This case embodies an argument about whether a cargo port and a causeway that Maine plans to build at Sears Island will "significantly affect[ ] the environment." 42 U.S.C. Sec. 4332(2)(C). Maine's Department of Transportation, the Federal Highway Administration, and the Army Corps of Engineers concluded that it would not; thus, they decided to fund, to grant permits for, and to proceed with, the project without preparing an Environmental Impact Statement ("EIS"). The Sierra Club sued the federal agencies, seeking to stop the project in the absence of an EIS. Sierra Club now appeals the federal district court's determination that the agencies' decision not to prepare an EIS was lawful. Having reviewed the administrative record, we disagree with the district court. In our view, the record reveals that the project will significantly affect the environment; and the agencies' contrary conclusion lies outside the legally permissible bounds laid down by relevant statutes. 42 U.S.C. Sec. 4332(2)(C); 5 U.S.C. Sec. 706(2)(A). Hence, the agencies must prepare an EIS.

* The legal framework in which this case arises is fairly simple. The National Environmental Policy Act of 1969 ("NEPA") says that federal agencies must prepare "a detailed statement ... on the environmental impact" of any proposed major federal action "significantly affecting the environment." 42 U.S.C. Sec. 4332(2)(C)(i). The federal Council on Environmental Quality ("CEQ") has issued detailed regulations and explanations of NEPA's statutory terms which tell federal agencies how to decide when an EIS is needed and how to go about preparing one. See 40 C.F.R. Sec. 1500 et seq. (1984); Forty Most Asked Questions Concerning CEQ's National Environmental Policy Act Regulations, 46 Fed.Reg. 18026 (1981). The CEQ regulations permit federal agencies to make a preliminary "Environmental Assessment" ("EA") aimed at determining whether the environmental effects of a proposed action are "significant." 40 C.F.R. Secs. 1501.3, 1501.4, 1508.9, 1508.27 (1984). According to these regulations, the EA is a "concise" document that "briefly" discusses the relevant issues and either reaches a conclusion that preparation of an EIS is necessary or concludes with a "Finding of No Significant Impact" (called, in environmental jargon, a "FONSI"). Id. Secs. 1508.9, 1508.13.

In this case, the Corps of Engineers and the Federal Highway Administration prepared or adopted EA's and concluded with "FONSI's." They therefore granted necessary permits and funding for the Sears Island project without an EIS. The Sierra Club claims that these findings of "no significant impact" were "arbitrary, capricious, an abuse of discretion," and therefore unlawful under the Administrative Procedure Act. 5 U.S.C. Sec. 706(2)(A). The district court, after reviewing the fairly lengthy and complex administrative record, decided that the agencies' conclusions were sufficiently reasonable to be lawful. Our legal duty, strictly speaking, is to review this district court conclusion.

The purely legal issue that we face at the outset is what standard to apply when reviewing the district court. In part the issue involves the question of what standard courts normally should apply to the agency's decision. We have previously said that one challenging a decision not to prepare an EIS must show a "substantial possibility that agency action 'could significantly affect the quality of the human environment.' " Quinonez-Lopez v. Coco Lagoon Development Corp., 733 F.2d 1, 2 (1st Cir.1984) (quoting Winnebago Tribe v. Ray, 621 F.2d 269, 271 (8th Cir.), cert. denied, 449 U.S. 836, 101 S. Ct. 110, 66 L. Ed. 2d 43 (1980) ). If the record reveals such a "substantial possibility" with sufficient clarity, the agency's decision (not to produce an EIS) violates NEPA. Depending upon the agency's reasons, a reviewing court might say that such an agency made a mistake interpreting NEPA or the CEQ regulations, or the court might say that the agency's "no significant impact" finding was simply "arbitrary, capricious, an abuse of discretion," 5 U.S.C. Sec. 706(2)(A). Often these questions cannot be easily separated one from the other. But whatever verbal formulation it applies, the court in a case like this must essentially look to see if the agency decision, in the context of the record, is too 'unreasonable' (given its statutory and factual context) for the law to permit it to stand. Quinonez-Lopez, 733 F.2d at 2-3 (citing National Wildlife Federation v. Appalachian Regional Commission, 677 F.2d 883, 889 (D.C.Cir.1981); Committee for Auto Responsibility v. Solomon, 603 F.2d 992, 1002 (D.C.Cir.1979), cert. denied, 445 U.S. 915, 100 S. Ct. 1274, 63 L. Ed. 2d 599 (1980) ); see generally Gee v. Boyd, --- U.S. ----, 105 S. Ct. 2123, 85 L. Ed. 2d 487 (1985) (White, J., dissenting from denial of certiorari); Shea, The Judicial Standard for Review of Environmental Impact Statement Threshhold Decisions, 9 B.C.Envtl.Aff.L.Rev. 63 (1980).

This case poses the additional (and somewhat difficult) problem of determining what standard we, an appellate court, should apply to the district court's decision upholding the agency. The reason this matter is a little difficult is that administrative review as typically practiced by the federal courts differs from the expectation of those who framed the Administrative Procedure Act in an important respect. Initially, the Act's authors apparently believed that district courts would review the reasonableness of agency decisionmaking on the basis of a record created in the district court. See generally Nathanson, Probing the Mind of the Administrator: Hearing Variations and Standards of Judicial Review Under the Administrative Procedure Act and Other Federal Statutes, 75 Colum.L.Rev. 721 (1975). Under this arrangement, the courts of appeals would play a limited role, reviewing, for example, a district court's factual findings for "clear error." Because of the way in which the findings and characterization of facts affect one's judgment about the 'reasonableness' or 'arbitrariness' of agency behavior, appellate review of a finding about 'arbitrariness' might also have been limited. Evolving practice, however, under special review statutes has led to courts of appeals normally reviewing the legality of agency decisionmaking on the basis of a record created in the agency. See, e.g., 47 U.S.C. Sec. 402 (providing for court of appeals review of FCC decisions); 29 U.S.C. Sec. 160(l ) (same, NLRB); 15 U.S.C. Sec. 77i (same, SEC). Courts of appeals thus apply, say, the APA's 'arbitrary and capricious' standard to the agency record directly, without the intervening set of findings of a district court, which in turn has reviewed the agency's findings.

Under these circumstances, a court of appeals review of a district court review of an administrative agency's record is an awkward legal animal. Are we to set aside such a district court decision only if it is "clearly erroneous"? Fed.R.Civ.P. 52(a). Or, are we to ignore the district court and simply conduct our own review of the administrative record? Other cases, containing what we have described as a "luxuriant jungle of differing review descriptions," Quinonez-Lopez v. Coco Lagoon Development Corp., 733 F.2d at 3, do not settle the matter. Compare Lange v. Brinegar, 625 F.2d 812, 815 (9th Cir.1980) ("In determining whether the evidence was sufficient to establish [the need for an EIS] ..., we are bound by the findings of the trial court unless they are clearly erroneous."), with Save Our Wetlands, Inc. v. Sands, 711 F.2d 634, 644-45 (5th Cir.1983) (relying on district court trial testimony and administrative record in determining that agency acted reasonably in failing to require EIS), and Concerned Citizens on I-190 v. Secretary of Transportation, 641 F.2d 1, 3 (1st Cir.1981) (court of appeals should review both agency's conclusions and those of the district court).

We believe our answer to this type of question should be practical. We should be more willing, or be less willing, to differ with a district court about the 'reasonableness' or 'arbitrariness' of an agency decision, depending upon the particular features of the particular case that seem to make a more independent, or a less independent, appellate court scrutiny of the administrative record appropriate. Where, for example, the district court's judgment turns on matters of fact that it has determined, or upon evidence presented by witnesses in court, or even upon lengthy district court proceedings in which knowledgeable counsel explain the agency's decisionmaking process in detail, we will show appropriate hesitation to overturn that judgment. Cf. Fed.R.Civ.P. 52(a) ("due regard shall be given to the opportunity of the trial court to judge the credibility of the witnesses"); Anderson v. City of Besemer, --- U.S. ----, ---- - ----, 105 S. Ct. 1504, 1512, 84 L. Ed. 2d 518 (1985) ("clearly erroneous" standard applies to factfinding even when district court factfinding is based on "documentary evidence or inferences from other facts"). But, where the district court simply reviews a set of agency documents and, applying the same legal standard as we apply here, reaches a particular legal conclusion about the 'reasonableness' of an agency's action, we have greater legal freedom to differ with the district court's ultimate characterization of agency behavior.

We have before us a case in which we and the district court are to apply the same legal standard to the same agency record. The court made no independent findings of fact and heard no witnesses. The district court had somewhat greater opportunity than did we to explore the administrative record with the help of counsel; and it is more familiar than are we with the geographical area and circumstances out of which this suit arises. But these latter, offsetting factors are not very weighty where, as here, the written administrative record is so complete. On balance, we believe we should exercise a considerable degree of independence in reviewing the administrative record and deciding whether the issuance of FONSI's was legally proper.

* The record shows that Sears Island is an undeveloped, wooded 940-acre island in upper Penobscot Bay. (See Appendix for maps of the island.) The island is connected to the mainland by a gravel bar exposed only at low tide. The mainland area adjacent to the island has been developed for industrial use: a chemical plant and a petroleum storage area sit on either side of the point leading to the gravel bar. Indeed, Searsport (where the island is located) is one of the busiest ports in Maine. Those interested in developing Sears Island for industrial purposes have suggested using it as a site for a nuclear-powered electric generating plant, a coal-fired power plant, a petroleum refinery, and an aluminum reduction plant. These proposals have received varying degrees of consideration, but each has ultimately been withdrawn. The most recent proposal for Sears Island consists of three parts: (1) a 1,200-foot solid-fill causeway that would connect Sears Island to the mainland with a railroad line and a two-lane road; (2) a dry-cargo marine terminal designed principally for the shipping of lumber and agricultural products, containerized cargo, and, possibly at a later stage in the project, coal; and (3) an industrial park in an area adjacent to the cargo port. The precise nature of the industrial park is now uncertain, for the park's eventual shape depends on what businesses choose to locate there. The plans for the other two components, however, are definite, and the impact of those plans has been discussed in great detail by various federal agencies which have examined the proposal.

Maine voters and government agencies have shown considerable support for the Sears Island project. Maine voters have twice approved bond referenda to finance the state's share of costs for the causeway and port. State agencies have recommended reconciling concerns about the environment with those about economic growth by concentrating industrial development in Maine within several selected coastal areas; these agencies have designated the Searsport area (and two other municipal areas in Maine) "for the 'most constraining' heavy industrial development" (Maine DOT EA, vol. 3, at 4). The town of Searsport has prepared "A Municipal Response Plan for the Industrial Development of Sears Island," which predicts substantial economic benefits from the causeway/port/industrial park project but recommends that the town enact appropriate land use laws to control the level of industrial development. The town has recently enacted such laws and has zoned the island for industrial use. The town has also borrowed $400,000 for its share of the cost of constructing the causeway.

The record also reveals a lengthy, complex, and controversial history underlying the creation of the EA's (and FONSI's) here at issue. Federal review of the Sears Island project began in 1981 when the Maine DOT circulated a preliminary study for construction of a causeway from the mainland to Sears Island. At least one federal agency, the Fish and Wildlife Service, said the study was too narrow; it said that any environmental study should include the effects of the proposed port facility and related industrial development as well. Maine DOT, however, prepared an EA that focused only on the causeway. The Federal Highway Administration adopted this EA as its own, 42 U.S.C. Sec. 4332(2)(D), issued a FONSI, and approved federal funding for the causeway. At this point, at least four federal agencies objected--three 'environmental' agencies (the Fish and Wildlife Service, EPA, the National Marine Fisheries Service) and the Coast Guard. See 42 U.S.C. Sec. 4332(2)(C) ("Prior to making any detailed statement, the responsible Federal official shall consult with and obtain the comments of any Federal agency which has jurisdiction by law or special expertise with respect to any environmental impact involved.") (emphasis added). These four agencies said that the EA was inadequate and that all three parts of the proposed development (causeway, port, and industrial development) should be considered together in an EIS.

Maine DOT then prepared another EA, this time on the port facility alone; the Federal Highway Administration adopted this EA and then issued another FONSI. Again, the three federal environmental agencies objected, and this time the Commerce Department's Economic Development Administration joined them. Again, these federal 'commenting' agencies asked for an EIS; and the Economic Development Administration stated that it would not consider Maine's request for funding without a combined EIS on "the three major components of the project: causeway, port facility, and industrial park." Responding to this criticism, the Federal Highway Administration adopted a new document, prepared by Maine DOT, called an "Environmental Assessment Summary." The new document considered both causeway and port, but it expressly disclaimed any need for consideration of "development on Sears Island outside of the current marine terminal project." On December 16, 1983 the Federal Highway Administration issued yet another FONSI for the causeway/port project.

On the same day, the Army Corps of Engineers (the agency responsible for issuing permits for the project) released its own EA, in respect to port and causeway, and, on the basis of that EA, issued a FONSI. The three environmental agencies criticized the Corps' EA. The Corps responded with a further "Statement of Findings and Environmental Assessment," and, after renewed criticism, issued a "Supplement to Statement of Findings." The Corps, like the Federal Highway Administration, concluded with a FONSI. Then, without preparing an EIS, it issued a permit allowing causeway and port construction to begin. At that point, the Sierra Club filed suit.

As this history indicates, we do not have before us the "brief" or "concise" EA of which the CEQ's regulations speak. 40 C.F.R. Sec. 1508.9. Rather, we have an 'EA' consisting of at least seven documents containing 350 pages of text, plus numerous pages of diagrams, maps, and technical drawings. They include:

EA on causeway (prepared by Maine DOT for Federal Highway Administration), as revised (Sept. 1982)--30 pages

EA on port facility (prepared by Maine DOT for Federal Highway Administration) (Aug. 1983)--138 pages, plus 61 pages of appendices and 34 pages of technical drawings

Summary EA on both causeway and port facility (prepared by Maine DOT for Federal Highway Administration) (Nov. 1983)--17 pages plus 29 pages of responses to various agencies' objections

Maine DOT study (prepared at the request of the Corps) reevaluating the environmental effects of building the causeway (April 1984)--110 pages

Corps' statement of findings and EA (July 1984)--15 pages

Corps' supplement to statement of findings (Aug. 1984)--7 pages

The agencies claim these documents support one conclusion, namely that the Sears Island project will have no significant impact on the environment.

The very length and detail of these documents have posed something of a dilemma to the agencies, the parties, and the reviewing courts. On the one hand, one is tempted to argue that the very complexity of the documents shows that an EIS is needed. The EA's are far longer than the CEQ recommends, see Forty Most Asked Questions, supra, 46 Fed.Reg. at 18037 (calling on agencies "to limit EA's to not more than approximately 10-15 pages"), and significantly longer than the typical full-blown EIS, see 40 C.F.R. Sec. 1502.7 (1984) (EIS should "normally be less than 150 pages and for proposals of unusual scope or complexity ... less than 300 pages"). The CEQ has stated:

Agencies should avoid preparing lengthy EAs except in unusual cases, where a proposal is so complex that a concise document cannot meet the goals of [CEQ regulations] and where it is extremely difficult to determine whether the proposal could have significant environmental effects. In most cases, however, a lengthy EA indicates that an EIS is needed.

40 Fed.Reg. at 18037 (emphasis added). In addition, the EA's reflect considerable disagreement among federal agencies over the documents' findings. Compare Quinonez-Lopez v. Coco Lagoon Development Corp., 733 F.2d at 3 (affirming agency's decision not to prepare EIS where five governmental agencies agreed that impact on affected area was of no ecological significance), with Silva v. Lynn, 482 F.2d 1282, 1285-86 (1st Cir.1973) (finding EIS inadequate where proposal "drew heavy fire" from three federal agencies "with expertise ... equal to or greater than that of [the lead agency]"); see also Grazing Fields Farm v. Goldschmidt, 626 F.2d 1068, 1071 & n. 1 (1st Cir.1980); 40 C.F.R. Sec. 1508.27(b)(4) (in evaluating "significance" of project's impacts, agency should consider "the degree to which the effects [of the project] ... are likely to be highly controversial"). To announce that these documents--despite their length and complexity--demonstrate no need for an EIS is rather like the mathematics teacher who, after filling three blackboards with equations, announces to the class, "You see, it is obvious."

On the other hand, one tends to recognize, as the appellees point out, that the lengthy documents reflect a thorough consideration of potential impact on the environment. Since these documents, at least arguably, already amount to an EIS in all but name, what is the practical point of requiring additional preparation of another document? This argument is well summarized in a Federal Highway Administration memorandum contained in the record, which says:

The critical issue is whether or not an EIS should be prepared. Several Federal agencies including, EPA, U.S. Fish & Wildlife, and the National Marine Fisheries feel that the project requires an EIS.... Thus, the Maine DOT and the [Federal Highway Administration] are faced with a decision. From a practical standpoint, there is nothing to gain from going with an EIS since there has been excellent coordination and exposure/consideration of environmental, social and economic issues. From a purely legal perspective, an EIS may be required per the CEQ regulations Section 1508.27, paragraphs (b)(1), (b)(6), and (b)(7), attached. Mr. [William] Richardson [the head of the Maine Division of Federal Highway Administration] is well aware of the critical issues of this project as well as CEQ requirements. Deep down, Bill feels that the public may be done a great injustice if the decision for an EIS is made. The processing would take an absolute minimum of 6 months if not more reasonably 9-12 months without any foreseeable benefits.

(A. 373-74 (emphasis added).)

After considering both of these arguments, we conclude that we should accept neither. We should not give conclusive weight, one way or the other, to the simple facts of EA length, complexity, and controversy. These facts do not by themselves show that the EAs' conclusion--"no significant impact"--is correct, nor do they show it is incorrect. At most they show the practical wisdom of CEQ's advice: the agencies would have saved time in the long run had they devoted their considerable effort to the production of an EIS, instead of the production of documents seeking to prove that an EIS is not needed. See Maryland-National Capital Park and Planning Commission v. U.S. Postal Service, 487 F.2d 1029, 1040 (D.C.Cir.1973) ("an agency that relies on an 'assessment' to dispense with an impact statement may well run risks not warranted by any countervailing benefit"); Hanly v. Kleindienst, 471 F.2d 823, 836 (2d Cir.1973).

Moreover, under NEPA and its implementing regulations, we cannot accept the EA's as a substitute for an EIS--despite the time, effort, and analysis that went into their production--because an EA and an EIS serve very different purposes. An EA aims simply to identify (and assess the 'significance' of) potential impacts on the environment; it does not balance different kinds of positive and negative environmental effects, one against the other; nor does it weigh negative environmental impacts against a project's other objectives, such as, for example, economic development. This latter balancing job belongs to the officials who decide whether to approve the project; and (where there are 'significant effects') those officials should make the decision in light of an EIS. An EIS helps them make their decision by describing and evaluating the project's likely effects on the environment. The purpose of an EA is simply to help the agencies decide if an EIS is needed.

To treat an EA as if it were an EIS would confuse these different roles, to the point where neither the agency nor those outside it could be certain that the government fully recognized and took proper account of environmental effects in making a decision with a likely significant impact on the environment. For one thing, those outside the agency have less opportunity to comment on an EA than on an EIS. See 40 C.F.R. Secs. 1503.1, 1506.6 (1984); Massachusetts v. Watt, 716 F.2d 946, 951 (1st Cir.1983) and authorities cited there. For another thing, those inside the agency might pay less attention to environmental effects when described in an EA than when described in an EIS. At the same time, if we assume that the EA's at issue here offer nearly as thorough an analysis as would an EIS, the agencies will not find it difficult to comply with the additional procedural EIS requirements that NEPA imposes. (Indeed, at oral argument, counsel for the appellees said that the agencies need take only one additional step to meet EIS requirements, namely to offer additional opportunity for public comment.)

For these reasons, we will not treat the EA's before us as if they were an EIS or as if they reflected an ultimate decision to proceed with the project. Relevant decisionmakers may well find that the project's economic good outweighs any environmental harm. But, at this point, we consider only the lawfulness of the agencies' finding that the project will have no significant impact on the environment. The legal issue, as we have said, is whether the record reveals that conclusion to be unreasonable, to the point where the decision not to prepare an EIS either violates NEPA or (what here comes to the same thing) is "arbitrary, capricious, an abuse of discretion." 5 U.S.C. Sec. 706(2)(A).

* The EA's before us concern the likely environmental effects of building the causeway and the port. Our reading of the record indicates that the major environmental bones of contention have included the following:

1. Clam flats. Building the causeway will eliminate 1.5 to 2 acres of clam flats; construction of the port will eliminate another 1.5 to 2 acres. Maine's DOT, however, will replace 2.14 acres of this habitat by seeding an area east of the causeway. And the Corps' permit for the port/causeway project requires that the state of Maine submit a plan for additional mitigation before proceeding with the project.

2. Lobsters, scallops, and other marine animals. Building the marine terminal will require dredging and filling 90 acres now inhabited by lobsters, scallops, and other marine animals. Thirty-five acres of this habitat will be permanently lost; the balance--including scallop fishing grounds and lobster fishing grounds (with up to 500 lobster pots)--will be dredged in order to create deep water access to the port, and therefore may be lost only temporarily. Maine's DOT, however, noted that many of the lobsters can move elsewhere and the scallop grounds are not very productive. Moreover, the Corps has conditioned its permit on Maine's preparing a report on the need for (and, if necessary, a plan for) additional mitigation of these habitat losses.

3. Waterfowl. The appellants and several environmental agencies said the project would adversely affect certain birds by encroaching on their current habitat. Maine's DOT, after consulting with Maine wildlife agencies, concluded that it would not do so because the birds' winter feeding on the island is limited to areas that do not freeze, and the project will not deprive them of a significant amount of such habitat.

4. Seals. One of the federal environmental agencies argued that the project would drive seals from the area. Maine's DOT, however, concluded that the harbor seals are already accustomed to the shipping traffic and would not be significantly disturbed.

5. Upland habitat. The parties agree that the port terminal would eliminate at least 40 acres of wooded upland habitat which supports several kinds of mammals and birds (including foxes, white-tailed deer, Osprey, and Woodcock). Maine's DOT and the Corps concluded, however, that the loss was not significant because the 40 acres represent only 4 percent of the island's total 'upland habitat'; displaced animals could go elsewhere; and the area has an abundance of such resources.

6. Run-off. Several federal environmental agencies and the Corps noted that the 'run-off' of water contaminated by oil, grease, and toxic substances from the marine terminal could pollute water in the harbor. Consequently, the Corps included in its permit for the project the condition that the terminal include

facilities capable [of] preventing the discharge of sediment grease and oil associated with storm drainage discharge to the satisfaction of the Division Engineer.

7. Tidal exchange. During each tidal cycle, approximately 22 million cubic feet of water crosses the gravel bar along which the state intends to build the causeway. The Corps has required Maine to include a 24-inch diameter culvert through the causeway to permit some tidal exchange between the two harbors that the causeway will separate. In light of this requirement, and the fact that the 22 million cubic feet of water represents only 3 percent of the total amount of tidal flow in and out of the two harbors, the Corps concluded that any impact on tidal exchange would not prove significant.

8. Dredging and 'spoil' disposal. Construction of the port will require the disposal of over 2 million cubic yards of dredged material (called 'spoils')--1.3 million in the initial phases of the project; 750,000 in the later stages. The Fish and Wildlife Service feared that Maine's plan to dump the spoils at a special ocean dump site would destroy a "benthic community"--organisms on which fish and other sea animals feed. The Corps concluded that this possibility was not environmentally significant because the dredged material would cover only 65 acres, the 'communities' could reestablish themselves, and Maine agreed to consider other ways of disposing of some of the material.

Whether or not these environmental effects, when considered together, do, or do not, show a "significant impact" is arguable. We note that the federal agencies, including the project's agency sponsors, differ among themselves about the significance of some of these effects. The Corps' EA, for example, criticized Maine DOT's environmental assessment of the project, for (in the Corps' view) it did "not adequately address impacts" (A. 378). We also note that certain factual matters are in dispute. For example, the environmental agencies have argued that the Corps' estimate of the number of undersea acres that will be covered by the dredging spoils (first, 20--40 acres, then 65 acres) is too low; and, the record contains a Corps memo suggesting that dredged material may cover 194 undersea acres, not 65. Finally, we note that the Corps evidently believes that promises to mitigate certain environmental impacts in the future mean that these impacts lack 'significance.' The CEQ, however, has written:

Mitigation measures may be relied upon to make a finding of no significant impact only if they are imposed by statute or regulation, or submitted by an applicant or agency as part of the original proposal. As a general rule, the regulations contemplate that agencies ... should not rely on the possibility of mitigation to avoid the EIS requirement.

If a proposal appears to have adverse effects which would be significant, and certain mitigation measures are then developed during the scoping or EA stages, the existence of such possible mitigation does not obviate the need for an EIS.... [Preparation of an EIS] is essential to ensure that the final decision is based on all the relevant factors and that the full NEPA process will result in enforceable mitigation measures through the Record of Decision.

Forty Most Asked Questions, supra, 46 Fed.Reg. 18028, 18038 (1981) (citations omitted) (emphasis in original). Regardless, were only the above-mentioned impacts at issue, we doubt that we could say that the 'FONSI' conclusions of the Corps and the Federal Highway Administration were "arbitrary, capricious, an abuse of discretion." 5 U.S.C. Sec. 706(2)(A).

The problems just noted become significant, however, when combined with a more serious omission by the Corps and the Federal Highway Administration--their failure to consider adequately the fact that building a port and causeway may lead to the further industrial development of Sears Island, and that further development will significantly affect the environment. The CEQ says that agencies must take account of such "indirect effects," which it defines as those that are

caused by the action and are later in time or farther removed in distance, but are still reasonably foreseeable. Indirect effects may include growth inducing effects and other effects related to induced changes in the pattern of land use, population density, or growth rate, and related effects on air and water and other natural systems, including ecosystems.

40 C.F.R. Sec. 1508.8 (emphasis added). Of course, agencies need not consider highly speculative or indefinite impacts. See, e.g., Kleppe v. Sierra Club, 427 U.S. 390, 402, 96 S. Ct. 2718, 2726, 49 L. Ed. 2d 576 (1976). But, here the 'impacts' seem neither speculative nor indefinite.

Whether a particular set of impacts is definite enough to take into account, or too speculative to warrant consideration, reflects several different factors. With what confidence can one say that the impacts are likely to occur? Can one describe them 'now' with sufficient specificity to make their consideration useful? If the decisionmaker does not take them into account 'now,' will the decisionmaker be able to take account of them before the agency is so firmly committed to the project that further environmental knowledge, as a practical matter, will prove irrelevant to the government's decision? See Massachusetts v. Watt, 716 F.2d at 952-53.

In this case, the record contains clear answers to these questions. And those answers show that the agencies should have taken account of the "secondary impacts." First, the record makes it nearly impossible to doubt that building the causeway and port will lead to further development of Sears Island. Local planners have considered the port, causeway, and industrial park to be components of an integrated plan. The town of Searsport, for example, begins its "Municipal Response Plan" as follows:

The 517 acres comprising the southern half of Sears Island have become the focus of considerable attention over the last few years as efforts to develop both a major cargo port and an industrial park have crystallized.

This two part development package has emerged in response to identified needs both for cargo port development in Maine and for economic development in the Waldo County area.

(Emphasis added). This theme is echoed in several of the EA's prepared for the Sears Island project. Maine DOT's August 1983 EA, for example, projected further industrial development after construction of the cargo port:

Development of the cargo terminal will ... act as the principal stimulus to further industrial development on the island itself. An estimated 350 acres on the island is expected to be made available by the present land owner (Bangor Investment Company [a subsidiary of Bangor & Aroostock Railway]. Central Maine Power already has a lease on an additional 423 acres.... Industries likely to be attracted to the island include those heavily dependent on port and rail facilities.... Several forest product and food industries are also expected to have facilities on the island, as well as suppliers of paper-making machinery and machinery components.

* * *

Industrial development, indirectly stimulated by constructing the cargo terminal, will generate increased revenues [for the town].... The eventual fiscal impact on the town will, of course, depend upon the degree and timing of the expected co-development of the island.

(A. 243, 248 (emphasis added).) Cf. Citizens for Responsible Area Growth v. Adams, 477 F. Supp. 994, 1001-02 (D.N.H.1979) (comprehensive EIS required where history of project shows that community and various agencies considered the three parts of airport improvement project as related). Maine DOT says that Searsport has "actively pursued such development for the Island" (A. 354). And it pointed to this development as one of the important factors in its selection of Searsport as the site for the marine terminal:

While development in Portland and other sites investigated would require less natural habitat destruction, development at these sites would not stimulate further industrial development to the extent that developing a cargo terminal on Sears Island would.

(A. 289.) See Chelsea Neighborhood Associations v. U.S. Postal Service, 516 F.2d 378, 388 (2d Cir.1975) (construction of new housing was a "selling point" for proposed postal facility, so EIS must consider it); Sierra Club v. Sigler, 695 F.2d 957, 979 (5th Cir.1983) (bulk cargo activities a "selling point" for oil project, so EIS must consider them); City of Davis v. Coleman, 521 F.2d 661, 676-77 (9th Cir.1975) (EIS must include consideration of "growth-inducing effects" of proposed highway construction project, where those effects are the project's "raison d'etre"). Moreover, the record reveals that, after negotiations with town officials, the owner of the lower half of the island has already agreed in principle to accept a four-fold increase in its real estate tax assessment (from $2,360/acre to $10,000/acre) in exchange for town financing of part of the causeway construction costs--an increase that provides a powerful incentive to develop the land.

Second, the plans for further development are precise enough for an EIS usefully to take account them into account. The record contains, for example, a 35-page "Land Use Plan/Industrial Marketing Study" prepared for the owner of the southern half of the island, and the town's 50-page "Municipal Response Plan for the Industrial Development of Sears Island." These documents provide detailed descriptions of likely further development, including a plot plan of the proposed industrial park that shows measured lots of 15, 25, and 50 acres, and locations for railroad and secondary road loops (see Appendix). The reports analyze the physical characteristics of the lower half of the island, discuss the feasibility of construction at various sites on the island, and outline development options; they also discuss the likely impact of industrial development on employment, housing, medical services, municipal services, the environment, and so forth. Actual further development may not follow the precise course that these documents suggest, but the CEQ has developed practical methods for dealing with such "secondary impacts." It states:

[T]he agency is not required to engage in speculation.... But ... it will often be possible to consider likely purchasers [of land] and the development trends in that area or similar areas in recent years.... The agency has the responsibility to make an informed judgment, and to estimate future impacts on that basis.... The agency cannot ignore these uncertain, but probable, effects of its decisions.

Forty Most Asked Questions, supra, 46 Fed.Reg. at 18031 (emphasis added). The 'land use' and 'response' plans are detailed enough for an EIS to describe the type of development likely to occur, even if it is pointless to analyze precise details.

Third, once Maine completes the causeway and port, pressure to develop the rest of the island could well prove irreversible. There is some indication in the record that if the Sears Island terminal is built, there will probably not be another major port facility built in Maine for a long time, and whatever new shipping traffic and industrial development is created by the project will be funneled into the Searsport area. See 40 C.F.R. 1508.27(b)(6) (in deciding whether an EIS is necessary, agency must consider the "degree to which an action may establish a precedent for future actions with significant effects ") (emphasis added); cf. 42 U.S.C. Sec. 4332(2)(C)(v) (EIS must consider "any irreversible and irretrievable commitments of resources which would be involved in the proposed action should it be implemented"). Even if federal authorities were to have an opportunity to consider the environmental effects of the industrial park at a later time, that later consideration would be unlikely to offer the decisionmaker a meaningful choice about whether to proceed. See Massachusetts v. Watt, 716 F.2d at 952-53 (describing each step in proposed offshore leasing program as "a link in a chain of bureaucratic commitment that will become progressively harder to undo the longer it continues").

The appellees point to Kleppe v. Sierra Club, 427 U.S. 390 at 402, 96 S. Ct. at 2726, where the Supreme Court wrote:

Where no ... plan exists, any attempt to produce an impact statement would be little more than ... [an] estimate[ ] of potential development and attendant environmental consequences. There would be no factual predicate for the production of an environmental impact statement of the type envisioned by NEPA.

But Kleppe is not contrary to the CEQ or to our analysis here. In Kleppe, federal agencies had prepared EIS's for individual mining operations and for a national mining program; appellants had argued that certain federal agencies should have prepared an additional, separate EIS for regional mining activity in the Northern Great Plains. The Court pointed out that there was "no proposal for a regional plan or program of development." 427 U.S. at 404, 96 S.Ct. at 2727 (emphasis added). Thus, it saw a separate EIS as impractical and unnecessary. Here, plans and a proposal for the industrial park exist; we are not considering an additional EIS for the industrial park, but whether its likely construction should be taken into account in an EIS for the Sears Island project. And, we conclude that taking this aspect of the project into account, unlike the Kleppe EIS, is both practical and necessary.

Although Maine DOT says that likely industrial development fell outside the scope of its EA, the record shows that each of the EA's discussed this development to some extent. Thus, in fairness to the agencies, we should not rest simply upon the fact that the Maine DOT says that it did not take full account of the industrial development; rather, we must go on to decide whether the consideration of industrial development that the EA's actually made was adequate--i.e., could the agencies reasonably conclude that the project (including the likely further development) would have no significant impact on the environment?

We do not see how the agencies could reach such a conclusion given the town's report and other record items that show that industrial development of the island would add (even under the town's 160-acre, 'moderate development' scenario) 2,750 new jobs in a town with a population of under 2,500; that show increased traffic on Route 1 (the town's principal artery) by 50-71 percent; that show additional lost scallop beds and clam flats, more soil erosion and aesthetic harm, a need for additional waste disposal and water supply, an added threat to water quality, and (under the town's 300-acre, 'full development' scenario) a loss of at least 23 percent of the island's upland animal habitat.

We have read the agency EA's, searching for information that might allow a conclusion that these impacts are not significant. But we have not found it. We find the explanations offered by the agencies as to why these impacts are not significant unpersuasive. First, some of the EA's suggest that secondary impacts might be "mitigated." But, as we noted above, mitigation cannot, by itself, render impacts 'insignificant' unless the mitigation measures are "imposed by statute or regulation, or submitted ... as part of the original proposal." 46 Fed.Reg. at 18038. Here, the agencies have not said what mitigation measures they have in mind with respect to the industrial park.

Second, some of the EA's point to the economic benefits of industrial development. The Corps' EA, for example, says that

The area is economically depressed and the benefits that would accrue from development of the island would outweigh any loss to wildlife or habitat.

(A. 413.) This type of argument, however, is not relevant to the question of the existence of significant environmental effects. It says that adverse effects (even if significant) are warranted --a matter that must, under NEPA, be decided in light of an EIS. As the CEQ regulations make clear, "[a] significant effect may exist even if the Federal agency believes that on balance the effect will be beneficial." 40 C.F.R. 1508.27(b)(1).

Third, the Corps' points to various land use regulations as a safeguard, noting that "[t]here is adequate local, state and Federal regulations to prevent complete devastation of the island" (A. 413). Of course, the drafters of NEPA had more in mind than avoiding "complete devastation." The Corps also points to the fact that Searsport has itself zoned the island for industrial use--a fact that it finds to be "evidence supporting a finding of no significant impact." Preservation Coalition Inc. v. Pierce, 667 F.2d 851, 861 (9th Cir.1982). The cited case, however, rested on a finding that the zoning there (in Pierce ) was consistent with the existing pattern of (urban) land use--a use that the proposed urban renewal project would not have altered. The project before us, however, would radically alter existing land use. See 32 C.F.R. Sec. 651.28(e) (1984) (requiring Corps to prepare EIS for projects "normally requiring EISs," such as "actions which may lead to significant change in land use ") (emphasis added); 23 C.F.R. Sec. 771.115(a)(3) (1984) (similar requirement for Federal Highway Administration).

Fourth, the Corps' EA dismisses the impact on 'upland habitat' as insignificant on the ground that there is adequate habitat elsewhere in the area. The force of this argument depends upon the meaning of the word "area." If the Corps means "elsewhere on Sears Island," its conclusion is weak. The EA's indicate that 'full development' of the southern half of the island will increase the amount of upland habitat 'taken' from 4 percent of the island to 23 percent. Since this latter figure is based on a calculation that counts as 'taken' only the land which is actually cleared for construction of buildings and parking lots, the percentage of 'upland habitat' that the development will render unsuitable for wildlife would likely be far higher. If the Corps means to include in the relevant "area" other land on the coast of Maine, its conclusion is more reasonable. Yet we doubt that the Corps can include so wide an area in its calculation. The CEQ's regulations state that "in the case of a site-specific action, significance [of environmental impacts] would usually depend upon the effects in the locale rather than in the world as a whole." 40 C.F.R. Sec. 1508.27(a) (1984) (emphasis added). Here, the nature of the action, and the geographical character of Sears Island, suggest that the appropriate "locale" is Sears Island and its immediate surroundings.

Fifth, the Corps' EA concludes that because the secondary development being considered is "light dry industrial," "there should be no significant air or water quality impacts from this type of industrial operation" (A. 412). This conclusion, however, is not supported or explained, nor does the record indicate why this is so.

As we read the record, the relevant federal agencies' conclusions that the project would have no significant impact rest not so much on their belief that the industrial park would not affect the environment, as upon their belief that they need not take account of the industrial park's effects. And, as we have indicated, we do not believe that NEPA permits the agencies here to ignore these impacts. See Scientists' Institute for Public Information v. AEC, 481 F.2d 1079, 1087 (D.C.Cir.1973) (EIS required for overall project where individual parts of project are related logically or geographically); City of Rochester v. U.S. Postal Service, 541 F.2d 967, 972 (2d Cir.1976) ("To permit noncomprehensive consideration of a project divisible into smaller parts, each of which taken alone does not have a significant impact but which taken as a whole has cumulative significant impact, would provide a clear loophole to NEPA."); Susquehana Valley Alliance v. Three Mile Island, 619 F.2d 231, 240 (3d Cir.1980) (same), cert. denied, 449 U.S. 1096, 101 S. Ct. 893, 66 L. Ed. 2d 824 (1981); West Chicago v. U.S. Nuclear Regulatory Commission, 701 F.2d 632 (7th Cir.1983) (illegal "piecemealing" or "segmentation" allows agency to avoid requirements of the Act); Alpine Lakes Protection Society v. Schlapfer, 518 F.2d 1089, 1090 (9th Cir.1975) (per curiam) (such segmentation merits "close scrutiny to prevent the policies of NEPA from being nibbled away"); see generally W. Rodgers, Environmental Law Secs. 7.9, at 787-92 (1977) ("NEPA discourages the cramped look that segmentation represents.").

In sum, given the likely secondary effects of the Sears Island project and the other effects previously described, the record in this case cannot support a FONSI, and therefore an EIS must be prepared. We reach this conclusion not because preparation of an EIS is merely a technical requirement which, under NEPA and its implementing regulations, we must here enforce. Rather, this requirement reflects NEPA's underlying purpose in requiring agencies to determine and assess environmental effects in a systematic way--namely, having decisionmakers focus on these effects when they make major decisions. That is to say, the requirement flows not only from the letter, but also from the spirit, of NEPA.

The decision of the district court is vacated and the case remanded for proceedings consistent with this opinion.

NOTE: OPINION CONTAINS TABLE OR OTHER DATA THAT IS NOT VIEWABLE

NOTE: OPINION CONTAINS TABLE OR OTHER DATA THAT IS NOT VIEWABLE

NOTE: OPINION CONTAINS TABLE OR OTHER DATA THAT IS NOT VIEWABLE