Opinion by Judge WILLIAM A. FLETCHER; Dissent by Judge N.R. SMITH.
ORDER
This court’s opinion filed on February 3, 2012, and reported at 668 F.3d 609 (9th Cir.2012), is withdrawn, and is replaced by the attached Opinion and Dissent.
With the filing of the new opinion, Judges Reinhardt and W. Fletcher vote to deny the petition for rehearing and the petition for rehearing en banc. Judge N.R. Smith votes to grant the petition for rehearing and the petition for rehearing en banc.
The full court has been advised of the petition for rehearing en banc and no judge of the court has requested a vote on whether to rehear the matter en banc. Fed. R.App. P. 35.
The petition for rehearing and the petition for rehearing en banc, filed on April 18, 2012, are DENIED.
No further petitions for rehearing or rehearing en banc will be accepted.
*1015OPINION
W. FLETCHER, Circuit Judge:The national forests of the Sierra Nevada Mountains (“the Sierras”) are home to a rich array of fauna, including at least 61 species of fish and 35 species of amphibians. The Sierra Nevada Ecosystem Project, a study commissioned by Congress, concluded in 1996 that their environment has been severely degraded: “The aquatic/riparian systems are the most altered and impaired habitats in the Sierra.”
The national forests in the Sierras are managed under eleven Forest Plans (“the Forest Plans”). In January 2001, the United States Forest Service (“Forest Service”) issued a Final Environmental Impact Statement (“2001 EIS”) recommending amendments to the Forest Plans in the Sierras. The amendments were intended, among other things, to conserve and repair the aquatic and riparian ecosystems. In January 2001, under the administration of President Clinton, the Forest Service adopted a modified version of the preferred alternative recommended in the 2001 EIS. The parties refer to this as the 2001 Framework.
In November 2001, under the administration of newly elected President Bush, the Chief of the Forest Service asked for a review of the 2001 Framework. In January 2004, the Forest Service issued a Final Supplemental Environmental Impact Statement (“2004 EIS”) recommending significant changes to the 2001 Framework. The Forest Service adopted the preferred alternative in the 2004 EIS. The parties refer to this as the 2004 Framework.
Plaintiff-Appellant Pacific Rivers Council (“Pacific Rivers”) brought suit in federal district court challenging the 2004 Framework as inconsistent with the National Environmental Protection Act (“NEPA”) and the Administrative Procedure Act (“APA”). The gravamen of Pacific Rivers’ complaint is that the 2004 EIS does not sufficiently analyze the environmental consequences of the 2004 Framework for fish and amphibians. On cross-motions for summary judgment, the district court granted summary judgment to the Forest Service.
Pacific Rivers timely appealed the grant of summary judgment. For the reasons that follow, we conclude that the Forest Service’s analysis of fish in the 2004 EIS does not comply with NEPA. However, we conclude that the Forest Service’s analysis of amphibians does comply with NEPA. We therefore reverse in part, affirm in part, and remand to the district court.
I. Background
Stretching along a north-south axis for more than 400 miles, the Sierra Nevada Mountains form one of the longest continuous mountain ranges in the lower 48 states. The Forest Service manages nearly 11.5 million acres of land under the Forest Plans. Each Forest Plan is a Land and Resource Management Plan (“LRMP”) formulated and promulgated pursuant to the National Forest Management Act (“NFMA”). See 16 U.S.C. § 1604. NFMA requires the Forest Service to provide for and to coordinate multiple uses of the national forests, including “outdoor recreation, range, timber, watershed, wildlife and fish, and wilderness.” 16 U.S.C. § 1604(e)(1). An LRMP adopted pursuant to NFMA guides all management decisions within the forests subject to that LRMP. Individual projects are developed according to the guiding principles and management goals expressed in the LRMP. See Ohio Forestry Ass’n, Inc. v. Sierra Club, 523 U.S. 726, 729-31, 118 S.Ct. 1665, 140 L.Ed.2d 921 (1998).
The Forest Plans govern the eleven national forests that run the length of the *1016Sierras from Southern California to the California-Oregon border — the Sequoia, Inyo, Sierra, Stanislaus, Humboldt-Toiyabe, Eldorado, Tahoe, Plumas, Lassen, and Modoc National Forests, and the Lake Tahoe Basin Management Unit. The area encompassed by the Plan amounts to more than 5% of the total forest land managed by the Forest Service. See http://www.fs. fed.us/r5/sierra/about/ (National Forests encompass 191 million acres). The forests support substantial economic activity, including logging and grazing, as well as recreation. The forests comprise dozens of complex ecosystems. They include iconic natural landmarks such as Mt. Whitney, Mono Lake, Lake Tahoe, and giant sequoia trees.
As part of its mandate to manage the national forests, the Forest Service took major steps in the 1990s to improve the ecological health of the Sierras. In November 1998, the Forest Service published a Notice of Intent to prepare an Environmental Impact Statement (“EIS”) analyzing a number of proposed changes to the Forest Plans for the Sierras. The Forest Service cited the need to “improve national forest management direction for five broad problems: (1) conservation of old-forest ecosystems, (2) conservation of aquatic, riparian, and meadow ecosystems, (3) increased risk of fire and fuels buildup, (4) introduction of noxious weeds, and (5) sustaining hardwood forests.”
In 2000, after nearly a decade of study, the Forest Service proposed a number of changes to the Forest Plans to ensure “the ecological sustainability of the entire Sierra Nevada ecosystem and the communities that depend on it.” The Forest Service issued a Draft EIS evaluating eight alternatives for implementing the objectives outlined in the Notice of Intent. Following public comment, scientific review and consultation with other agencies, the Forest Service released a Final EIS in January, 2001.
The 2001 EIS designated the “Modified Alternative 8” as the preferred alternative. In a Record of Decision issued January 12, 2001, the Forest Service adopted this alternative. This is the “2001 Framework.”
The Forest Service received over 200 timely administrative appeals. The Chief of the Forest Service, newly appointed by the incoming administration, did not respond directly to the appeals. Rather, he directed the Regional Forester to reevaluate the 2001 Framework with respect to three fire-related issues. First, the Chief directed him “to re-evaluate the decision for possibilities of more flexibility in aggressive fuels treatment.” Second, he directed him “to re-evaluate the decision based on possible new information associated with the National Fire Plan,” a ten-year strategy developed by Congress, federal agencies, Indian Tribes and western States to restore fire-adapted ecosystem health. Third, he directed him to re-evaluate limitations placed by the 2001 Framework on the Herger-Feinstein Quincy Library Group Forest pilot project dealing with fire prevention.
In December 2001, the Regional Forester appointed an Amendment Review Team. The Regional Forester added non-fire-related issues to the issues identified by the Chief. In addition to the fire-related issues, he asked the Review Team to “identify opportunities” in three areas: first, to “reduce the unintended and adverse impacts [of the 2001 Framework] on grazing permit holders”; second, to “reduce the unintended and adverse impacts [of the 2001 Framework] on recreation users and permit holders”; and, third, to “reduce the unintended and adverse impacts [of the 2001 Framework] on local communities.”
*1017In June 2003, the Forest Service issued a Draft Supplemental EIS, based on the work of the Review Team. The Draft focused on a comparison of two alternatives. “Alternative SI” was the 2001 Framework. “Alternative S2” was the “preferred alternative.” Alternative S2 proposed substantially more logging and associated activities than the 2001 Framework. It also proposed to reduce restrictions on grazing by commercial and recreational livestock.
The Draft was criticized by the staff of the Forest Service’s Washington Office for Watershed, Fish, Wildlife, Air and Rare Plants. The staff wrote a letter complaining that there was no discussion of the effects of the logging and logging-related activities on fish:
Aquatic and Riparian: There needs to be a discussion of the effects of the new alternatives on riparian ecosystems, streams and fisheries. It is not sufficient to dismiss these effects as within the range of impacts discussed in the [2001] framework ... without further analysis, given the activities proposed in Alternative S2. If the treatments [proposed in Alternative S2] will be sufficient to have their intended effect, there is a high likelihood that there will be significant and measurable direct, indirect and cumulative effects on the environment, which need to be analyzed and disclosed in this document.
(Emphasis added.) The letter also raised concerns that the Draft did not adequately analyze the impact of changed grazing standards on riparian environments, streams and fisheries.
The Forest Service issued the 2004 EIS in January 2004 without adding the discussion of “riparian ecosystems, streams and fisheries” that the staff letter had said was needed. The Regional Forester adopted Alternative S2 shortly afterwards in a Record of Decision. Over 6,000 administrative appeals were filed objecting to the Record of Decision. The Forest Service Chief approved the Record Of Decision without change in November 2004. This is the “2004 Framework.”
Both the 2001 and 2004 Frameworks are written in general terms, rather than addressing specific sites at which the logging and logging-related activities will take place. But there are substantial differences between the 2001 and 2004 Frameworks. Relevant to this appeal are changes in authorized logging and logging-related activities, and changes in grazing standards for commercial and recreational livestock.
The most substantial changes are in logging and logging-related activities. The 2004 Framework allows the harvesting of substantially more timber than the 2001 Framework. The 2001 Framework allowed the harvesting of 30 million board feet of salvage timber per year during the Framework’s first and second decades. By contrast, the 2004 Framework allows the harvesting of three times that amount of salvage timber — 90 million board feet per year during its first and second decades. The 2001 Framework allowed the harvesting of 70 million board feet of green timber per year during its first decade and 20 million board feet per year during its second decade. By contrast, the 2004 Framework allows the harvesting of 4.7 and 6.6 times that amount of green timber — 329 million board feet per year during its first decade and 132 million board feet per year during its second decade. The totals for salvage timber for the two decades are 600 million board feet under the 2001 Framework, and 1.8 billion board feet under the 2004 Framework. The totals for green timber for the two decades are 900 million board feet under the 2001 Framework, and 4.6 billion board feet under the 2004 Framework. Stated differently, compared to the 2001 Framework, the 2004 Framework allows the har*1018vesting of an additional 4.9 billion board feet of timber — 1.2 billion board feet of salvage timber and 3.7 billion board feet of green timber — during its first two decades.
The 2004 Framework also allows the harvesting of larger trees than the 2001 Framework. For example, under the 2001 Framework, trees up to 30 inches in breast-height-diameter could be harvested in the wetter west side of the Sierras, but only up to 24 inches in the drier east side. Under the 2004 Framework, trees up to 30 inches in breast-height-diameter can be harvested on both the west and east sides.
The 2004 Framework substantially increases the total acre-age to be logged. Under the 2004 Framework, about 15% fewer acres will be subject to prescribed burns than under the 2001 Framework, but about 250% more acres will be logged “mechanically.” Further, under the 2004 Framework, more logging will be conducted close to streams than under the 2001 Framework. The 2004 EIS states, with more than the usual amount of obfuscating bureaueratese:
The spatial location of strategically placed area treatments1 under Alternatives SI [the 2001 Framework] and S2 [the 2004 Framework] are the same, but they are different than previously considered. For example, analysis in the [2001 EIS] was based on the assumption that the area treatments would be placed2 primarily on the upper two-thirds of slopes, thus minimizing overlap with RCAs3 associated with perennial, intermittent, and ephemeral streams. However, this assumption is no longer valid. Consequently, under Alternatives SI and S2, treatments are not limited to any geographic position.4 As a result, more treatments within RCAs are expected.5 Alternative SI requires that portions of treatment areas be left in an untreated condition.6 It is likely that riparian areas would be priorities for retention to meet this requirement.7 Alternative S2 does not require retention of untreated areas within treatment units so that fire behavior and fire effects are effectively reduced within the entire unit.8 *1019The 2001 Framework limited soil “compaction” in project areas close to streams to 5% of the area, but the 2004 Framework places no limit on “disturbances” in such areas.
The 2004 Framework allows substantially more construction of new, and reconstruction of existing, logging roads than the 2001 Framework. Under the 2001 Framework, 25 miles of new roads were to be constructed, and 655 miles of existing roads were to be reconstructed, during the first decade. Under the 2004 Framework, 115 miles of new roads are to be constructed, and 1,520 miles of existing roads reconstructed, during the first decade. However, under the 2001 Framework, 950 miles of roads were to be decommissioned, compared with 1,175 miles of old roads that are to be decommissioned under the 2004 Framework. The 2004 Framework also allows an additional 215 miles of temporary roads (43 miles of temporary roads per year for 5 years) and slates an additional 3,200 miles of roads for maintenance (640 miles per year for five years).
Finally, grazing restrictions under the 2001 Framework are reduced in the 2004 Framework. Under the 2001 Framework, commercial livestock (cattle and sheep), as well as recreational livestock (pack and saddle stock used by commercial outfitters) were to be excluded from meadows known to be occupied by Yosemite Toads during the toads’ breeding and rearing seasons, as well as from meadows where surveys to determine the presence (or absence) of Yosemite Toads had not yet been performed. The 2004 Framework allows commercial livestock to graze in meadows where surveys to determine the presence of Yosemite Toads have not yet been performed. Further, the 2004 Framework eliminates the categorical exclusion of recreational pack stock and saddle stock from toad-occupied meadows during the breeding and rearing season, and allows managers to develop project-based plans to mitigate effects on the toad.
Other restrictions on grazing have also been reduced. The 2004 Framework divides habitat-protecting restrictions on grazing into several categories based on the adverse impacts on the grazing permit-tee: the greater the adverse impact, the more habitat-protecting effort is required on the part of the permittee. The 2004 EIS describes the effect of the 2004 Framework on 47 grazing permittees (amounting to 11% of the “active allotments”). Under the 2001 Framework, the regulations had “no adverse impact” on any permittee. The regulations had a “low adverse impact” on 11 permittees, a “medium adverse impact” on 17, a “high adverse impact” on 12, and a “very high adverse impact” on 7. Under the 2004 Framework, those numbers are, respectively, 14 (no adverse impact), 7 (low), 10 (medium), 9 (high), and 7 (very high). That is, a total of 14 grazing permittees who had been adversely impacted by habitat-protecting regulations under the 2001 Framework are not adversely impacted at all under the 2004 Framework. For 3 of those 14 permittees, the change effected by the 2004 Framework is to move from a high adverse impact to no impact at all — that is, to move from regulations requiring “substantial” habitat-protective effort by the permittee to regulations requiring no effort whatsoever.
The 2004 EIS predicts that the 2004 Framework will reduce the annual acreage burned by wildfires. Under the 2001 *1020Framework, the estimated annual acreage of wildfires was 64,000 acres during the first decade, and 63,000 acres during the fifth decade. Under the 2004 Framework, the estimated annual acreage of wildfires is 60,000 acres during the first decade, and 49,000 acres during the fifth decade, resulting in a total reduction of 18,000 acres over two decades.
Pacific Rivers filed suit in May 2005, alleging that the 2004 Framework was adopted in violation of NEPA and the APA. On appeal, Pacific Rivers contends that the 2004 EIS fails to take a “hard look” at the environmental impact of the 2004 Framework on fish and amphibians. We conclude that the 2004 EIS does not comply with NEPA with respect to fish, but does comply with respect to amphibians.
II. Standard of Review
We review de novo questions of Article III justiciability, including standing. Porter v. Jones, 319 F.3d 483, 489 (9th Cir.2003). We also review de novo a district court’s decision on summary judgment that an agency complied with NEPA. Or. Natural Desert Ass’n v. Bureau of Land Mgmt., 531 F.3d 1114, 1130 (9th Cir.2008). Judicial review of an agency’s compliance with NEPA is governed by the APA, which requires this court to set aside the agency’s action if it is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” Id. (quoting 5 U.S.C. § 706(2)(A)).
“[W]e will reverse a decision as arbitrary and capricious only if the agency relied on factors Congress did not intend it to consider, ‘entirely failed to consider an important aspect of the problem,’ or offered an explanation ‘that runs counter to the evidence before the agency or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise.’ ” The Lands Council v. McNair (Lands Council II), 537 F.3d 981, 987 (9th Cir.2008) (en banc) (quoting Earth Island Inst. v. U.S. Forest Serv., 442 F.3d 1147, 1156 (9th Cir.2006)), overruled on other grounds by Winter v. Natural Res. Def. Council, 555 U.S. 7, 129 S.Ct. 365, 172 L.Ed.2d 249 (2008).
“In reviewing the adequacy of an EIS, we employ a rule of reason to determine whether the EIS contains a reasonably thorough discussion of the significant aspects of probable environmental consequences.” Kern v. Bureau of Land Mgmt., 284 F.3d 1062, 1071 (9th Cir.2002) (internal quotation marks omitted). “Once an agency has an obligation to prepare an EIS, the scope of the analysis of environmental consequences in that EIS must be appropriate to the action in question.... If it is reasonably possible to analyze the environmental consequences in an EIS ..., the agency is required to perform that analysis.” Id. at 1072.
III. Discussion
A. Standing
The Forest Service argues for the first time on appeal that Pacific Rivers lacks standing under Article III of the Constitution. Questions of Article III jurisdiction can be raised at any time. See Ctr. for Biological Diversity v. Kempthorne, 588 F.3d 701, 707 (9th Cir.2009).
To have standing under Article III, a plaintiff must establish that
(1) it has suffered an “injury in fact” that is (a) concrete and particularized and (b) actual or imminent, not conjectural or hypothetical; (2) the injury is fairly traceable to the challenged action of the defendant; and (3) it is likely, as opposed to merely speculative, that the injury will be redressed by a favorable decision.
*1021Friends of the Earth, Inc. v. Laidlaw Envtl. Sews., 528 U.S. 167, 180-81, 120 S.Ct. 693, 145 L.Ed.2d 610 (2000). To have standing to seek injunctive relief under Article III
a plaintiff must show that he is under threat of suffering “injury in fact” that is concrete and particularized; the threat must be actual and imminent, not conjectural or hypothetical; it must be fairly traceable to the challenged action of the defendant; and it must be likely that a favorable judicial decision will prevent or redress the injury.
Summers v. Earth Island Inst., 555 U.S. 488, 129 S.Ct. 1142, 1149, 173 L.Ed.2d 1 (2009). An organization may sue on behalf of its members
when its members would otherwise have standing to sue in their own right, the interests at stake are germane to the organization’s purpose, and neither the claim asserted nor the relief requested requires the participation of individual members in the lawsuit.
Laidlaw, 528 U.S. at 181, 120 S.Ct. 693.
The Forest Service contends that because Pacific Rivers challenges amendments to Land and Resource Management Plans rather than a specific project under an LRMP, it has failed to allege a threat of a “concrete and particularized” injury that is “actual or imminent.” The Forest Service also contends that Pacific Rivers’ members have not specified which parts of the national forests in the Sierras they use.
The Forest Service relies heavily on the Supreme Court’s decision in Summers. The plaintiffs in Summers challenged nationwide regulations promulgated by the Forest Service that exempted sales of salvage timber of 250 acres or less from NEPA requirements to prepare an EIS or an Environmental Assessment (“EA”). Id. at 1147. The plaintiffs initially challenged a specific sale of salvage timber. After the district court issued a preliminary injunction, the parties settled the dispute over that sale. Id. at 1148. On appeal, both before the Ninth Circuit and before the Supreme Court, the plaintiffs continued to challenge the validity of the exemption for 250 acres or less, though now there was no specific sale at issue. They could make only a general statement that they would visit national forests in the future and might come in contact with a parcel of 250 acres or less on which a salvage-timber sale had been conducted without an EIS or an EA. Id. at 1149-50.
The Supreme Court concluded that there was only a remote chance, “hardly a likelihood,” that such visits would bring plaintiffs into contact with land affected by the challenged regulations. Id. at 1150. The Court noted that the regulation at issue applied to all national forest land (190 million acres) and that the size of the affected parcels was small (250 acres or less). Id. “Accepting an intention to visit the National Forests as adequate to confer standing to challenge any Government action affecting any portion of those forests would be tantamount to eliminating the requirement of concrete, particularized injury in fact.” Id.
One year after Summers, we held in Wilderness Society, Inc. v. Rey, 622 F.3d 1251 (9th Cir.2010), that a plaintiff organization lacked standing to challenge the same nationwide Forest Service regulations at issue in Summers. A member of the plaintiff organization expressed a general intention to return to the Umpqua National Forest for recreational use. Id. at 1256. We held that Summers demands more than a showing of a general intention of returning to a national forest. The member must “show[ ] that he is likely to encounter an affected area of the Umpqua National Forest in his future visits.” Id. (emphasis in original). In addition, the member feared that the Ash Creek Project *1022in the Umpqua National Forest might threaten the plaintiff-organization’s interests in protecting wilderness lands. Id. But the court found this fear insufficient to obtain standing: The member did not “allege that his future enjoyment is in any way threatened by the Ash Creek Project.” Id. (emphasis added).
Summers and Wilderness Society are substantially different from this case. Pacific Rivers introduced into evidence in the district court a declaration of its Chairman, Bob Anderson. Anderson declares that he lives in South Lake Tahoe, that he and his wife own property at Mono Lake, and that they “frequently hike and climb in the Sierra Nevada Range.” Anderson declares further that Pacific Rivers has over 750 members, some of whom live in California. He states:
My first Sierra Nevada backpacking trip was to the Mineral King area in 2000, during which time I also fished. I plan to continue these activities as long as the management of Sierra Nevada national forests does not prevent me from doing so. I have garnered great personal solace in the knowledge that Sierra Nevada native species and the watersheds that support them persist despite over a century’s worth of impacts from grazing, mining, logging, road building, dam construction, and related activities. The same is true for the membership of [Pacific Rivers], many of whom recreate in, fish throughout, and derive much satisfaction from the Sierra Nevada.
He writes specifically with respect to members:
[Pacific Rivers] members participate in recreational activities, such as fishing, hiking, backpacking, cross-county skiing, nature photography, and river and lake boating throughout the Sierra Nevada.
The Forest Service challenged Pacific Rivers’ Article III standing for the first time in this court. If the Forest Service had objected to standing in the district court, Pacific Rivers could easily have supplemented Anderson’s declaration with declarations of individual members who use and enjoy the Sierras, specifying particular national forests and particular patterns of use. Given the timing of the Forest Service’s objection to standing, if we were to hold on the current record that Pacific Rivers has not sufficiently established threats of harm to its members who use the Sierras for recreation, we would remand to the district court to allow further development of the record. But we think such additional development is unnecessary. Anderson has clearly stated that he and a number of Pacific Rivers’ members have used, and will continue to use, the national forests in the Sierras in a variety of places and in a variety of ways.
During the first two decades, the 2004 Framework allows the harvesting of approximately 4.6 billion board feet of green timber and approximately 1.8 billion board feet of salvage timber. This harvesting will take place in every one of the 11 national forests in the Sierras. The smallest amount of green timber harvesting during the two decades — 35 million board feet — will take place in the Lake Tahoe Management Basin. The Lake Tahoe basin is relatively small and is subject to the most intensive recreational use of the 11 national forests covered by the 2004 Framework. Anderson lives in the Lake Tahoe basin. The greatest amount of harvesting — 1.4 billion board feet — will take place in Plumas National Forest. Harvesting in quantities between these two amounts will take place in each of the other nine national forests covered by the 2004 Framework.
Under the 2004 Framework, much of the timber harvesting will be in the upper two thirds of slopes. The result of that harvesting will therefore likely be visible from *1023great distances. Significant timber harvesting will also take place near streams, where recreational users of forests spend much of their time. The 2004 Framework authorizes the construction of 115 miles of new roads and the reconstruction of 1,520 miles of existing roads during the first decade. Grazing restrictions on commercial and recreational livestock will be reduced throughout the Sierras.
There is a concrete connection between the interests of Pacific Rivers’ members in enjoying the forests of the Sierras and the effect of the 2004 Framework. See Lujan v. Defenders of Wildlife, 504 U.S. 555, 562-64, 112 S.Ct. 2130, 119 L.Ed.2d 351 (1992). There is little doubt that members of Pacific Rivers will come into contact with affected areas, and that the implementation of the 2004 Framework will affect their continued use and enjoyment of the forests. By contrast, the regulation at issue in Summers affected only small and widely scattered parcels of land throughout the entire United States, and the plaintiffs had not shown any realistic likelihood that they would come into contact with those parcels.
There are two relevant cases in this circuit, both controlling. In Salmon River Concerned Citizens v. Robertson, 32 F.3d 1346 (9th Cir.1994), we held that an environmental organization had standing to bring a challenge under NEPA to an LRMP that applied to 6 million acres of national forest land in the Sierras. Id. at 1349-55. The challenged LRMP allowed the “use of all methods to treat competing vegetation ... [in order] to meet the timber yield objectives,” and delegated the decision to use herbicides to the district foresters. Id. at 1351.
The Forest Service’s standing argument in Salmon River was essentially the same as its standing argument here — that plaintiff lacked standing because it failed “to demonstrate that the members would be harmed by a specific project using herbicides.” Id. at 1352. Members of the organization lived next to or within the boundaries of the area where herbicides had previously been banned but would now be permitted, and they frequently used the area for recreation. Id. at 1353. These members contended that their health and recreational interests were adversely affected by the Forest Service’s decision to permit herbicide use. Id. We characterized the members’ injury as the risk “that environmental consequences” of herbicide use “might be overlooked[ ] as a result of deficiencies in the government’s analysis under environmental statutes.” Id. at 1355 (internal quotation marks omitted). That risk constituted a concrete, specific and imminent injury sufficient to challenge an EIS because “unfettered use of herbicides ... in the absence of NEPA compliance will cause harm to visitors’ recreational use and enjoyment, if not to their health.” Id.
We specifically held that the plaintiffs did not have to “wait to challenge a specific project when their grievance is with an overall plan.” Id. We explained why:
[I]f the agency action only could be challenged at the site-specific development stage, the underlying programmatic authorization would forever escape review. To the extent that the plan pre-determined the future, it represents a concrete injury that plaintiffs must, at some point, have standing to challenge. That point is now, or it is never.
Id. (quoting Idaho Conservation League v. Mumma, 956 F.2d 1508, 1516 (9th Cir.1992)). See also Res. Ltd., Inc. v. Robertson, 35 F.3d 1300, 1303 (9th Cir.1993) (rejecting, in a challenge to a forest plan, the argument that plaintiffs must “point to the precise area of the park where their injury will occur”).
*1024Another Ninth Circuit panel has recently addressed a separate NEPA challenge to the same 2004 Framework at issue in our case. In Sierra Forest Legacy v. Sherman, 646 F.3d 1161 (9th Cir.2011), we held that an environmental organization had standing to challenge the 2004 Framework. Id. at 1179-80. We noted Ohio Forestry Ass’n v. Sierra Club, 523 U.S. 726, 118 S.Ct. 1665, 140 L.Ed.2d 921 (1998), in which the Supreme Court held that the Sierra Club’s challenge to an LRMP under the National Forest Management Act was unripe as a prudential matter, but did not hold that the Sierra Club lacked Article III standing. Ohio Forestry, 523 U.S. 726, 118 S.Ct. 1665. The Court in Ohio Forestry specifically noted that despite the “considerable legal distance between the adoption of the Plan and the moment when a tree is cut, the Plan’s promulgation nonetheless makes logging more likely in that it is a logging precondition; in its absence logging could not take place.” Id. at 730, 118 S.Ct. 1665. We held in Sierra Forest Legacy that the harm flowing from a failure to comply with NEPA in formulating the 2004 Framework was sufficient to confer standing on plaintiff “to bring a facial NEPA challenge to the 2004 Framework, independent from specific implementing projects.” 646 F.3d at 1179.
We therefore conclude that Pacific Rivers has Article III standing to challenge the 2004 Framework under NEPA.
B. NEPA
“The National Environmental Policy Act has ‘twin aims. First, it places upon [a federal] agency the obligation to consider every significant aspect of the environmental impact of a proposed action. Second, it ensures that the agency will inform the public that it has indeed considered environmental concerns in its decisionmaking process.’” Kern, 284 F.3d at 1066 (quoting Baltimore Gas & Elec. Co. v. Natural Res. Def. Council, Inc., 462 U.S. 87, 97, 103 S.Ct. 2246, 76 L.Ed.2d 437 (1983)) (internal quotations and citations omitted, alteration in original). NEPA is not substantive. It does not require that agencies adopt the most environmentally friendly course of action. Kern, 284 F.3d at 1066. Rather, “[t]he sweeping policy goals ... of NEPA are ... realized through a set of ‘action-forcing’ procedures that require that agencies take a ‘hard look at environmental consequences.’ ” Robertson v. Methow Valley Citizens Council, 490 U.S. 332, 350, 109 S.Ct. 1835, 104 L.Ed.2d 351 (1989) (quoting Kleppe v. Sierra Club, 427 U.S. 390, 410 n. 20, 96 S.Ct. 2718, 49 L.Ed.2d 576 (1976)).
Taking a “hard look” at environmental consequences of major federal actions includes “considering all foreseeable direct and indirect impacts. Furthermore, a ‘hard look’ should involve a discussion of adverse impacts that does not improperly minimize negative side effects.” N. Alaska Envtl. Ctr. v. Kempthorne, 457 F.3d 969, 975 (9th Cir.2006) (internal quotation marks and citations omitted); see also Or. Natural Res. Council Fund v. Brong, 492 F.3d 1120, 1133 (9th Cir.2007) (“[GJeneral statements about possible effects and some risk do not constitute a hard look absent a justification regarding why more definitive information could not be provided.”) (internal quotation marks omitted).
Pacific Rivers alleges that the 2004 EIS does not take a hard look at environmental consequences of the 2004 Framework on fish and amphibians. For the reasons that follow, we agree with Pacific Rivers with respect to fish, but disagree with respect to amphibians.
1. Fish
The 2001 EIS contained a 64-page detailed analysis of environmental consequences of the 2001 Framework for indi*1025vidual species of fish. In stark contrast to the 2001 EIS, the 2004 EIS contains no analysis whatsoever of environmental consequences of the 2004 Framework for individual species of fish. The 2004 EIS incorporates by reference the analysis contained in the 2001 EIS, but contains no analysis of additional or different environmental consequences of the 2004 Framework even though the new framework authorizes substantially more environment-altering activities than the old framework. Of particular importance, the 2004 Framework allows an additional 4.9 billion board feet of green and salvage timber harvesting during the first two decades, much of it conducted nearer streams, compared to the 2001 Framework. The 2004 EIS also incorporates by reference two biological assessments (“BAs”) of the consequences of the 2001 and 2004 Frameworks on listed fish under the Endangered Species Act. But it neither summarizes the findings of the BAs nor includes them in an appendix.
The Forest Service contends that the 2004 EIS takes a sufficiently hard look at environmental consequences of the 2004 Framework on fish. It makes two arguments. First, it points out that the 2004 Framework is an amendment to each of the Forest Plans in the Sierras. The Forest Service argues that because the Forest Plans are LRMPs, it is not reasonably possible for the 2004 EIS to provide an analysis of environmental consequences of the 2004 Framework on individual species. Second, it argues that the 2004 EIS’s incorporation by reference of the BAs concerning environmental consequences of the 2001 and 2004 Frameworks on listed fish satisfies the hard look requirement. We consider these arguments in turn.
a. Level of Required Analysis in the 2004 EIS
During the time frame at issue here, federal law required preparation of an EIS in conjunction with the preparation of a programmatic-level plan such as an LRMP. See 36 C.F.R. § 219.10(b) (1983) (“A draft and final environmental impact statement shall be prepared for the proposed plan according to the NEPA procedures.”); see also 36 C.F.R. § 219.5(a)(2)(i) (2012) (“A new plan or plan revision requires preparation of an environmental impact statement.”). The 2004 Framework is not, in itself, an LRMP; rather, it is an amendment to the LRMPs for the Sierras. Some amendments to LRMPs may be so insignificant that they do not require preparation of an EIS. But the 2004 Framework is a fundamental revision of the Forest Plans in the Sierras. See 36 C.F.R. § 219.10(f) (1983) (requiring Forest Service, when making a “significant change in the plan,” to “follow the same procedure as that required for development and approval of a forest plan”); see also 36 C.F.R. § 219.5(a)(2)(h) (2012) (“The appropriate NEPA documentation for an amendment may be an [EIS], an [EA], or a categorical exclusion, depending on the scope and scale of the amendment and its likely effects.”). The Forest Service does not argue that an EIS was not required. But the Forest Service does argue that, because of the programmatic nature of the 2004 Framework, it was not required in its EIS to perform an analysis of environmental consequences for the individual species of fish.
The required level of analysis in an EIS is different for programmatic and site-specific plans. We wrote in Friends of Yosemite Valley v. Norton, 348 F.3d 789 (9th Cir.2003):
An agency’s planning and management decisions may occur at two distinct administrative levels:
(1) the “programmatic level” at which the [agency] develops alternative management scenarios *1026responsive to public concerns, analyzes the costs, benefits and consequences of each alternative in an [EIS], and adopts an amendable [management] plan to guide management of multiple use resources; and (2) the implementation stage during which individual site specific projects, consistent with the [management] plan, are proposed and assessed.
Ecology Ctr., Inc. v. United States Forest Serv., 192 F.3d 922, 923, n. 2 (9th Cir.1999). An EIS for a programmatic plan ... must provide ‘sufficient detail to foster informed decision-making/ but “site-specific impacts need not be fully evaluated until a critical decision has been made to act on site development.” N. Alaska Envtl. Ctr. v. Lujan, 961 F.2d 886, 890-91 (9th Cir.1992)....
Although NEPA requires that the [agency] evaluate the consequences of its action at an early stage in the project’s planning process, that requirement is tempered by (1) ‘the statutory command that [a reviewing court] focus upon a proposal’s parameters as the agency defines them/ and (2) ‘the preference to defer detailed analysis until a concrete development proposal crystallized the dimensions of a project’s probable environmental consequences.’ [California v. Block, 690 F.2d 753, 761 (9th Cir.1982)].
Id. at 800-01.
Regardless of whether a programmatic or site-specific plan is at issue, NEPA requires that an EIS analyze environmental consequences of a proposed plan as soon as it is “reasonably possible” to do so. Kern, 284 F.3d at 1072. At issue in Kern were two things: an EIS for a Resource Management Plan (“RMP”) for the Coos Bay District in Oregon, and an Environmental Assessment (“EA”) for a site-specific project in that district. The RMP in Kern was a programmatic plan, like the LRMP in the case before us. We wrote:
An agency may not avoid an obligation to analyze in an EIS environmental consequences that foreseeably arise from an RMP merely by saying that the consequences are unclear or will be analyzed later when an EA is prepared for a site-specific program proposed pursuant to an RMP. “[T]he purpose of an [EIS] is to evaluate the possibilities in light of current and contemplated plans and to produce an informed estimate of the environmental consequences.... Drafting an [EIS] necessarily involves some degree of forecasting.” City of Davis v. Coleman, 521 F.2d 661, 676 (9th Cir.1975) (emphasis added). If an agency were to defer analysis ... of environmental consequences in an RMP, based on a promise to perform a comparable analysis in connection with later site-specific projects, no environmental consequences would ever need to be addressed in an EIS at the RMP level if comparable consequences might arise, but on a smaller scale, from a later site-specific action proposed pursuant to the RMP.
Once an agency has an obligation to prepare an EIS, the scope of its analysis of environmental consequences in that EIS must be appropriate to the action in question. NEPA is not designed to postpone analysis of an environmental consequence to the last possible moment. Rather, it is designed to require such analysis as soon as it can reasonably be done. See Save Our Ecosystems v. Clark, 747 F.2d 1240, 1246 n. 9 (9th Cir.1984) (“Reasonable forecasting and speculation is ... implicit in NEPA, and we must reject any attempt by agencies to shirk their responsibilities under NEPA by labeling any and all discussion of future environmental effects as ‘crys*1027tal ball inquiry,’” quoting Scientists’ Inst. for Pub. Info., Inc. v. Atomic Energy Comm’n, 481 F.2d 1079, 1092 (D.C.Cir.1973)). If it is reasonably possible to analyze the environmental consequences in an EIS for an RMP, the agency is required to perform that analysis. The EIS analysis may be more general than a subsequent EA analysis, and it may turn out that a particular environmental consequence must be analyzed in both the EIS and the EA. But an earlier EIS analysis will not have been wasted effort, for it will guide the EA analysis and, to the extent appropriate, permit “tiering” by the EA to the EIS in order to avoid wasteful duplication.
Id. at 1072 (emphasis added). See also 40 C.F.R. § 1501.2 (“Agencies shall integrate the NEPA process with other planning at the earliest possible time to insure that planning and decisions reflect environmental values, to avoid delays later in the process, and to head off potential conflicts.”); New Mexico ex rel. Richardson v. Bur. of Land Mgmt., 565 F.3d 683, 707-08, 716 (10th Cir.2009) (relying on Kern to find NEPA violation with respect to programmatic EIS).
Our dissenting colleague contends that we overruled Kern with respect to programmatic-level plans in our en banc decision in Lands Council II, 537 F.3d 981. We do not believe that Lands Council II overruled the “reasonably possible” requirement of Kern. At issue -in Lands Council II was an EIS for a site-specific project. In our en banc opinion, we specifically overruled Ecology Center, Inc. v. Austin, 430 F.3d 1057 (9th Cir.2005), cert. denied sub nom. Mineral County v. Ecology Ctr., Inc., 549 U.S. 1111, 127 S.Ct. 931, 166 L.Ed.2d 702 (2007). Our holding in Lands Council II was that the analysis in the site-specific EIS at issue was sufficiently supported by studies and on-the-ground analysis. Our opinion nowhere mentioned Kern, nowhere mentioned a programmatic EIS, and nowhere suggested that environmental consequences need not be analyzed in a programmatic EIS if it is “reasonably possible” to perform that analysis.
Nor does the Forest Service believe that Lands Council II overruled the “reasonably possible” requirement of Kern. The Forest Service nowhere contends that we wrongly decided Kern, or that Lands Council II overruled Kern’s “reasonably possible” requirement. The Forest Service recognizes in its brief that Kern requires it to perform reasonably possible analyses of environmental consequences in a programmatic EIS. See Appellee’s Br. at 25 (“Pacific Rivers correctly notes that this Court has held [in Kern ] that a programmatic EIS should analyze environmental consequences where ‘reasonably possible.’ ”). The Forest Service argues under Lands Council II that a court owes deference to its determination of what is reasonably possible because, in its view, “[w]hat scientific analysis is ‘reasonably possible’ at the programmatic stage is a methodological question within the expertise of the agency.” Id. But the Forest Service nowhere argues that it need not comply with Kern.
The 2004 EIS at issue in this case recommends extensive changes to the 2001 Framework and even more extensive changes to the underlying Forest Plans in the Sierras. We have described the principal changes above. Briefly, they include harvesting 4.9 billion more board feet of timber than under the 2001 Framework (6.4 billion more than under the unamended Forest Plans). They include logging and burning near streams that would not have been permitted under the 2001 Framework. They include the construction of 90 more miles of new roads than under the 2001 Framework (115 more *1028miles than under the unamended Forest Plans), and reconstruction of 855 more miles of existing roads than under the 2001 Framework (1,520 more miles than under the unamended Forest Plans). And they include reduction of restrictions on grazing by commercial and recreational stock.
The 2004 EIS contains no analysis whatsoever of environmental consequences of these changes on individual fish species in the Sierra. The 2004 EIS promises, in Section 4.2.3 (“Aquatic, Riparian, and Meadow Ecosystems”), that it will provide such an analysis. The EIS states, “Effects of the alternatives on species dependent on aquatic, riparian, and meadow habitats are explained elsewhere in this [EIS] (Section 4.3.2).” (Emphasis added.). But that promise is not fulfilled. Section 4.3.2 contains 67 pages of analysis of the environmental consequences of the framework for a number of individual species of mammals, birds and amphibians who are dependent on riparian and meadow habitats in the Sierras. But nowhere in that section (or anywhere else in the 2004 EIS) is there any analysis of individual species of fish. The explicit promise to analyze effects “on species dependent on aquatic ... habitats” in Section 4.3.2, and the absence of any such analysis in that section (or anywhere else), is puzzling. It is possible that the absence of the promised analysis is nothing more than a simple mistake. But if a mistake, it was a mistake that was specifically brought to the attention of the Forest Service in the letter written by its Washington staff. As described above, that letter stated, “There needs to be a discussion of the effects of the new alternatives on riparian ecosystems, streams and fisheries.”
In striking contrast to the 2004 EIS, the 2001 EIS contained 64 pages of detailed analysis of environmental consequences of the 2001 Framework on individual fish species. The 2001 EIS devoted 28 pages to individualized analyses of nine “federally threatened and endangered fish species” — the Little Kern Golden Trout, the Paiute Cutthroat Trout, the Lahontan Cutthroat Trout, the Modoc Sucker, the Warner Sucker, the Shortnose and Lost River Suckers, the Central Valley Chinook Salmon, and the Central Valley Steelhead Trout. It then devoted 21 pages to individualized analyses of 11 “sensitive fish species” — the Goose Lake Lamprey, the Fall Run Chinook Salmon, the Eagle Lake Rainbow Trout, the Volcano Creek Golden Trout, the Goose Lake Redband Trout, the Warner Valley Redband Trout, the Goose Lake Sucker, the Lahontan Lake Tui Chub, the Goose Lake Tui Chub, and the Hardhead. Finally, it devoted 13 pages to individualized analyses of 14 “moderate and high vulnerability fish species” — the Kern Brook Lamprey, the Pacific Lamprey, the Kern River Rainbow Trout, the Owens Sucker, the Mountain Sucker, the Eagle Lake Tui Chub, the Pit River Tui Chub, the Sacramento Hitch, the Owens Speckled Dace, the Pit River Roach, the San Joaquin Roach, and the Rough Sculpin.
The 2001 EIS analyzed the environmental consequences to fish of each of the eight alternatives identified in the EIS. See, e.g., 2001 EIS, vol. 3, ch. 3, at 262 (“Timber harvesting may be conducted in riparian areas, following different guidelines, under Alternatives 3, 4, 6, 7, and Modified 8. Alternatives 3 and 5 prohibit road building in riparian zones; Alternative 5 further addresses negative effects of roads on streams by requiring that failed road crossings and culverts be identified and have priority for rehabilitation.”); see also id. at 63, 122 (same). The 2001 EIS also described the environmental consequences of grazing. See, e.g., id. (“One of the greatest risk factors, within the control of the Forest Service, to Forest Service sensitive fish species in the western United *1029States has been the degradation of the aquatic environment, especially those resulting from long term livestock grazing.”); see also id. at 63, 122 (same).
The 2001 EIS also analyzed particular environmental risks for individual species of fish. For example, for both Paiute and Lahontan Cutthroat Trout, “risk factors” included “the immediate loss of individual fish ... specific habitat features such as undercut banks use[d] for cover, increases in sedimentation leading to changes in spawning bed capacity, and the loss of riparian vegetation necessary to maintain adequate temperature regime[s].” For Shortnose and Lost River Suckers, risk factors included “[djecreases in water quality resulting from timber harvest, dredging activities, removal of riparian vegetation, and livestock grazing.” For Central Valley Steelhead, “habitat destruction” was listed as a “risk factor.” The 2001 EIS noted that “timber harvest, road building, agriculture, livestock grazing, and urban development” all “affect[ ] steelhead habitat.” For Volcano Creek Golden Trout, risk factors included “increases in sedimentation leading to changes in spawning bed capacity, and the loss of riparian vegetation necessary to maintain adequate temperature regime. The risk factors identified are primarily a result of historic and current grazing practices.” For Goose Lake Suckers, risk factors included the fact that “many of the streams have experienced some habitat loss due to the effects of logging, grazing and other factors that can degrade watersheds.”
The adequacy of the 2001 EIS with respect to fish is not at issue. What is at issue is the adequacy of the 2004 EIS. Whether or not the analysis in the 2001 EIS was adequate (a question that is not before us), the 2001 EIS shows that an analysis of environmental consequences of the 2004 Framework for individual species of fish was “reasonably possible.” There is no explanation in the 2004 EIS of why it was not reasonably possible to provide any analysis whatsoever of environmental consequence for individual species of fish, when an extensive analysis had been provided in the 2001 EIS. There is also no explanation in the 2004 EIS of why it was “reasonably possible” to provide an extensive analysis of environmental consequences to individual species of mammals, birds, and amphibians in 2004, but not reasonably possible to provide any analysis whatsoever of environmental consequences to individual species of fish in 2004.
An agency has flexibility in deciding when to perform environmental analyses. But an environmental analysis must “provide ‘sufficient detail to foster informed decision-making,’ ” Friends of Yosemite Valley, 348 F.3d at 800 (citation omitted), and so cannot be unreasonably postponed. In 2002, the Council on Environmental Quality (“CEQ”) established a Task Force to review agency practices under NEPA. The Task Force wrote in its September 2003 report to CEQ, “Reliance on programmatic NEPA documents has resulted in public and regulatory agency concern that programmatic NEPA documents often play a ‘shell game’ of when and where deferred issues will be addressed, undermining agency credibility and trust.” The NEPA Task Force, Modernizing NEPA Implementation 39 (2003), available at http://ceq.hss.doe.gov/ ntf/report/frontmats.pdf. An agency’s compliance with the “reasonably possible” requirement in a programmatic EIS, resulting in an appropriate level of environmental analysis, ensures that a “shell game” or the appearance of such a game is avoided. Judicial review under the arbitrary and capricious standard of the Administrative Procedure Act, 5 U.S.C. § 706(2)(A), in turn ensures that an agency does not improperly evade its re*1030sponsibility to perform an environmental analysis when such an analysis is “reasonably possible.”
In some cases, the appropriate level of environmental analysis in a programmatic EIS is fairly debatable. In such cases, our obligation is to defer to the expertise of the agency. But in this case the Forest Service has largely resolved the debate for us. In its 2001 EIS, the Forest Service performed an extensive analysis of the likely environmental impact of the 2001 Framework, including 64 pages of detailed analysis of the likely impact on individual fish species. In stark contrast, the Forest Service performed no analysis whatsoever in its 2004 EIS of the likely impact of the 2004 Framework on fish. The Forest Service provided no analysis despite the fact that the 2004 Framework allows much more logging, burning, road construction, and grazing than the 2001 Fi’amework, and despite the fact that it had provided a detailed analysis in a programmatic EIS only three years earlier.
We do not require the Forest Service to provide in the 2004 EIS precisely the same level of analysis as in its 2001 EIS. We recognize that it may be appropriate to have fewer than 64 pages of detailed analysis of environmental consequences for individual species of fish in the 2004 EIS. Indeed, if the Forest Service had explained its reasons for entirely omitting any analysis of the impact of the 2004 Framework on individual species of fish, it might have been able to show that it is reasonable to postpone such analysis until it makes a site-specific proposal. But the Forest Service has provided no explanation. Compare 40 C.F.R. § 1502.22 (requiring that an agency “always make clear” if it lacks information to conduct environmental analysis). The Forest Service has provided almost the opposite of an explanation, for it promised such an analysis and then failed to provide it. As we noted above, Section 4.2.3. of the 2004 EIS promises an analysis of the “[e]ffects of the alternatives on species dependent on aquatic, riparian, and meadow habitats” in Section 4.3.2. Section 4.3.2 contains a detailed analysis of the environmental effects on individual species of mammals, birds and amphibians. But Section 4.3.2. contains no analysis whatsoever of individual species of fish, even though fish are the quintessential “species dependant on aquatic ... habitat[ ].”
In light of the extensive analysis of the environmental consequences on individual fish species in the 2001 EIS, and of the extensive analysis of the environmental consequences on individual species of mammals, birds, and amphibians in the 2004 EIS, we conclude, contrary to the Forest Service’s contention, that it was “reasonably possible” to provide some analysis of the environmental consequences on individual fish species in the 2004 EIS. The failure of the 2004 EIS to provide any such analysis is a failure to comply with the hard look requirement of NEPA.
b. Incorporation by Reference of the Biological Assessment
The Forest Service’s fail-back argument is that even if an analysis of environmental consequences of the 2004 Framework for individual fish species was “reasonably possible,” the hard-look requirement is satisfied by two Biological Assessments (“BAs”), incorporated by reference in the 2004 EIS. We disagree.
Section 7 of the Endangered Species Act requires a federal agency to consult with the U.S. Fish and Wildlife Service (“USFWS”) if a proposed action by that agency “may affect” a “listed” species or its critical habitat. 16 U.S.C. § 1536(a)(2); 50 C.F.R. § 402.14(a). Pursuant to Section 7, the Forest Service sent two BAs to *1031the USFWS to initiate the consultation process. The first BA, sent in December 2000, indicated that the alternatives considered in the 2001 EIS “may affect” the Little Kern Golden Trout, California Golden Trout, Lahontan Cutthroat Trout, Paiute Cutthroat Trout, Owen’s Tai Chub, Modoc Sucker, Lost River Sucker, Short-nose Sucker, and Warner Sucker. The second BA, sent in July 2003, indicated that the alternatives considered in the 2004 EIS “may affect” all of the species listed in the 2000 BA except the California Golden Trout.
The 2004 EIS does not include the texts of the BAs, but it refers to them twice, once in the text and once in an appendix. First, Section 4.3.1 discusses “Threatened, Endangered, and Proposed Species.” With respect to the BAs, it states, in its entirety:
[T]he biological assessments] for the [2001 EIS] and for the [2004 EIS] contain a more thorough analysis of effects and was [sic] used in evaluating effects on each species. They are hereby incorporated by reference.
The text does not identify the individual species of fish included in the BAs.
Second, Appendix C of the 2004 EIS is a “Consistency Review” that compares the 2001 and 2004 Frameworks to determine whether a supplemental environmental analysis is needed in the 2004 EIS. With respect to “Endangered, Threatened, and Proposed Species” of fish, the Consistency Review concluded:
Implementing the proposed changes considered in the [2004 EIS] would not be expected to produce appreciably different results. Effects on these species are documented in the Biological Assessment for the [2004. EIS], July 30, 2003.
The Appendix identifies the species of fish covered by the 2003 BA.
The Forest Service’s argument fails for three independently sufficient reasons.
First, depending on its nature, material should be in the text of an EIS, should be in an appendix to the EIS, or should be incorporated by reference in the EIS. In descending order of importance: (1) Discussion of significant environmental impacts must appear in the text of an EIS. 40 C.F.R. § 1502.1. (2) Material that
“substantiates any analysis fundamental to the [EIS]” may appear in an appendix. Id. § 1502.18. (3) Material may be incorporated by reference so long as its omission from the EIS does not “imped[e] agency and public review.” Id. § 1502.21; see also Forty Most Asked Questions Concerning CEQ’s National Environmental Policy Act Regulations, 46 Fed.Reg. 18026, 18033-34 (March 17, 1981) (“FAQs”). If the BAs were intended to serve as the analysis of the environmental consequences of the 2004 Framework for fish, the 2004 EIS needed to do more than incorporate them by reference. They should have been described and analyzed in the text of the 2004 EIS, and the BAs themselves should have been included in an appendix.
This is not a mere formality. The purpose of an EIS is to inform decisionmakers and the general public of the environmental consequences of a proposed federal action. That purpose would be defeated if a critical part of the analysis could be omitted from an EIS and its appendices. The EIS is circulated to the general public. “If at all possible,” the appendices are also circulated to the public. Id. at 18034 (FAQ 25a). The material that is incorporated by reference is not circulated to the public; it need only be “made available.” Id. Material that is incorporated by reference must be “briefly described” in the body of the EIS, 40 C.F.R. § 1502.21, but a brief description cannot fulfill the pur*1032pose of the EIS if the substance of what is incorporated is an important part of the environmental analysis.
Second, even if they had been fully described and analyzed in the 2004 EIS, the BAs could not have satisfied the “hard look” requirement. The BAs functioned as a trigger to the consultation process required under Section 7 of the Endangered Species Act. They merely enumerated the several species of “listed” fish that may have, been affected by the alternatives considered in the 2001 and 2004 EISs. There was no analysis in either of the BAs of the manner or degree to which the alternatives may have affected these fish. To the degree that any analysis was performed, it was performed by the Fish and Wildlife Service when it prepared Biological Opinions in response to the BAs. The 2004 EIS makes no reference, in any form, to either of the Biological Opinions.
Third, even if the BAs could have satisfied the hard look requirement, they applied to only one group of fish species. As described above, the 2001 EIS analyzed the environmental consequences for three groups: (1) “federally threatened and endangered fish species” (9 species); (2) “sensitive fish species” (11 species); and (3) “moderate and high vulnerability fish species” (14 species). The BAs analyzed only the individual species in the first group. They said nothing about the individual species in the second and third groups.
2. Amphibians
The 2004 EIS contains an extensive analysis of individual amphibians. It specifically analyzes six species of amphibian: the California Red-legged Frog, the Foothill Yellow-legged Frog, the Mountain Yellow-legged Frog, the Northern Leopard Frog, the Cascades Frog and the Yosemite Toad. For each species, the 2004 EIS identifies changes between the 2001 and the 2004 Frameworks that are likely to affect that species. The 2004 EIS discusses the impact of livestock grazing, prescribed fire, mechanical fuels treatments and road maintenance.
One of the major differences between the 2001 and 2004 Frameworks is the latter’s emphasis on logging, rather than prescribed burning, as a means of reducing the risk of wildfires. The 2004 EIS describes the impact of the changed emphasis on the Foothill Yellow-legged Frog. It states that the 2001 Framework posed some risk to the frog because prescribed burning often results in the destruction or dispersal of coarse woody debris that the frog uses for shelter. By decreasing the amount of prescribed burning, the 2004 Framework will provide some benefit to the frog. However, the 2004 EIS also identifies the use of mechanical logging as a risk. For example, the frogs sometimes seek shelter beneath parked vehicles. When logging operations begin on any particular' day, the vehicles may crush frogs sheltered beneath the tires.
Similarly, the 2004 EIS considers the impact of changed grazing standards on the Yosemite Toad. It states that risk factors to the Yosemite Toad from grazing include
decreased growth rate of tadpoles as a result of increased bacteria from livestock fecal matter; mortality from being-buried by livestock feces; reduced vegetative hiding cover for metamorphs, juveniles, and adults, which increases their vulnerability to predation by snakes and birds; and the collapse of rodent burrows from livestock hoof punching, thereby entrapping or burying individuals that use burrows for hiding cover.
The 2004 EIS notes that allowing grazing in meadows that have not yet been surveyed for Yosemite Toads “may contribute to localized extirpations.”
*1033The 2004 EIS also discusses a number of mitigation strategies to minimize the environmental consequences of the 2004 Framework. For example, the Forest Service will use “Best Management Practices” for road construction and maintenance. These practices include designing stream crossings and replacement stream crossings for a 100-year flood; designing stream crossings to minimize the diversion of natural stream flow; and avoiding road construction in wetlands and meadows. The Forest Service will also continuously monitor grazing allotments if site-specific changes around Yosemite Toad breeding sites are authorized. The 2004 EIS states that such monitoring will allow the Forest Service to identify and mitigate threats to the Yosemite Toad.
Pacific Rivers contends that the Forest Service is required to provide further analysis of the changes that are authorized under the 2004 Framework. Pacific Rivers’ contention stems in part from the Forest Service’s decision under the 2004 Framework to delegate significant decisionmaking authority to local managers of amphibian habitats. For example, in a portion of Section 4.2.3 discussing livestock grazing on meadows, the 2004 EIS notes that the new framework makes changes designed to “allow flexibility to design management practices [to] address local conditions.” However, we are satisfied that the Forest Service’s analysis was sufficient, at this stage of the process, given that the EIS provides significant analysis of the environmental effects on amphibians, and that site-specific projects are not yet at issue.
The Forest Service has repeatedly committed itself to complying with NEPA for site-specific projects that will be proposed under the 2004 Framework. For example, in its brief in this court, it states that “additional NEPA analysis will occur at the project-level.” See Appellee’s Br. at 22. It states, further, that “because on-the-ground activities such as timber harvest and road, skid trail, and log landing construction would not occur prior to a future site-specific decision, the Forest Service will analyze the site-specific effects of those activities before allowing them.” Id. at 24. The brief states, still further, that “[w]hen the Forest Service makes a decision to authorize or reauthorize grazing on an allotment, it conducts a detailed NEPA analysis, where it can examine the effects of the particular proposed grazing, considering ... the allotment’s location ..., [and] the timing, scope, and intensity of proposed grazing.” Id. at 40 (emphasis deleted). The Forest Service makes similar commitments in the 2004 EIS. For example, the 2004 EIS states, “Site-specific decisions will be made on projects in compliance with NEPA ... following applicable public involvement and administrative appeal procedures.” The 2004 EIS states further, “Any site-specific actions taken to implement direction in the Forest Plan Amendment would require compliance with NEPA.” We are confident that when the Forest Service proposes to build, reconstruct or decommission roads; to conduct a logging or a prescribed burning operation for fuels management; to allow pack stock and/or saddle stock into Yosemite Toad-occupied meadows; to permit commercial livestock to graze near Yosemite Toad breeding and rearing sites; or to use pesticides in a riparian conservation area, that it will fully comply with the NEPA requirements applicable to such site-specific projects.
Conclusion
In Lands Council II, we wrote that we will hold that an agency has acted in an arbitrary and capricious manner in preparing an EIS when it has “ ‘entirely failed to consider an important aspect of the problem.’ ” 537 F.3d at 987. In this case, the *1034Forest Service “entirely failed to consider” environmental consequences of the 2004 Framework on individual species of fish. Given the detailed 64-page analysis of the likely impact on individual species of fish in the 2001 EIS, the complete lack of such analysis of the likely impact on individual species of fish in the 2004 EIS, and the lack of any explanation in the 2004 EIS why it is not “reasonably possible” to perform some level of analysis of such impact, we have no choice but to conclude that the Forest Service failed to take the requisite “hard look” at environmental consequences of the 2004 Framework for fish.
We hold that the Forest Service failed to take a hard look at environmental consequences on fish in the 2004 EIS, in violation of NEPA. We hold that the Forest Service did take a hard look at environmental consequences on amphibians in the 2004 EIS, in compliance with NEPA. We therefore reverse in part and affirm in part, and remand to the district court.
REVERSED in part, AFFIRMED in part, and REMANDED. Costs to Plaintiff-Appellant.
. There is no definitions section in the 2004 EIS. From usage in the EIS, it is apparent that “treatments” means logging and/or prescribed burns.
. In standard English, "placed” means “conducted.”
. "RCAs” are Riparian Conservation Areas. See January 2004 Record Of Decision approving the 2004 EIS, at 114 ("riparian conservation area (RCA)”).
In its brief to this court, the Forest Service misstates the meaning of the acronym. It indicates that RCAs are Resource Conservation Areas. See Response Brief at 33 ("Resource Conservation Areas (‘RCAs’)”). In the context of this case, the difference between “riparian” and "resource” is important. "Riparian” is a precise term, meaning something related to the bank of a river, stream, or other body of water. "Resource” is a general term, meaning anything from a natural resource such as trees to a financial resource such as a bank account.
. This sentence is misleading. "Treatments” (i.e., logging and burning) under Alternative SI (the 2001 Framework) are more geographically limited than "treatments” under Alternative S2 (the 2004 Framework).
. This sentence translated into standard English: "As a result, more logging and burning close to streams are expected under the 2004 Framework.”
. This sentence translated into standard English: "The 2001 Framework requires that certain areas not be logged or burned.”
. This sentence translated into standard English: "It is likely that under the 2001 Framework riparian areas would not be logged or burned.”
. This sentence translated into standard English: "The 2004 Framework allows logging and burning close to streams in order to eliminate trees everywhere in a given 'treatment unit' as a means of reducing the risk of fire.”
*1019We remind the Forest Service: "Environmental impact statements shall be written in plain language ... so that decisionmakers and the public can readily understand them. Agencies should employ writers of clear prose or editors to write, review, or edit statements!;.]” 40 C.F.R. § 1502.8.