NOT FOR PUBLICATION FILED MAR 19 2024
UNITED STATES COURT OF APPEALS MOLLY C. DWYER, CLERK
U.S. COURT OF APPEALS
FOR THE NINTH CIRCUIT
ALASKA WILDLIFE ALLIANCE; No. 23-35299 ALASKA WILDERNESS LEAGUE; CENTER FOR BIOLOGICAL D.C. No. DIVERSITY; DEFENDERS OF 3:21-cv-00209-SLG-KFR WILDLIFE; FRIENDS OF THE EARTH; SIERRA CLUB, MEMORANDUM*
Plaintiffs-Appellants,
v.
UNITED STATES FISH AND WILDLIFE SERVICE; U.S. DEPARTMENT OF THE INTERIOR; SHANNON ESTENOZ, in her official capacity as Assistant Secretary for Fish and Wildlife and Parks; DEB HAALAND, in her official capacity as Secretary of the Interior,
Defendants-Appellees,
and
ALASKA OIL AND GAS ASSOCIATION; STATE OF ALASKA,
Intervenor-Defendants-
Appellees.
*
This disposition is not appropriate for publication and is not precedent except as provided by Ninth Circuit Rule 36-3.
2
Appeal from the United States District Court for the District of Alaska Sharon L. Gleason, Chief District Judge, Presiding
Argued and Submitted February 8, 2024 Portland, Oregon
Before: McKEOWN, BYBEE, and BRESS, Circuit Judges. Partial Dissent by Judge BRESS.
This case concerns the fate of polar bears living along Alaska’s Beaufort Sea. Our task is to determine whether, in issuing its 2021 incidental take regulation for that region (the “2021 ITR”), the Fish and Wildlife Service satisfied the Marine Mammal Protection Act’s (“MMPA’s” or the “Act’s”) requirements that such take be of “small numbers” of bears and have a “negligible impact” on the Beaufort Sea subpopulation. 16 U.S.C. § 1371(a)(5)(A)(i), (I). We assume familiarity with the facts and applicable law. We have jurisdiction under 28 U.S.C. § 1291 and review de novo the district court’s grant of summary judgment for defendants. Pit River Tribe v. U.S. Forest Serv., 469 F.3d 768, 778 (9th Cir. 2006). We affirm in part and reverse and remand in part.
I. STANDARD OF REVIEW
Since the MMPA lacks a provision for judicial review, this case is subject to the “highly deferential” standard prescribed by the APA. Ctr. for Biological Diversity v. Bureau of Land Mgmt., 833 F.3d 1136, 1146 (9th Cir. 2016) (citation omitted). The scope of that review “is narrow and a court is not to substitute its
3
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 (1983). Agency action may not be invalidated unless “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706(2)(A). “[W]e will uphold a decision of less than ideal clarity if the agency’s path may reasonably be discerned.” Bowman Transp., Inc. v. Ark.-Best Freight Sys., Inc., 419 U.S. 281, 286 (1974).
Nevertheless, “we insist that an agency ‘examine the relevant data and articulate a satisfactory explanation for its action.’” FCC v. Fox Television Stations, Inc., 556 U.S. 502, 513 (2009) (quoting State Farm, 463 U.S. at 43). We reverse where, inter alia, “the agency has . . . entirely failed to consider an important aspect of the problem” or “offered an explanation . . . that runs counter to the evidence before [it].” State Farm, 463 U.S. at 43; see also City of Los Angeles v. FAA, 63 F.4th 835, 851 (9th Cir. 2023) (granting a petition for review where the agency “failed to perform the necessary calculations” that underlay its action and relied on “assumption[s that] defie[d] common sense”).
II. DISCUSSION
A. Subdividing Level A Harassment to Assess “Negligible Impact” Was Reasonable, but the Service’s Failure to Offer an Aggregate Figure Was Not
In making its “negligible impact” determination with respect to Level A harassment, the Service split this category into two: “serious Level A harassment or lethal takes and . . . non-serious Level A harassment.” Plaintiffs dispute this
4
choice as deviating from the MMPA’s two-part conception of take—Levels A and B. 16 U.S.C. § 1362(18)(C)–(D). Yet, we are unconvinced.
Plaintiffs direct us to no section of the MMPA that prohibits the Service from subdividing Level A harassment, and we are persuaded that it was reasonable to do so. The Service’s implementing regulations define a “negligible impact” as one “not reasonably likely to[] adversely affect . . . [the] annual rates of recruitment or survival” of bears. 50 C.F.R. § 18.27(c). Harassment that carries a high risk of killing bears may have a different impact than harassment that does not. Hence, it is rational for the Service to split out gradations of take when assessing how negligible the impact of such take may be.
Regardless of what subcategories the Service fashions, however, these cannot be a substitute for the Service analyzing Level A and Level B harassment in the aggregate according to the schema that Congress created. As noted, the MMPA provides that ITRs may issue only upon the Service’s finding that any “taking” will be of “small numbers of marine mammals” and “have a negligible impact” on the “species or stock” of such mammals.” 16 U.S.C. § 1371(a)(5)(A)(i), (I). “Take” is defined as, among other things, to “harass.” Id. § 1362(13). The Act in turn defines “harassment” as consisting of two categories: Levels A and B. Id. § 1362(18)(C)–(D). To contend, as the Service did at oral argument and our dissenting colleague now does, that the MMPA does not direct
5
the Service to analyze harassment in terms of these explicit, statutorily enumerated categories is to reduce the letter of the Act to mere surplusage. See Oral Argument at 24:21–45, Alaska Wildlife All. v. U.S. Fish & Wildlife Serv., No. 23-35299 (9th Cir. Feb. 8, 2024), https://www.youtube.com/watch?v=G9XiELvJKO4. It is inconsistent with the Service’s own prior interpretation of the Act, as suggested by the previous ITR for the Beaufort Sea. Marine Mammals; Incidental Take During Specified Activities, 81 Fed. Reg. 52,276, 52,306 (Aug. 5, 2016) (estimating harassment of polar bears in Level A and B terms). And it does not reflect the emphasis that Congress placed on these same categories in other portions of the Act. 16 U.S.C. § 1374(c)(3)(C) (directing the Service to “issue a general authorization . . . allowing bona fide scientific research that may result only in taking by Level B harassment of a marine mammal”); see also 50 C.F.R. § 216.45 (adopting said general authorization with respect to Level B harassment). We decline to approve of the Service upsetting Congress’s carefully wrought categories.
We further note that a contrary holding would enable the statutory whole to be reduced to less than the sum of its subdivided parts, obscuring foreseeable take that the MMPA otherwise directs the Service to investigate. No further example is needed than the model that informed the 2021 ITR. Here, the Service estimated the annual probabilities of serious and non-serious Level A harassment to be 45–
6
46% and 28–29%, respectively.1 Plaintiffs observe that adding these figures yields estimates approaching (and in some years, exceeding) 75% per year.2 That poses a problem for the 2021 ITR, which can issue only on the Service’s finding that any anticipated take will have a “negligible impact,” 16 U.S.C. § 1371(a)(5)(A)(i)(I)— unless it can be discounted as not “reasonably likely” to occur, 50 C.F.R. § 18.27(c). Three-in-four odds are hardly so low.
The Service offers two principal counterarguments, neither of which is compelling. First, it disputes Plaintiffs’ math, contending that their “addition of two separate probability outputs is statistically flawed because the probabilities of these different types of harassment are not independent of each other” (emphasis added)—contrary to the conclusion reached by Plaintiffs’ expert. However, the Service has neither denied Plaintiffs’ claim that the aggregate annual likelihood of Level A harassment exceeds 50%, provided its own aggregate figure, nor claimed to have calculated that number at all before issuing the 2021 ITR. Crediting the Service’s unsupported reasoning would allow Level A harassment to be subdivided
1 The model further indicates that a mean of 1.2 polar bears are apt to suffer serious Level A harassment each year. 2 Adding these numbers is possible, in the view of Plaintiffs’ expert, “[b]ecause ‘serious’ and ‘non-serious’ are mutually exclusive.” The Service disputes this proposition but has offered no indication of how much overlap exists between the two categories.
7
however much is necessary to reduce it to less than a “majority of model runs,” at which point it can be discounted as “[un]likely.”
Second, the Service asserts that it was entitled to deprioritize its model’s probability and mean outputs given the “skewed” nature of the results. Instead, it invokes the median value of zero Level A takes to justify its conclusion that no bears will suffer such take during the 2021 ITR. The decision to prioritize the median over other data is one best left to the Service’s expert judgment. Idaho Wool Growers Ass’n v. Vilsack, 816 F.3d 1095, 1109–10 (9th Cir. 2016). But as with the probability figures, the Service has neither provided an aggregate Level A median value nor justified its omission. Irrespective of how the Service chooses to assess the statutory Level A and B categories, it must actually do so and has not satisfied us that it did. This constitutes “fail[ure] to consider an important aspect of the problem”—namely, one of the statutory requirements to issue an ITR—and is arbitrary and capricious. State Farm, 463 U.S. at 43.
We therefore reverse the district court and remand to the Service to offer a fuller explanation for its determination that no Level A incidents are expected during the period covered by the 2021 ITR. See Los Angeles, 63 F.4th at 840 (remanding for the agency “to redo the deficient parts of its analysis”). In assessing the “negligible impact” prong on remand, the Service may, consistent with its expertise, emphasize certain outputs over others. However, given the
8
MMPA’s two-part conception of take, it must determine whether aggregating serious and non-serious Level A take yields a “reasonably likely” result. 50 C.F.R. § 18.27(c). If so (as the 75% figure proffered by Plaintiffs suggests), the Service will then need to determine (i) whether any Level A take predicted will affect only “small numbers” of bears and have a “negligible impact” on the subpopulation and, if so, (ii) whether to issue an updated ITR covering Level A take or no ITR at all. B. The Service Erred by Not Examining the Five-Year Risk of Level A Take Still on the “negligible impact” prong as applied to Level A take, Plaintiffs fault the Service for “[c]onsidering only the isolated probability of take in each individual year of the . . . [2021] ITR . . . but not the much higher total probability of a take occurring at least once over the five-year period.” The parties seem to agree that, in assessing this prong, the Service must examine take in the aggregate across all five years of the regulation. This is consistent with the language of the MMPA. Contrary to the “small numbers” prong, 16 U.S.C. § 1371(a)(5)(A)(i)(I) states that the “total . . . tak[e] during each five-year (or less) period” must have a “negligible impact” on the subpopulation (emphasis added).
While nothing in the MMPA requires the Service to quantify its “negligible impact” analysis, the Service chose to rely on quantitative estimates in issuing the 2021 ITR. Marine Mammals; Incidental Take During Specified Activities; North Slope, Alaska, 86 Fed. Reg. 42,982, 43,040 (Aug. 5, 2021) (“The Service does not
9
anticipate any lethal or injurious take that would [have more than a negligible impact because, inter alia,] . . . . the majority of the Service’s model runs result in no serious injury Level A harassment or lethal takes and the median of the model runs is 0.0.”). Hence, these estimates must not be arbitrary or capricious.
Plaintiffs argue that the Service cannot have satisfied the MMPA’s aggregation requirement because the 46% figure represents the annual risk of serious Level A harassment projected by the Service’s model but, viewed cumulatively across the five years of the 2021 ITR, jumps to almost a statistical certainty. To that end, they submit an expert report, which calculates the cumulative five-year probability of serious Level A take to be 94%. The Service disagrees with this figure but neither offers its own nor explains its failure to do so. That leaves us with only Plaintiffs’ 94% figure—caveated by the Service’s reference to vague “complexities and limitations” that prevent aggregation—to go from. The Service thus faces a catch-22. Either it determined that a 94% risk of Level A take across five years was too “[un]likely” to warrant assessment, see 50 C.F.R. § 18.27(c)—a conclusion that “defies common sense,” Los Angeles, 63 F.4th at 851—or it has “entirely failed to consider an important aspect of the problem,” State Farm, 463 U.S. at 43.
The Service’s counterarguments are unavailing. First, whatever “defer[ence]” is due to its determination of the point at which risk becomes likely,
10
a figure merely 6% shy of certain undoubtedly crosses it. Second, and relatedly, while the Service may credibly rely on the median output instead of probability, that credibility falters once probability approaches certainty. The Service justifies the 2021 ITR by contending that both the “median” and “the most likely result[s] every year [are] . . . zero Level A harassment.” Again, however, the Service does not claim to have calculated the five-year cumulative median; instead, it attempts to explain its decision to ignore the five-year probability (which it also does not appear to have calculated) based on median data for one year. Third and finally, the Service’s claim that Plaintiffs “extrapolate[d] a level of take over five years without accounting for all the . . . complexities and limitations” of the model fails because it cuts against the statistical evidence that we do have and is unsubstantiated. State Farm, 463 U.S. at 43.
In dissent, our colleague asserts that, to satisfy the statutory “total[ity]”
requirement, the Service need only have “look[ed] at the frequency of Level A harassment each year, for all five years.” For the reasons we have supplied, we agree that the Service ought to have provided a Level A estimate in connection with its “negligible impact” analysis. Yet, we disagree with the dissent’s interpretation of the word “total,” which the dissent understands to enable a year- by-year analysis of take. Per-year take is not the “total” when the license is for five years. See 16 U.S.C. § 1371(a)(5)(A)(i)(I). If the Service wishes to issue an
11
ITR for a year at a time, a one-year estimate of total take will suffice. It is not sufficient when the ITR is for a five-year period.
Since the Service either did not calculate the five-year cumulative risk of Level A take as the MMPA requires or unreasonably discounted a high likelihood of such take, we once more find that it “entirely failed to consider an important aspect of the problem.” Id. Hence, we reverse the district court and remand to the Service so that it may (i) aggregate serious and non-serious Level A take together, see supra Part II.A, and (ii) determine whether the five-year risk of such take of a denning cub is “reasonably likely,” 50 C.F.R. § 18.27(c). To the extent that it is, the Service must then evaluate whether the five-year impact of Level A take is “negligible” and whether such take will be of “small numbers” of bears and possibly amend or reverse the 2021 ITR.
C. Plaintiffs’ Remaining Claims Concerning the Service’s Take Calculations Are Unpersuasive
Plaintiffs raise two additional challenges to the methodology underlying the 2021 ITR, neither of which we find to be a basis for reversing the district court.
First, they claim that “[t]he Service misapplied the MMPA[’s] . . .
evidentiary standard . . . and instead required a preponderance of the evidence” for take to be deemed “reasonably likely.” Thus, Plaintiffs’ logic goes, the Service ignored a 46% chance of at least one serious Level A incident per year—a “substantial” risk amounting to “essentially a coin toss.” However, the Service
12
denies having set a preponderance standard at all. To aid in evaluating “negligible impact,” the Service’s “model generated three outputs”—mean, median, and probability. Finding the results to be skewed, the Service explained that it would prioritize the median value and so determined that serious Level A harassment (along with its even-less-probable non-serious counterpart) was not “reasonably likely.” In other words, the Service wrote off the 46% figure not because it was less than 50% (as Plaintiffs suggest) but because, in the agency’s experience, the median of 0.0 was more reliable.
As discussed, the Service’s judgment concerning which of its model’s outputs to prioritize is a technical one concerning the interpretation of statistical data, to which we owe deference. Idaho Wool Growers Ass’n, 816 F.3d at 1109– 10. We find that judgment to be reasonable in theory, although a high enough probability must at the very least be explained and not ignored. See supra Part II.B. In light of the Service’s failure to make the calculations described in Part II.A–B above, it may need to revisit its methodological choices.
Second, with respect to Level B take and the “small numbers” prong, Plaintiffs assert that this criterion cannot have been met because the 2021 ITR authorizes the take of approximately half of the Beaufort Sea subpopulation. Indeed, the Service’s model suggests that 443 of the area’s 907 polar bears (or 48.84%) may suffer Level B take during the 2021 ITR. The Service’s attempt to
13
avoid this result by using a much-lower per-year figure (92 bears, or 10.14%), Plaintiffs assert, was “arbitrary and unlawful.”
As a matter of statutory interpretation, we agree with the Service’s reading that an annual analysis is appropriate under the “small numbers” prong. Nothing in the MMPA requires the Service to conduct this assessment in the aggregate. The word “total” is absent from the “small numbers” portion of the statute. See 16 U.S.C. § 1371(a)(5)(A)(i). At the same time, it does appear in the “negligible impact” portion. See id. § 1371(a)(5)(A)(i)(I). We find that the lack of a “totality” requirement for “small numbers”—together with the presence of such a requirement in the immediately subsequent “negligible impact” provision—offers strong support for the reasonableness of an annual assessment.
Bearing in mind the propriety of the Service’s annual approach, we hold that it was reasonable to determine that only “small numbers” of bears would be subject to Level B take under the 2021 ITR. On both an annual and aggregate basis, the estimates underlying the 2021 ITR are in line with the immediately preceding ITR for the same region. See Marine Mammals; Incidental Take During Specified Activities, 81 Fed. Reg. at 52,304.
In explaining the slightly higher take allowed by the 2021 ITR, the Service found it significant that the model’s “conservative[] assum[ptions]” likely yield an overestimate. Notably, the Service’s projection of the number of bears to be
14
affected counts each instance of take as though it affected a different bear. But as the Service observes, this is highly improbable, considering that industrial activities will affect only 7% of the 19.8-million-acre area covered by the 2021 ITR. Bears living in the immediate vicinity of industrial operations can reasonably be expected to be harassed multiple times, meaning that the actual number of bears affected would likely be smaller than the share estimated. The model also assumes that bears will not flee the harassment, leaving them to be affected again (and counted as two bears if so). Once more, we find these explanations reasonable and affirm. D. The Service Has Satisfied the “Least Practicable Adverse Impact” Standard Plaintiffs’ last challenge to the 2021 ITR concerns the adequacy of the measures the Service has imposed to minimize the risk of harm to polar bears that it expects to occur. The MMPA requires that ITRs “set[] forth . . . permissible methods of taking . . . and other means of effecting the least practicable adverse impact” on the marine mammals affected. 16 U.S.C. § 1371(a)(5)(A)(i)(II) (emphasis added); 50 C.F.R. § 18.27(e). This requirement is subject to a “stringent standard,” compelling the Service to offer “meaningful discussion” whenever it determines that a proposed measure is not practicable. Nat. Res. Def. Council, Inc.v. Pritzker, 828 F.3d 1125, 1138, 1139 n.12 (9th Cir. 2016) (citation omitted).
15
The MMPA does not define “least practicable adverse impact.” Id. at 1134.
Hence, the Service has interpreted this standard to “require mitigation measures that reduce impacts from industry activities but are not so restrictive as to make [such] activities unduly burdensome or impossible to undertake and complete” (internal quotation marks omitted). Plaintiffs do not dispute the Service’s interpretation. Rather, they accuse the Service of offhandedly dismissing certain ameliorative measures that they proposed and that the Service had previously determined to be practicable.
We find the basis for Plaintiffs’ claims of practicability to be dubious and so need not independently assess the adequacy of each of their proposed measures. Plaintiffs attempt to analogize the 2021 ITR to a proposed incidental harassment authorization (“IHA”) granted to one company operating in an entirely different area of Alaska. However, as the Service notes, this tentative IHA “never resulted in a final agency action.” Even if it had gone live, the IHA concerned a single actor, covered a 356,282-acre tract of land, and would last for only eight months. By contrast, the 2021 ITR would apply to all 17 companies operating in a 19.8- million-acre area for five years. Whereas the “narrow scope” of the proposed IHA more feasibly enabled the Service to “mandate rolling, specific dates” for the one company concerned to “enter subsections of the project area,” the 2021 ITR forced the Service to confront the more onerous task of predicting and regulating the
16
temporal and geographical scope of five years’ worth of industrial activities being conducted by more than a dozen companies. We have no reason to doubt the Service’s “least practicable adverse impact” determination and so affirm. E. Remedy: Remanding, Without Vacatur, the 2021 ITR Having found the reasoning behind the 2021 ITR to be incomplete, we must decide on an appropriate remedy. Plaintiffs would have us “vacate the [2021] ITR and any approvals or decisions made in reliance upon it.”3 Yet, vacatur does not flow inevitably from the finding that agency action was arbitrary or capricious, turning instead “on how serious the agency’s errors are and the disruptive consequences of an interim change that may itself be changed.” Cal. Cmtys. Against Toxics v. U.S. EPA, 688 F.3d 989, 992 (9th Cir. 2012) (internal quotation marks and citation omitted). Here, we find that the Service’s errors pale in comparison to the consequences of vacatur for industrial actors in the Beaufort Sea and so remand without vacatur for the agency to supply the missing analysis.
The Service appears to have underestimated the risk of Level A take by cutting it in half and failing to aggregate it across the five years of the 2021 ITR.
3 The Service asserts that Plaintiffs’ request for vacatur was forfeited, having been “raised for the first time” in their opening brief. That contention is belied by the record. Complaint at 63, Alaska Wildlife All. v. U.S. Fish & Wildlife Serv., No. 21- cv-00209, 2023 WL 2674760 (D. Alaska Sept. 16, 2021) (demanding that the court “[v]acate and set aside as unlawful any and all agency approvals and underlying analysis documents, including the [2021] ITR” (emphasis added)).
17
Even assuming that “serious” and “non-serious” Level A take can be aggregated as Plaintiffs contend, however, the probability that a single bear will be taken in a single year is around 75%. Neither side has estimated the probability that two or more bears will be subject to Level A take in a given year, but it would obviously be lower. As noted, Plaintiffs’ expert calculated a 94% risk of serious Level A take occurring at least once during the 2021 ITR period—but only a 70.64% risk of two or more incidents. The available evidence thus suggests that, while the Service has ignored a high likelihood of Level A take, the actual amount of such take will likely be small.
The opposite goes for the economic repercussions of vacating the 2021 ITR.
Doing so would subject industrial actors to civil penalties of up to $10,000 per incident. 16 U.S.C. § 1375(a)(1). “[K]nowing[]” violations carry a fine of up to $20,000 and a year in prison. Id. § 1375(b). Given the wide breadth of Level B harassment, id. § 1362(18)(A)(ii), (D), vacating the 2021 ITR would in essence criminalize the act of knowingly operating as an oil company in the Beaufort Sea. That result is grossly disproportionate to the small amount of Level A take the Service’s model projects, and so we decline to vacate the 2021 ITR.
III. CONCLUSION
We reverse as to Issues 3 (asserting that the Service failed to determine the five-year cumulative risk of Level A take in assessing “negligible impact”) and 5
18
(concerning the Service’s failure to address the high likelihood of unauthorized Level A take) of Plaintiffs’ Opening Brief. We remand to the Service to furnish promptly the calculations discussed in Part II.A–B above and conduct any other necessary proceedings consistent with this memorandum. Finally, we affirm as to the remaining issues and remand without vacating the 2021 ITR. Each side shall bear its own costs. Exxon Valdez v. Exxon Mobil, 568 F.3d 1077, 1081 (9th Cir. 2009).
AFFIRMED in part, REVERSED AND REMANDED in part.
FILED
MAR 19 2024
Alaska Wildlife Alliance v. USFWS, No. 23-35299 MOLLY C. DWYER, CLERK
U.S. COURT OF APPEALS
BRESS, Circuit Judge, dissenting in part:
I agree with those aspects of the court’s decision that affirm the district court and find no error in the Fish and Wildlife Service’s 2021 incidental take regulation (ITR) for Alaska’s Southern Beaufort Sea. But I disagree with the majority’s conclusions that the Fish and Wildlife Service erred when conducting its negligible impact analysis by failing to (1) aggregate serious and non-serious Level A harassment, and (2) calculate the five-year cumulative risk of Level A take. I would affirm the district court’s grant of summary judgment to the Service in full.
The Service may only issue an ITR if it determines that the “total” take during the ITR term will have a “negligible impact” on the species at issue. 16 U.S.C. § 1371(a)(5)(A)(i)(I). A “negligible impact” is one that “cannot reasonably be expected to, and is not reasonably likely to, adversely affect the species or stock through effects on annual rates of recruitment or survival.” 50 C.F.R. § 18.27(c). Here, the Service gave multiple reasons why the total take under the 2021 ITR would have no more than a negligible impact on polar bears living in the South Beaufort Sea.
Some of those reasons were qualitative. The Service explained, for instance, that (1) polar bears’ habitat makes them “unlikely to interact with” industrial activities authorized by the ITR, (2) prior, similar ITRs in the South Beaufort Sea
2
had only “minimal effects” on polar bear behavior, and (3) the mitigation measures prescribed by the 2021 ITR would “limit potential Industry disturbance” of polar bears. One of the Service’s reasons for finding a negligible impact was quantitative. The Service explained that it did “not anticipate any lethal or injurious take that would remove individual polar bears from the population or prevent their successful reproduction” because it had generated a probability model in which “the majority of the . . . model runs result[ed] in no serious injury Level A harassment or lethal takes and the median of the model runs [wa]s 0.0.”
In my view, the above reasoning offers sufficient support for the Service’s negligible impact finding under the APA’s “highly deferential” arbitrary and capricious standard. Ctr. for Biological Diversity v. Bureau of Land Mgmt., 833 F.3d 1136, 1146 (9th Cir. 2016). The majority’s criticisms—which pertain only to the Service’s quantitative analysis—do not convince me otherwise.
First, nothing required the Service to sum together the probabilities of serious and non-serious Level A harassment (which its model calculated separately) before determining that the 2021 ITR would not produce lethal or reproduction-inhibiting take. By the majority’s own account, “[h]arassment that carries a high risk of killing bears [i.e., serious Level A harassment] may have a different impact than harassment that does not [i.e., non-serious Level A harassment].” And the majority further agrees that nothing in the MMPA prohibits the Service from subdividing Level A
3
harassment. The Service could thus conclude that, where the median occurrence of both serious and non-serious Level A harassment was 0 in all model runs, and the probability of serious Level A harassment did not exceed 50% in any model run, the total taking under the 2021 ITR was not “reasonably likely” to “adversely affect [polar bears] through effects on annual rates of recruitment or survival.” 50 C.F.R. § 18.27(c).
The majority deems the Service’s analysis incomplete because the overall probability of Level A harassment purportedly approached 75% in each model run— too high, in the majority’s view, to “be discounted as not ‘reasonably likely’ to occur.” But this logic misapprehends the Service’s obligations under the MMPA.
The Service did not need to show that any particular type of take was not reasonably likely to occur, but rather that the total take under the 2021 ITR was not reasonably likely to adversely affect polar bear recruitment or survival. See 16 U.S.C. § 1371(a)(5)(A)(i)(I); 50 C.F.R. § 18.27(c). Here, the Service assessed the total take under the ITR in both a qualitative and quantitative fashion, and then explained why it believed that take was unlikely to have drastic consequences. That the Service when doing so prioritized the probability of serious Level A harassment—the type of take most relevant to recruitment and survival rates—but not the probability of overall Level A harassment, does not indicate that the Service “entirely failed to consider an important part of the problem.” Motor Vehicle Mfrs.
Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983). In effect, this was no different than the Service initially considering everything that falls within the broader “Level A” harassment category, but then discounting some of this harassment because it is less serious in nature. The Service could decide this was a more precise method of examining the types of harassment that mattered for species survival and recruitment.
According to the majority, the Service “upset[] Congress’s carefully wrought categories” of Level A and B harassment when it reasoned in this way. The majority’s argument rests on the statutory term “taking.” The majority reasons that “take” “is defined as, among other things, to ‘harass,’” and “[t]he [MMPA] in turn defines ‘harassment’ as consisting of two categories: Levels A and B.” But focusing on that definitional chain misapprehends the situation. The issue here is not whether the Service considered Level A harassment in evaluating whether there was a “taking.” It did so. The question is instead whether this harassment has more than a “negligible impact” on the polar bear population. Nothing required the Service to treat all Level A harassment as equivalent when making that determination. The majority focuses on “take,” but the real issue is “negligible impact.”
Again, the Service did use the Level A/B schema when making its negligible impact determination; it simply created an additional level of specificity within that schema to reflect the fact that, as its ITR expressly states, “[n]ot all Level A
5
harassment events affect annual rates of recruitment or survival.” Contrary to the majority, that appears to be far more than the Service did in the negligible impact analysis for its previous Beaufort Sea ITR. See Marine Mammals; Incidental Take During Specified Activities, 81 Fed. Reg. 52,276, 52,306 (Aug. 5, 2016).
The majority also asserts that the Service’s approach “would allow Level A harassment to be subdivided however much is necessary” to reach a “negligible impact” result. I do not think that is the import of the Service’s position. At some point, assigning different weight to different levels of harassment could become arbitrary and capricious. But the Service’s division of Level A harassment into just two sub-categories, serious and non-serious, was rational and supported.
Second, and for similar reasons, nothing required the Service to calculate the five-year probability of Level A harassment before concluding that the 2021 ITR would not lead to lethal or reproduction-inhibiting take. As discussed, the MMPA mandates that the Service analyze the “total” of authorized take across the ITR term when making its negligible impact determination. 16 U.S.C. § 1371(a)(5)(A)(i)(I). The MMPA is silent, however, as to the particular method by which the Service should do so.
The Service’s quantitative model produced both a median and probability value for Level A harassment during each year of the ITR, and the Service assessed all of the model’s outputs. The majority faults the Service for failing to either tally
6
those probabilities together or calculate a “five-year cumulative median.” But beyond its interpretation of the word “total,” the majority does not explain why the Service needed to take this further computational step, as opposed to simply looking at the frequency of Level A harassment each year, for all five years. Indeed, it is the latter approach, and not the former, that is most consistent with the Service’s own unchallenged regulations, which require assessing the impacts of take on “annual rates of recruitment and survival.” 50 C.F.R. § 18.27(c). The majority may “disagree” that a “year-by-year analysis of take” is appropriate for a five-year ITR, but the Service’s regulation says “annual.” And in this case, an annual analysis over five years is further supported by the fact that even with an ITR, applicants must still obtain a letter of authorization (LOA) from the Service to conduct activities that may result in takes, and those LOAs are only valid for one year. In short, especially when considering the Service’s qualitative determinations, the Service did not act arbitrarily or capriciously by failing to perform the majority’s desired calculations. That is even more apparent where—as the majority admits—the Service was not obligated to conduct any quantitative analysis in the first place.
For these reasons, I would find the Service’s existing analysis sufficient and affirm the district court.