Opinion by Judge GRABER; Dissent by Judge Bybee.
OPINION
GRABER, Circuit Judge:The Kofa National Wildlife Refuge and Wilderness in southwest Arizona contains a desert ecosystem that is home to, among other species, bighorn sheep. After an unexpected decline in the population of the sheep, the United States Fish and Wildlife Service (“Service”) built two water structures (the Yaqui and McPherson tanks) within the wilderness area. Plaintiffs Wilderness Watch, Inc., Arizona Wilderness Coalition, Grand Canyon Wildlands Council, Western Watersheds Project, and Grand Canyon Chapter of the Sierra Club brought suit against the Service. Plaintiffs allege that the Service’s actions violated the express prohibition on the development of structures in the Wilderness Act, 16 U.S.C. §§ 1131-1133. The district court granted summary judgment to the Service, and Plaintiffs timely appeal. Reviewing de novo, High Sierra Hikers Ass’n v. Blackwell, 390 F.3d 630, 638 (9th Cir.2004), we reverse and remand.
FACTUAL AND PROCEDURAL HISTORY
Situated between two mountain ranges, the Kofa Refuge and Wilderness consists of more than 600,000 acres of land in the Sonoran Desert in southwest Arizona. The area contains steep slopes, sparse vegetation, poor soil, and an extremely dry ecosystem. Summer temperatures can reach 120 degrees. The average rainfall measured at one weather station in the Kofa Mountains is about seven inches a year. Most of the rainfall occurs during one month, followed by many hot summer months with no measurable precipitation. The vegetation is dominated by saguaro, creosote, ironwood, paloverde, and mesquite. Kofa is home to 45 mammal species, including the desert bighorn sheep and muledeer, and 47 species of reptiles.
President Franklin D. Roosevelt established the Kofa Game Range by executive order in 1939. The President’s order expressly designated the area for the “conservation and development of natural wildlife resources,” Exec. Order No. 8039, 4 Fed.Reg. 438 (Jan. 25, 1939), and it was understood that preservation of bighorn sheep was one of the principal reasons for establishing the refuge. See David Brown, Early History, in The Desert Bighorn Sheep in Arizona 5-7 (Raymond M. Lee, ed., Ariz. Game & Fish Dep’t 1993). Following the executive order, the Fish and Wildlife Service and the Bureau of Land Management managed the land until 1976, when the Service assumed sole jurisdiction and the reserve was renamed the Kofa National Wildlife Refuge. See An Act to amend the National Wildlife Refuge System Administration Act of 1966, Pub.L. No. 94-223, 90 Stat. 199.
As a wildlife refuge, the area is subject to the provisions of the Refuge Act. Among other things, the Refuge Act requires the Secretary of the Interior to *1027“provide for the conservation of ... wildlife,” “ensure that the biological integrity, diversity, and environmental health of the System are maintained,” and “assist in the maintenance of -adequate water quantity and water quality to fulfill the mission of the System and the purposes of each refuge.” 16 U.S.C. § 668dd(a)(4)(A), (B), (F). In 1990, Congress designated about 82% of the Kofa National Wildlife Refuge as wilderness, and it became the Kofa National Wildlife Refuge and Wilderness.1 Arizona Desert Wilderness Act of 1990, Pub.L. No. 101-628, § 301(a)(3), 104 Stat. 4469. In doing so, Congress subjected the area to the provisions of the Wilderness Act.
Under the Wilderness Act, the Service is “responsible for preserving the wilderness character of the area,” but it also must “administer such area for such other purposes for which it may have been established as also to preserve its wilderness character.” 16 U.S.C. § 1133(b). Congress specifically provided that “wilderness areas shall be devoted to the public purposes of recreational, scenic, scientific, educational, conservation, and historical use.” Id. “[Ejxcept as necessary to meet minimum requirements for the administration of the area,” the Wilderness Act prohibits any “temporary road, ... use of motor vehicles, motorized equipment or motorboats, ... landing of aircraft, ... mechanical transport, and ... structure or installation” within a wilderness area. Id. § 1133(c). The Act declares that the purposes of the Wilderness Act are “within and supplemental to the purposes for which ... [the] national wildlife refuge systems are established and administered.” Id. § 1133(a). Thus, in managing the Kofa Wilderness, the Service must comply with both the Wilderness Act and the Refuge Act.
Since the 1950s, the State of Arizona, non-profit organizations, and the federal government developed water sources such as catchments, wells, and tanks to augment the availability of water for the bighorn sheep that inhabit the area. There are now more than 100 water sources in the area. During the summer months, the distribution of bighorn sheep is restricted by water availability, and most sheep can be found within a two-mile radius of water sources. With the cooperation of the Arizona Game and Fish Department and the Arizona Desert Bighorn Sheep Society, refuge personnel build, maintain, and monitor the water sources. During times of “extensive drought,” refuge personnel transport water to the structures. Wildlife managers believe that these water sources have been instrumental in helping to restore the population of bighorn sheep.
A. The Management Plan for the Refuge
After Congress designated most of the Kofa Refuge as a wilderness area in 1990, and in an attempt to coordinate the dual purposes of the Kofa Wilderness and Refuge, the Service and the Bureau of Land Management (“BLM”) issued a management plan. Kofa National Wildlife Refuge and Wilderness Interagency Management Plan (1997) (“Management Plan”).2 The plan, which received public review and *1028comment, was intended to ensure that future management decisions and techniques concerning the Kofa Wilderness were compatible with the Wilderness and Refuge Acts. The plan recognized the purpose of the Kofa Wilderness in preserving bighorn sheep:
Historically, Kofa ... ha[s] played a central wildlife and wildlands conservation role in western Arizona. To counter dwindling populations of desert bighorn sheep in the earlier part of the century, a management theme relating to the recovery of the species had become necessary beyond the establishment of legal protection for the species under the Arizona State Game code. Thus, a clear and dominant strategy for the management of these historically “rocky, waterless sierras ...” was designed specifically for the recovery of bighorn sheep populations.
Management Plan at 2 (second ellipsis in original) (footnote omitted). The plan stated that the Service and the BLM would “continue important efforts on behalf of the bighorn sheep.” Id. at 3. The plan acknowledged that, although the wilderness designation of the land would “not chang[e] the purposes of these areas or the importance of current activities,” the designation would “call for the consideration of these activities within the larger ecological contexts and within national wilderness goals inherent in the Wilderness Act of 1964.” Id. The plan stated that “[t]he needs of the species and the requirements of the Act are not necessarily in conflict. In fact, the habitat management work done to benefit bighorn sheep, including water development, could have a positive influence on the natural cycles of predation and succession for a diversity of life in the desert without detraction of wilderness attributes and values.” Id. at 39^10.
After espousing a broad goal to reconcile the Wilderness and Refuge Acts in the Kofa Wilderness, the plan discussed comprehensive planning objectives for the area. The plan addressed many issues, including “protection of wilderness values,” “wildlife and habitat management,” “law enforcement and emergency services,” and “Native American religious access.” Id. at 5-7 (some capitalization omitted). With respect to wildlife protection, the plan explained that “the Service is responsible to carry out a dual, but nonetheless interrelated, role of managing for bighorn sheep within the context of wilderness.” Id. at 37. According to the plan, the Service would use “minimum tools” in order to “maintain! ] an optimal desert bighorn sheep population.” Id. at 53.
B. Decline in Population of Bighorn Sheep, and the Service’s Response
As the Management Plan described, the Kofa Wilderness historically has been home to the desert bighorn sheep. In recent decades, the estimated population of the bighorn sheep in the Kofa Wilderness and Refuge remained relatively stable, with fluctuations between 600 and 800 sheep. Surveys done in 1991 and 1997 (the period shortly after the refuge was first designated as a wilderness area and while the Service was preparing the Management Plan) revealed that the population remained relatively stable at around 700 sheep, and the Service determined that the carrying capacity of the area was about 800 sheep. In 2000, the estimated population registered on the high end of the range, at 813 sheep.
Due to the population’s stability, the Service and other government agencies have permitted certain activities that generally are viewed as inconsistent with population conservation. For example, since 1979, the area has served as a primary source of sheep for translocation programs to re-establish populations of bighorn sheep in Arizona, Colorado, New Mexico, *1029and Texas. For decades, the Service translocated sheep from the refuge on a nearly annual basis.3 The area also has been a hunting ground for bighorn sheep, and the Service has issued a limited number of hunting licenses (between 9 and 17) each year. The Service also permits hiking in known lambing areas, despite the sheep’s strong aversion to human disturbance. One factor historically benefitting the bighorn sheep within the Kofa Refuge was the absence of predators such as mountain lions. Studies conducted in the 1990s found no evidence of mountain-lion presence within the Kofa Refuge.
In 2003, the estimated population of bighorn sheep in the refuge was 623. Although this marked a decline since 2000, the estimated population nevertheless remained comfortably within the acceptable range of 600 to 800 sheep. Unconcerned, the Service continued issuing hunting permits, allowed hiking, and, encouraged by abundant rainfall in 2005, translocated 31 sheep that year.
In 2006, however, surveys indicated that the population of bighorn sheep had suffered an unexpected decline — to 390 sheep, or approximately 30% to 50% fewer sheep than in recent decades. In response, the Service and the Arizona Game and Fish Department prepared a document in 2007 titled “Investigative Report and Recommendations for the Kofa Bighorn Sheep Herd” (“Investigative Report”). The report came to no concrete conclusions concerning the recent decline in population. Instead, the comprehensive document examined the wide range of mortality factors of bighorn sheep and, for each factor, it presented prescriptive strategies intended to aid successful recovery of their population. The most prominent factors identified by the report are availability of water, predation, translocation, hunting, and human disturbance.
1. Availability of Water
The report identified availability of water as a “critical habitat variable.” Investigative Report at 8. The report concluded that, “[i]n addition to a need for better monitoring and maintenance of existing waterholes, a better distribution of permanent water supplies is needed to provide water in all areas of suitable sheep habitat.” Id. at 9. The report identified strategies such as identifying existing permanent water sources that could be improved by hauling water during low-water seasons, identifying existing water sources that could be redeveloped to improve capacity and efficiency, and identifying locations for new water sources. Id. at 10.
With respect to new water sources, the report stated that “[n]ew water developments can likely be constructed outside of wilderness, although construction in wilderness should remain an option if a wilderness location best meets wildlife:management needs.” Id. at 9. “Plans for ... new water developments will require an environmental assessment and minimum requirement analysis/minimum tool analysis.” Id. at 10.
2. Predation
The report examined predation, primarily by mountain lions. Id. at 11-15. The report noted that, until 2003, no mountain lions were known to exist within the Refuge. Id. at 11-12. Beginning in 2003, however, there have been mountain-lion sightings and confirmed mountain-lion kills of bighorn sheep. Id. at 12. “The overall impact of mountain lion predation on the sheep population is unknown. It is unlike*1030ly that lion predation alone accounts for the decline observed, but it may be additive to other sources of mortality or sufficient to prevent sheep population recovery. Predation by bobcats or coyotes may also be a contributing factor....” Id. at 13. “Limited removal of individual lions identified as regularly preying on sheep may help the bighorn population recover to historical levels.” Id. “Determining the cause of bighorn sheep mortality is another vital part of assessing the effect of predation on the Kofa population.” Id. The report recommended the strategies of determining the extent of predation and reducing predation by the removal of “offending” lions until the sheep population fully recovers. Id. at 14-15.
3. Translocations
The report discussed the high demand by other natural areas for translocated bighorn sheep and the history of translocations from the Refuge to distant natural areas. Id. at 18-19. The report discussed the translocation of 31 sheep in 2005 despite the relatively low population estimate in 2003. Id. at 18. The report found that, “[w]hile not the ultimate cause of the population decline, the 2005 transplant may have contributed to the low numbers seen in [one region of the Refuge] on the 2006 survey.” Id. The report concluded that no additional translocations will occur until the population of sheep returns to historical levels. Id. at 18-19.
4. Hunting
“Hunting for desert bighorn sheep in Arizona is a once-in-a-lifetime opportunity and the demand for bighorn sheep hunting exceeds the allowable harvest.” Id. at 19. The report described the very restrictive limits on permitted hunting and the financial benefit to certain conservation groups, which are granted a small number of annual permits for auction at fundraisers. Id. The government agencies have “issued anywhere from 5 to 17 bighorn sheep permits for the Kofa [Refuge] ... since 1960.” Id. The report recommended, without explanation, that the agencies “[c]ontinue to offer bighorn sheep hunting opportunities consistent with sheep conservation.” Id. at 20.
5. Human Disturbance
“Bighorn sheep tend to use the highest, most rugged areas within their home ranges for lambing.” Id. at 16. “Signal Peak and Castle Dome Peak are two of the most distinctive features of the refuge and as such are popular destinations for hikers.” Id. “Most use of these areas occurs in the cool winter months (November-March), which strongly overlaps the peak lambing season of January-March.” Id. The report cited documentation of “strong reactions (immediate running, left area and did not return) from Kofa sheep in response to 1 or 2 people.” Id. at 17. “Frequent human disturbance of ewes may cause them to abandon these areas for less optimal habitat, which could in turn affect lamb survival.” Id. The report recommended “[r]edue[ing] the negative impacts of human recreational activities on bighorn sheep” by monitoring sheep usage during lambing-season and, if necessary, closing popular hiking trails during lambing season. Id.
6. Conclusions and Additional Documents
The report contained no overall summary and came to no conclusions about the causes of the decline in the population of bighorn sheep, nor did it make a reasoned comparison among the different potential causes and recovery strategies. The report’s discussion of water does not mention the development of the two water structures at issue in this case — the Yaqui and McPherson tanks. Those structures are *1031listed only in a cost comparison chart near the end of the report. Id. at 23. The report lists projects “in priority order”— though the reasons for the priorities are unstated. Id. at 23 (capitalization omitted). In that list, the development of the Yaqui and McPherson tanks is listed sixth out of 14, below four strategies aimed at monitoring mountain lions and reducing mountain-lion predation and one overall strategy for recovery of bighorn sheep. Id.
The Service prepared two other documents before initiating work on the new water structures within the Kofa Wilderness. The first document, titled “Kofa National Wildlife Refuge Minimum Requirements Analysis,” is a two-page document that contains a series of yes/no questions and requires the preparer to circle, with a pen or pencil, either YES or NO. The questions on the second page provide a very small space for an explanation of a yes/no answer, but the first page contains no such area. The preparer — whose identity is unknown — dutifully circled answers to the questions and, on the second page, provided very short explanations for the selected yes/no answers to those questions. The final question asks if there are potential adverse effects on wilderness. The preparer circled “YES — PROCEED TO MINIMUM TOOL ANALYSIS.”
The second document, titled “Kofa National Wildlife Refuge Minimum Tool Analysis,” is a detailed explanation of the proposed action, an explanation for why it is necessary, alternative action plans (including the “no action” alternative), an analysis of the effects of the alternative action plans, and a summary description of the chosen alternative. In a section titled “Why Project Is Necessary,” the Minimum Tool Analysis states, in full:
Kofa NWR was established, in part, for the conservation of desert bighorn sheep and other wildlife, and the maintenance of this population of desert bighorn sheep is very important regionally for the conservation of sheep and as a source for transplants to other locations in order to establish and re-establish other sheep herds. Wildlife is an important component of Wilderness.
The three alternatives are (1) no action, (2) constructing the two structures with mechanized means, and (3) constructing the two structures without mechanized means. The Service selected alternative number 2: constructing the two structures with mechanized means. The Service rejected the no-action alternative because it would not help the sheep. The Service rejected the alternative that used non-mechanized means because it would increase the time needed to complete the project, resulting in increased disturbance to wildlife and human visitors alike.
Over a three-day period in 2007, the Service built the new Yaqui and McPherson water structures, consisting mostly of aerated PYC pipe buried underground. Designed to catch rainwater and run that water into small concrete weirs or troughs, each system is capable of holding approximately 13,000 gallons of water. The Service decided to enter the two areas with motorized vehicles and equipment because it concluded, in its Minimum Tool Analysis, that using motorized equipment was safer for the workers and would reduce the amount of time the workers would be in the wilderness. The workers used existing roads and removed tracks left by the vehicles. The workers also covered the troughs with local sand and rocks to blend the structures into the natural environment, so that only the troughs and small vent pipes are visible above ground. The Yaqui tank is located in the refuge, just outside the wilderness, but two or three water diversion weirs fall within the wilderness. The McPherson tank is located *1032well inside the wilderness, but within 0.1 mile of a designated road in the area.
C. The Current Suit
Plaintiffs filed suit against the Service shortly after the completion of the Yaqui tank,4 arguing that the structure violated the Wilderness Act because the Act prohibits any “structure or installation” within a wilderness area “except as necessary to meet minimum requirements for the administration of the area for the purpose of this chapter.” 16 U.S.C. § 1133(c). The district court granted summary judgment to the Service, holding that the Service had not violated the Wilderness Act.5 Plaintiffs timely appeal.
DISCUSSION
The Wilderness Act prohibits the development of “structure^] or installation^]” on the land “except as necessary to meet minimum requirements for the administration of the area for the purpose of this chapter.” 16 U.S.C. § 1133(c). Both parties (and we) agree that the two water tanks qualify as structures or installations in the Kofa Wilderness. At issue here is whether the water structures fall within the quoted exception in the Act. The Service argues that these structures fit within the exception in the Wilderness Act because (1) the conservation of bighorn sheep is a valid “purpose” of the Wilderness Act and (2) the Service adequately determined that the structures are necessary to meet the minimum requirements for conserving bighorn sheep. Plaintiffs disagree on both points.
Under the Administrative Procedure Act, we may “set aside agency action” only if we determine that the action was “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706(2)(a). We have held that this standard requires us to
carefully review the record to ensure that agency decisions are founded on a reasoned evaluation of the relevant factors, and may not rubber-stamp administrative decisions that [we] deem inconsistent with a statutory mandate or that frustrate the congressional policy underlying a statute. Nevertheless, we may not substitute our judgment for that of the agency....
Friends of Yosemite Valley v. Norton, 348 F.3d 789, 793 (9th Cir.2003) (citation, internal quotation marks, and alterations omitted).
A. Conservation of Bighorn Sheep as a Purpose of the Wilderness Act
As stated above, the Wilderness Act prohibits “structure^] or installation^]” unless they are necessary to meet the minimum requirements of a “purpose” of the Wilderness Act. 16 U.S.C. § 1133(c). We first must decide if the Service’s determination that a purpose of the Act includes conservation of the bighorn sheep is unambiguously contrary to the language of the Wilderness Act. If so, the structures violate the Act, and we will “give effect to the unambiguously expressed intent of Congress.” Chevron, U.S.A, Inc. v. Natural Resources Def. Council, Inc., 467 U.S. 837, 846, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). If not, we must then determine what level *1033of deference to grant to the Service’s interpretation, a determination that depends upon whether the Service’s interpretation has the force of law. Christensen v. Harris County, 529 U.S. 576, 587, 120 S.Ct. 1655, 146 L.Ed.2d 621 (2000).
The Act begins with a broad statement of purpose:
In order to assure that an increasing population, accompanied by expanding settlement and growing mechanization, does not occupy and modify all areas within the United States ..., leaving no lands designated for preservation and protection in their natural condition, it is hereby declared to be the policy of the Congress to secure for the American people of present and future generations the benefits of an enduring resource of wilderness.
16 U.S.C. § 1131(a). The Act defines wilderness as “an area where the earth and its community of life are untrammeled by man, where man himself is a visitor who does not remain,” and as “an area of undeveloped Federal land retaining its primeval character and influence, ... which is protected and managed so as to preserve its natural conditions.” Id. § 1131(c). The Act also states that the “agency administering any area designated as wilderness” must “administer such area for such other purposes for which it may have been established as also to preserve its wilderness character.” Id. § 1133(b). Had Congress stopped there, these strongly worded phrases would have suggested that wilderness areas were to remain untouched- — not merely untouched by development but, literally, untouched by humans. But Congress did not mandate that the Service preserve the wilderness in a museum diorama, one that we might observe only from a safe distance, behind a brass railing and a thick glass window. Instead, Congress stated that the wilderness was to be preserved as wilderness and made accessible to people, “devoted to the public purposes of recreational, scenic, scientific, educational, conservation, and historical use.” Id. Congress was specific about what it understood might be necessary to preserve the wilderness for such public purposes. Congress expressly authorized structures, motorized vehicles, and temporary roads if such things are necessary to meet the minimum requirements for administering the area, id. § 1133(c); indeed, the Act permits, under certain circumstances, aircraft and motorboat use and even mining, id. § 1133(d). Those uses are incompatible with a museum notion of wilderness.
Read as a whole, the Act gives conflicting policy directives to the Service in administering the area. The Service is “charged with maintaining the wilderness character of the land, providing opportunities for wilderness recreation, managing fire and insect risk, and even facilitating mineral extraction activities.” High Sierra, 390 F.3d at 647. It is charged with simultaneously devoting the land to “conservation” and protecting and preserving the wilderness in its natural condition. 16 U.S.C. §§ 1131(c), 1133(b). We cannot discern an unambiguous instruction to the Service. Rather, those competing instructions call for the application of judgment and discretion. We may be able to identify violations at the margins but, in this case, the Act is not so clear that we can identify precisely what the Service must do and must not do. We conclude that the purpose of the Wilderness Act with regard to conservation is ambiguous. See High Sierra, 390 F.3d at 647-48 (“Although we believe that Congress intended to enshrine the long-term preservation of wilderness areas as the ultimate goal of the Act, the diverse, and sometimes conflicting list of responsibilities imposed on administering agencies renders Congress’s intent arguably ambiguous.”).
*1034Our decision in Wilderness Society v. United States Fish & Wildlife Service, 353 F.3d 1051 (9th Cir.2003) (en banc), is not to the contrary. In Wilderness Society, the Service initiated a sockeye salmon enhancement project in a freshwater lake, located in the Kenai Wildlife Refuge and Wilderness, that flowed into the Gulf of Alaska. The Service planned to allow a private corporation to capture 10,000 sockeye salmon each year and transport about 10 million eggs to a hatchery outside the Kenai Wilderness. During the spring, the organization planned to return about 6 million salmon to the wilderness and to sell the rest. Id. at 1058. The plaintiffs argued that the corporation’s activities fell within the Wilderness Act’s ban on all “commercial enterprise[s]” within a wilderness area (subject to certain exceptions inapplicable in that case). Id. at 1061 (citing 16 U.S.C. § 1133(c)). Looking to the common meaning of the term and to the purpose and structure of the Act, we held that the corporation’s activities unambiguously fell within the Act’s prohibition on commercial enterprises.6 Id. at 1062.
Wilderness Society is easily distinguishable from the present case. There, we gave effect to the unambiguously expressed intent that, subject to certain inapplicable exceptions, commercial enterprises were prohibited. The Service argued that its interpretation of the undefined term “commercial enterprise” was entitled to some deference but we held that, because the primary purpose and effect of the action was purely commercial, no deference was due. Here, by contrast, we must analyze conflicting instructions in the Wilderness Act: The Service must preserve the wilderness character of the area while at the same time providing for “recreational, scenic, scientific, educational, conservation, and historical use.” 16 U.S.C. § 1133(b). Both the specific statutory mandate that “conservation” is a valid purpose of the Act and the historical focus of the area on the preservation of bighorn sheep render this case vastly different from the situation we analyzed in Wilderness Society.
Because we conclude that the term “conservation” is ambiguous, we turn to the question of what level of deference to grant the Service’s interpretation of “conservation” as including wildlife conservation and, more specifically, conservation of bighorn sheep. United States v. Mead Corp., 533 U.S. 218, 229-30, 121 S.Ct. 2164, 150 L.Ed.2d 292 (2001); Christensen, 529 U.S. at 587, 120 S.Ct. 1655. We apply Chevron deference “when it appears that Congress delegated authority to the agency generally to make rules carrying the force of law, and that the agency interpretation claiming deference was promulgated in the exercise of that authority.” Mead, 533 U.S. at 226-27, 121 S.Ct. 2164. “It is fair to assume generally that Congress contemplates administrative action with the effect of law when it provides for a relatively formal administrative procedure tending to foster the fairness and deliberation that should underlie a pronouncement of such force.” Id. at 230, 121 S.Ct. 2164. By contrast, “[i]nterpretations such as those in opinion letters — like interpretations contained in policy statements, agency manuals, and enforcement guidelines, all of which lack the force of law — do not warrant Chevron-style deference.” Christensen, 529 U.S. at 587, 120 S.Ct. 1655. “Such views, ... even if not authoritative for purposes of Chevron, are entitled to so-called Skidmore deference insofar as they ‘constitute a body of experience and informed judgment to which courts and liti*1035gants may properly resort for guidance.’ ” Vigil v. Leavitt, 381 F.3d 826, 835 (9th Cir.2004) (quoting Skidmore v. Swift & Co., 323 U.S. 134, 140, 65 S.Ct. 161, 89 L.Ed. 124 (1944)).
Here, the 1997 Management Plan was subject to public review and comment and was intended to “provide long-term management guidance.” But, other than stating that the plan was subject to public review and comment, the record is bereft of any other information describing the formality of the administrative procedure that fostered the plan. We are not convinced that the “management guidance” included in the plan carries the force of law. On this record, we are unable to distinguish with certainty the plan from “interpretations contained in policy statements, agency manuals, and enforcement guidelines, all of which lack the force of law.” Christensen, 529 U.S. at 587, 120 S.Ct. 1655. We therefore apply Skidmore deference. Under that standard, “the deference to be accorded ... depends upon ‘the thoroughness evident in its consideration, the validity of its reasoning, its consistency with earlier and later pronouncements, and all those factors which give it power to persuade, if lacking power to control.’ ” Wilderness Soc’y, 353 F.3d at 1060 (quoting Mead, 533 U.S. at 228, 121 S.Ct. 2164).
The plan demonstrated consistency in recounting the history of Kofa and its “conservation role in western Arizona.” Management Plan at 2. Preservation of the bighorn sheep in the area was one of the principal motivations for President Roosevelt’s establishing the Kofa Game Range in 1939. From the beginning, federal agencies cooperated with the Arizona Game and Fish Department to protect the bighorn sheep.7 The plan reiterated that preserving desert bighorn sheep was an important component of the management of the area and noted that the “management theme” of the area is devoted “[t]o counter dwindling populations of desert bighorn sheep” and to “continue important efforts on behalf of the bighorn sheep.” Id. at 2-3.
The plan also demonstrated thoroughness in addressing the new requirements that the wilderness designation would impose on the goal of conserving bighorn sheep. It concluded:
There is no question that management of this species remains as one of the princip[al] missions of the Kofa [National Wildlife Refuge].... However, the new considerations relative to the Wilderness designations require the Service and the BLM to review management techniques and their compatibility with wilderness principles.
Id. at 36. The plan acknowledged that “the Service is responsible to carry out a dual, but nonetheless interrelated, role of managing for bighorn sheep within the context of wilderness.” Id. at 37. The plan further recognized that the Service had “to maintain the natural character of the landscape” consistent with the Wilderness Act, a duty that required use of “the minimum tool necessary to accomplish the work” and “technologies ... as unobtrusive as possible.” Id. at 37, 39. The plan continued:
The needs of the species and the requirements of the Act are not necessarily in conflict. In fact, the habitat management work done to benefit bighorn sheep, including water development, could have a positive influence on the natural cycles of predation and succession for a diversity of life in the desert *1036without detraction of wilderness attributes and values.
Id. at 39-40.
In light of the historical purpose of the area to preserve bighorn sheep and the explicit purpose of “conservation” in the Act, we find that the Service’s reasoning was thorough, valid, consistent, and persuasive. We defer to the Service’s interpretation in the Management Plan that conservation of the bighorn sheep is consistent with the purposes of the Wilderness Act.
B. The Wilderness Act’s Exception for Structures that are “Necessary” to Meet the “Minimum Requirements” for Conserving Bighorn Sheep
The Wilderness Act prohibits the development of any structure within a wilderness area, subject to only one exception: “except as necessary to meet minimum requirements for the administration of the area for the purpose of this chapter.” 16 U.S.C. § 1133(c). Because the conservation of bighorn sheep is a valid purpose of the Wilderness Act, the relevant question is whether the Service made an adequately reasoned determination of necessity. See High Sierra, 390 F.3d at 646-47 (holding that, in order to invoke an exception to prohibited conduct in a wilderness area, the Wilderness Act requires the relevant agency to make a reasoned finding of necessity).
In High Sierra, we interpreted the similar provision 16 U.S.C. § 1133(d)(5), which permits commercial services “to the extent necessary for activities which are proper for realizing the recreational or other wilderness purposes of the areas.” We held that “[i]t is clear that the statutory scheme requires ... that the Forest Service make a finding of ‘necessity’ before authorizing commercial services in wilderness areas.” High Sierra, 390 F.3d at 646. We held that the Wilderness Act “does not specify any particular form or content for such an assessment” and that we must defer to the form selected by the agency. Id. at 646-47. Accordingly, we found that a “Needs Assessment” document prepared by the Forest Service sufficed as the required “necessity” finding. Id. at 647.
However, under the terms of the Wilderness Act, a finding of necessity is a necessary, but not sufficient, ground for permitting commercial activity in a wilderness area. The finding of necessity required by the Act is a specialized one. The Forest Service may authorize commercial services only “to the extent necessary.”
Id. (quoting 16 U.S.C. § 1133(d)(5)). Keying off the emphasized text, we held that “the Forest Service must show that the number of [commercial activity] permits granted was no more than was necessary to achieve the goals of the Act.” Id.
We held that the Needs Assessment document did not meet that requirement because “[n]owhere in the[document] does the Forest Service articulate why the extent of such [commercial] services authorized by the permits is ‘necessary.’” Id. The Needs Assessment document “examined independently three topics related to the need for commercial services: the types of activities for which commercial services are needed, the extent to which current permits are being used, and the amount of use the land can tolerate.” Id. “All of these are relevant factors to consider ... [but], at some point in the analysis, the factors must be considered in relation to one another.” Id. “If complying with the Wilderness Act on one factor will impede progress toward goals on another factor, the administering agency must determine the most important value and make its decision to protect that value. That is what the Forest Service failed to do in this case.” Id.
*1037Here, similarly, the relevant statutory provision prohibits the creation of structures within a wilderness area “except as necessary to meet minimum requirements for the administration of the area for the purpose of this chapter.” 16 U.S.C. § 1133(c). Like the similar provision in High Sierra, this provision requires the agency to make a finding of necessity. But a generic finding of necessity does not suffice, High Sierra, 390 F.3d at 647; the Service must make a finding that the structures are “necessary” to meet the “minimum requirements for the administration of the area for the purpose of [conserving bighorn sheep],” 16 U.S.C. § 1133(c) (emphasis added).8
The Service failed to make that required finding. The Service undoubtedly found that, assuming that improvements to water facilities were necessary, the development of the two water structures was necessary. The record contains the requisite necessity finding as to that narrow question and a reasoned analysis justifying the finding. But the key question — whether water structures were necessary at all — remains entirely unanswered and unexplained by the record, even though the Service’s own documentation strongly suggests that many other strategies could have met the goal of conserving bighorn sheep without having to construct additional structures within the wilderness area (for example, eliminating hunting, stopping translocations of sheep, and ending predation by mountain lions).
The Service points us to the three documents discussed in some detail in the factual background section, above: the 2007 Investigative Report, the Minimum Requirements Analysis, and the Minimum Tool Analysis. But they do not contain the required analysis.
The Service’s 2007 Investigative Report provided a thorough, neutral, and scientific assessment of the many factors that contribute to sheep mortality, and the report made recommendations pertaining to each mortality factor. But the Investigative Report did not — and did not purport to— assess the factors relative to each other. Nor did the Investigative Report reach legal conclusions or even cite the relevant legal standard that the Service must apply before developing structures in a wilderness area. What the report did conclude, however, is that many factors other than access to water, either alone or in combination, might suffice to restore the population of bighorn sheep.
Perhaps most importantly, the report concludes that mountain lions recently began inhabiting the Kofa Refuge and killing bighorn sheep; in fact, the arrival of mountain lions corresponds precisely to the time frame in which the population of sheep declined from its previously stable, optimal level. After decades of mountain-lion absence, the Service first received reports of mountain lions (and mountain-lion kills of bighorn sheep) in the Kofa Refuge during the crucial time period of 2003 to 2006. The report even concludes that mountain-lion predation may be “sufficient to prevent sheep population recovery.” In its summary chart, the report lists four different actions related to mountain-lion predation as higher priorities than the development of water structures.
Similarly, the report acknowledges that, despite the relatively low population of sheep in 2003, the Service carried out a translocation of 31 sheep from the Refuge in 2005 which, the report concluded, “may have contributed to the low numbers seen in [one region of the Refuge] on the 2006 *1038survey.” The report recommends that the temporary cessation of translocations be continued.
Next, the report notes the obvious fact that hunting results in a population decline. Yet the report recommends, without explanation, that hunting be continued in the Refuge.
Finally, the report explains that human disturbance may lead to a reduction in the survival rate of bighorn lambs, and the report notes that peak hiking season in prime lambing territory corresponds directly to peak lambing season. The report concludes that temporary trail closures may be advisable.
In short, the report identified many different actions that were likely to lead to an increase in the population of bighorn sheep: reduction in mountain-lion predation, cessation of translocations, moratorium on hunting, and temporary trail closures. Importantly, in contrast to the creation of new structures within the wilderness, the Wilderness Act does not prohibit any of those actions. The Service could have taken any or all of those actions without the need for a finding of necessity. Yet nowhere in the record does the Service explain why those actions, alone or in combination, are insufficient to restore the population of bighorn sheep.
The Investigative Report did, of course, identify the creation of the new structures as one possible strategy toward the recovery of the bighorn sheep population. But the report never stated that the two water structures are necessary, either in the abstract or with respect to the statutory standard (which the report nowhere cited). The report simply listed that action-item alongside the many other action-items, including those described above, all of which might contribute to the recovery of the bighorn sheep. Indeed, the report concluded that “[n]ew water developments can likely be constructed outside of wilderness,” which would not require a finding of necessity. To the extent that the report contemplated the construction of water structures within the wilderness, it stated that those structures “will require ... [a] minimum requirement analysis/minimum tool analysis.”
Before building the two water structures, the Service did complete a Minimum Requirements Analysis and a Minimum Tool Analysis. Those documents amply describe the reasons for the Service’s decision to construct these two particular water structures, assuming that water structures are necessary at all. But, again, nowhere does the Service address that underlying assumption. The documents leap from the worthy goal of conserving bighorn sheep to the need for additional water structures. The basis for that analytical leap is nowhere described.
In fact, with only one possible exception, the two documents provide no basis for the conclusion that the Service even considered the possibility of actions other than the construction of water structures. The documents as a whole demonstrate that the Service began with the assumption that water structures are necessary and reasoned from that starting point.
In only one place is it even arguable that the Service considered the possibility of other actions. On the first page of the yes/no checklist, the Service circled “NO” in response to the question: “Are there other less intrusive actions that can be taken or that should be tried first inside or outside wilderness that will resolve this issue? (i.e. signing, visitor education, information, regulations, use limits, law enforcement, are [sic] or trail closures, etc). Circle Yes or No.” Viewing the documents as a whole, the Service’s response to this question likely carried forward its primary underlying assumption: that water *1039structures are necessary. In context, the Service likely interpreted this question as asking whether “other less intrusive [water-based] actions” such as hauling of water or development of structures outside of wilderness “should be tried first.”
Even if one assumes that the Service understood this question as querying whether other non-water-based actions should be tried first, this lone, generic question cannot meet the requirement under the Wilderness Act that the Service explain its conclusion. As we wrote in High Sierra, 390 F.3d at 647, the Service must, “at some point in the analysis,” weigh the relevant factors “in relation to one another.” (Emphasis added.) As in High Sierra, “[n]owhere in the [record] does the ... Service articulate why ” the action taken is “necessary” to meet the “minimum requirements” of the Act. Id. (emphasis added). And, as in High Sierra, that failure is fatal. Where, as here, the record demonstrates that many alternative actions not prohibited by the Wilderness Act very well could have attained the Service’s goal, a single yes/no question cannot suffice to invoke the very limited exception for structures that are necessary to meet the minimum requirements for the administration of the purposes of the Wilderness Act. The Service’s decision “entirely fail[s] to consider an important aspect of the problem,” namely, the failure to consider whether new water structures are necessary at all. Lands Council v. McNair, 537 F.3d 981, 987 (9th Cir.2008) (en banc) (internal quotation marks omitted), overruled in other part as recognized by Am. Trucking Ass’ns v. City of Los Angeles, 559 F.3d 1046, 1052 (9th Cir.2009).
The Wilderness Act imposes a strong prohibition on the creation of structures, subject only to an exception for structures that are necessary to meet the Act’s minimum requirements. Just because a particular variable affects the sheep’s viability, the Service is not free to create structures addressing that variable without regard to any other variables at play. The Act certainly provides for some flexibility to address a given situation, even with imperfect information and time and budget constraints. But, unless the Act’s “minimum requirements” provision is empty, the Service must, at the very least, explain why addressing one variable is more important than addressing the other variables and must explain why addressing that one variable is even necessary at all, given that addressing the others could fix the problem just as well or better.
There is little question that improvements to the water supply likely will help the sheep recover. But, when the issue is a new structure, that conclusion is not good enough under this statute. The statute requires that a structure be “necessary to meet minimum requirements” of the wilderness. 16 U.S.C. § 1133(c). It is beyond dispute that, if addressing other variables will lead to satisfactory sheep recovery, then a new structure is not “necessary.” The Service’s complete failure to address that key question is fatal to its conclusion.9
CONCLUSION
The Wilderness Act requires a delicate balancing between Congress’ desire to *1040maintain lands untouched by humans and Congress’ recognition that such an idealistic view is subject to some practical limitations. The prohibition on the creation of permanent structures within a wilderness area is one of the strictest prohibitions, subject only to an exception for structures that are necessary to meet the minimum requirements for the administration of the purposes of the Act. It may be that the two water structures at issue here are, in fact, necessary to meet the minimum requirements for conserving bighorn sheep, which is a permissible purpose. But, in light of the many other potential avenues of achieving bighorn sheep conservation identified by the Service itself, the Service must provide enough evidence and explanation in the record to assure this court that it fully considered those avenues and nevertheless rationally concluded that new water structures are, in fact, necessary. Because the record is wholly inadequate to meet that requirement, we reverse the district court’s determination to the contrary and remand with instructions to determine the appropriate remedy. On remand, the court may accept briefing from the parties on whether to require the Service to dismantle the structures, to remand the matter to the Service for reconsideration regarding “necessity” under the Wilderness Act, or to fashion such other relief as may be appropriate.10
REVERSED and REMANDED with instructions.
. The record often fails to distinguish between the area designated as refuge and the area designated as wilderness. We will refer to the area as the Kofa Refuge or Kofa Wilderness as necessary in this opinion. When we refer to the Kofa Wilderness, we are referring to the part of the Kofa Refuge designated as wilderness, which is subject to provisions of the Wilderness Act and the Refuge Act. We use Kofa Refuge to refer to the entire area. We note that about 18% of the Kofa Refuge is subject only to the provisions of the Refuge Act, an arguably less restrictive statute than the Wilderness Act.
. The plan also covered the New Water Mountains Wilderness.
. The first translocation occurred in 1957, and a total of 569 sheep have been moved from the refuge.
. Plaintiffs also sought a temporary restraining order against construction of the McPherson tank. The parties settled that issue out of court, and the Service built the structure.
. At the district court and on appeal, Plaintiffs also argued that the Service violated the National Environmental Policy Act of 1969 by failing to prepare a public analysis of the environmental effects of the water structures. Because we hold that Plaintiffs prevail on their claim under the Wilderness Act, we need not and do not reach their claim under the National Environmental Policy Act.
. In the alternative, we held that the corporation's activities constituted a commercial enterprise, "even if we were to assume that the Wilderness Act’s prohibition on commercial enterprise within the wilderness is ambiguous.” Wilderness Soc'y, 353 F.3d at 1067-69.
. Arizona granted legal protection to bighorn sheep in 1913.
. It is that specific necessity finding, mandated by the statute, that the Service must make. We in no way hold that the Service must make a finding of "absolute necessity.” Dissent at 1050-51.
. As we have explained, and notwithstanding the dissent’s assertions to the contrary, the documents in the record simply do not contain an explanation for that key question. The dissent cobbles together its own explanation from disparate parts of the various documents at issue and from its own analysis of sheep recovery, but it is the Service's explanation, or lack thereof, that we must review, and we cannot supply missing elements for the agency.
. On appeal, Plaintiffs requested a specific remedy: removal of the structures via non-motorized means. The Service did not object and, indeed, did not address the issue of remedy at all. We express no opinion on whether the Service may be precluded, by waiver or otherwise, from objecting to that form of relief.
Because the Service did not brief the issue of remedy, we believe that it is prudent to remand the issue to the district court to address with the benefit of full advocacy. The dissent's view of the appropriate remedy is, as we state in text, one possible option. But we are perplexed by the dissent's insistence that we must reach an issue that one party failed to brief — particularly because our decision to remand prejudices neither party.