Earth Island Institute and the Center for Biological Diversity appeal the district court denial of their request for a preliminary injunction against implementation of a United States Forest Service restoration project involving two timber sales in the Sierra Nevada mountains. We reverse and remand.

* In late August 2001, a large wildfire began in the Sierra Nevada mountains and quickly swept through two adjacent national forests as well as some private lands. By the time it was extinguished in September, the "Star Fire" had consumed thousands of acres in both the Eldorado National Forest and the Tahoe National Forest. Over the course of the next year, Forest Service personnel responsible for each forest developed and implemented management plans responding to the blaze. This appeal focuses on the Star Fire Restoration Project, which covered the Eldorado forest.

In the immediate aftermath of the fire, the Forest Service prepared an initial Burned Area Emergency Rehabilitation (BAER) report, which estimated that some 11% of the burned area in the Eldorado forest had experienced high fire intensity, 57% moderate intensity, and 32% low intensity. Subsequently, a more intensive survey prepared in connection with the Environmental Impact Statement (EIS) estimated that 35% of the area had experienced high intensity burns, 45% moderate intensity, and 18% low intensity, leaving 2% unburned.1

In March 2002, the Forest Service released a draft EIS that proposed logging 1,714 acres of the Eldorado Forest using tractor, skyline, and helicopter methods. Following public comment, the Forest Service issued a Final Environmental Impact Statement (FEIS) in June that again recommended this option. The plan aimed to prevent another "stand replacing" fire by removing most of the dead trees and preventing the development of excessive woody debris. A "stand" is "an easily defined area of the forest that is relatively uniform in species composition or age and can be managed as a single unit." The plan also sought to prevent soil erosion by promoting appropriate ground cover, to preserve some dead trees (often referred to as "snags") and down logs for the use of dependent animal species, and to maximize the monetary value of dead trees by acting quickly to allow logging. Revenue from the timber sale would finance other elements of the plan.

According to the FEIS, 71% of the project area had experienced high severity burn. The report relied upon the mortality guidelines developed by Sherri Smith, a Forest Service entomologist. Based on cambium sampling2 of partially green trees and her review of both the scientific literature and some Forest Service data, Smith determined that trees with 35% green canopy or less were effectively dead; the remaining green leaves or needles would eventually turn brown and fall off. Thus, trees demonstrating either less than 35% live crown or three dead cambium samples would be considered dead and marked for removal.3 The preferred alternative in the FEIS called for the retention of approximately four to eight large dead trees per acre and the removal of the remainder in any general forest areas with greater than 50% tree mortality and in Old Forest Emphasis Areas (OFEAs) with greater than 75% mortality. The guidelines placed no limit on the diameter of tree to be removed.

As one of the eleven national forests in the Sierra Nevada range, the Eldorado forest remains subject to the Sierra Nevada Framework, a comprehensive forest plan published in January 2001. The Framework established a comprehensive conservation strategy for all national forests in the area, including special limitations on logging in OFEAs and specific protection zones designed to preserve the habitat of the California spotted owl.4 According to the Framework, the Forest Service must establish 300-acre Protected Activity Centers (PACs) around all known or suspected spotted owl nesting sites. In these areas, logging is severely restricted, generally to the reduction of surface and ladder fuels.5 The Framework provides that the PACs must be maintained "regardless of California spotted owl occupancy status, unless habitat is rendered unsuitable by a catastrophic stand-replacing event and surveys conducted to protocol confirm non-occupancy."

In addition, the Framework requires 1000-acre Home Range Core Areas (HRCAs) around each PAC in the Eldorado forest. These areas are supposed to encompass "the best available spotted owl habitat in the closest proximity to the owl PACs where the most concentrated owl foraging activity is likely to occur." Within these areas, which are managed with the same restrictions as OFEAs, trees of more than 12" in diameter generally may not be removed, though in some circumstances slightly larger trees may be thinned. Following a "stand-replacing event," the Framework calls for the retention of "all snags 15 inches or greater ... except to address imminent hazards to human safety." However, in these circumstances the Framework also permits the removal of dead trees "to the extent that project analysis recommends removal to benefit landscape conditions for old forest structure and function." If fire renders the PAC unsuitable, the Forest Service must attempt to relocate the PAC within the HRCA.

A Forest Service team surveyed the two Eldorado PACs within the Star Fire area, PAC055 and PAC075.6 Using Smith's mortality estimates, the report concluded that approximately 4% of PAC055 and approximately 13% of PAC075 remained at less than 75% tree mortality per acre. In addition, the report estimated that within the Star Fire area, only 5% of the PAC055 HRCA and 18% of the PAC075 HCRA remained at less than 75% mortality. As a result, the FEIS concluded that adjustment of PAC boundaries was impossible. The report recommended dropping both PACs from the forest plan.

In August 2002, Forest Supervisor John Berry adopted the recommendation of the FEIS but modified the preferred alternative so that no trees with green canopy would be removed from partially burned stands within the former PACs. He further determined that cambium sampling should not be used because of its relative inefficiency (though he also noted that this decision would satisfy any public concern over potential damage to trees from the sampling). Plaintiffs appealed this decision administratively, and the agency affirmed the decision in September. Drawing upon an earlier recommendation by the Regional Forester, the final administrative appeal extended the restriction on removing trees with any green canopy to all trees within the OFEAs. The Forest Service then divided the project area into two timber salvage sales and awarded the contracts to Sierra Pacific Industries.

On September 27, 2002, two environmental organizations challenged the timber sales in federal district court and requested a preliminary injunction against their implementation, arguing that the Forest Service had violated the National Environmental Policy Act (NEPA), 42 U.S.C. § 4321, et seq., and the National Forest Management Act (NFMA), 16 U.S.C. § 1600 et seq. Because logging was scheduled to begin immediately, Plaintiffs also requested a temporary restraining order on October 1. On October 3, the district court issued a temporary injunction against logging trees with any green canopy remaining. The court also permitted Sierra Pacific to intervene on behalf of the Forest Service.

On October 11, the district court denied Plaintiffs' request for a preliminary injunction. The court determined that the Plaintiffs were unlikely to succeed on their challenges to the Forest Service methodology and data because agencies are entitled to rely upon their own methodology and experts. The court found that the Forest Service had taken a "hard look" at the environmental issues raised by the restoration project and had not abused its discretion in preparing a separate EIS for each national forest. Finally, the court concluded that the Plaintiffs had failed to establish that the timber sales would result in irreparable harm to the California spotted owl or that the "balance of hardships" tipped in their favor. This appeal followed.

Earth Island seeks a preliminary injunction barring implementation of the Star Fire timber sales until their NEPA and NFMA claims are adjudicated. The Ninth Circuit has described two sets of criteria for preliminary injunctive relief. Under the "traditional" criteria, a plaintiff must show "(1) a strong likelihood of success on the merits, (2) the possibility of irreparable injury to plaintiff if preliminary relief is not granted, (3) a balance of hardships favoring the plaintiff, and (4) advancement of the public interest (in certain cases)." Johnson v. Cal. State Bd. of Accountancy, 72 F.3d 1427, 1430 (9th Cir. 1995). Alternatively, a court may grant the injunction if the plaintiff "demonstrates either a combination of probable success on the merits and the possibility of irreparable injury or that serious questions are raised and the balance of hardships tips sharply in his favor." Id. (citations omitted) (emphasis in original). "These two alternatives represent `extremes of a single continuum,' rather than two separate tests." Clear Channel Outdoor, Inc. v. City of Los Angeles, 340 F.3d 810 (9th Cir.2003). "Thus, the greater the relative hardship to [the party seeking the preliminary injunction,] the less probability of success must be shown." Id. (citation omitted) (alteration in original).

In general, we review the denial of a preliminary injunction for abuse of discretion. Bay Area Addiction Research & Treatment, Inc. v. City of Antioch, 179 F.3d 725, 730 (9th Cir.1999). The district court, however, "necessarily abuses its discretion when it bases its decision on an erroneous legal standard or on clearly erroneous findings of fact." Rucker v. Davis, 237 F.3d 1113, 1118 (9th Cir.2001) (en banc), rev'd on other grounds, Dep't of Hous. & Urban Dev. v. Rucker, 535 U.S. 125, 122 S. Ct. 1230, 152 L. Ed. 2d 258 (2002). When the district court is alleged to have relied on an erroneous legal premise, we review the underlying issues of law de novo. Does 1-5 v. Chandler, 83 F.3d 1150, 1152 (9th Cir.2002).

The scope of our review is generally limited to whether the district court employed the proper preliminary injunction standard and whether the court correctly apprehended the underlying legal issues in the case. Rucker, 237 F.3d at 1118. "We typically will not reach the merits of a case when reviewing a preliminary injunction.... By this we mean we will not second guess whether the court correctly applied the law to the facts of the case, which may be largely undeveloped at the early stages of litigation. As long as the district court got the law right, it will not be reversed simply because the appellate court would have arrived at a different result if it had applied the law to the facts of the case." Id. (citations and internal quotation marks omitted).

In this case, the district court applied an improper legal standard when assessing the "possibility of irreparable injury." Although the court began with an accurate recitation of the legal standard, it ultimately found that the Plaintiffs "failed to establish that the project will result in actual harm to the California spotted owl as opposed to speculation that some such harm could possibly occur." The court further noted that the Plaintiffs "failed to show that measures already in place ... will not afford sufficient protection" and that the Plaintiffs "failed to identify any concrete probability of irreparable harm in any other respect."

Each of these statements places a higher burden of proof on the plaintiff than is warranted. A preliminary injunction only requires plaintiffs to show probable success on the merits and the possibility of irreparable harm. As discussed below, the probability of success in this case is not so remote as to render the irreparable injury component irrelevant. Cf. Arcamuzi v. Continental Air Lines, Inc., 819 F.2d 935, 937 (9th Cir.1987) ("These two formulations represent two points on a sliding scale in which the required degree of irreparable harm increases as the probability of success decreases. If the plaintiff shows no chance of success on the merits, however, the injunction should not issue.") (internal citations and quotation marks omitted).

Moreover, we have often held that a Forest Service logging plan may, in some circumstances, fulfill the irreparable injury criterion because of the long term environmental consequences. See, e.g., Idaho Sporting Cong., Inc. v. Alexander, 222 F.3d 562, 569 (9th Cir.2000) (noting that the imminent and continuing logging activities presented "evidence of environmental harm ... sufficient to tip the balance in favor of injunctive relief"); Neighbors of Cuddy Mountain v. U.S. Forest Serv., 137 F.3d 1372, 1382 (9th Cir.1998) ("The old growth forests plaintiffs seek to protect would, if cut, take hundreds of years to reproduce.") (citation omitted). As we recognized in Idaho Sporting Congress, 222 F.3d at 569, the Supreme Court has held that insufficient evaluation of environmental impact under NEPA does not create a presumption of irreparable injury. However, the Supreme Court has also recognized that "[e]nvironmental injury, by its nature, can seldom be adequately remedied by money damages and is often permanent or at least of long duration, i.e., irreparable. If such injury is sufficiently likely, therefore, the balance of harms will usually favor the issuance of an injunction to protect the environment." Amoco Production Co. v. Village of Gambell, AK, 480 U.S. 531, 545, 107 S. Ct. 1396, 94 L. Ed. 2d 542 (1987). The irreparable injury question in this case, therefore, turns on the likelihood of that injury occurring.

In dismissing the likelihood of injury to the spotted owl, the district court focused on the Forest Service decision to prohibit logging of trees with any green canopy within the former PACs and the OFEAs. Plaintiffs, however, directed much of their attention upon trees with green canopy outside the PACs that are scheduled to be logged even under the modified plan. At the preliminary hearing, their attorney argued that:

The project area is 1400 acres. The only place where they are not removing green trees is in the 600 acres of PACs. That's it. Everywhere else they are taking green trees. That is why we're here because we feel at the very minimum that the Home Range Core Areas need to be protected as well. Because that's the only thing that gives the owl a fighting chance. They use their PAC for 50 percent of their foraging, and they use the Home Range Core Area for the other 50 percent. Even if you have this restriction in the PACs for not cutting any green trees, owls that are going to come back there ... that owl isn't going to survive.

Hearing Transcript at 45-46. There is no indication that the district court considered this possibility in its assessment of irreparable harm. Nor does the record indicate the exact amount of forest with green canopy that exists within the HRCA but outside of the OFEA or the PAC.7 Under these circumstances, the court's conclusion that Plaintiffs failed to show "actual harm" or a "concrete probability of irreparable harm" constitutes the application of an "erroneous legal standard" and thus "necessarily" an abuse of discretion. Rucker, 237 F.3d at 1118.

Even if the district court erred in its assessment of irreparable harm, remand might not be necessary if the Plaintiffs showed no possibility of success on the merits. We must therefore examine the district court's determination that the Plaintiffs have not shown any likelihood of success on their NEPA and NFMA claims. Moreover, the degree of irreparable harm required for a preliminary injunction increases as the probability of success on the merits decreases, and vice versa. See Idaho Sporting Cong., 222 F.3d at 565; see also Kootenai Tribe of Idaho v. Veneman, 313 F.3d 1094, 1124 (9th Cir.2002) ("If plaintiffs had demonstrated a strong likelihood of success on the merits, then plaintiffs would have needed only to make a minimal showing of harm to justify the preliminary injunction."). As a result, our assessment of the likelihood of success will also play a role in the district court's reevaluation of irreparable harm.

Earth Island raises a number of challenges to the Star Fire plan based upon both NEPA and the NFMA. NEPA requires that a federal agency "consider every significant aspect of the environmental impact of a proposed action... [and] inform the public that it has indeed considered environmental concerns in its decisionmaking process." Kern v. U.S. Bureau of Land Mgmt., 284 F.3d 1062, 1066 (9th Cir.2002) (quoting Baltimore Gas & Elec. Co. v. Natural Res. Def. Council, 462 U.S. 87, 97, 103 S. Ct. 2246, 76 L. Ed. 2d 437 (1983)). In order to accomplish this, NEPA imposes procedural requirements designed to force agencies to take a "hard look" at environmental consequences. Id. The NFMA requires the Forest Service to develop comprehensive forest management plans for each unit in the National Forest System. 16 U.S.C. § 1604(a). Subsequent "plans, permits, contracts, and other instruments for the use and occupancy" of the forests must be consistent with the governing forest plan, in this case the Sierra Nevada Framework. 16 U.S.C. § 1604(I); see also Wilderness Soc'y v. Thomas, 188 F.3d 1130, 1132 (9th Cir. 1999). Judicial review of agency decisions under both statutes is governed by the Administrative Procedure Act, which specifies that an agency action may only be overturned when it is "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law." 5 U.S.C. § 706(2)(a); see also Native Ecosystems Council v. Dombeck, 304 F.3d 886, 891-92 (9th Cir.2002).

* Plaintiffs first challenge the Forest Service conclusion that logging large trees reduces the potential damage caused by a future fire. They claim that the relevant scientific studies focus solely on smaller trees, while the few studies discussing larger trees ultimately advocate leaving most of them in place. Plaintiffs also argue that the Forest Service deliberately used scientifically questionable mortality standards to overestimate the level of tree destruction. These standards allowed the agency to remove owl habitat protections and permit more logging than the Framework would otherwise allow. According to the Plaintiffs, the scientific literature overwhelmingly suggests that a large percentage of trees categorized as dead under Smith's criteria would survive. They also suggest that Smith's own data contradict her conclusions and that the Forest Service ignored contradictory data contained in more recent studies, especially a 2001 study by Drs. Stephens and Finney. In addition, Plaintiffs offer expert evidence of many healthy trees on the ground more than one year after the fire, contradicting USFS predictions that trees with 65% or more crown scorch will not survive.

NEPA requires that "the public receive the underlying environmental data from which a Forest Service expert derived her opinion." Idaho Sporting Cong. v. Thomas, 137 F.3d 1146, 1150 (9th Cir.1988). An agency must also "identify any methodologies used" and "make explicit reference by footnote to the scientific and other sources relied upon for conclusions in the [EIS]." 40 C.F.R. § 1502.24. Failure to provide this information "either vitiates a plaintiff's ability to challenge an agency action or results in the courts second guessing an agency's scientific conclusions." Idaho Sporting Cong., 137 F.3d at 1150. However, an agency is entitled to wide discretion in assessing the scientific evidence, so long as it takes a hard look at the issues and responds to reasonable opposing viewpoints. See 40 C.F.R. § 1502.9(a)-(b). Because analysis of scientific data requires a high level of technical expertise, courts must defer to the informed discretion of the responsible federal agencies. Marsh v. Ore. Natural Res. Council, 490 U.S. 360, 377, 109 S. Ct. 1851, 104 L. Ed. 2d 377 (1989). "When specialists express conflicting views, an agency must have discretion to rely on the reasonable opinions of its own experts, even if a court may find contrary views more persuasive." Id. at 378, 109 S. Ct. 1851. At the same time, courts must independently review the record in order to satisfy themselves that the agency has made a reasoned decision based on its evaluation of the evidence. Id.

In this case, the Forest Service explicitly drew upon the existing literature when estimating the size and potential damage of various levels of future fuel loads. The agency concluded that higher numbers of larger downed logs would increase fire intensity and make fire control much more difficult. The FEIS discussed this decision at length and provided both data from the Star Fire area and detailed projections of future snag falls. The FEIS also directly responded to comments raising questions about the removal of larger snags. Based on this record, we cannot say that the Forest Service acted arbitrarily and capriciously in its assessment.

Similarly, the FEIS discusses at length the development of Smith's criteria for determining tree mortality, explaining that Smith and her colleagues merely updated a well-known set of tree mortality guidelines based on experience with recent fires. The FEIS also describes the relationship between Smith's criteria and other published findings (such as the fact that she recommends linear measurement of crown damage rather than volume measurement and thus arrives at different numbers) and the reasons for distinguishing apparently conflicting studies cited in public comments. Finally, the FEIS explains that the Forest Service had requested but not received the Stephens & Finney data and that one of the authors had noted privately that the study dealt with prescribed fires rather than post-fire salvage criteria. The Forest Service thus provided a reasoned explanation for its decision not to incorporate this data, which is all that NEPA requires. Even if we found Plaintiffs' interpretation of the overall scientific evidence more persuasive, that would not be enough to declare the alternative agency interpretation arbitrary and capricious.

However, Plaintiffs also charge that the USFS data is factually incorrect. That is, regardless of the specific methodology adopted for analyzing the data, Plaintiffs claim that the agency data simply does not match the reality on the ground. If true, this fact would do more than challenge the USFS methodology or its conclusion that partially green trees will likely die and thus may be counted as already dead. It would suggest that the USFS experts had relied upon factually inaccurate data or that FEIS did not follow the methodology it claimed to follow.

Plaintiffs present evidence that Monica Bond, a spotted owl scientist, found much higher tree canopy readings (using a densiometer) than the FEIS describes, and that she discovered large numbers of green trees where the FEIS specifically states that no green foliage exists. Similarly, Plaintiffs claim that Dr. Ed Royce used the Forest Service's own methodology and arrived at radically different results than those reported in the FEIS. Finally, the Plaintiffs argue that the sale area maps, issued after the final agency decision had been made, reclassify many "severely burned" areas as "cut tree" areas, meaning that the logging company may only cut marked trees (as opposed to leaving only marked trees). Plaintiffs claim that areas are only marked as "cut tree" when they have experienced mild or moderate burns, thus suggesting that the Forest Service had implicitly acknowledged a lower burn level.

At this stage, the record does not allow us to conclude that the Forest Service acted arbitrary and capriciously in relying on its own data and discounting the alternative evidence offered by the Plaintiffs. The agency plausibly explained that apparent differences in burn severity estimates across surveys and maps stemmed from the different levels of specificity and different document purposes. To the extent that the agency reasonably relied upon mortality estimates obtained in compliance with the Smith methodology and reasonably discounted alternative evidence as unreliable or preliminary, the Forest Service is entitled to use the data it collected.

We note, however, that if Plaintiffs are able to convince the district court that the agency unreasonably relied upon inaccurate data, they may be able to succeed on the merits of this claim. In Rybachek v. United States EPA, 904 F.2d 1276, 1297-98 (9th Cir.1990), we reviewed a claim that the agency had falsified data and ignored the results of other tests that would have undermined its analysis. We reviewed the data and the methodology and determined that the agency had used reasonable assumptions to estimate undetectable pollution levels and had reasonably disregarded an anomalous test result. Id. In Idaho Sporting Congress v. Rittenhouse, 305 F.3d 957, 972 (9th Cir.2002), by contrast, we invalidated the Forest Service's use of habitat as a proxy for population trends in part because its methodology produced obviously inaccurate habitat numbers. Because a claim of factual accuracy differs from an attack on the methodology itself, the district court should revisit this argument on remand. See 40 C.F.R. § 1502.24 ("Agencies shall insure the professional integrity, including the scientific integrity, of the discussions and analyses in environmental impact statements.")

Earth Island also argues that three separate Forest Service decisions violate the Sierra Nevada Framework and thus the NFMA: the logging of trees over 20" in diameter, the de-listing of the PACs within the Eldorado Forest, and the failure to readjust those PAC boundaries. On the first point, the agency plausibly concluded that the 20" limitation applied to undergrowth thinning in live forest stands rather than salvage logging after a severe fire. Thus, the mere fact that the Star Fire Plan allows logging of some larger trees does not constitute a violation of the Framework. Plaintiffs respond that even if this general forest limitation does not apply to severely burned areas, it should still apply to the moderately and mildly burned areas included in the sale areas. However, nothing in the Framework suggests that the diameter limitation must be maintained in certain portions of a general forest area when it has been legitimately lifted in other portions the same area and the Forest Service has produced a comprehensive plan for the entire area.

As for the PAC claims, the Framework calls for maintenance of PACs "regardless of California spotted owl occupancy status, unless habitat is rendered unsuitable by a catastrophic stand-replacing event and surveys conducted to protocol confirm non-occupancy." Earth Island must show that the USFS failed to meet these two criteria in order to establish probability of success on its claim that the USFS violated the Framework agreement.

Although the USFS characterization of this intense fire as a stand-replacing event appears reasonable, Plaintiffs raise considerable doubts as to whether the government can meet the second condition necessary to remove restrictions on logging. The government explains in great detail why the habitat appears unsuitable to spotted owl populations, largely relying upon the percentage of dead or dying trees in the area as determined under the Smith methodology. Unfortunately for the agency position, demographic surveys did not "confirm non-occupancy" of PAC075. In fact, the surveys actually turned up an owl pair. According to the FEIS, "[s]urveys for 2002 began in April. Survey of the area near the historic activity center for PAC075 has located the same banded male and a new sub-adult female near the Middle Fork American River." Plaintiffs also supply their own evidence of owl occupancy, as owl expert Monica Boyd discovered a feather, whitewash, and pellets in August 2002.8 Thus, the agency appears to have failed the second Framework condition for lifting the restrictions on logging within the PAC.

In response, the Forest Service argues that because the habitat was unsuitable, surveys were unnecessary. The Record of Decision confirms that the general determination of unsuitable habitat, rather than any occupancy survey, drove the decision to remove PAC protections. The administrative appeal acknowledges that wildlife animals, including the spotted owl, "occasionally return[] to a previously occupied home range that has been severely degraded by a stochastic, stand-destroying event such as a wild fire." The appeal concludes, however, that "this behavior does not warrant a change in the current regional definition of suitable habitat." In essence, while Plaintiffs point to the actual presence of owls, the Forest Service simply offers reasons to classify the habitat as unsuitable.9

As authority for the proposition that surveys are not always necessary despite the Framework language, the Forest Service points to a comment posted on the agency website in response to a question in October 2001:

If the entire PAC (and much of the home range core area) has experienced a high intensity burn (70 to 80 percent mortality) and there are obviously no green trees or very few green trees remaining, there is little need to survey to protocol. The probability of occupancy is extremely low. However, if there are portions of the PAC that have remained green, the remaining suitable habitat within the defined PAC and adjacent home range core area should be surveyed to determine occupancy. There is a high probability that individual birds or pairs will merely shift use patterns and relocate the activity center. A lost PAC does not always lead to the abandonment of a home range core area.

Regional Forester memorandum, Question E5-6 (Oct. 11, 2001). Even under the Smith methodology, the Forest Service admits that some 13% of the original PAC experienced less than 75% mortality. This suggests that "portions of the PAC ... have remained green," triggering the need for occupancy surveys under the October memorandum.

More importantly, even accepting the general approach outlined in the memorandum, in which post-catastrophe surveys are only required within "suitable" habitat areas, these surveys seem intended to find out whether or not owls are still in the area despite the catastrophe. If such surveys confirm that owls still reside within the area, the restrictions under the Framework should remain in place. After all, the Framework calls for maintenance of PACs "unless habitat is rendered unsuitable... and surveys confirm non-occupancy" (emphasis added). In this case, the Forest Service appears to treat post-catastrophe surveys as suitable for informational purposes only. According to this approach, the agency may independently determine that the habitat is unsuitable and proceed with declassification of PACs and HRCAs while ignoring any contrary information about occupancy it obtains from the surveys it still conducts. This procedure does not seem to be a plausible interpretation of the quoted passage or of the Framework in general.

A reviewing court is not permitted to substitute its own preferred interpretation of a forest plan for that of the Forest Service. Forest Guardians v. U.S. Forest Serv., 329 F.3d 1089, 1098 (9th Cir.2003). Instead, the agency's decision can be overturned only if the scheme is plainly erroneous or inconsistent with the forest plan. Id. In this case, given the unambiguous language in the Framework and the most plausible interpretation of the agency's earlier comments, we cannot agree that the Forest Service fully complied with the forest plan when it ignored the confirmed presence of owls and de-listed PAC075.10

Plaintiffs also argue that under NEPA the Forest Service was required to prepare a single EIS covering the Star Fire restoration projects and timber sales in both the Eldorado and Tahoe national forests. The Supreme Court acknowledged that NEPA may require a comprehensive impact statement when several concurrent proposals have a cumulative or synergistic impact. Kleppe v. Sierra Club, 427 U.S. 390, 409, 96 S. Ct. 2718, 49 L. Ed. 2d 576 (1976). As we recently reiterated, "[a] single NEPA review document is required for distinct projects when there is a single proposal governing the projects or when the projects are connected, cumulative, or similar actions under the regulations implementing NEPA." Native Ecosystems, 304 F.3d at 893-94 (citations and internal quotation marks omitted).

In this case, because there is no comprehensive plan covering both forests, Plaintiffs may only prevail by showing that the separate actions are "connected, cumulative or similar," and that the Forest Service acted arbitrarily and capriciously in failing to prepare a single, comprehensive EIS. Id. (citing 40 C.F.R. § 1508.25).11 Although federal agencies have considerable discretion to define the scope of NEPA review, some actions must be considered together to prevent an agency from "dividing a project into multiple `actions,' each of which individually has an insignificant environmental impact, but which collectively have a substantial impact." Thomas v. Peterson, 753 F.2d 754, 758 (9th Cir.1985).

The Ninth Circuit applies an "independent utility" test to determine whether actions are "connected" for purposes of NEPA review. See Native Ecosystems, 304 F.3d at 894. ("Where each of two projects would have taken place with or without the other, each has `independent utility' and the two are not considered connected actions."). As in Native Ecosystems, the two restoration projects in this case have independent utility in that they each generate revenue and implement distinct forest conservation measures, and each plan would go forward without the other. Id. Thus, the two projects do not meet the Ninth Circuit test for connected actions.

Because the two projects may have "cumulatively significant impacts" on spotted owl habitat, they might instead fall under the definition of "cumulative actions" in 40 C.F.R. § 1508.25(a)(2). We have required the Forest Service to prepare a single EIS for multiple timber sales when they formed part of a single timber salvage project, were announced simultaneously, were reasonably foreseeable, and were located in the same watershed. Blue Mountains Biodiversity Project v. Blackwood, 161 F.3d 1208, 1214-15 (9th Cir.1998). Many of the same factors do not seem to apply here. Unlike the multiple timber sales in Blue Mountain, the boundary between sale areas in this case predated the agency decision, and the sales and analyses proceeded on separate time schedules. Moreover, nothing in the record suggests that the agency intended to segment review to minimize cumulative impact analysis. Churchill County v. Norton, 276 F.3d 1060, 1079-80 (9th Cir.2001). Rather, the FEIS explicitly discusses the cumulative impact of many elements of Tahoe's corresponding Red Star Restoration Project. Although the regulations might support a decision to proceed with a comprehensive EIS for "cumulative actions," we cannot say that the Forest Service acted arbitrarily under this circuit's case law in deciding not to do so.

Plaintiffs insist that the two projects instead meet the definition of "similar actions" in 40 C.F.R. § 1508.25(a)(3): "actions which, when viewed with other reasonably foreseeable or proposed agency actions, have similarities that provide a basis for evaluating their environmental consequences together, such as common timing or geography." We agree that the many similarities in underlying cause, proposed solution, and general geography place the post-fire sales in the Tahoe and Eldorado national forests into this category.

Plaintiffs insist that "similar" actions must be considered together within a single EIS, a position echoed by language found in at least two Ninth Circuit cases. See Northwest Res. Info. Ctr. v. Nat'l Marine Fisheries Serv., 56 F.3d 1060, 1067 (9th Cir.1995) ("an agency is required to consider more than one action in a single EIS if they are `connected actions,' `cumulative actions,' or `similar actions'") (citing 40 C.F.R. § 1508.25); see also Native Ecosystems, 304 F.3d at 893-94. However, the regulations actually accord the agency more deference for "similar actions" than the other two categories. Compare 40 C.F.R. § 1508.25(a)(1) ("Connected actions... should be discussed in the same impact statement") and § 1508.25(a)(2) ("Cumulative actions ... should therefore be discussed in the same impact statement") with § 1508.25(a)(3) ("An agency may wish to analyze [similar] actions in the same impact statement. It should do so when the best way to assess adequately the combined impacts of similar actions or reasonable alternatives to such actions is to treat them in a single impact statement.") (emphasis added). Neither case actually involved a claim about "similar actions," nor did either decision analyze that particular regulatory provision. See Native Ecosystems, 304 F.3d at 894-95; Northwest Res. Info. Ctr., 56 F.3d. at 1067-68. Moreover, several other Ninth Circuit cases describe the regulations as requiring the preparation of a comprehensive EIS only for cumulative and connected actions. See Churchill County, 276 F.3d at 1076; Wetlands Action Network v. U.S. Army Corps of Eng'rs, 222 F.3d 1105, 1118 (9th Cir. 2000); Thomas v. Peterson, 753 F.2d at 758-59.

After comparing the clear language of the regulations with the isolated sentence contained in two cases that do not discuss "similar" actions, we are forced to conclude that the agency was not required to prepare a single EIS for these actions, at least under the circumstances presented here: prior administrative boundaries, different patterns of ownership and destruction, disparate timetables, and separate supervisory personnel. Given these factors, the agency may have reasonably concluded that two separate review documents constituted "the best way to assess adequately the combined impacts of similar actions." 40 C.F.R. § 1508.25(a)(3). We cannot say that the Forest Service acted arbitrarily.

Finally, Plaintiffs argue that the Eldorado study failed to consider adequately the cumulative impact of foreseeable actions in the Tahoe forest. Even if a single, comprehensive EIS is not required, the agency must still adequately analyze the cumulative effects of the projects within each individual EIS. See Native Ecosystems, 304 F.3d at 896 n. 2 (differentiating cumulative impact analysis from the discretionary determination of the EIS scope). Cumulative impact involves 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." 40 C.F.R. § 1508.7; see also Kern, 284 F.3d at 1075 (discussing this analysis). The government points to the many sections of "cumulative impact" analysis in the FEIS. However, the reviewing court needs to explore the specific parameters of this analysis; the fact that such a section exists is not enough. See Native Ecosystems, 304 F.3d at 896 (noting that the Environmental Assessment had a cumulative impact section but that it failed to discuss the road density amendments at issue).

In this case, the Plaintiffs suggest that the Eldorado Forest FEIS failed to consider adequately the cumulative impact on spotted owl PAC075 within the Tahoe National Forest. Because spotted owls depend upon the surrounding HRCA in addition to the PCA, the destruction of the Eldordo HRCA might significantly impact those owls, even if they "reside" exclusively within Tahoe.

Although the FEIS discusses cumulative impacts on several occasions, only twice does the FEIS mention the role of the Eldorado HRCA in supporting PAC075 within Tahoe. Critically, both passages assume the Tahoe PAC075 has been removed due to lack of suitable habitat. The report also acknowledges that the spotted owl exhibits "high site fidelity" and notes that a pair has returned to the area. The report states that, when partially burned stands are counted, some 304 acres of suitable habitat remain on the former HRCA within the Eldorado Forest. Because this amount was not enough to sustain a breeding pair, the report recommended that the PAC and HRCA protections be removed, though some green patches within the former PAC would be subject to a limited operating period pending further surveys.

The Tahoe Forest has now decided not to de-list PAC075 due to the presence of the owl pair. Given the difficulty of constructing an adequate HRCA within Tahoe Forest, the HRCA area within Eldorado Forest could play an important role for the pair of owls in the Tahoe PAC through the surviving patches of green cover and nearby snags. The Eldorado FEIS never assessed the potential role of the remaining suitable habitat within the former HRCA for a maintained Tahoe PAC075 despite the acknowledged presence of owls in the area. This omission amounts to an insufficient consideration of cumulative impact under NEPA.

The government first responds that the owl nest is in Tahoe, that the Tahoe PAC is being preserved, and that the Plaintiffs failed to prove how close the nest lies to Eldorado. However, the Eldorado FEIS itself reveals the close connection: "The activity center for owl territory PAC075 is actually located on the Tahoe National Forest.... Since the activity center for PAC075 is close to the Eldorado National Forest boundary, a 300-acre PAC for PAC075 was delineated on the Forest." The close proximity suggests the remaining viable portions of the former Eldorado HRCA may play a role in supporting the Tahoe PAC, triggering the requirement for analysis of the cumulative impact.

The government also argues that Tahoe made its listing decision in November 2002, some five months after Eldorado completed its FEIS.12 However, the government admits that the Tahoe decision was based upon the same May 2002 surveys showing actual owl presence that the Eldorado FEIS recognized. The Sierra Nevada Framework only permits de-listing of an owl PAC when "habitat is rendered unsuitable by a catastrophic stand-replacing event and surveys conducted to protocol confirm non-occupancy." Given that the Eldorado FEIS itself acknowledged the return of owls to Tahoe PAC075, see FEIS at M-16 & M-45, the agency should have reasonably foreseen that Tahoe would not — indeed, could not — de-list PAC075. Thus, the failure to consider the impact of Tahoe's likely decision appears to be a serious omission.

In supplemental briefing, the government stressed that both forest supervisors surveyed a great deal of territory and found very limited suitable habitat. The agency also emphasized that the FEIS explicitly acknowledged the presence of the two owls, considered the available habitat, and approved the sale nonetheless. Essentially, the government argues that because it reasonably determined the entire area to be unsuitable for owl habitat, it could treat the temporary presence of owls as an aberration. From this point of view, the Tahoe decision to maintain the PAC was irrelevant, since it did not imply the continuing presence of owls. However, at the time Eldorado issued the FEIS, surveys confirmed the presence of owls in their traditional nesting area within Tahoe.13 Given that the Sierra Nevada Framework calls for the maintenance of an HRCA surrounding a PAC, that the same Framework forbids de-listing of a PAC during owl occupancy, and that the record showed actual evidence of owl occupancy, the failure to consider maintaining additional snags and green canopy throughout the former HRCA is not reasonable.

Finally, the government argues that the agency did take reasonable precautions against the Tahoe decision by modifying the preferred alternative to prevent the removal of any trees with partial green canopy from former PAC and OFEA areas. However, Plaintiffs argue primarily that the owls will be damaged by the removal of snags within the PAC and removal of partially green trees from the surrounding HRCA. As mentioned, Plaintiffs claim that a great deal of green forest still exists within the HRCA but outside of the PAC and the OFEA. Thus, they have focused this portion of their NEPA claim on trees not affected by the administrative modifications. Because the FEIS failed to consider the effect that removal of these trees and snags might have on the Tahoe PAC, even though the agency could have reasonably anticipated that the Tahoe forest would be obliged to maintain the PAC, we cannot say that the FEIS sufficiently considered the cumulative impacts of the Star Fire sale.

The Plaintiffs have demonstrated a reasonable probability of success on the merits of some of their claims, and the district court applied an improper legal standard of irreparable harm. We also note that in its discussion of the balance of hardships, the district court only compared the Plaintiff's failure to demonstrate irreparable harm with the "loss in value of deteriorating timber" to Sierra Pacific. Under Ninth Circuit case law, however, it is also appropriate to consider the broader public interest in the preservation of the forest and its resources. See Kootenai Tribe, 313 F.3d at 1125 ("We have already decided that, in a case such as this one where the purpose of the challenged action is to benefit the environment, the public's interest in preserving precious, unreplenishable resources must be taken into account in balancing the hardships."). Those resources include any living trees outside the PACs and OFEA, as well as the California spotted owl. See also Sammartano v. First Judicial District, 303 F.3d 959, 974 (9th Cir.2002) (holding that the public interest is "better seen as an element that deserves separate attention in cases where the public interest may be affected."). We therefore reverse the decision of the district court and remand for further proceedings consistent with this opinion.

REVERSED AND REMANDED.

I concur in Judge Thomas's opinion. I write separately to state my belief that the Forest Service, because of its financial interest in the sale, may be disqualified from approving the sale of timber from the Eldorado Forest. The "general rule" has been long established: "officers acting in a judicial or quasi-judicial capacity are disqualified by their interest in the controversy to be decided." Tumey v. Ohio, 273 U.S. 510, 521, 47 S. Ct. 437, 71 L. Ed. 749 (1927). In deciding whether or not a sale should be made, the Forest Service determines the legal rights of a private corporation and the legal rights of those seeking to enforce the statutes protecting the environment. The Forest Supervisor and the Regional Forester making this determination are not judges in a black gown sitting on a bench, but as surely as such traditional figures they are applying law to resolve a legal controversy. Their function can be called judicial or quasi-judicial or even administrative without changing the relevant analysis because "[i]t has also come to be the prevailing view that `[m]ost of the law concerning disqualification because of interest applies with equal force to ... administrative adjudicators.'" Gibson v. Berryhill, 411 U.S. 564, 579, 93 S. Ct. 1689, 36 L. Ed. 2d 488 (1973), quoting Kenneth Davis, Administrative Law Text, § 12.04 (1972). It is scarcely surprising that, as administrative proceedings have come to determine more and more rights and responsibilities, the desire to purge administrative proceedings of possible financial bias should have become definitive. A bureaucracy, protecting its turf and cherishing the number of its employees and the extent of its empire, can have as lively a bias towards its budget as any old-fashioned venal politician might have in his pocketbook.

The general rule has been tempered by finding no financial bias where regional offices of the Department of Labor imposed fines amounting to less than 1% of the enforcement agency's budget, the fines went to the national office, and the amounts received back by the regions hinged not on fines but on the costs incurred by the regions. Marshall v. Jerrico, Inc., 446 U.S. 238, 245-47, 100 S. Ct. 1610, 64 L. Ed. 2d 182 (1979). It may be that the Forest Service officers making the determination have a similar de minimis interest in the sale. The money received from timber sales by the Forest Service is, by statute, to be deposited in a special fund to be used for construction of "needed roads" by the Forest Service and to cover "the cost for Forest Service sale preparation and supervision of the harvesting of such timber." 16 U.S.C. § 472a(h). The reference to "needed roads" is to the forest development roads authorized by 16 U.S.C. § 535. The agency that is the decision-maker is thereby funded in operations that it conducts. If the Forest Service sold no timber, a portion of its operations would shut down. Forest Service jobs would be lost. Forest Rangers would have to find other work. We are not informed, even after asking for and receiving supplemental briefing, how much income the Forest Service derives from the sales or what percentage of its sales budget comes from the sales.

Congress cannot by statute or long-standing custom turn a biased adjudicator into an impartial adjudicator. The requirement of impartiality is imposed by the constitution. In re Murchison, 349 U.S. 133, 136, 75 S. Ct. 623, 99 L. Ed. 942 (1955). Of course, the impartiality of the agency would not be an issue if all the money from the sales went to the Treasury. Even if a tiny fraction of the sales budget is derived from the sales, the potential for bias may be dismissed as de minimis. The relevant amounts to be considered are not the sums involved in a single sale, but the annual total sum derived from the sales and the percentage of the Forest Service sales budget constituted by that amount. A preliminary survey of the public information available on the budget of the Forest Service suggests that timber sales by the Forest Service generate many millions of dollars and that, to an extent not immediately determinable, the sales create a budget for the Forest Service that, in the conduct of more sales, make it independent of the normal appropriation process. Any governmental agency would put a premium on an operation that gives it a perpetual revolving fund not dependent on Congress. Further investigation of the budgetary process of the Forest Service and the impartiality of the service appears appropriate on remand.

My disagreements with the majority opinion are limited, but they serve to push my position over the line to a respectfully dissenting one. I would affirm.

The majority concludes that the district court applied an improper legal standard. The district court's order properly set forth both sets of criteria for preliminary injunctive relief recognized by our court, the "traditional" test and the "alternative" standard, citing to our decisions:

Certain prerequisites must be satisfied prior to issuance of a preliminary injunction. Under the so-called "traditional" standard, an injunction may be had if the court determines that (1) the moving party will suffer irreparable injury if the relief is denied; (2) there is a strong likelihood that the moving party will prevail on the merits at trial; (3) the balance of potential harm favors the moving party; and (4) the public interest favors granting relief. International Jensen, Inc. v. Metrosound U.S.A., Inc., 4 F.3d 819, 822 (9th Cir.1993). Under the "alternative" standard, an injunction properly issues when a party demonstrates either: (1) a combination of probable success on the merits and the possibility of irreparable injury if relief is not granted; or (2) the existence of serious questions going to the merits combined with a balancing of hardships tipping sharply in favor of the moving party. Id. The requirement for showing a likelihood of irreparable harm increases or decreases in inverse correlation to the probability of success on the merits, with these factors representing two points on a sliding scale. United States v. Nutri-cology, Inc. 982 F.2d 394, 396 (9th Cir.1985).

The concern identified by the majority is with the court's articulation of one element of the test, the "possibility of irreparable injury." What the district court said about that element in applying the "traditional" test was: Even if Plaintiffs were successful in demonstrating a likelihood of prevailing on the merits, they must still make some showing of irreparable harm under the "traditional" standard for granting a preliminary injunction. Plaintiffs have failed to meet that burden. They have failed to establish that the project will result in actual harm to the California spotted owl as opposed to speculation that some such harm could possibly occur. Plaintiffs have failed to show that measures already in place to restrict cutting of trees exhibiting any green canopy in either the PACs or within areas of Old Forest Emphasis will not afford sufficient protection. They have also failed to identify any concrete probability of irreparable harm in any other respect.

In applying the "alternative" test, the district court stated: "As discussed above, Plaintiffs have shown neither probable success on the merits or the possibility of any concrete irreparable injury."

The majority concludes that the district court thus placed a higher burden of proof on Plaintiffs than is warranted. Though reluctant to invite a debate over semantics, I do not believe that was the case. Rather, I understand the language used by the district court to have been intended to emphasize that Plaintiffs did not persuade it that the chance or probability of actual harm — as opposed to something speculative — was sufficient to tip the balance in favor of granting the requested preliminary injunction.

The fact that the district court used slightly different words does not mean that it failed to apply the proper legal standard for deciding whether or not to grant the requested relief. Our many cases discussing preliminary injunctions do not all use exactly the same words in articulating the criteria. That is illustrated by the difference in language used in the cases cited by the district court as compared with the language used in the cases cited in the majority opinion, ante at 1297-98, Johnson v. Cal. State Bd. of Accountancy, 72 F.3d 1427, 1430 (9th Cir.1995) and Clear Channel Outdoor, Inc. v. City of Los Angeles, 340 F.3d 810, 813 (9th Cir.2003). There is no indication or reason to believe that the slightly different language is intended to express different legal standards. We have not, by en banc process, substantively altered the standards expressed in the older cases.

Regardless of whether the "traditional" or "alternative" test is applied, the decision whether or not to issue a preliminary injunction inherently involves weighing and balancing competing factors. One of the factors is irreparable injury. In order to justify a preliminary injunction — in order to carry enough weight to tip the balance — the cited harm must be real. The law does not require the identified injury to be certain to occur, but it is not enough to identify a purported injury which is only theoretical or speculative. That is true no matter which standard is applied:

To obtain a preliminary injunction, the moving party must show either (1) a combination of probable success on the merits and the possibility of irreparable injury, or (2) that serious questions are raised and the balance of hardships tips sharply in its favor. These formulations are not different tests but represent two points on a sliding scale in which the degree of irreparable harm increases as the probability of success on the merits decreases. Under either formulation, the moving party must demonstrate a significant threat of irreparable injury, irrespective of the magnitude of the injury.

Big Country Foods, Inc. v. Bd. of Educ. of Anchorage Sch. Dist., 868 F.2d 1085, 1088 (9th Cir.1989) (citations omitted) (emphasis added); see Simula, Inc. v. Autoliv, Inc., 175 F.3d 716, 725 (9th Cir.1999); Goldie's Bookstore, Inc. v. Superior Court, 739 F.2d 466, 472 (9th Cir.1984) ("Speculative injury does not constitute irreparable injury."); 11A CHARLES ALAN WRIGHT, ARTHUR R. MILLER & MARY KAY KANE, FEDERAL PRACTICE & PROCEDURE: CIVIL 2D § 2948.1 at 153-56 (1995).

The balancing to be conducted by the court properly factors in how likely it is that the irreparable injury identified by the moving party will actually occur. In this case, the district court concluded that Plaintiffs had failed to establish that the owls would be put more at risk by implementation of the Forest Service's plan — that is, that denial of the injunction would actually result in an irreparable injury. That was not to say that the owl's habitat was secure. The fire put that habitat at grave risk. The Forest Service concluded that the areas previously set off as Protected Area Centers were no longer suitable for the owls and that the PACs could not effectively be relocated in what remained of the forest after the fire. The district court declined to hold that conclusion erroneous, and we have not held it to be erroneous, either. Thus, the prospects for long-term owl inhabitation of the forest or the PACs, without the restoration program, were not good. By itself, that meant that it was unlikely that permitting the reforestation plan to go forward would result in the irreparable injury cited by Plaintiffs, since preservation of the owl habitat was dubious, either way.

Nonetheless, the district court went on to consider whether the owls' prospects would be made worse if the preliminary injunction was denied. As to that, the conclusion of the district court was that the restoration plan did what it could to protect the remaining habitat: "Plaintiffs have failed to show that measures already in place to restrict cutting of trees exhibiting any green canopy in either the PACs or within areas of Old Forest Emphasis will not afford sufficient protection." On that basis, entry of the requested preliminary relief was denied.

That treatment by the district court did not represent the application of an erroneous legal standard. The court did not require Plaintiffs to establish a probability of irreparable injury, in the sense of more likely than not. The factual determination by the Forest Service that the remaining habitat is not suitable for owls by itself established that it was unlikely that the injunction would save the owls' habitat. But the district court went on to consider the protections provided in the restoration plan and concluded that they were sufficient, such that the injunction sought by Plaintiff would not have made an actual difference. If an actual injury is not going to result, then denial of an application for preliminary injunctive relief is appropriate.

After concluding that the district court applied an erroneous legal standard, the majority moves on to fault the court for failing to focus separately on the logging of trees with green canopy outside the PACs and the Old Forest Emphasis Areas. In particular, the majority quotes an argument made by Plaintiffs' counsel to the effect that the owls had no chance of surviving unless additional green canopy was preserved — with no reference to any evidence whatsoever — and concludes that "[t]here is no indication that the district court considered this possibility in its assessment of irreparable harm." Ante at 1299-1300. The burden was on Plaintiffs as the moving party. Argument is not evidence. The district court is not required to discuss and discount any argument which Plaintiffs happened to make.

More to the point, that assertion is simply wrong. As quoted above, the district court concluded, in so many words: "Plaintiffs have failed to show that measures already in place to restrict cutting of trees exhibiting any green canopy in either the PACs or within areas of Old Forest Emphasis will not afford sufficient protection." That indicates that the district court did consider the remaining green canopy and determined — as a matter of fact — that Plaintiffs had failed to prove their assertion. We have not held that determination to be clearly erroneous.

The district court concluded that Plaintiffs did not demonstrate a likelihood or a reasonable possibility that irreparable injury would result from denial of their application. That conclusion was not erroneous. The district court's denial of the application for a preliminary injunction was proper and should be affirmed.

3. Public Interest as a Preliminary Injunction Factor to Consider

The "traditional" standard for preliminary injunctions explicitly identifies the "public interest" as a factor to be considered. The "alternative" test does not, but that does not mean that it should not be considered. In considering an application for a preliminary injunction, the public interest should be considered as an "element that deserves separate attention in cases where the public interest may be affected." Sammartano v. First Judicial District, 303 F.3d 959, 974 (9th Cir.2002).

The majority faults the district court for failing "to consider the broader public interest in the preservation of the forest and its resources." Ante at 1308. But it did. The district court's order explicitly discussed this factor and appeared to conclude that the public interest weighed against granting the injunction.

That discussion followed the court's discussion concerning the "balance of hardships." The court's order referred to Plaintiffs' failure to demonstrate irreparable injury. It then went on to note that enjoining the program would result in a loss in value from deteriorating timber — and thus a reduction in the money available to fund the reforestation plan — and would delay implementation of replanting efforts. The court concluded that the balance of hardships appeared to favor permitting the restoration program to proceed.

Immediately after that discussion, the court moved to consideration of the public interest:

The fact that the balance of hardships does not weigh in Plaintiff's favor, as discussed above, is significant in assessing whether the public interest would be served by granting the proposed injunction. Delay in harvesting dead timber will result in degradation of wood quality and decreased sales revenue that, in turn, will reduce the only assured source of funding for forest rehabilitation as envisioned by the project. Restoration of those portions of the Eldorado National Forest damaged by the Star fire is clearly in the public interest.

The majority fails to identify precisely what is wrong with that evaluation. It similarly fails to explain why preservation of a burned-out forest and postponement of rehabilitation plans serves the public interest.

This is not an environmental case where some natural treasure is threatened with extinction because of commercial plans to harvest resources. The reality is that a fire devastated Eldorado National Forest, leaving the Forest Service to decide how to make the best of a bad situation. There is not inherently a public interest value in "preservation," when what is to be preserved is that bad situation. Reasonable people can disagree on what approach would be best for the forest, the owls, the environment, and the public interest generally. But the responsibility for making the decision has been assigned to the Forest Service, and it does not appear that the Forest Service — or the district court — disregarded the public interest or acted arbitrarily and capriciously in making the judgment that it did. The public interest does not support reversal of the district court's order. Rather, it favors affirming it.

The majority identifies two bases for concluding that Plaintiffs have demonstrated a reasonable probability of success on the merits of some of their claims. The second, that the Final Environmental Impact Statement (FEIS) for the project did not sufficiently anticipate the possibility that the Tahoe National Forest would subsequently decide not to give up the Protected Area Center known as PAC075, gives me pause and causes me to add a comment.

The majority concludes that the Forest Service was not required to prepare a single EIS covering the similar Star Fire restoration and timber sale projects planned for the two neighboring national forests, Eldorado and Tahoe. It also concludes that consideration should have been given in the Eldorado study to the cumulative impact of conditions and foreseeable actions in the Tahoe forest. I agree with both of those conclusions. The majority also acknowledges the many sections of "cumulative impact" analysis contained within the Eldorado FEIS, but holds that analysis to be insufficient for one reason: that the Eldorado FEIS and the subsequent decision regarding the Star Fire Restoration program did not anticipate that, months later, the persons responsible for Tahoe would change their minds about one element of the Tahoe plan.

At the time that the key decisions were made and the FEIS issued for Eldorado, it was assumed by the persons making those decisions that Tahoe was not going to maintain (or was going to "de-list") the relevant area within its boundaries known as PAC075, because that was the preliminary plan announced for Tahoe. That was still the case at the time that this matter was presented to and ruled on by the district court. That surely did not seem surprising to the officials responsible for Eldorado, because the conditions after the fire in Tahoe were generally the same as the conditions in Eldorado, and Eldorado had determined that the habitat after the fire, including the Eldorado PAC075, which adjoined the Tahoe PAC075, was unsuitable to support owls, a factual judgment we do not hold to have been erroneous.

It was not until November 2002, five months after Eldorado completed its FEIS and a month after the district court made its decision, that Tahoe made its decision not to de-list PAC075. The majority faults Eldorado and the district court, in effect, for not being better predictors of the future. In the terms used by the majority, "the agency should have reasonably foreseen that Tahoe would not — indeed, could not — de-list PAC075." Ante at 1308. If Tahoe had not changed its position and instead had gone forward with de-listing PAC075, there would be no basis for that objection.

I do not understand our decision to say in any general sense that one federal office should be expected to predict what another office will do, or more specifically, that one office should be required to predict that another office will change its position on an important element of an announced preliminary plan. The circumstances here are very unique, and they should be recognized as such by anyone reading our opinion. The decision by Tahoe which the majority concludes should have been foreseen by Eldorado was essentially identical to a decision which Eldorado had to make for itself — whether to maintain or to de-list Eldorado's adjacent PAC075. The majority has concluded that Eldorado's decision to de-list its PAC075 was erroneous because it ignored the presence of a pair of owls spotted in the area of PAC075 in an April 2002 survey. It is precisely that same spotting of the pair of owls which the majority cites as the basis for concluding that Eldorado should have reasonably foreseen that Tahoe would not de-list PAC075. As the majority points out, the Framework requires maintenance of PACs unless "habitat is rendered unsuitable ... and surveys confirm non-occupancy" (emphasis added). Ante at 1304.

In this instance, then, Eldorado is really not being faulted for not looking over Tahoe's shoulder and anticipating that Tahoe would change its mind about an element of its own plan. Instead, Eldorado has been faulted for de-listing its own PAC075 even though a pair of owls was spotted there. The discussion of Tahoe's PAC075 is simply a further ramification of that failure, since if Eldorado had realized that the Framework required the preservation of its PAC075, it would have anticipated the preservation of Tahoe's PAC075, as well.

I comment on this to underscore the unusual circumstance here. It should not ordinarily be expected that one government agency or office can reasonably anticipate that another will change its mind.