NATURAL RESOURCES DEFENSE COUNCIL, INC. et al v. SOBECK et al

Filing 54

MEMORANDUM OPINION granting Plaintiffs' motion for summary judgment 26 and denying Defendants' cross-motion for summary judgment 29 . See opinion for details. Signed by Judge Randolph D. Moss on 3/25/2017. (lcrdm3, )

UNITED STATES DISTRICT COURT
FOR THE DISTRICT OF COLUMBIA

NATURAL RESOURCES DEFENSE
COUNCIL, INC., et al.,
Plaintiffs,
v.

Civil Action No. 15-198 (RDM)

SAMUEL D. RAUCH, III, Acting Assistant
Administrator for Fisheries, National Marine
Fisheries Service, 1 et al.,
Defendants.

MEMORANDUM OPINION
Blueback herring (alosa aestivalis) are silver-colored fish, roughly a foot in length, that
inhabit much of North America’s Atlantic coast. The species is “anadromous,” meaning the fish
are born in inland rivers, then spend most of their adult lives at sea, while still returning to their
natal rivers for six to eight weeks each spring to spawn.
In 2011, the Natural Resources Defense Council (“NRDC”) petitioned the National
Marine Fisheries Service (“Service”) to list blueback herring as “threatened” under the
Endangered Species Act of 1973. The Service undertook a lengthy review in response, but
ultimately found that listing the species “[wa]s not warranted.” See Endangered Species Act
Listing Determination for Alewife and Blueback Herring, 78 Fed. Reg. 48,944 (Aug. 12, 2013)
(“Listing Decision”). The NRDC and others now challenge that determination. Because the
Court agrees that the Service failed to offer a rational connection between the facts and two of its
essential conclusions, and because the Service entirely failed to consider other important aspects
1

The current officeholders are substituted as defendants. See Fed. R. Civ. P. 25(d).

of the problem, the Court will VACATE the Listing Decision and will REMAND the matter to
the Service for further proceedings consistent with this opinion.
I. BACKGROUND
A.

The Endangered Species Act
The Endangered Species Act exists to conserve endangered and threatened species and

“the ecosystems upon which [they] depend.” 16 U.S.C. § 1531(b). As amended, the Act defines
“species” to include not just “any subspecies of fish or wildlife,” but also any “distinct
population segment” of any species of vertebrate fish or wildlife which interbreeds when mature.
Id. § 1532(16). A species is “endangered” if it “is in danger of extinction” either throughout
“all . . . of its range” or throughout “a significant portion of its range.” Id. § 1532(6). And a
species is “threatened” if it “is likely to become . . . endangered . . . within the foreseeable
future” throughout “all or a significant portion of its range.” Id. § 1532(20) (emphasis added).
The Act directs the Service, along with its counterpart in the Department of the Interior, 2
to “determine whether any species is . . . endangered . . . or . . . threatened,” id. § 1533(a), and to
publish lists of species designated as such, id. § 1533(c)(1). Listed species then receive
heightened protections under the Act. See generally id. §§ 1533–1538. Any “interested person”
may petition the Service to change a species’s status, see 5 U.S.C. § 553(e); 16 U.S.C.
§ 1533(b)(3); 50 C.F.R. § 424.14, and, if the petition “presents substantial scientific or

2

Two agencies—the National Marine Fisheries Service (in the Department of Commerce) and
the U.S. Fish and Wildlife Service (in the Department of the Interior)—“share responsibilities for
administering the Act.” 50 C.F.R. § 402.01(b); see also 16 U.S.C. § 1532(15). This case
concerns only the National Marine Fisheries Service, which asserts jurisdiction over nonlandlocked populations of blueback herring. See Listing Decision, 78 Fed. Reg. at 48,945–46
(citing U.S. Fish & Wildlife Serv. & Nat’l Marine Fisheries Serv., Memorandum of
Understanding Regarding Jurisdictional Responsibilities and Listing Procedures Under the
Endangered Species Act of 1973, at 2–3 (Aug. 28, 1974), available at http://www.nmfs.noaa.gov
/pr/laws/esa/mou_usfws.pdf).
2

commercial information indicating that the petitioned action may be warranted,” the Service
must “promptly” conduct a species status review, 16 U.S.C. § 1533(b)(3)(A). The Service must
then publish its findings in a listing determination, id., which must rest on any one or a
combination of the following factors:
(A)

the present or threatened destruction, modification, or curtailment of [the
species’s] habitat or range;

(B)

overutilization for commercial, recreational, scientific, or educational
purposes;

(C)

disease or predation;

(D)

the inadequacy of existing regulatory mechanisms; or

(E)

other natural or manmade factors affecting [the species’s] continued
existence.

16 U.S.C. § 1533(a)(1); see also 50 C.F.R. § 424.11(c). In addition, the Service must make its
listing determination “solely on the basis of the best scientific and commercial data available.”
16 U.S.C. § 1533(b)(1)(A).
B.

Factual Background
Blueback herring “use many different habits” throughout their life cycle, “including the

ocean, estuaries, rivers, and freshwater lakes and ponds.” Listing Decision, 78 Fed. Reg. at
48,945. They spend the first few months of their life in freshwaters. They then migrate to the
open sea, where they mature and spend most of their adult life (in what is called their “ocean
phase”), before returning “to estuarine and freshwater rivers, ponds, and lake habitats to spawn.”
Id.; see also Dkt. 40-1 at 104–06 (AR 2048–50) (describing blueback herring lifecycle). Adult
blueback herring “frequently return[] to their natal rivers for spawning,” but may, on occasion,
“stray[] . . . between rivers.” Listing Decision, 778 Fed. Reg. at 48,945. Evidence also suggests,
but has not conclusively determined, that blueback herring migrate large distances even during
3

their ocean phase, moving en masse with the seasons up and down the Atlantic coast. See, e.g.,
id. at 48,949–50; Dkt. 40-5 at 146 (AR 66,934). The species’s range reaches south to the St.
John’s River in Florida, north to the Miramichi River in New Brunswick, Canada, and east into
oceanic waters along the continental shelf. Listing Decision, 78 Fed. Reg. at 48,945, 48,948.
Figure 1: Approximate Blueback Herring Range in U.S. Waters (2007) 3

3

Nat’l Marine Fisheries Serv., Species of Concern: River Herring 2 (2009), http://www.nmfs.
noaa.gov/pr/pdfs/species/riverherring_detailed.pdf. This map does not appear to capture the
extent of the Canadian blueback herring range as described in the Listing Decision.

4

In August 2011, the NRDC petitioned the Service to list blueback herring as
“threatened.” 4 See Dkt. 40-1 at 86–192 (AR 2030–136) (Petition). Citing numerous studies, the
NRDC argued that blueback herring had “suffered dramatic population declines” from their
nineteenth-century peak and that those declines had continued over the past four decades. Id. at
87, 110 (AR 2031, 2054). The causes, the NRDC argued, were primarily “fishing-related
mortality, dams, dredging and blasting, water pollution, and global warming.” Id. at 88–89 (AR
2032–33). The Service deemed the petition supported by substantial scientific evidence, see
90-Day Finding on a Petition To List Alewife and Blueback Herring as Threatened Under the
Endangered Species Act, 76 Fed. Reg. 67,652, 67,656 (Nov. 2, 2011), and, pursuant to its
statutory mandate, embarked on a blueback herring status review.
As a launching point for its analysis, the Service relied on the May 2012 River Herring
Benchmark Stock Assessment prepared by the Atlantic States Marine Fisheries Commission
(“ASMFC”). 5 Listing Decision, 78 Fed. Reg. at 48,944. The ASMFC study analyzed data from
52 in-river United States “stocks,” i.e., discrete “part[s] of a fish population” with “particular
migration pattern[s]” and “specific spawning grounds.” Dkt. 40-3 at 146, 151 (AR 52,745,
52,750). Of those 52 stocks, 22 displayed evidence of declining river herring populations, 1

4

The NRDC’s petition also sought “threatened” status for a related fish, the alewife. Dkt. 40-1
at 86 (AR 2030). The Service reviewed the alewife’s status concurrently with that of the
blueback herring, using the same methods. See generally Listing Decision, 78 Fed. Reg. 48,944.
The Service found that the alewife is not threatened, id. at 48,992, and Plaintiffs do not challenge
that finding here, Dkt. 27-3 at 15 n.2.

5

The ASMFC is an interstate commission composed of representatives from the fourteen
Atlantic coast states. See Atlantic States Marine Fisheries Compact, Pub. L. No. 77-539, 56 Stat.
267 (1942), amended by Pub. L. No. 81-721, 64 Stat. 467 (1950). Its duties include periodic
“inquir[ies] . . . [into] such methods, practices, circumstances and conditions as may . . . bring[]
about the conservation . . . of the fisheries . . . of the Atlantic seaboard.” 56 Stat. at 268.

5

displayed evidence of an increasing population, and the remaining 28 lacked data to support
definitive quantitative conclusions about population trends. Id. at 151 (AR 52,750). On the
whole, the ASMFC determined that “[t]he coastwide meta-complex of river herring stocks on the
US Atlantic coast is depleted to near historic lows.” 6 Id.
The Service also “worked cooperatively with [the] ASMFC” to “identify information not
in the stock assessment that [would be] needed for [the] listing determination.” Listing Decision,
78 Fed. Reg. at 48,944. At the outset of the status review, they convened three sets of workshops
and working groups with experts in the field. Listing Decision, 78 Fed. Reg. at 48,944; see Dkt.
40-1 at 11 (AR 1). Two of those working group reports received peer review from the Center for
Independent Experts. Listing Decision, 78 Fed. Reg. at 48,944.
C.

The Listing Decision
On August 12, 2013, the Service published its fifty-page Listing Decision “determin[ing]

that listing blueback herring as threatened or endangered” was “not warranted at th[at] time.”
Listing Decision, 78 Fed. Reg. at 48,944. The decision turned on three central conclusions.
First, the Service found that it lacked sufficient data to deem any subset of blueback herring a
“distinct population segment.” Id. at 48,950, 48,993. Second, “as a result of [its] extinction risk
analysis,” the Service determined that the blueback herring “[was] not in danger of extinction
[throughout all of its range] or likely to become so in the foreseeable future.” Id. at 48,993.
And, third, without addressing whether the blueback herring was threatened or endangered

6

These ASMFC conclusions refer generally to “river herring,” a category that includes both
blueback herring and alewife.

6

throughout the “Mid-Atlantic” region, the Service concluded that the Mid-Atlantic region was
“not . . . a significant portion of the blueback herring’s range.” 7 Id.
The grounds for those conclusions appear to have been as follows.
1.

The Service’s “Distinct Population Segment” (“DPS”) Analysis

The Service first sought to identify any “distinct population segments” of blueback
herring. Because the Act’s definition of “species” includes “distinct population segment[s]” of
any species of vertebrate fish, 16 U.S.C. § 1532(16), each “distinct population segment” of
blueback herring requires its own status determination. The threshold issue for the Service,
therefore, was delineating exactly which “species” were the subjects of its review. See Listing
Decision, 78 Fed. Reg. at 48,948–50.
The 1996 DPS Policy
To make this threshold determination, the Service applied its Policy Regarding the
Recognition of Distinct Vertebrate Population Segments Under the Endangered Species Act, 61
Fed. Reg. 4722 (Feb. 7, 1996) (the “1996 DPS Policy”). Under the 1996 DPS Policy, a group of
animals of the same taxon (i.e., biological species) is a “distinct population segment” only if two
conditions are met. First, the group must be “discrete[].” Id. at 4725. A group is discrete under
the 1996 DPS Policy if it is either (1) “markedly separated from other populations of the same
taxon as a consequence of physical, physiological, ecological, or behavior factors” or (2)
“delimited by international governmental boundaries” across which meaningfully different
regulatory mechanisms exist. Id. Second, the group must be “significan[t] . . . to the [taxon] to

7

At the same time, however, the Service conceded that blueback herring are “at low abundance
compared to historical levels;” that the Service’s status review suffered from “significant data
deficiencies;” and that further “monitoring [of the] species is warranted.” Id. at 48,994. The
Service also noted that the “positive benefit” from the “many ongoing restoration and
conservation efforts” could not be “quantif[ied]” at that time. Id. “Given the[se] uncertainties
and data deficiencies,” the Service “commit[ted] to revisiting [the] species in 3 to 5 years.” Id.
7

which it belongs.” Id. Under the 1996 DPS Policy, “significance” depends on factors that “may
include”—but are “not limited to”—the following:
1.

Persistence of the discrete population segment in an ecological setting
unusual or unique for the taxon,

2.

Evidence that loss of the discrete population segment would result in a
significant gap in the range of a taxon,

3.

Evidence that the discrete population segment represents the only surviving
natural occurrence of a taxon that may be more abundant elsewhere as an
introduced population outside its historic range, or

4.

Evidence that the discrete population segment differs markedly from other
populations of the species in its genetic characteristics.

Id. “Because precise circumstances are likely to vary considerably from case to case,” the policy
adds, “it is not possible to describe prospectively all the classes of information that might bear on
the biological and ecological importance of a discrete population segment.” Id.
The Five “Discrete” Population Segments
The Service’s “discreteness” finding turned largely on genetics research by Dr. Eric P.
Palkovacs. Listing Decision, 78 Fed. Reg. at 48,948–49; see also Dkt. 40-6 at 233–47 (AR
73,485–99) (Palkovacs et al. report to the “stock structure” working group). Dr. Palkovacs
collected DNA samples from 1,183 blueback herring across 20 rivers in the United States and
performed statistical analyses to test them for genetic differentiation. 8 Listing Decision, 78 Fed.
Reg. at 48,948–49; Dkt. 40-6 at 234–35 (AR 73,486–87). He noted, however, that the “neutral
genetic markers” he tracked (i.e., the DNA sequences he compared across specimens)

8

The Palkovacs team’s initial presentation to the Service stated that they considered 1,201
blueback herring samples, Dkt. 40-3 at 88 (AR 52,337), and that number appears in the Listing
Decision, see 78 Fed. Reg. at 48,948. The Palkovacs report itself, however, states the number as
1,183. See Dkt. 40-6 at 234 (AR 73,486). The difference is immaterial to this opinion.
8

“represent[ed] the effects of gene flow and historical population isolation, but [did] not
[represent] the effects of adaptive processes.” Listing Decision, 78 Fed. Reg. at 48,949.
The Palkovacs analysis revealed a minimum of four geographic “clusters” of rivers in the
United States, within which blueback herring populations were “genetically homogenous,” Dkt.
40-6 at 235 (AR 73,487), and between which blueback herring populations were “genetically
distinguishable,” id. at 233 (AR 73,485). See also Listing Decision, 78 Fed. Reg. at 48,948–49.
These clusters also corresponded to “larger-scale geographic boundaries,” across which “gene
flow [was] extremely minimal.” Dkt. 40-6 at 237 (AR 73,489). Dr. Palkovacs therefore
concluded that these geographic boundaries “define[d] the higher-level population structure” for
blueback herring in the United States. Id. The four Palkovacs “stock complexes” were the
“Northern New England,” “Southern New England,” “Mid-Atlantic,” and “Southern” stock
complexes, Listing Decision, 78 Fed. Reg. at 48,949 (citing Dkt. 40-6 at 238 (AR 73,490)), as
shown in the figure below.

9

Figure 2: Blueback Herring U.S. Stock Structure as Defined by Palkovacs et al. 9

9

Listing Decision, 78 Fed. Reg. at 48,952 fig.2.
10

Applying the 1996 DPS Policy, the Service determined that the four stock complexes
identified by Dr. Palkovacs represented “discrete” population segments. Id. at 48,950. This
conclusion presumably relied on the policy’s first discreteness prong, which looks for
populations “markedly separated from other[s] . . . as a consequence of physical, physiological,
ecological, or behavior factors,” 1996 DPS Policy, 61 Fed. Reg. at 4725—factors that include the
genetic and geographic evidence that Dr. Palkovacs examined.
Notably, the Palkovacs analysis did not consider the entire blueback herring population:
it omitted (1) blueback herring in Canadian rivers and (2) blueback herring in their “ocean
phase,” where all blueback herring (regardless of stock) spend most of their adult lives. The
Service resolved the issue regarding Canadian blueback herring by designating them as a fifth
“discrete” population segment—not based on genetic differentiation, but based on the second
prong of the “discreteness” test, which keys to “international governmental boundaries.” Id. at
48,950 (applying 1996 DPS Policy, 61 Fed. Reg. at 4725). The oceanic population, however,
proved more difficult. For that group, the Service found that “[m]igration and mixing patterns
of . . . blueback herring in the ocean ha[d] not been determined,” id. at 48,949, but that experts
suspected the oceanic population to be “comprised of mixed stocks,” id. at 48,950. Given the
anticipated level of mixing, the Service decided that “the ocean phase” of blueback herring could
not be treated as its own discrete population segment “until further tagging and genetic data
become available.” Id. at 48,949.
The Service, accordingly, settled on five “discrete” blueback herring population
segments, which the Listing Decision and this opinion refer to—from north to south—as the (1)
Canadian, (2) Northern New England, (3) Southern New England, (4) Mid-Atlantic, and (5)
Southern stock complexes. Id. at 48,949–50. Although blueback herring spend the majority of

11

their adult lives in the ocean, see id. at 48,945, the Service could not conclude that the oceanic
population segment was “discrete,” id. at 48,949–50. All of these population segments
corresponded to geographic areas, as depicted (roughly) in the map below.
Figure 3: The Five “Discrete” In-River Stock Complexes and the
“Mixed” (Non-Discrete) Oceanic Population 10

10

Adapted from Dkt. 40-5 at 148 (AR 66,936) (opinion of expert Kevin Sullivan). The Court
has modified the original image by (1) updating it to reflect the final Palkovacs stock complexes,
12

No “Significant” Population Segments
Despite finding these five stock complexes to be “discrete,” the Service concluded that
none of them met any of the 1996 DPS Policy’s four (non-exclusive) criteria for “significance.”
See Listing Decision, 78 Fed. Reg. at 48,950. As an initial matter, the Listing Decision stated
that the Service “considered all [four] of the [1996 DPS Policy’s] criteria” and observed that
blueback herring “occupy many different ecological settings throughout their range.” Id. It then
devoted several paragraphs to the second and fourth criteria for “significance” under the policy.
As to the second criterion—whether “the loss of [a] discrete population segment” would
create a “significant gap” in the remaining population’s range, id. at 48,948—the Service
acknowledged that “the loss of [any one] stock complex would mean the loss of [the
corresponding] subpopulations” of herring that were born in and annually return to those rivers,
id. at 48,950. But, because the oceanic population is “comprised of mixed stocks,” the Service
reasoned, the oceanic territory “would most likely still be occupied by migratory river herring
from other stock complexes.” Id. The Service thus concluded:
As it has been shown that gene flow is greater among neighboring runs than among
distant runs, we might expect that river herring would re-colonize neighboring
systems over a relatively short time frame. Thus, the loss of one stock complex in
itself may not be significant; the loss of contiguous stock complexes may be. The
goal then for river herring stock complexes is to maintain connectivity between
genetic groups to support proper metapopulation function (spatially separated
populations of the same species that interact, recolonize vacant habitats, and occupy
new habitats through dispersal mechanisms).
Id.
As to the fourth criterion—whether any “discrete population segment differs
markedly . . . in its genetic characteristics”—the Service sought to distinguish its earlier finding

as depicted in the Listing Decision, 78 Fed. Reg. at 48,952 fig.2, and (2) replacing the label on
the “ocean phase” population, which originally read “Single At-Sea Mixing Stock.”
13

under the discreteness prong that the four Palkovacs stock complexes were “markedly separated
from other population segments” as a result of their genetic characteristics. See supra Part
I.C.1.b. To do so, the Service seized on Dr. Palkovacs’s “caveat” that the genetic markers his
study looked for “represented the effects of gene flow and historical population isolation, but not
the effects of adaptive processes.” Id. at 48,949. As the Service explained, “[n]eutral genetic
markers . . . have a longstanding history of utilization in stock designation for many anadromous
fish species.” Id. at 48,950. “[T]hese markers,” however, “represent the effects of gene flow and
historic population isolation and not the effects of adaptive behavior.” Id. That information, the
Service concluded, is “appropriately used in support of the discreteness criterion.” Id. But
because “neutral markers” of the kind Dr. Palkovacs measured do not “capture[]” “[t]he effects
of adaptive genetic and phenotypic diversity,” they have no bearing on the significance
determination. Id.
Based on this analysis, the Service declined to designate any “distinct population
segments” of blueback herring. See id. “The [four Palkovacs] stock complexes may be
discrete,” the Service wrote, “but under the DPS policy, they are not significant.” Id. The
Service further concluded that, “given the unknown level of intermixing between Canadian and
U.S. river herring in coastal waters, the Canadian stock complex should also not be considered
separately under the DPS policy.” Id.
2.

The Service’s Extinction Risk Analysis

Having determined that the only “species” at issue was the entire taxon of blueback
herring, the Service turned to its main inquiry: Are blueback herring “threatened”—i.e., “likely
to become [in danger of extinction] in the foreseeable future” throughout “all . . . of their range”?
16 U.S.C. §1532(20). This extinction risk analysis contained both a qualitative and a

14

quantitative component, although it is unclear whether and how the qualitative component
impacted the decision not to list the species.
The Qualitative Rankings
In the qualitative component, the Service first surveyed the threats to blueback herring
and categorized them under the five factors enumerated in 16 U.S.C. § 1533(a)(1). See Listing
Decision, 78 Fed. Reg. at 48,953–70. This part of the Listing Decision merely catalogued
potential threats to blueback herring, without analyzing the magnitudes of those threats or the
likelihood that they would cause extinction. See id. Instead, the Service promised after each
section that “the level of threat posed by these potential stressors [would be] evaluated further in
the qualitative threats assessment below.” Id. at 48,958; accord id. at 48,961, 48,963, 48,968,
48,970.
In the “qualitative threats assessment” itself, the Service asked nine of its team members
to rank each potential threat on a scale from zero to five, according to perceived severity. Id. at
48,970. For example, if a team member considered a threat to be of “low” severity, she would
assign it a “one;” if she considered it to be of “high” severity, she would assign it a “five.” Id.
After tabulating the results, the Service determined that its team believed “dams and other
barriers” to be the most serious threats. See id. at 48,978. Although this “qualitative analysis”
occupied thirty-two of the Listing Decision’s fifty pages, it nowhere discussed, analyzed, or drew
conclusions regarding whether these or any other threats posed a risk of extinction to the
blueback herring.
The Quantitative Models
The driver of the Service’s extinction risk conclusion—and, indeed, of the entire Listing
Decision—appears to have been the Service’s quantitative “trend analysis modeling.” See
Listing Decision, 78 Fed. Reg. at 48,985–92. This analysis began from the premise that, as of
15

1976, the blueback herring population had already undergone a significant decline. See id. at
48,987 (asserting that the risk analysis “assume[d] that there has already been a significant
decline in abundance” due to overfishing); id. at 48,989 (stating that, “[p]rior to 1976, fishing
intensity was much greater”); id. at 48,987 (explaining that the Service “use[d] data from 1976
through the present” to minimize the influence of the pre-1976 overfishing). “[G]iven the
extensive declines from historical levels,” the Service wrote, the blueback herring “is [not]
thought to be capable of withstanding continued high rates of decline.” Id. at 48,992. The
Service thus sought to answer the following question: Since 1976, has the blueback herring
population stabilized or has it continued to decline?
To explore this question, the Service retained Dr. Kiersten Curti of the Northeast
Fisheries Science Center (“NEFSC”), 11 and asked her to create a number of statistical models to
estimate the average annual change in the relative size of each blueback herring population
segment—a metric the Service calls “relative abundance.” Listing Decision, 78 Fed. Reg. at
48,989; see also Dkt. 40-2 at 71–113 (AR 2477–519) (Dr. Curti’s report). The analysis involved
the following steps.
i. The Models’ Design
Because Plaintiffs do not challenge the design of the models themselves, see, e.g., Dkt.
53 at 16, most of the technical details underlying those models are not material to this opinion.
For present purposes, the following simplified account will suffice.
As input, each model took some number of time series, i.e., lists of values recorded for a
given measurement each year. For example, some time series tracked the number of blueback

11

The NEFSC is “the scientific research arm of [National Oceanic and Atmospheric
Administration] Fisheries in the northeast.” Dkt. 29-1 at 42 n.23.
16

herring that passed by a given monitoring site during certain observational periods each year.
See Dkt. 47-1 at 5 (Damon-Randall Decl. ¶ 6(b)) (describing “run counts”). Other time series
tracked the total biomass of blueback herring caught in a given river each year, divided by the
total number of nets used. See id. (Damon-Randall Decl. ¶ 6(c)) (describing “catch per unit
effort”). Time series like these are valuable because they theoretically correlate with the size of
the regional blueback herring population—in other words, they are “indices of relative
abundance.” But, because they measured different types of data, often with units of different
dimensions (e.g., “fish per season” or “weight per net”), they were not always directly
comparable. So, whenever a model took more than one time series as an input, Dr. Curti
multiplied each subsequent time series by whatever number was necessary to “scale[]” it to the
same size as the first. Dkt. 40-2 at 75 (AR 2481). In other words, the time series within each
model were “normalized” to a common scale. Dkt. 53 at 53.
As output, each model then generated an estimate of the average change per year in the
modeled population’s “relative abundance.” 12 Exactly what this number represents is unclear;
neither the Service’s Listing Decision nor its submissions in this case have explained the units in
which relative abundance was measured. 13 As far as the Court can tell, relative abundance is

12

Technically, the output appears to represent the difference between the natural logarithm of
the modeled population’s estimated relative abundance in the most recent year modeled and the
natural logarithm of the modeled population’s estimated relative abundance in the first year
modeled, divided by the number of years modeled. See Dkt. 40-2 at 100 fig.9, 109–10 fig.16
(AR 2506, 2515–16); see also Dkt. 53 at 52 (Plaintiffs’ counsel’s explanation at oral argument).
Given that the agency’s subsequent approach considered only whether the confidence interval for
this number included zero, this distinction between relative abundance and the natural logarithm
of relative abundance does not appear material for present purposes.

13

At oral argument, the Court provided the Service with the opportunity to address this question
in a further filing. See Dkt. 53 at 8–9. The Service’s post-oral argument submission, however,
simply noted that “[r]elative abundance is different for every survey” and that the units of
17

measured in arbitrary “relative abundance” units, which do not directly represent any tangible
values (unlike, say, “fish per year” or “biomass per net per year”). Instead, the “relative
abundance” metric relays information only about the modeled population’s relative change in
size. See, e.g., Dkt. 40-6 at 104 (AR 68,439) (explaining that “[the] modeling will give us
percent decline over the time period modeled [but] not actual numbers”). So, for example, if the
modeled population had 10 “relative abundance” units in 1980, and 15 “relative abundance”
units in 1990, then the relative abundance grew by 5 units in 10 years, and the model’s output
would be “0.5.” See Dkt. 53 at 51–54 (Oral Arg. Tr.). The models also paired each estimate
with its 95% confidence interval and standard error. See Dkt. 40-2 at 77, 79 (AR 2483, 2485).
ii. The Models’ Inputs and Results
The Service originally intended that Dr. Curti would create a total of six models: one
model for each of the five “discrete” stock complexes (i.e., the Canadian, Northern New
England, Southern New England, Mid-Atlantic, and Southern stock complexes), and one
“coastwide” model that would incorporate data from the ocean near the Atlantic coast. See, e.g.,
Dkt. 40-2 at 72 (AR 2,478); Dkt. 40-6 at 204–05 (AR 68,929–30). Although the Listing
Decision sometimes referred to the “coastwide” model as the “rangewide” model, that is a
misnomer. Because the data in that model came only from the ocean—and not the species’s
entire range, which would include inland rivers—it is best thought of as a proxy for the blueback
herring’s “ocean phase,” where herring from all stock complexes mix together.
The Service was unable to obtain data from the Canadian government for the Canadian
stock complex model and, therefore, did not construct a model for the population in Canadian

measurement were “comparable between years within a specific survey.” Dkt. 47-1 at 4 (Damon
Randall Decl. ¶ 5) (emphasis added).
18

rivers. See Listing Decision, 78 Fed. Reg. at 48,989 (“No blueback herring abundance indices
were available for the Canadian stock.”); id. at 48,992 (“[W]e were unable to obtain data from
Canada to determine the population growth rate for rivers in Canada.”). For the “coastwide”
model, the Service used two sets of annual “trawl net” surveys from ships sailing along the coast.
Id. at 48,988–89. And for the other four stock-specific models, the Service used data from the
2012 ASMFC stock assessment. Id. at 48,989. Each time series in that dataset was specific to a
particular river or bay within a state or the District of Columbia. Most of the models used only
one or two time series as inputs. The Mid-Atlantic region, however, used twelve. See Dkt. 40-2
at 84 (AR 2490) (list of stock-specific inputs). It is unsurprising, then, that the Mid-Atlantic
model had the lowest margin of error of any of the models by an order of magnitude. See Listing
Decision, 78 Fed. Reg. at 48,991.
The figures below describe the models’ numerical results:

19

Table 1: The Models’ Inputs and Outputs 14
Outputs (Average Annual Change
in Relative Abundance)

Inputs (Time Series)

95% Confidence Interval
Modeled
Population
“Coastwide”
(mixed stock)
Canadian Stock
Complex
Northern New
England Stock
Complex
Southern New
England Stock
Complex

Mid-Atlantic
Stock Complex

Southern Stock
Complex

# of
Time
Series

Years

Std.
Error

Low
End

Point
Est.

High
End

2

States in Region

1976–
2012

0.040

–0.040

0.039

0.119

N/A
NO DATA
Maine
New Hampshire
Total
Massachusetts
Rhode Island
Total
Connecticut
New York
New Jersey
Delaware
Maryland
Washington, D.C.
Virginia
North Carolina
(northern)
Total
North Carolina
(southern)
South Carolina 15
Georgia
Florida
Total

N/A

1
1

N/A

1
2
2
1
0
1
1
2
3

0.058

–0.185

–0.076

0.041

1983–
2012

0.035

–0.101

–0.033

0.036

1979–
2011

2
1
0

1995–
2011

0.003

–0.054

–0.048

–0.042

–0.058

0.022

0.102

12
0
2
0
0

1979–
2011

0.041

2

14

The Court compiled this table from information in the Listing Decision, the Palkovacs report,
and Dr. Curti’s report. For the “coastwide” model, descriptions of the inputs are at Listing
Decision, 78 Fed. Reg. at 48,988–89 and Dkt. 40-2 at 100 fig.9 (AR 2506), and the outputs are at
Listing Decision, 78 Fed. Reg. at 48,990 tbl.15 (highlighted rows). For the stock-specific
models, the states are at Dkt. 40-6 at 238 (AR 73,490), the time series are at Dkt. 40-2 at 84 tbl.3
20

Figure 4: Estimates and 95% Confidence Intervals for Each Population’s
Average Annual Change in Relative Abundance 16

The Court notes, however, that while the time series were normalized within each model, the
Court has not located an explicit statement in the record stating that the time series were also
normalized across the five models. See, e.g., Dkt. 40-2 at 75 (AR 2481). In other words, it is
not obvious that a “0.05” decrease in one model is equivalent to a “0.05” decrease in another.

(AR 2490), the years are at id. at 78 (AR 2484), and the outputs are at Listing Decision, 78 Fed.
Reg. at 48,991 tbl.16 (highlighted rows).
15

It appears that one of the two time series for the Southern stock complex—the “SanteeCooper Min. Pop.” survey—may have been erroneously excluded from the model. The graph of
estimated relative abundance over time for that stock complex suggests that the model ignored
that time series, see Dkt. 40-2 at 109 (AR 2515), as do the graphs of the two time series’
residuals, see id. at 111 (AR 2517).
16

Adapted from Dkt. 40-2 at 108 (AR 2514). For clarity, the Court has replaced the
“Coastwide” label, which originally read “Range-wide,” and replaced the x-axis label, which
originally read “Growth rate ± 95% confidence intervals.”

21

iii. The Service’s Interpretation of the Models
In order to translate the models’ results into a concrete determination of blueback herring
extinction risk, the Service employed a three-step process of its own design, which does not
appear to have been the subject of any expert peer review. 17
In the first step, the Service assigned each modeled populations one of three possible
labels: “significantly increasing,” “significantly decreasing,” or “stable.” Listing Decision, 78
Fed. Reg. at 48,987. A modeled population was labeled “significantly increasing” if both ends
of the confidence interval around the model’s estimate were positive. Id. Conversely, a modeled
population was deemed “significantly decreasing” if both ends of the confidence interval were
negative. Id. And, if a confidence interval began at a negative number and ended at a positive
number (i.e., if it included zero), the Service considered that modeled population “stable.” Id.
As the Service explained:
Relative abundance of a stock is considered to be significantly increasing or
decreasing if the 95-percent confidence intervals of the population growth rate do
not include zero. In contrast, if the 95-percent confidence intervals do contain zero,
then the population is considered to be stable, as the increasing or decreasing trend
in abundance is not statistically significant.
Id. The Service, accordingly, assigned labels to the modeled populations as follows:

17

The Extinction Risk Analysis Working Group Report from the Service’s workshop and
working group with experts in the field makes no mention of the three-step process the Service
ultimately used to interpret Dr. Curti’s models. See generally Dkt. 40-5 at 89–128 (AR 66,800–
39).
22

Table 2: Trend Labels 18
Modeled Population

95% Confidence Interval for Average
Annual Change in Relative Abundance
Low End

“Coastwide”

Point Est.

High End

–0.040

0.039

0.119

Trend Label

Canadian Stock Complex

NO DATA

“Stable”
N/A

Northern New England
Stock Complex

–0.185

–0.076

0.041

“Stable”

Southern New England
Stock Complex

–0.101

–0.033

0.036

“Stable”

Mid-Atlantic Stock
Complex

–0.054

–0.048

–0.042

“Significantly
Decreasing”

Southern Stock Complex

–0.058

0.022

0.102

“Stable”

Because each trend label represented only a single blueback herring population segment,
however, the Service needed a further mechanism to assess the risk of extinction rangewide. To
do so, the Service created and applied two independent assessments, which it labeled “Tier A”
and “Tier B.” See id. at 48,986–87. Although the Service designed each tier to assess different
types of conservation metrics, see id., given the “significant uncertainties associated with the
available data” and the “significant number of data deficiencies,” the Service decided after
reviewing the models’ results that “it was not necessary to have information under both tiers in
order to make a risk determination.” Id. at 48,986.
“The goal of Tier A,” the Service wrote, “was to maintain three contiguous stock
complexes that are stable or increasing.” Id. As the results described above show, however, just
two adjacent stock complexes—the Northern and Southern New England stock complexes—

18

The Service described its labeling process for the modeled populations of blueback herring at
Listing Decision, 78 Fed. Reg. at 48,989–90.
23

received the designation “stable.” The only other “stable” complex—the Southern stock
complex—was separated from the others by the “significantly decreasing” Mid-Atlantic stock
complex. And the last stock complex—Canada—received no label due to insufficient data.
Thus, although the models failed to identify “three contiguous stock complexes” where the
blueback herring population was “stable or increasing,” the Service concluded that it was unable
to draw any results under Tier A because it lacked sufficient data regarding “the population
growth rate for rivers in Canada.” Listing Decision, 78 Fed. Reg. at 48,992. Without that data,
the Service wrote, “there [was] insufficient information available to make a conclusion under
Tier A.” Id.
The Service’s decision whether to list blueback herring as threatened throughout its
range, accordingly, came down to Tier B. The Service designed Tier B to “directly interpret the
results” of its trend analysis models. Listing Decision, 78 Fed. Reg. at 48,986–87. Unlike “Tier
A,” “Tier B” did not consider the stock complexes’ geographic arrangement; it looked only at
how many of the modeled populations were “stable” or “significantly increasing.” Id. at 48,987.
According to the Service:
[A] stable or significantly increasing trajectory suggests that [the blueback herring]
may be within the margins of being self-sustainable . . . . [But] [a] significantly
decreasing population growth rate for several stock complexes[, on the other hand,]
would be an indicator that the current abundance may not be sustainable relative to
current management measures and, therefore, may warrant further protections.
Id. The Service further stated that the level of extinction risk for blueback herring was a function
of two criteria: (1) the number of stock-specific models that were “significantly decreasing;” and
(2) whether or not the “coastwide” model was “significantly decreasing.” See id. The following
matrix summarizes how the Service used these two criteria to determine the species-wide
extinction risk:

24

Table 3: Tier B Outcome Matrix 19

One

Two

Three or more

“Significantly
Increasing” or
“Stable”

Coastwide Model’s Label

Zero

“Low risk” of
extinction

“Moderate-Low
risk” of
extinction

“ModerateHigh risk” of
extinction

“High risk” of
extinction

“Significantly
Decreasing”

Number of Discrete Stock Complexes Labeled “Significantly
Decreasing”

“High risk” of
extinction

“High risk” of
extinction

“High risk” of
extinction

“High risk” of
extinction

Thus, because only one stock complex—the Mid-Atlantic—was “significantly
decreasing,” and because the “coastwide trajectory” and three of the four stock complexes were
“stable,” the Service concluded that the blueback herring was at a “moderate low” risk of
extinction through all of its range. See id. at 48,992. The Service therefore “determined that, as
a result of the extinction risk analysis,” the blueback herring was “not in danger of extinction or
likely to become so in the foreseeable future,” and that “listing . . . blueback herring as either
endangered or threatened throughout all of [its] range[] [wa]s not warranted at th[at] time.” 20
Listing Decision, 78 Fed. Reg. at 48,993.

19

This outcome matrix is described in text at Listing Decision, 78 Fed. Reg. at 48,987.

20

Although the Service observed that “[t]he estimated population growth rates reflect[ed] the
impacts from the various threats [identified in the qualitative analysis],” id. at 48,987, it
otherwise appears that the Service derived its extinction-risk conclusion from the Tier B
quantitative analysis alone.
25

3.

The Service’s “Significant Portion of its Range” (“SPR”) Analysis

The Service then turned to the last of its three central questions: Is the blueback herring
threatened throughout a “significant portion of its range”? See 16 U.S.C. § 1532(20). For this
purpose, the Service took the candidate “portions of [the] range” to be the geographic boundaries
of the five discrete stock complexes which the Service had previously identified (but not the
ocean-phase population, which is comprised of mixed stock). See Listing Decision, 78 Fed. Reg.
at 48,993; see also supra Figure 3 (map of discrete stock complexes). To decide whether any of
these stock complexes were “significant,” the Service applied its Draft Policy on Interpretation
of the Phrase “Significant Portion of Its Range” in the Endangered Species Act’s Definition of
“Endangered Species” and “Threatened Species,” 76 Fed. Reg. 76,987 (Dec. 9, 2011) (the “2011
Draft SPR Policy”).
Under the 2011 Draft SPR Policy, “if a species is found to be endangered or threatened in
only a significant portion of its range, the entire species is listed as endangered or threatened,
respectively, and the Act’s protections applies across the species’s entire range.” Id. at 76,990
(emphasis added). The Service therefore adopted an interpretation of “significant” that used “the
species itself as the reference point.” Id. at 76,993. That interpretation is as follows:
Significant: A portion of the range of a species is “significant” if its contribution
to the viability of the species is so important that without that portion, the species
would be in danger of extinction.
Id. at 77,002. 21

21

After issuing the Listing Decision, the Service adopted a final SPR policy that differs
substantively from the 2011 draft. See Final Policy on Interpretation of the Phrase “Significant
Portion of Its Range” in the Endangered Species Act’s Definitions of “Endangered Species” and
“Threatened Species,” 79 Fed. Reg. 37,578 (July 1, 2014) (the “2014 Final SPR Policy”). Under
the old standard, “significance” existed only if, without the members of the species in the portion
of the range in question, the remainder of the species would be endangered. The new definition
26

The Service’s subsequent application of the 2011 Draft SPR Policy to the blueback
herring was brief. It concluded, in three sentences, that the risk of blueback herring extinction
throughout the Mid-Atlantic region was immaterial because that portion of the range was not
“significant.” It determined that the Mid-Atlantic region was not “significant” because, even
assuming the hypothetical loss of that portion of the range, “the overall coastwide trend is
stable.” Id. at 48,993. In other words, because “the relative abundance of the species throughout
its range (as demonstrated through the coast-wide population growth rate) [wa]s stable,” the
Service reasoned, “the [M]id-Atlantic stock complex does not contribute so much to the species
that without it, the entire species would be in danger of extinction.” Id. at 48,992. The Service
did not explicitly analyze the significance of any other regions. See id. at 48,992–93.
D.

The Present Proceeding
Plaintiffs are five non-profit organizations with interests in the continued existence of

blueback herring. They are the NRDC, the Anglers Conservation Network, the Delaware River
Shad Fishermen’s Association, the Great Egg Harbor River Council, and the Great Egg Harbor
Watershed Association. 22 Dkt. 1 at 4–6 (Compl. ¶¶ 10–14). Invoking the Act’s “citizen suit”
provision, see 16 U.S.C. § 1540(g), they ask the Court to declare the Listing Decision unlawful,
vacate it, and remand it to the Service, see Dkt. 1 at 3–4 (Compl. ¶¶ 5, 7); Dkt. 27-3 at 54. Both
sides have moved for summary judgment. Dkts. 26 & 29. The Court has also received an

allows for “significance” if the remainder would be merely threatened—an explicitly lower
standard. See id. at 37,579.
22

Plaintiffs have also submitted declarations attesting to their organizations’ conservationoriented missions, as well as their members’ concrete aesthetic and recreational interests in the
continued existence of blueback herring. See Dkt. 27-1 at 2–32. These declarations are
sufficient to establish Plaintiffs’ standing, see, e.g., Friends of the Earth, Inc. v. Laidlaw Envtl.
Servs. (TOC), Inc., 528 U.S. 167, 180–83 (2000), which the Service does not contest.
27

amicus brief, Dkt. 36, Plaintiffs’ response to that brief, Dkt. 38, both parties’ post-oral argument
briefs, Dkts. 48 & 50, and various notices of supplemental authority, Dkts. 41–46, 51–52.
II. ANALYSIS
Plaintiffs challenge all three of the Listing Decision’s central conclusions, which the
Court will address in the following order: First, did the Service reasonably conclude that the
Mid-Atlantic region is not a “significant portion” of the blueback herring’s range? Second, did
the Service reasonably conclude that blueback herring are not threatened throughout all their
entire range? And, third, did the Service reasonably conclude that the Mid-Atlantic stock
complex of blueback herring is not a “distinct population segment”?
In considering each of these contentions, the Court must be mindful of both the bounds
and demands of judicial review of administrative action. On the one hand, it is not the Court’s
role to substitute its judgment for that of agency, and this rule applies with particular force in a
case, like this one, involving scientific analysis and “technical expertise.” Marsh v. Or. Nat. Res.
Council, 490 U.S. 360, 377 (1989); see also Balt. Gas & Elec. Co. v. NRDC, 462 U.S. 87, 103
(1983) (“When examining this kind of scientific determination, . . . a reviewing court must
generally be at its most deferential.”). “When specialists express conflicting views,” for
example, “an agency must have discretion to rely on the reasonable opinions of its own qualified
experts even if, as an original matter, a court might find contrary views more persuasive.”
Marsh, 490 U.S. at 378; see also, e.g., Defs. of Wildlife v. Zinke, --- F.3d ----, 2017 WL 836089,
at *9 (D.C. Cir. Mar. 3, 2017) (“[C]ompeting views about scientific data and policy choices . . .
fail to show that the Service’s conclusions were arbitrary and capricious or contrary to law.”).
Nor is an agency required to explain its conclusions with crystal clarity; rather, the
Administrative Procedure Act (“APA”) merely requires that the reviewing court be able

28

“reasonably [to] discern[]” the “agency’s path.” Alaska Dep’t of Envtl. Conservation v. EPA,
540 U.S. 461, 497 (2004) (citation omitted).
On the other hand, it is the Court’s job to ensure that the agency has “examine[d] the
relevant data and [has] articulate[d] a satisfactory explanation for its action.” Motor Vehicle
Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983). The APA
“established a scheme of ‘reasoned decisionmaking,’” and thus the “the process by which [an
agency] reaches [its decreed] result must be logical and rational.” Allentown Mack Sales &
Serv., Inc. v. NLRB, 522 U.S. 359, 374 (1998) (quoting State Farm, 463 U.S. at 52); see also,
e.g., Tripoli Rocketry Ass’n v. Bureau of Alcohol, Tobacco, Firearms, & Explosives, 437 F.3d
75, 77 (D.C. Cir. 2006) (same); Fisher v. Pension Benefit Guarantee Corp., 151 F. Supp. 3d 159,
165 (D.D.C. 2016) (same). As the Supreme Court explained in State Farm, “[t]he scope of
review under the ‘arbitrary and capricious’ standard is narrow,” but the agency must “articulate a
satisfactory explanation for its action including a ‘rational connection between the facts found
and the choice made.’” 463 U.S. at 43 (quoting Burlington Truck Lines, Inc. v. United States,
371 U.S. 156, 168 (1962)); see also Sw. Power Pool, Inc. v. FERC, 736 F.3d 994, 997 (D.C. Cir.
2013) (same). Although review of the agency’s reasoned decision is deferential, where the
agency “entirely fail[s] to consider an important aspect of the problem” at issue, the Court must
set the agency’s action aside as “arbitrary and capricious.” State Farm, 463 U.S. at 43.
Applying these standards here, much of what the Service did is unobjectionable. In at
least three key respects, however, the Listing Decision either failed to offer a rational connection
between the facts and its conclusions or entirely failed to consider an important aspect of the
problem. The Court, accordingly, will vacate the Listing Decision and remand to the Service for
further consideration consistent with this opinion.

29

A.

Challenges to the “Significant Portion of Its Range” Analysis
The most salient of these errors occurred in the Service’s determination that the species is

not threatened in a “significant portion of its range.” As explained above, the Endangered
Species Act requires that the Service list a species it if it is threatened “throughout all or a
significant portion of its range.” 16 U.S.C. § 1532(20) (emphasis added). The Service
understood the relevant “portions of [the] range” to be the five geographic stock complexes it
had previously identified. See Listing Decision, 78 Fed. Reg. at 48,950, 48,993. According to its
2011 Draft SPR Policy, the Service was then required to ask “whether there is substantial
information suggesting that the hypothetical loss of any of the[se] individual stock complexes . . .
would reasonably be expected to increase the demographic risks to the point that the [entire]
species would then be in danger of extinction.” Id. at 48,993. Because the only stock complex
for which the Service had statistically significant evidence of population decline was the MidAtlantic stock complex, the Service performed the analysis only for that region. See id. If the
loss of the Mid-Atlantic region would not place the entire species in danger of extinction, the
Service reasoned, it “need not address the question whether the portion of the species occupying”
that region was threatened or endangered. Id.
Against this backdrop, the Service dispensed with the SPR inquiry in just a few
sentences. Given the centrality of this analysis to the Court’s holding, it is worth quoting in full:
For blueback herring, 3 of the stock complexes and the coastwide trend are all
stable, but the [M]id-Atlantic stock complex is decreasing. The [internal status
review team] determined that the [M]id-Atlantic stock complex is not significant to
the species, given that[,] even though it is decreasing, the overall coastwide trend
is stable. Thus, the loss of this stock complex would not place the entire species at
risk of extinction. We concur with this conclusion. Because the portion of the
blueback herring stock complex residing in the [M]id-Atlantic is not so significant
that its hypothetical loss would render the species endangered, we conclude that the
[M]id-Atlantic stock complex does not constitute a significant portion of the
blueback herring’s range.
30

Id.; accord id. at 48,992 (explaining same rationale). This analysis, however, is based on a
premise (that the coastwide trend was “stable”) and an assumption (that the Mid-Atlantic and
coastwide complexes did not overlap) that are not supported by the Listing Decision. The Court
will consider each of these propositions in turn. 23
1.

Inference That the Coastwide Trend Was “Stable”

The Service’s determination that the Mid-Atlantic region does not constitute a significant
portion of the blueback herring’s range depended on the factual premise that “the overall
coastwide trend is stable.” Id. at 48,993. The Service’s only stated basis for that claim was the
“coastwide” trend analysis model, which estimated the average annual change in relative
abundance of ocean-phase blueback herring, and generated a 95% confidence interval for that
result. See supra Part I.C.2.b. As explained above, that estimate and its confidence interval are
reflected in the top line of the following chart:

23

Plaintiffs also urge the Court to reject the 2011 Draft SPR Policy’s gloss on the statutory
phrase “significant portion of its range” on the grounds that it is contrary to the Act and that draft
policies deserve no deference. See Dkt. 27-3 at 33–40; Dkt. 34 at 18–23. The draft policy,
however, is no longer in effect. See 2014 Final SPR Policy, 79 Fed. Reg. at 37,578; see also
supra note 21. Plaintiffs agree that the new policy is more lenient, and that, if rationally applied,
it might qualify the Mid-Atlantic region as “significant.” See Dkt. 53 at 31–32. They also agree
that their arguments regarding deference do not apply to the new policy. See id. at 41. Thus,
because other flaws already necessitate a remand, because the remand will not involve the 2011
Draft SPR Policy, and because the 2014 Final SPR Policy presents a materially different set of
issues, the Court need not decide whether the 2011 Draft SPR Policy was lawful.

31

Figure 4 (repeated): Estimates and 95% Confidence Intervals for Each Population’s
Average Annual Change in Relative Abundance 24

In making its decision, however, the Service did not consider the value of the point
estimate (i.e., the midpoints identified above). Nor did it consider the length of the error bars.
Instead, it considered only the fact that the 95% confidence interval “contain[ed] zero.” Listing
Decision, 78 Fed. Reg. at 48,987. In other words, because the data did not show with 95%
confidence that the population size had changed, the Service “considered [the population] to be
stable.” Id.; accord id. at 48,989. Plaintiffs contend that this inference lacked the “rational
connection between the facts found and the choice made” that the APA requires. State Farm,
463 U.S. at 43; see Dkt. 27-3 at 21–29; Dkt. 34 at 11–17. In short, they argue that the absence of
statistically significant evidence that the coastwide population is declining (or increasing) does
not—absent more—show that it is stable. As explained below, the Court agrees with this

24

Adapted from Dkt. 40-2 at 108 (AR 2514). As stated above, for clarity, the Court has
replaced the “Coastwide” label, which originally read “Range-wide,” and replaced the x-axis
label, which originally read “Growth rate ± 95% confidence intervals.”
32

undisputed tenet and, accordingly, concludes that the Service has committed one of two errors:
It has either committed the logical error that Plaintiffs have identified or, at the very least, it has
failed to explain why it has not done so. See State Farm, 463 U.S. at 43.
It is helpful to begin with some background, none of which is disputed for present
purposes. In conducting a formal significance test, researchers start from the hypothesis they are
seeking to disprove, called the “null hypothesis” or “null.” David H. Kaye & David A.
Freedman, Reference Guide on Statistics, in REFERENCE MANUAL ON SCIENTIFIC EVIDENCE 211,
241 (Fed. Judicial Ctr. 3d ed. 2011) [hereinafter Reference Guide]. 25 They then ask: Assuming
the null hypothesis is true, what is the probability of observing data that conflict with the null
hypothesis to at least as great an extent as do the data actually observed? Id. at 249–51. To take
a simple example, suppose a study seeks to draw inferences about the color of apples in an
orchard, and begins from a random sample of 10 apples, 3 of which are green. If the null
hypothesis is “most apples in this orchard are green,” the question would be: “If most apples in
this orchard are, in fact, green, then what is the probability of choosing 10 apples at random from
the orchard and finding 3 or fewer green ones?” If the answer falls below a predetermined
significance threshold (often 5%), then the null hypothesis is sufficiently inconsistent with the
data that it is deemed “rejected,” and the result is deemed “statistically significant.” Id. at 251–

25

Judicial review is, of course, typically confined to the administrative record, but that does not
mean that courts cannot consult widely accepted and undisputed background materials for the
purpose of elucidating the issues presented. Any question of the Court’s ability to do so here,
moreover, is put soundly to rest by the D.C. Circuit’s decision in American Petroleum Institute v.
EPA, 684 F.3d 1342 (D.C. Cir. 2012), where it relied on the Federal Judicial Center’s Reference
Guide on Statistics to explain the concept of the null hypothesis in an APA case. See id. at 1350.
Although the Service objects to any use by the Court of “extra-record evidence,” see Dkt. 29-1 at
40–41 (emphasis added), it “[does] not take issue” with the Court relying on documents like the
Reference Guide (which the Service attaches to its brief) in order to “explain the [concept of] the
null hypothesis,” Dkt. 53 at 66; see Dkt. 29-2.
33

52; accord, e.g., Matrixx Initiatives, Inc. v. Siracusano, 563 U.S. 27, 39 n.6 (2011) (explaining
statistical significance). A “95% confidence interval,” then, is the interval containing all and
only the null hypotheses that could not be rejected using a 5% significance threshold. Richard D.
De Veaux et al., Stats: Data and Models 510–11 (2d ed. 2008); see also Reference Guide at 247.
Plaintiffs’ argument turns on the established principle that, when a statistical test fails to
reject the null hypothesis, that does not by itself constitute evidence that the null hypothesis is
true. See, e.g., Reference Guide at 251–52, 253–54; De Veaux et al. at 485, 501, 508. This
concept is familiar in law: A criminal verdict of “not guilty,” for example, does not establish
that the defendant did not commit the alleged conduct; it simply means that the government
failed to disprove the assumption of innocence beyond a reasonable doubt. Helvering v.
Mitchell, 303 U.S. 391, 397 (1938). So too in significance testing. Failing to reject the null does
not mean that the null hypothesis accurately describes the world; “it simply means that we don’t
have enough evidence to reject our assumption [that it does].” 26 De Veaux et al. at 485 (drawing
criminal law analogy). As the D.C. Circuit has explained, a “fail[ure] to reject the null
hypothesis” might mean that no trend in fact exists, but it is equally consistent with proposition
that a trend does exist and the test “merely lacked sufficient power to detect [it].” Am. Petroleum
Inst., 684 F.3d at 1350 (“[T]he study did not establish there was ‘no dose-response relationship’;
it simply failed to reject the null hypothesis to that effect.”); see also, e.g., Smith v. Wyeth-Ayerst

26

Professor De Veaux’s textbook admonishes students as follows:
When you have not found sufficient evidence to reject the null according to the
standard you have established, you should say that “The data have failed to
provide sufficient evidence to reject the null hypothesis.” Don’t say that you
“accept the null hypothesis.” You certainly haven’t proven or established it; it
was assumed to begin with.

De Veaux et al. at 508.
34

Labs. Co., 278 F. Supp. 2d 684, 693 n.14 (W.D.N.C. 2003) (“A study may fail to disprove the
null hypothesis, but that alone does not justify a conclusion that the null hypothesis has been
proved.”); cf. Matrixx Initiatives, Inc., 563 U.S. at 40 (“A lack of statistically significant
[evidence of an effect] does not mean that . . . experts have no reliable basis for [concluding that
no effect exists].”). As a result, just as a court cannot conclude that a criminal defendant did not
commit the underlying criminal conduct merely because the jury failed to find sufficient
evidence of guilt, reasoned decisionmaking does not permit an agency to conclude, based on a
failure to reject the null hypothesis and without further analysis, that the null hypothesis is true.
The Listing Decision, however, seems to have committed this very error. The Service
defined the null hypothesis as the proposition that the relative abundance of ocean-phase
blueback herring was the same in 2012 as it was in 1976 (i.e., that the average change in relative
abundance between those years was zero). The model, then, failed to reject the null. That is,
zero was among the values in the confidence interval for the estimated change in abundance over
the last 36 years—or, to put it in in different terms, the data did not show with a 95% degree of
confidence that the number of ocean-phase blueback herring had changed between 1976 and
2012. Listing Decision, 78 Fed. Reg. at 48,990; see supra Figure 4. And, based on the model’s
failure to reject the hypothesis that the relative abundance of the ocean-phase blueback herring
had not changed, the Service concluded (or at least seems to have concluded) that the oceanphase population was “stable.” Listing Decision, 78 Fed. Reg. at 48,987; see also Dkt. 29-1 at
44–45 (Defs.’ Brief) (confirming this description of the Listing Decision).
Before concluding that the Listing Decision was therefore unreasonable, the Court must,
of course, ensure that it is not substituting its judgment for that of the agency, see State Farm,
463 U.S. at 43, and that this not the type of technical issue that can reasonably lead to competing

35

expert perspectives, see, e.g., Marsh, 490 U.S. at 378. One could imagine, for example, that the
technical details of the model somehow accounted for and nullified this putative error. Or
perhaps Plaintiffs’ account of hypothesis testing—although “correct” enough to appear in college
textbooks and the Federal Judicial Center’s statistics manual—is at odds with other, more
complex statistical principles on which the Service relied. After three rounds of briefing and a
lengthy oral argument, however, the Service has not even argued that such an explanation exists.
To start, the Service does not appear to dispute the mathematical premise of Plaintiffs’
argument, i.e., that accepting an unrejected null hypothesis as true, absent further analysis, is
error. See Dkt. 29-1 at 39–48; Dkt. 37 at 12–17. To the contrary, the Service’s citations confirm
Plaintiffs’ description of the relevant principles. The Service directs the Court, for example, to a
“commonly cited scholarly article,” Dkt. 29-1 at 45–46 & n.28, which characterizes Plaintiffs’
position as “well known among applied scientists,” John M. Hoenig & Dennis M. Heisey, The
Abuse of Power: The Pervasive Fallacy of Power Calculations for Data Analysis, AMERICAN
STATISTICIAN, Feb. 2001, at 1, 1. That article’s opening paragraph states:
It is well known among applied scientists that a lack of impact or effect is
not sufficiently established by a failure to demonstrate statistical significance. A
failure to reject the null hypothesis of no effect may be the result of low statistical
power when an important effect actually exists and the null hypothesis of no effect
is in fact false.
Id. The Service also cites the website of Dr. Martha Smith, a retired statistics professor. See
Dkt. 29-1 at 46 n.30. Her website explains that “[a]ccepting a null hypothesis when a result is
not statistically significant, without taking power into account” is a “common [statistical]
mistake[].” Martha K. Smith, Common Mistakes Involving Power (Aug. 28, 2012), http://
www.ma.utexas.edu/users/mks/statmistakes/PowerMistakes.html. She continues:

36

Since smaller samples yield smaller power, a small sample size may not be able to
detect an important difference. If there is strong evidence that the power of a
procedure will indeed detect a difference of practical importance, then accepting
the null hypothesis may be appropriate; otherwise it is not—all we can legitimately
say then is that we fail to reject the null hypothesis.
Id. (footnote omitted) (emphasis added). Far from justifying the Listing Decision’s reasoning,
these sources support the view that it was flawed.
Nor does the Service argue that it relied on the type of additional analysis that might have
permitted it reasonably to draw affirmative conclusions from the rejection of the null. One such
type of analysis would evaluate the model’s “power,” which “is the chance that a statistical test
will declare an effect when there is an effect to be declared.” Reference Guide at 254; accord De
Veaux et al. at 515. The amount of power depends on both the size of the effect to be detected
and the amount of data considered. Reference Guide at 254. When a low-powered test fails to
reject the null hypothesis, the results are “more fairly described as inconclusive than negative.”
Id. But, when a high-powered test fails to reject the null hypothesis, that “can be persuasive
evidence” that the null is close to being correct. Id. Thus, demonstrating that the Service’s
significance test had high power would be one way to bridge the gap between failing to reject the
null and accepting it as true.
The Service, however, concedes that it did not conduct a power analysis. Indeed, it
argues that it would have been inappropriate to do so, because power analyses “are most useful
when designing a study (before data is collected),” and the data the Service used were collected
before the status review began. See Dkt. 29-1 at 45, 47 (citing Hoenig & Heisey at 5; Smith).
The statistical utility of a “post hoc power analysis” is beyond the judicial ken, but it is also
beside the point. Plaintiffs do not argue, nor do their views entail, that a power analysis was
required. Instead, they contend only that the Service impermissibly accepted an unrejected null
as true. See Dkt. 34 at 16–17. The fact that the Service considered a power analysis infeasible
37

does not cure that problem. A leaky ship is still leaky, even if it is not feasible to patch the leak
after the ship has gone to sea. The Service, moreover, has not identified anything in the
administrative record suggesting that it contemporaneously considered and rejected this—or any
other—means of assessing whether the null hypothesis is true. See Dkt. 29-1 at 45, 47. All
agree that the Court “must judge the propriety of [agency] action solely by the grounds invoked
by the agency.” SEC v. Chenery Corp., 332 U.S. 194, 196 (1947); see Dkt. 48 at 5. Because the
Service points to nothing in the administrative record suggesting that it even considered whether
or how it might be able to confirm the truth of the null hypothesis, the Court cannot credit the
Service’s “post hoc rationalization[]” now. Gerber v. Norton, 294 F.3d 173, 184 (D.C. Cir.
2002).
One could imagine the argument that the Service did not use the word “stable” to mean
“stable,” but rather used it to mean that it lacked data demonstrating a statistically significant
trend in population size. This line of reasoning, however, faces hurdles of its own. To the extent
the Service indeed takes that position now, it has contradicted itself in its briefs. Nor did the
Listing Decision itself state that it was employing such a counterintuitive definition. And, even
if the Service did mean to use “stable” in this way, it has not explained how a finding of
“stability”—understood to mean a lack of statistically significant evidence—would support the
conclusions it reached.
As an initial matter, the Service does not advance in its briefs the argument that “stable”
meant “unknown.” It writes, for example that “[t]he term ‘stable’ [in the Listing Decision]
mean[t] that the stock[’s population] [wa]s neither increasing nor decreasing, because the
estimated growth rate for the particular population [wa]s not statistically different from zero.”
Dkt. 29-1 at 44. There is a substantial difference between asserting that the data did not show

38

whether the population was changing, and affirmatively concluding, as the Service appears to
have done, that the population was, in fact, “neither increasing nor decreasing.” One assertion
posits a lack of knowledge, while the other asserts a fact based on the absence of statistically
significant evidence to the contrary.
The Service’s briefs go on to agree, moreover, that its inability to reject the null was
consistent with at least two possibilities: (1) that the null hypothesis (that there was no trend) was
actually true or (2) that there was a trend (i.e., the null was false) but that the data lacked
sufficient power to detect it. See Dkt. 37 at 13; Dkt. 53 at 69. The inability to distinguish those
two possibilities, however, is precisely the problem. As the Service’s own sources explain,
without some reason to favor the first possibility over the second, simply assuming the first to be
true constitutes an unreasoned choice. Cf., e.g., Allentown, 522 U.S. at 374 (requiring “the
process by which [an agency] reaches [its] result [to] be logical and rational). And, to the extent
the Service intended to rely on the absence of evidence—that is, to rely on its inability to discern
whether the coastwide population was increasing, decreasing, or stable—as a reason why it could
not render a listing decision on the existing record, that effort fails because the Service never
“articulate[d] [a] rational connection between” that lack of evidence and its listing decision. See
State Farm, 463 U.S. at 43.
When viewed in the context of the Listing Decision itself, this difficulty is acute. The
Service recognized that the population of blueback herring is, in fact, declining in the MidAtlantic region. The 2011 Draft SPR Policy required the Service then to ask, if the Mid-Atlantic
stock complex went extinct, would the remainder of the species would become endangered? 76
Fed. Reg. at 77,002. But the Service asserted that it would not, because “the overall coastwide
trend is stable.” Listing Decision, 78 Fed. Reg. at 48,993. That is, the Service concluded that the

39

abundance of blueback herring in the “coastwide” region (i.e., the ocean) was sufficient to
protect the species against the hypothetical loss of the Mid-Atlantic stock complex. Although
the Court does not foreclose the possibility that the Service might conclude that its lack of
knowledge about whether the coastwide population would be sufficient to protect against such a
loss required further study before the Service could find that listing the species was warranted,
that is not what the agency said. To the contrary, the Service concluded that the coastwide
population “is stable,” that “the loss of [the Mid-Atlantic] stock complex would not place the
entire species at risk of extinction,” and that “the [M]id-Atlantic stock complex does not
constitute a significant portion of the blueback herring’s range.” Id. (emphases added). If the
Service meant something else, it was required to say so and to offer a reasoned explanation for
that alternative.
Ultimately, the Service’s main argument seems to be that the Court should defer to the
Service’s statistical analysis because it involved complicated questions of statistics and ecology.
See Dkt. 29-1 at 39–40; Dkt. 37 at 12–13; Dkt. 53 at 71–72. The Court is unpersuaded. It is of
course true that “agency determinations receive ‘an extreme degree of deference [when] they
involve complex judgments about sampling methodology and data analysis that are within the
agency’s technical expertise.” Alaska Airlines, Inc. v. TSA, 588 F.3d 1116, 1120 (D.C. Cir.
2009) (alteration in original) (quoting Kennecott Greens Creek Min. Co. v. Mine Safety & Health
Admin., 476 F.3d 946, 956 (D.C. Cir. 2007)). But Plaintiffs’ challenge here is not particularly
technical. They dispute none of the choices the Service made in selecting its data. Nor do they
dispute the design of the models themselves. See Dkt. 53 at 15–16. Rather, Plaintiffs challenge
the Service’s interpretation of the models’ output—namely, the inference that, because the
models failed to detect a trend, no trend in fact existed. Id. at 60. This was hardly a “complex

40

judgment about sampling methodology and data analysis;” it was a simple error of logic. And
the law is clear that “the process by which [an agency] reaches [its] result must be logical and
rational.” Allentown, 522 U.S. at 374. In any event, even in the most technical cases, the
Service must still “provide[] a complete analytic defense” of its “assumptions and methodology.”
Nat’l Ass’n for Surface Finishing v. EPA, 795 F.3d 1, 18 (D.C. Cir. 2015). The Service has not
done so here. And not even the most deferential standard of review can save an agency action
for which the agency can articulate no reasoned basis. See ALLTEL Corp. v. FCC, 838 F.2d 551,
562 (D.C. Cir. 1988).
Finally, the Service contends that this ruling will “undermine a fundamental statutory
assumption of the [Endangered Species Act], i.e., the default that a species is not warranted for
listing unless the best available science indicates otherwise.” Dkt. 37 at 14–15; see also Dkt.
29-1 at 46–47. It is not obvious why the Service believes that to be so. The Service’s concern
appears to be that, on remand, if it cannot show with 95% certainty that blueback herring are not
threatened, the Service will be required to list the species. Nothing in this opinion mandates that
result. The Court merely holds that the Listing Decision failed to offer a “rational connection”
between the agency’s inability to reject the null hypothesis that the trend had not changed, and its
conclusion that the coastwide population was, in fact, stable. The Listing Decision’s conclusion
that the Mid-Atlantic stock complex does not constitute a significant portion of the blueback
herring range must, accordingly, be set aside. 27

27

This same logic also fells the Service’s “threatened throughout all of its range” analysis. See
infra Part II.B.
41

2.

Assumption That the Coastwide and Mid-Atlantic Stock Complexes Were Disjoint

The Service’s “significant portion of its range” analysis also suffers from a second,
critical lapse in reasoning. As explained above, the Service applied its 2011 Draft SPR Policy to
determine whether the Mid-Atlantic complex constituted a significant portion of the overall
range for blueback herring. Under that draft policy, the Service was required to hypothesize a
scenario in which the members of Mid-Atlantic stock complex perish, and then to ask whether,
without that stock complex, the rest of the species would become endangered. 2011 Draft SPR
Policy, 76 Fed. Reg. at 76,994, 77,002; see Listing Decision, 78 Fed. Reg. at 48,993. If so, the
Mid-Atlantic complex constitutes a “significant” portion of the range. But, if the loss of the
Mid-Atlantic complex would not pose a risk of extinction to the entire species, then that portion
of the range does not qualify as “significant” for purposes of the SPR analysis. Applying this
test, the Service concluded that the Mid-Atlantic complex was not “significant” because the
“coastwide trend is stable” and, thus, “the loss of [the Mid-Atlantic] stock complex would not
place the entire species at risk of extinction.” Listing Decision, 78 Fed. Reg. at 48,993.
Plaintiffs criticize this reasoning on a number of grounds, even apart from the use of the
term “stable.” See Dkt. 27-3 at 40–44; Dkt. 34 at 23–25; Dkt. 38 at 18–22. Most significantly
for present purposes, they contend that the Service “never actually examined what the effect
would be of extirpation of the Mid-Atlantic stock,” Dkt. 38 at 19–20, and that it failed to
“engag[e] in a meaningful analysis of what the loss of the Mid-Atlantic blueback population
would mean for the entire species,” Dkt. 34 at 23. The Court inquired about this issue at oral
argument, see Dkt. 53 at 34–37, 76–78, and the parties submitted supplemental briefs further
addressing the issue, see Dkt. 48 at 7–10; Dkt. 50 at 7–8.
As crystalized at oral argument and in the post-argument briefing, one logical flaw in the
Service’s reasoning stands out: The Service’s SPR analysis implicitly assumed that the Mid42

Atlantic stock complex and the coastwide population are disjoint—that is, that the two
populations do not overlap. By doing so, however, the Service did not truly hypothesize the loss
of the Mid-Atlantic stock complex, because Mid-Atlantic herring still contributed to the
coastwide trend analysis. In other words, the Service posited that the coastwide population
would compensate for the hypothetical total loss of the Mid-Atlantic stock complex, but it failed
to consider the fact that the loss of the Mid-Atlantic complex would invariably also diminish the
coastwide population, which is made up of a mix of fish that migrate from each of the stockspecific complexes. E.g., Listing Decision, 78 Fed. Reg. at 48,950.
In response, the Service argues that accounting for the overlap in population between the
Mid-Atlantic stock specific-complex and the coastwide mixed-stock population was, as a
practical matter, impossible. See Dkt. 48 at 6–10. It cites the Listing Decision for the
proposition that “[t]he coastwide [population] undisputedly represents a mixed stock (comprising
multiple stock complexes).” Id. at 8 (citing Listing Decision, 78 Fed. Reg. at 48,950). Thus, it
explains, in order to “subtract” the Mid-Atlantic stock complex from the coastwide model, the
Service would have needed to convert estimates of relative abundance into estimates of “absolute
abundance,” id. at 7, or else obtain genetic data from the ocean-phase population extending all
the way back to 1976, id. at 8. Neither option, the Service says, was viable. Id. at 10.
According to the Service, then, the only way to disentangle the “Mid-Atlantic” and “coastwide”
datasets would have been to “bas[e] its listing[] on speculation or surmise.” Id.
This leaves the Service in an untenable position. Faced with an unsupported assumption
on which its decision necessarily relied, the Service cannot identify any portion of the
administrative record in which it so much as acknowledged the gap in its reasoning. This
omission alone warrants remand. See, e.g., State Farm, 463 U.S. at 43 (remand is appropriate

43

when the agency “entirely failed to consider an important aspect of the problem”). The Service’s
assumption, however, was not merely unexplained; the Service now acknowledges that the
record “undisputedly” showed its critical, unstated assumption to be false. Dkt. 48 at 8. Suffice
it to say, it is arbitrary and capricious for an agency to base its decision on a factual premise that
the record plainly showed to be wrong. State Farm, 463 U.S. at 43 (an agency cannot “offer[] an
explanation for its decision that runs counter to the evidence before [it]”).
*

*

*

The Court, accordingly, will order the Service to conduct a new “signification portion of
[the] range” analysis, in which the Service (1) may not, at least without rational explanation,
accept the lack of statistically significant evidence of decline as proof that the population is
“stable”; and (2) may not assume that the ocean-phase population would be unaffected by the
loss of the Mid-Atlantic stock complex.
B.

Challenges to the Throughout “All of Its Range” Analysis
The Service’s determination that the blueback herring is not endangered “throughout all

of its range” shares at least one of the SPR determination’s fatal flaws. As described above, the
Service’s “throughout all of its range” analysis relied, at least in principal part, on the Tier B
interpretation of the quantitative models. Listing Decision, 78 Fed. Reg. at 48,992; see supra
Part I.C.2.b.iii. Under the Tier B approach, the Service’s assessment of species-wide extinctionrisk was a function of (1) the number of stock-specific models that showed a statistically
significant decline and (2) whether the coastwide model also showed a statistically significant
decline. See supra Table 3. Although the Service found statistically significant evidence that the
Mid-Atlantic population was declining, it was unable to reject the null hypothesis (i.e., the
hypothesis that there was no trend) with respect to the three other stock-specific models and the
“coastwide” model. The Service, accordingly, concluded that the coastwide population and
44

other stock-specific complexes were “stable,” and thus concluded that, overall, the blueback
herring faces only a “moderate-low risk” of extinction. Listing Decision, 78 Fed. Reg. at 48,992.
As with its SPR analysis, however, the Service appears to have committed one of two errors:
Either it committed a logical error—that is, it treated the models’ failure to reject the null
hypothesis as proof that the null hypothesis is true—or it failed adequately to explain the basis
for its decision.
As with the Service’s SPR determination, it is possible that the Service did not use the
word “stable” to mean “stable,” but rather meant to convey that it lacked sufficient information
to conclude that these stock complexes were decreasing. But, as with the SPR determination,
that is not what the Service said. To the contrary, the Service seems to have used the word
“stable” in something like the ordinary-language sense; it concluded that “a stable or
significantly increasing trajectory [in these complexes] suggests that the[] species may be within
the margins of being self-sustainable and thus, if all of the growth rates for the[se] . . . stock
complexes are stable or significantly increasing, the species is at low risk of extinction.” Id. at
48,987. And, of equal importance, the Service has failed to point to anything in the Listing
Decision or the administrative record explaining how evidence of a statistically significant
decline in the Mid-Atlantic population—combined with the lack of evidence regarding the
coastwide complex and the remaining three stock-specific complexes—supports a finding of a
“moderate-low” risk of extinction. Because the Service’s analysis, accordingly, either
committed a logical misstep, see Allentown, 522 U.S. at 374, or failed adequately to explain the
basis for the Service’s decision, see State Farm, 463 U.S. at 43, the agency’s finding with respect
to the blueback herring’s risk of extinction rangewide cannot stand.

45

The Court, as above, will order the Service on remand to conduct a new “through all of
its range” analysis in which it may not, at least without rational explanation, accept the lack of
statistically significant evidence of decline as proof that the population is “stable.” 28
C.

Challenges to the “Distinct Population Segment” Analysis
Finally, the Service’s “distinct population segment” (or “DPS”) analysis cannot withstand

APA scrutiny because the agency failed entirely to consider an important aspect of problem. See
State Farm, 463 U.S. at 43. Under the Service’s 1996 DPS Policy, any subset of blueback
herring qualifies as a “distinct population segment” if it is both “discrete” and “significant.” 61
Fed. Reg. at 4725. The Policy then lists four criteria to guide consideration of the “significance”
prong, “[a]ny one of [which] is sufficient” to establish “significance.” Am. Forest Res. Council
v. Ashe, 601 F. App’x 1, 4 (D.C. Cir. 2015); see also Dkt. 53 at 95 (statement of agency counsel
at oral argument) (“[A]ny of those four significance factors, if they’re positive, will result in
passing the significance test.”). Although the Listing Decision contains the conclusory assertion
that the agency “considered all of the [enumerated] criteria” for significance, 78 Fed. Reg. at
48,950 (emphasis added), Plaintiffs contend that the Service “provided no meaningful
evaluation” of the first such criterion—namely, whether any blueback herring stock
“[p]ersist[s] . . . in an ecological setting” that is “unusual or unique” among blueback herring.
1996 DPS Policy, 61 Fed. Reg. at 4725; see Dkt. 27-3 at 45. The Court agrees. Indeed, the
Service does not appear to have considered this criterion at all.

28

Plaintiffs also argue that the Service acted arbitrarily and capriciously when it determined that
“[t]here is insufficient information available to make a conclusion under Tier A as [it was]
unable to obtain data from Canada.” Listing Decision, 78 Fed. Reg. at 48,992; see Dkt. 27-3 at
29–32. In light of the Court’s other conclusions, it need not reach that issue.
46

In the Listing Decision itself, the Service points only to its statement that blueback
herring “occupy many different ecological settings throughout their range.” 78 Fed. Reg. at
48,950; see Dkt. 29-1 at 20–21. That statement, however, provides no support for the Service’s
contention that no population segment persists in an ecological setting that is “unusual or
unique”; to the contrary, if anything, it adds modest support for the proposition that one or more
of the regional stock complexes might contain an ecological setting that is “different” from the
settings in which other stocks reside. To the extent the Service considers this sentence in the
Listing Statement to be a “reasonabl[e] assessment of the first significance factor,” Dkt. 29-1 at
22, the Service is incorrect. It is, therefore, not surprising that the Service looks for support for
its conclusion in the broader administrative record.
There, however, the Service again fails to show that it considered the issue, much less
that it reached a reasoned conclusion. It directs the Court to (1) the map displayed below and (2)
an excerpt of an internal memorandum from the Service’s internal status review team. The map
appears to be a map of “ecoregions” (as defined by a third party in 2004), with circles overlaid to
represent the corresponding stock complexes for the alewife (a fish closely related to the
blueback herring). At least in relevant respects, the Mid-Atlantic stock complex for blueback
herring encompasses the Mid-Atlantic stock complex for alewife, so any “ecoregion” containing
alewife Mid-Atlantic stock should also contain blueback herring Mid-Atlantic stock. See Dkt.
40-6 at 238 (AR 73,490) (describing extent of both stock complexes). Compare Listing
Decision, 78 Fed. Reg. 48,951 fig.1 (alewife map) with id. at 48,952 fig.2 (blueback herring
map).

47

Figure 5: Alewife Stock Complexes Overlaid on
Map of “Terrestrial and Marine Ecoregions” 29

29

Dkt. 40-6 at 102 (AR 68,435). For the complete ecoregions map, see Dkt. 40-2 at 247 (AR
49,964).
48

The Service apparently intends these documents as evidence that it “considered the possibility of
[associating] ecoregions . . . with regional stock complexes.” Dkt. 29-1 at 21. It is difficult to
understand what the Service means by this, or how it is relevant. In any event, it is safe to
conclude that these material do not show that the Service considered whether the Mid-Atlantic
stock includes an ecoregion that is “unique or unusual” for blueback herring.
As above, this map, if anything, works against the Service. It suggests that the MidAtlantic stock complex wholly encompasses “unique” ecoregion No. 58, which the legend
identifies as the Chesapeake Bay Lowlands. See Dkt. 40-2 at 247 (AR 49,964). If so, then it
indicates that the Mid-Atlantic stock of blueback herring is the only stock to “persist” in the
Chesapeake Bay Lowlands—rendering it “significant” under the 1996 DPS Policy. Cf.
Threatened and Endangered Status for Distinct Population Segments of Atlantic Sturgeon in the
Northeast Region, 77 Fed. Reg. 5880, 5904–05 (Feb. 6, 2012) (concluding that, because only one
discrete population segment of Atlantic sturgeon lives in the Chesapeake Bay Lowlands
ecoregion, that population segment is “significant”).
Nor does the internal memorandum support the Service’s contention that it considered,
and rejected, the possibility that any of the four stock-specific complexes might contain a unique
ecoregion. The memorandum discusses what eventually became “Tier A” in the rangewide
extinction risk analysis. In a single sentence, the author suggests that the requirement that there
be three contiguous stock complexes that “are either stable or increasing” is a sensible one,
because “[i]f you have three contiguous stocks, they should span at least [two] ecoregions.” Dkt.
40-6 at 191–92 (AR 68,679–80). It contains no evidence that the author actually researched or
analyzed this question but, rather, notes: “Expand information with McElhany et al.” Id. at 192
(AR 68,680). If that reference was “expanded” elsewhere in the administrative record, the
49

Service has not directed the Court to it, and it is not the Court’s role to comb through the
voluminous joint appendix in search of it. As is, however, this single, incomplete sentence in an
internal memorandum does not vindicate the Service’s position that it did study the issue but
simply neglected to set forth its findings in the Listing Decision.
At bottom, there is no evidence that the Service gave any consideration whatsoever to the
first significance prong of its 1996 DPS Policy. It thereby “entirely failed to consider an
important aspect of the problem,” State Farm, 463 U.S. at 43, and the DPS finding must be set
aside. On remand, the Service will need to address this criterion, as well. 30
CONCLUSION
The Court will grant Plaintiffs’ motion for summary judgment and deny the Service’s
cross-motion for summary judgment. The Court will, accordingly, vacate the Listing Decision
with respect to blueback herring and will remand this matter to the Service for further
proceedings consistent with this opinion. A separate order will issue.

/s/ Randolph D. Moss
RANDOLPH D. MOSS
United States District Judge

Date: March 25, 2017

30

Plaintiffs also raise a number of challenges to the Listing Decision’s discussion of prongs two
and four of the 1996 DPS Policy. See Dkt. 27-3 at 47–54; see also Listing Decision, 78 Fed.
Reg. at 48,950. The Court need not consider those challenges at this time.
50