nor light-duty vehicles, governed by section 202(a)(1). See note 3 supra. This distinction does not affect our analysis of the present challenges to the particulate standard for light-duty trucks, however, because petitioners' objections are too general to implicate the varying nuances of the separate statutory provisions. NRDC essentially repeats its claims against the light-duty vehicle standard, while GM's attack focuses on the adequacy of the support in the record.

The EPA originally proposed the same particulate standards for light-duty trucks and light-duty vehicles, for both 1981 and 1983, see 44 Fed.Reg. 6650 (1979). The agency subsequently explained this proposal as reflecting the frequent congruity between light-duty truck and light-duty vehicle emission control:

It has been established in previous EPA rulemakings that manufacturers usually apply passenger car emission control technologies to light-duty trucks in order to comply with similar standards, since the engine configurations and type of use are very similar. For instance, GM's diesel light-duty trucks utilize the same diesel engines that are used in the GM 4,500 pound light-duty vehicles.

45 Fed.Reg. 48,133, 48,138 (1980) (footnotes omitted). The meager relevant data submitted in comments on the proposed rulemaking, however, suggested that light-duty trucks emit substantially more particulates than do passenger vehicles. See Regulatory Analysis at 53-58, J.A. 532-37.

The EPA analyzed these data, and concluded that the "higher inertia weight and aerodynamic drag" of light-duty trucks would necessarily result in higher particulate levels. 45 Fed.Reg. 14,496, 14,497 (1980). The higher road load horsepower of the trucks also contributed to greater emission levels. See 45 Fed.Reg. 48,133, 48,138 (1980); Regulatory Analysis at 54-55, J.A. 533-34. Furthermore, the expected improvements in light-duty vehicle emissions due to downsizing of vehicles and their engines would not be equalled by light-duty trucks. Regulatory Analysis at 58, J.A. 537. Finally, the increasing stringency over the next decade of the oxides of nitrogen standard for light-duty trucks would exacerbate particulate emissions.34 Id. at 57, J.A. 536. Taking into account all these factors, the EPA concluded that the particulate emissions from light-duty trucks, before aftertreatment, would be thirty percent greater than those from light-duty vehicles. Therefore, the appropriate particulate standard, after trap-oxidizer treatment, would be 0.26 gpm rather than 0.20 gpm. 45 Fed.Reg. 14,496, 14,497 (1980).

The EPA's adjustments to the particulate standard were based on actual industry data, not theoretical projections of the effect of the measured factors on particulate emissions. GM offers an alternative explanation of the empirical observations, claiming that truck exhaust flow is increased by the different operating conditions of truck-type service, and faults the EPA for not including GM's hypothesized factor in its analysis. GM presents no data suggesting that inclusion of this additional factor would have resulted in a different extrapolation from the current observed emissions, and presented none at the time of its petition to the agency for reconsideration, see GM Petition, J.A. 805-16. We perceive no justification in GM's conclusory allegations for condemning as unreasoned the EPA's attempt to determine the relevant differences between light-duty trucks and passenger vehicles.

GM also attacks as unduly speculative the agency's assumption that trap-oxidizer technology developed for light-duty vehicles can be applied to light-duty trucks with only minor adaptations. Again, GM offers no specific reason for believing that particulate emission control devices will be any less suitable for light-duty trucks than previous emission control devices have been. GM's own pessimistic speculations cannot be conclusively refuted until the trap-oxidizer is completed, but as we have stated before, the mere possibility that unforeseen complications will develop does not render a prediction unreasonable.35

Finally, NRDC's challenge to the particulate standard for light-duty trucks is more candidly a variation on its objections to the light-duty vehicle standard. NRDC denies that a standard based on the worst performing diesel can accomplish the "greatest degree of emission reduction achievable" within the meaning of section 202(a)(3)(A)(iii). We have already upheld the worst-performing-diesel standard for light-duty vehicles as within the EPA's regulatory discretion, and we find that the similar standard for light-duty trucks is equally so.

The NRDC also challenges the EPA's granting of waivers of the standards for oxides of nitrogen (NOx) to various manufacturers. The NRDC argues that the EPA did not properly interpret or apply the statutory prerequisites that must be satisfied before such waivers may be granted. We uphold the EPA's decisions.

The NOx waivers at issue are temporary relaxations of the statutory NOx standard of 1.0 gpm for individual light-duty diesel vehicle models that are unable to achieve that standard. The EPA granted waivers permitting emissions up to 1.5 gpm for the 1981 and 1982 model years for certain GM, Daimler-Benz, Volvo, and Peugeot models, and waivers at levels varying from 1.3 to 1.5 gpm for certain Volkswagen diesels. 45 Fed.Reg. 5480 (1980); 45 Fed.Reg. 34,71 9 (1980). The EPA's basic rationale in granting the waivers was that looser NOx standards were necessary to permit compliance with the new particulate standards.

The waiver provisions of the Clean Air Act Amendments of 1977 reflect Congress' concern that pollution control standards tailored to the virtues and vices of the traditional gasoline engine might stifle the development of innovative alternative technologies. Newer systems that could eventually offer distinct advantages might require a period of indulgence before demonstrating their full potential. This concern focused particularly on oxides of nitrogen, which were known to present a serious challenge for the diesel engine. See, e. g., H.R.Rep.No. 294, 95th Cong., 1st Sess. 249-50 (1977); 123 Cong.Rec. 16,92 4 (1977) (remarks of Rep. Dingell); 123 Cong.Rec. 18,184 (1977) (remarks of Sen. Baker).

The present NOx waiver provision for diesels, section 202(b)(6)(B) of the Act, emerged in conference, splitting a separate diesel waiver off from the general waiver for an "innovative power train technology or innovative emission control device or system."36 The duration of the waiver was limited to a maximum of four years, and a list of prerequisites to granting the waiver was added:

Such waiver may be granted if the Administrator determines

(i) that such waiver will not endanger public health,

(ii) that such waiver will result in significant fuel savings at least equal to the fuel economy standard applicable in each year under the Energy Policy and Conservation Act, and

(iii) that the technology has a potential for long-term air quality benefit and has the potential to meet or exceed the average fuel economy standard applicable under the Energy Policy and Conservation Act at the expiration of the waiver.

Act § 202(b)(6)(B). The House bill had contained a similar public health requirement,37 but the "air quality benefit" criterion was wholly new.

In its guidelines for waiver applications, the EPA set out its interpretation of these criteria, and listed in detail the information that manufacturers would have to submit in order to meet their burden of demonstrating that waivers should be granted. 43 Fed.Reg. 30341 (1978). The January 1980 announcement of waivers for GM, Daimler-Benz, and Volvo was accompanied by a rejection of Peugeot and Volkswagen applications due to insufficiency of information;38 after more data were submitted, those waivers were granted in May 1980. 45 Fed.Reg. 5480 (1980); 45 Fed.Reg. 34,719 (1980). Although all manufacturers had requested four-year waivers, the EPA concluded that the data only justified waivers for 1981 and 1982.

The NRDC argues that these waivers should never have been granted because diesel technology cannot meet two of the statutory criteria for the waiver: the public health standard and the air quality benefit standard. The NRDC does not dispute the EPA's findings that waivers are necessary for certain vehicles. Before scrutinizing these claims, it is important to note our agreement with a position that the EPA has emphasized throughout these proceedings. The NO x waiver process was not intended as an opportunity to decide the ultimate fate of diesel technology. It was meant to offer temporary encouragement to further development of diesels, so long as the public interest would not be unduly endangered. There will be other opportunities to evaluate the costs and benefits of that technology as more becomes known about it. In announcing the NO x waivers, the EPA repeatedly stressed that "this decision does not represent an open door to future production of light-duty diesels without great emphasis on developing methods to identify, characterize and control diesel-related emissions." And the agency cautioned manufacturers "to proceed with the full realization that EPA has authority to act in the future under other provisions of the Act to prevent diesel emissions from posing an unacceptable risk to public health."39 45 Fed.Reg. 5480, 5492-93 (1980).

A. The Waiver "Will Not Endanger Public Health"

The public health criterion for the NO x waivers does not require an evaluation of the long-term public health impact of diesel technology. It requires a finding "that such waiver will not endanger public health." Act § 202(b)(6)(B)(i) (emphasis added). A major focus is thus on the character of emissions during the period of the waiver. Furthermore, the EPA found it likely that diesel vehicles could be marketed even if no NO x waivers were granted. Given the industry's commitment of resources to diesel production, the EPA believed that the original NO x standard might be met using exhaust gas recirculation, which reduces NO x but aggravates other polluting characteristics of diesel engines.

The EPA rejected the NRDC's claim that NO x waivers would endanger the public health because of increased NO x emissions. After analyzing the relevant data, the agency concluded that diesel NO x waivers would slow the expected decreases in ambient NO x levels over the next decade, but that this impact "would not be significant." 45 Fed.Reg. 5480, 5488-89 (1980). We agree that Congress did not intend to preclude NO x waivers merely upon a showing that some impact on NO x levels would result. We cannot agree with the NRDC's contention that the EPA's determination that the effect was de minimis was "clearly erroneous," Brief for Petitioner NRDC at 59 n.93 (hereinafter cited as NRDC Brief).

The EPA's public health evaluation also addressed the effect of the NO x waivers on emissions of other pollutants. This factor weighed in favor of granting the waivers because the technology used for NO x reduction increases emissions of several other harmful substances. 45 Fed.Reg. 5480, 5489 (1980). Particulates were an object of special concern, since greatly increased particulate emissions are a major distinguishing feature of diesel vehicles.

The EPA concluded that, given current technology, there was necessarily a trade-off between NO x emissions and particulate emissions. The NRDC opposed this view, but its arguments before the agency failed to take into account the fact that particulate standards under section 202(a) must be "technology-based." It therefore insisted that the NO x waivers would increase particulate emissions by increasing sales of diesel vehicles. "For any given particulate standard, relaxing the NO x standard would lower the cost of producing diesel-powered cars." NRDC Comments on NO x Waiver Applications at 7 (1979), J.A. 1067. This assertion overlooks the fact that no "given particulate standard" can be assumed until the NO x level has been determined, for as we have recognized above, the applicable standards for other pollutants help determine the particulate levels that available technology can achieve. In setting the 0.6 gpm particulate standard for 1982-1984, the EPA repeatedly observed that it was assuming a 1.5 gpm NO x standard. Regulatory Analysis at 33, J.A. 512; 45 Fed.Reg. 14,496, 14,497 (1980).

It was perfectly proper for the EPA to bear in mind this trade-off in evaluating the public health impact of the NO x waivers. The NRDC itself has stressed the respiratory and carcinogenic dangers of particulates, and we cannot say that the EPA was arbitrary in concluding that the benefits of reducing particulate levels outweighed the dangers of slightly increased NO x levels.

On appeal, the NRDC makes a somewhat different argument that even if the particulate level will be raised by denial of the NO x waiver, the total particulate emissions nationwide from the fleet of diesel vehicles would still be less than if waivers had been granted. NRDC estimates that the marginal cost of achieving the 1.0 gpm NO x standard would increase the price of diesels, and thereby lower the demand for them, to so great an extent that the total fleet emissions of particulate would be decreased by denial of the waiver, even though individual vehicle particulate emissions would be greater. NRDC Brief at 61. NRDC claims that if the EPA had properly calculated this effect of the waiver decision it would have concluded that the public health requires that the waiver be denied. But, as the EPA's brief observes, the NRDC has never offered factual support for this contention. Brief for Respondent EPA at 119 n.111. Certainly the record does not compel NRDC's conclusion. Data submitted by the manufacturers, though not adjusted for this marginal cost argument, indicated substantial increases in total particulate levels if the waivers were denied.40 We cannot say that the cost of an exhaust gas recirculation system would necessarily dampen diesel demand enough to reverse this trend.

B. The Technology "Has a Potential for Long-Term Air Quality Benefit"

The meaning of the "long-term air quality benefit" criterion is less self-evident than that of the public health requirement. This prerequisite to the diesel waivers was contained in neither the House nor the Senate bill, but emerged in conference. The sole elaboration of its content is found in an explanation of the conference's efforts made to the Senate by the bill's sponsor, Senator Muskie:

And finally, in order to assure that the waiver will be targeted towards long term air quality benefits, the waivers may only be granted if there is a substantial likelihood that the engines will be able to comply at the end of the waiver period with the statutory emissions standards. Thus the waiver is not intended to provide a loophole in the statutory standards, but only to provide an opportunity for technology development which may lead to greatly improved emissions performance in the next decade.

123 Cong.Rec. 26,848 (1977). This provision represented a compromise between the Senate bill, which allowed only a two-year waiver of the 1.0 gpm NO x standard for innovative technologies, and the House bill, which provided indefinite waivers for four or more years as long as the public health would not be endangered.41

Relying on this legislative history, the EPA interprets the air quality benefit criterion as requiring only that there be "a substantial likelihood that the engines will be able to comply with the statutory standards applicable at the expiration of the waiver period." 45 Fed.Reg. 5480, 5493 (1980). As long as diesel vehicles have the potential for meeting those standards, no loophole has been opened and the statute is satisfied.

This is not the only possible interpretation of the air quality benefit provision, but it is not an implausible one. "Where different interpretations of the statute are plausible, so long as EPA's construction of the statute is reasonable we may not substitute our own interpretation for the Agency's." Lead Industries Ass'n v. EPA, 647 F.2d 1130 at 1147-1148 (D.C.Cir.1980), cert. denied, --- U.S. ----, 101 S. Ct. 621, 66 L. Ed. 2d 503 (1980). We decline the NRDC's invitation to hold that no waiver may be granted unless diesel emissions are found to be harmless.

The NRDC finds the EPA's interpretation illogical, asserting that its application to the particulate standards is totally circular. The NO x waivers are being justified by the ability of diesels to meet particulate standards that are themselves defined so that every diesel can meet those standards if given a NO x waiver. The NRDC regards it as absurd to see air quality benefit in compliance with standards so lacking in content.

It is true that there is a certain element of paradox in regarding achievement of technology-based standards as demonstrating a benefit. But this paradox is not an inevitable corollary of the EPA's interpretation of section 202(a)(6)(B) (iii). The effective circularity of this criterion as applied to particulates results specifically from the EPA's choice of the worst-performing diesel as the measure of particulate performance. As we have held, that choice is a permissible regulatory strategy but is in no way mandated by the statute. See Part II(C) supra. The air quality benefit criterion will be more substantial when applied to pollutants regulated under other schemes.

Finally, the NRDC complains that the EPA misconstrued Senator Muskie's reference to "statutory emissions standards" as implicating only the explicit numerical standards for hydrocarbons, carbon monoxide, and NO x, instead of including other emissions standards set by the EPA under its statutory authority. This claim is unfounded. The portion of the EPA's guidelines for waiver applications that relates to the air quality benefit criterion requires information concerning particulates and other unregulated pollutants, see 43 Fed.Reg. 30,341, 30,346-47 (1978), and the May 1980 waiver decisions explicitly recognized that the standards applicable at the end of the waiver period include the 0.6 gpm particulate standard, see 45 Fed.Reg. 34,719, 34,723-24 (1980). The first waiver decisions, in January 1980, did not mention this standard, but this was understandable, as it had not yet been promulgated.

In conclusion, we find no merit in the NRDC's allegations that the EPA granted NO x waivers unlawfully, at undue risk to the public health. The agency's actions were consistent with a reasonable interpretation of the statutory provisions.

One small issue remains. At the same time that the EPA issued its particulate regulations, it also modified the test procedures for measuring hydrocarbon emissions from diesel vehicles. See 45 Fed.Reg. 14,496, 14,500 (1980). GM challenges that method.

Application of heat during the hydrocarbon measuring process results in the release in free, gaseous form of hydrocarbons that would at lower temperatures have remained adsorbed to the surface of particulates. The EPA's hydrocarbon test procedure counts these as hydrocarbons; the particulate test counts them as particulates. GM objects to measurement of these emissions as hydrocarbons, asserting that they do not contribute to air pollution problems, and that GM is a victim of double counting.

The EPA rejected GM's complaints on several grounds; we agree with all of these. First, the statute requires the agency to measure total hydrocarbon emissions, regardless of the reactive potential of individual hydrocarbon compounds. 45 Fed.Reg. 48,133, 48,140 (1980). This court has already upheld the EPA's attention to total hydrocarbon emissions as a valid administrative interpretation of the Act, even in the context of admittedly nonreactive hydrocarbons. Ford Motor Co. v. EPA, 604 F.2d 685 (D.C.Cir.1979). Second, the EPA found GM's test data exonerating adsorbed hydrocarbons unconvincing, because "the tests performed were unrealistic in simulating atmospheric conditions." 45 Fed.Reg. 14,496, 14,505 (1980). The EPA thus regarded it as an open question whether particle-bound hydrocarbons contribute first to the particulate problem and then later to the hydrocarbon pollution problem as well. Analysis of Comments at 132, J.A. 454.

Finally, despite GM's characterization of the test as "double counting," GM has not shown that it is prejudiced by inclusion of adsorbed hydrocarbons in the particulate measurements. Adsorbed hydrocarbons are included in determining compliance with the absolute statutory hydrocarbon emission standard, and so their measurement is significant. But the particulate standards are technology-based, and therefore are set at a level that already includes any inflation due to adsorbed hydrocarbons. 45 Fed.Reg. 48,133, 48,140 n.76 (1980). We do not see any "fundamental unfairness for the vehicle manufacturers" in this test procedure.

In the Clean Air Act, Congress encouraged the EPA to set standards for the future without specifying the methodology the agency must follow to determine the probable course of future technological growth. In these circumstances, a reviewing court's role is to make sure that the agency has acted responsibly in formulating a reasoned prediction. It is not our task to decide whether the agency is correct, or to require proof to a mathematical certainty. We must be satisfied if the agency has undertaken its analysis with the degree of precision and clarity that the subject inherently permits. The EPA has done so in this case.

We find no merit in the NRDC's allegations that the EPA granted NO x waivers unlawfully, at undue risk to the public health. We similarly reject the NRDC's challenges to the particulate standards as inconsistent with the statutory mandate. We uphold the EPA's hydrocarbon testing procedure, and accept as sufficiently reasoned the EPA's prediction of technological availability and the particulate standards based thereupon. The regulations reviewed in this proceeding, in their entirety, are

Affirmed.

ROBB, Circuit Judge, concurring in part and dissenting in part:

I concur in Part III of the court's opinion, upholding the EPA's grant of NO x waivers to various diesel manufacturers. I must dissent, however, from that part of the opinion in which the court sustains the particulate standards for 1985. In my view the record does not support the EPA's prediction that the necessary technology will be available in time to meet the 1985 standard. My doubts focus in particular on the inability of the auto manufacturers to develop a filtering material for trap-oxidizers that possesses the durability needed to withstand periodic incineration of collected particulates for the useful life of the vehicle.

The majority states that "the EPA will have demonstrated the reasonableness of its basis for prediction if it answers any theoretical objections to the trap-oxidizer method, identifies the major steps necessary in refinement of the device, and offers plausible reasons for believing that each of those steps can be completed in the time available." Ante at 331. I have no quarrel with this standard, but I do not agree that the EPA has offered plausible reasons for believing that the critical step of achieving the required level of trap durability can be completed in the time available.

The record demonstrates that General Motors (GM) alone, which began its particulate control research program in 1974, has tested many different trap materials provided by at least 16 different manufacturers.1 Of these materials, 22 are characterized by GM as "the best materials" available,2 based on tests conducted with an Opel 2.1 liter diesel engine and an Oldsmobile 5.7 liter diesel engine. (J.A. 237) Yet the durability of even the best of these materials the metal mesh is far below what will be needed to meet the standard. The most successful test results were obtained by installing a metal mesh trap in a GM Opel, which was then driven on a non-typical schedule with no hard accelerations, hills, or speeds above 45 miles per hour. The filtering material survived only 12,800 miles, at which time it had a collection efficiency of 55 percent. GM also reported some particulate "blow-off" (i. e., particulate matter escaping through the exhaust system) and self-incineration. (J.A. 530) The statute, however, establishes a useful life for light duty vehicles and light duty vehicle engines of "five years or fifty thousand miles (or the equivalent), whichever first occurs." 42 U.S.C. § 7521(d)(1) (Supp. I 1977). Furthermore, the EPA stated in its Regulatory Analysis of the particulate standards that the trapping material "should last at least 100,000 miles." (J.A. 530)

After acknowledging the shortcomings of the Opel test, the EPA summarized the status of trap-oxidizer research as follows:

Clearly, more basic research still needs to be done in the areas of regeneration initiation and control, and trap durability. Enough progress has been achieved to convince EPA that a successful trap-oxidizer can be developed, but as of this time, no design has proven to have the required collection efficiency over the desired length of time. With the research that has been, and is, going on with regards to trap-oxidizer development, and a determined broad-based effort by the manufacturers to comply with the final standards, EPA's technical staff has concluded that it is very likely that a successful trap-oxidizer design can be optimized within the next 11/2 to 2 years.

(J.A. 531) In my opinion, these exhortations to the manufacturers and the EPA's vaguely articulated faith that a "design can be optimized" soon do not amount to "plausible reasons for believing that each of (the necessary) steps can be completed in the time available," within the meaning of the majority's standard. Pious hope and speculation cannot take the place of evidence. Accordingly, I must dissent from Part II of the court's opinion.