Opinion by Judge D.W. NELSON; Dissent by Judge O’SCANNLAIN.
D.W. NELSON, Senior Circuit Judge:
Elaine Brong, Oregon State Director of the Bureau of Land Management (“BLM”), and other parties1 appeal the district court’s decision invalidating the Timbered Rock Fire Salvage and Elk Creek Watershed Restoration Project (“Timbered Rock Project” or “Project”), a plan developed by the BLM to log nearly a thousand acres of protected land in southwest Oregon after a major forest fire. The district court held that the Timbered Rock Project violated both the Federal Land Policy and Management Act (“FLPMA”) and the National Environmental Policy Act (“NEPA”). We affirm.
I. FACTUAL & PROCEDURAL BACKGROUND
Following a series of lightning strikes, on July 13, 2002, the Medford District of the BLM was devastated by the “Timbered Rock” fire. This fire burned approximately 12,000 acres of land in the district, all within an area known as the Elk Creek Watershed.2 Under federal law, Elk Creek is a “Late-Successional Reserve,” which entitles the area to heightened environmental protection.
Following the Timbered Rock fire, the BLM began considering a range of options of how to revitalize the Elk Creek area.3 [1124]*1124The BLM considered the environmental impacts of various alternatives, ultimately-devising the Timbered Rock Project. On August 15, 2003, the BLM announced the availability of a Draft Environmental Impact Statement for the Timbered Rock Project, and indicated that it would accept public comment until October 14, 2003. On January 30, 2004, the BLM made public the Project’s Final Environmental Impact Statement (“Timbered Rock FEIS”), and on March 23, 2004, the BLM issued its Record of Decision for the Timbered Rock Project (“Timbered Rock ROD”).
Pursuant to the Project, the BLM proposes to log more than 961 acres of environmentally-protected land affected by the fire. Timbered Rock ROD at 3. Of the 961 acres, 282 are designated as “research units” for investigating the influence of post-fire salvage and salvage intensity on wildfire response, while the remaining acreage is designated for area salvage. Id. As a whole, the Project would allow salvage of approximately 23.4 million board feet of timber to be sold to private companies. Id.
The Oregon Natural Resources Council and other parties (collectively “ONRC”)4 challenged the Timbered Rock Project via an administrative protest filed on April 12, 2004. On May 18, 2004, the BLM responded to ONRC’s protest and affirmed its decision to proceed with the Project. ONRC challenged the agency’s decision in district court. ONRC argued that the BLM violated the Medford District Bureau of Resource Management Plan, as amended by the Northwest Forest Plan, which the BLM is required to follow pursuant to FLPMA. 43 U.S.C. § 1732; 43 C.F.R. § 1610.5-3(a). Specifically, ONRC alleged that the Project violated the Plan because it proposed the excessive removal of large diameter dead or dying trees, impermissible research logging, and timber removal in “non-suitable woodlands.” ONRC also alleged that the BLM failed to designate properly certain areas as “riparian reserves.”
ONRC also alleged the Project violated NEPA-because (1) the BLM failed to analyze the cumulative effects of fire suppression activities, private salvage logging, and salvage logging in deferred watersheds, and (2) the BLM employed a flawed methodology by using an unreliable tool, known as the Decayed Wood Advisor (“DecAID”), to calculate the effect of the Project on certain species.5
On June 15, 2004, the district court granted ONRC’s motion for a temporary restraining order. On November 10, 2004, the district court entered an opinion and order in favor of ONRC, and on November 23, 2004, it entered a judgment granting ONRC a permanent injunction. See Or. Natural Res. Council Fund v. Brong, No. Civ. 04-693-AA, 2004 WL 2554575 (D.Or. Nov.8, 2004). The BLM timely appealed.
II. STANDARD OF REVIEW
We review the BLM’s compliance with FLPMA and NEPA de novo. See Or. Natural Res. Council v. U.S. Bureau of Land Mgmt., 470 F.3d 818, 820 (9th Cir.2006); Klamath Siskiyou Wildlands Ctr. v. Boody, 468 F.3d 549, 554 (9th Cir.2006). Decisions that allegedly violate NEPA and FLPMA are reviewed under the Administrative Procedure Act [1125]*1125(“APA”), which “dictates that we should ‘hold unlawful and set aside agency action ... [that is] arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.’ ” Natural Res. Def. Council v. Nat’l Marine Fisheries Serv., 421 F.3d 872, 877 (9th Cir.2005) (quoting 5 U.S.C. § 706(2)(A)).
While the APA requires that we not substitute our own judgment for that of the agency, it nevertheless requires us to “engage in a substantial inquiry” and a “thorough, probing, in-depth review.” Native Ecosystems Council v. U.S. Forest Serv., 418 F.3d 953, 960 (9th Cir.2005) (quoting Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 415-16, 91 S.Ct. 814, 28 L.Ed.2d 136 (1971)). As we have said before, “[t]o have not acted in an arbitrary and capricious manner, the agency must present a ‘rational connection between the facts found and the conclusions made.’ ” Id. (quoting Nat’l Wildlife Fed’n v. U.S. Army Corps of Eng’rs, 384 F.3d 1163, 1170 (9th Cir.2004)). Though we normally afford deference to an administrative agency’s interpretation of its own regulations, “an agency’s interpretation ‘does not control, where ... it is plainly inconsistent with the regulation at issue.’ ” Id. (quoting Friends of Southeast’s Future v. Morrison, 153 F.3d 1059, 1069 (9th Cir.1998)).
III. THE FLPMA CLAIMS
The Federal Land Policy & Management Act, 43 U.S.C. §§ 1701-1785 (2006), establishes requirements for land use planning on public land. FLPMA requires that the BLM, under the Secretary of the Interior, “develop, maintain, and when appropriate, revise land use plans” to ensure that land management be conducted “on the basis of multiple use and sustained yield.” 43 U.S.C. §§ 1701(a)(7), 1712(a); see also Kern v. U.S. Bureau of Land Mgmt., 284 F.3d 1062, 1067 (9th Cir.2002) (holding that FLPMA “requires the BLM to prepare [resource management plans] for the various districts under its control”). The process for developing, maintaining, and revising resource management plans is controlled by federal regulations at 43 C.F.R. §§ 1601.0-1610.8 (2006). Once a land use plan is developed, “[a]ll future resource management authorizations and actions ... shall conform to the approved plan.” 43 C.F.R. § 1610.5-3(a).
The land use plan governing the Timbered Rock Project is the Medford District Resource Management Plan (“Medford RMP”), as amended significantly by the Northwest Forest Plan (“NFP” or “Plan”).6 The BLM interpreted the NFP as permitting the Timbered Rock Project. For the following reasons, however, we disagree.
A. The BLM’s Interpretation of the Northwest Forest Plan is Inconsistent with the Plan’s Mandate to Prioritize the Maintenance and Preservation of Late-Successional Ecosystems.
Because the NFP embodies the substantive management directives with which the BLM must comply under FLPMA, our review must start with, and remain anchored in, an understanding of the NFP. A careful reading shows that while the NFP as a whole seeks to strike a balance between environmental protection and resource extraction, its management directives for specified reserve areas give priority to environmental concerns. The [1126]*1126BLM’s interpretation of the Plan is plainly inconsistent with these directives.
Consider the NFP’s history. The Plan is a comprehensive response to a long and bitter legal battle over the scope of logging in old-growth forests, home to the endangered northern spotted owl. See Seattle Audubon Soc’y v. Lyons, 871 F.Supp. 1291, 1300-01 (W.D.Wa.1994), aff'd, Seattle Audubon Soc’y v. Moseley, 80 F.3d 1401 (9th Cir.1996) (per curiam). Indeed, “[i]t should be borne in mind that the NFP is not an ordinary government land-management strategy; instead, the history and care in its creation bespeak the massive effort that led to its birth.” Gifford Pinchot Task Force v. U.S. Fish and Wildlife Serv., 378 F.3d 1059, 1068 (9th Cir.2004).
To ensure that national forest timber sales would comply with legal conservation requirements, the NFP divided the approximately 24.5 million acres of federal land within the northern spotted owl’s range into several hierarchical allocations designated by the type of land use in each allocation. See id.; Northwest Forest Plan Standards and Guidelines (“NFP S & G”)7 at A-1-A-7, B-1. This hierarchy is the fundamental means by which the NFP achieves its primary goal of protecting and enhancing habitat for late-successional and old-growth forest-related species. See Seattle Audubon Soc’y, 871 F.Supp. at 1304-05; NFP S & G at A-1. Six of the allocations are “reserve areas in which logging and other ground-disturbing activities are generally prohibited.” Seattle Audubon Soc’y, 871 F.Supp. at 1304-05; see also NFP S & G at A-4-A-5 (summarizing what activities are permitted within each classification). The Plan designates the remaining “unreserved areas as ‘matrix,’ in which timber harvest may go forward subject to environmental requirements.” Seattle Audubon Soc’y, 871 F.Supp. at 1305.
As explained in Section I, supra, the Timbered Rock Project would permit logging in the Elk Creek Watershed (“Elk Creek”). Elk Creek is designated as a Late-Successional Reserve (“LSR”)—a protected area—under the NFP. LSRs lie at the heart of the NFP’s ecosystem-based conservation strategy for the northern spotted owl and other endangered species. “The objective of Late-Succes-sional Reserves is to protect and enhance conditions of late-successional and old-growth forest ecosystems, which serve as habitat for late-successional and old-growth related species including the northern spotted owl.” NFP S & G at C-9. The NFP plainly states that LSRs “are to be managed to protect and enhance conditions of late-successional and old-growth forest ecosystems.” NFP S & G at C—11; see also Northwest Forest Plan Record of Decision (“NFP ROD”)8 at 8 (“Late-successional reserves are to be managed to protect and enhance old-growth forest conditions.”).
Pursuant to these goals, the NFP makes programmed “stand management” activities, such as logging, impermissible in LSRs. See NFP ROD at 8 (“No programmed timber harvest is allowed inside the reserves.”). The NFP recognizes a narrow exception, however, following a stand-disturbing event, such as a massive fire. In these specific circumstances, the [1127]*1127Plan contains “Guidelines for Salvage”9 that prescribe the extent to which limited logging within the LSRs is permitted. NFP S & G at C-13-C-16. We must determine whether the Timbered Rock Project’s salvage plan is consistent with these guidelines.
Of crucial importance, however, is that the NFP does not in any way relax its LSR management goals for salvage operations. Instead, the LSR salvage guidelines reiterate that “[bjecause Late-Suc-cessional Reserves have been established to provide high quality habitat for species associated with late-successional forest conditions, management following a stand-replacing event should be designed to accelerate or not impede the development of those conditions.” Id. at C-14; see also NFP ROD at 8 (“Salvage guidelines are intended to prevent negative effects on late-successional habitat.”). Moreover, the guidelines state that “[wjhile priority should be given to salvage in areas where it will have a positive effect on late-succes-sional forest habitat, salvage operations should not diminish habitat suitability now or in the future.” Id. at C-13. Thus, while it permits salvage logging in limited circumstances, the NFP clearly prioritizes the preservation of LSR ecosystems over commercial benefits.
The BLM’s interpretation of the LSR salvage guidelines is inconsistent with the NFP’s clear direction. The BLM construes the guidelines as balancing environmental concerns and economic factors equally. To be sure, the Forest Service may consider economic interests in choosing how it will conduct LSR salvage operations; that it may do so is not only a matter of common sense, but it is also something explicitly contemplated by the Plan. See NFP S & G at C-13-C-14 (“[Mjanagement planning for Late-Sucees-sional Reserves must acknowledge the considerable value of retaining dead and dying trees in the forest as well as the benefits from salvage activities.”). However, the NFP does not permit a salvage project in an LSR for the purpose of recovering the economic value of timber without at least explaining—in the administrative record—how such action is compatible with the NFP’s direction to protect and enhance late-successional ecosystems. Not only would doing so run afoul of the NFP’s clear priority of protecting LSR ecosystems, but it would contradict the Plan’s directive:
In all cases, planning for salvage should focus on long-range objectives, which are based on desired future condition of the forest. Because Late-Successional Reserves have been established to provide high quality habitat for species associated with late-successional forest conditions, management following a stand-replacing event should be designed to accelerate or not impede the development of those conditions.
Id. at C-14.
In sum, “[sjalvage activities must be intended to prevent negative effects on late-successional habitat,” NFP ROD at 63, and the BLM’s interpretation is inconsistent with this directive. Therefore, it is not entitled to deference, Native Ecosystems, 418 F.3d at 960, and we must independently evaluate the Project to determine whether its specific elements comply with the NFP.
B. The Timbered Rock Project Violates the NFP.
ONRC challenges four components of the Timbered Rock Project, two of which [1128]*1128warrant attention here: snag retention and research logging.10 Because the BLM relied on an erroneous interpretation of the NFP’s managing directives for LSRs when developing the Project, it is not surprising that both of these components are inconsistent with the Plan and, consequently, violate FLPMA.
1. Snag Retention
As we explained in Section I, supra, the Timbered Rock Project proposes substantial salvage logging in the acreage affected by the fire in an effort to recover economic value from the timber therein. The salvage would remove from the Elk Creek LSR a significant number of large, standing dead or dying trees, known as “snags.”11 This is significant because snags play an integral role in the ecology of old-growth forests. Indeed, the NFP expressly states:
Tree mortality is an important and natural process within a forest ecosystem. Diseased and damaged trees and logs are key structural components of late-successional and old-growth forests. Salvage of dead trees affects the development of future stands and habitat quality for a number of organisms. Snag removal may result in long-term influences on forest stands because large snags are not produced in natural stands until trees become large and begin to die from natural mortality. Snags are used extensively by cavity-nesting birds and mammals such ,. as woodpeckers, nuthatches, chickadees, squirrels, red tree voles, and American marten. Removal of snags following disturbance can reduce the carrying capacity for these species for many years.
NFP S & G at B-8; see also id. at B-9 (“[T]rees injured by disturbance may develop cavities, deformed crowns, and limbs which are habitat components for a variety of wildlife species.”).
Given the importance of snags in late-successional ecosystems, it is not surprising the NFP restricts removal of snags in LSRs. In fact, the NFP’s salvage guideline no. 3 expressly limits the removal of such snags:
Snags provide a variety of habitat benefits for a variety of wildlife species associated with late-successional forests. Accordingly, following stand-replacing disturbance, management should focus on retaining snags that are likely to persist until late-successional conditions have developed and the new stand is again producing large snags. Late-successional conditions are not associated with stands less than 80 years old.
Id. at C-14 (emphasis added). Despite this clear directive against removing large snags (i.e., those likely to persist until late-successional conditions have developed), the BLM asserts that the Project, which entails the removal of a significant number of large snags in late-successional areas, is nonetheless consistent with the NFP. We disagree, and find the BLM’s reasoning lacking in multiple respects.
The BLM asserts that removing a significant number of large snags is consistent with the NFP because leaving some large snags—between eight and twelve per acre in the nonresearch units—would provide sufficient habitat for species that rely upon large snags for survival (i.e., a “some-is-enough” standard). The BLM [1129]*1129supports this conclusion with data produced by the DecAID Wood Advisor. De-cAID is a newly-developed model designed to “help managers evaluate effects! ] of forest conditions and existing or proposed management activities on organisms that use snags and down wood.” Timbered Rock FEIS at D-16. However, “DecAID does not specifically address effects of fire,” id. at D-17, and it “is not intended to predict occurrence of -wildlife species at the scale of individual forest stands or specific locations,” id. at D-19.
ONRC argues that the BLM’s use of DecAID is arbitrary and capricious,12 yet even if we assume, arguendo, that using DecAID is permissible, the BLM’s argument that the Project, in employing the model’s recommended snag-retention levels, complies with the NFP still fails for two principal reasons. First, the BLM can point to no part of the NFP to support its argument that using the some-is-enough standard satisfies the Plan. In light of the Plan’s clear directive against removing snags that will remain until the late-successional forest regenerates, this is not surprising. Indeed, the salvage guidelines expressly state that snags likely to persist “until late-successional conditions have developed and the new stand is again producing large snags” should not be removed. NFP S & G at C-14.13
Moreover, the importance of developing and retaining snags is emphasized in other parts of the NFP. For example, when discussing the role of silviculture in preserving LSRs, the Plan states that the “development of old-growth forest characteristics including snags” is a “principal objective.” NFP S & G at B-5. The Plan similarly states that “[d]esired late-successional and old-growth characteristics that will be created as younger stands change through successional development include ... moderate-to-high accumulations of large logs and snags.” Id. Again, the NFP’s emphasis on retaining snags in LSRs is not surprising given the Plan’s clear prioritization of preserving LSR ecosystems, see Section III.A, supra, and the key role snags play in that process.
Second, the amount of large snag retention the BLM claims to be “enough” to satisfy the NFP is only achieved by averaging salvaged and non-salvaged areas together across all the acres included in the logging. The Timbered Rock FEIS states:
The snags would be concentrated in portions of the units that receive no harvest. Approximately 1,004 acres would be included as harvest units, with 679 acres receiving harvest of all fire-killed trees and the remaining 325 acres retaining all trees, accounting for the 8-12 snags per acre.
Timbered Rock FEIS at 3-112. Thus, the BLM’s representation that between eight and twelve large snags per acre will [1130]*1130still be standing after the logging occurs is grossly misleading, as over two-thirds of the affected acreage will be completely stripped of all salvageable trees.
The BLM’s attempt to dilute the effects of its proposed activities by averaging the snag retention over such a wide area is inconsistent with the NFP and improper under our precedent. See Pac. Coast Fed’n of Fishermen’s Ass’ns v. Nat’l Marine Fisheries Serv., 265 F.3d 1028, 1035-37 (9th Cir.2001) (holding that an agency cannot try to “minimize” the environmental impact of an activity by simply adopting a scale of analysis so broad that it marginalizes the site-level impact of the activity on ecosystem health). In fact, if using such an approach was permitted, the Project could clear-cut all 1,004 acres and still claim to be retaining eight to twelve snags per acre by merely expanding the Project to “include” more land.14 Clearly, this would be unreasonable, as any adverse environmental effect could be “diluted to insignificance.”15 Id. at 1036.
Furthermore, even if the DecAID recommendations for snag density happen to approximate the density actually required to sustain late-successional forests, the BLM offers no evidence that a wide-scale averaging approach is compatible with these recommendations. The BLM notes it relied on two studies that recommended leaving snags in clumps rather than scattered across the landscape. The BLM does not explain, however, how it achieves the desired distribution of clumps by leaving roughly thirty large snags per acre in less than one-third of the land covered by the Project while clear-cutting the remaining lion’s share.
The BLM cites a specific passage in the NFP to support its argument that salvaging large snags at the level proposed in the Project is consistent with the Plan. The passage states:
Salvage is not required to be beneficial, but is designed to permit the recovery of timber volume in those instances where catastrophic events clearly kill more trees (resulting in more snags and down logs in the short and long term) than are needed to maintain late-successional conditions. For example, if a major blowdown event leaves dead trees 15 feet deep over the landscape, a determination could be made that only a portion of those logs are needed to meet the objectives of the reserve. The rest ... might be available for salvage.
NFP ROD at 66. Thus, under certain limited circumstances, salvage can occur in LSRs; indeed, we noted this in Section III.A, supra. However, the BLM does not claim or offer evidence to demonstrate that the Timbered Rock Fire “clearly kill[ed] more trees than are needed to maintain late-successional conditions,” or that only a portion of dead trees “are needed to meet the objectives 'of the reserve.” Instead, the BLM relies on its some-is-enough argument without making the threshold findings required by the NFP.
The BLM also cites the NFP’s statement that “some commercial wood volume removal” is permitted in LSRs. See NFP S & G at C-13. But again, the NFP clearly states that salvaging should be minimal, [1131]*1131that environmental concerns ought to take priority over potential commercial benefits, and that large snags should be retained so as to ensure the development and preservation of late-successional habitat. Despite these numerous mandates emphasizing that logging snags should not harm LSRs, the BLM neglects to explain how the Timbered Rock Project avoids doing just that.
In sum, we require an agency to “present a rational connection between the facts found and the conclusions made.” Native Ecosystems, 418 F.3d at 960 (internal quotation marks omitted). The BLM’s decision to preserve a baseline number of snags is insufficient in a fundamental way: it neglects to explain why the snag removal it does authorize, which undisputably harms late-successional habitat in the short term, will somehow maintain overall habitat suitability now or in the future, as expressly required by the NFP. See NFP S & G at B-8, C-14. Consequently, the Timbered Rock Project violates FLPMA.
2. Research Logging
ONRC also challenges the Project’s proposed research logging, which would occur on 282 acres of land affected by the fire within the Elk Creek LSR.16 The NFP establishes narrow guidelines identifying the circumstances in which research is permissible in LSRs, similar to the guidelines regarding salvage logging. Specifically, the Plan states:
A variety of wildlife and other research activities may be ongoing and proposed in late-successional habitat. These activities must be assessed to determine if they are consistent with Late-Succes-sional Reserve objectives. Some activities (including those with experimental forests) not otherwise consistent with the objectives may be appropriate, particularly if the activities will test critical assumptions of these standards and guidelines, will produce results important for habitat development, or if the activities represent continuation of long-term research. These activities should only be considered if there are no equivalent opportunities outside Late-Sucees-sional Reserves.
NFP S & G C-18. Thus, the NFP creates a two-step inquiry for assessing whether research activities are permitted in an LSR.
First, the research activity is examined to determine whether it is “consistent with Late-Súccéssional Reserve objectives.” One of the NFP’s stated goals for LSRs is “the development of old-growth forest characteristics including snags.” NFP S & G at B-5. However, the research logging proposed in the Project would itself result in the removal of hundreds (if not thousands) of snags, a fundamental component of LSRs. Moreover, the BLM states it is attempting to ensure that two-thirds of the dead trees in all size classes greater than twenty inches dbh would be maintained, but the BLM presents no evidence to support a claim that removing the remaining one-third would in fact assist with the “development of old-growth forest characteristics.” Thus, for the same reasons the Project falls short with regard to snag retention, it is also lacking with regard to its proposed research activities.
Under the second step of the inquiry, even if the proposed activities are inconsistent with these objectives, they might still be permitted if they “test -critical assumptions ..., produce results important for [1132]*1132habitat development, or ... represent continuation of long-term research” and there are “no equivalent opportunities outside Late-Successional Reserves.” The NFP mandates that while “[a]n important component of [the Plan] is the facilitation of research activities to gather information and test hypotheses in a range of environmental conditions, ... every effort should be made to locate non-conforming [research] activities in land allocations where they will have the least adverse effect upon the objectives of the applicable standards and guidelines.” NFP ROD at 15. The BLM has failed to satisfy this requirement.
Even if the BLM is correct that the Project would test various assumptions in the NFP,17 it has provided no evidence that equivalent opportunities are unavailable in non-LSR areas. What is more, although the Project’s “research proposals are related - to post-fire conditions and must be conducted in a recently burned area,” the BLM concedes “a number of these areas exist within southwest Oregon,” and the BLM makes little more than the conclusory statement that “conducting this research in an LSR is appropriate” in supporting its proposed research projects. Timbered Rock FEIS at 1-12. Thus, the research proposed under the Project fails to comply with the NFP’s guidelines.
In sum, the BLM’s interpretation of the Plan is plainly inconsistent with the NFP’s directives regarding Late-Successional Reserves. As a result, the snag removal and research activities proposed in the Timbered Rock Project are at odds with the Plan and, consequently, violate FLPMA.
IV. THE NEPA CLAIMS
In addition to its claims under FLPMA, ONRC also alleges the Timbered Rock Project violates the National Environmental Policy Act. NEPA requires agencies considering “major Federal actions significantly affecting the quality of the human environment” to prepare and issue an environmental impact statement. 42 U.S.C. § 4332(2)(C) (2006); Nw. Envtl. Advocates v. Nat’l Marine Fisheries Serv., 460 F.3d 1125, 1133 (9th Cir.2006). The statement “shall provide full and fair discussion of significant environmental impacts and shall inform decisionmakers and the public of the reasonable alternatives which would avoid or minimize adverse impacts or enhance the quality of the human environment.” 40 C.F.R. § 1502.1; Nw. Envtl. Advocates, 460 F.3d at 1134. Thus, it is more than a mere “disclosure document.” 40 C.F.R. § 1502.1. Our job in reviewing an EIS “is to ensure that the agency has taken a ‘hard look’ at the potential environmental consequences of the proposed action.” Klamath-Siskiyou Wildlands Ctr. v. Bureau of Land Mgmt., 387 F.3d 989, 993 (9th Cir.2004) (citing Churchill County v. Norton, 276 F.3d 1060, 1072 (9th Cir.2001)). By focusing agency and public attention on the environmental effects of proposed agency action, “NEPA ensures that the agency will not act on -incomplete information, only to regret its decision after it is too late to correct.” Marsh v. Or. Natural Res. Council, 490 U.S. 360, 371, 109 S.Ct. 1851, 104 L.Ed.2d 377 (1989).
One of the specific requirements under NEPA is that an agency must consider the effects of the proposed action in the context of all relevant circumstances, such that where “several actions have a cumula[1133]*1133tive ... environmental effect, this consequence must be considered in an EIS.” Neighbors of Cuddy Mountain v. U.S. Forest Serv., 137 F.3d 1372, 1378 (9th Cir.1998) (quoting City of Tenakee Springs v. Clough, 915 F.2d 1308, 1312 (9th Cir.1990)). A cumulative effect is “the impact on the environment which results from the incremental impact of the action when added to other past, present, and reasonably foreseeable future actions regardless of what agency (Federal or non-Federal) or person undertakes such other actions.”18 40 C.F.R. § 1508.7.
ONRC asserts that the BLM failed to analyze in the FEIS the cumulative impact of the Timbered Rock Project and three other categories of activity-fire suppression efforts, salvage logging on interspersed private lands, and salvage logging on deferred watersheds. Our cases firmly establish that a cumulative effects analysis “must be more than perfunctory; it must provide a useful analysis of the cumulative impacts of past, present, and future projects.” Klamath-Siskiyou, 387 F.3d at 994 (emphasis added) (quoting Ocean Advocates v. U.S. Army Corps of Eng’rs, 361 F.3d 1108, 1128 (9th Cir.2004)). To this end, we have recently noted two critical features of a cumulative effects analysis. First, it must not only describe related projects but also enumerate the environmental effects of those projects. See Lands Council v. Powell, 395 F.3d 1019, 1028 (9th Cir.2005) (holding a cumulative effects analysis violated NEPA because it failed to provide “adequate data of the time, place, and scale” and did not explain in detail “how different project plans and harvest methods affected the environment”). Second, it must consider the interaction of multiple activities and cannot focus exclusively on the environmental impacts of an individual project. See Klamath-Siskiyou, 387 F.3d at 996 (finding a cumulative effects analysis inadequate when “it only considers the effects of the very project at issue” and does not “take into account the combined effects that can be expected as a result of undertaking” multiple projects).
We are not persuaded that the Timbered Rock FEIS includes an adequate discussion of the direct effects of fire suppression activities or salvage logging on private lands. The FEIS mentions that fire suppression efforts increased erosion and sedimentation, that they might have increased fish mortality, and that private logging has had an impact on fish, aquatic insects, and the accumulation of woody debris-but this is not enough. The BLM must do more than merely state that past projects contributed to environmental harms. See Lands Council, 395 F.3d at 1027-28. In Lands Council, we found a cumulative effects analysis inadequate because it “should have provided adequate data of the time, type, place, and scale of past timber harvest and should have explained in sufficient detail how different project plans and harvest methods affected the environment.” 19 Id. at 1028.
The efforts to suppress the Timbered Rock Fire involved the construction of nearly thirty-three miles of new firelines and the aerial application of nearly 40,000 gallons of chemical retardant. After the [1134]*1134fire, Boise Corporation logged nearly 6,000 acres of private land interspersed with the Elk Creek LSR, requiring new roads to be constructed to reach the private holdings. Although the Timbered Rock FEIS recites these facts, it does not contain the level of detailed explanation required by Lands Council.20 The FEIS contains a table summarizing cumulative effects in broad terms, but it does not offer quantified or detailed data about these effects. As we have observed on multiple occasions, “general 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.” Klamath-Siskiyou, 387 F.3d at 993-94 (quoting Ocean Advocates, 361 F.3d at 1128). Even if the BLM was unable to indicate with any great degree of certainty the results of the Project, because the cumulative effects analysis requires an agency to predict future conditions, uncertainty is an inherent part of the process. Therefore, a general statement about uncertainty does not satisfy the procedural requirement that an agency take a hard look at the environmental effects of an action. The BLM can certainly explain specific projections with reference to uncertainty; however, it may not rely on a statement of uncertainty to avoid even attempting the requisite analysis.
Finally, the BLM’s analysis of cumulative effects in deferred watersheds presents a distinct problem. Deferred watersheds are areas identified in the Medford RMP as already sustaining significant cumulative effects. In light of these effects, the RMP defers further management activities in these areas for a period of time, although it notes that salvage and other limited activities “could be permitted if the effects will not increase the cumulative effects.” Medford RMP Record of Decision at 42. By virtue of this specific direction, the BLM is required to analyze the cumulative effects of additional logging in the five deferred watersheds in the Timbered Rock Project area. This analysis appears nowhere in the FEIS. While the FEIS does mention background facts about deferred watersheds, it fails to consider what the likely impact on the watersheds will be when the environmental impacts of the Timbered Rock Project are added to the pre-existing deteriorated state of the watersheds, which led to the initial “deferred” designation. Thus, as in Lands Council, the FEIS is inadequate because “there is no discussion of the connection between individual harvests and the prior environmental harms from those harvests that the [agency] now acknowledges.” Lands Council, 395 F.3d at 1027.
In sum, .the BLM failed to analyze the impact of the Timbered Rock Project when combined with previous fire suppression efforts, salvage logging on interspersed [1135]*1135private lands, and salvage logging on deferred watersheds. Because the BLM is required to consider these effects, Neighbors of Cuddy Mountain, 137 F.3d at 1378, by failing to do so it did not take the requisite “hard look” at the impact the Project would have on the environment, Klamath-Siskiyou, 387 F.3d at 993. Consequently, the BLM violated NEPA.
Y. CONCLUSION
For the foregoing reasons, we hold that in developing the Timbered Rock Project, the BLM violated (1) the NFP and, consequently, FLPMA, and (2) NEPA. Because the APA “dictates that we should ‘hold unlawful and set aside agency action ... not in accordance with law,’ ” Natural Res. Def. Council, 421 F.3d at 877 (quoting 5 U.S.C. § 706(2)(A)), we affirm the district court’s decision to enjoin the Project from going forward.
AFFIRMED.