dissenting:
The majority oversteps the limited role of a court reviewing an agency’s decision. First, it holds that the Forest Service’s 216-page Environmental Assessment, six sage-grouse project-area surveys, and bevy of supplemental reports all amount to a “ ‘clear error of judgment’ that would render its action ‘arbitrary and capricious’ ” under the National Forest Management Act. The Lands Council v. McNair, 537 F.3d 981, 993 (9th Cir.2008). Second, on the basis of these same substantive criticisms, the majority holds that the Service’s analysis was too flawed to satisfy the National Environmental Protection Act’s “hard look” requirement. See, e.g., Native Ecosystems Council v. U.S. Forest Serv., 428 F.3d 1233, 1239 (9th Cir.2005) (“Determining whether the Forest Service took the requisite ‘hard look’ is *939judged against the APA’s arbitrary and capricious standard.”).
But ours is only the modest task of ensuring the Service didn’t perform an arbitrary and capricious analysis or come to an arbitrary or capricious conclusion. The Service did neither so the majority is wrong to overturn its decision.
I
The National Forest Management Act requires the Forest Service to develop a general resource management plan for every forest in the National Forest System. 16 U.S.C. § 1604. Forest Plans permit productive uses of forests (such as grazing) while preserving the habitat to support viable populations of all pre-existing plant and animal life. M; 36 C.F.R. § 219.19 (2000). To keep track of animal populations, the Forest Service designates a small number of “management indicator species” to monitor as proxies for all the forest’s animals. 36 C.F.R. § 219.19(a)(1). And, rather than go out and individually count every animal, the Service often monitors indicator species’ habitat as a proxy for their population. We’ve long endorsed this proxy-on-proxy approach. See, e.g., Inland Empire Public Lands Council v. U.S. Forest Serv., 88 F.3d 754, 761 (9th Cir.1996).
Forest Plans, by definition, cover entire forests. Beaverhead-Deerlodge National Forest is the largest national forest in Montana and, at 3.36 million acres, it is roughly the size of Connecticut. In the Beaverhead-Deerlodge National Forest Plan, sage grouse are the indicator species for sagebrush dependent animals forest-wide. The Antelope Basin/Elk Lake project site, at 48,000 acres (1.4% of the total forest), is just a bit larger than the District of Columbia. The Service analyzed the project’s potential effects on the Forest Plan as NFMA required and, as the Forest Plan required, it used sage grouse as the indicator species for sagebrush-dependent animals.
But that isn’t good enough for the majority, which holds that it was arbitrary and capricious for the Service to rely on sage grouse because grouse are “virtually non-existent in the project area.” Maj. at 933. When grouse are “absent from the project area ... the Forest Service cannot reasonably argue that the proxy-on-proxy approach allows it to avoid separately monitoring sage grouse population trends.” Id. at 933-34. This rule — that a project analysis is arbitrary and capricious unless there is proof that each indicator species lives in the project area — is entirely new. It has at least four problems.
First, the majority doesn’t apply its new requirement fairly to the record. The record does not show, and the Service does not concede, that there are no sage grouse in the Antelope Basin/Elk Lake project site. The record shows only that no sage grouse “have been observed in the project area in the past 15 years and sage grouse surveys in 2001-2003 did not find sage grouse or their sign.” On the other hand, there is a report of two sage grouse being taken illegally from the project area in 2002. Rather than meaning that few sage grouse live in the project area, this could just mean the sage grouse in the project area are difficult to find. We’ve repeatedly approved habitat monitoring when the indicator species is difficult to detect. See, e.g., McNair, 537 F.3d at 998. The majority notes that “the government does not cite any ‘monitoring difficulties’ that prevent detection of the sage grouse.” Maj. at 933 n.9. But why would it have? Courts have never required that showing before. We can hardly fault the Service for not providing an unnecessary justification in defense of a well-established method. Creating new requirements and applying them retroactively is the kind of *940“gotcha” jurisprudence we may not engage in.
Second, the majority’s rule conflicts with our cases approving proxy-on-proxy analysis without requiring the indicator species to be present in the project area. For example, in Native Ecosystems Council v. United States Forest Service, plaintiffs challenged the Service’s performing proxy-on-proxy analysis without showing a viable population of the indicator species. 428 F.3d at 1250. We sided with the Service, explaining that “[o]ur case law permits the Forest Service to meet the wildlife species viability requirements by preserving habitat,” provided that the habitat is monitored using reliable methods. Id.; see also 36 C.F.R. § 219.19 (“[Hjabitat shall be managed to maintain viable populations....”). This is plainly inconsistent with the majority’s requirement that the Service prove that each indicator species lives in the project area. Indeed, requiring the Service to prove that each indicator species lives in the project area effectively requires it to directly monitor each animal’s population — precisely the cumbersome task that proxy-on-proxy is meant to avoid.
Third, the majority’s new rule will make it much harder for the Service to plan on a forest-wide scale, rely on forest-wide indicator species and administer general Forest Plans. Whenever one of the Forest Plan’s indicator species is absent or impossible to detect, the Service will have to make an ad hoc exception. Today’s majority requires a stand-in indicator species for a project area that is 1.4% of the total forest. Tomorrow’s might do so for 0.14%. Pretty soon, before the Service can allow grazing in the head ranger’s backyard it’ll have to prove no adverse impact on the gophers. This flies in the face of NFMA. See, e.g., 16 U.S.C. § 1604(f)(1) (The Forest Service shall “form one integrated plan for each unit of the National Forest System.”) (emphasis added); Idaho Sporting Cong., Inc. v. Rittenhouse, 305 F.3d 957, 962 (9th Cir.2002) (“[T]he Forest Service must conduct an analysis of each ‘site specific’ action, such as a timber sale, to ensure that the action is consistent with the forest plan.”).
The majority argues that “[i]f the Forest Service decided to select a project site consisting of the head ranger’s backyard ... it must indeed analyze that particular site to determine the effects of the proposed action.” Maj. at 934. Nobody disputes that. See p. 939-40 supra. It’s at the next step of the analysis where we differ: the majority’s requirement that the Service come up with all-new indicator species if it can’t prove that the species listed in the Forest Plan live in the project site. The entire point of a forest-wide plan is that complying with its forest-wide standards will ensure forest-wide species and habitat preservation. Although the Service must ensure that grazing in the backyard doesn’t hurt any sage grouse habitat, it’s not required to demonstrate that sage grouse actually live in the yard.
Fourth, it’s not clear that an indicator species’ absence from a particular project area undermines the Service’s habitat analysis. NFMA’s goal is preservation of animal life across the forest, which is necessarily determined using the Forest Plan and the Forest Plan’s indicator species. See, e.g., Inland Empire Pub. Lands Council v. U.S. Forest Serv., 88 F.3d 754, 757 (9th Cir.1996) (“[S]ite-specific projects must be consistent with the stage-one, forest-wide plan.”). If sage grouse are a proxy for all the animals that rely on sagebrush, and the Service can show that it’s reliably maintaining sagebrush at the levels required for sage grouse, then NFMA has been satisfied. The absence of grouse in a particular part of the forest may make it harder to double-check these *941methods, but in this case nobody disputes the reliability of Connelly’s Guidelines, which are part of the administrative record. John W. Connelly, et al., Guidelines to Manage Sage Grouse Populations and Their Habitats, 28 Wildlife Society Bulletin 967 (2000).
The majority also criticizes more specific aspects of the Service’s scientific analysis. But bare disagreement doesn’t make the Service’s analysis arbitrary and capricious. See Marsh v. Or. Natural Res. Council, 490 U.S. 360, 378, 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 qualified experts even if, as an original matter, a court might find contrary views more persuasive.”). This is especially true where, as here, the court bases its objections on the findings of a scientist who has actually endorsed the project being reviewed. The majority’s central criticism is that “the very guidelines used by the Forest Service militate against the Forest Service’s assertion” that sage grouse can be monitored by monitoring sagebrush. Maj. at 934. The majority bases this assertion almost entirely on supposed inconsistencies between the Service’s analysis and the work of Dr. Connelly (the Guidelines’ author). E.g., id. (“Connelly and his co-authors recognized that ‘populations of sage-grouse have been extirpated at places throughout their former range concomitant with habitat loss and degradation, so that the species’ current distribution is less closely aligned with that of sagebrush.’ ”).
But Dr. Connelly published a lengthy review of the Service’s project analysis (also in the administrative record) explaining that “[t]he actions authorized by the Antelope Basin/Elk Lake Environment Assessment are generally consistent with the current sage-grouse management guidelines.” John W. Connelly, A Review of USFS Management Activities and Their Relationship to Sage-Grouse in the Antelope Basin/Elk Lake Area of Southwestern Montana (September 15, 2004) at 14 (hereinafter Connelly Review). Connelly found that “[t]he conclusions in the Antelope Basin/Elk Lake Allotment Management Plan EA ... are reasonable and supported by the available evidence. Effects to sage grouse resulting from project implementation will likely be minimal.... The actions authorized ... appear [to] take a pro-active approach to managing for potential sage grouse habitat.” Id. at 14. Courts are ill-equipped to second guess scientists, particularly scientists who are interpreting their own scientific evidence.
II
As the majority acknowledges, NEPA imposes no substantive requirements but merely “exists to ensure a process.” Maj. at 936 (quotation omitted). Nonetheless, the majority concludes that the Service’s use of the Forest Plan’s management indicator species to evaluate the Antelope Basin/Elk Lake Allotment Management Plan did not constitute “the requisite ‘hard look’ mandated by NEPA.” Maj. at 937. This NEPA holding doubles down on the same point the majority makes about NFMA— that no project can be undertaken without studying animals that actually live in the project site. But the majority never explains why violating NFMA’s substantive requirements is necessarily enough to fail the more lenient NEPA requirement of a “hard look” that doesn’t “rely on incorrect assumptions or data.” Native Ecosystems Council v. U.S. Forest Serv., 418 F.3d 953, 964 (9th Cir.2005). Even if NFMA limited the Service to indicator species that live in the project area (it doesn’t), a species’ absence doesn’t necessarily make the Service’s extensive analysis of the project area totally unreliable.
*942The majority also criticizes the Service for not supplementing its Environmental Assessment in the wake of the Connelly Review. But this criticism is hollow on two levels. First, it ignores the fact that the Connelly Review endorsed the Service’s position. See p. 941 supra. Why (and how) would the Service “respond” to an unqualified endorsement? Take a bow? Second, the majority’s premise — that the Service never addressed the Review’s finding that some of the project area was suitable for nesting, maj. at 937 — -is factually inaccurate. The Supplemental Information Report explains on page five that “[cjattle do not enter the Elk Lake Allotment, where 1,900 acres of possible nesting habitat was identified, until June 26 or later.” (emphasis added). It reiterates the point on page twelve: “likelihood of disturbance by livestock to nesting sage grouse appears to be minimal as normally the range is not ready for livestock grazing until after the peak of sage-grouse egg incubation.” The Service clearly considered the Review’s finding. It’s just that the Service — like the Review’s author' — concluded that the project was consistent with these findings. A hard look is a hard look no matter what the Service sees, even if judges see something else.
The majority acts as both legislature and biologist. It acts as legislature by inventing new NFMA requirements and as biologist by dissecting reports about sage grouse in a misguided effort at second-guessing those reports’ authoring scientists. We should abstain from this sort of law office science. The Service has already printed hundreds of pages analyzing the Antelope Basin/Elk Lake’s suitability for summer grazing. Both NFMA and NEPA were satisfied. We have no authority to stand in the way.