Opinion by Judge FLETCHER; Partial Concurrence and Partial Dissent by Judge RYMER.
FLETCHER, Circuit Judge:The plaintiffs appeal the district court’s dismissal of their suit under the National Environmental Policy Act (NEPA), 42 U.S.C. § 4321 et seq. The district court determined that a Second Environmental Impact Statement Supplement (EISS-2), prepared by the Army Corps of Engineers (“Corps”) in response to our decision in ONRC v. Marsh, 832 F.2d 1489 (9th Cir.1987) (“Marsh II”), rev’d and remanded in part, 490 U.S. 360, 109 S.Ct. 1851, 104 L.Ed.2d 377 (1989), had taken a “hard look” at the cumulative environmental effects of the proposed Elk Creek Dam. The plaintiffs also appeal the district court’s denial of their request for attorneys fees pursuant to the Equal Access to Justice Act (EAJA), 28 U.S.C. § 2412. We have jurisdiction, and we reverse in part and affirm in part.
I
In 1962, Congress authorized the Corps to construct three dams in southern Oregon’s Rogue River Basin. Flood Control Act of 1962, Pub.L. No. 87-874, § 101, 76 Stat. 1173 (1962). Two of the projects, the Lost Creek and Applegate River Dams, have been completed. The subject of this litigation is the third dam, the Elk Creek Project, which has been standing uncompleted since 1987 at one-third of its designed height.
Elk Creek is a tributary of the Rogue River and is located upstream of the segment of the Rogue River that Congress has desig*1487nated as Wild and Scenic under the Wild and Scenic Rivers Act (WSRA), 16 U.S.C. 1271, et seq. Elk Creek supports the wild coho salmon and steelhead trout runs which pass through the Wild and Scenic portion of the Rogue River. The river’s wild coho and summer steelhead populations have declined since 1968, and the species are now in danger of extinction due to actual or threatened destruction of their spawning and rearing habitat.
In 1982, the Corps approved construction of the Elk Creek Dam based on a 1980 Environmental Impact Statement Supplement (EISS-1). The Oregon Natural Resources Council (“ONRC”)1 brought this action under NEPA in October 1985 to prevent the Corps from building the dam, alleging in part that the Corps had violated NEPA by failing to prepare adequate documentation of the environmental effects of the proposed Elk Creek Dam. In its first look at the case, the district court held that the Corps had complied with NEPA in all respects and denied ONRC’s request for an injunction to stop construction of Elk Creek Dam. ONRC v. Marsh, 628 F.Supp. 1557, 1563-69 (D.Or.1986) (“Marsh I”).
We reversed in Marsh II, in part because EISS-1 had failed to discuss the cumulative impact of the Lost Creek, Applegate, and Elk Creek dams taken together. 882 F.2d at 1498. Accordingly, we remanded to the district court for entry of appropriate injunctive relief. The district court enjoined further construction of the dam while the Corps prepared a new EISS. ONRC v. Marsh, 677 F.Supp. 1072, 1078 (D.Or.1987) (“Marsh III”). Although the Corps sought and obtained certiorari to the United States Supreme Court, it did hot petition for review of our holding that it had failed to discuss the cumulative impacts of the three Rogue River dams. Thus, although the Supreme Court reversed Marsh II on the issues for which certiorari was granted, Marsh v. ONRC, 490 U.S. 360, 369-85, 109 S.Ct. 1851, 1857-65, 104 L.Ed.2d 377 (1989) (“Marsh 7F’), our decision that EISS-1 had failed to discuss cumulative impacts was left intact. See ONRC v. Marsh, 880 F.2d 242, 242 (9th Cir.1989) (remand to the district court after Marsh IV) (“Marsh V’).
The Corps issued EISS-2 in May 1991. EISS-2 considers the environmental effects of a No Action Alternative, under which the Corps would not complete construction of the Elk Creek Dam, and of constructing and operating the dam under three different operating alternatives. Two of the operating alternatives, the Full Pool Alternative and the Minimum Pool Alternative, would involve using the dam for water conservation purposes as well as for flood control. Under the third, No Conservation Pool Alternative, the dam would be used only for flood control, with the possibility of conservation use at some time in the future pursuant to new environmental impact studies. EISS-2 recommends completion of construction under the No Conservation Pool Alternative.
During the public comment period, ONRC criticized the Corps’ draft of EISS-2 in part for failing to discuss adequately the cumulative impacts of the dams. Nevertheless, on January 24, 1992, Ernest Harrell, the Corps’ Division Engineer, issued the Corps’ Record of Decision approving construction of the dam under the No Conservation Pool Alternative.. Subsequently, on July 22, 1992, the Corps moved to dissolve the outstanding injunction against further construction, arguing that EISS-2 discussed cumulative environmental impacts in accordance with this court’s mandate in Marsh II.
Meanwhile, ONRC asked the Secretaries of the Forest Service and Bureau of Land Management to issue a determination under section 7(a) of the Wild and Scenic Rivers Act. See 16 U.S.C. § 1278(a). On November 5, 1992, the Secretaries issued a joint determination that, due to reduction of the coho and steelhead’s spawning and rearing habitat and the impediment to the fishes’ migration, the uncompleted Elk Creek Dam unreasonably diminishes the values for which the Wild and Scenic portion of the Rogue River was so designated. The Secretaries determined, however, that the diminishment of the Wild and Scenic values of the Rogue *1488River would not be unreasonable were fish passage assured.
In response to the Secretaries’ determination, ONRC sought modification of the injunction issued in Marsh III, asking the district court to order Corps either to demolish the dam or to remove the danfs spillway on the grounds that the dam, although incomplete, unreasonably diminishes the anadro-mous fish populations of the Rogue River. ONRC also alleged that Corps had violated NEPA in several regards, including failing to comply with this court’s mandate that EISS-2 analyze cumulative environmental impacts. ONRC’s claims under WSRA and claims of additional NEPA violations were filed as part of a new action, and we discuss them separately in an opinion filed today. ONRC v. Harrell, 52 F.3d 1499 (9th Cir.1995).
Discussing the remaining issue in Marsh, EISS-2’s analysis of cumulative environmental impacts, the district court held that EISS-2 had taken a “hard look” at the cumulative effects of the Rogue River Basin Projects on streamflows, water temperature and turbidity, fisheries, wildlife, local recreation, and the local'economy by employing an incremental approach under which the Corps described the base, pre-dam conditions; the effects of adding the Lost Creek and Apple-gate River Dams; and, finally, the predicted effects of Elk Creek Dam under each of the four proposed alternatives. ONRC v. Marsh, 845 F.Supp. 758 (D.Or.1994) (“Marsh VF). Accordingly, the district court dismissed the Marsh action and refused either to order the destruction of the dam or the removal of its spillway, or to require the Corps to prepare a new EISS. The district court also denied ONRC’s request for attorneys fees under the Equal Access to Justice Act (EAJA), 28 U.S.C. § 2412, on the ground that the Corps’ position in Marsh I and Marsh II, that EISS-1 included an adequate cumulative impact analysis, although erroneous, was substantially justified.
II
We review de novo the district court’s determination that EISS-2 complied with NEPA. Marsh II, 832 F.2d at 1492. Our role is to assure that the Corps took a “hard look” at the environmental consequences of its decision to complete the Elk Creek Dam under the No Conservation Pool Alternative. Marble Mountain Audubon Soc’y v. Rice, 914 F.2d 179, 182 (9th Cir.1990). Although “we must defer to ‘the informed discretion of the responsible federal agencies,”’ Marsh IV, 490 U.S. at 377, 109 S.Ct. at 1861 (quoting Kleppe v. Sierra Club, 427 U.S. 390, 96 S.Ct. 2718, 49 L.Ed.2d 576 (1976)), and are not to “‘fly speck’ environmental impact statements,” Lathan v. Brinegar, 506 F.2d 677, 693 (9th Cir.1974), we will reverse an agency’s decision where it is contrary to procedures required by law, 5 U.S.C. § 706(2)(D); Marble Mountain Audubon Soc’y, 914 F.2d at 182, or where it is arbitrary or capricious, 5 U.S.C. § 706(2)(A); Greenpeace Action v. Franklin, 14 F.3d 1324, 1330-32 (9th Cir.1992).
A
ONRC argues that, although the Corps has discussed the cumulative environmental effects of the three Rogue River Basin projects to some extent, the scope of EISS-2’s cumulative impacts analysis is too narrow. The Corps argues that our decision in Marsh II required it only to discuss the cumulative impacts on two specific water quality factors — temperature and turbidity— and their effects on fish and fisheries production. The Corps maintains that “the scope of [EISS-2] was properly limited to an evaluation of these specifically identified issues of water quality” because they were the only factors explicitly mentioned by this court in Marsh II.
We wrote in Marsh II:
[ONRC] contends that the Corps unreasonably limited the scope of [EISS-1] by failing to consider the cumulative effects of all three dam projects, Lost Creek Dam, Applegate Dam, and Elk Creek Dam, the first two of which already have been completed.
We disagree with the district court that the Corps took a hard look at the cumulative environmental impacts. Plaintiffs point out two examples of the Corps’ fail*1489ure to consider the cumulative impacts. First, although [EISS-1] conceded that the overall basin project would increase the turbidity of the Rogue River, it concluded that there would be no major adverse effect on fish production from increased turbidity. In reaching this conclusion, however, it considered only the turbidity created by the individual project. Similarly, in response to a comment made by the United States Environmental Protection Agency suggesting that the EIS should discuss the effects of Lost Creek Dam on downstream water quality, the Corps responded that the Elk Creek EIS was not the proper place to discuss the effects of Lost Creek Lake. Insofar as the effects of the Lost Creek Dam were necessary to a complete presentation of the cumulative impact of construction of Elk Creek Dam, we disagree.
The synergistic impact of the project should be taken into account at some stage, and certainly before the last dam is completed.
832 F.2d at 1497-98.
We agree with ONRC that the Corps has read our mandate too narrowly. We mentioned the factors of turbidity and downstream water quality as “examples” demonstrating the Corps’ failure to discuss cumulative impacts at all. Having noted two instances where the Corps demonstrated its unwillingness in EISS-1 to discuss anything other than the effects of the Elk Creek Dam, considered in isolation, we held:
While a separate EIS is not required, the Corps should consider the impact of the Elk Creek Dam in conjunction with Lost Creek Dam and, if appropriate, Ap-plegate River Dam. The Corps is building Elk Creek Dam in an area that already has two dams. It must consider the area as it finds it and take into consideration the cumulative impact of the basin project.
Id. On remand, the Corps should have discussed, at a minimum, those areas of environmental impact that were discussed in EIS and EISS-1 and analyzed how those factors would be affected by the Elk Creek Dam, in combination with the Lost Creek and Apple-gate River Dams.
B
The Corps maintains that, even though it believed that our decision in Marsh II required it to discuss cumulative impacts only with regard to water temperature and turbidity, its erroneously narrow interpretation of our remand did not detrimentally affect the scope and quality of EISS-2’s cumulative impact analysis. We disagree. The Corps’ responses to comments submitted during the public comment period demonstrate that the Corps’narrow interpretation of our mandate in Marsh II was not merely a litigation position, but also defined the scope of the Corp's’ analysis. Rather than examine on its own initiative the cumulative impacts of the Rogue River Basin projects with regard to each of the environmentál factors considered in EIS and EISS-1, the Corps included within the scope of its cumulative impact analysis only the issue of water quality, measured only in terms of temperature and turbidity; discussed additional factors only if they were raised as concerns during the scoping process; and then refused to broaden the scope of its analysis when concerns were raised during the public comment period.
For example, ONRC submitted a comment suggesting that the scope of EISS-2 was too narrow and that the Corps should discuss additional environmental factors, including water quality parameters other than temperature and turbidity. The Corps responded that “[t]he scope of [EISS-2] was determined through a public scoping process which identified the significant issues to be addressed. The elements of the environment which you mention were not identified as significant issues and were therefore not included in the scope.” When the Oregon Department of Environmental Quality (DEQ) notified the Corps during the comment period that EISS-2 failed to describe water quality in relation to each of the applicable State water quality standards and that such a discussion was necessary to “make the analyses more meaningful and to properly evaluate and consider the cumulative effects,” Corps responded by stating that temperature and turbidity were the only water quality standards raised as concerns during the scoping process.
*1490Similarly, the Corps ignored suggestions that it consider the cumulative impacts on flows during flood events. The Corps recognized in EISS-2 that the effects of reservoirs on flows are most noticeable during spring filling, summer/fall drawdown, and flood events, but EISS-2 discusses the cumulative impacts of the dams only during spring filling and summer/fall drawdown. When ONRC noted during the comment period that the Corps should also describe the effect of the dams on flows during the flood control season, Corps’ only response was that the effect of the dams on flows during flood events was not “identified as an issue” by our opinion in Marsh II or during the scoping process.
That Corps limited the areas of its cumulative impacts analysis is apparent even in the summary of EISS-2, which is required to stress “areas of controversy (including issues raised by agencies and the public).” See 40 C.F.R. § 1502.12. Under “ISSUES ADDRESSED IN THIS EIS SUPPLEMENT,” EISS-2 reads:
“The issues identified by the Ninth Circuit Court and through the scoping process are the effects of the project on flows, water temperatures, and turbidity and the related effects on fish production and fishing in the Rogue River; the effects of the project on wildlife, and the proposed mitigation of those effects; ‘ and the effects of the project on recreation and the local economy. These issues are all addressed in this Supplement.”
Although the scoping process will normally identify most of the significant areas of discussion, see 40 C.F.R. § 1501.7 (describing the scoping process), the Corps cannot forever omit a factor from the scope of an EIS solely because the factor was not raised as a concern during the scoping process. An agency preparing an EIS has a duty to assess, consider, and respond to all comments, even those relating to environmental factors not mentioned during the scoping process. See id. § 1503.4(a). Although the agency'is not required to include in its final analysis every factor raised by such a comment and may respond, for example, by explaining why the comment does not warrant agency response, id. § 1503.4(a)(5), the mere fact that the comment raises concerns not mentioned during the scoping process does not establish that the comment warrants no further discussion. To hold otherwise would undermine “NEPA’s purpose of ensuring well-informed government decisions and stimulating public comment on agency actions.” Competitive Enter. Inst. v. National Highway Traffic Safety Admin., 901 F.2d 107, 123 (D.C.Cir.1990).
Because of the Corps’ refusal even to consider environmental factors not specifically identified by this court in Marsh II or raised during the scoping process, EISS-2’s cumulative impacts analysis omits issues of critical importance. Although we already have noted various deficiencies in EISS-2’s discussion of the inanimate environment, we are most disturbed by EISS-2’s inadequate analysis of Elk Creek Dam’s effects on the Rogue River’s wild coho and summer steelhead populations. For instance, Elk Creek provides valuable spawning and rearing habitat for wild coho and steelhead in the Rogue Basin, but EISS-2 does not discuss the effect of losing Elk Creek’s habitat on these populations and on fisheries beyond the Rogue River Basin. This omission is particularly disturbing in light of the Corps’ failure to mention in EISS-2’s description of fisheries that wild coho salmon are classified as a “sensitive” species by the State of Oregon.
The Corps attempts to excuse these and other omissions by pointing out that few wild coho still spawn in the upper portions of the Rogue River Basin, where Lost Creek and Elk Creek are located, and that EISS-2 included this fact in its description of the existing animate environment. The Corps’ reliance is misplaced. Discussing the decline of the wild coho in Elk Creek is not a substitute for disclosing that the species is classified as “sensitive” for the Rogue River Basin as a whole. More importantly, the scarcity of wild coho within Elk Creek renders a discussion of the dams’ detrimental effects on fish and fish production more imperative, not less. If the species generally was abundant but found relatively infrequently in Elk Creek, then preserving the habitat within the Elk Creek tributary for the relatively few number of fish found there might not be *1491critical; however, given that the species as a whole is considered “sensitive,” protection of these fish takes on new importance.
Because EISS-2 does not comply with our mandate in Marsh II, the Corps must prepare an additional EISS discussing the impact of the Elk Creek Dam in conjunction with the other Rogue River Basin projects, with regard not just to those factors specifically identified by us, but to all environmental factors essential to an informed agency decision. ONRC urges us to go further and to require the Corps either to demolish the partially-built Elk Creek Dam or to remove the dam’s spillway. As discussed in our opinion filed today in Harrell, 52 F.3d at 1508, such extraordinary relief is not warranted at this time in light of the less than fully developed record regarding the necessity, cost, and potential consequences of destruction.
Ill
ONRC also appeals the district court’s denial of its request for attorneys fees under the EAJA. It argues, first, that it was the prevailing party on issues other than cumulative impacts; therefore, ONRC argues, the district court should have awarded attorneys fees unless the Corps established that its litigation positions on these issues were substantially justified. Second, ONRC argues that the district court abused its discretion in determining that the Corps’ argument in Marsh I and II — that EISS-1 adequately discussed cumulative impacts — was substantially justified.
A
ONRC argues that our decision on cumulative impacts was not the only holding in Marsh II to survive Marsh IV intact. It contends that the Supreme Court did not reverse our decision in Marsh II to the extent that we held that the Corps should have discussed the effectiveness of the proposed mitigation measures and that scientific uncertainty existed with regard to the impact of the Elk Creek Dam on turbidity. ■
The district court held correctly that ONRC was not a prevailing party on these issues. Despite ONRC’s attempt to portray it as a separate “holding,” our decision in Marsh II to require the Corps to discuss the effectiveness of the proposed mitigation measures was one component of our call for a more thorough discussion of mitigation. The Supreme Court summarized Marsh II correctly when it stated that we held that EISS-1 “was defective because it did not include a complete 'mitigation plan!” See Marsh IV, 490 U.S. 360, 109 S.Ct. 1851. In Robertson v. Methow Valley Citizens Council, 490 U.S. 332, 348-53, 109 S.Ct. 1835, 1844-47, 104 L.Ed.2d 351 (1989), the companion case to Marsh IV, the Supreme Court rejected a requirement for a broad mitigation discussion. Accordingly, the Court in Marsh IV reversed our holding on the mitigation issue as “erroneous.” 490 U.S. at 369, 109 S.Ct. at 1857. We cannot say that our decision regarding EISS-l’s analysis of the effectiveness of mitigation measures survived Marsh TV.
ONRC also argues that, although the Supreme Court reversed our decision to require the Corps to conduct a “worst case” analysis, it left standing our determination that scientific uncertainty existed regarding the impact of the Elk Creek project on turbidity. However, even if this determination can be said to survive Supreme Court review, its relevance does not. The Supreme Court held that we should not have relied on a rescinded regulation, under which the duty to conduct a worst case analysis or further research was triggered by “scientific uncertainty,” see 40 C.F.R. § 1502.22(a) (1985), because, contrary .to our reasoning, the regulation was not a codification of prior NEPA ease law. Marsh TV, 490 U.S. at 369, 109 S.Ct. at 1857; Robertson, 490 U.S. at 354-56, 109 S.Ct. at 1848-49. Rather, we should have applied the amended regulation, under which only “incomplete or unavailable information” triggered the agency’s additional obligations. Id.; see 40 C.F.R. § 1502.22(a) (1988). Because we reached the issue of scientific uncertainty only because we applied an inappropriate legal standard, we reject ONRC’s characterization of that determination as an independent “holding” on which it was the prevailing party.
*1492B
The more difficult issue is whether the district court should have granted ONRC’s request for attorneys fees associated with litigating EISS-l’s failure to include an analysis of cumulative, impacts, an issue on which ONRC prevailed in Marsh II and which was not reviewed by the Supreme Court.' The district court recognized that ONRC was the prevailing party on the issue of cumulative impacts, but denied ONRC’s request for attorneys fees on the ground that the Corps’ argument that it had adequately discussed cumulative impacts was substantially justified.2
The burden is on the Corps to show that its position in Marsh I and Marsh II, though erroneous, was substantially justified. Yang v. Shalala, 22 F.3d 213, 217 (9th Cir.1994). We review for an abuse of discretion the district court’s decision that the Corps has met this burden. Williams v. Bowen, 966 F.2d 1269, 1260 (9th Cir.1991). “The district court abuses its discretion when its ‘decision is based on an erroneous conclusion of law or when the record contains no evidence on which [it] rationally could have based that decision.’ ” Kali v. Bowen, 854 F.2d 329, 331 (9th Cir.1988) (quoting In re Hill, 775 F.2d 1037, 1040 (9th Cir.1985)).
The district court held that, although we determined in Marsh II that EISS-l’s cumulative impacts discussion was deficient, the Corps had “considered the existing conditions and anticipated future consequences with and without the Elk Creek project.' The ‘existing conditions,’ at this third' and final stage of the broader ongoing project; included the already completed portion of the Rogue River Basin Project.” The district court appeared to believe that the Corps was required to prove substantial justification only with regard to the two deficiencies specifically identified by this court. The district court abused its discretion in relying on this finding as an indicator that the Corps had given some consideration to cumulative impacts. It was bound, as are we, by our decision in Marsh II that the Corps failed to discuss the environmental impacts of the Elk Creek Dam in conjunction with the Apple-gate River and Lost Creek Dams. As we have already explained in discussing the scope of the analysis that should have been included in EISS-2, we did not hold in Marsh II that the Corps’ cumulative impacts analysis was insufficiently thorough. Rather, our mandate that “Corps should consider the impact of Elk Creek Dam in conjunction with Lost Creek Dam and, if appropriate, Apple-gate River Dam,” 832 F.2d at 1498, reflects our conclusion that the Corps had not addressed cumulative impacts at all.
The duty to discuss cumulative impacts in an EIS is mandatory and is not within the agency’s discretion. See 40 C.F.R. §§ 1502.16, 1508.7-8. Although the government can lose on the merits but still' have a substantially justified position, our opinion in Marsh II, holding that the Corps completely omitted an area of analysis clearly required by the regulations implementing NEPA, establishes that this is not such a ease. See ONRC v. Madigan, 980 F.2d 1330, 1332 (9th Cir.1992) (decision that statutory language clearly set forth a duty to issue regulations foreclosed government’s position that it was substantially justified in arguing that the language was subject to different interpretations).
Moreover, the district court indicated that the Corps was required to prove substantial justification only with regard to the two deficiencies specifically identified by this court. To the extent that the district court relied on this excessively narrow interpretation of our mandate in Marsh II in assessing substantial justification, the district court erred. In fact, the Corps’ position on the two water quality issues specifically identified by the court as representative of EISS-l’s general defectiveness establish that the Corps’ approach to the cumulative impacts discussion was not substantially justified. The Corps had responded to an EPA comment concerning the cumulative impact on water quality not by arguing that its existing discussion was adequate, but by stating that the Elk Creek *1493EISS-1 was an inappropriate forum to discuss the effect of the Lost Creek Lake Dam. 832 F.2d at 1498. Moreover, the EISS-1 did not include any consideration of the environmental impact of the increased turbidity from the Elk Creek Dam in combination with the Lost Creek Lake and Applegate River Dams. Id. The Corps argues that in Marsh II we read its response to the EPA comment “out of context.” Similarly, the Corps argues that our second observation in' Marsh II, that EISS-1 failed to consider the Elk Creek Dam’s effects on turbidity in conjunction with the Applegate River and Lost Creek Dams, was “contrary to the record.” However, Corps’ arguments amount to a misplaced attempt to relitigate issues already resolved in Marsh II.
The district court also erred in holding that the Corps could reasonably have concluded that EISS-l’s reference to 1974 and 1979 water quality studies, which discussed the effects of Lost Creek Dam on water quality, adequately addressed the cumulative impact of the Rogue River Basin Projects on water quality. We agree with ONRC that, if the Corps did believe that reliance on the 1974 and 1979 studies was a sufficient substitute for including within EISS-1 an analysis of cumulative impacts, its belief was unreasonable because neither of those studies addressed the cumulative impact of turbidity changes upon fish production.
The Corps relies on the following language from Marsh TV to argue that its position was substantially justified: “Based on [1974 and 1979 water quality] studies, [EISS-1] predicted that changes in the ‘turbidity regime’ would not have any major effect on fish production, but that the combined effect of the Lost Creek and Elk Creek Dams on the turbidity of the Rogue River might, on occasion, impair fishing.” 490 U.S. at 365-66,109 S.Ct. at 1855. The Corps argues that this language evidences the Supreme Court’s implicit recognition that EISS-1 contained a reasonably thorough cumulative impacts discussion.
We are not convinced. The statement is taken from the Supreme Court’s summary of the background of the appeal and carries no indication of the Court’s intention to undermine our holding in Marsh II. Moreover, the statement simply describes the fact that Corps relied on the 1974 and 1979 studies to conclude that the cumulative impact of the dams on turbidity might impair fishing, but says nothing regarding whether EISS-1 considered how the cumulative impact on turbidity might affect fish production or whether the Corps’ reliance on the water quality studies as a substitute for án independent analysis of cumulative impacts was reasonable.
Although we affirm the district court’s denial of ONRC’s request for attorneys fees on the issues of mitigation of impacts and scientific uncertainty, we reverse the denial of fees on the issue of cumulative impacts. Corps was not substantially justified in arguing that EISS-1 fulfilled Corps’ duty to discuss cumulative impacts. The district court’s contrary decision was the result of its mistaken interpretation of our decision in Marsh II and was an abuse of discretion.
IV
The Corps must prepare an additional EISS evaluating the impacts of adding Elk Creek Dam to the existing dams in the Rogue River Basin Project. In doing so, the Corps should consider any environmental factors essential to informed agency decision-making. ONRC should be awarded attorneys fees incurred both in the district court and this court litigating EISS-l’s failure to discuss cumulative impacts. We transfer to the district court the determination of the appropriate amount of fees.
AFFIRMED IN PART, REVERSED IN PART, REMANDED.
. Other plaintiffs, referred to collectively as ONRC, are Oregon Guides and Packers Association, Inc., Rogue Flyfishers, Inc., and Rogue River Guides Association,
. The district court initially denied ONRC’s request without explanation other than that the Corps’ position was substantially justified, and we vacated and remanded for the district court to enter specific findings.