Dissent by Judge Callahan
OPINION
MURGUIA, Circuit Judge:Plaintiffs Turtle Island Restoration Network and the Center for Biological Diversity challenge the decision of the National Marine Fisheries Service (“NMFS”) to allow a Hawaii-based swordfish fishery to increase its fishing efforts, which may result in the unintentional deaths of endangered sea turtles. Plaintiffs also challenge the decision of the U.S. Fish and Wildlife Service (“FWS”) to issue a “special purpose” permit to the NMFS, which authorizes the fishery to incidentally kill migratory birds.
Plaintiffs brought suit against the agencies under various environmental statutes that the NMFS and the FWS are charged with administering, including'the Magnu-son-Stevens Fishery Conservation and Management Act (the “Magnuson-Stevens Act”), the Endangered Species Act of 1973 (“ESA”), the Migratory Bird Treaty Act (“MBTA”), and the National Environmental Policy Act (“NEPA”). The Hawaii Longline Association subsequently intervened to represent the interests of the swordfish fishery in defense of the agencies’ actions. We have jurisdiction under 28 U.S.C. § 1291, and we affirm in part, and reverse and remand in part.
BACKGROUND
7. Regulatory Fmmetoork
In response to concerns about overfishing, Congress enacted the Magnuson-Ste-vens Act to promote- the long-term biological and economic sustainability of marine fisheries in U.S. federal waters. See 16 U.S.C. § 1801(b). Under this Act, the NMFS and eight regional councils devélop “management plans” for the nation’s fisheries, which the Secretary of Commerce may approve, partially approve, or reject. Id. §§ 1801(b)(4), 1852(h)(1), 1854(a)(3). The Magnuson-Stevens Act demands that a management plan be consistent with the national standards set out in the Act and “any other applicable law,” id. § 1853(a)(1)(C), including the ESA, id. §§ 1531-43, and the MBTA, id. §§ 703-12.
The ESA provides for the conservation of fish, wildlife, and plant species that are at risk of extinction by requiring federal agencies to ensure that actions they authorize, fund, or carry out are “not likely to jeopardize the continued existence” of any ESA-listed ' species. 16 U.S.C. § 1536(a)(2). Agencies proposing actions that may affect an ESA-listed species must consult with either the NMFS or the FWS—depending on ■ the species involved—which then reviews the proposed action and prepares a "biological opinion” (“BiOp”) that evaluates whether and the extent to which the action may impact the species. Id. § 1536(b); 50 C.F.R. § 402.12. If the NMFS or the FWS finds that the proposed action would riot jeopardize any species’ continued existence, it issues a statement permitting the “taking” 6f a particular number of protected animals “if such taking is incidental to, and not the purpose of, the carrying out of' an otherwise lawful activity.”' 16 U.S.C. § 1539(a)(1)(B).'
The FWS also has authority to enforce the MBTA, id. §§ 703-12; 50 C.F.R. § 10.1, which strictly prohibits the taking of any migratory bird the Act protects except under the terms of a valid permit issued by the Secretary of the Interior, id. § 703(a). The Secretary of the Interior has issued regulations authorizing various types of exemptions to the MBTA permitting the taking of migratory birds under certain circumstances. See 16 U.S.C. § 704(a).
In addition to the substantive mandates of the ESA and the MBTA, both the NMFS- and the FWS are subject to NEPA’s procedural requirements. See Robertson v. Methow Valley Citizens Council, 490 U.S. 332, 348, 109 S.Ct. 1835, 104 L.Ed.2d 351 (1989). NEPA is- concerned with process alone and “merely prohibits uninformed—rather than unwise—agency action.” Id. at 351, 109 S.Ct. 1835. NEPA requires federal agencies to prepare environmental impact statements (“EIS”) ■ detailing the effects of any proposed action-that stands to have a significant impact on the environment. See 42 U.S.C. § 4332(C); Robertson, 490 U.S. at 350, 109 S.Ct. 1835. An agency may also prepare an environmental assessment (“EA”) to determine whether an EIS is needed. 40 C.F.R. §§ 1501.4(b), 1508.9(a)(1); Te-Moak Tribe of W. Shoshone of Nev. v. U.S. Dep’t of Interior, 608 F.3d 592, 599 (9th Cir. 2010). If the EA shows that the proposed action may significantly affect the environment, then the agency must prepare a full EIS. W. Watersheds Project v. Abbey, 719 F.3d 1035, 1050 (9th Cir. 2013). Otherwise, the agency issues a finding of no significant impact and the proposed action can proceed without further study. Id.
II. The Hawaii-Based Longline Fishing Industry
“Longline” fishing is a. commercial fishing method that involves -reeling out—or. “setting”—a single, horizontal-mainline to which shorter “branchlines” are attached at intervals. Each dangling branchline carries baited .hooks. .A typical longline set can use several hundred or thousand individual hooks, allowing a single fishing vessel to spread its efforts over a large area. While .the mainline is in the water* • the fishing equipment often ensnares birds, sea turtles, and other marine wildlife in addition to the target fish. This incidental taking of non-target animals is known as “bycatch.”
The NMFS collects bycatch statistics by tracking the number of times a non-target animal is hooked or entangled by fishing gear. The most commonly observed non-target animal interactions are with Northern Pacific loggerhead and leatherback sea turtles, both of which are currently listed under the ESA as “endangered.” See' 50 C.F.R. § 17.11. In addition, several types of albatross interact often with the longline fisheries, including the black-footed albatross and the Laysan albatross.
There -are two separately regulated longline fisheries based out of Hawaii: the deep-set fishery—-which targets tuna—and the shallow-set fishery, which targets swordfish. The two fisheries are managed by the Fishery Ecosystem Plan for Pelagic Fisheries of the Western Pacific Region (“Pelagics FMP”), developed by the-Western Pacific Fishery Management Council (“Council”) in accordance with the Magnu-son-Stevens Act and implemented by the NMFS. In 2001, the shallow-set • fishery was elosed by court order due to the NMFS’s failure to prepare an EIS analyzing the impact of longline fishing on the sea turtle population, which- the court found was a • violation of the agency’s NEPA .obligations. See Leatherback Sea Turtle v. Nat'l Marine Fisheries Serv., No. 99-00152, 1999 WL 33594329 (D. Haw. Oct. 18, 1999). In response, the NMFS issued an EIS and a BiOp in which the agency- concluded that the shallow-set fishery was adversely affecting several species of sea turtles. In -2002, the NMFS issued regulations prohibiting all Hawaii-based swordfish longlining.
The Council' subsequently developed various measures to minimize turtle by-catch, and in 2004 the NMFS reauthorized shallow-set longlining subject- to new restrictions designed to reduce the number and severity of interactions between protected turtles and fishing gear. In part, the NMFS strictly limited- the number- of interactions the fishery could have with leatherback and loggerhead sea turtles to a maximum of 16 and 17, respectively, per fishing season. Further, the NMFS imposed an annual limit of 2,120 shallow sets; which represents fifty percent of the average number of sets deployed prior to the fishery’s closure in 2001.
In 2008, the NMFS proposed an amendment to the Pelagics FMP (“Amendment 18”) that would remove the 2,120 annual set limit, allowing gear deployments to increase to- their pre-2001 máximums, and also increase the number of sea turtle interactions allowed-each year. After consulting internally pursuant to the ESA, the NMFS produced a BiOp concluding that Amendment 18 would not jeopardize the se’a turtles. The NMFS issued a final rule implementing Amendment 18 in December 2009. 74 Fed. Reg. 65,640 (Dec. 10, 2009).
Plaintiffs initiated suit against the NMFS on the grounds that the 2009 rule violated the ESA and the MBTA. See Turtle Island Restoration Network v. U.S. Dep’t of Commerce, 834 F.Supp.2d 1004, 1007 (D. Haw. 2011). Plaintiffs’ MBTA claim was based on the fishery’s incidental take of migratory seabirds without an MBTA permit. The parties settled the case, and the NMFS entered into a consent decree that required it to withdraw its no jeopardy BiOp, reinstate the 2004 annual turtle-interaction caps, and issue a new BiOp after deciding whether to reclassify various population segments of sea turtles under the ESA. Id. at 1023-25. The other remaining provisions of the 2009 rule remained in effect, including the removal of annual set limits.
The NMFS later proposed raising the shallow-set fishery’s annual turtle interaction cap to 26 (with leatherbacks), and 34 (with loggerheads) and otherwise continuing to operate the fishery in accordance with the provisions of Amendment 18 to the Pelagics FMP. In January 2012, the NMFS issued a new BiOp concluding that the shallow-set fishery would not jeopardize the continued existence of either the loggerhead or leatherback turtles if it operated under higher caps on turtle interactions.
While it was engaged in the re-consultation process, the NMFS submitted an application to the FWS for a special purpose permit that would allow the shallow-set fishery to take migratory seabirds in connection with swordfish longlining.- The FWS issued a final EA in which it considered denying the permit, granting the permit as requested, and granting the permit while requiring the NMFS to conduct new research on additional ways to avoid seabird interactions. See 77 Fed. Reg. 1501 (Jan. 10, 2012). The FWS ultimately concluded that none of the alternatives would have a significant adverse impact on the seabirds’ population levels. Accordingly, the FWS issued a finding of “no significant impact.” In August 2012, the FWS granted a three-year special purpose permit authorizing the shallow-set fishery to kill a maximum of 191 black-footed albatross, 430 Laysan albatross, 30 northern fulmars, 30 sooty shearwaters, and one short-tailed albatross. Of those birds, only the short-tailed albatross is listed under the ESA, 50 C.F.R. § 17.11(h).
Plaintiffs subsequently filed this lawsuit under the ESA, the MBTA, and their implementing regulations, challenging the NMFS’s final rule approving the continued operation of the shallow-set fishery and the FWS’s issuance of a migratory bird permit to the NMFS. After the parties moved for summary judgment, the district court ruled in the agencies’ favor on all of Plaintiffs’ claims. Plaintiffs timely appealed.
STANDARD OF REVIEW
We review challenges to final agency action decided on summary judgment de novo and pursuant to Section 706 of the Administrative Procedure Act (“APA”). Turtle Island Restoration Network v. Nat’l Marine Fisheries Serv., 340 F.3d 969, 973 (9th Cir. 2003). Review is based on the administrative record. Camp v. Pitts, 411 U.S. 138, 142, 93 S.Ct. 1241, 36 L.Ed.2d 106 (1973).
The APA requires courts to “hold unlawful and set aside agency action, findings, and conclusions found to be ... arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law,” “in excess of statutory jurisdiction,” or “without observance of procedure required by law.” 5 U.S.C. § 706(2)(A), (Q(D). “The scope of review under the ‘arbitrary and capricious’ standard is narrow and a court is not to substitute its judgment for that of the agency.” Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983). Nevertheless, we require the agency to “examine the relevant data and articulate a satisfactory explanation for its action,” and we will strike down agency action as “arbitrary and capricious if the agency has relied on factors which Congress has not intended it to consider, entirely failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency,” or if the agency’s decision “is so implausible that it could not be ascribed to a difference in view or the product of agency expertise.” Id.
Separate from the APA, we also give deference to an agency’s interpretation of the statutes and regulations that define the scope of its authority. Chevron, U.S.A., Inc. v. Natural Resources Defense Council, Inc. compels us to defer to an agency’s reasonable interpretation of its enabling legislation. 467 U.S. 837, 843, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). Under the Chevron analysis, we must first exhaust the traditional tools of statutory construction to determine whether Congress has “directly spoken to the precise question at issue.” Id. at 842,104 S.Ct. 2778. If we determine that the statute is silent or ambiguous on the question at hand, then at Chevron step two we must respect the agency’s interpretation so long as it “is based on a permissible construction of the statute.” Id. at 843, 104 S.Ct. 2778. A permissible construction is one that is not “arbitrary, capricious, or manifestly contrary to the statute.” Id. at 844, 104 S.Ct. 2778; see also Judulang v. Holder, 565 U.S. 42, 132 S.Ct. 476, 483 n.7, 181 L.Ed.2d 449 (2011) (recognizing that Chevron step two is equivalent to the APA’s arbitrary and capricious standard).
Chevron deference applies only to agency decisions rendered through formal procedures. United States v. Mead Corp., 533 U.S. 218, 226-27, 121 S.Ct. 2164, 150 L.Ed.2d 292 (2001). However, under Auer v. Robbins, we must also defer to an agency’s interpretation of its own ambiguous regulations, which controls unless “plainly erroneous or inconsistent with the regulation,” or where there are grounds to believe that the interpretation “does not reflect the agency’s fair and considered judgment of the matter in question.” Christopher v. SmithKline Beecham Corp., 567 U.S. 142, 132 S.Ct. 2156, 2159, 183 L.Ed.2d 153 (2012) (quoting Auer v. Robbins, 519 U.S. 452, 461-62, 117 S.Ct. 905, 137 L.Ed.2d 79 (1997)). Similarly, “we must ensure that the interpretation is not inconsistent with a congressional directive; a court need not accept an agency’s interpretation of its own regulations if that interpretation is inconsistent with the statute under which the regulations were promulgated.” Marsh v. J. Alexander’s LLC, 869 F.3d 1108, 1116-17 (9th Cir. 2017) (internal changes, quotation marks and citations omitted). Our review of an agency’s construction of a statute or regulation that does not qualify for either Chevron or Auer deference is de novo, although we may still accord the agency’s opinion some weight. Satterfield v. Simon & Schuster, Inc., 569 F.3d 946, 952-53 (9th Cir. 2009) (citing Skidmore v. Swift & Co., 323 U.S. 134, 140, 65 S.Ct. 161, 89 L.Ed. 124 (1944)).
DISCUSSION
I. “Special Purpose” Permit
Plaintiffs argue that the FWS acted arbitrarily and capriciously by issuing a special purpose permit to the NMFS on behalf of a commercial operation—the shallow-set fishery.—that provides no benefit to migratory birds. Plaintiffs specifically contend that, in issuing this permit, the FWS ignored or violated its obligations under the MBTA.
The MBTA is a strict liability criminal statute that Congress enacted for the “object and purpose ... to aid in the restoration of [game and other wild] birds.” 16 U.S.C. § 701. The MBTA states in expansive language that, unless otherwise permitted by the Secretary of the Interior, “it shall be unlawful at any time, by any means or in any manner, to pursue, hunt, take, capture, kill, [or] attempt to take, capture, or kill ... any migratory bird.” 16 U.S.C. § 703(a). The MBTA also limits the FWS’s authority to authorize the killing of migratory birds absent specified regulations “[sjubject to the provisions and in order to carry out the purposes of the conventions” underlying the Act. Id. §. 704(a). The conventions underlying the MBTA stipulate that migratory birds may only be killed under “extraordinary conditions,” where birds have “become seriously injurious to the agricultural or other interests in any particular community.” Humane Soc’y of the U.S. v. Glickman, 217 F.3d 882, 885 (D.C. Cir. 2000) (internal quotation marks omitted).
Pursuant to the MBTA, the FWS has enacted a permitting program for narrow categories of migratory bird takings, such as scientific collecting, rehabilitation, hunting, and depredation control. See 16 U.S.C. §§ 704(a), 712(2) (empowering the FWS to promulgate implementing regulations); 50 C.F.R. §§ 21.21-21.61 (authorizing the issuance of various types'of permits). The FWS has also established a “special purpose” permit that allows a person to “lawfully take ... migratory birds ... for any purpose not covered by the standard form permits” included elsewhere in the regulations. 50 O.F.R. § 21.27(a). The FWS may issue such a permit for “special purpose activities related to migratory birds,” where the applicant “makes a ■ sufficient showing” that the activity would be “of benefit to the migratory bird resource, important research reasons, reasons of human concern for individual birds, or other compelling justification.” Id.
Here, the FWS interpreted § 21.27 as authorizing it to grant a special purpose permit sanctioning the incidental take of migratory birds ¡to the NMFS, thereby allowing a commercial activity—longline fishing—that does not concern bird conservation. In its decision to issue the permit, the FWS found that the “commercial fishery carries no intrinsic benefit for migratory bird resources,” “the take that occurs is neither directed by, nor is the result of, important research,” and that “the take that occurs does not result from concern for individual birds.” However, the FWS found that “compelling justification!’ existed to permit the continued operation of the shallow-set fishery, which the FWS believed “provides a net benefit to the Nation” economically and “serves as a benchmark internationally for employing effective seabird mitigation techniques and serves as an example of responsible conservation practices by a fishery.” The FWS also noted that “[closure of this fishery would likely result in replaced effort by foreign longline fleets to supply swordfish demand, where use. of bycatch mitigation methods would not likely follow international best practices.”
We conclude that the FWS’s decision to issue a special purpose permit to the NMFS oh behalf of a commercial fishery was arbitrary and capricious. Although the FWS’s interpretation of § 21.27 would ordinarily deserve deference, see' Mead, 533 U.S. at 226-27, 121 S.Ct. 2164, we cannot conclude that such deference is appropriate in this case. Deference to the FWS’s interpretation is not warranted because the plain languagé of this regulation is not reasonably susceptible to the FWS’s new interpretation. The other “standard form permits” the MBTA regulations authorize govern discrete types of takings, such as scientific collecting, taxidermy, and rehabilitation, and although ■§ 21.27 is intended to allow the FWS to authorize activities not otherwise permitted by the regulations, it is still a narrow exception to the MBTA’s general prohibition on killing migratory birds. See Marsh, 869 F.3d at 1116-17 (“[W]e must always ensure that the interpretation is not inconsistent-with a congressional directive .... ”); Ctr. for Biological Diversity v. Salazar, 706 F.3d 1085, 1092 (9th Cir. 2013) (“[W]e must interpret [a] regulation as a whole, in light of the overall statutory and regulatory scheme ...(internal quotation marks omitted)). The FWS’s construction of § 21.27’s “special purpose activit[y]” exception as applying to basic commercial activities like fishing that have no articula-ble “special purpose” is therefore inconsistent with the existing permitting scheme that the FWS has enacted. The FWS must read the “special purpose’,’ provision in the context of the regulation’s other requirements that, taken together, fail to turn § 21.27 into a general incidental take exception: the permit must “relate[] to migratory birds” and may issue only upon a “sufficient showing of ... [a] compelling justification.” 50 C.F.R. § 21.27.
The FWS unpersuasively argues that the phrase “related to migratory birds” is not a restriction on its permitting authority, but merely a description of what can be permitted. The FWS specifically maintains that longline fishing is “related to migratory birds” because it incidentally interacts with them. Although nothing in the regulation requires that the permitted activity directly concern migratory birds, it nevertheless strains reason to say that .every activity that risks killing migratory birds “relate[s] to” those birds. See 50 C.F.R. § 21.27. The FWS’s approach to the regulation renders the majority of its text superfluous. See Nat’l Ass’n of Home Builders v. Defs. of Wildlife, 551 U.S. 644, 669, 127 S.Ct. 2518, 168 L.Ed.2d 467 (2007) (cautioning against reading an agency regulation in a way that renders part of it redundant).
The FWS’s interpretation of § 21.27 as authorizing it to grant an incidental take permit to the NMFS does not conform to either the MBTA’s conservation intent or the plain language of the regulation. We therefore conclude that the FWS’s grant of a special purpose permit to the NMFS was arbitrary and capricious.1
II. 2012 “No Jeopardy” BiOp
Plaintiffs also, argue that the NMFS violated the ESA by failing to properly assess the shallow-set fishery’s impacts on endangered sea turtles. The ESA permits federal agencies to authorize actions that will result in the taking of endangered or threatened species only if the projected take “is not likely to jeopardize the continued existence of’ any listed species. 16 U.S.C. § 1536(a)(2). “Jeopardize the continued existence of means to engage in an action that reasonably would, be expected, directly or indirectly, to reduce appreciably the likelihood of both the survival and recovery of a listed species in the wild by reducing the reproduction, numbers, or distribution of that species.” 50 C.F.R. § 402.02 (emphasis added).
Where listed marine species are concerned, the NMFS prepares a BiOp evaluating the effects of the proposed action on the survival, and recovery of the listed species. 16 U.S.C. § 1536(c). The agency specifically considers the proposed action’s direct, indirect, and cumulative effects on a listed species in relation to the environmental baseline, and opines on whether the action is likely to jeopardize the species’ survival. 50 C.F.R. § 402.14(g)(4); see also Nat’l Wildlife Fed’n v. Nat’l Marine Fisheries Serv., 524 F.3d 917, 924 (9th Cir. 2008). Where a species is already in peril, an agency may not take an action that will cause an “active change of status” for the worse. Nat’l Wildlife Fed’n, 524 F.3d at 930.
When formulating a BiOp, the NMFS must base its conclusions on evidence supported by “the best scientific and commercial data available.” 16 U.S.C. § 1536(a)(2); 50 C.F.R. § 402.14(g)(8). This requirement “prohibits [an agency] from disregarding available scientific evidence that is in some way better than the evidence [it] relies on.” San Luis & Deltar-Mendota Water Auth. v. Jewell, 747 F.3d 581, 602 (9th Cir. 2014) (citation and internal quotation mark omitted). “The determination of what constitutes the ‘best scientific data available’ belongs to the agency’s ‘special expertise’ ....” Id. (emphasis in original) (citation omitted).
In 2012, the NMFS issued a BiOp concluding that the removal of the annual limit of 2,120 shallow-set lines in the fishery might result in the incidental “take” of Northern Pacific loggerhead and leather-back sea turtles, but would not jeopardize the continued existence of either species for the next 25 years. To establish the environmental baseline, the NMFS used existing studies on loggerhead and leather-back interactions with all Pacific longline fisheries (domestic and international) from 2000 to 2009. The NMFS ultimately found that the Hawaii-based shallow-set fishery is currently responsible for killing two to three loggerheads and leatherbacks (each) per year. The NMFS also determined that the impacts associated with anthropogenic climate change were likely beginning to affect both sea turtle species, but lacked sufficient data to quantify the threat that climate change posed to the turtles.
The NMFS then attempted to predict the impact that allowing the fishery to deploy 5,500 longline sets per year—the approximate maximum annual number of sets before the fishery was first closed out of concern for the sea turtle populations— would have on the loggerheads and leath-erbacks. The NMFS ultimately projected that setting 5,500 lines would kill no more than one adult, female loggerhead turtle and four adult, female leatherback turtles. The NMFS then employed population viability assessment models to forecast the risk that killing small numbers of adult, female sea turtles would lead to the species’ extinction. The NMFS concluded from the results that the proposed action could not reasonably be expected to appreciably reduce the likelihood of survival of either the loggerhead or the leatherback turtles.
The NMFS’s “no jeopardy” conclusion was not affected by the agency’s consideration of the cumulative effects of worsening climates. And, the NMFS’s analysis of “spillover” trends suggested that the proposed increase in Hawaii-based swordfish-ing would benefit sea turtles overall. Because domestic fisheries operate under more stringent conservation measures than foreign fleets that compete to provide swordfish to U.S. consumers, the NMFS predicted that increasing domestic fishery yields would displace foreign fishing activities in the same area that the Hawaii-based shallow-set fishery operates, resulting in a net decrease in mortalities for the affected turtle species. However, because the NMFS concluded that the projected decrease in turtle deaths from this “spillover” effect was not precise enough to incorporate into its population assessment models, the NMFS did not incorporate these benefits into its no jeopardy finding.
A. Population Viability Assessment Models
Plaintiffs argue that the 2012 BiOp’s conclusion that the proposed action would not appreciably impact loggerhead and leatherback sea turtles is unsupported by the scientific methods the FWS relied on. To project the impact of the shallow-set fishery’s operations on the sea turtle species’ likelihood of survival, the NMFS ran a climate-based population forecast model and relied primarily on the results of this model, “along with inputs from multiple experts and sources, where available.” The climate-based model showed a significant decline in loggerhead numbers over the next generation even without the proposed action of removing the fishery’s set limits: 99.5% of the tests showed the loggerhead falling below the quasi-extinction threshold within 25 years. When the model was run incorporating the anticipated mortality associated with the fishery’s operations without set limits, the results were similar. The NMFS specifically found that “[v]irtually all the loggerhead climate model runs ... indicated] high extinction risk with high model confidence.” The additional loss to the loggerhead population from the proposed action ranged from 4 to 11%. As for the leatherback turtles, the climate-based model showed an increase in leatherback population over the next generation without a change in the fishery’s set limits, and even with the proposed action the “extinction risk remained] in the low category,” although the results predicted a “measurable loss to the population” of 16 to 30%.
Based on the results from the model, the NMFS decided that it did not “believe that the small effect posed by the lethal takes in this fishery, when considered together with the environmental baseline and the cumulative effects, will be detectable or appreciable” and “that the additional risk to the [loggerhead turtles] that would result from loss of one adult female annually is considered negligible.” Similarly, the NMFS concluded “that the proposed action would have a negligible impact on the risk to ... the western Pacific leatherback population as a whole.” Therefore, the NMFS opined that increasing the maximum annual number of sets at the fishery would not jeopardize either species.
1. Loggerhead Turtles
With respect to the loggerhead turtles, the NMFS violated the APA’s requirement that the agency articulate a rational connection between the population viability model upon which the NMFS relied and its no jeopardy conclusion. The BiOp acknowledged that the climate-based model predicted a decline in loggerhead populations to a level that “represents a heightened risk of extinction,” but still upheld a finding of “no jeopardy” on the grounds that there was “little to no difference in the extinction risk when the annual removal of one adult female loggerhead resulting from the proposed action is considered in the model.” We rejected similar logic in National Wildlife Federation, holding that “where baseline conditions already jeopardize a species, an agency may not take action that deepens the jeopardy by causing additional harm.” 524 F.3d at 930 (noting that listed species’ “slow slide into oblivion is one of the very ills the ESA seeks to prevent”). In National Wildlife Federation, the NMFS had prepared a BiOp in which it determined that hydro-power dam operations would not jeopardize threatened and endangered salmon populations. Id. at 925. NMFS, however, had already determined that baseline environmental conditions posed a risk of jeopardy to the species. Id. Therefore, to reach a conclusion of “no jeopardy,” the agency completely excluded from the environmental baseline all impacts from “nondiscre-tionary” federal activities such as operations relating to irrigation, flood control, and power generation. We held that this exclusion was improper and that baseline conditions must be factored into the jeopardy analysis, cumulatively with the entirety of agency actions. The relevant inquiry is therefore whether the “action effects, when added to the underlying baseline conditions,” are such that they would cause jeopardy. Id. at 929,
Here, the NMFS improperly minimized the risk of bycatch to the loggerheads’ survival by only comparing the effects of the fishery against the baseline conditions that have already contributed to the turtles’ decline. The BiOp’s no jeopardy opinion is premised on the proportionally low risk that the shallow-set fishery poses to the loggerheads relative to other threats, such as international fishing and climate change: the NMFS specifically found that although “any level of take and mortality can have an adverse effect on the overlying population ... the expected level of take from the action, including a small' number of mortalities, is extremely small when considered together with all impacts considered in the Status of the Species, Baseline arid Cumulative Effe'cts sections, includirig other federally authorized fisheries and foreign fisheries.” As in National Wildlife Federation, the agency reached an arbitrary conclusion by only comparing the prospective harm to the loggerheads that is attributable to the proposed action—the death of a single adult, female loggerhead per year—to the much greater harm resulting from factors-, beyond the fishery. Based on this impermissible comparison, the agericy concluded that- the proposed action’s adverse impacts would not appreciably reduce the loggerheads’ likelihood of survival-,. See' Nat’l Wildlife Fed’n, 524 F.3d at 930.
The NMFS relies heavily on the conservative nature of its calculations to support the difference between its conclusion and the climate-based model’s results. The NMFS asserts that it rounded up its calculation of maximum adult female -mortality, modeled the viability of turtle populations using the maximum potential number of annual interactions opposed to the average number of interactions reported in previous years, and estimated the number of sea turtle deaths based on the assumption that the shallow-set fishery would immediately operate at 5,500 sets- each year. In reality, the increase in sets is expected to be gradual over many years. The ESA, however, requires agencies to rigorously ensure their actions will not “tip [the loggerhead] species from a state of precarious survival into a state of likely extinction.” See Nat’l Wildlife Fed’n, 524 F.3d at 930. The agency may not reject the “best scientific data” in- favor of its belief that “incidental . take ... would be reduced to the best extent possible” and “the vast majority of the loggerhead sea turtle takes from the proposed action are expected to be non-lethal.”
The NMFS also notes that the climate-based model used an assumed fraction of the current turtle population size (50%) as a proxy for extinction, and explains that “population decline below that” number “does not necessarily mean that” the species is “unrecoverable” or “functionally extinct.” But, given the agency’s endorsement of the climate-based model and its expert’s decision to use a “quasi-extinction threshold” to" reflect a decline in the turtle population to numbers insufficient to ensure the population’s viability, this logic does not support the NMFS’s determination that the projected population declines would not . appreciably threaten the loggerheads’ survival.
Another rationale presented in the BiOp is-that “spillover effect is reasonably certain to contribute to a reduction in loggerhead mortalities ... due to reduced effort in foreign fisheries.” Shortly thereafter, however, the NMFS noted that data on foreign fishery bycatch are “likely incomplete or inaccurate.” The NMFS went on to state that “mortality reduction data associated with spillover effects are not -as robust as -those analyzed for the direct effects of the proposed action.” For those reasons, the NMFS did not incorporate the estimated sea turtle mortalities that would be avoided due to a potential spillover effect into its population assessment models.
The NMFS’s model showed the loggerhead species are on a path toward extinction, which accords with the fast that the NMFS recently raised the Pacific- loggerhead’s ESA listing from “threatened” .to “endangered.” The NMFS also found that “effects” to the loggerhead-“are likely to occur as a result of worsening climate change,” which the NMFS “expect[s] to continue and therefore may impact sea turtles and their habitats in the future.” Rising levels of marine debris “could also increase entanglements.” Even though, the NMFS was unable to-quantify the risks of climate change and its associated impacts, the agency recognized that they would be detrimental to the loggerheads.
The climate-based model predicted that the proposed action would exacerbate the loggerheads’ decline, and the BiOp is structurally flawed to the extent the NMFS failed to incorporate those findings into its jeopardy analysis. Nat’l Wildlife Fed’n, 524 F.3d at 927. Because the NMFS has not articulated a rational connection between the best available science and its conclusion that the loggerhead sea turtles would not be affected by-the increased fishing efforts, the agency’s determination that the loggerhead “population will remain large enough to retain the.potential for recovery” is arbitrary and capricious.
2. Leatherback Turtles
Plaintiffs also argue that the 2012 BiOp improperly concluded that the fishery would have no appreciable impact on the leatherback turtle population. Unlike its conclusion concerning the loggerheads, however, the NMFS’s no jeopardy conclusion regarding the leatherback turtles finds support in the scientific record and, therefore, is sufficient to withstand judicial review.
Plaintiffs specifically argue that the NMFS erred in limiting the “temporal scale” of its analysis to 25 years, despite the fact that- the fishery’s operations have no related limitation and the NMFS determined that impacts on the sea turtles due to increasing temperatures “are expected to occur slowly over the next century.” However, the NMFS was entitled to rely on the climate-based population assessment model, even though that model could only predict changes in the turtle population for 25 years. See San Luis & Deltas Mendota Water Auth. v. Locke, 776 F.3d 971, 997 (9th Cir. 2014) (“[T]he agehéy has substantial discretion to choose between available scientific models, provided that it explains its choice.”); The Lands Council v. McNair, 537 F.3d 981, 988 (9th Cir. 2008) (explaining that the court may not “act as a panel of scientists that instructs the [agency] how to ... choose[ ] among scientific studies”), overruled ■ on other grounds by Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7, 129 S.Ct. 365, 172 L.Ed.2d 249 (2008). The constraints in the available data supply a reasonable justification for the NMFS to limit its analysis. Accordingly, we cannot conclude that the 2012 BiOp violated the ESA or that the NMFS otherwise acted arbitrarily and capriciously in determining that the fishery would have no appreciable effect on the leatherback turtle population.
B. Consideration of the Effects of Climate Change
■ Lastly, Plaintiffs argue that the 2012 BiOp failed to evaluate the impacts of global climate change. Plaintiffs specifically maintain that the NMFS acted arbitrarily by dismissing the effects of global warming on sea turtles as uncertain without further study.
In the 2012 BiOp, the NMFS explained that the effects from climate change on listed turtle species include rising sand temperatures and sea levels, beach erosion, increased storm activity, and changes in ocean temperature and chemistry. The BiOp also summarized studies anticipating that climate change will impact, among other traits and behaviors, turtle gender ratios, nesting habitat, and reproductive capacity. However, the NMFS determined that there was no available data from which it could credibly project the impacts that climate change would have on the loggerhead or leatherback turtle survival rates. With respect to the loggerhead turtles, the NMFS explained that “current scientific methods are not able to reliably predict the future magnitude of climate change and associated impacts or the adaptive capacity of this species.” The NMFS also stated that “leatherbacks are probably already beginning to be affected by impacts associated with anthropogenic climate change in several ways,” but noted that it did “not have information to predict what the population would do” or “what impact other climate-related changes may have such as increasing sand temperatures, sea level rise, and increased storm events.” As the NMFS observed elsewhere in the BiOp, the effects of climate change will not be globally uniform, and the uncertainty of the rate, magnitude, and distribution of such effects on different temporal and spatial scales—not to mention the turtles’ ability to adapt to these effects—have not been comprehensively studied. Consequently, the NMFS decided that climate change effects could not be “reliably quantified” nor “qualitatively described or predicted” by the agency at the time.
Here, we cannot conclude from the NMFS’s lack of precision that it failed to adequately consider the effects of climate change on the sea turtles. On the whole, the BiOp demonstrated that the NMFS considered a variety of ways in which climate change may áffect the sea turtles, but simply concluded that the data available was too indeterminate for the agency to evaluate the potential sea-turtle impacts with any certainty. Cf. Greenpeace Action v. Franklin, 14 F.3d 1324, 1326-27, 1336 (9th Cir. 1993) (holding that the agency’s no jeopardy conclusion was not arbitrary because the BiOp at issue demonstrated that the agency had based its no jeopardy decision on the best available scientific data, even though the data was “uncertain”); Stop H-3 Ass’n v. Dole, 740 F.2d 1442, 1460 (9th Cir. 1984) (sustaining a BiOp that stated “we have very little data for providing an opinion, but feel it would be unreasonable to request [an additional] study which would be unlikely to provide definitive results.... Based on the available information, which we grant is weak, it is our opinion the proposed project is not likely to jeopardize the continued existence of the Oahu Creeper”). Plaintiffs have failed to sufficiently refute the NMFS’s stated inability to offer more specific predictions on the effects of climate change, and they have not alleged that less speculative scientific information is available that the agency overlooked. San Luis & Deltar-Mendota, 747 F.3d at 602 (“[W]here [superior] information is not readily available, we cannot insist on perfection: [T]he ‘best scientific ... data available,’ does not mean the best scientific data possible.” (citation and internal quotation marks omitted)). Accordingly, the NMFS’s consideration of climate change in the BiOp was neither arbitrary, capricious, nor contrary to the NMFS’s obligation to base its jeopardy decision on the best scientific data it could obtain. See 16 U.S.C. § 1536(a)(2).
CONCLUSION
We conclude that the FWS’s grant of an incidental take permit to the NMFS in reliance on the “special purpose permit” provision in 50 C.F.R. § 21.27 was arbitrary and capricious because the FWS’s interpretation of § 21.27 does not conform to either the MBTA’s conservation intent or the plain language of the regulation. We therefore reverse the district court’s grant of summary judgment affirming the FWS’s decision to issue the permit.
We also conclude that NMFS’s 2012 BiOp’s no jeopardy finding as to the loggerhead sea turtles was arbitrary and capricious because the scientific data suggested that the loggerhead population would significantly decline, and the agency failed to sufficiently explain the discrepancy in its opinion and the record evidence. We therefore reverse the district court’s grant of summary judgment upholding this portion of the BiOp. We otherwise affirm the district court’s grant of summary judgment to Defendants.
AFFIRMED in part; REVERSED in part; and REMANDED. Each party shall bear its own costs on appeal.
. Because we conclude that the FWS acted arbitrarily and capriciously in issuing the incidental take permit to the NMFS under § 21,27, we need not reach Plaintiffs’ additional argument concerning whether the FWS’s action also violated NEPA.