Richard Alan Kamp, Petitioner,the Environmental Defense Fund, Inc., Petitioner/intervenor, v. John Hernandez, in His Capacity As Acting Administrator Ofthe United States Environmental Protection Agency,and the Environmental Protection Agency,respondents,arizona Department of Health Services, Asarco, Incorporated,kennecott, Inspiration Consolidated Copper Co.,magma Copper Co., and Phelps Dodgecorp., Respondent/intervenors, 752 F.2d 1444 (9th Cir. 1985)

On January 14, 1983, the Environmental Protection Agency ("EPA") approved Arizona's plan for the control of sulphur dioxide (SO2) emissions from copper smelters. Richard Kamp challenges that approval on four grounds. He contends that Arizona's plan creates an excessive risk that emissions from the smelters will violate the national ambient air quality standards for SO2 ; that the plan improperly uses dispersion techniques instead of emission limitations to regulate SO2 emissions; that the attainment date set by the plan is unlawful; and that EPA should have promulgated regulations to control fugitive SO2 emissions before it approved Arizona's plan. We reject those challenges and affirm EPA's action.BACKGROUND

Copper smelters heat copper ore in order to separate the copper metal from other materials present in the ore. Because copper ore typically contains as much sulphur as it does copper, the smelting process generates as a by-product tremendous amounts of sulphur, most of it in the form of sulphur dioxide (SO 2) . Ninety percent of the SO2 is discharged into the atmosphere through the smelter's smokestacks, but 10% of the SO2 accidentally escapes from the smelter as "fugitive emissions." See generally Kennecott Copper Corp. v. Train, 526 F.2d 1149, 1150 n. 3 (9th Cir. 1975), cert. denied, 425 U.S. 935, 96 S. Ct. 1665, 48 L. Ed. 2d 176 (1976).

The attempt to bring Arizona's copper smelters into compliance with EPA's SO2 standards has had a long history. Arizona submitted its first state implementation plan to EPA in 1972. That plan, however, like four state and federal plans that followed it, never received final approval. A sixth plan, one promulgated by EPA, did receive final approval on January 4, 1978, see 43 Fed.Reg. 755 (1978), only to be stayed indefinitely about a month later, see 43 Fed.Reg. 6945 (1978). Arizona then developed a seventh implementation plan and submitted it to EPA for approval on September 20, 1979. That final plan is the subject of this litigation.

A valid implementation plan must insure attainment and maintenance of the national ambient air quality standards (NAAQSs) that have been set by EPA. See 42 U.S.C. §§ 7409(a), 7410(a) (2) (B). Those standards include an annual mean SO 2 concentration, see 40 C.F.R. Sec. 50.4(a) (1983), which is not in issue here. They also include two further standards that are in issue: a maximum 24-hour SO2 concentration that is not to be exceeded more than once per year and a maximum 3-hour concentration not to be exceeded more than once per year. See 40 C.F.R. Secs. 50.4(b), 50.5 (1983). Historically, compliance with the latter two SO2 standards has been achieved through use of the single-point rollback (single-point) system. In constructing a single-point implementation plan, the regulator attempts to predict the second worst meteorological conditions that will be confronted in an average year. It then sets a fixed, never-to-be-exceeded emission limitation at a level which insures that there will not be an exceedence of the maximum concentrations even on that day with the second-worst conditions of the year. Under the single-point approach, if the worst atmospheric conditions occur no more than once during the year, there will necessarily be compliance with the 24-hour and 3-hour NAAQSs, since there can be at most one exceedance of each maximum concentration during the year.

Arizona perceived two problems with the single-point technique. First, it regarded accurate implementation of the technique to be difficult. In particular, it doubted the regulator's ability to identify the day with the second worst meteorological conditions of the year and then to select an emission limitation which would insure compliance with the air standards on that day. Second, it regarded the single-point method as unduly harsh when applied to pollution sources which emit SO2 at highly variable rates, as copper smelters do. Arizona recognized that when a fixed, never-to-be-exceeded emission limitation is applied to a variable source, the source is effectively required to pollute well below the fixed limit most of the year in order to insure compliance on its few days of high emissions. In an attempt to avoid these problems, Arizona developed a new control method, the multi-point rollback (multi-point) technique.

The multi-point method permits more days of high emissions because, unlike the single-point method, it does not assume that those high emissions will necessarily occur on days of the worst or next-worst meteorological conditions. A regulator using the multi-point method to regulate 24-hour SO2 concentrations begins by selecting the rate at which it will permit the maximum 24-hour concentration to be exceeded. The rate selected by Arizona was once per year, in reflection of the NAAQSs. In the second step, the regulator measures SO2 concentrations around the SO2 source over an extended period of time. The data are organized into a graph that shows the rate at which different 24-hour SO2 concentrations are exceeded, and the regulator then proportionately reduces this observed air quality curve until it arrives at a new curve reflecting the desired rate of exceedance.¹ 

NOTE: OPINION CONTAINS TABLE OR OTHER DATA THAT IS NOT VIEWABLE

In the third step of the process, the regulator records the emissions from the SO2 source and in a similar fashion organizes the data into a graph that shows the rate at which various emission levels are exceeded. The regulator then reduces the observed emission curve until it arrives at a new emission curve or profile that will produce the air quality curve selected by the regulator in the second step of the multi-point process. In determining how much the observed emission curve must be reduced, the regulator assumes that if it reduces the observed emission curve by the same proportion as it reduced the observed air quality curve, then the new emission profile will produce the desired air quality curve.²  Finally, once the new emission profile is obtained, the regulator requires that the SO2 source conform its emissions to it. This requirement achieves the desired rate of exceedance, but at the same time allows the source to emit at the varying levels established by the emission profile.

NOTE: OPINION CONTAINS TABLE OR OTHER DATA THAT IS NOT VIEWABLE

Arizona used the multi-point system to regulate both 24-hour and 3-hour SO2 concentrations. Following the steps outlined above, Arizona identified for each smelter an emission curve that would be expected to cause one exceedance of the maximum 24-hour concentration per year. Arizona repeated the process with respect to 3-hour concentrations, similarly identifying for each smelter an emission curve that would be expected to cause one exceedance of the maximum 3-hour concentration per year. The state then compared the two emission curves it identified for each smelter and required each smelter to comply with the more stringent one. Arizona gave the smelters three years to achieve full compliance. In short, then, by the time full compliance occurs, Arizona's implementation plan will reduce smelter emissions to the point that each smelter can be expected to exceed the maximum SO2 concentrations no more than once per year.

On November 30, 1981, EPA announced its intention to give the Arizona implementation plan conditional approval. See 46 Fed.Reg. 58098 (1981). EPA proposed to approve Arizona's use and application of the multi-point methodology, as well as the state's three-year implementation period. The implementation plan, however, had failed to include specific regulations that reduced fugitive emissions, which can account for up to 10% of a smelter's SO2 output, see Kennecott Copper Corp. v. Train, 526 F.2d 1149, 1150 n. 3 (9th Cir. 1975), cert. denied, 425 U.S. 935, 96 S. Ct. 1665, 48 L. Ed. 2d 176 (1976).³  EPA proposed to condition its approval of Arizona's implementation plan upon the state's submitting to EPA by December 31, 1982, a plan for controlling fugitive emissions.

On January 14, 1983, EPA published its final rule. See 48 Fed.Reg. 1717 (1983). EPA continued its approval of the multi-point methodology and of the three-year attainment date. Regarding fugitive emissions, EPA dropped its proposed conditional approval in favor of combining full, unconditional approval of the implementation plan with a requirement that Arizona promulgate fugitive emission regulations by August 1, 1984. EPA informed Arizona that if it failed to promulgate adequate and timely regulations, EPA would do so.

Richard Kamp, a resident of Arizona who lives and works near copper smelters, petitioned for review of EPA's decision on March 15, 1983, pursuant to Sec. 307(b) (1) of the Clean Air Act, 42 U.S.C. § 7607(b) (1). The Environmental Defense Fund intervened in support of his petition, and the state of Arizona and a number of copper smelters intervened on behalf of EPA. As we have indicated, Kamp challenges four aspects of EPA's approval. He claims that the multi-point implementation plan does not insure attainment and maintenance of the NAAQSs; that the plan improperly relies on dispersion techniques instead of emission limitations; that the three-year attainment date is unlawful; and, finally, that EPA should have promulgated a control plan for fugitive emissions.

ANALYSIS

I. Insuring Attainment and Maintenance of the National Ambient Air Quality Standards for SO2

Section 110(a) (2) (B) of the Clean Air Act, 42 U.S.C. § 7410(a) (2) (B), requires that EPA shall approve an implementation plan only if it determines that the plan will "insure attainment and maintenance" of the NAAQSs. On two grounds, Kamp argues that EPA's approval of the multi-point implementation plan is inconsistent with that requirement. Both of Kamp's arguments assume the accuracy of the multi-point technique. We are therefore presented with legal questions rather than highly technical disputes about the validity of the science underlying the multi-point method of emission regulation. Cf., e.g., Connecticut v. EPA, 696 F.2d 147, 157-59 (2d Cir. 1982).

First, Kamp argues that the multi-point implementation plan creates an excessive risk that the NAAQSs will be violated. Kamp does not contest EPA's claim that the multi-point implementation plan produces expected rates of exceedance of no more than once per year. His objection is that an expected rate of exceedance of once per year is not enough to insure attainment and maintenance of the NAAQSs. Those standards require that there be no more than one exceedance of the maximum concentrations per year. See 40 C.F.R. Secs. 50.4(b), 50.5 (1983). Although over a ten-year period the multi-point system can be expected to allow no more than ten exceedances of each maximum concentration, there is no guarantee that the ten exceedances will occur exactly 365 days apart from each other. To the contrary, there is a 26% chance that in any given year there will be at least two exceedances bunched together.⁴  In any year in which such bunching occurs, there will of course be a violation of a NAAQS. In sum, Kamp argues that because the multi-point implementation plan produces a 26% chance of violation in any given year, that plan does not insure attainment and maintenance of the NAAQSs.

We find no indication that Congress addressed this issue raised by Kamp. The Act does not define the requirement that the implementation plan must "insure attainment and maintenance," nor does the legislative history elaborate on its meaning. See, e.g., H.R.Rep. No. 1146, 91st Cong., 2d Sess. 8 (1970) (stating only that the state plan must "assure [ ] achieving [the] standard [s]"), reprinted in 1970 U.S.Code Cong. & Ad.News 5356, 5363. We therefore review EPA's action to determine whether it is based on a reasonable interpretation of the Act. See Chevron, U.S.A., Inc. v. Natural Resources Defense Council, --- U.S. ----, 104 S. Ct. 2778, 2782-83, 81 L. Ed. 2d 694 (1984).

EPA approved the implementation plan and rejected Kamp's argument on the ground that allowing a 26% probability of random bunching within any given 365-day period is not inconsistent with the statutory requirement that the plan insure attainment and maintenance of the NAAQSs. That is a reasonable position. We would be troubled if the implementation plan had an expected rate of exceedance greater than once per year. We would also be troubled if the implementation plan could be expected to produce twelve or fifteen exceedances in any given ten-year period. That is not our case, however. What causes the 26% chance of violation here is only the possibility that exceedances may occasionally bunch together in a 365-day period. Similarly, some years may go by without any exceedance. It is not unreasonable for EPA to conclude that accepting the 26% risk of exceedence in any given year comports with its obligation to approve only implementation plans which insure attainment and maintenance of the NAAQSs.

Our conclusion that EPA has interpreted the Clean Air Act in a reasonable manner is bolstered by prior administrative practice. See Udall v. Tallman, 380 U.S. 1, 17-18, 85 S. Ct. 792, 801-802, 13 L. Ed. 2d 616 (1965). As we have noted, SO2 emissions have historically been regulated by the single-point method, which sets a fixed never-to-be-exceeded emission limitation that will insure that no exceedance will occur even on the day with the second-worst meteorological conditions expected in an average year. The federal implementation plan of 1978, which was retracted a month after its promulgation and which Kamp states that he prefers to the multi-point plan, used the single-point technique. As a theoretical matter, however, a single-point implementation plan generates the same possibility of bunching as does the multi-point plan. Under the single-point approach, if the worst atmospheric conditions occur no more than once during the year, there will be compliance with the NAAQSs. There is no guarantee, however, that the worst atmospheric conditions expected in an average year will occur only once. Although over a ten-year period there may be only ten days of the worst conditions, there will be a significant possibility--26% if the occurrence of the worst conditions is a truly random event, see supra note 4--that any given year will have two or more days of the worst conditions. When that happens, assuming the SO2 sources constantly emit up to the fixed, never-to-be-exceeded limit, there will necessarily be a NAAQS violation. Consequently, we find our conclusion that EPA acted reasonably confirmed by the fact that the multi-point implementation plan tends to be as protective of air quality as were plans approved under prior administrative practice.

Kamp raises a second argument based on the requirement that the multi-point implementation plan must insure attainment and maintenance of the NAAQSs. The multi-point plan allows the smelters to emit at high levels for a limited number of days each year. See supra pp. 1447-48 & n. 2. EPA admits that the levels are high enough that a NAAQS violation will occur if the high emissions twice coincide with a day having either the worst or the second worst weather conditions expected in an average year. In that sense, the high emissions exceed what Kamp calls the "threshold of violation." Kamp argues that permitting emissions above the "threshold of violation" is inconsistent with EPA's obligation to insure attainment and maintenance of the NAAQSs.

We reject Kamp's argument. The concept of a "threshold of violation," upon which the argument is based, has no foundation in the Clean Air Act. The Act requires simply that the state implementation plan insure attainment and maintenance of the NAAQSs. See 42 U.S.C. § 7410(a) (2) (B). We have already indicated that the Arizona's multi-point plan complies with this requirement. It is therefore irrelevant that on some days emissions are at such a level that as a theoretical matter a NAAQS violation could occur in an average year. Furthermore, even if emissions were always below the "threshold of violation" as defined by Kamp, that requirement would not remove the possibility of NAAQS violations. There would remain the possibility that in the nonaverage years in which two days of the worst conditions bunch together, a NAAQS violation could occur.

We review to determine whether EPA has given a reasonable interpretation to the requirement that the attainment and maintenance of the NAAQSs be insured. We conclude that EPA's position, which risks only the possibility that exceedances may randomly bunch together, constitutes a reasonable interpretation of the Clean Air Act.

II. Dispersion Techniques and Emission Limitations

Section 123(a) of the Clean Air Act, 42 U.S.C. § 7423(a), provides that EPA cannot approve an implementation plan which relies on "dispersion techniques" to attain compliance with the NAAQSs. Kamp argues that Arizona's multi-point plan uses dispersion techniques and therefore could not lawfully be approved by EPA.

There are two types of dispersion control techniques: (1) "tall stacks," which reduce ambient air concentrations by dispersing emissions high into the atmosphere, and (2) intermittent or supplemental control techniques. See Bunker Hill Co. v. EPA, 572 F.2d 1286, 1291 n. 4 (9th Cir. 1977); Kennecott Corp. v. EPA, 684 F.2d 1007, 1010 n. 4 (D.C. Cir. 1982). Obviously Arizona's multi-point plan does not rely on tall stacks. Therefore, if the multi-point plan uses dispersion techniques as Kamp charges, it must do so by employing intermittent control techniques.

The statute defines an "intermittent control technique" as one which varies emissions with atmospheric conditions. 42 U.S.C. § 7423(b). The cases discussing the term have used the same definition. See, e.g., Bunker Hill, 572 F.2d at 1291 n. 4. The multi-point implementation plan, however, does not set emission levels which vary with atmospheric conditions. So long as the smelters' emissions are within the specified emission profile, the source will be in compliance with the implementation plan regardless of the prevailing atmospheric conditions. It follows that the multi-point plan does not use dispersion techniques.

Kamp recognizes that the multi-point implementation plan does not set emissions which expressly vary with atmospheric conditions. He argues, however, that the multi-point plan should nonetheless be treated like an intermittent control system because its effectiveness would be enhanced if combined with intermittent controls. Kamp is correct that if intermittent controls were added to the multi-point plan, greater protection against pollution would result. It would be an improvement from the point of view of air quality if the smelters were prohibited from polluting at the high end of the emission spectrum during periods of bad conditions. That argument, however, proves too much; in all probability any control strategy could be enhanced through the addition of intermittent controls. Indeed, the 1978 federal single-point implementation plan, which Kamp concedes does not use dispersion techniques, would be improved by the addition of intermittent controls that require further reduction of emissions on days of extremely bad atmospheric conditions. We therefore cannot accept Kamp's arguments.

Section 110(a) (2) (B) of the Clean Air Act, 42 U.S.C. § 7410(a) (2) (B), requires that implementation plans use emission limitations to the maximum extent feasible in achieving compliance with the NAAQSs. See, e.g., Kennecott Copper Corp. v. Train, 526 F.2d 1149, 1150 (9th Cir. 1975), cert. denied, 425 U.S. 935, 96 S. Ct. 1665, 48 L. Ed. 2d 176 (1976); National Resources Defense Council v. EPA, 489 F.2d 390, 406 (5th Cir. 1974), rev'd on other grounds sub nom. Train v. National Resources Defense Council, 421 U.S. 60, 95 S. Ct. 1470, 43 L. Ed. 2d 731 (1975). Kamp argues that the multi-point implementation plan does not use emission limitations; that the use of such limitations is feasible; and that therefore the multi-point plan violates Sec. 110(a) (2) (B).

The definition of "emission limitation" in the Act has two elements: (1) an emission limitation must limit "the quantity, rate, or concentration of emissions of air pollutants"; and (2) it must operate "on a continuous basis." See 42 U.S.C. § 7602(k). The first requirement is clearly met, for by lowering each smelter's emission curve, the multi-point plan reduces the overall quantity of emissions. The controversy centers on the second requirement. Kamp argues that because the multi-point plan allows the acceptable emission level to fluctuate, it does not continuously regulate emissions.

EPA has a broader definition of "continuous" than does Kamp. Under the Agency's definition, an implementation plan operates continuously so long as some limitation on emissions, although not necessarily the same limitation, is always imposed on the SO2 source. Under that definition, the multi-point plan operates continuously, since the SO2 source is always subject to the specified emission profile.

We cannot say that EPA's definition of "continuous" is unreasonable. In the first place, the requirement of regulation on a continuous basis does not necessarily imply that the source always be subject to precisely the same limitation. Equally important, EPA's definition is supported by the legislative history. Congress's primary purpose behind requiring regulation on a continuous basis was to exclude intermittent control techniques from the definition of emission limitations. The House Report states:

By defining the term [ ] "emission limitation," ... the committee has made clear that constant or continuous means of reducing emissions must be used.... By the same token, intermittent or supplemental controls or other temporary, periodic, or limited systems of control would not be permitted....

H.R.Rep. No. 294, 95th Cong. 1st Sess. 92 (1977), reprinted in 1977 U.S.Code Cong. & Ad.News 1077, 1170. Accord id. at 6, reprinted in 1977 U.S.Cong. & Ad.News at 1084. EPA's definition of "continuous" accommodates the legislative history by giving "continuous" an essentially negative definition: any control technique is continuous which does not operate on an intermittent basis. Since EPA has reasonably interpreted the Act, we affirm its conclusion that the multi-point implementation plan uses "emission limitations." See Chevron, U.S.A., Inc. v. Natural Resources Defense Council, --- U.S. ----, 104 S. Ct. 2778, 2782-83, 81 L. Ed. 2d 694 (1984).

Arizona's implementation plan gives the copper smelters three years from the date of EPA approval to achieve full compliance with its emission limitations. Kamp challenges that provision. He argues that because Arizona has been designated a nonattainment area for SO2 under Sec. 107(d) of the Act, 42 U.S.C. § 7407(d), see 40 C.F.R. Sec. 81.303 (1983), Arizona's plan should be considered a nonattainment implementation plan under Part D of the Clean Air Act, 42 U.S.C. §§ 7501-7508. If the plan is so treated, then its three-year implementation period would be unlawful, for Part D implementation plans are required to have December 31, 1982, as their attainment date. See 42 U.S.C. § 7502(a) (1).

EPA argues that we should not reach the merits of Kamp's argument because making a past date like December 31, 1982, the attainment date for Arizona's plan would be a futile gesture. EPA maintains that regardless of what we order, the smelters cannot achieve compliance with the plan's emission limitations prior to the end of the three-year implementation period.

We are willing to accept EPA's assumption about the inability of the smelters to achieve immediate compliance with the required emission limitations. Despite that fact, something important does turn on the attainment date issue: whether in the interim it is likely that the smelters will have to use dispersion techniques to attain compliance with the NAAQSs. If the attainment date is a past date, the smelters will be immediately subject to enforcement actions under Sec. 113 of the Clean Air Act, 42 U.S.C. § 7413. To avoid being assessed a penalty for their noncompliance, the smelters must obtain a nonferrous smelter order under Sec. 119 of the Act, 42 U.S.C. § 7419. A prerequisite to such an order is that the smelter adopt a combination of emission limitations and dispersion techniques sufficient to bring it into immediate compliance with the NAAQSs, see 42 U.S.C. § 7419(d) (1). Thus, unlike the three-year period in Arizona's plan, a past attainment date would give the smelters a substantial incentive to use dispersion techniques to achieve immediate compliance with the NAAQSs.

When we reach the merits, however, we reject Kamp's argument. Part D of the Clean Air Act specifies the alterations to its implementation plan that a nonattainment area must adopt before new major sources of air pollution may be constructed within the nonattainment area. See 42 U.S.C. §§ 7410(a) (2) (I), 7502. The implementation plan involved in this case has nothing to do with the concerns of Part D: its objective is to bring existing sources of SO2 pollution into compliance with the NAAQSs, not to allow for the construction of new major sources of emissions. Therefore, we will not evaluate Arizona's plan as if it were a Part D plan.⁵ 

Since Arizona's plan is not a Part D plan, we look to Sec. 110(a) (2) (A) of the Act, 42 U.S.C. § 7410(a) (2) (A), to determine whether the three-year implementation period is lawful. See Northern Ohio Lung Association v. EPA, 572 F.2d 1143, 1148-49 (6th Cir. 1978). Under Sec. 110(a) (2) (A), the attainment date set by Arizona's plan is lawful. The section requires that the implementation plan adopt a date which "provides for the attainment of [the] primary standard [s] as expeditiously as practicable but ... in no case later than three years from the date of approval...." Here the Arizona plan's attainment date is exactly three years from the date of approval, and no argument is made that an earlier attainment date would be practicable.⁶ 

IV. The Failure to Regulate Fugitive Emissions

EPA concedes that at least with respect to some of the smelters, fugitive emissions may cause violations of the NAAQSs. See 48 Fed.Reg. 1717-18 (1983). Despite that concession and the absence of specific regulations directed at fugitive emissions in Arizona's plan, EPA approved the plan. EPA required only that Arizona promulgate regulations to control the problem of fugitive emissions by August 1, 1984. Kamp argues that EPA unlawfully gave Arizona additional time to develop a control strategy for fugitive emissions. He maintains that the lack of fugitive emission controls in Arizona's plan means that the plan does not insure attainment and maintenance of the NAAQSs, see 42 U.S.C. § 7410(a) (2) (B), and that therefore EPA should have promulgated its own plan for the control of fugitive emissions under Sec. 110(c) (1) of the Act, 42 U.S.C. § 7410(c) (1).

EPA first argues that we have no jurisdiction to review its failure to promulgate a control strategy for fugitive emissions because that failure should have been challenged in the district court in the first instance. It cites as authority Sec. 304(a) (2) of the Clean Air Act, 42 U.S.C. § 7604(a) (2), which provides for suits in district court "against the Administrator where there is alleged a failure of the Administrator to perform any act or duty ... which is not discretionary with the Administrator."

EPA's jurisdictional argument was considered and rejected by the Seventh Circuit in Indiana & Michigan Electric Co. v. EPA, 733 F.2d 489 (7th Cir. 1984). Section 307(b) (1) of the Act, 42 U.S.C. § 7607(b) (1), provides for judicial review in the circuit courts of EPA's approval of a state implementation plan. Relying on that provision, the Seventh Circuit held that when the challenge to agency inaction "is embedded in a challenge to the validity of an implementation plan," jurisdiction lies in the circuit court reviewing the implementation plan. Id. at 490. Otherwise, of course, there would be a danger that two proceedings involving essentially the same agency action could occur simultaneously. Id. at 491. We find the reasoning of the Seventh Circuit persuasive and so hold that Kamp's challenge to the failure to regulate is properly before us.

The intervenor copper smelters argue that despite the Arizona plan's lack of a specific control strategy for fugitive emissions, that plan does insure attainment and maintenance of the NAAQSs. They point out that the plan requires the smelters to submit studies of the fugitive emission problem to Arizona and that EPA has required Arizona to promulgate detailed regulations by August 1, 1984. They maintain that those provisions provide ample guarantee that fugitive emissions will not prevent the attainment and maintenance of the NAAQSs.

Section 110(a) (2) (B) of the Clean Air Act, 42 U.S.C. § 7410(a) (2) (B), requires that implementation plans rely on emission limitations to the maximum extent feasible in insuring the attainment and maintenance of the NAAQSs. See, e.g., Kennecott Copper Corp. v. Train, 526 F.2d 1149, 1150 (9th Cir. 1975), cert. denied, 425 U.S. 935, 96 S. Ct. 1665, 48 L. Ed. 2d 176 (1976); Natural Resources Defense Council v. EPA, 489 F.2d 390, 406 (5th Cir. 1974), rev'd on other grounds sub nom. Train v. Natural Resources Defense Council, 421 U.S. 60, 95 S. Ct. 1470, 43 L. Ed. 2d 731 (1975). Accordingly, the Arizona plan's reliance on future regulation would fully comply with Sec. 110(a) (2) (B) only if it were infeasible at the time the plan was promulgated to resolve the fugitive emission problem through emission limitations. EPA's final rulemaking indicates that perhaps it would have been infeasible. See 48 Fed.Reg. 1718 (1983).

We do not resolve whether Arizona's plan is in complete compliance with Sec. 110(a) (2) (B) because an implementation plan need not be in absolute compliance with Sec. 110(a) (2) before it receives federal approval. EPA can approve a substantially complete implementation plan if it has assurance that the state will promptly complete the plan and if the approval of the incomplete plan does not circumvent any of the Act's substantive requirements. See Connecticut Fund for the Environment, Inc. v. EPA, 672 F.2d 998, 1005-07 (2d Cir.), cert. denied, 459 U.S. 1035, 103 S. Ct. 445, 74 L. Ed. 2d 601 (1982); City of Seabrook v. EPA, 659 F.2d 1349, 1353-57 (5th Cir. 1981), cert. denied, 459 U.S. 822, 103 S. Ct. 51, 74 L. Ed. 2d 57 (1982); Friends of the Earth v. EPA, 499 F.2d 1118, 1124 (2d Cir. 1974). As the Second Circuit reasoned, the demands of its "difficult and complex job" require that EPA be given some flexibility to approve nearly complete implementation plans. Connecticut Fund for the Environment, 672 F.2d at 1006 (quoting Friends of the Earth, 499 F.2d at 1124).

Applying the approach of Connecticut Fund for the Environment, we conclude that EPA acted lawfully. Arizona's plan is substantially complete, since fugitive emissions are only one part of a large SO2 pollution problem otherwise covered by the plan. Arizona has assured that the incomplete part of its plan will be promptly completed, for EPA required Arizona to submit a detailed fugitive emission control plan by August 1, 1984. Most important, EPA's action does not circumvent the Act's substantive requirements by delaying the full attainment and maintenance of the NAAQSs. EPA has required that the state's plan for fugitive emissions be fully implemented within the plan's three-year attainment period. See 48 Fed.Reg. 1718 (1983). In light of these factors, EPA could properly approve Arizona's plan without itself promulgating a control plan for fugitive emissions.

CONCLUSION

For the foregoing reasons, we affirm the approval of Arizona's multi-point implementation plan for the control of SO2 emissions from copper smelters.