Opinion by Judge FERGUSON; Concurrence by Judge MILAN D. SMITH, JR.; Concurrence by Judge FERGUSON.
FERGUSON, Circuit Judge:The Lands Council and the Wild West Institute (collectively, “Lands Council”) appeal the district court’s denial of their motion for a preliminary injunction to halt the Mission Brush Project (“Project”). Under the Project, the United States Forest Service (“Forest Service” or “Service”) plans to allow the selective logging of 3,829 acres of forest in the Idaho Panhandle National Forests (“IPNF”) for the purpose of restoring portions of the forest to historic conditions. Lands Council alleges that the Project violates the Administrative Procedure Act (APA), 5 U.S.C. § 706 et seq., the National Forest Management Act (NFMA), 16 U.S.C. § 1600 et seq., the National Environmental Policy Act (NEPA), 42 U.S.C. § 4321 et seq., and Standard 10(b) of the IPNF Forest Plan. The district court held that Lands Council was unlikely to prevail on its claims and that the balance of hardships favored the Forest Service.
We reverse.
FACTUAL BACKGROUND
The Mission Brush Area
The Project assessment area is in the Bonners Ferry Ranger District in the northern portion of the IPNF. The area is home to abundant plant and animal species, including grizzly bears, Canada lynx, and flammulated owls. Due to decades of unsustainable forestry practices, however, the area has deviated significantly from its historical composition and structure, which consisted of open ponderosa pine and Douglas-fir stands. For decades, logging companies cut down these old growth trees and, along with the Forest Service, suppressed the frequent, low-intensity fires that formerly contributed to the cyclical process of healthy forest ecology. As a result, much of the historic forest conditions have been replaced by dense, crowded stands of younger Douglas-firs and other mid- and late-successional species. These overcrowded forests, dominated by shade-tolerant trees, can lead to insect infestations, diseases, and stand-replacing fires. According to the Forest Service, “the densely stocked stands we see today are causing a general health and vigor decline in all tree species.” U.S. Forest Serv., Mission Brush Supplemental Final Environmental Impact Statement 3-15 (2006) [hereinafter SFEIS].
The Mission Brush Project
The Project would perform silvicultural treatments and commercial logging on 3,829 acres of forest, including restoration cutting within 277 acres of old growth stands, with the goal of trending the forest toward historic conditions. The Forest Service has divided the Project into three commercial timber sales, the Brushy Mission Sale, the Haller Down Sale, and the Mission Fly By Sale, comprising in total 23.5 million board feet of timber. The first two sales have been sold to private timber companies, but there were no bids on the third. The Service’s contracting officer has stated that he does not intend to award the Mission Fly By Sale until this litigation concludes, although logging under the Brushy Mission and Haller Down sales began several months ago.
*775PROCEDURAL HISTORY
In June 2004, the Forest Service released the Mission Brush Final Environmental Impact Statement (“EIS”) and the Record of Decision, which adopted the Project. Lands Council, along with several other environmental groups, appealed to the Regional Forester, who upheld the Project in August 2004 but ordered the preparation of a supplemental EIS in light of our decision in Lands Council v. Powell (Lands Council I), 379 F.3d 738 (9th Cir.2004), amended by 395 F.3d 1019 (2005).1 In April 2006, the Forest Service released its Supplemental Final EIS (“SFEIS”) and Record of Decision (“ROD”). Lands Council filed an administrative appeal, which the Forest Service denied in July 2006.
In October 2006, Lands Council filed suit challenging the Project in the U.S. District Court for the District of Idaho. Lands Council filed a motion for a temporary restraining order and preliminary injunction to halt the Project. The district court denied the motion on December 18, 2006, and Lands Council timely appealed.
DISCUSSION
I. Preliminary Injunction Standard
We review a district court’s denial of a preliminary injunction for an abuse of discretion. Earth Island Inst. v. U.S. Forest Serv. (Earth Island Inst. II), 442 F.3d 1147, 1156 (9th Cir.2006). A district court abuses its discretion if it “base[s] its decision on an erroneous legal standard or clearly erroneous findings of fact.” Id.
A preliminary injunction should issue when the plaintiff shows “either: (1) a likelihood of success on the merits and the possibility of irreparable injury; or (2) that serious questions going to the merits were raised and the balance of hardships tips sharply in [the plaintiffs] favor.” Lands Council v. Martin (Lands Council II), 479 F.3d 636, 639 (9th Cir.2007) (quoting Clear Channel Outdoor Inc. v. City of Los Angeles, 340 F.3d 810, 813 (9th Cir.2003)). These two alternatives are “extremes of a single continuum” in which “the greater the relative hardship to the party seeking the preliminary injunction, the less probability of success must be shown.” Clear Channel Outdoor Inc., 340 F.3d at 813 (internal punctuation and quotation omitted).
II. Likelihood of Success on the Merits
National Forest Management Act
NFMA requires the Forest Service to develop a forest plan for each unit of the National Forest System. 16 U.S.C. § 1604(a). These plans must include provisions for public participation, while adopting “a systematic interdisciplinary approach to achieve integrated consideration of physical, biological, economic, and other sciences.” § 1604(b), (d). Once a forest plan is developed, subsequent agency actions must be consistent with the plan. § 1604(f).
In addition to these procedural components, NFMA imposes substantive requirements on the Forest Service. In particular, “the forest plan must comply with substantive requirements of the Forest Act designed to ensure continued diversity of plant and animal communities and the continued viability of wildlife in the forest.” Idaho Sporting Cong. v. Rittenhouse, 305 F.3d 957, 961 (9th Cir.2002) *776(citing 16 U.S.C. § 1604(g)(3)(B)). The Forest Service must also “demonstrate the reliability of its scientific methodology.” Ecology Ctr. v. Austin, 430 F.3d 1057, 1064 (9th Cir.2005). A reliable scientific methodology is one that the Forest Service has “verified with observation” and “on the ground analysis.” Lands Council I, 395 F.3d at 1035. The Forest Service may not rely on a methodology that “is predicated on an unverified hypothesis.” Ecology Ctr., 430 F.3d at 1064.
The Forest Service has not proven the reliability of its scientific methodology with regard to wildlife habitat restoration in the Mission Brush Project. In particular, the Service has failed to demonstrate that the Project will not harm the flammu-lated owl, the northern goshawk, the fisher, and the western toad, all of whom the Forest Service has designated as “sensitive species” whose viability is of special concern. See Friends of the Clearwater v. Dombeck, 222 F.3d 552, 556 n. 2 (9th Cir.2000) (explaining “sensitive species” designation). As in Ecology Center, the Forest Service is relying on the “unverified hypothesis” that “treating old-growth forest is beneficial to dependent species.” Ecology Ctr., 430 F.3d at 1064.
In Ecology Center, the Forest Service, as part of another project, sought to engage in rehabilitative treatment of old growth stands “to correct uncharacteristic forest development resulting from years of fire suppression.” Id. at 1063. We concluded that the Forest Service did “not offer proof that the proposed treatment benefits — or at least does not harm — old-growth dependent species.” Id. We held that the Forest Service’s methodology was unreliable since it had not been verified, and that the treatments therefore violated NFMA. Id. at 1063-64.
The Forest Service argues that the present case is distinguishable from Ecology Center because the Service has provided sufficient scientific data on the effects of the Project on wildlife habitat. None of the documents it cites, however, demonstrates the reliability of the Forest Service’s hypothesis that restoration treatment will benefit dependent species.
The Forest Service relies primarily on the Dawson Ridge Study, Dawson Ridge Flammulated Owl Habitat Monitoring (2006) [hereinafter Dawson Ridge Study], the only study it has conducted since our decision in Ecology Center. The Dawson Ridge Study monitored a “relatively small area” of flammulated owl habitat: five 1/5 acre plots in an area totaling only eighteen acres. Id. at 2-3. The researchers received a single response in the 2006 survey. Id. at 1. Based on this solitary hoot, and the fact that the area had been logged in 2000 and underburned in 2002, the report concluded that “owls are using the area after harvest.” Id. at 3. The report admitted that it was “inappropriate” to conclude that the treatments had improved owl habitat, but found it “encouraging” that an owl response had been received in the area. Id. Such responses, it concluded, “imply” that the harvesting practices “are at least maintaining suitable habitat.” Id.
This report is insufficient to meet the requirements of Ecology Center. See 430 F.3d at 1063(single report of observation of bird species in formerly-treated old growth stand was insufficient to prove reliability of scientific methodology). Lands Council rightly points out that the Dawson Ridge Study made no ultimate conclusion about one of the underlying hypotheses of the Project: “that treating old-growth forest is beneficial to dependent species.” Ecology Ctr., 430 F.3d at 1064. The study also says nothing about whether such treatment can create suitable habitat that dependent species will actually use. Its *777conclusion that such treatment could maintain habitat is circumspect at best. By its own statement, there is merely an “encouraging” “implication].” Dawson Ridge Study at 3. This is hardly sufficient to justify “grant[ing][the Forest Service] the license to continue treating old-growth forests while excusing it from ever having to verify that such treatment is not harmful.” Ecology Ctr., 430 F.3d at 1064.2
The other studies fall even shorter of meeting the Ecology Center standards. In none of those studies was any observation made of the actual dependent species in order to determine whether the species will use the habitat if the Forest Service engages in the process it proposes. Compare R. Richard Howie and Ralph Ritcey, Distribution, Habitat Selection, and Densities of Flammulated Owls in British Columbia, USDA Forest Serv. General Technical Report RM-142 (1987) (twenty-year-old study from Canadian forest) with Lands Council I, 395 F.3d at 1034, 1035, 1031(holding that Forest Service methodology was unreliable because it did not “walk ... the land,” it relied on a “model with no on-site inspection,” and its data was “stale”). Other documents relied on by the Forest Service are not studies at all, but rather position papers and “conservation plans.” See, e.g., Montana Partners in Flight, Montana Bird Conservation Plan (2000); Idaho Partners in Flight, Idaho Bird Conservation Plan (2000). These documents are not “on the ground analysis” sufficient to prove the reliability of the Project’s methodology. Lands Council I, 395 F.3d at 1035.3
Accordingly, the Forest Service has not demonstrated the reliability of its methodology. We conclude that Lands Council was, therefore, likely to succeed on its NFMA claim.
National Environmental Policy Act
NEPA requires federal agencies to take a “hard look” at the potential environmental impacts of their actions. Idaho Sporting Cong., 305 F.3d at 963. An agency must prepare a detailed EIS for each action “significantly affecting the quality of the human environment.” 42 U.S.C. § 4332(C). The EIS must “provide full and fair discussion of significant environmental impacts” so as to “inform decision-makers 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. The EIS must “be supported by evidence that the agency has made the necessary environmental analyses,” id., *778and must “address in [a] meaningful way the various uncertainties surrounding the scientific evidence.” Ecology Ctr., 480 F.3d at 1065 (quoting Seattle Audubon Soc’y v. Espy, 998 F.2d 699, 704 (9th Cir.1993)).
Lands Council contends that the Project violates NEPA because the Forest Service failed to include a full discussion of the scientific uncertainty surrounding its strategy for improving wildlife habitat. For the reasons already stated, supra at 776-77, we agree. As in Ecology Center, “[t]he EIS ... treats the prediction that treatment will benefit old-growth dependent species as a fact instead of an untested and debated hypothesis.” Id.
In responding to the public comment that the Service had “failed to [cjite any evidence that its managing for old growth habitat strategy will improve old growth species habitat over the short-term or long-term,” the SFEIS does not address the scientific uncertainty described above, nor does it cite the sources now relied on by the Forest Service. SFEIS at F-3. The SFEIS cites only sources discussing the historical conditions of the forest, the role of fire in the forest’s ecology, and the health of old growth trees following treatment. Id. There is no discussion of the uncertainties regarding wildlife and then-use of these habitats following treatment. Id. In fact, the SFEIS’s direct and indirect effects analysis simply “assumes that active management through regeneration and selective tree cutting can help restore natural processes in an ecological system.” SFEIS at 4-68 (emphasis added).
For these reasons, Lands Council was likely to succeed on its NEPA claim.
IPNF Plan Standard 10(b)
NFMA requires the Forest Service to comply with the forest management plan for each national forest. 16 U.S.C. § 1604(i). Standard 10(b) of the IPNF Forest Plan requires the Forest Service to maintain at least ten percent old growth throughout the forest. IPNF Forest Plan Standard 10(b).
Lands Council contends that the IPNF is not currently meeting the ten percent requirement and that the Forest Service therefore must address how cutting mature, future old-growth trees will affect its future compliance with Standard 10(b). Lands Council bases its contention on its own report, which concluded that seventy percent of designated old growth stands did not actually meet the Forest Service’s own standards for old growth. Ellen Pick-en, The Lands Council, Lost Forests: An Investigative Report on the Old-Growth of North Idaho (2005).
The Forest Service disagrees with these results. In determining the percentage of old growth in the IPNF, the Service has used two independent monitoring tools, each of which concluded that approximately twelve percent of the forest met old growth criteria. Arthur C. Zack, Review of Old Growth Assessments for the Idaho Panhandle National Forest 6 (2006) (referencing the Forest Inventory Analysis data, finding 11.8 percent old growth, and the IPNF stand map, finding 12.1 percent old growth). The Forest Service’s expert, Dr. Arthur Zack, specifically considered and evaluated Lands Council’s report, but disagreed with its methodology and conclusions.
Where an agency is presented with conflicting data, it “must have discretion to rely on the reasonable opinions of its own qualified experts.” Marsh v. Oregon Natural Res. Council, 490 U.S. 360, 378, 109 S.Ct. 1851, 104 L.Ed.2d 377 (1989). Accordingly, “[a]t this stage, the record does not allow us to conclude that the Forest Service acted arbitrar[il]y and capriciously in relying on its own data and *779discounting the alternative evidence offered by the [p]laintiffs.” Earth Island Inst. v. U.S. Forest Serv. (Earth Island Inst. I), 351 F.3d 1291, 1302 (9th Cir.2003). The Service explained the differences between its findings and those of Lands Council, and it “is entitled to use the data it collected.” Id.4
Accordingly, the district court properly concluded that Lands Council was not likely to succeed on the merits of its Standard 10(b) claim.
III. Balance of Hardships
Because Lands Council has demonstrated a strong probability of success on the merits of its NFMA and NEPA claims, “it need only show the possibility of irreparable injury if preliminary relief is not granted, and that the balance of hardships tips in its favor.” Earth Island Inst. II, 442 F.3d at 1177 (internal punctuation omitted). “Environmental 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.” Amoco Prod. Co. v. Vill. of Gambell, 480 U.S. 531, 545, 107 S.Ct. 1396, 94 L.Ed.2d 542 (1987). In addition to balancing the hardships to the parties, we must also consider the public interest. Kootenai Tribe of Idaho v. Veneman, 313 F.3d 1094, 1125 (9th Cir.2002).
The Project, if not enjoined, would allow treatment of 2,326 acres of capable flammulated owl habitat and 202 acres of suitable flammulated owl habitat, the latter of which is more than half of the owls’ 364 current suitable habitat acres within the Project area. The Project would treat 2,503 acres of capable northern goshawk habitat and 561 acres of suitable northern goshawk habitat. It would convert 255 acres of suitable goshawk habitat to unsuitable habitat and would prevent 757 acres of capable goshawk habitat from becoming suitable. The Project would also treat 1,839 acres of capable fisher habitat and 449 acres of suitable fisher habitat.
On the other side of the balance are the potential environmental harms to the forest caused by delaying the Project, as well as the potential economic harms to the local community from enjoining logging.
As to the risk to the forest of delaying the project, the permanent and certain harm of violating the environmental laws outweighs the speculative harm that might result from a failure to engage in a statutorily prohibited activity. That determination was made by Congress when it enacted the statutes which prohibit the type of activity in which the Forest Service wishes to engage unless and until the Service complies with those statutes.
The potential economic hardships, however, are more troubling. According to Intervenors, enjoining the project will force the timber companies that purchased the sales to lay off some or all of their workers. One of the companies employs fifteen people and the other employs twenty-two. The Project area is located in Boundary County, which has one of Idaho’s highest unemployment rates and an average wage that is below the national average. Since 2003, the county has lost two major employers (accounting for 400 jobs), including a Louisiana Pacific mill. *780These concerns implicate the public interest.
While balancing environmental harms and economic harms is not easy, it is not unprecedented. We have held time and again that the public interest in preserving nature and avoiding irreparable environmental injury outweighs economic concerns. Earth Island Inst. II, 442 F.3d at 1177; Earth Island Inst. I, 351 F.3d at 1308-09; National Parks & Conservation Ass’n v. Babbitt, 241 F.3d 722, 738 (9th Cir.2001); see also Sierra Nev. Forest Prot. Campaign v. Tippin, No. 06-00351, 2006 WL 2583036, at *21 (E.D.Cal. Sept. 6, 2006) (“The environment is a vital constituent public interest that must be recognized and protected by federal law even in the face of adverse economic consequences.”).
Accordingly, Lands Council demonstrated a threat of irreparable injury sufficient to warrant granting the preliminary injunction.
CONCLUSION
Lands Council demonstrated a probability of success on the merits and a possibility of irreparable injury. Lands Council further showed that the balance of hardships and the public interest favored granting the preliminary injunction. For these reasons, we reverse the district court and remand for entry of a preliminary injunction of the contested portions of the Mission Brush Project.
To the extent that either party believes that any further factual development is required and appropriate in light of this opinion, the district court may engage in such further factual determinations, including by way of trial, as it deems proper. The mandate shall issue forthwith.
REVERSED and REMANDED.
. Lands Council I involved a different project in a different area of the IPNF. 395 F.3d 1019.
. The Forest Service argues that continued monitoring pursuant to the ROD will confirm the reliability of its methodology. While ongoing monitoring is certainly a good idea, this “authorize first, verify later” approach was roundly rejected in Ecology Center as inconsistent with both NFMA and NEPA. 430 F.3d at 1071.
. Intervenors contend that "the proxy on proxy method” of using ' changes to old growth habitat to assess environmental effects on wildlife is a reliable methodology. However, we have never held that manufacturing wildlife habitat through invasive commercial harvesting allows the Forest Service to assume that such habitat will subsequently be occupied by the species at issue. Rather, the proxy on proxy method permits the Service to assume only that “maintaining the acreage of habitat necessary for survival would in fact assure a species’ survival.” Envtl. Prot. Info. Ctr. (EPIC)v. U.S. Forest Serv., 451 F.3d 1005, 1017 (9th Cir.2006) (emphasis added); see also Native Ecosystems Council v. U.S. Forest Serv., 428 F.3d 1233, 1250 (9th Cir.2005) ("Our case law permits the Forest Service to meet the wildlife species viability requirements by preserving habitat ....”) (emphasis added); Ecology Ctr., 430 F.3d at 1064 (distinguishing between "maintaining ... old-growth habitat" and "altering the composition of old-growth habitat through an invasive process”).
. Lands Council may nevertheless revisit this issue on the merits before the district court should further development on remand be appropriate. Infra at 780; see Earth Island Inst. I, 351 F.3d at 1302 ("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.”).