The parties disagree about the kind of pipe to be used in monitoring a cleaned-up landfill operated by plaintiff-appellant Browning-Ferris Industries of South Jersey, Inc. (BFI). They use this dispute to pose fascinating and far-reaching questions of first impression concerning environmental substantive law, procedural law and jurisdiction of the courts and relevant administrative agencies. We see the dispute as a pipe case. In the interests of judicial restraint and deference to the agency charged with primary responsibility for protecting the environment, we prefer to assume colorable jurisdiction for the purposes of this appeal only, without deciding the jurisdictional questions. On the merits, BFI’s legal and equitable claims must fail.

I.

FACTS

BFI operated the South Brunswick, New Jersey landfill from 1958 until its closing in 1978. In 1980 contaminated groundwater was discovered near the landfill. In 1981 BFI and the Environmental Protection Agency (EPA) entered into an administrative Consent Order pursuant to section 7003 of the Resource Conservation and Recovery Act (RCRA). 42 U.S.C. § 6973. This RCRA Order required BFI to prevent contamination of the groundwater due to migrating pollutants from the landfill.

The RCRA Order included a monitoring program, to consist of “appropriate sampling and analysis to be conducted over a three-year period.” The type of material to be used in the monitoring wells from which water samples would be drawn is not mentioned in the Order, but there is no dispute that the monitoring agreed upon requires wells.

BFI spent over $8,000,000.00 in cleaning up the dump. According to EPA, BFI “effectively remediated [the site], thereby mitigating the threat of release of contaminants into the environment which could present an imminent and substantial endangerment to human health and the environment.”

In 1982 EPA placed the landfill on its Superfund National Priorities List and continued its investigation pursuant to the Comprehensive Environmental Response, Compensation, and Liability Act (CERC-LA). 42 U.S.C. §§ 9605(8)(B) & 9606(a). Under CERCLA, EPA “shall, as appropriate,” prepare a “Remedial Investigation/Feasibility Study” to “determine the nature and extent of the threat” presented by the landfill. 40 C.F.R. § 300.68(d) (1989). In this case, EPA did not prepare a new study, but instead reviewed the investigation that had prompted the 1981 RCRA Order.

In April 1987 EPA notified BFI by letter that it required monitoring wells at the site *153to be made of stainless steel. In June 1987 BFI submitted a proposed monitoring plan to EPA providing for wells made of polyvinyl chloride (PVC).

In August 1987 EPA disseminated for public comment a Record of Decision pursuant to CERCLA. The Decision described the appropriate remedy for the site and the status of BFI’s clean-up effort. While EPA expressed satisfaction with BFI’s cleanup, it did not discuss the type of wells that should be installed for the monitoring program. EPA did note that “[i]t is anticipated that BFI will implement this program, under EPA supervision, beginning in the fall of 1987,” and indicated that “EPA may modify the monitoring program ... as necessary in order to fully evaluate the remedy’s protectiveness of human health and the environment.” The Decision was issued in September 1987.

In December 1987 EPA advised BFI by letter that the monitoring wells must be made of stainless steel. BFI responded in February 1988 in a letter indicating that it would install PVC wells beginning that month. EPA reiterated its insistence on stainless steel in a telephone call to BFI. The record does not reflect whether this call occurred before BFI began installation.

BFI installed PVC wells in February 1988. EPA continued to object. Both sides continued oral discussions and exchanged correspondence on the subject.

In November 1988 BFI suggested that four stainless steel wells be constructed alongside some of the PVC wells so that their performance could be compared. EPA did not agree.

The dispute continued until March 1989 when EPA issued an order claiming as authority section 106 of CERCLA. 42 U.S.C. § 9606(a). The CERCLA Order required BFI to submit a new monitoring plan that provided for the installation of stainless steel wells. Once the plan was approved by EPA, it would be incorporated into the Order so that a violation of the plan would be, according to EPA, a violation of the CERCLA Order. The CERCLA Order by its terms, and in EPA’s submission, “superseded” that part of the RCRA Order relating to monitoring.

Violation of a CERCLA Order may result in civil penalties of up to $25,000 per day and punitive damages up to three times the amount of costs incurred by EPA as a result of the violation. See 42 U.S.C. §§ 9606(b), 9607(a) & 9607(c).

II.

PROCEDURE IN DISTRICT COURT

BFI commenced an action in district court seeking declaratory and injunctive relief. It claimed that EPA breached the RCRA Consent Order by issuing the CERCLA Order; that under the Administrative Procedure Act (APA), 5 U.S.C. § 706, this breach was arbitrary, capricious, an abuse of discretion and not in accordance with law; that, under the fifth amendment to the Constitution, the use of the CERCLA Order to avoid resolution of the RCRA dispute denied BFI due process of law; that EPA’s election to proceed under RCRA estopped EPA from using the CERCLA Order to supersede the RCRA Order; that EPA’s determination that conditions at the landfill presented an imminent and substantial danger was unsupported by the record, was not made in good faith, denied BFI due process, and was an abuse of discretion; that EPA’s requirement of stainless steel wells as if the requirement were pursuant to an EPA rule denied BFI due process, and was an abuse of discretion under the APA; that the damage provisions of CERCLA denied BFI due process; that BFI had sufficient cause for refusing to comply with the CERCLA Order; and that, finally, it could install PVC instead of stainless steel well casing since PVC was a preferable material because stainless steel would corrode and contaminate the water samples.

In an opinion and order dated May 10, 1989, the district court granted EPA’s motion to dismiss the complaint for lack of subject matter jurisdiction. See Browning-Ferris Industries of South Jersey, Inc. v. Muszynski, 89 CV 1929, slip op., 1989 WL 51916 (S.D.N.Y. May 10, 1989). *154The court decided that a provision of the Superfund Amendments and Reauthorization Act of 1986 (SARA), 42 U.S.C. § 9613(h), explicitly prohibits pre-enforcement challenges to CERCLA Orders. It held the CERCLA Order proper, and found nothing in the RCRA Order restricting EPA from taking additional action under CERCLA. The court denied BFI’s motion for a preliminary injunction, holding that BFI had not made a sufficient showing of irreparable harm, and it cited this fact as an independent ground for granting the motion to dismiss.

The district court concluded that BFI had a sufficient remedy if it complied with the CERCLA Order and then sued EPA for reimbursement, pursuant to 42 U.S.C. § 9606(b)(2). If BFI’s compliance resulted in further contamination of the groundwater, the court pointed out, BFI could raise the alleged breach of the RCRA Order and its warning about using stainless steel as an equitable defense in any EPA-initiated action. We do not answer these questions decided by the district court since we have not reached the issue of the authority of the EPA under CERCLA.

III.

CONTENTIONS ON APPEAL

BFI claims that the district court erred as a matter of law and fact in finding no subject matter jurisdiction and in concluding that there were insufficient grounds for a preliminary injunction. EPA asserts that the district court correctly dismissed the complaint because there is no pre-en-forcement judicial review of CERCLA orders. Nothing, it contends, in RCRA, the APA, the Constitution or common law provides a basis for pre-enforcement judicial review of a decision characterized by it as a CERCLA order.

Both sides urge the court to reach a decision in their favor on the broadest grounds possible. BFI would have the court declare that CERCLA can never be used to modify a RCRA consent order. EPA would have the court rule that CERC-LA gives it complete freedom to modify such orders.

Similarly broad rulings are sought on the question of subject matter jurisdiction. BFI would like the court to find that pre-enforcement judicial review of a RCRA consent order is always available, even if this review necessarily requires pre-en-forcement review of a CERCLA order. Conversely, EPA would have the court find that the promulgation of a CERCLA order, with its express statutory preclusion of pre-enforcement review, automatically precludes pre-enforcement review of prior EPA action pursuant to other statutes such as RCRA.

Resolving these jurisdictional questions would involve the court in a complex analysis of congressional design and the relationship among CERCLA, RCRA and SARA. A comprehensive ruling on the jurisdictional issues would necessarily have a broad impact on future EPA pollution remediation efforts. While not dispositive, the express congressional prohibition on pre-enforcement judicial review of CERC-LA orders, 42 U.S.C. § 9613(h), suggests a general congressional design to minimize judicial intervention in the hazardous waste cleanup process.

IV.

ASSUMPTION OF JURISDICTION

Although it is customary to resolve subject matter jurisdiction issues before reaching the merits, there are exceptions to this rule that have been spelled out in a long line of decisions. See, e.g., Duke Power Co. v. Carolina Environmental Study Group, Inc., 438 U.S. 59, 72 n. 16, 98 S.Ct. 2620, 2629 n. 12, 57 L.Ed.2d 595 (1978) (with jurisdiction over claims against one defendant established, court assumed jurisdiction as to same claims against other defendant, because outcome would not be affected); Norton v. Mathews, 427 U.S. 524, 96 S.Ct. 2771, 49 L.Ed.2d 672 (1976), discussed infra; Philbrook v. Glodgett, 421 U.S. 707, 721-22, 95 S.Ct. 1893, 1902-03, 44 L.Ed.2d 525 (1975) (“complex question of federal jurisdiction” avoided where *155jurisdiction existed over claims as to one defendant); McLucas v. De Champlain, 421 U.S. 21, 32, 95 S.Ct. 1365, 1372, 43 L.Ed.2d 699 (1975) (court-martial review case in which Court noted doubt that the district court had subject matter jurisdiction, but ignored this question, and disposed of the case on merits); Secretary of the Navy v. Avrech, 418 U.S. 676, 677-78, 94 S.Ct. 3039, 3039-40, 41 L.Ed.2d 1033 (1974) (per curiam) (same), discussed infra; Chandler v. Judicial Council of the Tenth Circuit, 398 U.S. 74, 86-89, 90 S.Ct. 1648, 1654-56, 26 L.Ed.2d 100 (1970) (motion for leave to file writ of mandamus or prohibition; Court avoided “knotty jurisdictional problem” and denied motion on merits); United States v. Augenblick, 393 U.S. 348, 350-52, 89 S.Ct. 528, 530-32, 21 L.Ed.2d 537 (1969) (collateral review of court-martial; court assumes jurisdiction arguendo and decides case on merits); Grunenthal v. Long Island Railroad, 393 U.S. 156, 89 S.Ct. 331, 21 L.Ed.2d 309 (1968) (avoids jurisdictional issues of power of court of appeals in order to avoid constitutional question); Neese v. Southern Railway, 350 U.S. 77, 76 S.Ct. 131, 100 L.Ed. 60 (1955) (per curiam) (same). See generally, Comment: Assuming Jurisdiction Arguendo: The Rationale and Limits of Hypothetical Jurisdiction, 127 U.Pa.L.Rev. 712 (1979).

For example, in Secretary of the Navy v. Avrech, 418 U.S. 676, 677-78, 94 S.Ct. 3039, 3039-40, 41 L.Ed.2d 1033 (1974), the Court considered a void-for-vagueness challenge to a statute that was the basis for appellant’s court martial. The Court stated, “[W]e are unwilling to decide the difficult jurisdictional issue which the parties have briefed.” Id. at 677, 94 S.Ct. at 3040. It assumed the existence of jurisdiction in order to reach the merits of the case, denying appellant’s claim based on prior controlling case law.

In Philbrook v. Glodgett, 421 U.S. 707, 721-22, 95 S.Ct. 1893, 1902-03, 44 L.Ed.2d 525 (1975), appellees sued both a state and a federal agency in connection with the construction of a state regulation affecting administration of the federal Social Security Act. It was unsettled whether the district court had jurisdiction as against the federal agency. The Court noted that the jurisdictional question was complex, and that the parties’ briefs were inadequate. Jurisdiction as to the state agency was established, and the substantive issues decided as to it would be dispositive. Since the federal agency had indicated its intention to comply with the district court’s ruling, the Court assumed pendent jurisdiction as to the federal agency and found for the plaintiff. Even though this case is unusual in assuming jurisdiction and then holding in favor o/the party asserting jurisdiction, it is consistent with the other cases in that the assumption of jurisdiction did not affect the outcome.

The Court enunciated most clearly its practice of avoiding difficult and sweeping jurisdictional questions in favor of deciding clear-cut and narrow issues on the merits in Norton v. Mathews, 427 U.S. 524, 96 S.Ct. 2771, 49 L.Ed.2d 672 (1976). Appellant was denied survivor benefits under the Social Security Act because he was born out of wedlock. The subject matter jurisdiction of the three-judge district court that denied the benefits was unclear. The jurisdiction of the Supreme Court in turn depended on the jurisdiction of the district panel. The Supreme Court noted that appellant’s claim would necessarily be denied on the merits according to prior controlling case law, “rendering] the merits in the present case a decided issue and thus no longer substantial in the jurisdictional sense.” Id. at 530-31, 96 S.Ct. at 2775. The outcome of the case would be the same if the court found jurisdiction and ruled against the appellant on the merits, or found no jurisdiction and dismissed the appeal.

It thus is evident that, whichever disposition we undertake, the effect is the same. It follows that there is no need to decide the theoretical question of jurisdiction in this case. In the past, we similarly have reserved difficult questions of our jurisdiction when the case alternatively could be resolved on the merits in favor of the same party. See Secretary of the Navy v. Avrech, 418 *156U.S. 676 [94 S.Ct. 3039, 41 L.Ed.2d 1033] (1974).

Id. 427 U.S. at 532, 96 S.Ct. at 2775. The three dissenting justices preferred to find jurisdiction and uphold appellant’s claim on the merits.

Courts of appeals have followed the Supreme Court’s lead in assuming jurisdiction and ruling on the merits against the party invoking jurisdiction. See, e.g., Edwards v. Carter, 580 F.2d 1055, 1056-57 (D.C.Cir.), cert. denied, 436 U.S. 907, 98 S.Ct. 2240, 56 L.Ed.2d 406 (1978), discussed infra; Adams v. Vance, 570 F.2d 950 (D.C.Cir. 1978) (per curiam), discussed infra; Ripon Society v. National Republican Party, 525 F.2d 567, 576 n. 26 & 578 n. 28 (D.C. Cir. 1975) (assuming, without deciding, jurisdiction), cert. denied, 424 U.S. 933, 96 S.Ct. 1147, 47 L.Ed.2d 341 (1976); Kaiser v. Armstrong World Industries, Inc., 872 F.2d 512, 514 (1st Cir. 1989) (court assumes jurisdiction arguendo and holds that plaintiff’s damages claim is time barred); Federal Deposit Insurance Corp. v. Caledonia Investment Corp., 862 F.2d 378, 381 (1st Cir. 1988) (“[s]ince we affirm on the merits, however, we need not decide the jurisdictional issue because the result is the same”); Switlik v. Hardwicke Co., 651 F.2d 852 (3d Cir.), cert. denied, 454 U.S. 1064, 102 S.Ct. 614, 70 L.Ed.2d 601 (1981), discussed infra; Mitchell v. West Feliciana Parish School Board, 507 F.2d 662, 666-67 (5th Cir. 1975), discussed infra; Southern Pacific Transportation Co. v. Usery, 539 F.2d 386, 389 n. 1 (5th Cir. 1976), cert. denied, 434 U.S. 874, 98 S.Ct. 222, 54 L.Ed.2d 154 (1977) (consolidated cases in which court avoided challenge to jurisdiction as to one case because other cases were clearly within court’s jurisdiction); Forster v. County of Santa Barbara, 896 F.2d 1146 (9th Cir. 1990) (ignoring jurisdictional question because of factual dispute, unresolved at district court level, over whether appellant filed timely notice of appeal); Wolder v. United States, 807 F.2d 1506, 1507 (9th Cir. 1987) (per curiam) (“where the jurisdictional question is complex and the appeal is clearly without merit,” court will avoid jurisdictional question and rule on merits); Lehner v. United States, 685 F.2d 1187 (9th Cir. 1982) (court avoids question of whether jurisdiction exists over claims for money damages, because jurisdiction over equitable claims was clear, and merits would not be affected), cert. denied, 460 U.S. 1039, 103 S.Ct. 1431, 75 L.Ed.2d 790 (1983). Cf. Friedman v. Beame, 558 F.2d 1107, 1110 n. 3 (2d Cir. 1977) (assuming standing in order to reach the merits); Franklin Savings Bank of New York v. Levy, 551 F.2d 521, 528 (2d Cir. 1977) (characterizing as “jurisdictional” the question of whether a note in issue was a security under section 10(b) of the Securities and Exchange Act of 1934, and avoiding the question to reach the “merits”).

In Edwards v. Carter, 580 F.2d 1055, 1056-57 (D.C.Cir.), cert. denied, 436 U.S. 907, 98 S.Ct. 2240, 56 L.Ed.2d 406 (1978), for example, the court of appeals considered a challenge by members of the House of Representatives to the transfer of the Panama Canal to Panama pursuant to a self-executing treaty. The district court had dismissed the case for lack of standing. The court of appeals affirmed, but on the merits, for failure to state a claim upon which relief could be granted. It assumed the existence of jurisdiction, because the jurisdictional issues were complex (going beyond standing), and the merits were clearly against the parties asserting jurisdiction. The court suggested that its decision to avoid the jurisdictional question was justified by efficiency, noting that deciding only the narrow jurisdictional question presented by the parties would result in remand to consider still other jurisdictional issues as well as the merits.

In Adams v. Vance, 570 F.2d 950 (D.C. Cir. 1978), the court of appeals avoided constitutional justiciability questions, reasoning that

when the merits of a case are clearly against the party seeking to invoke the court’s jurisdiction, the jurisdictional question is especially difficult and far-reaching, and the inadequacies in the record or briefing make the case a poor vehicle for deciding the jurisdictional question, we may rule on the merits *157without reaching the inappropriately presented jurisdictional contention.

Id. at 954 n. 7 (citations omitted).

The court of appeals declined, in Switlik v. Hardwicke Co., 651 F.2d 852, 856 n. 3 (3d Cir.), cert. denied, 454 U.S. 1064, 102 S.Ct. 614, 70 L.Ed.2d 601 (1981), “[f]or reasons of policy and considerations of judicial economy,” to decide the “complex constitutional question” of jurisdiction; instead it assumed jurisdiction arguendo. The court added, “assuming without deciding that we have jurisdiction permits a proper and lawful resolution of the dispute without facing the dangers lurking in the murky waters surrounding the [jurisdictional] question.” Id.

The Court of Appeals for the Fifth Circuit has also avoided difficult jurisdictional questions when the merits are clearly against the party asserting jurisdiction. In Mitchell v. West Feliciana Parish School Board, 507 F.2d 662, 666-67 (5th Cir. 1975), the court reviewed a denial of appellants’ civil rights claim. In refusing to decide the question of whether the defendant was a “person” under section 1983 of title 42, the court declined to “attempt such a difficult issue, or one having potential for such far-reaching consequences, without being fully advised.” Id. at 667. Noting the pointlessness of a remand to consider the jurisdictional issue in the face of the result dictated by the merits, the court simply decided the merits.

The Court of Appeals for the Ninth Circuit has also found occasion to avoid jurisdictional issues and reach the merits. See, e.g., Forster v. County of Santa Barbara, 896 F.2d 1146 (9th Cir. 1990) (ignoring jurisdictional question because of factual dispute, unresolved at district court level, over whether appellant filed timely notice of appeal); Wolder v. United States, 807 F.2d 1506, 1507 (9th Cir. 1987) (“where the jurisdictional question is complex and the appeal is clearly without merit,” court will avoid jurisdictional question and rule on merits); Lehner v. United States, 685 F.2d 1187, 1189-90 (9th Cir. 1982) (court avoids question of whether jurisdiction exists over claims for money damages because jurisdiction over equitable claims was clear and merits would not be affected), cert. denied, 460 U.S. 1039, 103 S.Ct. 1431, 75 L.Ed.2d 790 (1983).

Some of these cases do not make it clear whether the holding is properly characterized as jurisdictional or on the merits. In Norton v. Mathews, for example, the court expressly noted that it was not deciding the issue of jurisdiction; at the same time, however, the court expressed the opinion that the case was “no longer substantial in a jurisdictional sense.” 427 U.S. 524, 530-31, 96 S.Ct. 2771, 2775, 49 L.Ed.2d 672 (1976). This language is reminiscent of a long line of Supreme Court cases dismissing claims as meritless, frivolous, or for want of a “substantial federal question.” See, e.g., Hagans v. Lavine, 415 U.S. 528, 536-37, 94 S.Ct. 1372, 1378-79, 39 L.Ed.2d 577 (1974); Bell v. Hood, 327 U.S. 678, 66 S.Ct. 773, 90 L.Ed. 939 (1946); The Fair v. Kohler Die and Specialty Co., 228 U.S. 22, 33 S.Ct. 410, 57 L.Ed. 716 (1913); Deming v. Carlisle Packing Co., 226 U.S. 102, 105, 33 S.Ct. 80, 81, 57 L.Ed. 140 (1912); Equitable Life Assurance Society v. Brown, 187 U.S. 308, 311, 23 S.Ct. 123, 124, 47 L.Ed. 190 (1902).

In The Fair v. Kohler Die and Specialty Co., 228 U.S. 22, 25, 33 S.Ct. 410, 411-12, 57 L.Ed. 716 (1913), Justice Holmes wrote, “No doubt if it should appear that ... the claim of right were frivolous, the case might be dismissed.... [T]he jurisdiction would not be denied, except possibly in form.”

The most frequently cited of these cases is probably Bell v. Hood, 327 U.S. 678, 66 S.Ct. 773, 90 L.Ed. 939 (1946). Here the Court stated, “a suit may sometimes be dismissed for want of jurisdiction where the alleged claim under the Constitution or federal statutes clearly appears to be ... wholly insubstantial and frivolous. The accuracy of calling these dismissals jurisdictional has been questioned.” Id. at 682-83, 66 S.Ct. at 776. See also Fogel v. Chestnutt, 668 F.2d 100, 106 (2d Cir. 1981) (citing The Fair and Bell v. Hood); Harper v. McDonald, 679 F.2d 955 (D.C.Cir. 1982) (citing Bell). Cf. Block v. Community Nu*158trition Institute, 467 U.S. 340, 353 n. 4, 104 S.Ct. 2450, 2457-58 n. 4, 81 L.Ed.2d 270 (1984) (“congressional preclusion of judicial review is in effect jurisdictional”) (emphasis added).

The rationales for avoiding difficult jurisdictional issues to reach the merits are those of judicial efficiency and restraint. These rationales suggest an analogy to eases in which the court avoids deciding prudential issues such as standing in favor of deciding the case on the merits. In Franchise Tax Board of California v. Alcan Aluminum, Ltd., — U.S. -, -, 110 S.Ct. 661, 665-67, 107 L.Ed.2d 696 (1990), for example, the court noted a circuit split as to the prudential (as opposed to article III) standing of respondents, but assumed standing without deciding the issue, and rejected the claim on the merits. See also Moe v. Confederated Salish & Kootenai Tribes, Inc., 425 U.S. 463, 479-80, 96 S.Ct. 1634, 1644-45, 48 L.Ed.2d 96 (1976) (assuming standing to decide the merits).

The judicial restraint rationales also suggest an analogy to the doctrine that constitutional claims should not be decided unnecessarily. See Neese v. Southern Railway, 350 U.S. 77, 76 S.Ct. 131, 100 L.Ed. 60 (1955) (avoiding jurisdictional questions in order to avoid constitutional questions); Ashwander v. Tennessee Valley Authority, 297 U.S. 288, 345, 56 S.Ct. 466, 482, 80 L.Ed. 688 (1936) (Brandéis, J., concurring) (“ ‘the power to declare a legislative enactment void is one which the judge, conscious of the fallibility of the human judgment, will shrink from exercising in any case where he can conscientiously and with due regard to duty and official oath decline the responsibility’ ”) (quoting 1 T. Cooley, Constitutional Limitations 332 (8th ed.)); Poe v. Ullman, 367 U.S. 497, 503, 81 S.Ct. 1752, 1755, 6 L.Ed.2d 989 (1961) (explaining Ash-wander as derived from the “historically defined, limited nature and function of courts,” and the theory that the adjudicatory process works best under circumstances making “resolution of the controverted issue a practical necessity”); Siler v. Louisville Railroad, 213 U.S. 175, 29 S.Ct. 451, 53 L.Ed. 753 (1909); Comment, Assuming Jurisdiction Arguendo: The Rationale and Limits of Hypothetical Jurisdiction, 127 U.Pa.L.Rev. 712, 753 (“Hypothetical jurisdiction stems from the general policy of avoiding constitutional adjudication unless necessary to the disposition of a case. It conserves increasingly limited judicial resources and fosters thoughtful and principled handling of important jurisdictional questions.”). Thus, a court will decide a case solely on state law claims if doing so would avoid the decision of a constitutional claim. Hagans v. Lavine, 415 U.S. 528, 546-47, 94 S.Ct. 1372, 1383-84, 39 L.Ed.2d 577 (1974); Siler v. Louisville Railroad, 213 U.S. 175, 29 S.Ct. 451, 53 L.Ed. 753 (1909). Construction of statutes in a way that avoids a constitutional question is another related technique for avoiding difficult questions whose resolution is unnecessary and would have broad impact. See Crowell v. Benson, 285 U.S. 22, 62, 52 S.Ct. 285, 296-97, 76 L.Ed. 598 (1932).

One of the controlling rationales of judicial restraint in these cases is the desire to minimize the intrusion of the federal judiciary into the business of the states and branches of the federal government. See Poe v. Ullman, 367 U.S. 497, 503, 81 S.Ct. 1752, 1755-56, 6 L.Ed.2d 989 (1961); Rescue Army v. Municipal Court of Los Angeles, 331 U.S. 549, 570-71, 67 S.Ct. 1409, 1420-21, 91 L.Ed. 1666 (1947).

Finally, an analogy is also suggested to the common-law tradition of requiring hard edged, sharply defined real controversies as the basis for shaping developing law. See, e.g., A. Bickel, The Least Dangerous Branch: The Supreme Court at the Bar of Politics 115 (1962) (“[Tjhere are sound reasons, grounded not only in theory but in the judicial experience of centuries, here and elsewhere, for believing that the hard, confining and yet enlarging context of a real controversy leads to sounder and more enduring judgments.”); K. Llewellyn, The Common Law Tradition: Deciding Appeals 109 (1960) (“almost unnoticed changes” of discrete cases may be more significant than large changes in “crisis-character” cases); Brilmayer, Judicial Review, Justiciability and the Limits of the Common Law Method, 57 B.U.L.Rev. 807, *159821 (1977) (“Justiciability limitations under the common law case method serve generally to prevent the use of the judicial process for the sole purpose of articulating principles of law rather than for settling the rights of injured parties.”); Vining, Judicial Review and the Doctrine of Ripeness in Administrative Law, 69 Mich.L. Rev. 1443,1449-50 (1971) (at common law a case “limited and organized thought.... There was no place for the determination of legal rules as such.... In the interest of individual freedom as well as judicial economy, courts refused to go beyond what the ‘case’ required”).

In drawing these analogies, we do not mean to suggest that difficult jurisdictional issues should be avoided whenever possible. A court is obliged to consider challenges to subject matter jurisdiction sua sponte at any stage of the proceedings. E.g., Mansfield, Coldwater & Lake Michigan Railway v. Swan, 111 U.S. 379, 381-82, 4 S.Ct. 510, 510-11, 28 L.Ed. 462 (1884); Fed.R.Civ.P. 12(h)(3) (requiring dismissal whenever it appears that jurisdiction is lacking). See generally P. Bator, D. Melt-zer, P. Mishkin & D. Shapiro, Hart and Wechsler’s The Federal Courts and the Federal System 1703 (3d ed. 1988) (discussing duty of court to ensure that jurisdiction exists). In applying the rule of judicial restraint and conservation of energy to avoid a difficult jurisdictional issue, the following principles need to be observed in order to avoid restraint becoming lethargy and efficiency mere avoidance.

First, we reiterate that the normal rule is to decide jurisdiction before the merits. See generally, P. Bator, D. Meltzer, P. Mishkin & D. Shapiro, Hart and Wech-sler’s The Federal Courts and the Federal System 1703 (3d ed. 1988).

Second, the assumption of jurisdiction should not do injustice to the parties by determining the outcome of the case. Usually this will mean that the merits be against the party invoking jurisdiction. In rare cases where there is no effect on the result, such as Philbrook v. Glodgett, 421 U.S. 707, 721-22, 95 S.Ct. 1893, 1902-03, 44 L.Ed.2d 525 (1975), discussed supra, the merits can be against the party challenging jurisdiction.

Third, the merits should be clearly in favor of one party. If the case is close on the merits, the court may have to expend as much energy deciding the merits as it would deciding the jurisdictional issues.

Fourth, the jurisdictional issues should be difficult and far-reaching. This principle implicates the rationales of efficiency and restraint that are the heart of this approach to deciding cases. Without such a limitation the concept of jurisdiction — the power to decide — would be too greatly eroded. Jurisdictional questions may be considered difficult if they are of first impression and involve the interpretation of a complex statutory scheme, or if they are of constitutional stature. Conversely, a case that does not present a claim of even color-able jurisdiction is not appropriate for assuming jurisdiction arguendo. Jurisdictional questions may be considered far-reaching if they affect large numbers of potential litigants, implicate separation of powers or federalism concerns, or concern the activities of administrative agencies.

Fifth, inadequacies in the record or briefs may be a factor persuading a court to assume jurisdiction without deciding. This is not a necessary factor, since both parties may deeply desire to obtain a decision on the jurisdictional issue and their briefs may be exhaustive, yet the insub-stantiality of the claim on the merits may make the jurisdictional issue unsuitable for decision at the moment.

Under the principles just outlined, the case before us is an appropriate one in which to assume jurisdiction arguendo without deciding the issue. First, colorable federal question jurisdiction exists because the claims on their face involve the interpretation solely of federal law. Cf. Mount Healthy City School District Board of Education v. Doyle, 429 U.S. 274, 278-79, 97 S.Ct. 568, 571-72, 50 L.Ed.2d 471 (1977); Bell v. Hood, 327 U.S. 678, 682, 66 S.Ct. 773, 776, 90 L.Ed. 939 (1946). The jurisdictional issues are of first impression. They involve the interpretation of a complex and delicate web of environmental legislation *160consisting of several overlapping statutes promulgated by different Congresses under different Presidential administrations. Their resolution would necessarily have a broad impact on an important legislative scheme involving the expenditure of billions of dollars of tax and private money, and in particular on the actions of an important executive agency. A decision in BFI’s favor might have serious implications for future EPA remediation efforts and could hamper protection of the environment in a way that runs counter to congressional design. Other CERCLA orders that might be one component of an EPA cleanup project that relies upon several statutes could become subject to lengthy and costly legal challenges. Conversely, a decision in EPA’s favor could turn a powerful weapon in the fight against pollution into a bludgeon to be used against alleged polluters regardless of prior agreements or understandings.

The record was not appropriate for the broad rulings sought by the parties. Finally, the outcome of this case on the merits is clear, and would be consistent with either a finding or denial of jurisdiction. On the spectrum of cases suitable for finding hypothetical jurisdiction, the case before us is an “easy” one. See Comment, Assuming Jurisdiction Arguendo: The Rationale and Limits of Hypothetical Jurisdiction, 127 U.Pa.L.Rev. 712, 733 (1979).

V.

THE MERITS

BFI claims that EPA is bound by the terms of the RCRA Order and cannot unilaterally abrogate it by using a contrary CERCLA order. The subject of the CERC-LA Order, the composition of the monitoring wells, was not explicitly or implicitly covered in the parties’ consensual RCRA Order. Although it is self-evident that a groundwater monitoring system requires monitoring wells of some kind, and the parties do not dispute this, there is no mention of the monitoring wells at all in the RCRA Order. EPA’s subsequent insistence on stainless steel can be seen as a reasonable administrative expert interpretation of the RCRA agreement between BFI and EPA; it is not an abrogation of that agreement.

There is no conflict between the EPA’s requirement of stainless steel and the RCRA Order. EPA’s insistence is consistent with a fair administrative interpretation of the consensual RCRA Order. The fact that EPA, in the face of BFI’s disagreement, finally translated its request for stainless steel into what purported to be a CERCLA order is not dispositive.

On the merits this case presents a narrow issue: can the EPA exercise its discretion pursuant to its expertise in environmental matters to require the installation of stainless steel rather than PVC wells to monitor ground water pollution, when the parties have voluntarily agreed that wells of some kind are required but have not specified the type of material? Resolution of this issue does not require that we review what purports to be a CERCLA Order, and we do not do so.

Courts should be particularly reluctant to second-guess agency choices involving scientific disputes that are in the agency’s province of expertise. Deference is desirable. Baltimore Gas & Electric Co. v. NRDC, 462 U.S. 87, 103, 103 S.Ct. 2246, 2255, 76 L.Ed.2d 437 (1983) (“reviewing court must generally be at its most deferential” when examining “scientific determination[s]” by administrative agency “within its area of special expertise”); Federal Power Commission v. Florida Power & Light Co., 404 U.S. 453, 463, 92 S.Ct. 637, 643-44, 30 L.Ed.2d 600 (1972) (“Particularly when we consider a purely factual question within the area of competence of an administrative agency created by Congress, and when resolution of that question depends on ‘engineering and scientific’ considerations, we recognize the relevant agency’s technical expertise and experience, and defer to its analysis unless it is without substantial basis in fact.”); City of New York v. United States Department of Transportation, 715 F.2d 732, 745 (2d Cir. 1983) (citing Baltimore Gas as defining the “standards circumscribing our role”).

*161According to EPA, the choice of stainless steel over PVC was based on its conclusion that PVC may leach or sorb contaminants, thus resulting in inaccurate readings. BFI claimed that PVC was adequate, and cheaper than stainless steel. In addition, BFI claimed that stainless steel would corrode and itself contaminate the samples, as it had allegedly done at another dump site administered by BFI.

An appellate court may take judicial notice of the existence of a body of scientific literature. See Fed.R.Evid. 201(b) (judicial notice may be taken from “sources whose accuracy cannot reasonably be questioned”); Fed.R.Evid. 201(f) (judicial notice may be taken “at any stage of the proceeding”); Fed.R.Evid. 803(18) (admission of “periodicals ... on a subject of ... science” after “judicial notice” of reliability). Before an appellate court takes judicial notice it should advise the parties so they can object or furnish helpful documents. See Fed.R.Evid. 201(e) (opportunity to be heard). In the instant case the court at the time of argument requested the parties to submit published literature on the issue of proper well-casing materials. They have graciously complied, providing an extensive collection that is now part of the record.

It is the existence of the extensive scientific literature on well-casing material rather than the dispositive nature of any of the articles that establishes that there is a sufficient basis for the decision of the EPA. See R. Gillham, S. O’Hannesin & J. Barker, Sorption/Desorption of Soluble Constituents of Petroleum Products by Materials Used in Construction of Monitoring Wells (American Petroleum Institute Nov. 1988) (experiment showed that PVC sorbed less of six soluble constituents of petroleum than did other polymer casing materials, but more than stainless steel); M. Barcelona, G. George & M. Schock, Comparison of Water Samples from PTFE, PVC, and SS Monitoring Wells 8 (Feb. 1988) (experiment showed that PVC sorbed significant amounts of three of five organic compounds tested; stainless steel “should provide significant advantages over the use of PVC in ground-water monitoring applications”) (EPA contract report); United States Environmental Protection Agency, Standard Operating Procedure for Selecting Ground Water Well Construction Material 7-9 (Dec. 18, 1986) (“PVC will leach when in contact with-organics” and will “react with [various] aqueous organic mixtures”; recommended for all conditions “except organic media”); United States Environmental Protection Agency, RCRA Ground-Water Monitoring Technical Enforcement Guidance Document 78 (Sept. 1986) (stainless steel should be used when volatile organics may be tested during a 30-year monitoring period; PVC acceptable if only trace metals or nonvolatile organics are to be tested); Memorandum from D. Pérsico, E.I. DuPont de Nemours & Co., Polymer Products Dep’t, Re Defects in Poly(Vinyl Chloride) (PVC), Jan. 22, 1986 (describing PVC as highly unstable in presence of certain organics); G. Reynolds & R. Gillham, Absorption of Halogenated Organic Compounds by Polymer Materials Commonly Used in Groundwater Monitors, presented at The Second Annual Canadian/American Conference on Hydrogeology: Hazardous Wastes in Ground Water — A Soluble Dilemma, June 25-29, 1985 (experiment showed that four of five tested halogenated organic compounds were absorbed by PVC, but rate of absorption was slow enough that bias may not be significant “where the groundwater recovers rapidly and is sampled within the same day as well development”); Houghton & Berger, Effects of Well-Casing Composition and Sampling Method on Apparent Quality of Ground Water, in Proceedings of the Fourth National Symposium on Aquifer Restoration and Ground Water Monitoring, May, 1984, at 203, 206 (experiment showed that samples from PVC-cased wells (using adhesive) were enriched in organic carbon compared to stainless steel wells (presumably indicating leaching)); M. Barcelona, J. Gibb & R. Miller, A Guide to the Selection of Materials for Monitoring Well Construction and Ground-Water Sampling 43 (Aug. 1983) (noting possibility of adsorption and absorption of organic compounds, and recommending that in high-organic or *162unknown conditions, if PVC wells are used, they should be paired with some non-polymeric wells to determine potential bias in the PVC wells); Cook & Hartz, Adsorption of Chlorinated Methanes from Aqueous Solution on Selected Plastic Adsorbents, 1983 Research & Technology 423, 424 (experiment showed that PVC tubing adsorbed significant amounts of chlorometh-ane compounds); Miller, Uptake and Release of Lead, Chromium and Trace Level Volatile Organics Exposed to Synthetic Well Casings, in Proceedings for the Second National Symposium on Aquifer Restoration and Groundwater Rehabilitation, May, 1982, at 236 (experiment showed that schedule 40 PVC 1120 adsorbed trichloroe-thylene and chromium and adsorbed and leached tetrachloroethylene and lead, but did not adsorb several other organic compounds); Sosebee, Geizler, Winegardner & Fisher, Contamination of Groundwater Samples with Poly(Vinyl Chloride) Adhesives and Poly(Vinyl Chloride) Primer from Monitor Wells, in Hazardous and Industrial Solid Waste Testing: Second Symposium, January 28-29, 1982, ASTM PCN 04-805000-16, at 38, 39 (experiment showed that compounds in PVC adhesive and primer may mask the existence of groundwater contaminants); E. Boettner, G. Ball, Z. Hollingsworth, R. Aquino, Organic and Organotin Compounds Leached from PVC and CPVC Pipe 3 (Dep’t of Environ. & Industrial Health, School of Public Health, Univ. of Mich., Sept. 1981) (experiment showed that órga-no tins contained in heat stabilizers of PVC and organics in pipe sealing cement solvents can leach into water) (EPA contract report).

This literature indicates that there was ample justification for EPA’s choice. While not conclusive on the issue of the unsuitability of PVC for this particular dump site, the literature describes frequent occurrences of leaching, absorption and adsorption, all of which could skew the monitoring process. The Record of Decision indicated that the leachate at the site contained “low levels of volatile organics” that might, according to the literature, affect monitoring. EPA’s decision is sufficiently supported by the available scientific literature.

It cannot be said that EPA did not consider BFI’s position. BFI presented its preference for PVC to EPA in oral discussions and correspondence. In addition, BFI supplied EPA with scientific literature supporting PVC and critical of stainless steel, which supplemented EPA’s own literature comparing the suitability of both materials. See R. Gillham, S. O’Hannesin & J. Barker, Sorption/Desorption of Soluble Constituents of Petroleum Products by Materials Used in Construction of Monitoring Wells (American Petroleum Institute, Nov. 1988) (experiment showed that PVC sorbed less of six soluble constituents of petroleum than did other polymer casing materials, but more than stainless steel; report recommends that EPA establish performance criteria for monitoring wells, rather than adopting a general preference for one material over another.); United States Environmental Protection Agency, Standard Operating Procedure for Selecting Ground Water Well Construction Material 7-9 (Dec. 18, 1986) (stainless steel leaches chromium or nickel “after long exposure to very corrosive (pH < 4.5) conditions” and SS-304 stainless steel is corroded by hydrochloric and nitric acids); United States Environmental Protection Agency, RCRA Ground-Water Monitoring Technical Enforcement Guidance Document 78 (Sept. 1986) (fluorocarbon resins recommended over stainless steel “where high corrosion potential exists”); A. Sykes, R. McAllister & J. Homolya, Sorption of Organics by Monitoring Well Construction Materials, Ground Water Monitoring Rev., Fall 1986, at 44 (experiment showed no significant statistical difference in sorption of six organic compounds between stainless steel and PVC); H. Breard & M. Brady, Presentation to Groundwater Monitoring Guidance Review Subcommittee of the Environmental Engineering Committee, EPA Science Advisory Board, by CE-COS/BFI, Oct. 3, 1985, at 6-9 (summarizing scientific literature and claiming that adsorption problems of PVC can be solved by purging and sampling the well on the *163same day, and that stainless steel’s protective film cannot form in a dilute hydrochloric acid environment; recommends EPA listing of stainless steel and PVC as standard materials); G. Reynolds & R. Gill-ham, Absorption of Halogenated Organic Compounds by Polymer Materials Commonly Used in Groundwater Monitors, presented at The Second Annual Canadian/American Conference on Hydrogeology: Hazardous Wastes in Ground Water — A Soluble Dilemma, June 25-29, 1985 (experiment showed that four of five tested halogenated organic compounds were absorbed by rigid PVC, but rate of absorption was slow enough that bias may not be significant “where the groundwater recovers rapidly and is sampled within the same day as well development”); D. Pope, D. Duquette. P. Wayner & A. Johannes, Microbiologically Influenced Corrosion: A State-of-the-Art Review 39-44 (MTI Publication No. 13, June, 1984) (survey of literature noting evidence of microbiologieally induced corrosion of stainless steel); Houghton & Berger, Effects of Well-Casing Composition and Sampling Method on Apparent Quality of Ground Water, in Proceedings of the 4th National Symposium on Aquifer Restoration and Ground Water Monitoring May, 1984, at 203, 206 (samples from stainless steel wells were enriched in various metals compared to PVC wells, to the extent that stainless steel is inappropriate for sampling trace metals at low concentrations); M. Barcelona, J. Gibb & R. Miller, A Guide to the Selection of Materials for Monitoring Well Construction and Ground-Water Sampling 39 (Aug. 1983) (stainless steel may cause chromium or nickel contamination in very corrosive conditions; PVC “outperforms” stainless steel in high ionic strength acid environments) (EPA contract report); Curran & Tomson, Leaching of Trace Organics into Water from Five Common Plastics, Ground Water Monitoring Review, Summer 1983, at 68, 70 (experiment showed that PVC resulted in little leaching contamination, and was acceptable for monitoring wells if rinsed with room temperature water before installation); Unibell Plastic Pipe Association, Vinyl Chloride; The Control of Residual Vinyl Chloride Monomer in PVC Water Pipe (1981) (survey of research claiming that vinyl chloride monomer, the main constituent of PVC, does not migrate into water); United States Environmental Protection Agency, Office of Water & Waste Management, Procedures Manual for Ground Water Monitoring at Solid Waste Disposal Facilities 146 (1977) (PVC well-casing “is very attractive ... it will contribute little in the way of chemical constituents to a leachate sample except in the parts-per-billion range”; “steel pipe” (unclear if stainless or not) leaches iron and other ions, but contamination can probably be avoided by proper flushing before collection); G. Kobrin, Corrosion by Microbiological Organisms in Natural Waters, Materials Performance, July 1976, at 38, 40-42 (describing one case history of severe corrosion of stainless steel believed to have been caused by microbes).

BFI had a bona fide concern that acidic groundwater at the site might corrode stainless steel, but having considered the matter, EPA was satisfied that the groundwater would not be corrosive. The published literature available to EPA considered the merits of both stainless steel and PVC. It cannot be said that EPA failed to consider BFI’s concerns in interpreting the RCRA Order. The public documentation available to BFI and EPA, and the extensive discussions between EPA and BFI reflected in the record sufficiently supported EPA’s determination. See Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 408-10, 91 S.Ct. 814, 819-21, 28 L.Ed.2d 136 (1971).

BFI suggests that EPA’s decision was one that did not take into account the specific characteristics of the dump site. Even if this were true (and the Record of Decision suggests that it is not), the benefits that EPA might in its administrative expertise find derived from standardizing well-casing materials cannot be gainsaid by this court.

There are 848 landfills and hazardous waste sites on the Superfund National Priorities List as of March, 1989, 40 C.F.R. Part 300, App. B (1989), and countless more *164that are subject to some kind of EPA mandated remediation. Congress has delegated authority to EPA to deal with the problems presented by each of these sites. The judicial system lacks the resources and the expertise to address every scientific dispute that might arise from EPA’s actions. The monitoring process chosen by EPA in this instance is within the normal discretion of an administrative agency interpreting a voluntary agreement — the RCRA Order. While BFI’s position is not without scientific merit, deference to the agency and the need to keep out of the courts such detailed scientific disputes as the proper material for monitoring wells demands a finding on the merits for the EPA in this case.

BFI contends that the administrative record on which EPA based its decision is inadequate. In the special circumstances of this case, we are satisfied that EPA’s basis for reaching its decision was sufficiently a matter of record adequately made known to BFI. EPA had informed BFI of EPA Region II’s “Standard Operating Procedure for Selecting Groundwater Well Construction Material at CERCLA Sites,” its “Ground-Water Monitoring Technical Enforcement Guidance Document,” and its analysis of the chemical composition of the groundwater at the South Brunswick site.

Regardless of the standard of review, EPA’s decision to require stainless steel as the material for the monitoring well casing was well within its expert discretion. For a discussion of possible standards, see Administrative Procedure Act, 5 U.S.C. § 706 (“arbitrary and capricious” standard, urged by BFI); Motor Vehicle Manufacturer’s Association v. State Farm Mutual Automobile Insurance Co., 463 U.S. 29, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983) (same); Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 413-15, 91 S.Ct. 814, 822-23, 28 L.Ed.2d 136 (1971) (same); Association of Data Processing Service Organizations, Inc. v. Board of Governors of the Federal Reserve System, 745 F.2d 677, 683-84 (D.C.Cir. 1984) (conflating “arbitrary and capricious” and “substantial evidence” standards of review when applied to factual judgments). Compare U.S. v. Hardage, 663 F.Supp. 1280 (W.D.Okl. 1987) (de novo review appropriate in action for injunctive relief to compel compliance with EPA remedy pursuant to RCRA), with R. Pierce, S. Schapiro & P. Verkuil, Administrative Process § 7.3, at 357 (1985) (de novo review only appropriate when mandated by statute). Under any arguable standard, EPA’s decision withstands scrutiny.

CONCLUSION

The district court was correct in holding that BFI had not established equitable or legal grounds for relief.

The judgment of the district court is affirmed.