Plaintiffs in these consolidated cases challenge the spraying of herbicides on United States Forest Service (USFS) and Bureau of Land Management (BLM) lands. Both cases involve whether research on and disclosure of the potential carcinogenic, teratogenic and mutagenic effects of the herbicides is required under the National Environmental Policy Act of 1969, 42 U.S.C. Secs. 4331-4335 (1976) (NEPA).

The district court enjoined portions of both the BLM spraying program for the Eugene District of Oregon in Save Our Ecosystems v. Clark (SOS) and the USFS program for the State of Oregon in Merrell v. Block (Merrell). We affirm the district court's holdings in the two cases that the USFS and the BLM violated NEPA and the regulations of the Council on Environmental Quality (CEQ). However, we modify the injunctions to enjoin all spraying until the agencies comply with NEPA.

A. SOS v. Clark.

In 1978 the BLM prepared a programmatic environmental impact statement (PEIS) entitled "Vegetation Management With Herbicides: Western Oregon, 1978-1987." The statement discussed the environmental impacts of a ten-year program of herbicide spraying,1 intended to destroy undergrowth thereby increasing the growth rate of conifers. The PEIS was to be supplemented annually by an environmental assessment (EA), upon which would be based the decision whether to spray in the succeeding year, and, if so, how the spraying would be done.

In 1979 an organization called Southern Oregon Citizens Against Toxic Sprays (SOCATS ) filed suit to enjoin the BLM from spraying in the Medford District. In that case, Judge Frye enjoined the spraying because the BLM had failed to prepare a "worst case analysis" (WCA) under 40 C.F.R. Sec. 1502.22 (1981). That decision was affirmed by this court. See Southern Oregon Citizens Against Toxic Sprays v. Clark, 720 F.2d 1475 (9th Cir.1983) (SOCATS ).

In response to the district court decision in SOCATS, the BLM prepared a worst case analysis of its spraying program for the Eugene district. The plaintiffs in this case challenge its adequacy. Judge Belloni agreed with plaintiffs, but limited the injunction to prohibiting aerial spraying in a portion of the district and granted defendant's motion to stay the injunction pending appeal. We vacated the stay and reinstated the original injunction.

B. Merrell v. Block.

The Merrell case arises out of the USFS spraying program for its forests in Oregon, a program very similar to that of the BLM. In 1978 the Forest Service prepared a PEIS on "Vegetation Management With Herbicides" covering the Pacific Northwest Region.2 The PEIS was to be supplemented annually by an EA. Soon after spraying commenced in 1979, numerous and serious health problems were reported in the Five Rivers Valley, including spontaneous abortions, birth defects in humans and animals, and various other illnesses. The EPA began an investigation into these problems, but the Forest Service declined requests by the county health department and board of commissioners to delay the spraying.3 The Forest Service conducted no research of its own into these problems and, in its 1981 EA, concluded that the continued use of the herbicides would have no significant impact on the human environment and declined to prepare an EIS.

In 1981 Paul Merrell, a resident of the Five Rivers area of the Suislaw National Forest, filed a suit seeking an injunction against further spraying in that national forest. In response to cross-motions for summary judgment Judge Belloni held that the Forest Service could not rely solely on research done incident to the EPA registration of the chemicals under the Federal Insecticide, Fungicide, and Rodenticide Act, 7 U.S.C. Secs. 136-136y (FIFRA), and that the Forest Service must address the health effects of using the herbicides in the area to be sprayed. As in SOS, Judge Belloni enjoined only a portion of the spraying program and denied the motions of 42 additional individuals and organizations to intervene. They sought to broaden the injunction to prevent spraying in the areas where they lived (the injunction was limited to the area where the plaintiffs lived).

I. SOS v. Clark.

A. Worst Case Analysis.

1. NEPA Requires Analysis of Uncertain Risks.

CEQ regulations require an EIS to contain a "worst case analysis" when "the information relevant to adverse impacts is essential ... and is not known and the overall costs of obtaining it are exorbitant or ... the information ... is important and the means to obtain it are not known...." 40 C.F.R. Sec. 1502.22 (1981).4

40 C.F.R. Sec. 1508.27(b)(5) (1981) requires a similar analysis, although it is not specifically labelled a "worst case analysis." NEPA requires an impact statement for "major federal actions significantly affecting the quality of the human environment ...." 42 U.S.C. Sec. 4332(C) (1976). The CEQ defines "significantly" to include considerations of both context and intensity, and states that "intensity" includes "[t]he degree to which the possible effects on the human environment are highly uncertain or involve unique or unknown risks." 40 C.F.R. Sec. 1508.27(b)(5) (1981). See Foundation for North American Wild Sheep v. United States Department of Agriculture, 681 F.2d 1172, 1182 n. 47 (9th Cir.1982).

These regulations are binding on the BLM and entitled to substantial deference by the courts. Andrus v. Sierra Club, 442 U.S. 347, 358, 99 S. Ct. 2335, 2341, 60 L. Ed. 2d 943 (1979). They are not new requirements, but rather are a codification of prior case law that required analysis of the costs of proceeding without more and better information. SOCATS, 720 F.2d at 1478.

On their face these regulations require an ordered process by an agency when it is proceeding in the face of uncertainty. First, the agency must determine whether the information is important or essential5 and whether it can be obtained. If it cannot be obtained or if the costs of obtaining it are exorbitant, the agency must do a worst case analysis weighing the need for the action against all possible adverse impacts. The agency must consider the range of worst possible effects and the likelihood of these effects occurring. It must also consider the costs of proceeding without the information.

The BLM acknowledges that the issue of whether a WCA is required has been determined by our decision in SOCATS, where we held that the BLM must prepare a worst case analysis bottomed on the assumption that its herbicides are not safe. We noted that scientific uncertainty existed regarding the carcinogenacity of the herbicides and that "[w]hen uncertainty exists, it must be exposed." 720 F.2d at 1479. This is not all that is required, however. Besides exposing the fact of uncertainty, because of that uncertainty, a spectrum of possible events must be considered.

The CEQ interprets its regulation as follows:

The purpose of the analysis is to carry out NEPA's mandate for full disclosure to the public of the potential consequences of agency decisions, and to cause agencies to consider those potential consequences when acting on the basis of scientific uncertainties or gaps in available information. The analysis is formulated on the basis of available information, using reasonable projections of the worst possible consequences of a proposed action.

For example, if there are scientific uncertainties and gaps in the available information concerning the numbers of juvenile fish that would be entrained in a cooling facility, the responsible agency must disclose and consider the possibility of the loss of the commercial or sport fishery.

In addition to an analysis of a low probability/catastrophic impact event, the worst case analysis should also include a spectrum of events of higher probability but less drastic impact.

46 Fed.Reg. 18032 (Monday, March 3, 1981).

In Sierra Club v. Sigler, 695 F.2d 957 (5th Cir.1983), the Fifth Circuit required the Army Corps of Engineers to prepare a worst case analysis hypothesizing a massive oil spill in Galveston Bay. In a detailed analysis of the procedural history of the regulation, the court upheld the regulation under NEPA and said that " 'CEQ's interpretation of NEPA is entitled to substantial deference.' " 695 F.2d at 972, quoting Andrus v. Sierra Club, 442 U.S. 347, 358, 99 S. Ct. 2335, 2341, 60 L. Ed. 2d 943 (1979).

The court emphasized that the mere fact that the possibility of an event occurring is remote or unlikely does not obviate the necessity to do a worst case analysis.

All parties agree that a total cargo loss could occur and could wreak catastrophic environmental damage in the Bay. While this damage is a "significant adverse effect," there is considerable uncertainty about its likelihood, scope, and consequences; information on it is certainly important, if not essential, to the Corps' decision, yet that information is beyond the state of the art. However, there is a body of data with which a reasonable worst case analysis can be made that is not unreasonably speculative. Remoteness does not bar a worst case analysis so founded and should instead be weighed by the Corps when it applies the worst case analysis in its decisionmaking process. The Corps must at least consider information relevant to a "significant" effect of a proposal, if that information is both "important" to that decision and not based on "unreasonable speculation."

695 F.2d at 974. See also SOCATS, at 1479-80.6

Both the CEQ regulation and other NEPA cases contemplate analysis of a "spectrum of events."7 With these requirements in mind, we examine the WCA prepared by the BLM in this case.

2. Adequacy of the WCA.

The WCA prepared by the BLM is brief and cursory, and proceeds from an entirely wrong assumption. The BLM admits that no level of exposure to the herbicides has been proven safe, but assumes in the WCA that "a point is reached at which it becomes clear that no human health effect will occur." Judge Belloni said, "Plainly, the worst result that can occur as a result of proceeding in the face of uncertainty as to whether a herbicide causes cancer is that it does cause cancer." Slip op. at 5. We agree. We observed in SOCATS that, "The possibility that the safe level of dosage for herbicides is low or nonexistent creates a possibility of 'significant adverse effects on the human environment.' ... This potential calls for a worst case analysis." 720 F.2d at 1479, quoting 40 C.F.R. Sec. 1502.22.

The BLM argues that the analysis plaintiffs say is necessary would be pure guesswork because no credible data exist to support the proposition that cancer can occur at any dose. This contention is specious in light of the evidence presented by plaintiffs' experts and the holding of SOCATS that "[t]he agency may not omit the analysis only because it believes that the worst case is unlikely." 720 F.2d at 1479; see also Sigler, 695 F.2d at 974 ("that the possibility of a total cargo loss by a supertanker is remote does not obviate the requirement of a worst case analysis").8 The duty of an agency is to analyze the costs and environmental effects of the worst case and its costs and then to provide its assessment of the likelihood of the event occurring. The district court was correct in finding the WCA deficient.9

B. Public Comment Period for the WCA.

The BLM provided five days for public comment on the WCA before it made its final decision. It argues that the WCA was part of the EA and, accordingly, need be made available only at some time prior to the final decision.10

We agree with the district court that a longer comment period is required. As we stated in SOCATS, "[T]he label of the document is unimportant." 720 F.2d at 1480. For procedural purposes, we must examine the purpose which the document serves.

The BLM uses the annual EA to supplement its ten-year programatic EIS. Rather than using the EA simply to determine whether to prepare an EIS, the EA serves as the decisionmaking document to assess the environmental costs of each year's spraying program. The EA is an especially significant document where, as here, the WCA is included in it. See SOCATS, 720 F.2d at 1480. When an EA is the functional equivalent of an EIS, it is subject to the same procedures. See National Indian Youth Council v. Andrus, 501 F. Supp. 649, 657 (D.N.M.1980), aff'd 664 F.2d 220 (10th Cir.1981). Judge Belloni was correct in holding that the WCA was subject to the minimum 45-day comment period for draft EIS's contained in 40 C.F.R. Sec. 1506.10(c) (1983).11

II. Merrell v. Block.

A. NEPA Requires Research on the Environmental Effects of the USFS Program.

Judge Belloni found that "[w]hen USFS and BLM completed their EAR's [environmental assessment reports or EA's] they did not do any actual research on the health effects of using the listed herbicides in the area. Instead, they simply relied on research already completed by EPA when that agency registered the chemicals under FIFRA."

The Forest Service argues that Judge Belloni's decision requires it to do original research on the use of the herbicides and that such a requirement is beyond the scope of a court's authority under NEPA. Neither of these contentions is accurate.

Judge Belloni did not specify what the Forest Service must do, rather, he held that it could not rely solely on EPA registration under FIFRA.12 The Forest Service could appropriately consider EPA's data on the herbicides in the specific context of the area in which it proposes to spray; it could require the chemical companies, such as amicus Monsanto, to provide the data and necessary research on their herbicides for similar analysis; it could commission studies or undertake its own research or any combination of these options that would satisfy its obligations under NEPA to provide an adequate analysis of the effects of its spraying program in the targeted area.13

1. Reliance on EPA Registration is Improper.

The EPA registration process for herbicides under FIFRA is inadequate to address environmental concerns under NEPA, particularly where, as here, the registration is only conditional. Conditional registration is a recognition that less than complete data exists. FIFRA does not require or even contemplate the same examination that the Forest Service is required to undertake under NEPA. FIFRA registration is a cost-benefit analysis that no unreasonable risk exists "to man or the environment taking into account the economic, social and environmental costs and benefits of the use of any pesticide." 7 U.S.C. Sec. 136(bb).

Reliance on EPA data is clearly improper under this court's holdings in Oregon Environmental Council v. Kunzman, 714 F.2d 901 (9th Cir.1983) and SOCATS. In Kunzman we held that the Forest Service's spraying of carbaryl over populated areas violated NEPA where the PEIS and EA failed to address adequately the health effects of "spraying on people living or working in or near the areas sprayed." 714 F.2d at 904. We held that the Forest Service could not rely on EPA registration under FIFRA, citing Judge Skopil's opinion in Citizens Against Toxic Sprays v. Bergland, 428 F. Supp. 908, 927 (D.Or.1977). 714 F.2d at 905. We also relied on 40 C.F.R. Sec. 1508.27, discussed supra Section IA1, as an alternate ground for requiring a worst case analysis in SOS v. Clark.

We said in Kunzman that,

while the PEIS does briefly discuss the health risks of carbaryl, it does not consider the supposedly carcinogenic effects of an attendant by-product of carbaryl, nitrosocarbaryl .... The licensing of pesticides containing carbaryl does not "reflect a conclusion that a pesticide is safe under any condition," as the federal appellees assert in their brief. One agency cannot rely on another's examination of environmental effects under NEPA.... "Thus, the mere fact that a program involves use of substances registered under FIFRA does not exempt the program from the requirements of NEPA."

714 F.2d at 905 (citations omitted, emphasis added). We followed this decision in SOCATS, stating that "[t]he BLM must assess independently the safety of the herbicides that it uses." SOCATS, at 1480.

2. The Forest Service Must Do Research If No Adequate Data Exists.

We recognized in SOCATS that an agency may be required to do independent research on the health effects of a herbicide. This is not a new requirement.

In Foundation for North American Wild Sheep v. U.S. Dept. of Agriculture, 681 F.2d 1172 (9th Cir.1982), this court held an EIS inadequate because it failed to address the effect on bighorn sheep of opening a road when those effects were uncertain. We said, "the very purpose of NEPA's requirement that an EIS be prepared for all actions that may significantly affect the environment is to obviate the need for such speculation by insuring that available data is gathered and analyzed prior to the implementation of the proposed action." 681 F.2d at 1179 (emphasis added). Similarly, in Warm Springs Dam Task Force v. Gribble, 621 F.2d 1017 (9th Cir.1980), we held that an agency cured the defect in its EIS by commissioning a study about the effects of a newly discovered fault system on that dam. 621 F.2d at 1025-26. Other courts have imposed similar requirements on agencies. See, e.g., Rankin v. Coleman, 394 F. Supp. 647, 658 (highway project enjoined for inadequate EIS on effects and alternatives; alternatives must be "affirmatively studied"), mod. 401 F. Supp. 664 (E.D.N.C.1975); Montgomery v. Ellis, 364 F. Supp. 517, 528 (N.D.Ala.1973) ("NEPA requires each agency to undertake research needed adequately to expose environmental harms and, hence, to appraise available alternatives") (project enjoined pending preparation of an adequate EIS); Brooks v. Volpe, 350 F. Supp. 269, 279 ("NEPA requires each agency to indicate the research needed to adequately expose environmental harms"), supplemented, 350 F. Supp. 287 (W.D.Wash.1972), aff'd, 487 F.2d 1344 (9th Cir.1973); Environmental Defense Fund v. Hardin, 325 F. Supp. 1401, 1403 (D.D.C.1971) (interpreting section 102(2)(A) as making "the completion of an adequate research program a prerequisite to agency action .... The Act envisions that program formulation will be directed by research results rather than that research programs will be designed to substantiate programs already decided upon").

Furthermore, in SOCATS and in Warm Springs we recognized that such a duty also flowed from the worst case analysis regulation:

If the information relevant to adverse impacts is essential to a reasoned choice among alternatives and is not known and the overall costs of obtaining it are not exorbitant, the agency shall include the information in the environmental impact statement.

40 C.F.R. Sec. 1502.22(a) (emphasis added). Only if the costs are exorbitant or the means of obtaining the information is beyond the state of the art is the agency excused from compliance and allowed to perform a worst case analysis. 40 C.F.R. Sec. 1502.22(b).14 The Forest Service presents no evidence and makes no argument that the costs are exorbitant or that research is impossible. Rather, it argues that it cannot be forced to do it. Section 1502.22 clearly contemplates original research if necessary.

Federal agencies routinely either do their own studies or commission studies of the particular area in which a proposed project is to be located. Almost every EIS contains some original research. And, almost every time an EIS is ruled inadequate by a court it is because more data or research is needed. The Forest Service does not, and indeed cannot, cite any case which holds that an agency is not obliged to do research to comply with NEPA. The Forest Service cannot abdicate its responsibilities by relying on another agency. It must evaluate the impact of its own actions.

B. Preparation of a Joint EIS.

Plaintiffs argue that 40 C.F.R. Sec. 1501.5(a) requires the preparation of a joint EIS by the EPA, BLM, and USFS. This argument is unpersuasive in this case.

(a) A lead agency shall supervise the preparation of an environmental impact statement if more than one Federal agency either

(1) Proposes or is involved in the same action; or

(2) Is involved in a group of actions directly related to each other because of their functional interdependence or geographical proximity.

40 C.F.R. Sec. 1501.5(a).

Although both the BLM and the Forest Service are engaged in forest spraying programs, the programs are in separate areas. The EPA's function is clearly distinct. We are unwilling to find that the district court erred in refusing to mandate a joint EIS. This is not to say that the agencies and the public might not be better served were a joint EIS prepared. The programs of the BLM and the Forest Service are similar. The geographic areas in some instances are close. The terrain and conditions in the BLM and forest service areas may be very similar. Neither agency can ignore the program by the other. To the extent the program of one affects the other or the effects of the programs are cumulative neither agency can avoid assessing the impact of its program on the other. Joint participation by the EPA as a resource, consultant and expert would indeed be beneficial.

III. Scope of the Injunctions.

Judge Belloni consolidated SOS v. Clark and Merrell v. Block when considering the scope of the injunctions. We do likewise.

We are puzzled by the limited scope of the injunctions in these cases. Judge Belloni limited the injunctions in both cases in two ways, by limiting the territorial scope of the injunctions and by restricting the injunctions to aerial spraying (allowing ground spraying to go forward).

Irreparable damage is presumed when an agency fails to evaluate thoroughly the environmental impact of a proposed action. See Friends of the Earth Inc. v. Coleman, 518 F.2d 323, 330 (9th Cir.1975). Only in a rare circumstance may a court refuse to issue an injunction when it finds a NEPA violation.15 Indeed, the policies underlying NEPA "weight the scales in favor of those seeking the suspension of all action until the Act's requirements are met...." Alpine Lake Protection Society v. Schlapfer, 518 F.2d 1089, 1090 (9th Cir.1975). The "rare circumstance" arose in American Motorcyclist Association v. Watt, 714 F.2d 962 (9th Cir.1983) (AMA), where the court found more harm would result to the environment from granting the injunction than by refusing to grant it.16 See also Alpine Lake, 518 F.2d at 1090 (refusing to grant injunction because "unusual circumstances" meant more harm could occur to forest from disease if injunction not granted).

We find no such justification here that would excuse a total ban on spraying. The district court should have enjoined the spraying programs of the BLM and USFS in their entirety until NEPA requirements were fulfilled.

IV. Attorneys' Fees.

Plaintiffs in both cases seek fees under the Equal Access to Justice Act, 28 U.S.C. Sec. 2412 and 5 U.S.C. Sec. 504. In deciding whether fees should be awarded against the government in this case we are guided by our decision in SOCATS.

Relitigation of a previously decided issue is a strong factor weighing against the government in determining substantial justification. In SOCATS we held that the government was on notice that its reading of section 1502.22 was untenable after the appellate decision in Sigler. In accord with that decision, we hold that plaintiffs in SOS v. Clark are entitled to attorneys fees for legal services rendered at least from the date of that decision (January 20, 1983). Although prior adverse litigation is ordinarily the most important factor to consider, other relevant factors may warrant an award from an earlier date. See, e.g., Environmental Defense Fund v. Watt, 722 F.2d 1081-1086 (2d Cir.1983).

In Merrell v. Block, the Forest Service has known that reliance solely on EPA registration under FIFRA was improper since Judge Skopil's decision in Citizens Against Toxic Sprays, Inc. v. Bergland, 428 F. Supp. 908, 927 (D.Or.1977) ("the mere fact that a program involves the use of substances registered under FIFRA does not exempt the program from the requirements of NEPA"). Plaintiffs seek fees for this appeal only; their entitlement to fees before the district court is the subject of a separate appeal now pending before this court. We hold only that plaintiffs are entitled to fees on appeal because the government's position was not substantially justified before this court. We express no view as to the propriety of fees for the district court litigation.

We remand to the district court to take such further evidence as may be necessary to determine the amount of fees to be awarded. See, e.g., Kerr v. Screen Extras Guild, Inc., 526 F.2d 67, 70 (9th Cir.1975). In light of allegations made in respect to information withheld despite FOIA requests, the district court should consider such evidence in that regard as may have bearing on the government's bona fides and may affect the amount of attorneys' fees to be awarded.17

We affirm the district court's holdings that the WCA prepared by the BLM (in SOS v. Clark ) and the EA prepared by the Forest Service (in Merrell v. Block ) were inadequate. We affirm the district court's holding in SOS v. Clark that a 45-day comment period was applicable to the WCA and that no joint EIS was required. The district court erred, however, in limiting the scope of the injunctions in both cases: the entire spraying programs of both agencies should be halted until they comply with NEPA.18 The district court shall award attorneys' fees in both cases for services below and on appeal, in amounts to be determined by the trial court.

AFFIRMED in part, REVERSED in part, and REMANDED.