Full opinion text
RALPH B. GUY, JR., Circuit Judge. Navistar International Transportation Corporation (Navistar) petitions this court for review of the United States Environmental Protection Agency’s (EPA) decision to disapprove a proposed revision to the Ohio State Implementation Plan (SIP), which is a plan for implementing federal air quality standards pursuant to the Clean Air Act (CAA or Act). 42 U.S.C. §§ 7401 et seq. The revision, approved by the Ohio Environmental Protection Agency (OEPA), proposed to relax the requirements of the Ohio SIP regarding the limits on emissions of volatile organic compounds (VOCs) at Navistar’s truck manufacturing and assembly facility in Springfield, Ohio. The EPA’s determination is a final agency action subject to judicial review in the courts of appeals under section 307(b)(1) of the CAA. 42 U.S.C. § 7607(b)(1). The standard of review with regard to EPA action concerning a SIP is specified in the Administrative Procedure Act (APA), 5 U.S.C. § 706(2)(A), which provides that agency action may be set aside only if it is found to be “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” Because we do not find the EPA’s determination here to have been arbitrary or capricious, the petition will be denied. I. A. Statutory and regulatory background A combined state and federal program to control air pollution was formulated in 1970 through various amendments to the Clean Air Act. Pursuant to the mandates of sections 108 and 109 of the Act, the EPA promulgated national ambient air quality standards (NAAQS) for a variety of pollutants. The standards relevant to the case before us are those in effect for ozone. In 1979, the EPA established the primary and secondary standards for ozone at 0.12 parts per million. 40 C.F.R. § 50.9 (1988); see also id. at pt. 50, app. H (“Interpretation of the [NAAQS] for Ozone”). The states were given the primary responsibility for achieving the federal air quality standards and, absent special circumstances, were required to do so by 1975. In keeping with this responsibility, the states were required under section 110 of the Act to develop state implementation plans providing for “implementation, maintenance, and enforcement” of the federal ambient air quality standards within their borders. 42 U.S.C. § 7410(a). The SIPs must be approved by the EPA and must provide, among other things, for the attainment of primary NAAQS “as expeditiously as practicable” and insure the “maintenance” of those standards. CAA § 110(a)(2)(A) and (B), 42 U.S.C. § 7410(a)(2)(A) and (B). The Supreme Court explained the review process as follows: Under § 110(a)(2), the Agency is required to approve a state plan which provides for the timely attainment and subsequent maintenance of ambient air standards, and which also satisfies that section’s other general requirements. The Act gives the Agency no authority to question the wisdom of a State’s choices of emission limitations if they are part of a plan which satisfies the standards of § 110(a)(2)_ Thus, so long as the ultimate effect of a State's choice of emission limitations is compliance with the national standards for ambient air, the State is at liberty to adopt whatever mix of emission limitations it deems best suited to its particular situation. Train v. Natural Resources Defense Council, Inc., 421 U.S. 60, 79, 95 S.Ct. 1470, 1482, 43 L.Ed.2d 731 (1975) (emphasis in original). The states may also propose revisions to their SIPs. CAA § 110(a)(3), 42 U.S.C. § 7410(a)(3)(A). Proposed revisions also are reviewed under the standards of section 110(a)(2). Section 110(a)(3)(A) of the Clean Air Act provides that the EPA “shall approve any revision of an implementation plan” if the EPA “determines that it meets the requirements of paragraph (2)” of section 110. 42 U.S.C. § 7410(a)(3)(A). When evaluating a revision to a state plan, “the logical inquiry for the EPA is to assess whether the proposed change interferes with attainment.” United States Steel Corp. v. EPA, 633 F.2d 671, 674 (3d Cir.1980) (emphasis in original). As of 1975, many states had air quality control regions (AQCR) or portions thereof that did not meet the national standards, despite the existence of previously approved SIPs. If any region of a state fails to meet NAAQS, that region is designated as a “nonattainment area” for the particular pollutant exceeding the applicable standard. CAA § 171(2), 42 U.S.C. § 7501(2); CAA § 107(d)(1)(A)-(C), 42 U.S.C. § 7407(d)(l)(A)-(C). The Act was amended in 1977 in order to deal with these nonat-tainment areas. Under the amendment (now Part D of Title I of the Act), states were required to submit revisions to their SIPs demonstrating that NAAQS would be attained “as expeditiously as practicable,” though no later than the end of the 1982 calendar year. CAA § 172(a)(1), 42 U.S.C. § 7502(a)(1). Part D leaves to the states the primary responsibility for meeting NAAQS and allows considerable discretion in devising an appropriate mix of emission limitations. Connecticut Fund for Environment, Inc. v. EPA, 696 F.2d 169, 173 (2d Cir.1982). However, any state that did not have an approved SIP providing for the attainment of primary NAAQS not later than 1982 was subject to a moratorium on the construction and modification of any “major stationary source” of pollution in nonattainment areas. CAA § 110(a)(2)(I), 42 U.S.C. § 7410(a)(2)(I). These Part D plans must meet the requirements of section 172(b) of the Act, which provides in pertinent part, that the SIP shall, inter alia: (2) provide for the implementation of all reasonably available control measures as expeditiously as practicable; (3) require, in the interim, reasonable further progress (as defined in section 7501(1) of this title) including such reduction in emissions from existing sources in the area as may be obtained through the adoption, at a minimum, of reasonably available control technology. 42 U.S.C. § 7502(b). “Reasonably available control technology” (RACT) has been defined at 40 C.F.R. § 51.1(o) to mean “devices, systems, process modifications, or other apparatus or techniques, the application of which will permit attainment of the emission limitations set forth in Appendix B to this part.” Appendix B of 40 C.F.R. part 51 is entitled “Examples of Emission Limitations Attainable With Reasonably Available Technology.” Since 1976, the EPA has interpreted “reasonably available control technology” to be “the lowest emission limitation that a particular source is capable of meeting by the application of control technology that is reasonably available considering technological and economic feasibility.” The City of Springfield in Clark County, Ohio, was designated a nonattainment area for ozone NAAQS in 1978. 40 C.F.R. § 81.-336. As a consequence, Ohio proposed revisions to the ozone portion of its SIP in order to satisfy the requirements of Part D of the CAA, 45 Fed.Reg. 72,122 (1980); 47 Fed.Reg. 28,097 (1982). As part of the 1980 proposal, the state submitted a revision that combined Clark County with five other counties (Montgomery, Greene, Darke, Miami, and Preble) into the Dayton Urban Area. The revised Ohio SIP modified the state’s existing VOC regulations. In particular, the revised Ohio SIP modified section 3745-21-09(U) of the Ohio Administrative Code to provide that an owner or operator of a miscellaneous metal part coating line must use only those extreme performance coatings, i.e., paints, with a VOC content of 3.5 pounds of VOC per gallon of coating (excluding water) or less. Ohio Admin.Code § 3745-21-09(U)(l)(iii). Compliance with this VOC emission limitation was required by December 31, 1982. Id. at § 3745-21-04(C)(28). As part of the Part D SIP revision approval process, Ohio submitted documentation known as an “attainment and reasonable further progress demonstration” (attainment demonstration) for the greater Dayton area. The attainment demonstration contained an inventory of the VOC emissions occurring in a base year (1975), evidence as to anticipated emission reductions from the implementation of proposed revisions, and an estimate of the effectiveness of the proposed revisions to demonstrate attainment of the ozone NAAQS by December 31, 1982. 45 Fed.Reg. 72,122, 72,123. The EPA approved the attainment demonstration for the greater Dayton area, 40 C.F.R. § 52.1885, which established December 31,1982, as the date for compliance by the greater Dayton area with the ozone NAAQS. 45 Fed.Reg. at 72,130, 72,141. Finally, the EPA approved most portions of the proposed Ohio SIP revisions, including the ozone portions of section 3745-21-09 governing VOC emissions from surface coatings of miscellaneous metal parts and products. 45 Fed.Reg. at 72,142. B. Navistar’s proposed SIP revision Navistar maintains a plant in Springfield, Ohio. Its operations consist of the manufacture and assembly of medium and heavy-duty trucks. Navistar uses surface coating lines at its body and assembly plants to paint truck cabs, hoods, chassis and miscellaneous metal parts. The coating lines use paints containing VOCs, such as solvents, which maintain various operational properties of the paints. As the VOCs evaporate, they are emitted into the ambient air and contribute to the formation of ozone. The parties agree that Navis-tar’s operations are regulated by Ohio Administrative Code sections 3745-21-09(U)(l)(a)(iii) and 3745-21-04(0(28), which require Navistar to limit its emissions by December 31, 1982, to no more than 3.5 pounds of VOCs per gallon of paint applied. The parties further agree that the ten coating lines at issue here — nine at Navistar’s assembly plant and one at its body plant— were not in compliance with the Ohio SIP during the period the EPA was evaluating the proposed SIP revision. In March 1986, the OEPA submitted to the EPA a proposed revision (variance) to the Ohio ozone SIP for Navistar’s ten coating lines. The Navistar SIP revision sought a five-year compliance date extension to December 31, 1987, for three Navis-tar coating lines that were to be shut down and replaced with new and complying lines after that date. As to the remaining seven lines, OEPA proposed to extend compliance until December 31, 1987, at which time Navistar was to submit a study to the OEPA “discussing the feasibility of compliance with” three alternative emission limitations, including (1) complying with the 3.5 pounds/gallon emission limit, (2) complying with the state’s limits based on a compliance averaging method called a “bubble,” and (3) continuing to operate according to the limits in the variance. Before final submission of the SIP revision to the EPA, the EPA requested additional information and characterized the proposal as one consisting of a “compliance date extension” for three coating lines and “a relaxation of RACT” for the remaining coating lines. The EPA also indicated that permanent “relaxations” from RACT and temporary “extensions” from RACT compliance are governed by two different sets of approval criteria. The EPA explained these approved criteria and requested further documentation from OEPA as follows: The two main criteria for evaluating RACT compliance date extensions are expeditiousness and likelihood of success (April 16, 1984, letter from Region V to State Air Directors, and August 17, 1984, letter from former Assistant Administrator Joseph Cannon to Richard Lillquist). IH [Navistar] is proposing to replace its existing cab prime line with a new elec-trodeposition coat/interior finish line and to replace two existing spray booths with new first color lines controlled by thermal oxidation. These new lines should be installed by December 1987. For this compliance date extension to be approved, IH must document that it proceeded expeditiously to develop and implement these compliance plans from the time the regulations were adopted to the present with no significant periods of inaction. IH has provided no information concerning the development of the compliance plans. A relaxation from RACT can only be allowed when the RACT-level emission limit is technically and economically infeasible. Therefore, IH must document that complying coatings are not available and that add-on control is economically infeasible. IH has provided correspondence from only 4 coating suppliers regarding the availability of complying coatings. Most of this correspondence is dated October 1983 and states that low solvent topcoats are not available. However, in a March 26, 1984, letter Inmont indicates that it has developed a high solids topcoat that seems to be compatible with IH’s coating operations. IH should submit more recent correspondence from a larger number of coating suppliers. If a supplier can provide a low solvent topcoat but IH believes that it is not acceptable, IH should explain why the coating is unacceptable and, if possible, include the results of testing performed on the coating at the Springfield plants. IH submitted a feasibility study on control of VOC emission from its exhaust stacks which evaluated the cost effectiveness of 7 types of control equipment. According to the study, the cost effectiveness ranged from $2140 to $10,300 per ton of VOC removed. This falls within the range contained in the CTG [Control Technology Guidelines] for manual spray coating of miscellaneous metal parts and products. In addition, the situation evaluated in the feasibility study does not represent that of IH’s compliance plan and the values given for VOC content of coatings and actual and allowable emissions do not correspond to those contained in the Variances to Operate. Therefore, IH has not demonstrated that add-on control is economically infeasible for its Assembly Plant. Finally, a July 29, 1983, memorandum from Sheldon Meyers titled “Source Specific SIP Revisions” states that for any relaxation to be approved, the State must demonstrate that the SIP as a whole despite the relaxation, will continue to provide for attainment by the end of 1982. The 1979 SIP for the Dayton AQCR included six counties in the attainment demonstration. Although the SIP demonstrated attainment by the end of 1982, there were measured violations of the NAAQS for ozone in 1983 and 1984 in the Dayton area. Therefore, it appears that the 1979 attainment demonstration may no longer be valid and a full up-to-date attainment demonstration would be required before the proposed SIP relaxation for IH could be considered. (Letter of February 26, 1986, from Steve Rothblatt (EPA) to Charles M. Taylor (OEPA)). OEPA responded to the EPA’s requests and comments as follows: (1) as to the need for further information on the availability of compliance coatings, Navistar had already contacted the four leading suppliers of truck coatings, and “additional searches for complying coatings would be futile”; (2) as to the economic feasibility of control equipment, OEPA considered $1,600 per ton of VOC reduced “to be the highest cost-effectiveness threshold of all the RACT VOC categories,” Navistar calculated the range of cost-effectiveness for add-on controls to be $2,140 to $10,300, and the EPA values given in the Control Technology Guidelines (CTG) (ranging from $2,250 per ton to $20,154 per ton) were unreasonable; and (3) with respect to ozone violations since 1982 and the need for an attainment demonstration, OEPA wrote that, in 1983, measured exceedances of the ozone standard were recorded on a regional basis. An exceedence of the standard occurs when the maximum average one hour concentration of ozone exceeds 0.12 parts per million (PPM). A violation of the standard occurs when the number of ex-ceedances at any given monitoring site, when averaged over a three-year period, is greater than one. RAPCA operates equipment to measure ambient ozone concentrations at five sites in six counties, including one site in Clark County where the facility is located. During 1983, the monitoring equipment measured ozone exceedances totalling three, three, zero, zero, and six, respectively, for the five sites. During 1984, excee-dances totalling one, zero, one, zero, and zero, respectively, were measured at the five sites. During 1985, no exceedances were measured at any site. Ozone measurements for the years 1984 and 1985 reflect essentially total implementation of SIP attainment measures, indicating original air quality projections to be valid. (Letter of April 16,1986) (emphasis in original). The February EPA letter and OEPA’s response manifest disagreements that prevailed throughout the approval process and still persist today. The disagreements involve the feasibility of using coatings that comply with YOC emission limitations, the feasibility of add-on controls and the need for a new attainment demonstration. A further conflict would develop later concerning the EPA’s characterization of the variance as involving both a permanent “relaxation” from RACT and a temporary “extension” from RACT compliance, with Navistar and Ohio contending that the revision only proposed a compliance date extension for all ten lines. Consequently, from Ohio and Navistar’s perspective, the approval criteria for evaluating permanent relaxations from RACT do not apply. The EPA staff proceeded to evaluate the proposed SIP revision and Ohio’s comments, which evaluation is set forth in three internal technical support documents (TSDs). In the May 2,1986, TSD, the staff evaluated the compliance date extension requested for three of the coating lines according to the criteria of expeditiousness and likelihood of success and determined that Navistar had provided no information regarding its attempts to develop and implement “its compliance plan from the time the regulations were adopted to the present with no significant periods of inaction.” With respect to RACT relaxation for the remaining seven coating lines, the TSD reiterated that RACT can be relaxed only “when the RACT-level emission limit is technically or economically infeasible” and found, for the same reasons set forth in the February correspondence, that Navistar had not demonstrated that complying “coatings were unavailable or that add-on controls were not feasible. Regarding the need for a new attainment demonstration, the staff noted that Navistar’s air quality demonstration for the proposed revision was based on the 1979 ozone SIP for the Dayton AQCR, and, although that SIP demonstrated that attainment would be achieved by the end of 1982, measured violations of the ozone NAAQS in 1983 and 1984 demonstrated the need for an up-to-date attainment demonstration. While the May 2, 1986, TSD set forth the same arguments as the February letter, it did not address the objections raised by OEPA in its April 16, 1986, response to that letter. However, in a addendum to the May 2 TSD (May 30, 1986, TSD) the staff commented on some of the responses made in the April letter. The May 30 TSD pointed out that the revision regarding seven of the lines could be evaluated either as a proposal for a permanent RACT relaxation, as the May 2 TSD did, or as a temporary relaxation according to the criteria applicable to compliance date extensions. Accordingly, the TSD set forth the staff evaluation under both designations. Evaluating the revision under the RACT relaxation criteria, the staff addressed OEPA’s contention that Navistar had already contacted the four leading suppliers of truck coatings and rejected the contention that additional searches for complying coatings would be futile. Specifically, the TSD found that Navistar’s exchange of correspondence with the four coating suppliers occurred in October 1983 and merely indicated that complying topcoats were not available at that time but that research was continuing. The TSD stated that this was not sufficient documentation of technical or economic infeasibility — the RACT relaxation criteria — and directed that Navistar should supply more recent correspondence with a larger number of suppliers. The TSD acknowledged Navistar’s own efforts to develop complying coatings but found that, because Navistar was currently testing some promising coatings and had recently replaced some of its non-complying coatings, “there appears to be no justification for a permanent relaxation.” As to add-on controls, the EPA stated that it considers the CTG to be the appropriate cost-effective threshold and rejected OEPA’s contention that $1,600/ton of VOC should be the highest threshold for all RACT VOC categories. Accordingly, the EPA found the cost controls calculated by Navistar to be within the cost-effective range and therefore not economically infeasible. When treating the revision solely as a compliance date extension and evaluating the variance according to expeditiousness and likelihood of success, the TSD stated as follows: In order to demonstrate expeditiousness, a source should show that it progressed from the time the applicable regulations were adopted to [the] present with no significant periods of inaction. The high solids paint summary submitted by IH shows that it has been testing paints fairly consistently since 1984. However, little information was provided concerning any previous work done. This additional information is required for IH to demonstrate expeditiousness. A compliance plan can be considered likely to succeed if a source can provide tangible evidence of significant progress toward developing complying coatings. Although IH has shown that it was able to reformulate some of its coatings, it should also provide an estimate of the remaining work. In addition, if this is to be evaluated as a compliance date extension, the OEPA should provide an enforceable compliance schedule for IH. With respect to the attainment demonstration, the May 30 TSD did not alter the position taken in the May 2 TSD that a new attainment demonstration was needed. However, neither TSD addressed the distinctions made in the Ohio letter between “violation” and “exceedence.” The TSD recommended that the revision be disapproved for the following reasons: 1. [Navistar] has not demonstrated that its compliance plan was developed and implemented expeditiously. 2. [Navistar] had not demonstrated that RACT-level control is technically or economically infeasible. 3. The 1979 ozone SIP modeling demonstration for the Dayton area appears to be invalid. The EPA issued a third TSD (February 7, 1987) that responded to comments made by the OEPA following the previous TSD. This TSD again recommended disapproval and reiterated the reasons set forth in the previous TSDs. However, unlike the previous TSDs, it integrated its discussion of an attainment demonstration into the approval criteria for compliance date extensions. Although the EPA continued to rely on its claim that “violations of the ozone NAAQS in 1983 and 1984 in the Dayton AQCR” demonstrated the invalidity of the 1979 SIP air quality demonstration, in addition to evaluating expeditiousness and likelihood of success of the proposed compliance date extension for three lines, the EPA also stated that “exceedances” of the NAAQS and changes in the “margin of attainment” required a “revised attainment demonstration” as a threshold to demonstrating the section 110 requirements of attainment and maintenance of NAAQS. In addition, the TSD required that an evaluation of expeditiousness required the state to perform a national survey of the compliance status of sources in the same category. The EPA set forth these requirements as follows: First, the state must demonstrate that the extension will not interfere with timely attainment and maintenance of the ozone standard. This would generally be done by comparing the margin for attainment predicted by the approved ozone attainment demonstration and the increased emissions that would result under the proposed extension. However, if the State or USEPA believes that there has been a substantial change in the inventory since the ozone SIP was approved so that the margin of attainment has changed significantly, a revised demonstration in support of the revision is required. Such a demonstration would be necessary in areas originally demonstrating attainment by 1982, but for which post-1982 monitoring data are indicating exceedances of the ozone standard. Because IH is located in such an area, a revised demonstration of attainment for the Dayton area would be required before the compliance date extension can be approved. Second, time extensions must be consistent with the requirement that nonattainment area SIPs provide for implementation of all reasonably available control measures as expeditiously as practicable. A determination of expeditiousness should include an examination of the compliance status of other sources nationally in the same VOC source category and the most expeditious means of compliance available irrespective of the method proposed in the SIP revision. Unless it can be shown that the time-frame in the SIP did not allow sufficient time for an economically and technologically feasible compliance plan to be implemented, a compliance date extension can not be approved. This requirement has not been met because the State has not performed a national survey of the compliance status of sources in the same category. In addition, it has not been demonstrated that the timeframe in the SIP was not adequate to implement an economically and technically feasible compliance plan. The 1987 TSD also evaluated the RACT relaxation portion of the revision in light of additional information provided OEPA as follows: On November 17, 1986, the Ohio EPA submitted additional information concerning this revision which contains revised cost estimates for add-on control. The revised estimates are based on control of each of the coating lines. On this basis, the range of cost-effectiveness for add-on controls is determined to be $15,-630 to $46,720 per ton of VOC controlled. Although this cost-effectiveness is higher than what is considered reasonable by USEPA, the basis for the cost estimates has not been adequately explained and documented. On December 23, 1987, the EPA again analyzed the proposed Navistar SIP revision, in particular, the RACT relaxation proposed for seven lines. (Addendum to the Feb. 9, 1987 TSD). This time, the staff pointed to the findings of a January 1986 study of complying coatings (HOV Study), that documented the availability and actual use of certain complying coatings by the heavy-duty off-highway vehicle (HOV) manufacturing industry. According to this study, at least six coating companies supplied complying coatings, and at least six manufacturing facilities currently used those coatings. On July 7, 1988, the EPA finally issued a Federal Register Notice, proposing to disapprove the SIP revision. 53 Fed.Reg. 25,-509. In proposing disapproval, the EPA essentially relied upon the reasons stated in the TSDs. According to the agency, violations in 1983 and 1984, and recent ex-ceedances in 1987, required the SIP revision to be supported by an up-to-date attainment demonstration. In the absence of such an attainment demonstration, and given the inadequacy of research on the availability of complying coatings and the inadequacy of documentation explaining Navis-tar’s cost estimates for add-on controls, the EPA determined there was not a sufficient showing that the proposed SIP would not interfere with expeditious attainment and maintenance of the ozone standard or with reasonable further progress toward attainment. With respect to the RACT relaxation and add-on controls, the EPA added that OEPA had not considered the alternative that “it may be less expensive to control oven exhaust than spray booth exhaust” and that, even if the current emission limits were not economically and technologically feasible for sources like Navis-tar, there was no demonstration that Nav-istar’s proposed limits satisfied the RACT standard for those particular sources. Id. at 25,511. Navistar responded to the EPA’s proposed disapproval by filing detailed written comments with an attached technical report done by Roy F. Weston, Inc. (Weston Report). Navistar assailed the EPA’s reliance on the HOV study by citing portions of the study indicating that it was not intended to serve as “a site or facility-specific RACT determination or to establish RACT for coating heavy off-highway vehicle products.” Navistar also cited the study’s ac-knowledgement that “there is considerable variability within the HOV manufacturing industry that may affect the determination of reasonably available control technology for specific HOV facilities.” Navistar further asserted that there were substantial differences between the off-highway vehicle coatings in the HOV study and the high quality Navistar coatings that must meet more stringent customer specifications. In light of these observations, as well as the observation that the volumes of production and types of coatings were significantly different between the two types of facilities, Navistar disagreed with the EPA’s statement that “complying coatings are available and in use at similar coating facilities.” 53 Fed.Reg. 25,509, 25,511. Navistar also contested the EPA’s determination as to the feasibility of add-on controls by referring to the Weston Report’s conclusion that by the 1982 SIP compliance date no economically and technologically feasible controls were available. As to the costs of oven controls versus spray booth controls, Navistar responded that OEPA had concluded that it was fruitless to control the ovens due to the small amount of emissions from those areas. With respect to the air quality of the region, Navistar contended that there had been no ozone violations since 1983 and that the margin of attainment (the capacity for additional emissions which would not result in violation of NAAQS) was six times more than the increased emissions that would result from the Navistar variance. OEPA also submitted comments to the EPA’s notice of proposed rulemaking. OEPA noted that it did not consider the proposed variance to constitute a permanent relaxation of the Ohio SIP but, instead, viewed it as temporary and ending on December 31, 1987. OEPA also challenged the EPA’s argument that “a substantial change in the inventory” since the 1979 SIP, combined with post-1982 excee-dances, required a revised attainment demonstration. 53 Fed.Reg. at 25,510. The state referred to a technical report prepared by Q Source, Inc., submitted to the EPA earlier, that reported no significant change in the emissions inventory in the area. Although Ohio conceded that there had been exceedances of the ozone emissions standards since 1984, it argued that no violations of the standard occurred in the period from 1984 to 1987. With respect to the plan to extend compliance for three lines until they were phased out by December 81, 1987, OEPA disagreed with the EPA’s conclusion on expeditiousness. OEPA argued that Nav-istar had “conclusively demonstrated that low solvent coatings ... are not available for use and that it is not cost-effective to retrofit VOC emission control devices.” According to the state, the schedule to eliminate the three lines by the end of 1987 “must, therefore, be considered expeditious.” With respect to the EPA’s conclusion that the state had not demonstrated the lack of technical or economic feasibility of RACT, OEPA responded that it was not aware of any medium or heavy-duty truck manufacturers complying with the 3.5 pounds/gallons limit, either through low-VOC coatings or add-on controls. OEPA pointed out that the EPA had failed to identify any facilities similar to Navistar that were in compliance, and the state distinguished the HOV study on the basis of differences between the plant operations surveyed in that study and Navistar’s Springfield operations. In its Final TSD, issued in December 1988, the EPA noted that the early development of Navistar’s compliance efforts remained largely undocumented, despite the May 2, 1987, TSD statement that Navistar “must demonstrate that it proceeded expeditiously to develop and implement its compliance plan from the time the regulations were adopted to the present with no significant periods of inaction.” The EPA noted that the earliest indication of Navistar’s intent to replace three lines with new lines was November 1985, nearly three years after Navistar was required to comply with Ohio Administrative Code section 3745-21-09(U). The EPA also concluded the following: OEPA has not provided a complete survey of the availability of complying coatings and the compliance status of other similar sources. The State should provide evidence that it made all reasonable efforts to determine the availability of complying coatings or other kinds of control, as appropriate. Examples of these efforts include, but may not be limited to examining information that is or should be reasonably available to the State including whether sources operating in the State that were in an industry comparable to the source are using complying coatings, or other kinds of controls, that the source could adopt. Reasonable efforts also include seeking all information that is reasonably available to the source requesting the SIP revision. This would include contacting suppliers that the source uses or could [use], to determine if they have, or could develop, complying coatings or other controls. In addition, the State or source should contact regional or national trade associations for the industry, and review information the associations may have concerning coatings or other controls. The Final TSD further acknowledged Ohio’s and Navistar’s comments that the HOV study involved plants not comparable to Navistar’s operations, in that the HOV industry uses standard color coatings that do not have to meet the performance requirements of the high-quality finishes applied by Navistar. The EPA also acknowledged the differences in numbers and volumes of coatings used. However, the EPA emphasized that the reference to the HOV study in its notice of proposed rulemaking was not meant to suggest that Navistar must use the coatings used by the HOV industry. Instead, explained the EPA, because the two industries have similarities, some explanation of the possible differences in the operations that would make it not feasible for Navistar to use the coatings used by some HOV manufacturers was in order. The EPA pointed out that there was no showing that the coatings used by HOV plants would not meet the performance standards of Navistar, nor did Navistar or the state explain how the number and volume of coatings related to Nav-istar’s ability to use the complying coatings of the HOV industry. While the Final TSD concluded that Nav-istar had not demonstrated that complying coatings were unavailable, the staff did agree with Navistar that add-on control equipment was not economically feasible for the spray booths. However, the EPA maintained that Navistar still had not demonstrated that it would not be feasible to control oven exhaust. Acknowledging Navistar’s position that such controls would be ineffective due to the low VOC content of the oven exhaust, the EPA pointed out that the information submitted by Navistar failed to report the percentage of VOC emissions from five of the coating lines with ovens. While control of the bake ovens might not bring those lines into compliance, the EPA pointed out that such controls may still be a reasonable way to reduce emissions beyond the levels proposed in the variance. With respect to Ohio’s and Navistar’s contention that the greater Dayton area was in attainment of the ozone standard during the 1985-1987 period even without an emission reduction from Navistar, the Final TSD responded as follows: Although no violation of the ozone standard was monitored during the 1985-1987 period in the Dayton area, an ozone standard violation was monitored during the 1986-1988 period. In this period, the Urbana Road site in Clark County recorded seven exceedances of the standard, with five occurring in 1988. The Sprangler Road site in Clark County experienced six exceedances of the standard in the 1986-1988 period. Clearly Clark County and the rest of the Dayton area continue to experience violations of the ozone standard. More ozone precursor emission reductions are required in this area. Finally, the 1987 TSD addressed the objection to treating the variance for seven lines as a permanent relaxation. From the perspective of Ohio and Navistar, because the variance expired on December 31,1987, the revision should be considered a compliance date extension for all lines. The EPA provided the following response: The variances for the seven lines were treated as permanent relaxations because they do not require these lines to come into compliance with the SIP once the variances expire. Instead, the variances require Navistar to evaluate the feasibility [of] three different emission limitation[s] (one of which is a permanent relaxation). OEPA’s and RAPCA’s comments indicate that Navistar has submitted an alternative emission control strategy which is under review by OEPA. In addition, it should be noted that even if these variances could be considered compliance date extensions, they would still not be approvable. As discussed previously, it has not been demonstrated that compliance by December 31, 1987, is expeditious and that such an extension will not jeopardize attainment or maintenance. Finally, because the variances expired before they went into effect (they are effective upon approval by USEPA), even if they were approved as SIP revisions, they would have no effect. In conclusion, the Final TSD recommended that the proposed revision be disapproved “because the State has not demonstrated that Navistar [’s] compliance schedule is expeditious and that meeting the existing SIP limit is technically or economically infeasible.” In addition, “for both the extension and the relaxation, the state must demonstrate that approval will not jeopardize attainment or maintenance,” which the State has not done. For the same reasons enunciated in the Final TSD, the EPA finally disapproved Navistar’s SIP revision on September 13, 1989, 54 Fed. Reg. 37,795, and Navistar subsequently filed its petition for review. C. Related proceedings Pursuant to section 113(b) of the Clean Air Act, 42 U.S.C. § 7413(b), the EPA, on March 21, 1984, issued a notice of violations at Navistar’s assembly plant. 54 Fed. Reg. at 37,798. The United States instituted a civil proceeding for enforcement of that notice of violation on April 30, 1985. United States v. Navistar Int’l Transp. Corp., No. C-3-85-365 (S.D.Ohio, Rice, J.). On December 30, 1985, the EPA issued a second notice of violations at Navistar’s body plant. 54 Fed.Reg. at 37,798. A civil proceeding to enforce that notice of violation was instituted on November 30, 1986. United States v. Navistar Int’l Transp. Corp., No. C-3-86-541 (S.D.Ohio, Rice, J.). No decision has yet been rendered on these two enforcement proceedings. The EPA also instituted- administrative proceedings against Navistar under section 120 of the Clean Air Act, 42 U.S.C. § 7420, on September 24, 1984. The AU’s finding that Navistar had violated the Ohio SIP was upheld by this court. Navistar Int’l Transp. Corp. v. EPA, 858 F.2d 282 (6th Cir.1988), cert. denied, 490 U.S. 1039, 109 S.Ct. 1943, 104 L.Ed.2d 413 (1989). II. Our review of the EPA’s final action regarding the proposed revision to Ohio's state implementation plan is an extremely narrow one. The standards set out in the Act (APA), 5 U.S.C. § 706, are controlling. See Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 413, 91 S.Ct. 814, 822, 28 L.Ed.2d 136 (1971); South Terminal Corp. v. Environmental Protection Agency, 504 F.2d 646, 655 (1st Cir.1974). Those APA guidelines require us to determine whether the EPA followed the proper lawful procedures and acted within its statutory authority when it promulgated its final rulemaking and whether that rulemaking is constitutional. If so, we may set aside the EPA actions only if we find that the actual choices made were “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706(2)(A). National Steel Corp., Great Lakes Steel Div. v. Gorsuch, 700 F.2d 314, 320 (6th Cir.1983). See also Michigan v. Thomas, 805 F.2d 176, 181 (6th Cir.1986). The Supreme Court has explained the section 706(2)(A) review as follows: [T]he court must consider whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment.... Although this inquiry into the facts is to be searching and careful, the ultimate standard of review is a narrow one. The court is not empowered to substitute its judgment for that of the agency. Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 823, 28 L.Ed.2d 136 (1971) (citations omitted); see also Northern Ohio Lung Ass’n v. EPA, 572 F.2d 1143, 1148 (6th Cir.1978). Relying on the Supreme Court ruling that the EPA is “required to approve a state plan which provides for the timely attainment and subsequent maintenance of ambient air standards,” Train, 421 U.S. at 79, 95 S.Ct. at 1482 (emphasis in original), Navistar argues that the EPA has no choice but to approve its proposed SIP revision. According to Navistar, during the 22-month period from when Ohio submitted the SIP revision to the EPA up to the termination date of the variance— March 10, 1986 to December 31, 1987— Navistar operated according to the terms of the revision and no violations of the ozone standard occurred during 1986 and 1987. Therefore, according to Navistar, the SIP revision could not have interfered with the attainment or maintenance of NAAQS. See United States Steel Corp., 633 F.2d at 674 (“the logical inquiry for the EPA is to assess whether the proposed change interferes with attainment”) (emphasis in original). The EPA does not disagree with Navis-tar as to the standard to be employed, but does disagree with Navistar’s conclusion that the SIP revision is dealing with an area where the national standards have been met. The conflict, then, in part revolves around the method for determining whether the Dayton urban area — designated a nonattainment area in 1978 — remained a nonattainment area despite the 1979 SIP revision demonstrating compliance by 1982, and, if so, whether that nonattainment status meant that the attainment demonstration in the 1979 SIP revision was invalid, thus requiring Navistar or OEPA to submit a new attainment demonstration. Put another way, the question is what constitutes a sufficient demonstration of “attainment” for the purpose of achieving the standard of section 110(a)(2)(A). Navistar maintains that the EPA erred by using an improper method for calculating ozone NAAQ violations. According to the EPA, in order for an area to be in compliance with an ozone NAAQ, it may experience no more than three excee-dances for a given three-year period. Nav-istar points out that this standard allows the EPA to aggregate all exceedances at all monitoring sites in the AQCR, such that more than four exceedances over three years anywhere in the area constitutes a violation. Navistar contends that this is contrary to the EPA’s own regulations. We agree. Under the EPA regulations, four excee-dances at an individual monitoring site over three years establish noncompliance with a NAAQ. 40 C.F.R. pt. 50, app. H; see also 50 Fed.Reg. 42,735, 42,736 (1985) (“An area is considered in violation of the ozone NAAQS for any one year if the total number of exceedances, at a monitor, for that year and the previous two years is greater than three.”) (emphasis added); see also 54 Fed.Reg. 37,132 (1989). There is no dispute that in both 1982 and 1983 there were monitoring sites in the Dayton area that had more than three exceedances. (Q Source Engineering Report; OEPA letter of April 16, 1986). According to this data, the three-year periods 1982-1984 and 1983-1985 experienced an average number of exceedances per year of greater than one. Further, there is agreement that none of the individual monitoring sites in the greater Dayton area had more than one excee-dance per year over the period of 1984-1986 or 1985-1987. Consequently, when determining compliance for a given year by looking at that year and the preceding two years, we find that the greater Dayton area was in “violation” of the ozone NAAQS in 1984 and 1985, even though the data for those individual years demonstrates few exceedances, and the area was in compliance in 1986 and 1987. Thus, the EPA was correct in stating that “there were measured violations of the MAAQS for ozone in 1983 and 1984 in the Dayton area.” However, because the area was in compliance with ozone standards during 1986 and 1987 — the period covered by the proposed SIP revision according to Navistar — the question remains whether “the 1979 attainment demonstration may no longer be valid and a full up-to-date attainment demonstration would be required before the proposed SIP relaxation for [Navistar] could be considered.” (February 26, 1986, letter from EPA to OEPA). In its proposed disapproval, 53 Fed.Reg. 25,509, and its Final TSD, the EPA answered that question in the affirmative. We find the EPA’s answer reasonable in light of the relevant statutory provisions. The EPA judged the validity of the 1979 attainment demonstration according to the air quality for the period from December 31, 1982 (the deadline for compliance with the ozone NAAQS pursuant to Part D of the CAA and pursuant to the Ohio SIP itself) to December 31, 1987 (the date of the extension requested by Navistar for three lines and the date for submission of alternative control strategy for the other seven lines). This is not unreasonable given that the Act itself and the current approved Ohio SIP require compliance by December 31, 1982. Further, the proposed SIP revision would have authorized noncompliance with the existing VOC limitations from December 31, 1982, through December 31, 1987, not merely the 22-month period between the proposal’s submission and its extension date. Because there were both violations and exceedances subsequent to the 1982 compliance date, the EPA determined that the original prediction of attainment according to the current SIP was in serious doubt, and that it was necessary for Navis-tar’s source-specific SIP revision to be accompanied by a revised attainment demonstration. The EPA consistently indicated to Navistar and Ohio its requirement for an up-to-date attainment demonstration, which requirement is consistent with the EPA policy statements issued in 1983 and 1986 regarding SIP revisions seeking compliance date extensions for VOC sources in ozone designated nonattainment areas. (See Memorandum from J. Craig Potter to Regional Administrators, “Policy on SIP Revisions Requesting Compliance Date Extensions for VOC Sources”; Memorandum from Sheldon Meyers to Regions I-X, “Source Specific SIP Revisions”). We find this policy to be a reasonable implementation of the statutory requirement that a state plan must provide for the timely attainment and subsequent maintenance of ambient air quality standards. Indeed, it is difficult to perceive how the EPA could determine whether an individual compliance date extension would interfere with timely attainment and maintenance of a standard without a valid and approvable attainment demonstration. Requiring a revised demonstration does not preclude approval of a SIP revision, but merely allows for more effective review. However, where an up-to-date attainment demonstration is required and, as here, one is not submitted in support of the proposed SIP revision, the EPA could not conclude whether the revision provided for attainment or not and properly disapproved the proposal. III. Even were we to find that the EPA abused its discretion in requiring a revised attainment demonstration, Navistar would still have to show that its proposed revisions were consistent with the statutory-requirement that nonattainment area SIPs provide for the “implementation of all reasonably available control measures as expeditiously as practicable.” CAA § 172(b)(2), 42 U.S.C. § 7502(b)(2). The EPA found that Navistar did not meet this requirement, and we decline to rule that the EPA acted arbitrarily or capriciously in making that finding. Navistar argues that the requirements of section 172 (a Part D section) apply only to comprehensive SIP revisions and do not apply to site-specific SIPs. Therefore, according to Navistar, the RACT requirements of sections 172(b)(2) and (3) — expeditiousness and reasonable further progress — do not apply to Navis-tar’s proposed revision, and the EPA was not authorized to reject the revision on the basis that Navistar did not demonstrate RACT compliance. Indeed, Navistar points to a case where the court, taking the EPA’s position in that case, held that section 172 was designed to govern only comprehensive plan revisions under Part D of the 1977 amendments. United States Steel Corp., 633 F.2d at 675. The Third Circuit explained its holding as follows: Section 172 was designed to enable states that had failed to achieve the NAAQS by the deadline imposed in the original Act to submit a new comprehensive plan, in order to avoid the drastic consequence of closing existing facilities. Thus, when Part D and § 172 speak of “revised” plans, they contemplate a new and complete plan replacing the original SIP, rather than the minor revisions altering or updating particular emission standards which are covered by § 110(a)(3). The conclusion that only a comprehensive, redrawn SIP must meet § 172 requirements is underscored by § 129 of the 1977 amendments, which imposes the obligation to submit a § 172 plan. Section 129(c) provides that: ... each State in which there is any non-attainment area (as defined in Subpart D of the Clean Air Act) shall adopt and submit an implementation plan which meets the requirements of section 110(a)(2)(I) and subpart D of the Clean Air Act not later than January 1, 1979. (Emphasis added) (Uncodified). Id. (emphasis in original) (footnote omitted). Although the Third Circuit’s analysis was appropriate to the circumstances of the case before it, we decline to apply its holding as broadly as is urged. In United States Steel, the EPA was reviewing coke oven regulations as a revision to a pre-amendment SIP, before Pennsylvania had complied with the 1977 CAA amendments by submitting its comprehensive Part D state implementation plan. In proposing changes in the coke oven emission regulations, Pennsylvania represented to the EPA that the revision did not represent the revision package required for nonattainment areas and that its Part D plans would be forthcoming at a later date. United States Steel, 633 F.2d at 675. Thus, as the Third Circuit noted, Pennsylvania was proposing a revision to a plan covered only by section 110(a)(3), and, where Pennsylvania had not yet submitted and implemented a Part D plan, the revision proposal was not subject to the requirements of Part D. However, in the case before us now, Ohio and Navistar propose a revision to a Part D SIP that was approved according to the requirements of section 172. It would be a strange course of action to require Ohio to implement SIPs that meet the requirements of Part D, as Congress did when amending the Act in 1977, and then evaluate any further revisions to those SIPs without requiring continued adherence to the section 172 requirements in designated nonattainment areas. We believe that common sense and the purpose and design of both Part D and the CAA as a whole require that proposals to revise Part D SIPs be evaluated in accordance with the criteria by which the Part D SIPs were approved. While it is true that the Administrator is not to “concern himself with factors other than those specifically enunciated” in section 110(a)(2), Northern Ohio Lung Ass’n, 572 F.2d at 1147, that section includes subsection (a)(2)(I), which links Part D to the SIP revision process. Navistar next argues that, even if section 172 applies to its request for a compliance extension, it believes that it has met the requirements of that section. Nav-istar’s revision sought to extend for five years, from December 31, 1982, through December 31, 1987, Navistar’s time to bring ten of its coating lines into compliance with the Ohio SIP’s limits on YOC emissions. Based on the length of the requested compliance date extension, the EPA “closely scrutinized” the proposal for expeditiousness and required an examination of the compliance status of other sources in the same VOC category and an examination of the most expeditious means of compliance available. From the outset of the EPA review process, Ohio was on notice that the EPA required Navistar to document that it had proceeded expeditiously. Further, OEPA and Navistar knew that the EPA considered the proponents of the revision to have the burden of demonstrating that the timeframe for RACT compliance in the original Part D SIP did not allow enough time for Navistar to come into compliance. Navistar does not argue that it or Ohio conducted and submitted the comprehensive survey requested by the EPA, nor does it argue that the EPA erred by imposing that requirement. Rather, in its comments to the proposed disapproval, Navis-tar merely noted that it had examined the compliance status of an unspecified number of other sources, including two truck manufacturers. The EPA did not act in an arbitrary or capricious fashion when it concluded that this showing fell short of the obligation to provide “a complete survey of ... the compliance status of other similar sources.” 54 Fed.Reg. at 37,797. Even before Navistar’s proposed SIP revisions were formally presented to it for review, the EPA notified the state that Navistar had to show that it proceeded expeditiously to develop and implement compliance plans from the time the regulations were adopted to the present with no significant periods of inaction. Throughout the review period, the EPA continued to express its concern with Navistar’s failure to document its early compliance efforts. For example, the agency noted that Navistar had shown that it had been testing paints “fairly consistently since 1984” but had provided little information concerning work done before then. The EPA also expressed concern that the earliest indication of Navistar action toward replacement of old lines with new lines dated from November 6, 1985, nearly three years after Navistar was required to be in compliance with section 3745-21-09(U). In response, Navistar asserts that it conducted an “extensive series of coating development efforts and tests ... beginning in 1979.” Navistar also points to the OEPA finding that Navistar’s program of testing coatings was “active and continuous” between 1979 and 1986. However, the EPA’s finding that Navistar had failed to demonstrate expeditiousness was not a finding with respect to the reformulation and testing of coatings. Rather, the EPA found that Navistar had failed to demonstrate that it had been expeditious in developing and carrying out the plan to replace three of its old coating lines with new lines. 54 Fed.Reg. at 37,796. Navistar implies that, because of the state of technology, it was not possible to develop such a plan until 1985. However, no record evidence is cited to support that proposition, and, in the absence of such evidence, the EPA’s finding that Navistar failed to demonstrate expeditiousness is rational. IV. Unlike compliance extensions, which are based on the assertion that more time is needed to comply with RACT-level emission limits than was contemplated under the original SIP, a request for a relaxation of RACT contends that the relevant emission limit (in this case, 3.5 lbs. VOC per gallon of coating, excluding water) is not technologically and economically feasible— that the existing limit, purported to be RACT under the current SIP, is not in fact RACT. Therefore, under these standards, Navistar must not only show that it is not technologically and economically feasible for it to comply with the emission limit provided for in the current SIP, but also— due to the CAA’s command that nonattainment plans provide for the implementation of RACT and reasonable further progress toward reduction in emissions through RACT, 42 U.S.C. § 7502(b)(2), (3) — the revision to the SIP must be RACT, the best Navistar can do. Colloquially, Navistar must show both that the existing emission limit “is not RACT” and that its variance “is RACT.” Navistar first argues that these standards do not apply because the revision seeks only a compliance extension, not a relaxation of RACT. Navistar contends that the EPA treatment of the revision as a permanent extension for seven of the coating lines is contrary to the plain language of the proposed revision, which states that “[t]his variance will expire on December 31, 1987.” Acknowledging these arguments in the Final TSD and in its final disapproval of the revision, the EPA concluded that the proposed SIP revision was a permanent relaxation of the current SIP, as applied to the seven coating lines not to be shut down, because those lines were not required by the terms of the proposed revision to come into compliance when the proposed variance expired on December 81, 1987. The agency conclusion that the proposed variance did not commit Navistar to any compliance by that date is neither unreasonable nor a clear error of judgment. In fact, the variance required only that Navistar submit a feasibility study by that date, which study was to include a determination of the feasibility of either complying with section 3745-21-09(U) or simply continuing under the emission limitations in the variance. Concerning the technical feasibility of complying coatings, Navistar and Ohio stress the unique nature of Navistar’s operations and attack the EPA’s reliance on the Illinois HOV study. Navistar points out that the HOV study emphasized the site-specific differences in coating operations, which differences precluded general conclusions about the feasibility of compliance with emission limitations. Navistar also relies on the HOV study’s statement that owners of coating operations might have difficulty finding acceptable coatings. Thus, according- to Navistar, the EPA relied on a study that was both inapplicable to Navistar’s “unique” operations and contrary to the conclusions reached by the EPA concerning the feasibility of compliance by Navistar. However, as the administrative record demonstrates, the implication of the HOV study was merely one of several factors that the agency evaluated in determining whether Navistar had adequately demonstrated the infeasibility of compliance with the existing VOC limit. Moreover, the EPA expressly noted that it was “not suggesting that Navistar must use the coatings used by the HOV industry” but, rather, was asking for “some explanation of possible differences in the operations that would make it infeasible for Navistar to use the coatings used by some HOV manufacturers.” 54 Fed.Reg. at 37,797. Because Navistar failed to make such a showing, and because the EPA found that Navistar “has not demonstrated that it has been unable to reformulate a sufficient number of its coatings to achieve compliance with the applicable rule on a daily basis,” 54 Fed.Reg. at 37,798, the EPA neither overstepped the bounds of its authority nor acted unreasonably when it found that the company failed to demonstrate that compliance was not technically feasible. “While the state does have the primary responsibility for establishing RACT guidelines, the EPA’s statutory obligation before approving a SIP is