UNITED STATES of America, Plaintiff-Appellee, and State of Michigan: Frank J. Kelley, Attorney General, Intervening Plaintiff-Appellant, Cross-Appellee, v. AKZO COATINGS OF AMERICA, INC., et al., Defendants-Appellees, Cross-Appellants, 949 F.2d 1409

ENGEL, Senior Circuit Judge. This is an appeal by the State of Michigan from the entry of a consent decree between the United States Environmental Protection Agency (“EPA”) and twelve defendants pursuant to the Comprehensive Environmental Response, Compensation, and Liability Act of 1980 (“CERCLA”), as amended by the Superfund Amendments and Reauthorization Act of 1986 (“SARA”), 42 U.S.C. § 9601 et seq. The consent decree would require the defendants, or potentially responsible parties (“PRPs”), to engage in remedial work to clean up a hazardous waste site in Rose Township, Oakland County, Michigan (“Rose Site”).

The proposed remedial plan at the Rose Site calls for the excavation and incineration of surface soils contaminated with po-lychlorinated biphenyls (“PCBs”), lead, arsenic and other toxic materials and the flushing of the subsurface soils contaminated with a variety of volatile and semi-volatile organic compounds. The state challenges the legality of the remedial action, and seeks to prevent entry of the consent decree. The Natural Resources Defense Council, the Environmental Defense Fund and the Sierra Club have filed a brief as amici curiae supportive of the state’s position.

The majority of the state’s and amici’s objections to the decree focus on the effectiveness of soil flushing at the Rose Site, where layers of clay are interspersed among beds of sand and silt. The PRPs cross appeal the district court’s determination that the decree must comply with Michigan’s groundwater anti-degradation law. I. STATUTORY OVERVIEW By the late 1970s, Congress concluded that existing cleanup programs were inadequate to the task of taking care of literally thousands of sites across the country posing a serious threat to public health and the environment.

Consequently, in 1980, Congress enacted CERCLA, also known as “Superfund,” to ensure prompt and efficient cleanup of hazardous waste sites and to place the costs of those cleanups on the PRPs. See S.Rep. No. 848, 96th Cong., 2d Sess. 98, reprinted in, 1 Cong. Research Serv., 97th Cong., 2d Sess., A Legislative History of the Comprehensive Environmental Response, Compensation and Liability Act of 1980 (Superfund), at 405 (1980). Throughout the 1980s, the Superfund hazardous waste cleanup program enjoyed centerstage prominence in environmental law. Nevertheless, the early years of CERCLA were difficult.

CERCLA was a hastily-assembled bill which contained a number of technical flaws due to Congress’ limited understanding of the hazardous waste problem and its effects on the environment. See Grad, A Legislative History of the Comprehensive Environmental Response, Compensation and Liability (“Superfund”) Act of 1980, 8 Co-lum.J.Envtl.L. 1, 2, 34 (1982). Both Congress and EPA, for example, believed in the late 1970s that a site could be adequately cleaned up by “scraping a few inches of soil off the ground.” H.R.Rep.

No. 253, 99th Cong., 2d Sess., pt. 1, at 54 (1986), reprinted in 1986 U.S.Code Cong. & Admin.News 2835, 2836. Congress also grossly underestimated the number of sites requiring cleanup and the monies necessary to remedy the problem. Compare id. with H.R.Rep. No. 1016, 96th Cong., 2d Sess., pt. 1, at 18-20 (1980), reprinted in 1980 U.S.Code Cong. & Admin.News 6119, 6120-23.

EPA, as the delegatee of the President’s authority under CERCLA, 42 U.S.C. § 9615, was criticized for the slow pace of cleanups, for failing to provide remedies that would protect public health and the environment, and for alleged “sweetheart” deals that reduced cleanup costs for industry at public expense. As a result, in 1986 Congress passed SARA, which reauthorized and amended CERCLA in several important ways.

Congress sought to better define cleanup standards, to expand resources available to EPA for investigations and cleanups, to clarify EPA’s authority under Superfund law, and to expand and clarify the states’ role in any remedial action undertaken, or ordered, by EPA. CERCLA applies “primarily to the cleanup of leaking inactive or abandoned sites and to emergency responses to spills.” F. Anderson, D. Mandelker & A. Tarlock, Environmental Protection Law and Policy 568 (1984). The Act directs EPA to develop a National Priorities List (“NPL”) for response priority purposes. 42 U.S.C. § 9605(a).

After a site is placed on the NPL, a Remedial Investigation and Feasibility Study (“RI/FS”) is performed to define the nature and extent of the threat posed by the release and to evaluate proposed remedies. 42 U.S.C. §§ 9604, 9622; 40 C.F.R. § 300.68(d). Once EPA determines under CERCLA that a response action is needed at a particular hazardous waste site, it must publish a proposed remedial action plan (“RAP”) and provide an opportunity for comment. 42 U.S.C. § 9617.

EPA then issues a Record of Decision (“ROD”) setting forth the remedy selected for the site, including remedial technologies and cleanup standards. 42 U.S.C. § 9617. In implementing its RAP, EPA may pursue one of three possible courses of action. See generally Koppers Indus., Inc. v. EPA, 902 F.2d 756, 757 n. 1 (9th Cir.1990) (discussing the various options). EPA may undertake a response measure on its own, which may include removal and/or remedial action, and then sue PRPs it can find for reimbursement. 42 U.S.C. §§ 9604, 9607.

In the interim, or in the event it cannot locate any PRPs or they cannot be made to pay the cleanup costs, the government-initiated cleanup may be financed by the “Superfund,” 42 U.S.C. § 9611, a trust fund derived from general federal revenues and an excise tax on specified chemicals. See 42 U.S.C. § 9631. Secondly, EPA may, independent of fund-financed response ae-tions, issue an administrative order directing PRPs to implement removal or remedial action. 42 U.S.C. § 9606.

Alternatively, EPA may apply to the district court for an injunction to compel PRPs to clean up or abate an actual or threatened release of hazardous substances from a facility. Id. As a third option, EPA may enter into an agreement with PRPs to perform a response action, 42 U.S.C. § 9622. Such an agreement is at issue here. The federal legislative scheme and its history are persuasive that Congress did not intend to leave the cleanup under CERCLA solely in the hands of the federal government.

CERCLA, as amended by SARA, provides a substantial and meaningful role for the individual states in the selection and development of remedial actions to be taken within their jurisdictions. In this case for example, pursuant to 42 U.S.C. § 9621(

f) the State of Michigan had a reasonable opportunity to comment on the RI/FS, the RAP proposed in the amended ROD, and other technical data related to the implementation of the proposed remedy. The state was also entitled to and did participate in the settlement negotiations that led to the decree at issue. Id. Further, CERCLA is designed to accommodate more stringent “applicable or relevant and appropriate requirements” (“ARARs”), i.e. environmental standards of the state in which a site is located. 42 U.S.C. § 9621(d).

Once a consent decree is proposed by EPA, see id. § 9622(a), the state can challenge it if EPA has proposed implementation of a remedy for which the federal agency has waived a valid and more stringent state requirement. Id. § 9621(d)(4), (f)(2)(B). The state may also enforce a decree to the extent the remedial action fails to comply with any state environmental requirements which have not been waived by EPA. Id. § 9621(e).

If no PRPs can be located, or if they are insolvent, a state or political subdivision may enter into a contract or cooperative agreement with EPA, whereby both may take action on a cost-sharing basis. 42 U.S.C. § 9604(c), (d). A state may also sue PRPs for remedial and removal costs if such efforts are consistent with the National Contingency Plan (NCP). Id. § 9607(a)(4)(A). However, assuming it is not the “lead” agency, the state is limited in its ability to require alternative relief if and when a consent decree is entered into between PRPs and EPA. See id. § 9621(f).

Under CERCLA, Congress expressed its preference for thorough yet cost-effective remedies at hazardous waste sites. Compare 42 U.S.C. § 9621(a) (“the President shall select appropriate remedial actions ... which provide for cost-effective response.”) with 42 U.S.C. § 9621(b) (“Remedial actions in which treatment which permanently and significantly redress the volume, toxicity or mobility of the hazardous substances ... are to be preferred over remedial actions not involving such treatment.”).

CERCLA’s statutory scheme and legislative history reflect two other principal and related concerns: First, Congress intended that the federal government be immediately given the tools necessary for a prompt and effective response to problems of national magnitude resulting from hazardous waste disposal. Second, Congress intended that those responsible for problems caused by the disposal of chemical poisons bear the costs and responsibility for remedying the harmful conditions they created. United States v. Reilly Tar & Chemical Corp., 546 F.Supp. 1100, 1112 (D.Minn.1982).

These concerns must be kept in mind as we analyze the challenges to the consent decree. II. FACTS The Rose Site consists of about 110 acres on which liquid and solid industrial wastes were illegally dumped in the late 1960s. In 1979, the Michigan Toxic Substance and Control Commission declared a toxic substance emergency at the Site, and 5,000 drums of toxic waste were immediately removed. Investigation disclosed that the drums contained, among other chemical compounds, PCBs, phthalates, organic solvents, oil and grease, phenols and heavy metals. In 1983, the Rose Site was placed on the NPL.

All sites placed on the NPL must undergo a Remedial Investigation and Feasibility Study (“RI/FS”) to determine the extent of contamination and possible remedies. 42 U.S.C. § 9620(e)(1). Under a cooperative agreement with EPA, the Michigan Department of Natural Resources (“MDNR”) began the RI/FS evaluation of the Rose Site in 1984.

That study, completed in June of 1987, showed two primary areas of contamination: (1) an area which is less than one acre in size but contains groundwater contaminated by vinyl chloride and surface soils having elevated levels of arsenic; and (2) twelve acres in the southwest corner of the Site that contain surface soils contaminated with PCBs, lead, arsenic and other toxic metals; subsurface soils contaminated with a variety of volatile organic compounds (“VOCs”) and semi-volatile organic compounds (“SVOCs”); and groundwater contaminated with PCBs, metals, VOCs and SVOCs. A.

The RI/FS and the Original ROD After a detailed screening of possible remedies, the 1987 RI/FS recommended excavation and on-site thermal destruction to remedy the soil contamination, plus ground water treatment to cleanse the water under the Rose Site. Soil flushing, a method by which the contaminated soil is flushed with water and the resulting flus-hate is treated to designated cleanup levels and reinjected into the soil, was found to be ineffective at this Site due to the variable permeability of the Rose Site soils. RI/FS, Exh. 3.1a, Table 9-1, at 146. Pursuant to

section 117(

a) of CERCLA, 42 U.S.C. § 9617(a), which requires that the public be given a reasonable opportunity to comment on a proposed cleanup, EPA published a notice of the remedy and held a public meeting near the Site. In September 1987, EPA issued a Record of Decision (“ROD”), setting forth its proposed remedy as recommended in the RI/FS. The State of Michigan concurred in the ROD, which required, among other steps:

(1) Excavation of approximately 50,000 cubic yards of contaminated soil, incineration of the excavated soils that were contaminated with PCBs, VOCs and SVOCs, and proper treatment and disposal of the resulting incinerated ash; and

(2) Extraction and on-site treatment of contaminated ground water with diversion to adjacent marshlands or an alternate location. The 1987 ROD issued by EPA included a detailed explanation of the reasons for selecting the proposed remedy, and included specific findings that the remedy satisfied the requirements of CERCLA, complied with federal and state ARARs, and was cost effective. Soil flushing, though not adopted in the 1987 ROD, was not ruled out completely.

The ROD listed eight criteria EPA would consider before substituting soil flushing for thermal incineration: economies of scale, community acceptance, cleanup time, land regulations, reliability of soil flushing, implementability, complete site remediation, and cost effectiveness. B. The Proposed Consent Decree In June of 1987, shortly before issuance of the original ROD, EPA began settlement negotiations with the PRPs. The State of Michigan participated in these discussions.

In the course of the negotiations, EPA was persuaded that the soil flushing method might be a viable, less costly alternative to the incineration of the VOC/SVOC contaminated soil, and could still result in a cleanup that would comply with all federal and state ARARs. In August of 1988, EPA and the twelve PRPs who are defendants in this action signed the consent decree which included a soil flushing remedy for the site.

While under the original plan 50,000 cubic yards of contaminated soil were to be incinerated, the consent decree calls for incineration of only half that amount, augmented by soil flushing for the remaining 25,000 cubic yards. In economic terms this is represented as effecting savings of roughly $12 million.

To offset the danger that this process might be insufficient, the decree requires the PRPs to prove, both in a laboratory and at the Rose Site, that soil flushing is capable of meeting Phase I water target cleanup levels (“TCLs”) for the subsurface soils contaminated with VOCs and SVOCs within ten years after implementation of the system. Absent such proof, the PRPs would be required to fund and implement an alternate, permanent remedy designed to meet Phase I TCLs.

Under the proposed consent decree, EPA is required to review the remedial action at the site at least every five years, and is permitted to seek further response action from the defendants if EPA determines that supplemental remedies are necessary. The settling defendants are also required to provide EPA with monthly progress reports, and are subject to fines for failure to provide the reports or for delays in the implementation of the proposed remedial action.

The consent decree retains a requirement that PCBs above 10 parts per million (p.p.m.). at the Rose Site be incinerated either on-site or off-site, but as noted above, it does not incorporate the original remedial action plan’s requirement for incineration of all of the otherwise-contaminated soil. Under the decree, the settling defendants would be required to:

(1) Implement supplemental hydrogeo-logical studies regarding well placement, aquifers, permeability and porosity of unsaturated soil, placement of water extraction systems and characteristics of the soil;

(2) Install and maintain a ground water monitoring program;

(3) Excavate and incinerate all soils at the site containing PCBs in excess of 10 p.p.m.;

(4) Treat and bury soils containing lead in excess of 70 p.p.m.;

(5) Install and maintain a ground water extraction/treatment system that includes air stripping and carbon adsorption;

(6) Locate and treat wetlands on the site that contain PCBs in concentrations greater than 10 milligrams per kilogram;

(7) Construct and maintain a six-foot chain link fence around the site;

(8) Install and maintain a soil flushing system designed to remedy VOC and SVOC contaminated subsurface soil and, if the system proves ineffective, to submit within six months an alternate remedy; and

(9) Prepare the soil flushing plan; see details supra. In consideration of the work to be performed and the payments to be made by the settling defendants, the United States agrees in the proposed consent decree not to sue them, with some exceptions, for claims available under sections 106 and 107 of CERCLA and other federal and state environmental laws which are based on facts about the Site and its contamination known to EPA at the time of the entry of the decree.

The covenant does contain reopening provisions which would allow EPA to seek further injunctive relief or cost recovery if conditions unknown until after entry of the decree reveal that the remedial action is not protective of human health and the environment. See Consent Decree XVII. C. Proceedings in the District Court and the Amended ROD In September of 1988, EPA filed the proposed consent decree with the U.S. District Court for the Eastern District of Michigan pursuant to 42 U.S.C. § 9622(d)(1)(A).

As required by 42 U.S.C. § 9622(d)(2) and 28 C.F.R. § 50.7, notice of the proposed consent decree was published in the Federal Register on September 26, 1988. At the same time, EPA published a three page document entitled Proposed Settlement Plan — Explanation of Significant Differences (“ESD”). The ESD was published to comply with

section 9617(c), which requires EPA to explain why a settlement or consent decree to which the agency agrees differs in any significant respect from the final plan or ROD previously issued for a particular site. In this case the ESD explained the basis for the decision to allow defendants to try soil flushing at the Rose Site in conjunction with incineration, when the 1987 ROD had called for soil incineration only. As required by 42 U.S.C. § 9617(a), EPA provided a period for public comment on the proposed changes to the ROD.

EPA received written comments from the Michigan Department of Natural Resources and the Michigan Toxic Substances Control Commission, two congressmen, two private environmental organizations (the Environmental Defense Fund and the Michigan Environmental Council), several residents of Rose Township, and the settling defendants. Only the comments from the settling defendants expressed support for the terms of the consent decree.

Those who objected to soil flushing were concerned that it was not a well-demonstrated technology, especially in Michigan’s cold weather climate; that flushing may take as long as fifteen years to clean up the site as opposed to two years for incineration; that monitoring of soil flushing’s effectiveness is extremely difficult, and that flushing may violate Michigan’s groundwater anti-degradation laws. There were also concerns that the consent decree did not adequately define defendants’ obligations in the event soil flushing failed to achieve established cleanup levels within the required time frame.

The settling defendants asserted that the proposed consent decree would protect human health and the environment, and included a study by the Gradient Corporation, an environmental consulting firm, which estimated that approximately 12,325 pounds of organic chemicals would be removed by the soil incineration method and that approximately 12,234 pounds of organic chemicals would be removed by the soil flushing method.

The study added that the two amounts would be even closer in volume than this, because an additional amount of soil that was not to be incinerated under the original remedy would be subjected to soil flushing under the consent decree. On January 18, 1989, after considering the comments received, EPA issued an amended ROD for the Rose Site. The amended ROD formally adopts soil flushing as a remedy for VOC and SVOC-eon-taminated subsurface soils, but only if pilot testing proves that flushing is as protective as thermal destruction.

In adopting the remedy it originally ruled out, EPA reasons that (1) the excavation of PCB contaminated soils will remove most of the un-flushable contaminants; (2) the geology of the contaminated area may not be as complex as initially thought; and (3) pilot testing has not yet been performed to rule out soil flushing.

EPA, in the amended ROD, further asserts that (1) if Phase II target cleanup levels are achieved, flushing will have done as well as incineration was required to do under the original ROD, and will have brought the Site into compliance with all federal and state ARARs; (2) flushing is more cost effective than incineration; (3) assuming the groundwater treatment system uses granular activated carbon to capture the contaminants, soil flushing will satisfy CERCLA’s preference for remedies utilizing permanent and innovative treatments; and (4) soil flushing will reduce toxicity, mobility, and the volume of contaminants to the same extent as thermal destruction.

The State of Michigan filed a complaint with the district court and moved to intervene in the action between EPA and the settling defendants on February 14, 1989, pursuant to 42 U.S.C. § 9621(f)(2)(B). This provision allows a state to challenge a proposed consent decree which allegedly fails to meet the state’s environmental protection standards. On May 4, 1989, the U.S. District Court held that Michigan could intervene in order to challenge entry of the consent decree. On June 8, 1989, Michigan filed a brief opposing entry of the consent decree, and appended the affidavit of Robert A. Hayes, in which Mr.

Hayes discussed his scientific evaluation of soil permeability and the possible ineffectiveness of flushing at the Rose Site. The U.S. District Judge declined to consider the affidavit, as well as a memorandum drafted and submitted by EPA, concluding that the court’s review was limited to the administrative record as that record existed at the time EPA amended the ROD.

The district court did grant the motion of the Natural Resources Defense Council, the Environmental Defense Fund, and the Sierra Club to file a brief with the court as amici curiae, and their brief objecting to the entry of the consent decree was filed on June 30, 1989. On July 18, 1989, one day after oral argument, the district court granted EPA’s motion for entry of the consent decree. The State of Michigan moved for a rehearing, requesting the district court to remand to EPA or to grant an evidentiary hearing to determine whether the amended ROD complies with Michigan’s ARARs.

The court denied this motion when it issued a final Memorandum Opinion and Order on August 9, 1989, approving the decree and ordering its enforcement. United States v. Akzo Coatings of America, Inc., 719 F.Supp. 571 (E.D.Mich.1989). In its opinion, the district court held that Michigan’s groundwater anti-degradation law does represent an ARAR for purposes of CERCLA, but found that the consent decree embodying a soil flushing remedy did not violate the state ARAR.

The court found that Michigan’s concerns about the complex geology of the Site had been adequately addressed by EPA, and observed that soil flushing had been used, with state approval, at other Michigan sites. The district court concluded that, on the administrative record, EPA’s decision to enter into the consent decree was not arbitrary or capricious, and was reasonable, fair and not contrary to relevant federal and state laws.

In addition, the district court held that CERCLA’s provisions allowing EPA to settle claims for remedial action with the PRPs preempted the State of Michigan from imposing additional remedial action requirements on defendants under Michigan’s Water Resources Commission Act, M.C.L.A. § 323.6; Michigan’s Environmental Protection Act, M.C.L.A. § 691.1201 et seq.; and the common law of public nuisance. III. ISSUES ON APPEAL The State of Michigan now appeals the entry of the consent decree, and the district court’s finding that CERCLA preempts some of the state’s environmental remedies against these defendants.

The PRPs cross-appeal the district court’s finding that Michigan’s anti-degradation law is an ARAR. EPA appears as appellee in this action, and does not challenge the judgment of the district court. The specific issues on appeal are: A. What is the proper standard of review for consent decrees, and should the court consider supplemental evidence not appearing in the administrative record? B. Is the consent decree arbitrary and capricious? C. Is the consent decree fair, reasonable and adequate? D. Does the consent decree comply with CERCLA’s and Michigan’s applicable environmental provisions? E.

To what extent, if any, does the consent decree preempt state law claims for additional relief? These issues will be discussed in the order listed above. IV. THE STANDARD OF REVIEW A. The Consent Decree and the Administrative Record We must initially determine whether the district court applied the appropriate standard of review to the consent decree reached by EPA and the PRPs. The State of Michigan argues that the district court should have reviewed the consent decree under a de novo standard, rather than the more lenient arbitrary and capricious standard.

A court’s review process of a response action undertaken by EPA is guided by 42 U.S.C. § 96130, which provides:

(1) Limitation In any judicial action under this chapter, judicial review of any issues concerning the adequacy of any response action taken or ordered by the President shall be limited to the administrative record. Otherwise applicable principles of administrative law shall govern whether any supplemental materials may be considered by the court.

(2) Standard In considering objections raised in any judicial action under this chapter, the court shall uphold the President’s decision in selecting the response action unless the objecting party can demonstrate, on the administrative record, that the decision was arbitrary and capricious or otherwise not in accordance with law. Under a series of executive orders, the latest of which is codified at 42 U.S.C. § 9615, the functions of the president under CERCLA and SARA are delegated to EPA’s administrator, with the authority to redelegate.

EPA entered into the consent decree at issue in this ease under the authority of this provision.

CERCLA empowers the President, and those to whom he lawfully delegates authority, to remedy environmental problems such as the one at issue in Rose Township: In addition to any other action taken by a State or local government, when the President determines that there may be an imminent and substantial endangerment to the public health or welfare or the environment because of an actual or threatened release of a hazardous substance from a facility, he may require the Attorney General of the United States to secure such relief as may be necessary to abate such danger or threat, and the district court of the United States in the district in which the threat occurs shall have jurisdiction to grant such relief as the public interest and the equities of the case may require. 42 U.S.C. § 9606(a).

The State of Michigan argues that the consent decree negotiated between EPA and the PRPs was agreed to, but not “taken or ordered,” as described in section 96130(1), supra, or “selected]” as described in section 96130(2), supra. While these verbs are not defined in the CERCLA statute, we believe EPA, acting on the President’s behalf, did select the chosen remedy embodied in the consent decree, and has ordered that the terms of the agreement be carried out by the PRPs.

CERCLA authorizes the President to “remove or arrange for removal of, and provide for remedial action relating to” hazardous substances at a site, or to “take any other response measure” deemed necessary “to protect the public health or welfare or the environment.” 42 U.S.C. § 9604(a).

Section 9604(c)(4) authorizes the President to select remedies which will further his efforts to remove these hazardous substances and protect the environment. The statute specifically authorizes the President to enter into consent decrees with PRPs in order to achieve these goals. See 42 U.S.C. § 9622. We find that EPA’s decision to enter into a consent decree does represent a selection by the President of a remedy. As a result, CERCLA’s limitation of judicial review to the administrative record does apply here, and the district court properly declined to engage in a de novo review of the consent decree.

Our finding on this question of the standard of review is consistent not only with the language of CERCLA itself, but also with congressional intent concerning the role of agency expertise, and with the case law that has developed since the enactment of

section 9613(j). Ours should not be the task of engaging in a de novo review of the scientific evidence pro and con on each proposed remedy in the hazardous substance arena. The federal courts have neither the time nor the expertise to do so, and CERCLA has properly left the scientific decisions regarding toxic substance cleanup to the President’s delegatee, the EPA administrator and his staff. “When examining this kind of scientific determination ... a reviewing court must generally be at its most deferential.” Baltimore Gas & Electric Co. v.

Natural Resources Defense Council, Inc., 462 U.S. 87, 103, 103 S.Ct. 2246, 2255, 76 L.Ed.2d 437 (1983). Our role, as the CERCLA statute makes clear, is one of review on the administrative record, searching for errors of procedure and for glaring omissions or mistakes which indicate that EPA has acted arbitrarily and capriciously.

As the House Report on the SARA amendments notes: “limiting judicial review of response actions to the administrative record expedites the process of review, avoids the need for time-consuming and burdensome discovery, reduces litigation costs, and ensures that the reviewing court’s attention is focused on the criteria used in selecting the response.” H.R.Rep. No. 253, Pt. 1, 99th Cong., 1st Sess. 81 (1985), reprinted in 1986 U.S.Code Cong. & Admin.News 2835, 2863.

When reviewing a consent decree, a court need only “satisfy itself that the settlement is reasonable, fair, and consistent with the purposes that CERCLA is intended to serve.” H.R.Rep. No. 253, Pt. 3, 99th Cong., 1st Sess. at 19 (1985), reprinted in 1986 U.S.Code Cong. & Admin.News 2835, 3038, 3042 [hereinafter “H.R.Rep. No. 253, Pt. 3”]. Other courts have viewed this limited, yet important, role as we do. In United States v. Cannons Engineering Corp., 899 F.2d 79 (1st Cir.1990), the circuit court affirmed the district court’s approval of proposed consent decrees reached between EPA and PRPs under CERCLA.

The First Circuit stated in that case: “While the district court should not mechanistically rub-berstamp the agency’s suggestions, neither should it approach the merits of the contemplated settlement de novo.” 899 F.2d at 84. We recognize that one court has found that EPA’s selection of a remedy to clean up a hazardous waste site should be reviewed de novo in the district court. See United States v. Hardage, 663 F.Supp. 1280 (W.D.Okla.1987). However, we believe that court misinterpreted the plain language of CERCLA and the congressional intent behind the statute.

The Hardage court found that the environmental plan in that case had not been “selected,” but was merely a proposal which the court, not the President, was in a position to enforce. Therefore, reasoned the court, review of the proposed cleanup plan was not limited to the administrative record by

section 9613(j), since the court, not the President, was ordering a remedy. The Hardage distinction is without merit, and we believe the court improperly engaged in a de novo review of the remedy in that case. “If the Hardage court’s

interpretation of the statute is correct, Congress has enacted an unusual statutory scheme, one in which the scope and standard of review seems to hinge on whether EPA formally issues an order for a response plan or merely asks the court to enforce such a plan.” In re Acushnet River & New Bedford Harbor, 722 F.Supp. 888, 892 (D.Mass.1989). Another court has described the Hardage court’s distinctions as “hypertechnical and inconsistent with the plain meaning of the statutory language.” United States v. Bell Petroleum Servs., Inc., 718 F.Supp. 588, 591 (W.D.Tex.1989).

We agree with the view that the Hardage distinctions are improper, given 42 U.S.C. § 9621(a)’s description of actions “secured under

section 9606” as presi-dentially-selected remedial actions. The President, acting through EPA, does select a remedy when he presents a consent decree for court approval, and the standard of review found in

section 9613(

j) should apply to such decrees. The consent decree, as a judicial act, requires court approval. However, the court’s role is limited to approval or rejection of the decree, and it remains EPA’s responsibility to select the remedy and to take the steps necessary to bring the decree to the court for approval. We must respect Congress’ intent that the President develop such decrees, and that the courts review them on the administrative record under an arbitrary and capricious standard. Another case interpreting the standard of review under CERCLA is United States v. Ottati & Goss, Inc., 900 F.2d 429 (1st Cir.1990).

There the court held that review of an EPA request for injunctive relief to force several companies to clean up a hazardous waste site was not limited to the arbitrary and capricious standard. The court drew a distinction, with which we agree, between the court’s duty to enforce a “lawful (nonarbitrary) EPA order,” and the court’s discretion to accept or reject a “remedial injunction that EPA (lawfully and nonarbitrarily) decides is proper.” 900 F.2d at 434. As the Ottati & Goss court said, a reviewing court is not required to adopt EPA’s chosen remedy just because EPA believes the remedy is proper. Id.

However, we disagree with that court’s acceptance of the Hardage case’s distinction between remedial actions taken or ordered by the President, and proposed remedies which EPA asks a court to implement. Relying upon the language of

section 9606(a), which allows the President to require the Attorney General to secure such relief as may be necessary to abate the danger of a hazardous waste problem, and gives to the district court “jurisdiction to grant such relief as the public interest and the equities of the case may require,” the Ottati & Goss court upheld the district court’s modifications to EPA’s requested injunctive relief. We believe

section 9613(

j) reflects Congress’ intent that in this highly technical area, decisions concerning the selection of remedies should be left to EPA, and those decisions should be accepted or rejected — not modified — by the district court under an arbitrary and capricious standard. While Congress has directed district courts to “grant such relief as the public interest and the equities of the case may require,” § 9606(a), Congress has also left the crafting of that relief in the hands of qualified experts to whom the President delegates authority.

A reviewing court should not attempt to substitute its judgment for the expertise of EPA officials. Ours is the task of searching for errors of procedure, and serious omissions of substantive evidence, not the job of reformulating a scientific clean-up program developed over the course of months or years. Aside from the Hard,age and Ottati & Goss courts, federal courts have consistently reviewed environmental remedies formulated under CERCLA on the basis of the administrative record under an arbitrary and capricious test. United States v.

Wastecontrol of Florida, Inc., 730 F.Supp. 401 (M.D.Fla.1989); In re Acushnet River & New Bedford Harbor, 722 F.Supp. 888 (D.Mass.1989); United States v. Bell Petroleum Servs., Inc., 718 F.Supp. 588 (W.D.Tex.1989); United States v. Seymour Recycling Corp., 679 F.Supp. 859 (S.D.Ind.1987). Some courts, properly relying on the legislative history accompanying the statute, have also applied a three-part test of (1) fairness, (2) reasonableness, and (3) consistency with CERCLA’s goals. See H.R.Rep. No. 253, Pt. 3 at 19, supra. This test is similar to the standards applied before the 1986 SARA amendments.

United States v. Conservation Chemical Co., 628 F.Supp. 391, 400 (W.D.Mo.1985); United States v. Seymour Recycling Corp., 554 F.Supp. 1334, 1337-38 (S.D.Ind.1982). Review of consent decrees in our court has generally been conducted under similar standards. See United States v. Jones & Laughlin Steel Corp., 804 F.2d 348, 351 (6th Cir.1986); Williams v. Vukovich, 720 F.2d 909, 920-23 (6th Cir.1983). We view the standard of fairness, reasonableness and consistency with the statute — our court’s general test for consent decrees — coupled with the arbitrary and capricious standard of

section 9613(j), to be the proper tests for EPA’s proposed decree. We apply these tests to the administrative record before us, as did the district court. We must “consider whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment.” Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 823-24, 28 L.Ed.2d 136 (1971) (applying the arbitrary and capricious test of the Administrative Procedure Act, 5 U.S.C. § 706(2)(A)).

Though judicial review of a proposed consent decree under CERCLA must be performed under the arbitrary and capricious test, with an evaluation of the fairness and reasonableness of EPA’s decision, our review must be thorough and penetrating. One court, discussing its role in subjecting scientific evidence to the arbitrary and capricious standard of review has written: There is no inconsistency between the deferential standard of review and the requirement that the reviewing court involve itself in even the most complex evidentiary matters____ The close scrutiny of the evidence is intended to educate the court.

It must understand enough about the problem confronting the agency to comprehend the meaning of the evidence relied upon and the evidence discarded; the questions addressed by the agency and those bypassed; the choices open to the agency and those made. The more technical the case, the more intensive must be the court’s efforts to understand the evidence, for without an appropriate understanding of the case before it the court cannot properly perform its appellate function. Ethyl Corp. v. EPA, 541 F.2d 1, 36 (D.C.Cir.1975) (en banc).

In sum, in evaluating the efforts of an agency charged with making technical judgments and weighing complex data, we must give a proper degree of deference to the agency’s expertise, see Lile v. University of Iowa Hospitals and Clinics, 886 F.2d 157 (8th Cir.1989), yet also ensure that the agency has considered all of the relevant evidence in the record and has acted in the public interest. B. Evidence Submitted Outside of the Administrative Record The State of Michigan contends that the affidavit of Robert A. Hayes should have been considered by the district court during its review process.

This affidavit was filed with the district court on June 8, 1989. The affidavit indicates that Mr. Hayes is the Senior Hydrogeologist in the Compliance and Enforcement

Section of the Environmental Response Division of the Michigan Department of Natural Resources in Lansing, Michigan. He holds bachelors degrees from Wayne State University, and is certified as a Professional Geologist by the American Institute of Professional Geologists. Hayes conducted geophysical tests in March 1989, studying the subsurface soils in six locations at the Rose Site.

The State of Michigan submitted the Hayes affidavit with its brief to the district court, but the court refused to consider the affidavit because it had not been part of the administrative record considered by EPA and the PRPs at the time they drafted and filed the consent decree for the court’s approval on May 4, 1989. The district court held that since its review was limited to the administrative record, no additional supplementary materials could be considered by the court. Akzo Coatings, 719 F.Supp. at 582. The failure to consider the affidavit was erroneous under the circumstances of this case.

The district court allowed the State of Michigan to intervene in this action for entry of the consent decree on May 4, 1989, pursuant to Fed.R.Civ.P. 24 and 42 U.S.C. § 9621(f)(2)(B). The affidavit was filed the following month. CERCLA specifically provides that the State in which a remedial action plan is to be implemented should be given a reasonable opportunity to review and comment on the supporting technical data and engineering design of the plan. 42 U.S.C. § 9621(f)(1). The statute also provides for a public comment period before the court enters the consent decree as a final judgment.

The affidavit was filed over one month before the district court held a hearing on July 18, 1989 to consider approval of the consent decree. In light of the congressional intent expressed in the statute that public comment and state participation are to be encouraged and considered, we believe the district court improperly refused to accept the affidavit in June of 1989. Moreover,

section 9613(j)(

l) of CERCLA indicates that “[ojtherwise applicable principles of administrative law shall govern whether any supplemental materials may be considered by the court.” Our reading of related administrative law cases suggests that a reviewing court may consider materials supplementary to the administrative record in order to determine the adequacy of the government agency’s decision, even when the court’s scope of review is limited to the administrative record. In Norwich Eaton Pharmaceuticals, Inc. v.

Bowen, 808 F.2d 486 (6th Cir.1987), this court held that the district court had properly admitted evidence not found in the administrative record in reviewing a decision of the Food and Drug Administration. The additional evidence was required to determine whether the administrative record was adequate, and the district court based its decision on its review of the record.

As the Ninth Circuit has stated: It will often be impossible, especially when highly technical matters are involved, for the court to determine whether the agency took into consideration all relevant factors unless it looks outside the record to determine what matters the agency should have considered but did not. Asarco, Inc. v. EPA, 616 F.2d 1153, 1160 (9th Cir.1980).

Other courts have similarly held that a reviewing court evaluating agency action on the administrative record may consider additional evidence as either background information to aid the court’s understanding, or to determine if the agency examined all relevant factors or adequately explained its decision. See Missouri Coalition for the Environment v. Corps of Engineers of the U.S. Army, 866 F.2d 1025 (8th Cir.1989); Love v. Thomas, 858 F.2d 1347 (9th Cir.1988); Abington Memorial Hosp. v. Heckler, 576 F.Supp. 1081 (E.D.Pa.1983), aff'd, 750 F.2d 242 (3d Cir.1984).

However, the reviewing court “must be careful not to allow such evidence to change the character of the hearing from one of review to a trial de novo.” Town of Burlington v. Dep’t. of Educ., 736 F.2d 773, 791 (1st Cir.1984), aff'd on other grounds, 471 U.S. 359, 105 S.Ct. 1996, 85 L.Ed.2d 385 (1985). The court in Sterlingwear of Boston, Inc. v.

United States, 11 Cl.Ct. 879 (1987), held that a court conducting record review of agency proceedings may make findings of fact de novo where a party has demonstrated that proposed evidence is newly discovered or was unavailable to the agency at the time of its administrative action. Even so, we decline to transform the entire review process into a de novo consideration of the evidence in light of the Hayes affidavit because of the express language of CERCLA, which limits our review to an arbitrary and capricious standard on the record.

Furthermore, we believe that in a highly technical area such as the one at issue, federal courts are ill-equipped to engage in de novo review of such evidence presented to them during the public comment period prior to final entry of the decree. The district court should have admitted the Hayes affidavit into evidence, but only for the purpose of determining the adequacy of EPA’s decision, not in order to determine whether the decision was the best one available.

We must inquire, as the district court should have done, whether the information contained in the Hayes affidavit is of such significance that the agency must reconsider its decision in light of the new information, or whether the affidavit, when weighed against all of the other evidence available to EPA at the time it agreed to the consent decree, is insufficient to overcome the deference accorded EPA’s actions by a reviewing court applying the arbitrary and capricious test.

We recognize that this places the federal courts in the delicate position of weighing the technical strength of new evidence, while at the same time asking them to defer to agency expertise in scientific matters. Nonetheless, this treatment of new evidence reflects the intent of the statute, and comports with the proper limited role of courts in reviewing CERCLA consent decrees, even as those courts accommodate the arrival of new information or significant adverse public comment. Certainly our decision on the proper consideration of the Hayes affidavit is not reached without some reservations.

In other contexts, the Supreme Court has observed: “Administrative consideration of evidence ... always creates a gap between the time the record is closed and the time the administrative decision is promul-gated____ If upon the coming down of the order litigants might demand rehearing as a matter of law because some new circumstance has arisen, some new trend has been observed, or some new fact discovered, there would be little hope that the administrative process could ever be consummated in an order that would not be subject to reopening.” Vermont Yankee Nuclear Power Corp. v.

Natural Resources Defense Council, Inc., 435 U.S. 519, 554-55, 98 S.Ct. 1197, 1217, 55 L.Ed.2d 460 (1978) (quoting ICC v. Jersey City, 322 U.S. 503, 514, 64 S.Ct. 1129, 1134, 88 L.Ed. 1420 (1944)). Nevertheless, it is our opinion that additional evidence can properly be considered in order to ensure the soundness of the agency’s action. The standard may well prove much easier to apply than to define.

If, in the court’s admittedly unscientific judgment, some new evidence which was unavailable to the agency seems so significant that the agency’s original action now seems questionable, the reviewing court should remand the consent decree so that EPA’s experts can consider the new information. However, if the court finds that EPA would still have acted as it did even had the agency considered the new information, then the court may proceed to evaluate the consent decree on the administrative record using the arbitrary and capricious test.

Aside from the language of CERCLA and the case law concerning a district court’s consideration of additional evidence when reviewing an agency’s actions, our finding that the affidavit should have been considered by the district court is also sound from the viewpoint of public policy. In the technical and still-developing field of environmental science, new data regarding the efficacy of various cleanup remedies continues to come forward.

We can imagine a situation in which EPA enters into a consent decree embodying a proposed remedy, and only after this decree is filed with the court does a scientific study come forward discrediting the selected remedy because the agency reached erroneous conclusions, or perhaps because cutting-edge studies indicate that the proposal is technically flawed. In such a case, Congress cannot have intended, and sound principles of justice cannot allow a reviewing court to close its eyes and ears to the new evidence.

Were the court to enforce the decree simply because the information available to EPA at the time the decree was filed indicated that the remedial plan was not arbitrary or capricious, the court would surely not be acting in the best interests of the public if in fact the new data clearly showed that the plan would fail in its purpose. Undoubtedly, a district court’s response to new data in general, and the Hayes affidavit in this case, must be guided by the contents of the new evidence and its relationship to the balance of the evidence that was before the administrative agency.

We must engage, as the district court in this case should have engaged, in an evaluation of the Hayes affidavit and its possible effect on EPA’s decision to enter into the consent decree. Y. WHETHER THE CONSENT DECREE IS ARBITRARY AND CAPRICIOUS The State of Michigan first argues, along with amici curiae, that EPA’s decision to modify its ROD and consent decree to include soil flushing as a remedy for the Rose Site was arbitrary and capricious because the record does not support EPA’s conclusion that the Site is conducive to soil flushing.

Under the arbitrary and capricious standard, a lower court’s discretionary action “cannot be set aside by a reviewing court unless it has a definite and firm conviction that the court below committed a clear error of judgment in the conclusion it reached upon the weighing of the relevant factors.” McBee v. Bomar, 296 F.2d 235, 237 (6th Cir.1961); Cf. Motor Vehicle Manufacturers Ass’n v.

State Farm Mutual, 463 U.S. 29, 43, 103 S.Ct. 2856, 2866, 77 L.Ed.2d 443 (1983) (In articulating the arbitrary and capricious standard, the Supreme Court stated that it would “uphold a decision of less than ideal clarity if the agency’s path may reasonably be discerned.”). The 1987 RI/FS and ROD identified soil flushing as not applicable at the site for the following reasons: a. The soils are marginally suitable for this technology because of variable permeabilities; b.

The soils contain both soluble and insoluble chemicals — flushing is only reliable for soluble chemicals and would have to be used with another technology to remove the entire source; c. Pilot testing would have to be performed before such a remedy is implemented; and d. Flushing is not well demonstrated, especially in cold weather environments like that of Michigan. For the reasons that follow, we believe that the concerns noted in the RI/FS have all been adequately addressed by EPA in the ESD it published when it filed the decree with the district court.

See generally Exh. 3.18, Explanation of Significant Differences. As evidenced by its placement at the top of EPA’s concerns in 1987, there is no question that EPA originally considered the soil conditions at the Rose Site to be the prime deterent to the use of soil flushing. However, after the RI/FS was performed and after more soil samples were taken, EPA found that the soils to be flushed were not as complex as once thought: Recall from the RI/FS and the [original] ROD that any contaminated soils below the water table would not be excavated and incinerated (since no PCBs are present in those soils).

This includes clays. While the entire site may be geologically complex, only a small portion is intended to be evaluated for flushing. In that area some clays are present — either on the surface ([where] PCB removal [will be done by excavation and incineration]) or at or below the water table. Thus, a good portion of the clay problem either will be addressed or treated as in the [original] ROD. Exh. 3.22b, Responsiveness

Summary, at 21 (emphasis added). The ROD amendment likewise specifies: If one examines the logs of soil borings taken at Rose ..., it can be seen that clay lenses are not present in the entire contaminated area which would need to be flushed. Although the geology of the entire site as a whole is rather complex, the geology of the contaminated subsurface soils may not be. In some areas clay zones are found only in the water table. Other areas find clay at the surface which may have to be excavated due to the presence of PCB contamination. Exh. 3.22a, ROD amendment, at 3 (emphasis in original).

Amici admit that “the contaminated zones are not underlain by clay layers 10 to 40 feet thick, as are some of the non-contaminated areas,” but nevertheless argue that Figure 5-7, one of the soil boring charts on which the district court relied, “indicates continuation of clay layers through the areas of contamination.” That chart alone, however, does not persuade us that soil flushing should be avoided at the Rose Site. We recognize that at least six of the two dozen or more soil borings taken in the Southwest area of the site reveal a more complicated geology than does Figure 5-7.

However, two of those soil borings, one of which is located near the outer perimeter of the area to be flushed, did not detect volatile organic compounds. Compare Exh. 3.1a, Figure 5-7, Profile A-A’ with Exh. 3.1b, Soil Boring Logs RW 6D, RW 8D. Hence, the effectiveness of soil flushing in those areas of the site is not of prime importance. While two other borings revealed clay mixed with silt and sand from ground level down to six and seven feet respectively, Exh. 3.1b, Soil Boring Logs RW4, RWD 5, the excavation of PCB-contaminated soil may take care of most, if not all, of those clay layers.

One of the soil borings near the center of the area to be flushed does reveal quite a bit of clay at depths from 4.5 to 14.5 feet deep and 14.5 to 19.5 feet deep. The groundwater level at that location was measured at 22.52 feet. Id. Soil Boring RW 7. Thus, the clay layer in that location can be expected to interfere with the infiltration of flush water to the groundwater.

In addition to clay, this boring — as well as most of the soil borings previously discussed — reveals varying amounts of silt, sand and/or rock, all of which vary the soil’s permeability and raise the issue as to exactly how much interference with the flushing process there will actually be. It must be emphasized, however, that the Remedial Action Plan (“RAP”) annexed to the consent decree expressly requires that the settling defendants demonstrate to EPA, both in a laboratory and on-site, that soil flushing will work before it is implemented.

The required demonstration includes additional field tests to further define the permeability of the soils. Based on our thorough review of the scientific evidence in the record, we do not find EPA’s decision to experiment with soil flushing at the Rose Site to be arbitrary or without foundation. Nor does the Hayes affidavit {see supra), when viewed in light of the other evidence available to EPA at the time it agreed to the consent decree, render EPA’s most recent conclusions on the soil permeability of the Rose Site inadequate. As stated in the fact

summary, Hayes and two other Michigan Department of Natural Resources (“MDNR”) geologists visited the Rose Site in March of 1989 to conduct geophysical tests (gamma, neutron and gamma-gamma logs) on six monitoring wells previously installed on the site. Five of the wells are located within the area where the PRPs propose to use soil flushing, and the sixth well is less than one hundred feet from that area. “The purpose of these tests was to analyze the accuracy of the well driller’s logs for those six well locations.” Hayes Affidavit, Jt. App. at 533.

The geophysical logs, according to Hayes, “indicate that the drilling logs of the monitor wells tested are not very accurate with regard to detailed descriptions of the site geology.” Id. However, the difference in accuracy between the two types of logs performed on the Site does not appear to be substantial enough to overcome the deference accorded to EPA’s decision.

While it is claimed that the geophysical logs differentiate sand and clay zones precisely whereas the drilling logs conglomerate them into a stew of clay, silt, sand, and gravel, in general the former logs reveal the existence of clay layers at approximately the same depths as do the drilling logs. Consequently, much of the evidence the Hayes affidavit presents is only “supplementary” rather than “new,” and not necessary to our determination whether the agreement embodied in the decree is adequate. See, e.g., Asarco, Inc. v. United States EPA, 616 F.2d 1153, 1160 (9th Cir.1980).

Furthermore, whether the geophysical logs are capable of providing heightened soil type and permeability differentiation (as well as revealing any “previously undiscovered" clay layers) boils down to a credibility determination.

The state does not argue that EPA’s methods at the Rose Site to determine soil permeability are unconventional, but only that the geophysical tests are more accurate. “When specialists express conflicting views, an agency must have discretion to rely on the reasonable opinions of its own qualified experts even if, as an original matter, a court might find contrary views more persuasive.” Marsh v. Oregon Natural Resources Council, 490 U.S. 360, 378, 109 S.Ct. 1851, 1861, 104 L.Ed.2d 377 (1989).

We are not equipped to engage in the same technical evaluations of conflicting evidence that EPA and other experts are daily performing, especially when the evidence was submitted outside the administrative record. Fortunately, CERCLA does not ask us to do so, and we decline to proceed beyond the review procedures established in the statute. In sum, we do not find the Hayes affidavit to be of such significance that EPA must reconsider its selection of remedies for cleanup of the Rose Site.

We also believe that EPA’s responses to comments about the decree offered by residents, state agencies and politicians demonstrates the sufficiency of EPA’s, reconsideration of soil flushing as a potentially viable remedy for the Rose Site subsurface soils. For example, one of the main concerns expressed to EPA was that the clay lenses at the site may render grossly inaccurate the required testing defendants must conduct to measure soil flushing’s effectiveness.

Illustrative are the following comments: “Using flushate monitor wells to determine what can or cannot be flushed from the site soils ... is neither logical nor scientifically defensible” because “[f]lus-hate monitor wells receive flushate that percolates downward only in the vicinity above and around the well.” Accordingly, the wells “are not representative of the entire soil contamination” due to “unequal distribution of contamination coupled with the potential effect of channelization of flushwater [which] may result in varied concentrations and types of contaminants in samples collected only a few feet apart.” Exh. 3.22b, Responsiveness

Summary, at 7. EPA, however, has always recognized its responsibility to make sure the potential problems associated with the testing, as well as the implementation, of soil flushing are resolved before the cleanup begins: “The above comments involve design criteria. All these and more will need to be satisfactorily addressed by the PRPs before EPA allows flushing to occur at Rose. EPA believes that an adequate confirma-tional sampling system can be devised, however.” Id. at 7-8.

Actual placement and locations “of extraction and monitoring wells will be discerned after the hydrogeo-logic studies called for in the RAP are performed.” Id. at 14. The groundwater extraction and treatment system “will be in place to retain chemicals that may be missed by the flushing operation.” Id. at 8. Moreover, adequate soil sampling should reveal whether and to what extent there are individual pockets of missed residuals, which “will need to be addressed by the PRPs before they finish flushing.” Id. at 8.

While we have some concern for the current lack of concrete data as to exactly how effective soil flushing will be at the Rose Site, we are satisfied that EPA will obtain sufficient information to decide whether or not to implement soil flushing, and if so, to maintain its effectiveness. A conceptual model (drawing) of the flushing remedy was handed out by defendants at a public meeting in October, 1988. EPA stated that an actual working model would be made after laboratory testing was finished and the results would then be compiled and presented to the public. “At that time, U.S.

EPA will decide as to whether flushing may be performed in the field.” Id. The gathering of concrete data would then continue after implementation of soil flushing was allowed. Under the decree, defendants must provide EPA with monthly progress reports; they are required to update EPA annually on the effectiveness of the soil flushing system; and “are obliged to make all adjustments necessary to maximize” its effectiveness.

After five years of soil flushing, defendants must demonstrate to EPA that the soil flushing system will clean up the subsurface soil within ten years, or else they must adopt some other method. Exh. 3.18, Explanation of Significant Differences, at 3. In addition, defendants are subject to fairly heavy fines for failure to provide the reports and for delays in implementation of the proposed remedial scheme. We find EPA’s information-gathering timetable to be rational. Cf. United States v.

Cannons Engineering Corp., 899 F.2d 79, 88 (1st Cir.1990) (“[I]t would disserve a principal end of the statute — achievement of prompt settlement and a concomitant head start on response activities — to leave matters in limbo until more precise information was amassed.”). We also believe the consent decree provides adequate safeguards to ensure the veracity and timeliness of the required test results. Cf. United States v.

Hooker Chemicals & Plastics Corp., 540 F.Supp. 1067, 1074 n. 3 (W.D.N.Y.1982) (the decree and proposed remedy should not await completion of the tests required to better delineate the extent of chemical migration because the agreement provides adequate safeguards to ensure the veracity of the test results). For instance, the settling defendants must “assure that U.S. EPA personnel or authorized representatives are allowed access to any laboratory utilized by [them] in implementing this Consent Decree.” Consent Decree VIII. Moreover, the record reveals that soil flushing is already occurring at the Rose Site.

The proposed soil flushing program will basically accelerate the natural process of flushing. It is logical to assume that low permeable soils would have also redirected contaminants away as they migrated towards the water table. In essence, flush water may follow nearly the same path(

s) as the contaminants. Thus, highly permeable soils that are more heavily contaminated will be more intensely flushed than less permeable soils which are less contaminated. Id. at 21. The important distinction between the natural process now occurring and the proposed remediation program is that the flushate will be captured by extraction wells after it passes through the contaminated soils, and then returned to the surface for treatment.

In sum, we feel EPA has adequately explained the reversal of its prior conclusion that the Rose Site “soils are marginally suitable due to variable permeability.” The second reason EPA originally rejected soil flushing has also been satisfactorily resolved by the amended ROD. The ROD recognized that soil flushing, if used at all, would need to be used in conjunction with other technology to effectively clean up the Site and remove the insoluble chemicals in the soils.

The amended ROD complies with that concern, as soil flushing is to complement excavation and incineration and not to wholly substitute for the original remedy. Thus, most of the insoluble chemicals in the surface soils which cannot be flushed out, namely the PCBs and lead, will be excavated before soil flushing of the subsurface soils contaminated with soluble chemicals occurs. The decree also unquestionably resolves the third reason why soil flushing was screened out, i.e. pilot testing had not been performed.

As noted earlier, pursuant to the ROD amendment and the decree, pilot testing of the proposed remedy will be performed on the site. It must “be shown through laboratory or pilot studies that flushing would remove hazardous chemicals to such a degree that the operation would be as protective as removal of hazardous chemicals by excavation and thermal destruction. Otherwise, full scale flushing activities may not occur.” Exh. 3.22a, ROD Amendment, at 3 (emphasis added).

Finally, while EPA originally expressed some concern that soil flushing may not work well in cold climates, that remedy has already been selected at three sites in Michigan with the state’s concurrence. See, e.g., Exh. 3.8, ROD for U.S. Aviex Site, at 25-26. Even assuming it is a significant impediment to soil flushing, cold weather does not pose a problem all year long. Moreover, EPA will apparently require that the “[ejquipment ... be designed to enable its operation during the winter.” Exh. 3.22a, Responsiveness

Summary, at 21. Accordingly, we are satisfied that EPA has adequately explained its change of position. As evidenced by language in the 1987 ROD, in which the state concurred, EPA contemplated reopening the decision if soil flushing was found to be “practical” and “less expensive.” The public was therefore put on notice that soil flushing might be re-examined in the near future.

In evaluating the potential effectiveness of soil flushing at this particular site, EPA considered the RODs from four other sites at which soil flushing was used as a remedy as well as numerous articles discussing the use of soil flushing. The agency has also published and performed its own studies on the use of soil flushing to cleanup hazardous waste sites some of which had clayey soils, and thus is keenly aware of the remedy’s limitations. See Exh. 3.2-3.11; Jt.App. at 357-410.

Further, the Gradient Corporation, an environmental consulting firm, concluded that soil flushing would remove virtually the same amount of chemicals from the subsurface soil at the Rose Site as would incineration. Exh. 3.21i, Gradient Corporation Memorandum, at 11 (Oct. 26, 1988). Finally, it should be emphasized that while relatively new, soil flushing is nevertheless a “proven” technology. Exh. 3.8, ROD for U.S. Aviex Site, at Table 7. In addition to three sites in Michigan, as of March 1988 soil flushing was in use in various foreign countries and at a total of thirteen different Superfund sites.

An administrative agency should not be, and is not under CERCLA, estopped by its prior precedent from altering its decisions due to increased expertise. Michigan v. Thomas, 805 F.2d 176, 184-85 (6th Cir.1986); 42 U.S.C. § 9617(c).

We believe EPA’s conclusion that soil flushing (1) may work as well as incineration for the VOC-contaminated soils, (2) is protective of the human health and environment, and therefore (3) satisfies CERCLA’s preference for remedies that utilize permanent and innovative treatment to the maximum extent practicable to reduce toxicity, mobility or volume of hazardous substances, see exh. 3.22a, at 1, is rational and supported by the record. No clear error of judgment was made by the district court in approving EPA’s change of position. See McBee, 296 F.2d at 237. VI.

WHETHER THE DECREE IS FAIR, REASONABLE AND ADEQUATE As we have observed earlier, in addition to determining whether a decree is rational and not arbitrary or capricious, we must satisfy ourselves that the terms of the decree are fair, reasonable and adequate — in other words, “consistent with the purposes that CERCLA is intended to serve.” H.R.Rep. No. 253, 99th Cong., 1st Sess. Pt. 3 at 19, supra; United States v. Hooker Chemical & Plastics Co., 607 F.Supp. 1052, 1057 (W.D.N.Y.1985) (citation omitted). While we are to “eschew any rubber stamp approval in favor of an independent evaluation,” City of Detroit v.

Grinnell Corp., 495 F.2d 448, 463 (2d Cir.1974), we may not substitute our own judgment for that of the parties to the decree. United States v. Jones & Laughlin Steel Corp., 804 F.2d 348 (6th Cir.1986) (a reviewing court may not modify but only approve or reject a consent decree). Protection of the public interest is the key consideration in assessing whether a decree is fair, reasonable and adequate. Acushnet River & New Bedford Harbor: Proceedings re Alleged PCB Pollution, 712 F.Supp. 1019, 1028 (D.Mass.1989); United States v. Ketchikan Pulp Co., 430 F.Supp. 83, 86 (D.Ala.1977).

In determining whether a decree is “fair,” courts have considered the following: “the strength of plaintiffs case, the good faith efforts of the negotiators, the opinions of counsel, and the possible risks involved in the litigation if the settlement is not approved.” Hooker Chemical & Plastics Co., 607 F.Supp. at 1057; U.S. v.

Cannons Engineering Corp., 720 F.Supp. 1027, 1039-40 (D.Mass.1989). “Fairness should be evaluated from the standpoint of signatories and nonparties to the decree.” Conservation Chemical Co., 628 F.Supp. at 401. “The effect on non-settlers should be considered, but is not determinative in the court’s evaluation.” Cannons Engineering Corp., 720 F.Supp. at 1040 (citing Acushnet River, 712 F.Supp. at 1029).

The good faith efforts of the parties to the decree are evidenced by the voluminous record, the arms-length negotiation process and the manifested willingness of EPA to thoroughly consider all oral and written comments made with regard to the proposed decree. The State of Michigan itself was involved for more than six months in the post-1987 ROD negotiations that led to an amended ROD and the consent decree before it eventually withdrew from the settlement discussions. See Exh. 3.12, Rose Township Chronology of Events.

We note and emphasize that at one time during the negotiation process between EPA and the PRPs, the state was willing to consider soil flushing under conditions similar to those imposed by the consent decree. See Exh. 3.14, Letter from Gary Guenther, Chief Environmental Response Division (MDNR) to Basil Con-stantelos, Director of Waste Management Division (EPA) (May 20, 1988) [hereinafter “Letter from Guenther”]; Exh. 3.12, Rose Township Chronology of Events.

The only significant difference between the settlement contemplated by the state and the consent decree at issue is that the latter does not incorporate a specified remedy should soil flushing fail. Id. Exh. 3.14, Letter from Guenther. The respective legal positions of the parties also militate in favor of the settlement. The strength of the government’s case against defendants is not well documented at this point. The settling defendants have maintained throughout this litigation that the evidence linking them to the Rose Site is extremely tenuous.

Consequently, if the decree is overturned, the parties will no doubt engage in a protracted legal battle over liability and the appropriate remedy for the Site. In enacting the 1986 amendments to CERCLA, however, Congress sought to “expedite effective remedial actions and minimize litigation.” 42 U.S.C. § 9622(a). We have stated before that one of CERCLA’s main goals is “the prompt cleanup of hazardous waste sites.” Walls v. Waste Resource Corp., 761 F.2d 311, 318 (6th Cir.1985).

Given that (1) liability of the settling defendants is presently uncertain and (2) dumping at the site started some twenty years ago, we are sensitive to EPA’s desire to finalize a remedial plan and force the PRPs to get on with the job and clean up a long-standing mess. Cf. United States v. Cannons Engineering Corp., 899 F.2d 79, 90 (1st Cir.1990) (“[I]f the case is less than robust, or the outcome problematic, a reasonable settlement will ordinarily mirror such factors.”); United States v.

McGraw-Edison Co., 718 F.Supp. 154, 159 (W.D.N.Y.1989) (settlement reasonable in light of prospect of protracted litigation as contrasted to expeditious reimbursement and remedy). Moreover, we are faced with a presumption in favor of voluntary settlement. That presumption is particularly strong where a consent decree has been negotiated by the Department of Justice on behalf of a federal administrative agency like EPA which enjoys substantial expertise in the environmental field. United States v. Cannons Engineering Corp., 899 F.2d 79, 84 (1st Cir.1990).

We note that a settlement “is not an opportunity to avoid any of the cleanup requirements or procedures of the act.” 132 Cong.Rec. S 14,918 (daily ed. Oct. 3, 1986) (Statement of Sen. Mitchell). As discussed in subsection VII B, infra, however, the decree incorporates all “cleanup requirements” imposed by CERCLA and SARA, and the “procedural requirements” of these acts are not contested. In evaluating the decree, it is not our function to determine whether this is the best possible settlement that could have been obtained, but only whether it is fair, adequate and reasonable. See, e.g., Durrett v.

Housing Auth., 896 F.2d 600, 603-04 (1st Cir.1990). Accordingly, based on the legal posture of the parties, the nature of the negotiation process that led to the decree, and the need to expedite the cleanup at the Rose Site, we agree with the district court that this settlement is fair. In determining whether a consent decree is “reasonable” courts have considered the following: the nature/extent of hazards; the degree to which the remedy will adequately address the hazards; possible alternatives for remedying hazards; and the extent to which the decree furthers the goals of the statute.

Cannons Engineering Corp., 720 F.Supp. at 1038. See also Conservation Chemical, 628 F.Supp. at 391; United States v. Seymour Recycling Corp., 554 F.Supp. 1334, 1339 (S.D.Ind.1982). “These criteria reflect the court’s ‘limited duty’ to inquire into the technical aspects of the cleanup program proposed by a consent decree in order to ensure that the proposed settlement adequately addresses environmental and public health concerns.” Seymour Recycling Corp., 554 F.Supp. at 1038 (citing Hooker Chemicals & Plastics Corp., 540 F.Supp. at 1072).

The most important of these “reasonableness” factors, the decree’s likely effectiveness as a vehicle for cleansing the Rose Site, has already been addressed under the arbitrary and capricious standard and thus will not be reexamined. See Marsh, 490 U.S. at 377 n. 23, 109 S.Ct. at 1861 n. 23. In our opinion the decree is binding on the settling defendants. Therefore, contrary to amici’s suggestion, the fact that the decree fails to spell out an alternate remedy in the event soil flushing fails does not give settling defendants carte blanche.

Should soil flushing prove to be unfeasible, the selection of a significantly different alternative remedy would be subject to the public participation requirements in

section 9617 and state participation requirements in

section 9621(f), as well as judicial review under

section 9613(h)(4). Moreover, regardless of the effectiveness of soil flushing which must be proven within specific time limits, defendants are required to remedy the site to Phase I TCLs. Once Phase I TCLs have been met, cleanup to Phase II TCLs (ARARs) will be performed by EPA. In other words, should soil flushing be rejected, the settling defendants would still have an incentive to make sure that whatever remedial action were implemented in its place attained all ARARs and did not further degrade the environment. As long as ARARs will be attained, no CERCLA provision prohibits the use of open-ended remedial schemes. In fact,

section 9621(d)(4)(A), discussed infra in subsection VIIB, implicitly provides support for such decrees. We find that the Rose Site decree is carefully structured so as to ensure the protection of human health and the environment while providing reasonable flexibility to the PRPs. For the hazardous chemicals common to both the sites, the Rose Site TCLs are at or below (more stringent than) the levels agreed to by the State of Michigan at the U.S. Aviex Site, where soil flushing is also to be used.

Both the Rose Site Phase I and Phase II cleanup standards are at or substantially below the MCLs, which are promulgated water standards under the Safe Drinking Water Act (“SDWA”). When remediated to the Phase I levels, the groundwater at the site will be as clean or cleaner than water which is acceptable for the 245 million people in this country to drink. See Exh. A, attached to Brief for Settling Defendants. Other courts have accepted open-ended decrees.

See Partial Consent Decree, Cannons Engineering Corp., 720 F.Supp. at 1027 (“If the sampling results indicate that the remedial goal has not been attained, a decision will be made to either continue vacuum extraction for a specified length of time and resample, or to complete the remedy with some other suitable technology.”); Hooker Chemicals & Plastics Corp., 540 F.Supp. at 1076-77 (while the decree is not specific as to methods to be utilized to prevent further contamination should remedies already specified fail, it is nevertheless reasonable and in the public’s best interest).

We find no reason to strike down the open-ended provision of this decree. As CERCLA recognizes, the cleanup of hazardous wastes involves too many variables not to allow the settling defendants to carefully test viable, cost-effective remedies at a particular site. Indeed, the original ROD to which the state consented also allowed for the substitution or modification of the proposed remedy, i.e. excavation and incineration, in the event cost became a factor.

The decree’s failure to specify an alternative remedy may in fact be in the public’s best interest, as it allows the parties to consider the nature of remaining contamination, the effect of changed soil conditions, and the use of new remedial technologies. Cf. Akzo Coatings, 719 F.Supp. at 585. Already there are other methods to remove VOCs from soil besides incineration, such as thermal aeration and vacuum extraction. In determining the reasonableness of a consent decree, we must also consider the cost effectiveness of its proposed remedial action. See 42 U.S.C. § 9621(b).

Assuming soil flushing proves to be feasible at the Rose Site, it will result in the accomplishment of the same TCLs set forth in the 1987 ROD at 29% of the cost of the original remedy. While a lot of money may be saved through the use of soil flushing, we find no evidence of a sweetheart deal here between EPA and defendants. The settling defendants must meet Phase I TCLs regardless of cost.

It is estimated that the $500,000 to be placed by defendants in a trust fund to pay for the costs of cleanup from Phase I TCLs to Phase II TCLs will grow to about $1,200,000 before it will be needed, and that sum, EPA concludes, should be sufficient to finish the cleanup. With those funds, EPA will operate the water extraction and monitoring systems already in place.

In response to the public comment that the PRPs should themselves perform the entire cleanup, EPA stated: We believe that it was important to reach a settlement for this case to save the Superfund $30 million, although it means taking on the small burden of continuing the cleanup using the trust funds. Since the Settling Defendants are paying for the trust fund as well as the site work, the main burden of cleanup is in fact on them. Exh. 3.22b, Responsiveness

Summary, at 13. In accordance with CERCLA’s goals, the primary and the ultimately forseeable financial responsibility is on the settling defendants. See Walls v. Waste Resource Corp., 761 F.2d 311, 318 (6th Cir.1985) (“[I]t is clear that the statute was designed primarily to facilitate the prompt cleanup of hazardous waste sites by placing the ultimate financial responsibility for cleanup on those responsible for the hazardous wastes.”).

In addition, we find that the time required for the total cleanup of the site, while most likely extended a few years due to the incorporation of the soil flushing remedy, is not unreasonable. Indeed, the time required to comply with the decree’s remedial requirements may not be significantly longer than that required for the remedy as originally proposed. Soil flushing may extend the cleanup to 10 to 15 years. However, “[t]wo or three years of incine

[…]