CUDAHY, Circuit Judge.
[1142]*1142In 1970, Congress passed the provisions that form the basis of what we know today as the Clean Air Act. Pub.Law No. 91-604, 84 Stat.1676 et seq. (1970) (codified at 42 U.S.C. § 7401 et seq.). At the same time the Environmental Protection Agency (“EPA”) was created by Executive Order. 35 Fed.Reg. 15623 (1970).
In 1977 Congress amended the Act, adding, among other things, a Part D: “Plan Requirements for Nonattainment Areas.” §§ 171-178, 42 U.S.C. §§ 7501-08. The notion of a nonattainment area was a new one introduced with the 1977 amendments; the idea was to allow the construction of new facilities even in areas not in attainment with national standards for pollution, provided that certain conditions were met. It also required the imposition of certain controls — for example, inspection and maintenance programs for vehicles — in areas that continued to fall short of the standards after certain deadlines.
The standards were to be set by the EPA, and ways of determining attainment and nonattainment were also up to EPA. The states, on the other hand, were responsible for determining the geographic boundaries of the attainment and nonat-tainment areas, subject to EPA’s approval; and they were also responsible for submitting to EPA, in 1978, a list of all such areas, with an indication, for each area, whether it was in attainment with respect to the standards. Since the standards were different for different pollutants, and since the appropriate size of the areas might differ for different pollutants, the states really submitted a number of lists, one for each pollutant. The EPA undertook to provide guidance as to the appropriate size of the areas for different pollutants, and was in a position to enforce those guidelines, since final approval of the lists rested with it.
When a nonattainment area reached attainment (according to EPA criteria), the state was entitled to revise and resubmit its list. If EPA agreed with the redesig-nation to attainment, it approved the revision. If it did not, it was required to turn down the request within a certain time period, to give reasons for the refusal and to invite comments before taking final action. (§ 107(5), 42 U.S.C. § 7407(5); § 107(2), 42 U.S.C. § 7407(2).)
In 1978 Illinois submitted its lists and they were approved. The list for ozone— or rather, at that time, for photochemical oxidants1 — divided the Chicago urban area into counties, all of which were listed as nonattainment.
In 1983, Illinois revised the list to show that Will and McHenry Counties, two counties near Chicago, were in attainment, and EPA approved the change. A short time later Illinois proposed to upgrade Kane and Du Page Counties, two counties closer to the geographical center of Chicago. This time EPA denied the proposal. By way of explanation EPA did not claim that the counties had failed to meet the ozone standard; instead, EPA pointed to the ozone problem in the larger Chicago urban area. Appeals were taken (directly to this court under § 307(b)(1) of the Clean Air Act, 42 U.S.C. § 7607(b)(1)) both by the Illinois Environmental Protection Agency, the agency entrusted with carrying out the state’s part in the regulation process, and by the Illinois Chamber of Commerce.2
[1143]*1143I
Ozone is a significant component of the health-endangering smog that is one product of modern industry and transportation. Ozone can cause serious respiratory problems; when it reaches levels of moderate concentration in the air around us, normal men and women may begin to cough and may feel their chests tighten; they may begin to experience eye irritation; the very sensitive may have asthma attacks. At higher concentrations hospitalization becomes a possibility. Ozone is one of the substances that make up what we call air pollution, and Congress has given the Environmental Protection Agency the task of bringing it under control.
Ozone, unlike many other pollutants, is not discharged directly into the atmosphere by the offending source. Ozone production depends on the emission of volatile organic compounds3 which, under the appropriate meteorological conditions, react to form ozone. Since this reaction requires a certain amount of time, ozone concentrations are typically not found overhead of the sources that produce them, but rather downwind.
This process of ozone formation is now rather well documented. Scientists were at first surprised to find ozone concentrations over remote rural areas; there was some speculation that nature itself might be responsible for these formations. But at least by the 1970s it was clear that ozone in rural areas was largely due to the airborne “transport” of ozone precursor chemicals— volatile organic compounds — together with meteorological conditions that eventually would support the production of ozone. In general, ozone appeared to reach peak concentrations downwind of the sources.
Thus, studies carried on in 1973 showed that sources in Houston were responsible for ozone concentrations about 15-25 miles downwind, and showed similar, though not so pronounced, results in the Phoenix area. Martinez & Meyer, Urban-Nonurban Ozone Gradients and Their Significance, Symposium Paper, March 12, 1976 (reprinted in Appellee’s Appendix at A-76). In 1974, peak values of ozone were found to arise 5 to 15 miles downwind of Columbus, Ohio. Id. A 1975 study of sites in the Pennsylvania and New Jersey areas tended to show that ozone concentrations would occur 27 to 49 kilometers downwind of the source areas. “This accounts for some of the ozone previously observed in the nonindustrial, low-traffic density area of Ancora, N.J., where concentrations of primary pollutants are low, but where ozone daily max-ima frequently exceed the federal standard. Thus, photochemical air pollution [ozone] in this area is a regional rather than a local phenomenon, and ozone resulting from emissions from the urban complex is widespread and not confined to the complex itself.” Kleiner, Transport of Photochemical Air Pollution, 9 Environmental Science and Technology 886 (1975). See also EPA 600/3-77-017, Proceedings of a Symposium on 1975 Northeast Oxident Transport Study, Research Triangle Park, N.C. (1976); EPA, Mt. Storm, W.Va.-Gor-man, Md.-Keyser, W.Va.' Air Pollution Abatement Activity, No. APTD-0656 (April 1971).
Given these studies, it is natural to suppose that the Chicago-northwest Indiana area contributes much of the ozone pollution that occurs along the western shore of Lake Michigan; and indeed one study concludes that that is precisely the case: along Lake Michigan, through Milwaukee and beyond, “Chicago-northwest Indiana acts as a giant volume source resulting in a significant fraction of the ozone monitored on alert-level days.” Lyons & Cole, Photochemical Oxidant Transport, 15 J. Applied Meteorology 733 (1976). In 1976 [1144]*1144hourly average values as high as .30 ppm (parts per million) (the national standard is now set at .12 ppm), with excursions above .40 ppm, recorded in southeastern Wisconsin were attributed by the study to sources in the Chicago area. Nevertheless, we will not, in this opinion, make any factual assumptions about the direction in which ozone is transported away from the Chicago area; our only assumption, one which is not denied by either party, is that ozone tends to show up at varying distances downwind from the source.
II
A.
Prior to the passing of the 1977 Amendments to the Clean Air Act, the EPA had developed an “offset” policy to moderate some of the harsh consequences of a state’s failure to attain a standard set under the 1970 Act. 41 Fed.Reg. 55528-29 (1976); D. Currie, Air Pollution 6-2 (1981). Instead of prohibiting the construction of all new facilities in areas that had not attained federal standards, the EPA was prepared to permit such construction provided that the pollution produced by the new source was at least offset by reductions in emissions from existing sources. Because of the regional nature of the ozone problem, EPA advocated allowing ozone created by new sources to be offset by reductions anywhere within a large, urban, ozone-producing area. EPA, Effectiveness of Organic Emission Control Programs as a Function of Geographic Location (April 1977 memorandum) (reprinted in Appellee’s Appendix at A-249). EPA saw little benefit from focusing VOC (volatile organic compounds) programs on urban areas with fewer than 200,000 inhabitants; instead, ozone control was aimed at the larger areas. In the memorandum just cited, a map indicated the urban-ozone problem areas with shaded circles; Chicago and northwest Indiana were together in one large shaded area. Of the offset policy, the memorandum says that
a major new VOC source locating within one of the shaded areas should be required to obtain emission offsets from existing sources within that same area.
Id. at 32 (emphasis in original). The assumption that the offset policy would apply in all urban areas — that is, the assumption that urban areas failed to meet the ozone standard (at the time, .08 ppm) — was not based on any monitoring of the air in those areas, but on general knowledge of the process of ozone production.
One important reason for the 1977 Amendments to the Clean Air Act was to provide a statutory basis for the offset policy. See D. Currie, supra, at 6-2. This Congress did by introducing the notion of a “nonattainment area.” § 171, 42 U.S.C. § 7501. Each state was, with EPA approval, to provide a list dividing the state into areas — perhaps a different division for different pollutants — showing for each area whether it was in attainment of the standard for a certain pollutant; nonattainment of the standard; or as of yet unclassifiable. § 107(d)(1), 42 U.S.C. § 7407(d)(1). Within the nonattainment areas provisions of the statute based on the offset policy were to apply. See, e.g., § 173, 42 U.S.C. § 7503.
It was up to the states to determine the appropriate geographic size of these areas; but EPA had to approve and promulgate the list, and it was up to EPA to offer guidance as to what sizes would be acceptable, and what sizes would not. H.R. Rep. No. 294, 95th Cong., 1st Sess., 1977 U.S. Code Cong. & Ad.News 1077, 1292; § 107(d), 42 U.S.C. § 7407(d). For some pollutants, EPA discharged this duty with a fair amount of precision. Thus, for sulfur dioxide:
Generally where EPA promulgated a designation for S02 the minimum area was to be the county in which the violating monitoring site was located.
And for carbon monoxide:
A designation of nonattainment for the entire urban core area ... was desirable, but smaller area designations were ac[1145]*1145ceptable since CO violations are most pervasive in downtown areas----
43 Fed.Reg. 8962-63 (1978).
But in the matter of ozone, perhaps because of the essentially different way ozone is produced, EPA gave no clear guidance. There was talk that suggested that an entire urbanized area should be considered one nonattainment area for ozone, because all of the sources in a city were assumed to contribute to the ozone problem in the area. 44 Fed.Reg. 20376 (1979); 43 Fed.Reg. 45997 (1978); 43 Fed.Reg. 8962 (1978). See also Raffle, Env.Rep. (BNA) Monograph # 27 (1979) (most major urban areas in the U.S. designated as nonattainment areas); 9 Env.Rep. — Current Developments (BNA) 1812 (1979) (most cities nonattainment in spite of lowered standard). On that theory, the larger Chicago urban area should have been one nonattainment area; and yet when Illinois submitted its first list, the Chicago urban area was divided into counties, each county listed as — apparently—a separate nonattainment area. 43 Fed.Reg. 8962, 8988-89 (1978). EPA also advanced a slightly different theory, that a nonattainment area for ozone must be large enough to include both the polluted area and all major sources of the ozone pollution in that area, a theory obviously rooted in the offset theory out of which the notion of a nonattainment area had come. See Memorandum from David Hawkins to Regional Administrators of USEPA, January 3, 1978 (reprinted in Ap-pellee’s Appendix at A-39) (for ozone, regional offices should require “that the designated area be of sufficient size to include most of the hydrocarbon sources”); 46 Fed. Reg. 55724 (1981) (“A nonattainment area should be as small as possible while encompassing all areas of expected violations and all sources of significant impact on those violations.”); 48 Fed.Reg. 46084 (1983) (“[T]he designated nonattainment area should be of sufficient size to include most of the significant hydrocarbon sources.”) But it is clear that EPA never took that theory seriously. It never proposed that southeastern Wisconsin and Chicago be part of the same nonattainment area, for example, as it should have under the theory, if Chicago sources cause Wisconsin ozone. This theory is also inconsistent with the “urbanized area” theory, since on the urbanized area theory the peak ozone concentration area, miles downwind, would not be included in the nonattainment area for a city; but under the “polluted area plus sources” theory, it must be. Again, under this theory EPA should not have approved the division of the area into counties, since pollution in one county may be caused by sources in another. And finally, to top off the confusion, in 1983 EPA granted a change in status from nonattainment to attainment for Will and McHenry counties, 48 Fed.Reg. 21947 (1983); yet later EPA conceded that Will County, at least, contained significant sources of VOC emissions contributing to ozone pollution in the Chicago area. 49 Fed.Reg. 24130 (1984). If sources are to be in the same nonattainment area as the pollution they produce, Will County ought to be in the Chicago nonattainment area.
We might have supposed that EPA was not serious about either theory, having allowed Illinois to treat the separate counties in the Chicago metropolitan area as separate nonattainment areas, if it were not for the EPA action that gives rise to the present litigation. For EPA has refused to upgrade two counties in the Chicago area, in spite of the fact that no violations have been monitored in those counties.
B.
Section 107(d)(5) of the Clean Air Act, as amended in 1977, permits states to propose a revision in the status of nonattainment areas to the EPA:
A State may from time to time review, and as appropriate revise and resubmit, the list required under this subsection. The Administrator shall consider and promulgate such revised list in accordance with this subsection.
42 U.S.C. § 7407(d)(5). A change in status is called for when an area is in attainment [1146]*1146with the national standard set by EPA. For ozone the following standard applies:
The level of the national primary and secondary ambient air quality standards for ozone ... is 0.12 part per million____ The standard is attained when the expected number of days per calendar year with .maximum hourly average concentrations above 0.12 part per million ... is equal to or less than one____
40 C.F.R. § 50.9 (1979). As proof of attainment EPA requires data showing no violations over the previous year, or on the average less than one full day in violation per year, over three years. Violations are measured at monitoring sites. Guidelines for the Interpretation of Ozone Air Quality Standards (EPA 450/4-79-003).
In January, 1983, Illinois submitted a revised list in which Kane and Du Page Counties each would have been upgraded to attainment status for ozone, along with data showing that no violations had been monitored in those counties. EPA refused to approve the change. In the proposed rulemaking that denied Illinois’ request, EPA said:
Because the prevailing winds during the ozone season are from south through west, ozone precursor emissions from DuPage and Kane Counties can contribute significantly to ozone NAAQS [national ambient air quality standards] ex-ceedances which continue to be observed in the Chicago area____ These counties must continue to be considered as part of the Chicago urbanized ozone nonattainment area for the purpose of the 1982 [state implementation plan].
48 Fed.Reg. 46084. Later, in the final rule-making, answering objections raised by the state, EPA continued to rely on two separate theories, first, that a nonattainment area must include all the sources that contribute to pollution in that area; and second, that an urban ozone nonattainment area must include the entire urbanized area:
[I]t is important that the nonattainment areas be of sufficient size to include all emissions sources ... which contribute significantly to the violation of NAAQS.
49 Fed.Reg. 24130.
EPA’s determination is based upon the fact that these counties are part of major urbanized areas____
49 Fed.Reg. 24130. As we have pointed out already, these theories are not only apparently inconsistent with one another (the first would require Chicago sources to be part of the southeastern Wisconsin non-attainment area, if EPA meteorological information is right, whereas the second evidently would not), they are both inconsistent with the action EPA took in originally approving the Illinois ozone list for the Chicago area by counties.
Illinois, having gotten its list by counties approved in 1978, and having gathered data to show that two of the original nonattainment areas, Kane and Du Page Counties, are now in attainment, brings this appeal to protest the EPA action. (Although separate actions were brought by the Illinois Environmental Protection Agency and the Illinois Chamber of Commerce, for simplicity we will treat their arguments together as arguments of the Illinois litigants.) Illinois argues that, since no violations have been monitored in Kane or Du Page Counties, and since those counties have been approved as separate attainment-nonattainment areas, EPA is either basing a nonattainment status for those areas on air quality monitored in other areas; or else it is now trying to change the borders of the nonattainment areas to make all of the Chicago area — including Kane and Du Page Counties — one nonat-tainment area. If it is doing the first, the state argues, then it is doing something not authorized by the statute, which authorizes the EPA to use only local air quality to determine the status of an area; and in addition, according to Illinois, EPA has not shown that emissions in Kane and Du Page actually contribute to the ozone problem in the core Chicago area and downwind. If, on the other hand, it is trying to change the borders, then under Bethlehem Steel v. EPA, 723 F.2d 1303 (7th Cir.1983) (as lili-[1147]*1147nois reads it), it is doing something it may not do. As the state sees it, EPA’s authority over the boundaries of nonattainment areas is limited to modifying changes in those boundaries proposed by the states, and does not extend to changing the boundaries when the state has proposed a change in attainment status.
C.
Since the statute does not require a hearing before EPA promulgates a rule denying a state’s redesignation request, the proper standard of review, under § 706(2)(A) of the Administrative Procedure Act, 5 U.S.C. § 706(2)(A), is the “arbitrary and capricious” standard.
In this case EPA announced a proposed rulemaking that was apparently in conflict with earlier decisions. EPA is entitled to change its policy, but it must do so on the basis of a reasoned analysis. Motor Vehicle Manufacturers Assoc. v. State Farm Mutual Automobile Ins., 463 U.S. 29, 57, 103 S.Ct. 2856, 2874, 77 L.Ed.2d 443 (1983). If EPA changed its policy, it did not say so; and neither did it explain how its justification of its most recent action could be made consistent with its earlier actions, if no change in policy was involved. It must do one or the other. If it has changed its policy, it must explain how and why; if it has not, it must articulate an explanation that will account for both the earlier and the most recent actions it has taken. Until it has done one or the other, its actions will appear to be arbitrary, and rational review in this court will be impossible.4
Ill
We are aware of at least four different ways in which EPA might justify the action it has taken. Each is more or less consistent with the available evidence concerning ozone; but each is also inconsistent with something EPA has said or done in the past. It is not our duty, in reviewing EPA’s action, to pick and choose among the various justifications that might be given. State Farm, 463 U.S. at 43, 103 S.Ct. at 2866. We feel obliged, however, to say enough about the various theories to show how they fall short of making a recognizable pattern out of the pieces in this puzzle, [1148]*1148and to suggest how those shortcomings might be remedied.
Although the choice of a theory rests ultimately with EPA, we have already seen that one of these theories does not have much to recommend it: the theory which would make a polluted area, together with all areas containing sources contributing to pollution in that area, one nonattainment area. For one thing that theory does not explain the original division into counties, or the separate upgrading of Will and McHenry Counties, acknowledged sources of pollution in the Chicago area. For another thing, in the case of ozone the areas would have to be large; and since ozone pollution occurs downwind of the sources, in some cases the polluted area itself might contribute nothing to the pollution. It should be within EPA’s discretion to allow the upgrading of such areas (something it could not do if both the polluted area and the source made up one nonattainment area). Moreover, our beliefs about the movement of ozone and ozone precursors are subject to revision, as the environment and our knowledge of the environment changes; but the limits of a nonattainment area ought not to be subject to constant revision. And finally a look at the boundaries originally approved by EPA in the eight largest metropolitan areas shows that, with the exception of the Chicago area, the boundaries were large enough in each case to include the entire urbanized area, but that no attempt was made to draw up the boundaries in such a way as to track the movement of ozone away from the sources, as this theory would require. 40 C.F.R. §§ 81.333 (New York), 81.305 (Los Angeles), 81.314 (Chicago), 81.339 (Philadelphia), 81.323 (Detroit), 81.305 (San Francisco), 81.322 (Boston), 81.309 (Washington, D.C.). Hence the theory fails to explain not only what EPA did in the Chicago area, but what it did in those other areas as well.
The remaining three theories are more plausible. On the assumption that EPA did approve the counties as separate nonat-tainment areas in 1978, EPA may now be arguing that ozone attainment must be determined at monitors downwind and outside the area itself; or EPA may be arguing that attainment of the ozone standard is to be determined on the basis of ozone precursors monitored within the area itself. Both these theories are consistent with EPA’s desire to make the source area a nonattainment area, without involving the apparently unworkable notion that a polluted area together with all its sources must make up one nonattainment area. Under either of these theories, the present county boundaries remain the boundaries of the nonattainment areas in Chicago, but the attainment status of those areas will depend on ozone pollution elsewhere.
On the other hand, EPA may be arguing that it never approved the division of the area into counties, and that Kane and Du Page are part of a unitary Chicago nonat-tainment area. This theory requires an explanation of the original listing, and for the fact that Will and McHenry Counties, arguably part of the larger Chicago area, were upgraded individually.
The first of the theories — that attainment of the standard in one area may be determined by the presence of ozone in another area — is not inconsistent with the statute. The relevant provision is a definition:
The term “nonattainment area” means, for any air pollutant, an area which is shown by monitored data or which is calculated by air quality modeling (or other methods determined by the Administrator to be reliable) to exceed any national ambient air quality standard for such pollutant.
§ 171(2), 42 U.S.C. 7501(2). Nothing in the statute says that monitoring must be done within the area itself; indeed, the facts suggest that the best, and perhaps only, way to monitor for ozone would be to monitor downwind. (The word “ambient” is defined in the statute to mean roughly the same as “outdoor;” it does not carry the connotation of “local.”) Since the statute gives the Administrator a certain amount of discretion (“other methods determined [1149]*1149by the Administrator to be reliable”), and since other provisions of the Act do not limit that discretion, it would seem to be within the authority of the agency to monitor ozone production at sites downwind.5
EPA has not set down standards for this sort of off-location measuring, however; existing guidelines suggest a determination of status by means of monitoring within the area. See Guideline for the Interpretation of Ozone Air Quality Standards (EPA 450/4-79-003). If EPA has something different in mind, it should make clear how ozone pollution is to be measured downwind, and that is something it so far has not done.6
Something similar is true of the second theory, which would base attainment status on the measurement of ozone precursors within the area itself: as the monitoring guideline is currently written, it is written in terms of ozone, and not ozone precursors.
Nevertheless, the guidelines are within the power of EPA to change. The current state of scientific knowledge suggests that it would be worse than foolish to start with geographically small areas, for ozone, and then base the attainment status of each area on the monitoring of ozone within that area. Attainment and nonattainment are meant to reflect the contribution of an area to the pollution problem; and perhaps the worst way to determine the contribution of a small area to the ozone problem is to measure for ozone within the area itself. It cannot have been the intention of Congress that EPA choose the worst possible way to measure for the production of pollutants, and nothing prevents EPA from changing its guidelines to reflect one of these views.
What we have said so far is on the assumption that EPA in fact approved a division of the Chicago area into counties. As we will see, that is not so clear as Illinois would have us believe. EPA has argued, and we find the argument persuasive, that although the Chicago area is listed by counties, it has been clear all along that insofar as any county was part of the urbanized area or its fringe of development it was to be part of a single nonattainment area, not to be upgraded until the entire area reached attainment.
Suppose, for example, Illinois had submitted the original list with Du Page County marked as “attainment” because monitors had shown no violations. It is clear as these things can be that EPA would have disapproved the rating, on the grounds that Du Page was, for purposes of ozone pollution, part of the larger Chicago area. That is precisely what happened with Porter County, Indiana. When EPA solicited comments on the original list, there was an objection to the nonattainment status of Porter County in northwest Indiana. In its reply, EPA did not allege that there had been monitored violations in Porter; instead, it said:
The nonattainment designation of Porter County was based on the fact that portions of Porter County are part of the Chicago-northwest Indiana urbanized area.
43 Fed.Reg. 45997 (1978). EPA, as we have seen, had suggested that all urbanized areas were to be treated as units, for ozone, and were to be assumed at the outset not to have attained the standard.
Thus, the original listing by counties was very possibly no more than an accident of record-keeping. Congress demanded a list from the states. The EPA asked the states to treat urban areas as unitary nonattainment areas for ozone, as it was authorized to do. Illinois submitted a list, by counties, [1150]*1150in which all the counties in the Chicago urban area — and thus the urban area itself — were listed as not in attainment. Although such a list, out of context, is ambiguous between the two interpretations — is Chicago one area consisting of the listed counties, or is each county a separate area? —the circumstances surrounding the promulgation of the list suggest that EPA’s interpretation may be the proper one. When EPA proposed to turn down the request to upgrade Kane and Du Page Counties, it used language that echoed the 1978 comment explaining Porter County’s status:
Because the prevailing winds during the ozone season are from south through west, ozone precursor emissions from DuPage and Kane counties can contribute significantly to ozone NAAQS excee-dances which continued to be observed in the Chicago area____ These counties must continue to be considered as part of the Chicago urbanized ozone nonattainment area for the purpose of the 1982 SIP.
48 Fed.Reg. 46084. If we could be certain that this was indeed EPA’s practice all along, we would be unwise to reverse agency action on the basis of the happenstance of a particular way of listing the Chicago urban area in a 1978 list. This “urbanized area” theory would also explain the different treatment of Will and McHenry Counties which, although EPA concedes they contain significant sources of ozone precursors, do not — unlike Kane and Du Page County — contain any part of the Chicago urbanized area, as defined by EPA on the basis of the 1970 census.
While this last theory may be the least problematic under existing standards, it suffers from problems of its own. Although large cities are prime contributors to the problem of ozone, so that it makes sense to impose ozone controls on them, there remains the question of how the attainment status of an urban area is to be changed. Will it depend on monitoring within the area itself? For reasons we have explained, a good deal of the ozone created in the Chicago area apparently finds its way downwind into Wisconsin. Although making the area larger reduces the problem — much of the ozone is also found within the urban area itself — the theoretical basis for monitoring for ozone only within an urban area as defined by the census is not presently clear enough to allow for proper review of EPA’s action in this Court.
IV
Illinois argues that there are other ways for EPA to do what it feels must be done without denying Illinois’ application to upgrade Kane and Du Page Counties. The fact that the parties have brought this litigation to the court of appeals shows that neither side believes that any of the alternatives will provide exactly equivalent results. If EPA could get the same results without blocking the upgrading of Kane and Du Page Counties, then Illinois would have nothing to gain, and EPA nothing to lose, however the appeal came out. Clearly the outcome of the appeal does make a difference; for one thing, certain controls apply to nonattainment areas by operation of the statute. It is essential to the success of the Clean Air Act that these controls apply in the appropriate areas, and we refuse to construe the law in such a way that the controls will apply to areas that suffer from but do not produce ozone pollution but not to areas that produce but do not suffer from such pollution. The automatic imposition of controls gives EPA a certain leverage over the states, and we think such leverage should be available where, but only where, it will be useful.
Moreover, since Bethlehem Steel it has been clear that if these areas are mistakenly upgraded, the EPA will not be able unilaterally to return them to their nonat-tainment status. The consequences of a mistaken upgrading are extremely serious, therefore, and we cannot treat the revision process as a game which EPA must lose if it has failed on a previous occasion to be clear about its rationale. The states are entitled to a clear understanding of how [1151]*1151areas will be evaluated; but Congress clearly did not intend technical missteps to defeat EPA’s ability to implement the best of current opinions about the nature of pollution.
The state suggests that we vacate, and, since the record supports redesignation, reverse the decision of the EPA. It is not true that the record supports redesignation; but since EPA has not made clear the rationale for its action, and since certain of the possible rationales rely on empirical assumptions that should be subjected to the give and take of notice and comment we therefore vacate the decision in its present form and remand for further proceedings consistent with this opinion, State Farm, supra; Atchison, Topeka & Santa Fe Railway Co. v. Wichita Board of Trade, 412 U.S. 800, 93 S.Ct. 2367, 37 L.Ed.2d 350 (1973).
Our remand is for reconsideration by EPA and for clarification of the grounds upon which it has dealt or will deal with the Illinois request. EPA will, of course, give Illinois an appropriate opportunity to resPond-