OPINION

Between the late 1800s and the early 1900s, the number of bighorn sheep in North America declined dramatically, falling from a high of 1.5 to 2 million individuals to approximately 10% of that, number. Scientists have generally attributed the decline to over-harvesting, habitat loss, competition for food, and disease transmission from domestic sheep.

In response to concerns regarding disease transmission to immunologically vulnerable bighorn sheep, the Chief of the U.S. Forest Service ordered further analysis of the effects of grazing domestic sheep in the Payette National Forest of west-central Idaho (“Payette”). In response, the Forest Service prepared a draft supplemental environmental impact statement (“DSEIS”), an update to' the DSEIS, and, eventually, a final supplemental environmental impact statement (“FSEIS”) and Record of Decision (“ROD”). Concluding that there is a significant risk of fatal disease to the small and insular populations of bighorn sheep in the Payette, the Forest Service decided in the ROD to close to domestic sheep grazing approximately 70% of the allotments on which grazing had been permitted.

The Idaho Wool Growers Association, other state and national trade associations, and two" sheep ranchers (collectively, “Wool Growers”) challenged the Forest Service’s decision, objecting to the Forest Service’s (1) failure to consult the Agricultural Research Service (“ARS”), a federal agency within the U.S. Department of Agriculture, before preparing the FSEIS and ROD; (2) failure to supplement the FSEIS and ROD in light of the publication in 2010 of a certain study of the transmission of disease from domestic to bighorn sheep, the “Lawrence study”; and (3) choice and use of particular models to evaluate the risk of contact between domestic and bighorn sheep and the effects of disease transmission from domestic to bighorn sheep. The district court entered summary judgment in favor of the Forest Service. Wool Growers appealed.

We conclude that any error in failing to consult ARS was harmless. As the Forest Service did not otherwise act arbitrarily or capriciously or abuse its discretion, we affirm.

I.

A. Background

Bighorn sheep are currently found in two populations in the Payette—one in Hells Canyon and the other in the Salmon River Mountains. In approximately 1870, major die-offs of bighorn sheep began tó occur in the Salmon River Mountains. The die-offs roughly coincided with the onset of wide-spread grazing of domestic sheep in the Payette.

Over the years, the Payette’s bighorn populations have continued to experience periodic “large-scale, rapid, all-age die-offs.” FSEIS xx; see also ROD 6-7. Since 1981, despite efforts at transplanting sheep from elsewhere, the total population in the Payette has decreased by 47%. Although at one time more than 10,000 bighorn sheep lived in Hells Canyon and the surrounding mountains, “they were extirpated by the mid-1940s.” ROD 6. Between 1971 and 2004/ 474 bighorn sheep were transplanted into Hells Canyon. Seven die-offs in the Hells Canyon population have been reported since 1971. At the time the FSEIS was written, that population numbered 850 sheep.

The Salmon River population was never extirpated. According to surveys in 2001, 2003, and 2004, that population numbers roughly 700 sheep.

B. The NEPA Environmental Review Process

The administrative process underlying this appeal began in 2003, when the Forest Service, pursuant to the National Environmental Policy Act (“NEPA”), issued the Southwest Idaho Ecogroup Land and Resource Management Plans" Final Environmental Impact Statement and Record of Decision (“EIS”), which revised the 1988 Payette National Forest Land and Resource Management Plan. The EIS was appealed, appellants urging that the EIS “violated the [National Forest Management Act] and Hells Canyon National Recreation Area (HONRA) Act on the Pay-ette National Forest .by providing for grazing of domestic sheep within or near the range of bighorn sheep, thus threatening the viability of bighorn sheep though [sic]. disease transmission.” ROD 1. In March 2005, the Chief of the Forest Service agreed .that the EIS “did not adequately address viability [of bighorn sheep populations in the Payette] or the potential' for disease transmission.” Id., The Chief therefore rejected the EIS’s analysis.

‘The Chief then “instructed the Regional Forester to reanalyze the potential impacts of domestic sheep grazing on bighorn sheep-viability on the Payette National Forest to ensure'habitat is available to support a viable population of bighorn sheep.” ROD 1. The Chiefs decision reflected general concerns regarding disease transmission—in particular, the spread of various strains of pneumonia-causing or-contributing bacteria—from domestic to bighorn sheep, as confirmed by anecdotal evidence and a multitude of studies.

There is uncertainty regarding the particular mechanics of disease transmission, and the evidence of transmission is largely circumstantial. Pneumonia-causing bacteria are commonly found in domestic sheep, with, the worst outbreaks- killing 2.5% of domestic sheep in a herd. The impact of pneumonia on bighorn sheep is considerably more catastrophic. Episodic pneumonia outbreaks appear to be the current limiting factor in bighorn sheep abundance and distribution, both because large-scale die-offs caused by pneumonia kill most or all of a given population, and because female bighorns who survive die-offs experience low reproduction and high lamb mortality rates for years following an outbreak. Consequently, a number of state and federal agencies with jurisdiction over bighorn sheep have expressed concern and modified their management plans to address disease transmission from domestic to wild sheep. ROD 6; see, e.g., John Beecham et al., Rocky Mountain Bighorn Sheep (Ovis Canadensis): A Technical Conservation Assessment, Feb. 12, 2007; Timothy Schommer & Melanie Woolever, A Review of Disease Related Conflicts Between Domestic Sheep and Goats and Bighorn Sheep, 2008; Montand Department of Fish, Wildlife and Parks, Montana Bighorn Sheep Conservation Strategy, Draft, August 2009; Idaho Department of Fish & Game, Draft Bighorn Sheep Management Plan, May 2010. . As the FSEIS explained, “[scientists from both sides of the issue .... recommend that the .species be kept separate until the disease transmission science is better understood.” ROD 11.

In September 2008, the Forest Service released the DSEIS, which “proposed to modify, delete, and add to the current Forest Plan direction in response to the Chiefs instructions.” ROD 1. The DSEIS precipitated over 14,000 public comments.

In 2009, because of declining bighorn sheep numbers, the Forest Service designated bighorn sheep a sensitive species in the Intermountain Region, which includes the Payette. Under that designation, the objectives of bighorn management “are to prevent listing under the Endangered Species Act (ESA), avoid or minimize .impact to [the] speciesf,] whose viability has been identified as a concern, maintain viable populations of [the] species, and to develop and implement management objectives for [the species’] populations and habitat.” ROD 8.

The Forest Service released an update to the DSEIS in January 2010, which “provided interested stakeholders and the public an opportunity to review and comment on improved analyses and alternatives.” ROD 1. In the update, the Forest Service responded to criticisms of the DSEIS’s qualitative analysis by providing a quantitative analysis. See FSEIS xvii; e.g., A-85. That analysis used modeling to study the potential risks and effects of contact and disease transmission. More than 11,-000 comments on the update to the DSEIS were submitted.

In July 2010, the Forest Service completed and released the FSEIS and ROD. To analyze the effects of various alternative plans on the Payette’s bighorn sheep, the Forest Service used three models developed in conjunction with researchers at the University of California, Davis Center for Animal Disease Modeling and Surveillance. The models—called the source habitat, risk of contact,, and disease models— incorporated telemetric location data, including more than 54,000 data points collected in the Payette over the course of twelve years from over 400 radio-collared bighorn sheep. Even with the use of these models, the Forest Service recognized, “[determining the probability that a bighorn sheep will reach an occupied [domestic sheep grazing] allotment [in the Pay-ette] and that contact between the species will result in disease transmission is problematic,” and “there is ... essentially no research that would allow [ ] estimation” of the likelihood of contact causing disease transmission. ROD 12. To account for that difficulty, the Forest Service ran the disease model using a range of probabilities of contact resulting in disease transmission—5%, 10%, 25%, 50%, 75%, and 100%. This approach allowed the Forest Service to assess the effects of various management alternatives on bighorn sheep subpopulations despite its inability to estimate accurately the degree of risk of disease-conveying contact.

After applying the models, the Forest Service selected one of the alternatives analyzed in the FSEIS, Alternative 70. Alternative 70 would result in the termination of domestic sheep grazing on approximately 69,000 acres in the Payette, with' implementation to occur over three years. The models showed that this alternative reduced the risk of extirpation of all but one 'of the bighorn subpopulations to low-to-moderate at the 5% transmission level; the models predicted a 100% likelihood of extirpation of the Sheep Mountain subpopulation at all transmission levels, due to that subpopulation’s demographics and proximity to domestic sheep grazing allotments. At the higher transmission levels, the probabilities of extirpation under Alternative 70 ranged from 4% to 76%. Ultimately, the Forest Service found that large distances between domestic and bighorn sheep are necessary to assure protection of bighorns from disease, as bighorn sheep travel long’ distancés across rugged terrain; domestic sheep are known to stray from their herds, remaining on allotments at unpermitted times; and bighorn and domestic sheep are attracted to each other and seek out each other’s company.

The FSEIS incorporated findings from the then-unpublished Lawrence study, the first experiment to demonstrate directly that transmission of bacteria from domestic to bighorn sheep is implicated in causing pneumonia in bighorn sheep. The study’s authors extracted bacteria from domestic sheep, genetically modified the bacteria to be antibiotic' resistant and to radioactively fluoresce (that is, give off light). The domestic sheep were then inoculated with the modified bacteria. The domestic sheep were initially placed in pens separated by ten meters from pens of bighorn sheep. Later, they were moved so that they were' separated only by a fence. Still later, all the sheep were commingled in a single pen.

• By the end of the study, three of the four bighorn sheep had contracted pneumonia and died, and the fourth was eu-thanized after displaying telltale and severe signs of pneumonia. Post-mortem examinations of the bighorn sheep showed that they carried the radioactively-labeled, antibiotic-resistant bacteria with which only the domestic sheep had been inoculated.

C. This Litigation

Wool Growers appealed the ROD and FSEIS to the Intermountain Regional Forester. During the appeal process, Dr. Donald Patrick Knowles, a co-author of the Lawrence study and a research scientist within ARS, submitted a letter discussing his interpretation of the study’s -results (“Letter”). Dr. Knowles maintained that (1) the study did not prove that fence-line contact between domestic and bighorn sheep resulted in transmission of pneumonia-causing bacteria between the species; (2) transmission of a disease-causing or ganism, as well as the mechanisms of disease, are complex processes; hnd (8) “transmission of an organism doesn’t necessarily lead to disease.” Letter 1-2. The Regional Forester, having considered Dr. Knowles’ input, denied Wool Growers’ appeal.

Wool Growers then filed a complaint for declaratory, and injunctive relief in the district court, challenging the Forest Service’s decision. The court granted In-tervenors, a group of. environmental organizations, leave to enter the action as defendants. Wool Growers moved to expand the record pursuant to Asarco, Inc. v. U.S. EPA, 616 F.2d 1153 (9th Cir. 1980). The district court granted the motion, permitting the parties to submit expert declarations in support of them cross-motions for. summary judgment. After a hearing, the district court entered summary judgment in favor of the Forest Service and Intervenors and dismissed the action.

Wool (Growers timely appealed the district court’s decision. On appeal, Wool Growers challenges three aspects of the Forest Service’s decision under NEPA: its (1) failure to consult ARS prior to completing the FSEIS; (2) failure to supplement the FSEIS; and (3) choice and use of models.

II.

NEPA is a purely procedural statute; it does not impose, any substantive requirements on an agency undertaking environmental review. Lands Council v. McNair, 537 F.3d 981, 1000 (9th Cir.2008) (en banc), overruled on other grounds by Winter v. Nat. Res. Def. Council, Inc,, 555 U.S. 7, 20 , 129 S.Ct. 365 , 172 L.Ed.2d 249 (2008). Rather, “NEPA aims to make certain that the agency will have available, and will carefully consider, detailed information concerning significant environmental impacts, and that the relevant information will be made available to the larger public audience.” Id. (citation and alterations omitted). Accordingly, federal agencies must “take a ‘hard lobk’ at the environmental consequences of their actions by preparing an EIS for each ‘major Federal action significantly affecting the quality of the human environment.’ ” Id. at 1000-01 (alteration omitted) (quoting 42 U.S.C. § 4332 (2)(C)). An EIS “must ‘provide [a] full and fair discussion of significant environmental impacts’ so as to ‘inform deci-sionmakers and the public of the reasonable alternatives which would avoid or minimizé adverse impacts or enhance the quality of the human environment.’ ” Id. at 1001 (quoting 40 C.F.R. § 1502.1 ).

, Under NEPA and the Administrative Procedure Act .(“APA”), our review is limited to determining whether the Forest Service’s analysis, was “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” Id. at 987 (quoting 5 U.S.C. § 706 (2)(A)). We conclude that although the Forest Service may have acted contrary to NEPA in one respect, any such error was harmless. The Forest Service did not otherwise act arbitrarily or capriciously, nor did it abuse its discretion.

A. Consultation

1. Duty to consult

NEPA imposes on federal agencies conducting environmental review a duty to consult with certain other agencies. “Pri- or to making any detailed statement, the responsible Federal official shall consult with and obtain the comments of any Federal agency which has jurisdiction by law or special expertise with respect to any environmental impact involved [in the proposed action].” 42 U.S.C; § 4332(2)(C). Further,- to promote NEPA’s policies of public participation and informed decision-making, copies of the EIS and comments thereon from other agencies “shall accompany the proposal through the existing agency review processes.” Id.

The regulations implementing these provisions state that “[a]fter preparing a draft environmental impact statement and before preparing a final environmental impact statement the agency shall .,. [obtain the comments of any Federal agency which has jurisdiction bylaw or special expertise with respect to any environmental impact involved....” 40 C.F.R. § 1503.1 (a)(1); see also id. §-1500.1(b) (“Accurate scientific analysis, expert agency comments, and public scrutiny are essential to implementing NEPA.” (emphasis added)). “Special expertise” is defined as “statutory responsibility, agency mission, or related program experience.” Id. § 1508.26. Under the statute and its implementing regulations, the Forest Service may have had a duty to consult with ARS before issuing the FSEIS.

The pivotal question is whether ARS has “special expertise” concerning one .significant aspect of the proposed decision, the mechanics of pathogen transmission in domestic sheep, Wool Growers, argues that it does. They note, for example, that 7 C.F.R. § 2.65 delegates to ARS, among other matters, the authority to “[c]on(juct research concerning domestic animals and poultry, their protection and use, [and] the causes of contagious, infectious, and communicable diseases.” Also, ARS’s mission statement proclaims: “ARS conducts research to develop and transfer solutions to agricultural problems of high national priority and provide information access and dissemination to ... enhance the natural resource base and the environment____” U.S. Department of Agriculture, Agricultural Research Service, ARS: About US, http://www.ars.usda.gov/aboutus/aboutus..-htm.

The Forest Service argues, in response, that it had no duty to consult with ARS because that agency has no expertise in wildlife management. Although ARS’s expertise does center on domestic, not wild, animals, the development within and movement of pathogens in domestic sheep is of some relevance to concerns regarding disease transmission to bighorn sheep. The Forest Service’s assessment of the pertinence of that expertise may be too narrow an interpretation of its consultation duty under NEPA. And the language establishing NEPA’s consultation requirement is expansive. It mandates consultation with any federal agency that has “special expertise with respect to any , environmental impact involved.” 42 U.S.C. § 4332 (2)(C) (emphasis added). See also. 40 C.F.R. § 1503.1 (a)(1) (“[T]he agency shall ... [o]btain the comments of any Federal agency which has jurisdiction by law .or special expertise with respect to any environmental impact involved....” (emphasis added)).

.Further, Warm Springs Dam Task Force v. Gribble suggests that for the consultation requirement to apply, the particular expertise of an agency does not have to encompass- the proposed project as a whole or the issue the proposed project was designed to address. Rather, the expertise need relate only to one, of the project’s anticipated environmental effects. See 621- F.2d .1017, 1020-21 (9th Cir.1980) (per curiam). . , .- .

In the end, we need not resolve in this case the precise parameters of the consultation requirement, or whether it extended to ARS on the record before us.- Any violation of the consultation duty that occurred here, we are persuaded, was harmless.

2. Prejudice

The APA directs us to take “due account ... of the rule of prejudicial error.” 5 U.S.C. § 706 ; see Nevada v. Dep’t of Energy, 457 F.3d 78, 90 (D.C.Cir.2006). Accordingly, we consider whether the Forest Service’s failure to consult was harmless. See Warm Springs, 621 F.2d at 1022-23. The harmless-error analysis asks whether the failure to consult materially impeded NEPA’s goals—that is, whether the error caused the agency not to be fully aware of the environmental consequences of the proposed .action, thereby precluding informed decisionmaking and public participation, or otherwise materially affected, the substance of the agency’s decision. See Tucson Herpetological Soc. v. Salazar, 566 F.3d 870, 880 (9th Cir.2009).

Here, Wool Growers contends that, had consultation occurred, ARS and Dr. Knowles would have conveyed to the Forest Service information regarding the uncertainties of disease transmission mechanisms from domestic sheep and other contributors to bighorn disease. It is fair to assume Dr. Knowles would have offered this information—he did, offer it to the Forest Service, both in his letter on appeal to the Forest Supervisor, as well as in his declarations in the expanded record before the district court. But, as reflected in the FSEIS, such information had already been amply communicated, and the Forest Service and the public already had considered it.

For example, the FSEIS acknowledged:

FSEIS xxi. The FSEIS also specifically acknowledged comments by “scientists and others, primarily from agricultural disciplines,” regarding the effect of stressors on bighorn sheep disease, as well as uncertainties regarding disease transmission. Id. at 3-12-3-14. And in the ROD, the Forest Supervisor took those views and comments into consideration, stating: “Some scientists and others, primarily from agricultural disciplines, contend that disease transmission between bighorn sheep and domestic sheep is not a relevant factor in bighorn sheep distribution and population declines in the wildland environments I have taken these arguments into consideration while making my decision.” RODIL

Nor did these' concerns escape public comment.' The public provided extensive comments showing awareness of these' considerations, including the comments that: “The FSEIS should be particularly thorough when assessing the literature on the causal relationship of disease transmission between domestic sheep and bighorn sheep,” FSEIS A-81; “[e]vidence linking disease outbreaks to domestic sheep is inconclusive,” “[s]tudies documenting disease transmission between the two species have been done in controlled environments,” and “[s]tress could be a precursor to the onset of sickness”, id. at A-87; and “[t]he Forest Service-should disclose plans for reduction or elimination of stressors, such as bad weather or lack of nutrition, that can predispose bighorn sheep to disease or exacerbate risk of mortality,” id. at A-186.

In Warm Springs, we found no prejudice arising from the Army Corps of Engineers’ failure to consult the U.S. Geological Survey because the Corps subsequently considered the study that consultation would have revealed. See 621 F.2d at 1028. Similarly here, the Forest Service’s failure to consult did not prejudice its review process. The Forest Service took into account essentially the same' comments—arguments from “scientists and others, primarily from agricultural disciplines” contesting the inference that domestic sheep can and do transmit disease to wild bighorn sheep—that ARS, through Dr. Knowles, would have'provided by formally consulting with the Forest Service during the environmental review process.

The Forest Service’s failure to consult ARS is immaterial for another reason'as well. The precise mechanisms of disease transmission did not affect the Forest Service’s decision. As the FSEIS states: “The exact means by which the disease is transferred [from domestic to bighorn sheep] is beyond the scope of this analysis and should be conducted by qualified researchers.” FSEIS A-187. The FSEIS’s conclusion was that the scientific consensus is that disease transmission from domestic sheep—by whatever mechanism and involving whatever confounding- factors—poses a sufficient risk to bighorn sheep viability to merit separation of the bighorns from the domestic animals. Confirming that conclusion, numerous agencies with jurisdiction over bighorn sheep have found the risk significant enough to compel modification of their management plans. See section I.B., supra.

Because the lack of consultation did not prevent the Forest Service or the public from considering information about the uncertainties' in transmission of disease from domestic to bighorn sheep such as ARS would have offered (and which Dr. Knowles later did offer), and’because-information about the precise mechanisms of such transmission was not a basis of the Forest-Service’s-decision, no prejudice resulted from the lack of consultation.

Wool Growers maintains that its member wool companies were prejudiced— some of them went out of business due to the Forest Service’s ultimate decision, they claim. But this is not the type of prejudice with which NEPA is concerned. Rather, the question is whether the failure to consult somehow-materially altered the environmental review process, not whether a constituent body was harmed by the agency’s ultimate decision. See Cnty. of Del Norte v. United States, 732 F.2d 1462 , 1466-67 (9th Cir.1984). Almost invariably, some individuals or entities are negatiyely affected by an agency decision. As long as “[t]he integrity of the decision making process within the government and the public’s opportunity to comment in accordance with all legal requirements were not com^ promised in any way,” id., there is no prejudice of the sort with which NEPA is concerned.

Accordingly, we conclude that any error arising from the Forest Service’s failure to consult was harmless.

B. Supplementation

Wool Growers next challenges the Forest Service’s failure to supplement the FSEIS. “Agencies ... [s]hall prepare supplements to either draft or final environmental impact statements if ... [tjhere are significant new circumstances or information relevant to environmental concerns and bearing on the proposed action or its impacts.” 40 C.F.R. § 1502.9 (c). But “an agency need not supplement an EIS every time new information comes to. light after the EIS is finalized.” Marsh v. Or. Nat. Res. Council, 490 U.S. 360, 373 , 109 S.Ct. 1851 , 104 L.Ed.2d 377 (1989). So requiring “would render agency decisionmaking intractable, always awaiting updated information only to find the new information outdated by the time a decision is made.” Id. Thus, only “if the new information is sufficient to show that the remaining action will affect the quality of the human environment in a significant manner or to a significant extent not already considered, [must] a supplemental EIS ... be prepared.” Id. at 374 , 109 S.Ct. 1851 (alteration and citation omitted).

Here, Wool Growers contends that publication of the Lawrence study in 2010, after the FSEIS issued, triggered the Forest Service’s duty to supplement. But: (1) the Forest Service cited and discussed the study, in unpublished form, numerous times in the FSEIS; (2) the study bolstered the Forest Service’s decision by confirming that pneumonia-linked bacteria were transmitted from domestic to bighorn sheep, and that the transmitted bacteria likely caused the pneumonia that killed the bighorns in the study; and (3) the study, while confirming bacterial transmission between the species in a manner, not yet definitively .proven, reaffirmed the scientific consensus regarding the risk domestic sheep present to bighorn sheep, as discussed in the FSEIS and ROD.

The FSEIS stated, for example, that “the vast majority of literature supports the potential for disease transmission between the species, documents bighorn die-offs near domestic sheep, and supports the management option of keeping these species separate to prevent disease transmission.” FSEIS 3-14. The Lawrence study was fully in accord with the literature reported, and indeed was taken into account by the FSEIS, in making that statement. See id. It was eminently reasonable for the Forest Service to determine that the publication of the Lawrence study in essentially the form relied upon in the FSEIS did not provide significant information not already considered.

Wool Growers also argues that the Forest Service should have supplemented the FSEIS in light of Dr. Knowles’ critiques of the Lawrence study. But Dr. Knowles’ critiques either: (1) conflict with the express findings of the study, of which he was an author; or (2) state once more his views regarding the lack of certainty in disease transmission mechanisms. As to the former, it was reasonable for the Forest Service to decide that certain of Dr. Knowles’ comments—which contradict the study’s conclusion that its “results unequivocally demonstrate transmission of M. haemolytica from domestic to bighorn sheep, resulting in pneumonia and death of bighorn sheep”—did not represent “significant” new information. The responses from other authors of the Lawrence study disputed Dr. Knowles’ characterization of the results of the study, indicating that the Forest Service reasonably concluded that Dr. Knowles’ critiques did not undermine the relevance of the study’s conclusion. And as to the latter, the Forest Service could reasonably have decided that Dr. Knowles’ comments regarding the unproven and unknown mechanisms. of disease transmission were already well addressed in the FSEIS, as discussed above, and therefore that his additional comments to the same effect did not trigger its duty to supplement the FSEIS. Supplementation is not required where the agency, having taken a “hard look” at reevaluation, “determines that the new impacts will not be ... significantly different from those already considered.’’ N. Idaho Cmty. Action Network v. U.S. Dep’t of Transp., 545 F.3d 1147, 1154-55 (9th Cir.2008) (per curiam).

We conclude that the Forest Service did not act arbitrarily or capriciously or abuse its discretion by declining to supplement the FSEIS.

C. Models

Wool Growers’ final challenge is to the modeling the Forest Service used to analyze bighorn sheep home ranges and movement, and the potential impacts of various management alternatives.

The framework for evaluating such challenges is well-established. “Agencies [must] insure the ... scientific integrity[ ] of the discussions and analyses in environmental impact statements.” 40 C.F.R. § 1502.24 . Moreover, “NEPA requires that [EISs] contain high-quality information and accurate scientific analysis.” Lands Council v. Powell, 395 F.3d 1019 , 1031 (9th Cir.2005) (citing 40 C.F.R. .§ 1500.1(b)). If relevant data is incomplete or unavailable, the EIS “must disclose this fact.” Id. Thus, when an agency uses models in its NEPA analysis, it must provide “up-front disclosures of relevant shortcomings in the data or models.” Id. at 1032. When an agency undertakes technical scientific. analyses, as with the development, of models to help analyze a problem, the court’s deference to the agency’s judgment is at its peak. Lands Council, 537 F.3d at 993 .

1. Risk of contact model

The first model with which Wool Growers takes issue is the risk of contact model. Wool Growers argues that model failed, adequately to take into account obstacles to bighorn sheep mobility, including rivers and mountains.

The.Forest Service used one aspect of the risk of contact, model—the core herd home range analysis—to map the home ranges for the fifteen herds in the Payette, as well as one “area of concern.” The home range boundaries were developed using telemetry data from radio-collared bighorn sheep. A second aspect of the model, the foray analysis, was, used to map the distances bighorn sheep travel outside their home ranges. The Forest Service used the model, overall, to predict whether a given foray by a bighorn from its core home range would intersect a grazing allotment between May and October, the months during which domestic sheep were permitted to graze in the Payette. Contrary to Wool Growers’ argument, the risk of contact model does take into account obstacles to mobility. The model incorporates telemetry data of actual bighorn sheep movements within the Payette) That data necessarily accounts for any topographical features that impede bighorn mobility, as it indicates the distances bighorn sheep actually move across the landscape. Also, bighorns’ chosen habitat is rugged, including steep, rocky terrain that permits them to escape from predators. Thus, much of the Payette’s mountainous terrain does not serve as a barrier to bighorn sheep movement, although it might be a barrier to the movement of other species.' And the Forest Service has pointed to evidence indicating that rivers are not obstacles to bighorn sheep, as they have been known to swim across the largest ones in the region. See W. Watersheds Project v. BLM, No. 09-0507-E-BLW, 2009 WL 3335365 , at *4 (D.Idaho Oct. 14, 2009). Wool Growers’ argument that the risk of contact model failed to account for barriers to movement therefore fails.

Wool Growers next contends that ’the risk of contact model should have been validated by comparing the model’s predicted values with actual values to assess the model’s reliability! Neither the Forest Service nor Intervenors directly respond to this argument, although both generally note that the models: (1) were developed by leading experts in the field; (2) were derived from peer-reviewed and published models used to study bighorn sheep movements in the Sierra Nevada mountains of California; and (3) rely on a large data set reflecting actual bighorn sheep movements in the Payette.

We reject Wool Growers’ argument. The Forest Service is owed greater-than-average deference as it relates to its choice of technical methodologies. . Also, as the methodologies used to construct the risk of contact model were peer-reviewed and used successfully elsewhere, it was not unreasonable for the Forest Service to rely on the model, adjusted to fit local circumstances in the Payette. The model was reliable as a predictor of actual movements because it was predicated on data depicting actual bighorn sheep movements. Given the model’s Payette data-based origin, the Forest Service could reasonably assume that its predictions were sufficiently reliable to satisfy NEPA.

Ultimately, the Forest Service used top-rate model designers; relied on peer-reviewed methodologies applied by other bighorn researchers addressing similar issues; and incorporated on-the-ground data of bighorn sheep movements within the Payette. Given the foregoing, and in light of the deference owed to the agency when undertaking technical analysis within its purview, the Forest Service’s reliance on the risk of contact model was not arbitrary, capricious, or an abuse of discretion. Cf. Lands Council, 537 F.3d at 990-94 (holding that the Forest Service did not act arbitrarily and capriciously by failing to verify its model with on-the-ground data).

2. Disease model

Wool Growers asserts two final points of error, related to the Forest Service’s disease model. Wool Growers contends the model did not consider the effects of time on disease transmission, including both when and for how long bighorn and domestic sheep will be in contact with one another in the Payette, as well as the precise timing and period of contact necessary for transmission to occur.

Determining both the amount of time bighorn and domestic sheep spend cohabiting and the effects of timing on disease transmission involve uncertainties acknowledged by the Forest Service in the FSEIJ3. The degree to which both kinds of uncertainty affect the usefulness of the disease model can be resolved by the same three responses, as the Forest Service explained in the FSEIS:

(1) The disease model was not designed to predict actual disease outbreak probabilities. Instead, the Forest Service used the model to predict the likelihood that extirpation of bighorn sheep subpopula-tions would result from various alternative management plans, assuming probabilities of disease transmission risk ranging from 5% to 100%. The Forest Service did not use the disease model to predict a particular level of actual risk.

(2) Because the steps of disease transmission are uncertain—including the duration of contact between bighorn and domestic sheep and the precise timing and coincidence of events necessary for transmission to occur—the Forest Service ran the model at multiple different probability values. By using the model to test various proposed plans at different risk levels, the Forest Service accounted for the uncertainty of the variables involved.

(3) Most importantly, the Forest Service clearly explained the assumptions on which it built the model and the uncertainties inherent in it, thereby identifying the model’s limitations. The FSEIS states, for example: “We do not understand all of the mechanisms involved in potential disease transmission between the species,” FSEIS 3-13; “[There is] sb much uncertainty surrounding [the probability of contact resulting in disease transmission] and essentially no research ... that would allow its estimation,” id. at 3-43; and “[t]he complexity of the [disease transmission] model and the number of variables whose estimation was necessary to run it ... imply a high degree of uncertainty of its results,” id. at 3-56. Those explanations and acknowledgments are all that NEPA requires. Powell, 395 F.3d at 1031-32 (citing 40 C.F.R. § 1500.1 (b)). Were that not the case, government actions affecting the environment, positively or negatively, could be hamstrung by the need for-unattainable scientific certainty.

Wool Growers points to additional available information regarding domestic sheep movement within allotments—grazing permits, operating instructions, and post-season actual-use reports—the Forest Service could have used to evaluate more accurately the likelihood that bighorn sheep would cohabit with domestic sheep-. But as the FSEIS explained, so much uncertainty surrounds contact probabilities that the Forest Service chose not to estimate the likelihood of actual contact. Specifically, the FSEIS observed that, while this information was relevant, it could not be incorporated into the modeling because domestic sheep: (1) graze differently on the allotments each year; and (2)- could be anywhere in the Payette between May and October, and could stray during that time or be left behind after the grazing season. Given these uncertainties, the Forest Service did not -act unreasonably by using a range of probabilities to model the risk of disease transmission.

Wool Growers’ contention that the Forest Service should have modeled a disease transmission probability of 0% fares no better. If the disease transmission probability used was 0%, the models would necessarily show a 0% risk of extirpation due to disease -transmission for any possible management plan. Because of the obviousness of this outcome, a 0% value would not inform the environmental review process. Perhaps the Forest Service could have used a 1% probability to flesh out more fully the effects of disease transmission, given various management alternatives. But in the face of competing reasonable methodologies, we do not substitute our judgment for that of the agency. That the Forest Service could have chosen a different methodology does not render the extensive analysis it . undertook arbitrary or capricious.

In sum, the disease model was' used for a limited purpose, and, for that purpose, was sufficiently grounded in scientifically acceptable methodology to shed some light on the choice of an appropriate management plan. As the ROD explained: “Determining the probability that a bighorn sheep will reach an occupied [domestic sheep grazing] allotment [in the Payette] and that contact between the species will result in disease transmission is problematic,” partly because “there is ... essentially no research that would. allow ... estimation” of the likelihood of disease transmission. ROD 12. Given the Forest Service’s open acknowledgment of the model’s limitations, the uncertainties, inherent in estimating contact and disease transmission, and the actual use pf the model—which was not to estimate actual probabilities of disease transmission—the Forest Service’s development and use of the disease model was not arbitrary, capricious, an abuse of discretion, or otherwise contrary to law.

III.

In light of the foregoing, we conclude that the Forest Service committed no reversible error in preparing the FSEIS and ROD. Accordingly, we AFFIRM.

. Under Asarco, an administrative record can be expanded upon review by a federal court to explain what’s already in the record but not to add substantive evidence. 616 F.2d at 1159-60 . . Because we decide that any consultation error was harmless, see section H.A.2., infra, we do not address the Forest Service’s argument that the “special expertise” question does not apply to Dr. Knowles as an individual. • . The Forest Service also argües that any prejudice was cured by the thorough vetting in the district court of disease transmission questions, as expressed in the parties’ expert declarations. This extra-record evidence does not cure the prejudice—the question of harmlessness is as to the effect of an error on the NEPA process. The expert declarations were submitted well after the NEPA process was complete. They cannot, therefore, cure any prejudice, although they are informative as to whether prejudice occurred. . Wool Growers has -not argued that the -published version of the study differs materially from the unpublished form to which the FSEIS cited, and the Forest Service maintains there is no material difference. . The FSEIS also notes that, for various reasons, the telemetry data does not include some of the longest forays. It therefore underestimates the distances moved by bighorn sheep and thus the extent to which they are able to traverse challenging terrain. See FSEIS 3-35-36. . Wool Growers also challenge the risk of contact model on the ground that two of the Forest Service’s experts expressed conflicting views regarding whether the model takes into account barriers to bighorn sheep . forays. Even assuming the experts' statements conflict—and it is not clear they do—those statements, which were presented only at the district court, did not in any way affect the analysis in the FSEIS, which we find reasonable. . The peer-reviewed Clifford study of bighorn sheep in the Sierra Nevada assumed that, given contact between domestic and bighorn sheep, .the likelihood of disease transmission ranged from 50% to 100%. Here, the Forest Service Chose much more conservative probabilities for use in its disease model.