OPINION OF THE COURT
The Pennsylvania Department of Environmental Resources issued a Delayed Compliance Order (DCO) extending the time for Bethlehem Steel Corporation (Bethlehem) *865to bring the fugitive particulate emissions from certain of its blast furnace cast houses into compliance with the requirements of the Clean Air Act, Pub.L.No. 91 — 604, 84 Stat. 1676 (1970), as amended, 42 U.S.C. §§ 7401-7508 (1976 & Supp. I 1977). The Administrator of the United States Environmental Protection Agency (EPA) disapproved that order in a final action authorized by section 113(d)(2) of that Act, 42 U.S.C. § 7413(d)(2) (Supp. I 1977). Bethlehem petitions this court to review the EPA’s action. This entails an analysis of whether the EPA erred in finding that the Pennsylvania DCO failed to comply with the requirements of section 113(d)(4) of the Act, 42 U.S.C. § 7413(d)(4).
I.
Bethlehem operates integrated iron and steel facilities, including four blast furnaces, at its facility in Bethlehem, Northampton County, Pennsylvania. Each of these furnaces has an attached cast house, within which the products of the furnace are drawn off, separated into molten iron and slag, and moved out of the area in cars running on tracks. While the iron and slag are being drawn, separated and moved, they are exposed to the air and emit particulate pollution into it. Since these particulates are not channeled into smoke stacks or chimneys, they are described as “fugitive.” Fugitive particulate emissions during casting represent about 0.4% of the total particulate emissions associated with blast furnace operation. Fifty to eighty percent of these fugitive emissions are in turn associated with the drawing off and initial separation phases of the cast house operations, which take place within the areas of the cast house closest to the furnace. A more detailed description of the process is set forth in the margin.1
Section 109(a) of the Clean Air Act, 42 U.S.C. § 7409(a), requires the federal government to establish nationwide air quality standards, including standards for particulate emissions. Under section 110(a)(1) of the Act, 42 U.S.C. § 7410(a)(1), each state was required to establish a plan for “implementation, maintenance, and enforcement” of these air quality standards for all areas of the state. These State Implementation Plans were to be submitted to the Environmental Protection Agency for review under statutory criteria set out in section 110(a)(2), 42 U.S.C. § 7410(a)(2).
Under the State Implementation Plan (SIP) originally established by the Pennsylvania Department of Environmental Resources (“Pennsylvania”), no fugitive air contaminant shall be emitted “into the outdoor atmosphere .. . from any source other than ... sources determined by the Department [of Environmental Resources] to be of minor significance with respect to the achievement and maintenance of ambient air quality standards or with respect to causing air pollution.” 25 Pa.Code § 123.-1(a), (a)(9) (1972). Further, under the SIP fugitive emissions from Bethlehem’s cast houses had to be brought into compliance *866with air quality standards by 1972. However, in 1973 Bethlehem received a variance from Pennsylvania which extended the time for compliance until July 31, 1975. The extension was approved by the EPA and was meant to give Bethlehem time to investigate and develop appropriate control technology. Bethlehem did not meet the 1975 deadline for compliance, and Pennsylvania commenced an enforcement action against it.
That enforcement action was eventually resolved by a negotiated consent order. Bethlehem agreed, inter alia, to install a partial roof monitor enclosure system and bag houses on its blast furnace cast houses by July 31, 1980, at an estimated cost of $12.5 million. Such a system would pull air from the cast house through a filtering system, and would thereby remove particulates emitted during the drawing and separating phases of the casting process. These particulates would otherwise escape through the open sides of the cast house structure. Bethlehem projected that operating the system would cost about $500,000 annually. Based on prototypes and models tested by Bethlehem, Bethlehem concluded that this control system was more efficient than the alternative technology, which used complete enclosure of the cast houses to achieve total capture of particulates from all phases of the casting process.
By the time the consent order was negotiated between Pennsylvania and Bethlehem, Congress had enacted the Clean Air Act Amendments of 1977 which required that Pennsylvania submit the consent order to the EPA for approval. These amendments reflected congressional concern that the goals of the Clean Air Act were not being met, in part because the states and the EPA had developed a practice of negotiating consent orders that permitted compliance timetables more extended than those originally contemplated by the Act or that omitted deadlines altogether. Section 113(d), 42 U.S.C. § 7413(d), added in 1977, set uniform national standards for the use of delayed compliance orders. Clean Air Act Amendments of 1977, Pub.L.No. 95-95, § 112(a), 91 Stat. 705. Under section 113(d)(1), a state may issue a DCO if it determines that a stationary source, such as Bethlehem, is unable to comply with the applicable State Implementation Plan by the time specified therein and if the order meets certain criteria. Among other requirements, the DCO must contain a schedule and timetable for compliance, § 113(d)(1)(B); must require compliance with applicable interim requirements, § 113(d)(1)(C); and must require final compliance no later than July 1, 1979 or three years after the date for final compliance under the State Implementation Plan, whichever is later, § 113(d)(1)(D).
A longer ¡extension, up to five years after the date on which the source would otherwise be required to be in final compliance with the applicable SIP, is available under section 113(d)(4) for sources proposing to use a “new means of emission limitation.” To be entitled to this longer extension, the source must meet the requirements of section 113(d)(1) and the following additional conditions: (1) the EPA must determine that the new means is likely to be “adequately demonstrated” upon expiration of the order, § 113(d)(4)(A); (2) the new means must not be likely to be used unless an order is granted under this subsection, § 113(d)(4)(B); (3) the EPA must find there is a substantial likelihood the new means will achieve either greater continuous emission reduction than the means that would otherwise be used, § 113(d)(4XC)(i), or equivalent emission reduction at lower cost “in terms of energy, economic, or nonair quality environmental impact ...,” § 113(d)(4)(C)(ii); and (4) compliance with the requirement of the State Implementation Plan prior to or during the installation of such new means must be impracticable, § 113(d)(4)(D).
Before a DCO becomes effective for a major stationary source like Bethlehem, with or without new means, the EPA must review it and determine that a proposed DCO “has been issued in accordance with the requirements of this Act.” § 113(d)(2). Thereafter, while a DCO is in effect and the source complies with it, no federal or citizens’ enforcement action may be pursued *867for noncompliance with the requirements covered by the DCO. § 113(d)(10).
Pursuant to section 113(d), Pennsylvania issued the consent order negotiated with Bethlehem as a proposed DCO and submitted it to the EPA in June 1979. On July 30, 1979 the EPA published notice of a “Proposed Rule” disapproving the order. 44 Fed.Reg. 44572. In connection with this Proposed Rule, the EPA furnished to Bethlehem a Rationale Document setting forth the EPA’s reasoning for its disapproval. In the Proposed Rule and the Rationale Document, the EPA gave four grounds for its conclusion that the proposed DCO was not in accord with section 113(d). First, the order did not impose the interim requirements as required by section 113(d)(1)(C) and (d)(7). Second, the proposed DCO did not satisfy section 113(d)(1)(D), because it did not call for complete elimination of fugitive emissions, for EPA approval of a determination that all emissions except those of “minor significance” had been eliminated, or for final compliance with the SIP by July 31,1980. Third, the technology did not in fact qualify as a “new means” entitled to the five year extension under section 113(d)(4). Fourth, the technique to be used by Bethlehem would not achieve continuous emission reduction equivalent to existing technology at lower cost or greater reduction at same cost as required by section 113(d)(4)(C).
In response to the Proposed Rule and Rationale Document, Bethlehem filed written comments arguing for approval of the DCO. The EPA discussed and rejected Bethlehem’s arguments and published a Final Rule on October 2,1979, reiterating the four grounds for its disapproval. 44 Fed. Reg. 56696. On October 26, 1979 Bethlehem filed its Petition for Review in this court.2
II.
Our standard of review of a final order of the EPA is that established by section 10(e) of the Administrative .Procedure Act, 5 U.S.C. § 706 (1976). United States Steel Corp. v. EPA, 633 F.2d 671, 673 (3d Cir. 1980); Delaware Citizens for Clean Air, Inc. v. EPA, 480 F.2d 972 (3d Cir. 1973). That section states that a reviewing court “shall decide all relevant questions of law, interpret constitutional and statutory provisions, and determine the meaning or applicability of the terms of an agency action.” Further, the reviewing court shall hold unlawful and set aside agency action, findings, and conclusions found to be “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law,” or “in excess of statutory ... authority, or limitations .... ” 5 U.S.C. § 706(2). This is also the standard set out in the Clean Air Act, 42 U.S.C. § 7607(d)(9)(A), (C) (Supp. 1 1977).
Under this standard this court has primary responsibility for determining whether the EPA has correctly interpreted the statutory provision upon which it has based its decision. See St. Joe Minerals Corp. v. EPA, 508 F.2d 743, 746 (3d Cir. 1975), vacated as moot, 425 U.S. 987, 96 S.Ct. 2196, 48 L.Ed.2d 812 (1976). We must determine that the agency had statutory authority for the action it undertook. Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 823, 28 L.Ed.2d 136 (1971). We must then consider whether the EPA has considered the relevant information brought to its attention, Delaware Citizens for Clean Air, Inc. v. EPA, 480 F.2d 972 (3d Cir. 1973), and made a reasonable choice from among the alternatives presented to it. Train v. Natural Resource Defense Council, Inc., 421 U.S. 60, 75, 95 S.Ct. 1470, 1480, 43 L.Ed.2d 731 (1975); National Industrial Sand Ass’n v. Marshall, 601 F.2d 689, 699-700 (3d Cir. 1979).
While deference is due to the interpretation of a statutory term by the officers or agency charged with its administration, Udall v. Tallman, 380 U.S. 1, 16, 85 S.Ct. 792, 801, 13 L.Ed.2d 616 (1965); National Industrial Sand Ass’n v. Marshall, *868supra, 601 F.2d at 698, not all administrative interpretations of regulatory statutes will be accepted. See Udall v. Tallman, supra, 380 U.S. at 16, 85 S.Ct. at 801. We are constrained by our obligation “to honor the clear meaning of a statute, as revealed by its language, purpose and history.” International Brotherhood of Teamsters v. Daniel, 439 U.S. 551, 566 n.20, 99 S.Ct. 790, 800 n.20, 58 L.Ed.2d 808 (1979).
III.
A. WHETHER THE DCO MET THE STATUTORY REQUIREMENTS FOR INTERIM EMISSION REDUCTIONS AND STANDARDS
The EPA’s first ground for disapproving the DCO is that it does not meet the requirements of Section 113(d)(1)(C) and (d)(7) regarding interim emission reductions and standards. Section 113(d)(1)(C) provides that a state may issue a DCO if
the order requires compliance with applicable interim requirements as provided in ... paragraph (6) and (7) ....
Paragraph (6) states that a DCO shall set forth a compliance schedule “containing increments of progress which require compliance with the requirement postponed as expeditiously as practicable.” Paragraph (7) provides:
A source [subject to a DCO] shall use the best practicable system or systems of emission reduction (as determined by the [EPA] taking into account the requirement with which the source must ultimately comply) for the period during which such order is in effect and shall comply with such interim requirements as the [EPA] determines are reasonable and practicable. Such interim requirements shall include—
(A) such measures as the [EPA] determines are necessary to avoid an imminent and substantial endangerment to health of persons, and (B) a requirement that the source comply with the requirements of the applicable implementation plan during any such period insofar as such source is able to do so (as determined by the [EPA]).
Thus, the DCO must (1) require that Bethlehem, as the subject source, use the “best practicable system” of emission reduction during the interim period while the DCO is in effect and (2) require that Bethlehem comply with interim requirements approved by the EPA or contained in the applicable implementation plan.
Bethlehem appears to be arguing that because section 113(d)(1)(C) contains the language “applicable interim requirements,” it does not incorporate all of paragraph (7) but only that portion of paragraph (7) which refers to “applicable interim requirements.” In other words, Bethlehem argues that the provision in paragraph (7) regarding “systems of emission reduction” is not incorporated. Under our view of this issue, we can assume without deciding that section 113(d)(1)(C) incorporates all of paragraph (7), both the language relating to a “system” and the language relating to “requirements.”
The EPA has apparently not imposed any interim requirements which are applicable to Bethlehem. With regard to interim requirements imposed by the state, the DCO states that the cast houses are unable to comply with Pennsylvania regulations prior to the final compliance date, that the order provides for that final compliance as expeditiously as practicable, “that the compliance schedule set forth herein contains reasonable and practicable interim requirements,” and that the provisions for monitoring emissions during the DCO are in accord with law. The order requires Bethlehem to meet a schedule for submission of plans, placement of purchase orders, and construction. Bethlehem must also submit quarterly progress reports on this installation schedule. In the Notice accompanying publication of this proposed order, Pennsylvania stated that “Bethlehem has agreed to take reasonable measures during the period of installation of the partial roof monitor enclosure system to comply with the opacity and particulate regulations of [Pennsylvania], as best as is possible during this construction period.” 9 Pa.Bull. 1152, 1153 (1979).
*869The EPA does not contend that the DCO fails to meet the statutory obligation concerning applicable interim “requirements.” Instead, it argues that the DCO is deficient with regard to the “best practicable system” of emission reduction. It is true that the DCO imposes no obligation on Bethlehem with regard to the use of any interim system of emission reduction. Bethlehem argues that because the DCO is for “new means” technology, no other practicable system could exist. We agree with the EPA that the mere fact that new means are involved does not prove that there is no practicable system which can be used in the interim, since other practices may in fact permit an interim reduction in emissions. However, the EPA does not suggest that there exists a practicable system which the DCO should have required Bethlehem to install. In fact, the EPA has itself indicated the existence of such a system is unlikely. It acknowledged in its Rationale Document in connection with the proposed denial that “it is arguable .. . that no practicable system of emission reduction exists” and that “no reasonable and/or practical interim limitations” could be imposed in Bethlehem’s case. In effect, the EPA is arguing that the absence of an express determination by Pennsylvania that no “practicable system” exists means that the proposed DCO is not in accord with the requirements of the Clean Air Act, and so cannot be approved by the EPA.
We believe it is significant that the statutory language does not require that the DCO must make such a negative finding if there is no “practicable system of emission reduction.” The only reason the EPA gives for its position that such a negative finding is necessary is that Congress intended to insure a thorough examination of all possible emission control options during the period of the DCO. However, if Congress believed that a negative finding by the state were necessary to insure such a thorough examination, we may assume that Congress would have imposed such a requirement. In the absence of such a requirement, we are unwilling to read it into the statute.
Furthermore, we believe that the EPA position cannot be justified under the circumstances of this case. The EPA has not utilized the authority it has under § 113(d)(7) to determine those interim requirements which it deems reasonable and practicable. Similarly, if a practicable system of emission reduction existed which could be imposed for the period during which the DCO was in effect, the EPA could have referred to it in its rationale for disapproval. Its failure to do so, notwithstanding its experience with the type of emissions in question, is a strong indication that no such “practicable system” exists. The EPA cannot impose on the state a requirement that its proposed DCO make a negative finding which the statute itself does not impose. Cf. Union Electric Co. v. EPA, 427 U.S. 246, 96 S.Ct. 2518, 49 L.Ed.2d 474 (1976) (EPA may not add feasibility requirement). Thus its disapproval of the DCO for this reason was not in accordance with law.
B. WHETHER THE DCO REQUIRED COMPLIANCE BY JULY 31, 1980
The EPA’s second ground for disapproval is that the DCO proposed by the state does not comply with the requirement of section 113(d)(4) that the DCO “shall provide for final compliance” with the State Implementation Plan by July 31, 1980, the last date by which a DCO directed to Bethlehem’s cast houses could require final compliance.3 The explicit language of paragraph (1) of the DCO requires that Bethlehem
*870(d) Complete installation for operation of the approved air pollution control system at all four of its blast furnaces on or before July 31, 1980.
and
(g) On and after July 31, 1980, operate each source in compliance with any final conditions of the plan approval and pursuant to the authorization granted by the operating permit issued for each source.
The EPA contends that the DCO’s requirement of installation of the system by July 31, 1980 and its subsequent operation in compliance with the SIP does not constitute a provision for “final compliance” with the SIP by July 31, 1980. According to the EPA, the missing ingredient is the failure of the DCO to provide that a minor significance determination be made by that date.
The focus of the inquiry must be directed to the SIP, since it is the EPA’s position that the DCO does not comply with Pennsylvania’s implementation plan. The applicable provision of the State Implementation Plan, set out in 25 Pa.Code § 123.1 (1981), provides:
(a) No person shall cause, suffer, or permit the emission into the outdoor atmosphere of any fugitive air contaminant from any source other than the following:
(9) Sources and classes of sources . .. for which the operator has obtained a determination from [Pennsylvania] that ... (i) the emissions are of minor significance with respect to causing air pollution .... (emphasis added).
The EPA’s argument is that until Pennsylvania determines that the emissions from the Bethlehem Steel casting operations will be of minor significance after the air pollution control system is in operation, there can be no “final compliance” within the meaning of section 113(d)(4). Because the DCO clearly does not contemplate that a minor significance determination will be made by July 31, 1980, the EPA concludes that the DCO fails to comply with the time requirement of section 113(d)(4).
There is no dispute that under the DCO, no minor significance determination would be made by July 31, 1980. The DCO requires Bethlehem and the state to begin a one year study no later than January 1, 1981 to determine whether “residual fugitive particulate emissions from the blast furnace casting operations are interfering with the attainment or maintenance of ambient air quality standards or are causing air pollution.” The one year time period is necessary because these are not discrete emissions which can be measured and instantaneously evaluated. Under the Pennsylvania standard, fugitive emissions are measured in terms of their impact on ambient air quality which must be studied over a period of time to provide representative data reflecting operating conditions and various weather conditions.
We fail to find any statutory support, either in language or in intent, which would require that this lengthy period of testing must be completed before July 31, 1980. There is no specific provision, either in the Clean Air Act or in the state’s implementation plan, which requires that a minor significance determination must be made as a condition precedent to compliance with the requirements of the SIP. We recognize that there is a danger that on the date set for final compliance, a source which has been granted a DCO may fail to meet the applicable emission standards.4 If Bethlehem fails to meet the state’s standards, it will be subject to the enforcement provisions of the Act. We do not read section 113(d)(4) as requiring a guarantee by the state that the new means technology will in fact result in compliance. To require such a guarantee would be inconsistent with the language in subparagraph (A) of the same provision authorizing issuance of a DCO for new means technology which “is likely to be adequately demonstrated ... upon expiration of the order.” (emphasis added). See *871Currie, Relaxation of Implementation Plans Under the 1977 Clean Air Act Amendments, 78 Mich.L.Rev. 155, 172 (1979).
While it might have been preferable for the DCO to have expressly articulated the state’s determination that the new means technology is likely to achieve the required standard, we believe that determination was evidenced by Pennsylvania’s notice of proposed issuance of the order which stated that “the Order proposed to be submitted establishes an expeditious schedule requiring compliance by July 31, 1980.” 9 Pa. Bull. 1152 (1979). Furthermore, Pennsylvania’s transmittal letter submitting the Order to the Administrator for approval also stated that the air pollution abatement order which was enclosed “requires compliance with emission limitations in the federally-approved SIP ... by July 31, 1980.” In addition, Pennsylvania is precluded by its own regulations from approving the installation of air pollution controls unless it is satisfied that the application meets the requirements of both its SIP and those promulgated by the EPA. See 25 Pa.Code § 127.12(a)(4), (b) (1981).
Finally, we are concerned that the EPA’s interpretation of the “final compliance” language of the statute could materially decrease the time for deployment of new means for emission control and hence could operate to deter the technological inventiveness which the new means exception was designed to encourage. Where a lengthy period of time is required before ultimate determination of the success of the new means technology can be made, as in this case, the EPA’s approach would require installation of the new means a full year earlier than the statutory date for compliance. Although the legislative history of this provision is sparse and neither of the parties nor our independent research has uncovered any helpful comments, it would appear that Congress enacted its deadline extension for new means because of its recognition that time is required to resolve the technological problems which accompany an installation of new means. Without an express congressional indication that “final compliance” is not met until completion of subsequent testing or compliance determinations, we are unwilling to read such a requirement into the statute.
Another basis given by the EPA in its Rationale Document for disapproval for failure to meet the final compliance requirement was that “DER expressly waived its right to enforce Section 123.1 against the blast furnaces which are the subject of this order until January 1, 1982, the latest possible date of completion of the air quality study.” The provisions of the Order to which the EPA objects relate to the mix of remedies chosen by the state to be used in the enforcement phase, a question primarily for the states, See Union Electric Co. v. EPA, 427 U.S. 246, 96 S.Ct. 2518, 49 L.Ed.2d 474 (1976); Train v. Natural Resources Defense Council, 421 U.S. 60, 95 S.Ct. 1470, 43 L.Ed.2d 731 (1975). As the Court of Appeals for the Seventh Circuit recently observed in Bethlehem Steel Corp. v. EPA, 638 F.2d 994, 1006 (7th Cir. 1980), the EPA may not deny approval of a DCO because EPA prefers that the state use different means of enforcement. In that case the EPA disapproved a DCO in which a state had waived its right to seek certain civil and criminal penalties for specified types of delays in compliance. Noting that this waiver did not affect the right of the state to seek the non-compliance penalties provided in section 120 of the Act and did not reduce the EPA’s own enforcement powers under the Act, the court found the basis of the EPA’s decision went beyond the statutory authority granted to it.
The case before us is analogous to that considered by the Seventh Circuit. It may take the state a considerable period of time to determine what the level of Bethlehem’s emissions is, but that does not alter the requirement that the level in fact be in compliance by July 31, 1980. If the state determines at some later date that Bethlehem’s residuary emissions were too great to be regarded as of minor significance, then the state may act upon its conclusion. Similarly, the EPA retains power to enforce the standard from July 31, 1980.
*872For these reasons, we reject the EPA’s position that the DCO fails to meet the statutory requirement of timely “final compliance” by July 31, 1980.
C. WHETHER THE CANOPY HOOD SYSTEM QUALIFIES AS A NEW MEANS
The EPA’s third and fourth grounds for its disapproval relate to its conclusion that Bethlehem’s canopy hood system did not qualify as a “new means” of emission limitation under section 113(d)(4). If so, Bethlehem would be entitled at most to an extension of its compliance deadline until July 1, 1979 under section 113(d)(1)(D), and the DCO clearly does not require any compliance by that date.
The EPA’s first ground for concluding that the Bethlehem system does not satisfy paragraph (4) rests on the definition of “new means.” Although acknowledging that the legislative history is not helpful in establishing the meaning of this phrase, the EPA argues that it is most reasonable to infer that Congress intended “to limit the issuance” of such DCO’s “to those instances where the technology to be utilized is truly innovative.” The EPA reasons that the technique used here is well-documented, has been used on electric arc furnaces by Bethlehem, was completely installed on one blast furnace by October 1977, and was found in Pennsylvania’s DCO to represent “reasonably available control technology.” Therefore, according to the EPA, the system cannot be a new means of emission limitation, and might have been installed on all the cast houses by October 1977.
Bethlehem responds that under all the information available to the EPA in making its determination, Bethlehem’s system amounts to a new means. It points out that its system is the first blast furnace cast house control system in the United States, and also the first retrofitted to a previously operating cast house. It assert; that the system is within the contemplation of the Act as meriting the longer period of development allowed under paragraph (4), and that the EPA has taken an untenably narrow view of what constitutes a new means.
Section 113(d)(4)(A) provides for up to a five-year delay of compliance for an existing stationary source if “the source will expeditiously use new means of emission limitation which the [EPA] determines is likely to be adequately demonstrated . . . upon expiration of the order....’’ Further, section 113(d)(4)(B) requires that such new means must not be likely to be used unless a DCO is granted under subsection (d). The EPA apparently does not contend that the Bethlehem system will not be “adequately demonstrated” or that its use would have been inevitable without the prospect of a DCO. Rather, the EPA relies on what it perceives to be the lack of novelty in the system Bethlehem proposes to install and operate.
Both the statute and legislative history are uncommunicative, if not silent, on what standard of novelty is necessary in a new means. The House Report refers to the extension of deadlines for sources “which are willing to test new innovative technology which has not yet been adequately demonstrated.” H.R.Rep.No.294, 95th Cong., 1st Sess. 4 (1977), reprinted in [1977] U.S.Code Cong. & Ad.News 1077, 1081. Such deadline extensions were to apply only to existing sources that were currently unable to comply; sources not yet built had no need of such extensions. Id. at 57, [1977] U.S.Code Cong. & Ad.News at 1135. Thus, the paragraph is concerned with the problems of retrofitting technology, and not with the construction of new sources with state of the art technology.
The House Report also speaks of the “development of new and superior technology which will result in significant environmental benefits or achieve equivalent emission reduction with lower economic, energy, or environmental cost associated.” The bill proposed by the committee authorizes a DCO “for the purpose of encouraging use of innovative continuous emission reduction technology or processes.” Id. at 60 [1977] U.S.Code Cong. & Ad.News at 1138. By requiring that it be “adequately demon*873strated” the committee intended “that the existence of purely theoretical, experimental, or speculative technology will not serve as a justification for a DCO. . . . Nor may unreliable or inefficient technology provide such a justification.” Id. By requiring that a source be otherwise “unlikely to use” a new technology or process, the committee indicated that it meant that “[wjhere the remaining problems with a technology may be such as would make its successful installation and operation improbable without an extension, then a DCO could be issued.” Id. Finally, the committee reiterated its concern with “the burden sources may encounter in attempting to obtain and install new and superior technology,” particularly with regard to heavy and potentially prohibitive capital outlays. Id. at 61, [1977] U.S.Code Cong. & Ad.News at 1139. From this, it appears that Congress did not intend to restrict the benefits of paragraph (4) to the developer of a technology, but meant to extend it also to those who would put such technology into operation.
The Senate version of the delayed compliance provision proposed an extension to July 1, 1981 for sources which intended to comply by means of innovative control techniques which has “potential industrywide application at significantly lower costs than demonstrated control technology.” H.R. Rep.No.564, 95th Cong., 1st Sess. 133 (1977) (Conference Report), reprinted in [1977] U.S.Code Cong. & Ad.News 1502, 1514. The Conference Committee report discussed only the compromise on the date to which compliance would be delayed; apparently the House and Senate saw no differences in the scope of the new means they sought to protect. Id.
In view of the repeated references in the legislative history to use of technology, to retrofitting, to “adequate demonstration” of reliable and efficient technology, to successful installation and operation, and to potential industrywide application, we are unable to accept the EPA’s reading of paragraph (4) as limited to the development of technology superior to the state of the art. Rather section 113(d)(4), reasonably construed, must encompass innovative use of technology and application in changed circumstances. Thus, the fact that the system proposed by Bethlehem was “well documented” in scientific literature is insufficient to establish that the system was not a new means when applied for the first time in an industry in which no other system had yet proved practicable. Similarly, use of an analogous system on electric arc furnaces, absent any explanation by the EPA of how these furnaces resemble blast furnace cast houses, does not justify rejection of the system as not a new means.5
The EPA contends that the fact that Bethlehem had installed this system on one east house by October 1977 suggests that Bethlehem was technologically capable of building three more such systems by the same date. However, neither the statute nor the legislative history indicate that Congress intended to make such capacity the determinative factor. Congress in fact required that technology be found efficient and reliable before a DCO would issue. Hence, the statute cannot be interpreted to mean that the technological capacity, once established as practicable, would deprive a source of eligibility for a DCO. Here application of the new technology requires capital outlays of over $12 million. It is unrealistic to assume that Bethlehem would have undertaken such an investment without testing its technology. The EPA’s reliance on such testing as the basis for denial of the “new means” benefit would frustrate the intent of Congress in providing an exten*874sion for use of efficient, reliable technology. We find equally significant the fact that EPA never responds to Bethlehem’s argument that the system in place in 1977 was merely experimental and incomplete “since it did not include any air cleaning device, and since the ductwork and fans were only temporary and were being used to evaluate the system to develop additional data for design of the total control system for all four furnaces.” Thus, we cannot agree that Bethlehem’s completion of a prototype in 1977, absent more, removes the system from the realm of “new means” contemplated by Congress.
The fourth factor cited by the EPA, the finding by Pennsylvania that the system is “reasonably available control technology,” also does not support its refusal to find the Bethlehem system to be a new means.6 Under sections 171-78 of the Clean Air Act, 42 U.S.C. §§ 7501-08 (Supp. I 1977), a state that cannot comply with national air quality standards by July 1, 1979 must establish a plan for expeditious attainment of those standards, or be barred from permitting the construction or modification of any major stationary source that contributes to emission levels. As part of its plan, the state must show that it requires existing sources in the area to make “reasonable further progress,” including “adoption, at a minimum, of reasonably available control technology.” Section 172(b)(3). In its Order Pennsylvania found that the system proposed by Bethlehem represents reasonable available control technology, as that term is used in Section 172(b)(3) of the Clean Air Act. The EPA contends that this finding precludes the issuance of the new means DCO, but nothing in the statute or legislative history hints that Congress intended to deprive the states most in need of innovation of the benefits of section 113(d)(4) by this process of circular logic.
On the basis of the factors relied upon by the EPA we cannot sustain its conclusion that the system proposed by Bethlehem is not a “new means” under section 113(d)(4). For example, the EPA has not provided us with any basis for comparing use of the canopy hood system on Bethlehem’s electric arc furnaces to its use in the steel furnace cast houses at issue. Without such a background, we cannot accept ipse dixit that use in the electric arc furnace defeats the new means use on these cast houses. While we do not pretend to the technical expertise of the agency to which Congress has committed the problems of regulating environmental pollution, we must have before us the technical information necessary to our review. This requires of us “enough steeping in technical matters to determine whether the agency ‘has exercised a reasoned discretion.’ ” Portland Cement Ass’n v. Ruckelshaus, 486 F.2d 375, 402 (D.C.Cir.1973), cert. denied, 417 U.S. 921, 94 S.Ct. 2628, 41 L.Ed.2d 226 (1974). We consequently find that the EPA’s disapproval of the DCO on the ground that it does not involve a new means to be both erroneous as a matter of statutory interpretation, and arbitrary and capricious in its treatment, of the evidence available.
D. WHETHER THE SYSTEM MEETS THE REQUIREMENT OF EQUIVALENT EMISSION REDUCTION
The EPA’s second rationale for disapproving the Pennsylvania order as a proposed new means DCO is that the Bethlehem system cannot satisfy section 113(d)(4)(C), which requires that a new means DCO must involve a means determined by the EPA “to have a substantial likelihood” of
(i) achieving greater continuous emission reduction than the means of emission lim*875itation which, but for such order, would be required; or
(ii) achieving an equivalent continuous reduction at lower cost in terms of energy, economic, or nonair quality environmental impact ... (emphasis added).
Pennsylvania expressly found that Bethlehem’s proposed system satisfied the requirement set forth in subparagraph (C)(ii). The EPA however rejected this conclusion, stating:
The Region III Enforcement Division believes that the existing technology for control of blast furnace cast house emissions, capable of achieving compliance with the Pennsylvania SIP, is a complete enclosure of the cast house, a control technique Bethlehem utilizes at its Johns-town ferromanganese furnace, and presently utilized at one of the basic blast furnaces (Blast Furnace # 1) at Dominion Foundries and Steel Company (DO-FASCO) in Hamilton, Ontario, Canada. The technology being installed at Bethlehem clearly will not capture all of the cast house fugitive emissions, and therefore, will not affect emissions control equivalent to full cast house control at a lower cost, or greater control than full cast house enclosure at the same cost, as required by Section 113(d)(4)(C).
As an initial matter we note that application of the relevant statutory language entails certain inherent difficulties. One of the two subjects of comparison is the result of the operation of a new means, for which by definition no final operating results could be available. In obvious recognition of this fact, Congress provided that the new means must only show “a substantial likelihood” that it can meet the standard. Proof, in a courtroom sense, would be an impossible burden.
The statute fails to suggest how the EPA is to determine the other subject of comparison, the means of emission limitation which would otherwise be required. Where the applicable standard is framed in terms of result or performance, as it is in the case of the Pennsylvania SIP which requires elimination of all fugitive air contaminants except those of minor significance, the necessity of predicting the means of emission limitation which would otherwise be required seems an inapposite inquiry. As a practical matter, in Pennsylvania any means of emission limitation which satisfied its performance standard would have been required.
In framing the statutory standard, Congress did not intend that the inquiry should be limited only to a performance comparison without consideration of relevant costs. While subsection (i) authorizes new means treatment for those technologies which achieve greater emission reduction than the subject of comparison, subsection (ii) requires only equivalent reduction if accomplished at lower energy, economic or other environmental costs and was obviously inserted to broaden the range of technologies which would be accorded such treatment. It is also significant that Congress did not limit new means treatment to the best available option. If there is more than one option, any control system which meets the statutory standard will qualify for new means treatment.
In deciding whether subparagraph (C)(ii) could be satisfied by the proposed DCO, the EPA was required first to determine the means “which, but for such order, would be required,” thus establishing the standard for comparison. Second, the EPA had to determine what level of continuous reduction was achieved by that standard’s means. If the reduction likely to be achieved by the new means system was equivalent, the EPA should then have compared the costs of the two systems in terms of energy, economic or other environmental consequence. In this case, the EPA never reached the third step because of its conclusions as to the first two inquiries.
The EPA’s conclusion that the canopy hood system failed to meet the statutory requirements for new means was predicated on its determination that the standard for comparison must be a complete enclosure system. The EPA Administrator found that “the most probable alternative control” technology for cast houses was full building evacuation, a system which encloses the entire source and funnels all emis*876sions to one exhaust point. However, nothing in the Rationale Document or in the Pennsylvania order suggests that Pennsylvania would in fact have required installation of a complete evacuation system. Before requiring Bethlehem to install such a system, Pennsylvania would certainly have considered both its cost and feasibility. Neither factor is discussed by the EPA in supporting its conclusion that such a system is the appropriate standard of comparison. Bethlehem asserts that the economic cost of maximum operation of the DOFASCO-type system is more than 400% of the canopy hood system, and the EPA does not dispute this. Further, the EPA has failed to show that the complete evacuation system would be feasible for the cast houses in question. The EPA has itself recognized that technology used in one context may not be transferable in every case to other applications. A similarly bare assertion of the EPA’s belief in the existence of applicable technology was rejected by the Seventh Circuit in its recent decision in Bethlehem Steel Corp. v. EPA, 638 F.2d 994, 1004 (7th Cir. 1980). In the analogous context of water pollution controls, the Second Circuit refused to accept the EPA’s determination that technology developed in one industry necessarily sets the standards for another, stating:
[E]ven if technology which is not presently in use can-be treated as available and achievable, there must be some indication in the administrative record of the reasons for concluding that such technology is feasible and may reasonably be expected to yield the effluent reduction mandated when applied to the particular industry .... Agencies must present sufficient support for their conclusions in order that the Court can properly review their decisions.... Judged in that light, we conclude that the EPA has failed to adequately identify the facts underlying and explain the reasons for its conclusion that full recycle and inert casting are available technologies which could completely eliminate waste water discharge.
Hooker Chemicals & Plastics Corp. v. Train, 537 F.2d 620, 636 (2d Cir. 1976) (citations omitted). On the basis of the record before us, we see no basis to support the EPA’s assumption that Pennsylvania would have required Bethlehem to install the complete evacuation system and the EPA’s correlative conclusion that the complete evacuation system is the appropriate standard of comparison.
There is yet another difficulty with the EPA’s conclusion that DOFASCO’s full building evacuation system is the appropriate standard. The EPA bases this conclusion on test results indicating that the DO-FASCO system is capable of achieving 100% capture.7 Bethlehem notes, however, that the applicable Pennsylvania SIP does not require 100% capture but instead permits emissions of minor significance. If the DO-FASCO system can be operated at a lower efficiency level which captures all but emissions of minor significance, then the canopy hood system which is also projected to achieve that level of emission reduction is in fact substantially equivalent. If so, the EPA was required to compare the costs of the respective systems, which it failed to do.
The EPA fails to respond to Bethlehem’s contention about the operation of the DO-FASCO system at lower efficiency levels. Its fallback position is that even if the level of control required by the SIP is the correct measure of equivalence, it is not possible to say that Bethlehem’s system will achieve a level of control equal to that required by the SIP. If the EPA’s position were correct, there would be no new means qualified under the section because absolute certainty as to its ultimate results cannot be achieved in advance. Thus, the EPA has applied an incorrect standard since, as noted supra, the statute only requires finding of a “substantial likelihood” of achievement of the requisite standard.
Even under the review standard of “arbitrary, capricious” or in disregard of *877law, agency action will not be upheld when inadequacy of the agency’s explanation frustrates meaningful review. Camp v. Pitts, 411 U.S. 138, 93 S.Ct. 1241, 36 L.Ed.2d 106 (1973); Citizens to Preserve Overton Park Inc. v. Volpe, 401 U.S. 402, 91 S.Ct. 814, 28 L.Ed.2d 136 (1972). We believe the congressional intent in providing for delayed compliance orders requires that EPA provide a more thorough analysis than a purely arithmetical comparison of the potential capacities of the systems under consideration. The Rationale provided by the EPA fails to explain why a comparison with the DOFASCO system is appropriate, and fails to undertake the additional inquiries required by the statute.
IY.
In consideration of this matter, we believe it is important to note what is really at issue. The EPA’s attorney candidly admitted at argument that even if this court were to approve the Administrator’s action in disapproving the DCO, it was not seeking to close Bethlehem’s steel mill. We are not, therefore, dealing with a situation where the administrative agency to which Congress has delegated authority in this area believes that the emissions in question are so dangerous to the health and safety of the persons affected that immediate action is necessary to stop the emissions. If the DCO disapproval is upheld, the agency will be in a position to proceed with an enforcement action which would subject Bethlehem to fines which could be assessed up to $25,-000 a day depending on the seriousness of the violations for each day for which it is found to have been in violation. 42 U.S.C. § 7413(cXl).
Underlying the EPA’s various technical objections to the DCO which we have discussed, we discern two principal threads in its position. First, the EPA views a delayed compliance order for new means technology to be an exception to the general mandate of the Clean Air Act which must be narrowly construed. Second, it asserts that it has made a reasonable decision in an area delegated to its expertise, and we are therefore obliged to sustain its action. We believe neither of these contentions is persuasive. Congress has not indicated that a new means exception should be narrowly construed. In this case, for example, there was no technology which Bethlehem believed to be available to cure the serious problem of fugitive emissions from cast houses, and it proceeded upon a program of experimentation to develop technology designed to solve this particular problem. This is precisely the type of situation which appears to merit treatment as a new means. Secondly, the statute specifies the prerequisites for a new means DCO. We do not believe the EPA has the discretion to add to or alter those prerequisites, or to construe them in a manner which would frustrate congressional intentions.
In summary, we conclude that the bases used by the EPA to disapprove the DCO were in some instances based on an erroneous interpretation of the statute and in other instances were arbitrary and capricious. Therefore, we will grant Bethlehem’s petition for review, and remand this matter to the EPA for further proceedings in conformance with this opinion.