Standing Rock Sioux Tribe v. United States Army Corps

 United States Court of Appeals
         FOR THE DISTRICT OF COLUMBIA CIRCUIT



Argued November 4, 2020            Decided January 26, 2021

                        No. 20-5197

           STANDING ROCK SIOUX TRIBE , ET AL.,
                      APPELLEES

                             v.

        UNITED STATES A RMY CORPS OF ENGINEERS,
                      APPELLANT

                   DAKOTA A CCESS LLC,
                        INTERVENOR




                 Consolidated with 20-5201


        Appeals from the United States District Court
                for the District of Columbia
                    (No. 1:16-cv-01534)



    James A. Maysonett, Attorney, U.S. Department of Justice,
argued the cause for appellant United States Army Corps of
Engineers. With him on the briefs were Jeffrey Bossert Clark,
Assistant Attorney General, Jonathan D. Brightbill, Principal
Deputy Assistant Attorney General, Eric A. Grant, Deputy
                               2
Assistant Attorney General, and Andrew C. Mergen and Erica
M. Zilioli, Attorneys.

    Miguel A. Estrada argued the cause for appellant Dakota
Access LLC. With him on the briefs were William S. Scherman
and David J. Debold.

    Wayne K. Stenehjem, Attorney General, Office of the
Attorney General for the State of North Dakota, and Matthew
A. Sagsveen, Solicitor General, were on the brief for amicus
curiae the State of North Dakota.

     Tim Fox, Attorney General, Office of the Attorney General
for the State of Montana, Curtis T. Hill, Jr., Attorney General,
Office of the Attorney General for the State of Indiana, Thomas
M. Fisher, Solicitor General, Tom Miller, Attorney General,
Office of the Attorney General for the State of Iowa, Derek
Schmidt, Attorney General, Office of the Attorney General for
the State of Kansas, Daniel Cameron, Attorney General, Office
of the Attorney General for the Commonwealth of Kentucky,
Jeff Landry, Attorney General, Office of the Attorney General
for the State of Louisiana, Doug Peterson, Attorney General,
Office of the Attorney General for the State of Nebraska, Dave
Yost, Attorney General, Office of the Attorney General for the
State of Ohio, Jason Ravnsborg, Attorney General, Office of
the Attorney General for the State of South Dakota, Patrick
Morrisey, Attorney General, Office of the Attorney General for
the State of West Virginia, and Bridget Hall, Attorney General,
Office of the Attorney General for the State of Wyoming, were
on the brief for amici curiae the States of Indiana, Montana,
and 9 other states in support of appellants.

   David H. Coburn, Joshua H. Runyan, Richard S.
Moskowitz, Tyler J. Kubik, Stephen J. Obermeier, Wesley E.
Weeks, John P. Wagner, Steven M. Kramer, Steven P.
                               3
Lehotsky, and Michael B. Schon, were on the brief for amici
curiae American Fuel & Petrochemical Manufacturers, et al. in
support of appellants.

    Jared R. Wigginton and Kent Mayo were on the brief for
amici curiae North Dakota Farm Bureau, et al.

     Christopher O. Murray was on the brief for amicus curiae
for appellant North Dakota Water Users Association in support
of appellants.

    Jan Hasselman argued the cause for appellees Standing
Rock Sioux Tribe, et al. With him on the brief were Patti A.
Goldman, Nicole E. Ducheneaux, Jennifer S. Baker, Rollie E.
Wilson, Jeffrey Rasmussen, Michael L. Roy, Jennifer P.
Hughes, and Elliott A. Milhollin. Jeremy J. Patterson entered
an appearance.

    Joel West Williams was on the brief for amici curiae the
Great Plains Tribal Chairmen’s Association, et al. in support of
appellees.

    Maura Healey, Attorney General, Office of the Attorney
General for the Commonwealth of Massachusetts, Seth G.
Schofield, Senior Appellate Counsel, Xavier Becerra, Attorney
General, Office of the Attorney General for the State of
California, Jamie B. Jefferson and Joshua R. Purtle, Deputy
Attorneys General, Kathleen Jennings, Attorney General,
Office of the Attorney General for the State of Delaware,
Christian Douglas Wright, Director of Impact Ligitation,
Aaron M. Frey, Attorney General, Office of the Attorney
General for the State of Maine, William Tong, Attorney
General, Office of the Attorney General for the State of
Connecticut, Clare Kindall, Solicitor General, Kwame Raoul,
Attorney General, Office of the Attorney General for the State
                              4
of Illinois, Brian E. Frosh, Attorney General, Office of the
Attorney General for the State of Maryland, Dana Nessel,
Attorney General, Office of the Attorney General for the State
of Michigan, Elizabeth Morrisseau, Assistant Attorney
General, Gurbir S. Grewal, Attorney General, Office of the
Attorney General for the State of New Jersey, Letitia James,
Attorney General, Office of the Attorney General for the State
of New York, Aaron Ford, Attorney General, Office of the
Attorney General for the State of Nevada, Hector Balderas,
Attorney General, Office of the Attorney General for the State
of New Mexico, Ellen Rosenblum, Attorney General, Office of
the Attorney General for the State of Oregon, Paul Garrahan,
Attorney-in-Charge, Steven Novick, Special Assistant Attorney
General, Peter F. Neronha, Attorney General, Office of the
Attorney General for the State of Rhode Island, Tricia K.
Jedele, Special Assistant Attorney General, Robert W.
Ferguson, Attorney General, Office of the Attorney General
for the State of Washington, Noah Guzzo Purcell, Solicitor
General, Leevin T. Camacho, Attorney General, Office of the
Attorney General for the Territory of Guam, Thomas J.
Donovan, Jr., Attorney General, Office of the Attorney
General for the State of Vermont, Nicholas F. Persampieri,
Assistant Attorney General, Karl A. Racine, Attorney General,
Office of the Attorney General for the District of Columbia,
Loren L. AliKhan, Solicitor General, Jacqueline R. Bechara,
Appellant Litigation Fellow, and Sarah Utley were on the brief
for amici curiae States of Massachusetts, et al. in support of
appellees.

    Douglas P. Hayes was on the brief for amici curiae Sierra
Club, et al. in support of appellees.

    Kenneth Rumelt and James G. Murphy were on the brief
for amicus curiae Members of Congress in support of
appellees.
                              5

    Mary Kathryn Nagle was on the brief for amicus curiae
National Indigenous Women’s Resource Center, Inc. in
support of appellees.

   Before: TATEL and MILLETT, Circuit Judges, and
SENTELLE , Senior Circuit Judge.

    Opinion for the Court filed by Circuit Judge TATEL.

     TATEL, Circuit Judge: Lake Oahe, created when the
United States Army Corps of Engineers flooded thousands of
acres of Sioux lands in the Dakotas by constructing the Oahe
Dam on the Missouri River, provides several successor tribes
of the Great Sioux Nation with water for drinking, industry,
and sacred cultural practices. Passing beneath Lake Oahe’s
waters, the Dakota Access Pipeline transports crude oil from
North Dakota to Illinois. Under the Mineral Leasing Act, 30
U.S.C. § 185, the pipeline could not traverse the federally
owned land at the Oahe crossing site without an easement from
the Corps. The question presented here is whether the Corps
violated the National Environmental Policy Act, 42 U.S.C.
§ 4321, by issuing that easement without preparing an
environmental impact statement despite substantial criticisms
from the Tribes and, if so, what should be done about that
failure. We agree with the district court that the Corps acted
unlawfully, and we affirm the court’s order vacating the
easement while the Corps prepares an environmental impact
statement. But we reverse the court’s order to the extent it
directed that the pipeline be shut down and emptied of oil.

                              I.
    “In order to ‘create and maintain conditions under which
man and nature can exist in productive harmony,’ the National
Environmental Protection Act (NEPA), 42 U.S.C. § 4331(a),
                               6
requires any federal agency issuing a construction permit,
opening new lands to drilling, or undertaking any other ‘major’
project to take a hard look at the project’s environmental
consequences, id. § 4332(2)(C) . . . .” National Parks
Conservation Association v. Semonite, 916 F.3d 1075, 1077
(D.C. Cir. 2019). “To this end, the agency must develop an
environmental impact statement (EIS) that identifies and
rigorously appraises the project’s environmental effects, unless
it finds that the project will have ‘no significant impact.’” Id.
(quoting 40 C.F.R. § 1508.9(a)(1)). “If any ‘significant’
environmental impacts might result from the proposed agency
action[,] then an EIS must be prepared before agency action is
taken.” Grand Canyon Trust v. FAA, 290 F.3d 339, 340 (D.C.
Cir. 2002) (quoting Sierra Club v. Peterson, 717 F.2d 1409,
1415 (D.C. Cir. 1983)). Preparing an EIS is a significant
undertaking, requiring the agency to “consult with and obtain
the comments of” other relevant agencies and publish a
“detailed statement” about the action’s environmental effects.
42 U.S.C. § 4332(2)(C).

     “Whether a project has significant environmental impacts,
thus triggering the need to produce an EIS, depends on its
‘context’ (regional, locality) and ‘intensity’ (‘severity of
impact’).” National Parks, 916 F.3d at 1082 (quoting 40 C.F.R.
§ 1508.27 (2018)). The operative regulations (since amended,
Update to the Regulations Implementing the Procedural
Provisions of the National Environmental Policy Act, 85 Fed.
Reg. 43,304 (July 16, 2020)) enumerate ten factors that “should
be considered” in assessing NEPA’s “intensity” element. 40
C.F.R. § 1508.27(b) (2019). “Implicating any one of the factors
may be sufficient to require development of an EIS.” National
Parks, 916 F.3d at 1082. This case concerns the fourth factor—
“[t]he degree to which the effects on the quality of the human
environment are likely to be highly controversial.” 40 C.F.R.
§ 1508.27(b)(4) (2019).
                               7
     The Dakota Access Pipeline (DAPL), nearly 1,200 miles
long, is designed to move more than half a million gallons of
crude oil from North Dakota to Illinois each day. Standing
Rock Sioux Tribe v. U.S. Army Corps of Engineers (Standing
Rock III), 255 F. Supp. 3d 101, 114 (D.D.C. 2017). DAPL
crosses many waterways, including Lake Oahe, an artificial
reservoir in the Missouri River created when the Corps
constructed a dam in 1958. The dam’s construction and Lake
Oahe’s creation flooded 56,000 acres of the Standing Rock
Reservation and 104,420 acres of the Cheyenne River Sioux
Tribe’s trust lands. Id. The Tribes now rely on Lake Oahe’s
water for drinking, agriculture, industry, and sacred religious
and medicinal practices. Id. As the Standing Rock Sioux Tribe
explained:

    Lake Oahe is the source of life for the Tribe. It
    provides drinking water for over 4,200 people on the
    Reservation. It is the source of water for irrigation and
    other economic pursuits central to the Tribal
    economy. And it provides the habitat for fish and
    wildlife on the Reservation upon which tribal
    members rely for subsistence, cultural, and
    recreational purposes. Moreover, the Tribe’s
    traditions provide that water is more than just a
    resource, it is sacred—as water connects all of nature
    and sustains life.

Letter from Dave Archambault II, Chairman, Standing Rock
Sioux Tribe, to Lowry A. Crook, Principal Deputy Assistant
Secretary for Civil Works, Office of the Assistant Secretary for
the Army, and Col. John Henderson, P.E., District
Commander, U.S. Army Corps of Engineers—Omaha District
(Mar. 24, 2016), Appendix (A.) 318.
                              8
    Oil pipelines crossing federally regulated waters like Lake
Oahe require federal approval. See Standing Rock III, 255 F.
Supp. 3d at 114. In June 2014, Dakota Access, formed to
construct and own DAPL, notified the Corps that it intended to
construct a portion of DAPL under Lake Oahe, just half a mile
north of the Standing Rock Reservation. Id. To do so, Dakota
Access needed, among other things, a real-estate easement
from the Corps under the Mineral Leasing Act (MLA), 30
U.S.C. § 185.

     In December 2015, the Corps published and sought public
comment on a Draft Environmental Assessment (EA) finding
that the construction would have no significant environmental
impact. Standing Rock III, 255 F. Supp. 3d at 114–15. The
Tribes submitted comments voicing a range of concerns.
Relevant here, the Tribes contended that the Corps had
insufficiently analyzed the risks and consequences of an oil
spill.

     Two federal agencies also raised concerns. The
Department of the Interior requested that the Corps prepare an
EIS given the pipeline’s potential impact on trust resources,
criticizing the Corps for “not adequately justify[ing] or
otherwise support[ing] its conclusion that there would be no
significant impacts upon the surrounding environment and
community.” Letter from Lawrence S. Roberts, Acting
Assistant Secretary—Indian Affairs, U.S. Department of the
Interior, to Brent Cossette, U.S. Army Corps of Engineers,
Omaha District (Mar. 29, 2016), A. 385–86. The
Environmental Protection Agency (EPA) registered its concern
that the Draft EA “lack[ed] sufficient analysis of direct and
indirect impacts to water resources,” though it requested
additional information and mitigation in the EA rather than
preparation of an EIS. Letter from Philip S. Strobel, Director,
NEPA Compliance and Review Program Office of Ecosystems
                              9
Protection and Remediation, EPA, to Brent Cossette, U.S.
Army Corps of Engineers, Omaha District (Jan. 8, 2016),
Reply Supplemental Appendix 1. But after becoming aware of
the pipeline’s proximity to the Standing Rock reservation, EPA
supplemented its comments to note that, while it agreed with
the Corps that there was “minimal risk of an oil spill,” it
worried, based on its “experience in spill response,” that a
break or leak could nonetheless significantly affect water
resources. Letter from Philip S. Strobel, Director, NEPA
Compliance and Review Program, Office of Ecosystems
Protection and Remediation, EPA, to Brent Cossette, U.S.
Army Corps of Engineers, Omaha District (Mar. 11, 2016), A.
389–90.

     On July 25, 2016, the Corps published its Final EA and a
“Mitigated Finding of No Significant Impact” (Mitigated
FONSI). The Mitigated FONSI explained that, given the
Corps’s adoption of various mitigation measures, including
horizontal directional drilling, the Lake Oahe crossing would
not “significantly affect the quality of the human environment”
and that an EIS was therefore unnecessary.

    Shortly after the Final EA’s release, Standing Rock sued
the Corps for declaratory and injunctive relief under NEPA
(and several other federal laws not at issue in this appeal).
Standing Rock III, 255 F. Supp. 3d at 116–17. Dakota Access
and the Cheyenne River Sioux Tribe intervened on opposing
sides, and Cheyenne River filed a separate complaint adding
additional claims. Id. at 117. Though the district court denied
the Tribes’ request for a preliminary injunction on September
9, 2016, the Departments of Justice, Interior, and the Army
immediately issued a joint statement explaining that the Corps
would not issue an MLA easement and that construction would
not move forward until the Army could determine whether
                               10
reconsideration of any of its previous decisions was necessary.
Id.

     Following that statement, Standing Rock submitted
several letters to the Assistant Secretary of the Army for Civil
Works, who oversees the portion of the Corps’s mission that
includes issuing permits for pipelines like DAPL. Those letters
raised concerns about the EA’s spill risk analysis. The tribe
also submitted an expert review of the EA from an experienced
pipeline consultant who concluded that the assessment was
“seriously deficient and [could not] support the finding of no
significant impact, even with the proposed mitigations.”
Accufacts Review of the U.S. Army Corps of Engineers
Environmental Assessment for the Dakota Access Pipeline
(Oct. 28, 2016), A. 837–46. Following the Corps’s internal
review, the Assistant Secretary stood by her prior decision, but
nonetheless concluded that the historical relationship between
the affected tribes and the federal government “merit[ed]
additional analysis, more rigorous exploration and evaluation
of reasonable siting alternatives, and greater public and tribal
participation and comments.” Memorandum from Jo-Ellen
Darcy, Assistant Secretary of the Army (Civil Works) (Dec. 4,
2016), A. 260; see Standing Rock III, 255 F. Supp. 3d at 117–
18.

     During the ensuing review, both Standing Rock and the
Oglala Sioux Tribe submitted additional comments and
analysis. The Corps solicited Interior’s opinion on the pipeline,
Interior’s Solicitor responded with a recommendation that the
Corps prepare an EIS, and the Secretary of the Army for Civil
Works issued a memorandum directing the Army not to grant
an easement prior to preparation of an EIS. See Standing Rock
III, 255 F. Supp. 3d at 118–19. On January 18, 2017, the
Assistant Secretary of the Army for Civil Works published in
the Federal Register a notice of intent to prepare an EIS. See
                               11
Notice of Intent to Prepare an EIS in Connection with Dakota
Access, LLC’s Request for an Easement to Cross Lake Oahe,
North Dakota, 82 Fed. Reg. 5,543 (Jan. 18, 2017).

     Two days later, a new administration took office, and the
government’s position changed significantly. In a January 24
memorandum, the President directed the Secretary of the Army
to instruct the Corps and the Assistant Secretary for Civil
Works to expedite DAPL approvals and consider whether to
rescind or modify the Notice of Intent to Prepare an EIS.
Memorandum of January 24, 2017, Construction of the Dakota
Access Pipeline, 82 Fed. Reg. 8,661 (Jan. 30, 2017). The Army
in turn concluded that the record supported granting an
easement and that no EIS or further supplementation was
necessary.

     The Corps granted the easement on February 8, 2017, and
after the district court denied Cheyenne River’s motion for a
preliminary injunction and temporary restraining order, both
the Tribes and the Corps moved for partial summary judgment
on several claims. The district court concluded that the Corps’s
decision not to issue an EIS violated NEPA by failing to
adequately consider three issues: whether the project’s effects
were likely to be “highly controversial,” the impact of a
hypothetical oil spill on the Tribes’ fishing and hunting rights,
and the environmental-justice effects of the project. Standing
Rock III, 255 F. Supp. 3d at 111–12. It accordingly remanded
the matter to the agency to address those three issues. Id. at
160–61.

     After the Corps completed its remand analysis in February
2019, the parties again moved for summary judgment, with the
Tribes arguing that the Corps failed to remedy its NEPA
violations and pressing several other non-NEPA claims.
Standing Rock Sioux Tribe v. U.S. Army Corps of Engineers
                                12
(Standing Rock V), 440 F. Supp. 3d 1, 11 (D.D.C. 2020). Based
on its examination of four topics of criticism out of “many
. . . to choose from,” id. at 17, the district court concluded that
“many commenters in this case pointed to serious gaps in
crucial parts of the Corps’[s] analysis,” demonstrating that the
easement’s effects were “likely to be highly controversial,” id.
at 26 (internal quotation marks omitted). It therefore remanded
to the agency for it to complete an EIS but reserved the question
whether the easement should be vacated during the remand. Id.
at 29–30. Following additional briefing, the court concluded
that vacatur was warranted, Standing Rock Sioux Tribe v. U.S.
Army Corps of Engineers (Standing Rock VII), 471 F. Supp. 3d
71, 87 (D.D.C. 2020), and ordered that “Dakota Access shall
shut down the pipeline and empty it of oil by August 5, 2020,”
Order, Standing Rock Sioux Tribe v. U.S. Army Corps of
Engineers, No. 16-cv-01534-JEB, at 2 (D.D.C. July 6, 2020),
ECF No. 545.

     The Corps and Dakota Access now appeal the district
court’s order remanding for preparation of an EIS, as well as
its separate order granting vacatur of the pipeline’s MLA
easement and ordering that the pipeline be shut down. While
this appeal was pending, a motions panel denied the Corps’s
request to stay the vacatur of the easement but granted its
request to stay the district court’s order to the extent it enjoined
the pipeline’s use. Order, Standing Rock Sioux Tribe v. U.S.
Army Corps of Engineers, No. 20-5197, at 1 (D.C. Cir. Aug. 5,
2020) (August 5 Order).

                                II.
     The Corps, together with Dakota Access, challenges the
district court’s conclusion that the effects of the Corps’s
easement decision were “likely to be highly controversial”
under NEPA. A decision is “highly controversial,” we
explained in National Parks Conservation Association v.
                                13
Semonite, if a “substantial dispute exists as to the size, nature,
or effect of the major federal action.” 916 F.3d at 1083 (internal
quotation marks omitted). But not just any criticism renders the
effects of agency action “highly controversial.” Rather,
“something more is required for a highly controversial finding
besides the fact that some people may be highly agitated and
be willing to go to court over the matter.” Id. (internal quotation
marks omitted).

     In National Parks, we clarified what more is required.
There, we considered the Corps’s decision to forgo an EIS
before approving a permit authorizing an electrical
infrastructure project in a historically significant area. “[T]he
Corps’s assessment of the scope of the Project’s effects ha[d]
drawn consistent and strenuous opposition, often in the form of
concrete objections to the Corps’s analytical process and
findings, from agencies entrusted with preserving historical
resources and organizations with subject-matter expertise.” Id.
at 1086. Because those criticisms reflected “the considered
responses . . . of highly specialized governmental agencies and
organizations” rather than “the hyperbolic cries of . . . not-in-
my-backyard neighbors,” we found the effects of the Corps’s
decision “highly controversial.” Id. at 1085–86. “[R]epeated
criticism from many agencies who serve as stewards of the
exact resources at issue, not to mention consultants and
organizations with on-point expertise, surely rises to more than
mere passion.” Id. at 1085. And while the Corps “did
acknowledge and try to address [those] concerns,” that was not
enough to put the controversy to rest. Id. at 1085–86. “The
question is not whether the Corps attempted to resolve the
controversy, but whether it succeeded.” Id. Indeed, an EIS is
perhaps especially warranted where an agency explanation
confronts but fails to resolve serious outside criticism, leaving
a project’s effects uncertain. “Congress created the EIS process
to provide robust information in situations . . . where, following
                                14
an environmental assessment, the scope of a project’s impacts
remains both uncertain and controversial.” Id. at 1087–88.

     The Corps and Dakota Access advance two arguments:
that, in relying on National Parks, the “district court applied
the wrong legal standard,” Appellant’s Br. 14, and that the
Corps adequately addressed the four specific disputes on which
the district court relied in finding the effects of the Corps’s
easement decision likely to be highly controversial. We
disagree as to both.

     The Corps offers two bases for distinguishing this case
from National Parks. First, it argues that here, in contrast to in
National Parks, “the Corps’[s] efforts to respond to the Tribes’
criticisms were not ‘superficial.’” Appellant’s Br. 19. That
distinction, however, rests on an inaccurate description of
National Parks. Contrary to the Corps’s claim that we deemed
“superficial and inadequate” the Corps’s response to criticisms,
we pointedly explained that we took “no position on the
adequacy of the Corps’s alternatives analyses.” National
Parks, 916 F.3d at 1088. Instead, we noted only that other
agencies had expressed concerns about the superficiality and
inadequacy of the Corps’s efforts. Id. Furthermore, the Corps’s
position that a response to criticism suffices so long as it is not
“superficial” is hard to square with our statement in National
Parks that “[t]he question is not whether the Corps attempted
to resolve the controversy, but whether it succeeded.” Id. at
1085–86. The decisive factor is not the volume of ink spilled
in response to criticism, but whether the agency has, through
the strength of its response, convinced the court that it has
materially addressed and resolved serious objections to its
analysis, a matter requiring us to delve into the details of the
Tribes’ criticisms—to which we shall turn momentarily.
                               15
      As a second basis for distinguishing National Parks, the
Corps emphasizes that the “opposition here has come from the
Tribes and their consultants, not from disinterested public
officials.” Appellant’s Br. 20. But the Tribes are not, as Dakota
Access suggested at oral argument, “quintessential . . . not-in-
my-backyard neighbors.” Oral Arg. Tr. 97:17–18. They are
sovereign nations with at least some stewardship responsibility
over the precise natural resources implicated by the Corps’s
analysis. “Indian tribes within Indian country are,” the
Supreme Court has declared, “a good deal more than private,
voluntary organizations.” Merrion v. Jicarilla Apache Tribe,
455 U.S. 130, 140 (1982) (internal quotation marks omitted).
Rather, they are “domestic dependent nations that exercise
inherent sovereign authority over their members and
territories” and the resources therein. Oklahoma Tax
Commission v. Citizen Band Potawatomi Indian Tribe of
Oklahoma, 498 U.S. 505, 509 (1991) (internal quotation marks
omitted); see also New Mexico v. Mescalero Apache Tribe, 462
U.S. 324, 335 (1983) (“We have held that tribes have the power
to manage the use of [their] territory and resources by both
members and nonmembers . . . .”); Merrion, 455 U.S. at 140
(“Indian tribes . . . . are unique aggregations possessing
attributes of sovereignty over both their members and their
territory.” (internal quotation marks omitted)).

     The Tribes’ unique role and their government-to-
government relationship with the United States demand that
their criticisms be treated with appropriate solicitude. Of
course, as the Corps points out, the Tribes are not the federal
government. But in National Parks, we emphasized the
important role played by entities other than the federal
government. There, criticism came from “highly specialized
governmental agencies and organizations,” including the
Virginia Department of Historic Resources and several
conservation groups. National Parks, 916 F.3d at 1084–85; see
                               16
also North Carolina v. Federal Aviation Administration, 957
F.2d 1125, 1131–33 (4th Cir. 1992) (finding “legitimate
controversy” present where “[s]tate, local and federal officials,
interested individuals,” and a federal agency “expressed
concern”); Foundation for North American Wild Sheep v. U.S.
Department of Agriculture, 681 F.2d 1172, 1182 (9th Cir.
1982) (finding that criticism from “conservationists,
biologists,” two state agencies, and “other knowledgeable
individuals” demonstrated the existence of “precisely the type
of ‘controversial’ action for which an EIS must be prepared”);
Friends of the Earth, Inc. v. U.S. Army Corps of Engineers, 109
F. Supp. 2d 30, 43 (D.D.C. 2000) (finding that a project was
“genuinely and extremely controversial” where “three federal
agencies,” “one state agency,” and the public “all disputed the
Corps[’s] evaluation”). The Tribes are of at least equivalent
status.

    With the proper legal framework in mind, we turn to the
four disputed facets of the Corps’s analysis that the district
court found involved unresolved scientific controversies for
purposes of NEPA’s “highly controversial” factor.

               DAPL’s Leak Detection System
     The district court found that serious unresolved
controversy existed concerning the effectiveness of DAPL’s
leak detection system. Specifically, it found that the 2012
Pipeline and Hazardous Materials Safety Administration
(PHMSA) study submitted with Standing Rock’s expert report
“indicated an 80% failure rate in the type of leak-detection
system employed by DAPL.” Standing Rock V, 440 F. Supp.
3d at 18. The court went on to note that “the system was not
even designed to detect leaks that constituted 1% or less of the
pipe’s flow rate,” which could amount to 6,000 barrels a day.
Id. Because the Corps “failed entirely to respond to” those
deficiencies, the court found that the Corps had not succeeded
                              17
in resolving the controversy presented by the study. Id. at
17–18.

     On appeal, the Corps correctly points out that the 2012
PHMSA study does not reflect an 80% “failure rate.” Rather,
the study indicates that in 80% of all incidents where it was in
use and “functional,” the “computational pipeline monitoring”
(CPM) system used by DAPL was not the first system to detect
a leak. That the CPM system was commonly eclipsed by visual
identification, however, casts serious, unaddressed doubt on
the Corps’s statement that the system will “detect the pressure
drop from a pipeline rupture within seconds.” Appellant’s Br.
21 (internal quotation marks omitted). As the PHMSA study
explains, “CPM systems by themselves did not appear to
respond more often than personnel . . . or members of the
public passing by the release incident.” U.S. Department of
Transportation, Pipeline and Hazardous Materials Safety
Administration, Final Report Leak Detection Study 2-11 (Dec.
10, 2012). The Corps has failed to address the apparent
disconnect, suggested by the PHMSA study, between the CPM
system’s historic performance and the agency’s representations
about its future utility. Indeed, the Corps acknowledges that it
“did not explicitly discuss the 2012 PHMSA report” in its
review. Appellant’s Br. 22. The consequences of that oversight
are especially significant since DAPL is buried deep
underground and visual identification is therefore unlikely to
make up for deficiencies in the CPM system, as it apparently
has in the incidents included in the PHMSA study.

     Attempting to discount the significance of the Corps’s
failure to consider the 2012 PHMSA study, the Corps and
Dakota Access observe that the study included older pipelines
and that the type of pinhole leaks the study suggests the CPM
system might initially miss are rare. But as the district court
noted, the Tribes’ expert observed that “more recent
                               18
investigations” corroborated the study’s leak detection data.
Standing Rock V, 440 F. Supp. 3d at 17 (internal quotation
marks omitted). The Corps’s failure to address the study cannot
be justified by the mere fact that the study’s data set includes
some older pipelines.

     As for the rarity of pinhole leaks, the Tribes pointed to
“numerous examples of pipelines that leaked for hours or days
after similar detection systems failed.” Appellees’ Br. 27. In
one such instance, DAPL’s own operator spilled 8,600 barrels
of oil during a 12-day-long slow leak in 2016, even though the
monitoring system in use there showed the exact same type of
“detectable meter imbalance” that the Corps here claims will
quickly alert DAPL’s operators to a slow leak. See
Supplemental Appendix (S.A.) 317–18. That same year, at
another pipeline buried deep underground in North Dakota, an
operator’s leak detection system “registered an imbalance” and
“notified the control room”—but the control room
“misinterpreted its own data[.]” PHMSA, Post-Hearing
Decision Confirming Corrective Action Order, Belle Fourche
Pipeline Co. 5 (Mar. 24, 2017), https://primis.phmsa.dot.gov/
comm/reports/enforce/documents/520165013H/520165013H_
HQ%20Post%20Hearing%20Decision%20Confirming%20C
AO_03242017.pdf. That led to a slow release of more than
12,600 barrels of oil into a nearby creek over at least a two-day
period, until it was discovered by a rancher at the release site.
Id. at 1–2; S.A. 711. So there is ample reason to believe that
the magnitude of harm from such a leak could be substantial.

     Appearing to acknowledge those troubling examples, the
Corps discounts their significance by asserting that leaks will
eventually be found. But how rapidly such leaks would be
detected and their potential severity are key factors underlying
the Corps’s EA and precisely the issues called into question by
the Tribes’ unaddressed criticism. We also note that the volume
                                19
of a one percent spill from a pinhole leak would double if the
volume of oil placed in the pipeline were itself to double. And
DAPL’s operator has represented to its investors that it intends
to double the amount of oil it places in the pipeline as early as
this coming summer. See Illinois approves expansion of
Dakota Access oil pipeline, Reuters, Oct. 15, 2020,
https://www.reuters.com/article/us-energy-transfer-oil-
pipeline-illinois-idUSKBN2702DL. In any event, when asked
why the EA did not evaluate the potential consequences of an
undetected slow pinhole leak, the Corps responded that “there
was no particular reason” it did not do so. Oral Arg. Tr. 12:8–9,
Standing Rock Sioux Tribe v. U.S. Army Corps of Engineers,
No. 16-cv-01534-JEB (D.D.C. Mar. 18, 2020), ECF No. 498.
The Tribes’ criticisms therefore present an unresolved
controversy requiring the Corps to prepare an EIS.

               DAPL’s Operator Safety Record
     The district court found that the Corps’s decision to rely in
its risk analysis on general pipeline safety data, rather than
DAPL’s operator’s specific safety record, rendered the effects
of the Corps’s decision highly controversial. We agree.

     To analyze the Corps’s risk assessment, Standing Rock
retained as an expert “an attorney, investigator, and process
safety practitioner with many decades of experience.”
Holmstrom Decl. ¶ 1, S.A. 79–80. The expert explained that
“PHMSA data shows Sunoco,” DAPL’s operator, “has
experienced 276 incidents in 2006–2016,” which the expert
described as “one of the lower performing safety records of any
pipeline operator in the industry for spills and releases.” Id. ¶ 9.

     Here, as in the district court, “[t]he Corps focuse[s] its
responses on defending the operator’s performance record
itself rather than on justifying its decision to not incorporate
that record into its analysis.” Standing Rock V, 440 F. Supp. 3d
                               20
at 19. In so doing, the Corps and Dakota Access make two
arguments.

     First, the Corps emphasizes that “70% of [DAPL’s]
operator’s reported accidents on other pipelines were minor
and limited to the operator’s property.” Appellant’s Br. 31. But
that does nothing to address the “[t]wo central concerns” on
which the district court based its decision: “(1) the 30% of
spills—about 80 of them—that were not limited to operator
property; and (2) the criticism that the spill analysis should
have incorporated the operator’s record.” Standing Rock V, 440
F. Supp. 3d at 20. For its part, Dakota Access argues that while
Sunoco’s number of leaks is high, its number of spills per mile
of pipeline operated “is in line with industry averages.”
Intervenor’s Br. 22. Not only has Dakota Access failed to
identify record evidence supporting that assertion, the relevant
evidence that does exist suggests a serious risk that Sunoco’s
record is worse than the industry average. The Corps’s own
analysis concluded that, industry-wide, there were 0.953
onshore crude oil accidents per 1,000 miles of pipeline in 2016
and 0.848 in 2017. U.S. Army Corps of Engineers, Analysis of
the Issues Remanded by the U.S. District Court for the District
of Columbia Related to the Dakota Access Pipeline Crossing
at Lake Oahe 13 (Aug. 31, 2018). By contrast, Dakota Access’s
expert explained that Energy Transfer, Sunoco’s parent
company following a merger, experienced 1.42 “reportable
incidents per 1,000 miles of pipeline”—after a 50% decline in
incidents on Sunoco lines since 2017. Second Godfrey Decl.
¶ 7, A. 1612. If anything, comparing that figure to the industry-
wide average understates the safety gap between Sunoco and
other operators because, as Dakota Access and its expert
observe, Sunoco is “one of the largest pipeline operators,”
Intervenor’s Br. 22, and its own incidents are included in the
average. See Appellant’s Br. 32 (“The Corps also considered
                                21
PHMSA’s historical data on oil spills, which necessarily
includes this operator’s safety record.”).

     Nor are we persuaded by the Corps’s second argument,
that it had no need at all to address the operator safety
controversy. Though the Corps may have considered “other
objective measures of the operator’s safety practices,”
Appellant’s Br. 31, the cited materials—industry-wide spill
data and a questionnaire about Sunoco’s safety practices—fall
short of resolving the controversy. The Corps contends that its
“decision to use all data on oil spills, and not just the operator’s
safety record, is the kind of technical judgment that is entrusted
to the agency and entitled to deference from the Court.”
Appellant’s Br. 32. That is not at all clear. For example, it
would be strange indeed if we were to defer to the Federal
Aviation Administration’s decision to renew the operating
certificate of an airline with an extremely poor safety record on
the basis that the airline industry, on average, is safe. The
Supreme Court, moreover, has “frequently reiterated that an
agency must cogently explain why it has exercised its
discretion in a given manner,” Motor Vehicle Manufacturers
Ass’n of the United States, Inc. v. State Farm Mutual
Automobile Insurance Co., 463 U.S. 29, 48 (1983), and the
Corps has made no effort to do so here. To treat the Corps’s
unadorned plea for deference as a sufficient basis for ignoring
well-reasoned expert criticism would vitiate National Parks.

                       Winter Conditions
     The district court found the Corps’s response insufficient
to resolve criticism of the agency’s “failure to consider the
impact of harsh North Dakota winters on response efforts in the
event of a spill.” Standing Rock V, 440 F. Supp. 3d at 20. In
particular, the Tribes’ experts explained that shut-off valves
might be more prone to failure and response efforts hindered
by freezing conditions. Elaborating, Oglala’s expert explained
                               22
that “winter conditions create significant difficulties” because,
among other things, “workers require more breaks and move
slower due to the bundling of clothing,” “daylight hours are
shorter,” and “slip-trip-fall risk increases significantly.”
Earthfax Report at 7, A. 830.

     The Corps argues that it had no need to engage in a
quantitative evaluation of a winter spill scenario because its
non-quantitative response was adequate. Appellant’s Br.
29–30. In the Corps’s view, it adequately considered winter
conditions by noting that ice coverage could “have a mixed
effect on efforts to contain an oil spill” and by ordering DAPL’s
operator to conduct winter spill response training exercises at
Lake Oahe as a condition of the easement. Appellant’s Br. 29.
But the Corps’s passing reference to winter conditions’
“mixed” effects, without more, provides little comfort. The
Corps’s point might have been more forceful had the agency
estimated just how much time during a spill would be saved by
the oil-containing properties of ice and compared that to the
additional time required to identify oil pockets and adjust work
methods to extreme conditions. Indeed, it seems that such an
analysis is precisely what the Tribes believe the Corps ought to
have done, and such a reasoned weighing of the evidence
would have been entitled to substantial deference. But instead,
faced with serious expert criticism, the Corps simply declared
the evidence “mixed” and offered no attempt at explaining its
apparent conclusion that winter’s countervailing effects
measured out to zero. Moreover, we agree with the district
court that while winter response training may be “prudent and
perhaps a good avenue for producing data as to how exactly
winter conditions would delay response efforts,” such exercises
do “not get to the point of addressing the concern that the spill
model does not currently take that kind of data into account.”
Standing Rock V, 440 F. Supp. 3d at 21.
                               23
     The Corps next argues that the Tribes failed to present a
“specific alternative methodology” for incorporating winter
conditions into its spill response modeling. Appellant’s Br. 30.
But the fact that an established methodology for assessing the
consequences of a unique type of risk is not readily apparent to
commenters hardly means an agency can discount relevant,
serious criticism of its method of analysis. Although the Corps
emphasizes in its brief that “no one has identified any way to
calculate exactly how much more difficult” a clean-up would
be during winter, Appellant’s Br. 30, our review “is limited to
the grounds that the agency invoked when it took the action,”
Department of Homeland Security v. Regents of the University
of California, 140 S. Ct. 1891, 1907 (2020) (internal quotation
marks omitted), and the Corps does not suggest that, during its
environmental review process, it actually applied its technical
expertise to consider whether it was possible to identify such a
method. Had the Corps considered the problem and concluded
that no comprehensive analysis was possible, that might have
amounted to “successfully” resolving the controversy. But the
Corps cannot foist its duty to consider such technical matters
onto commenters who point out valid deficiencies.

                    Worst Case Discharge
    The district court considered the “largest area of scientific
controversy” to be “the worst-case-discharge estimate for
DAPL used in the spill-impact analysis.” Standing Rock V, 440
F. Supp 3d at 21. The regulations set forth a detailed formula
for calculating the worst-case discharge, 49 C.F.R.
§ 194.105(b)(1), but we need not delve into its specifics here.
“The idea,” the district court succinctly explained, “is to
calculate the maximum amount of oil that could possibly leak
from the pipeline before a spill is detected and stopped.”
Standing Rock V, 440 F. Supp. 3d at 21.
                               24
     According to the Corps, we need not consider the Tribes’
criticisms because “an accident leading to a full-bore rupture
of the pipeline is extremely unlikely” and, in any event, no
statute or regulation required the Corps to calculate the worst-
case discharge at all. Appellant’s Br. 26. The thrust of both
arguments is that because the Corps need not have calculated a
worst-case discharge in the first place, it is unimportant
whether it did so in a reasonable manner. But we agree with the
district court that because the Corps chose to perform such a
calculation and then relied on it throughout its analysis, it
cannot dispel serious doubts about its methods by explaining
that it could have forgone such a calculation in the first place.
See Sierra Club v. Sigler, 695 F.2d 957, 966 (5th Cir. 1983)
(“The purpose of judicial review under NEPA is to ensure the
procedural integrity of the agency’s consideration of
environmental factors in the EIS and in its decision to issue
permits. If the agency follows a particular procedure, it is only
logical to review the agency’s adherence to that procedure, not
to some altogether different one that was not used.”). We
therefore turn to the Tribes’ criticisms of the Corps’s
calculations.

     The Corps estimated that, for purposes of a worst-case
discharge, it would take 9 minutes to detect a leak and 3.9
minutes to close the shut-down valves. Appellant’s Br. 26–27.
Before the district court, the Corps suggested that its nine-
minute figure included one minute of detection time, with the
remaining eight minutes devoted to shutting down the mainline
pumps. Standing Rock V, 440 F. Supp. 3d at 23. But as the
district court observed, the Tribes pointed to “many experts
who commented that hours, rather than minutes, were more
accurate figures for the [worst-case discharge].” Id. The Tribes’
expert explained that “[m]ajor spill incidents typically occur
with multiple system causes, when people, or equipment, or
systems do not function exactly as they are expected to.”
                               25
Holmstrom Decl. ¶ 11, S.A. 83. The Corps’s explanation that
its response time estimates were mildly conservative does not
begin to explain its choice to ignore the real-world possibility
of significant human errors or technical malfunctions, see
supra at 18–19, in calculating what it claimed was a worst-case
estimate. Although the PHMSA formula did not require the
Corps to model a complete doomsday scenario in which every
possible human error and technical malfunction occurs
simultaneously, we agree with the district court that the Corps’s
failure to explain why it declined to consider any such
eventualities leaves unresolved a substantial dispute as to its
worst-case discharge calculation.

     The Corps also argues that, even if, as the Tribes claim,
some aspects of the model are unduly optimistic, the model is
nonetheless sufficiently conservative because it assumes the
pipeline lies directly on top of the water rather than beneath
ninety-two feet of overburden. Appellant’s Br. 25–26. In effect,
the Corps tries to defend its decision to develop a model that
assumes away significant risks by explaining that, despite those
omissions, it analyzed an imaginary pipeline of roughly
equivalent risk to DAPL—one laying directly on top of Lake
Oahe, but with superior leak detection and shut-down valve
systems. The Corps, however, never explains why its one
conservative assumption accurately counterbalances the
particular risks the Tribes identify. Accordingly, the model’s
assumption that DAPL lies directly on the water fails to resolve
the controversies raised by the Tribes’ criticisms.

                             ***

     Having determined that several serious scientific disputes
mean that the effects of the Corps’s easement decision are
likely to be “highly controversial,” we turn to one other issue
before considering the appropriate remedy. The Corps and
                               26
Dakota Access repeatedly urge that, whatever the merits of the
Tribes’ criticisms, the Corps’s easement decision cannot be
highly controversial because the risk of a spill is exceedingly
low and because the pipeline’s location deep underground
provides protection against the consequences of any spill. That
argument faces two major hurdles.

     First, the claimed low risk of a spill rests, in part, on the
Corps’s use of generalized industry safety data and its
optimism concerning its ability to respond to small leaks before
they worsen—precisely what the Tribes’ unresolved criticisms
address. Second, as our court made clear in New York v.
Nuclear Regulatory Commission, 681 F.3d 471, 478–79 (D.C.
Cir. 2012), “[u]nder NEPA, an agency must look at both the
probabilities of potentially harmful events and the
consequences if those events come to pass.” Id. at 148. A
finding of no significant impact is appropriate only if a grave
harm’s “probability is so low as to be remote and speculative,
or if the combination of probability and harm is sufficiently
minimal.” Id. at 147–48 (internal quotation marks omitted).
Doing away with the obligation to prepare an EIS whenever a
project presents a low-probability risk of very significant
consequences would wall off a vast category of major projects
from NEPA’s EIS requirement. After all, the government is not
in the business of approving pipelines, offshore oil wells,
nuclear power plants, or spent fuel rod storage facilities that
have any material prospect of catastrophic failure. In this case,
although the risk of a pipeline leak may be low, that risk is
sufficient “‘that a person of ordinary prudence would take it
into account in reaching a decision’” to approve the pipeline’s
placement, and its potential consequences are therefore
properly considered here. Sierra Club v. FERC, 827 F.3d 36,
47 (D.C. Cir. 2016) (quoting City of Shoreacres v. Waterworth,
420 F.3d 440, 453 (5th Cir. 2005)).
                               27
                               III.
    This brings us to the Corps’s challenge to the district
court’s remedy, and specifically to its orders (1) requiring that
the Corps prepare an EIS, (2) vacating the easement pending
preparation of an EIS, and (3) ordering that the pipeline be shut
down and emptied of oil.

     As already explained, “[i]mplicating any one of the
[intensity] factors may be sufficient to require development of
an EIS.” National Parks, 916 F.3d at 1082. Dakota Access
argues that because implicating the “highly controversial”
factor does not itself mandate preparation of an EIS, the district
court erred in ordering the Corps to prepare one. In National
Parks, however, we ordered the Corps to prepare an EIS where,
as here, it “failed to make a ‘convincing case’ that an EIS is
unnecessary.” Id. at 1087 (quoting Myersville Citizens for a
Rural Community, Inc. v. FERC, 783 F.3d 1301, 1322 (D.C.
Cir. 2015)). National Parks thus forecloses the idea that we
must ordinarily remand to the agency to weigh the intensity
factors anew whenever we find that it improperly analyzed one
of them.

     That National Parks involved multiple intensity factors is
at most a superficial distinction between this case and National
Parks. For one thing, as explained above, the effects of the
Corps’s easement decision are “highly controversial” in four
distinct respects, and we see no good reason for treating
differently a decision that implicates multiple significance
factors and a decision that implicates a single factor in several
important ways. Moreover, both National Parks and this case
present “precisely” the circumstances in which Congress
intended to require an EIS, namely “where, following an
environmental assessment, the scope of a project’s impacts
remains both uncertain and controversial.” Id. at 1087–88.
Finally, as in National Parks, the “context” of this case—“a
                                28
place of extraordinary importance to the Tribes, a landscape of
profound cultural importance, and the water supply for the
Tribes and millions of others”—weighs in favor of requiring an
EIS. Appellees’ Br. 40–41. And in at least one sense, the case
for ordering production of an EIS is stronger here than in
National Parks or the cases on which Dakota Access relies,
Intervenor’s Br. 29–30, given that, unlike in those cases, the
district court has already given the Corps an opportunity to
resolve the Tribes’ serious criticisms and it failed to do so.

       The Corps and Dakota Access next argue that, even if the
district court properly ordered the Corps to prepare an EIS, the
court abused its discretion by vacating the pipeline’s easement
in the interim. “The ordinary practice,” however, “is to vacate
unlawful agency action,” United Steel v. Mine Safety & Health
Administration, 925 F.3d 1279, 1287 (D.C. Cir. 2019) (citing 5
U.S.C. § 706(2)), and district courts in this circuit routinely
vacate agency actions taken in violation of NEPA. See, e.g.,
Humane Society of the United States v. Johanns, 520 F. Supp.
2d 8, 37 (D.D.C. 2007) (observing that vacatur is the “standard
remedy” for an “action promulgated in violation of NEPA”);
Greater Yellowstone Coalition v. Bosworth, 209 F. Supp. 2d
156, 163 (D.D.C. 2002) (“[P]laintiffs . . . seek a vacatur of the
permit . . . until the [agency] complies with NEPA. As a
general matter, an agency action that violates the APA must be
set aside. . . . Based on this authority, I shall vacate the permit
. . . .”).

     “While unsupported agency action normally warrants
vacatur, [a] court is not without discretion” to leave agency
action in place while the decision is remanded for further
explanation. Advocates for Highway and Auto Safety v.
Federal Motor Carrier Safety Administration, 429 F.3d 1136,
1151 (D.C. Cir. 2005) (citation omitted). In Allied-Signal,
Inc. v. U.S. Nuclear Regulatory Commission, 988 F.2d 146
                               29
(D.C. Cir. 1993), our court set forth the two factors governing
that exercise of discretion: “The decision whether to vacate
depends on the seriousness of the order’s deficiencies (and thus
the extent of doubt whether the agency chose correctly) and the
disruptive consequences of an interim change that may itself be
changed.” Id. at 150–51 (internal quotation marks omitted).
The “seriousness” of a deficiency, we have explained, is
determined at least in part by whether there is “a significant
possibility that the [agency] may find an adequate explanation
for its actions” on remand. Williston Basin Interstate Pipeline
Co. v. FERC, 519 F.3d 497, 504 (D.C. Cir. 2008). “We review
the district court’s decision to vacate . . . for abuse of
discretion.” Nebraska Department of Health & Human
Services v. Department of Health & Human Services, 435 F.3d
326, 330 (D.C. Cir. 2006).

     As to the first factor, the district court concluded that the
Corps was unlikely to resolve the controversies on remand
because the court had previously remanded without vacatur for
just that purpose and the Corps had nonetheless failed to
resolve them. Standing Rock VII, 471 F. Supp. 3d at 79–80. The
court also explained that the Corps focused on the wrong
question: whether, on remand, it would be able to justify its
easement decision rather than its decision to forgo an EIS. Id.
at 81. (“Looking at the first Allied-Signal factor, the Court does
not assess the deficiency of the ultimate decision itself—the
choice to issue the permit—but rather the deficiency of the
determination that an EIS was not warranted.” (internal
quotation marks omitted)).

     With respect to the disruptive consequences of vacatur, the
district court understood that shutting down pipeline operations
would cause Dakota Access and other entities significant
economic harm. But for four reasons it concluded that those
effects did not justify remanding without vacatur. First, the
                              30
Corps’s expedited timeline for preparing an EIS “would cabin
the economic disruption of a shutdown.” Id. at 84. Second,
though economic disruption is properly considered, it is not
commonly a basis, standing alone, for declining to vacate
agency action. Id. at 84–85. Third, Dakota Access’s approach
would subvert NEPA’s objectives. “[I]f you can build first and
consider environmental consequences later, NEPA’s action-
forcing purpose loses its bite.” Id. at 85. And finally, the
countervailing risk of a spill—difficult to quantify in part
because of the Corps’s failure to prepare an EIS—counseled in
favor of vacatur. Id. at 85–86. The district court discounted as
“inconclusive” Dakota Access’s evidence that if DAPL were
inoperative, more oil would be transported by rail, a riskier
alternative. Id. at 87.

     On appeal, Dakota Access takes primary responsibility for
arguing against vacatur. It contends first that the Corps can
“easily substantiate its easement decision on remand even if it
must prepare an EIS.” Intervenor’s Br. 33. But that is not the
question. As the district court explained, the question is
whether the Corps is likely to justify its issuance of a FONSI
and refusal to prepare an EIS. Dakota Access argues that
Heartland Regional Medical Center v. Sebelius, 566 F.3d 193
(D.C. Cir. 2009), supports its contrary view that the Allied-
Signal factors look to whether an agency can justify the action
the court is considering whether to vacate, rather than the
challenged procedural decision. There, we sought to determine
whether an earlier district court decision had, by declaring a
regulatory requirement invalid for failing to consider certain
public comments, necessarily vacated the regulation. In
making that determination, we concluded that the Allied-Signal
factors would have directed remand without vacatur. Id. at
197–98. But because the agency had not elected to forgo a
procedural requirement (in that case, notice and comment),
only one agency action—the decision to promulgate the
                               31
challenged rule—was implicated at all. Heartland Regional
therefore says nothing one way or the other about the proper
focus of the Allied-Signal inquiry in cases, like this one, where
we confront a distinct challenge to an agency’s decision to
forgo a major procedural step in its path to its ultimate action.
Cf. id. at 199 (“Failure to provide the required notice and to
invite public comment—in contrast to the agency’s failure here
adequately to explain why it chose one approach rather than
another for one aspect of an otherwise permissible rule—is a
fundamental flaw that normally requires vacatur of the rule.”
(internal quotation marks omitted)). Besides, the district
court’s view is more sensible.

     Consider the consequences of Dakota Access’s contrary
approach. If, when an agency declined to prepare an EIS before
approving a project, courts considered only whether the agency
was likely to ultimately justify the approval, it would subvert
NEPA’s purpose by giving substantial ammunition to agencies
seeking to build first and conduct comprehensive reviews later.
If an agency were reasonably confident that its EIS would
ultimately counsel in favor of approval, there would be little
reason to bear the economic consequences of additional delay.
For similar reasons, an agency that bypassed required notice
and comment rulemaking obviously could not ordinarily keep
in place a regulation while it completed that fundamental
procedural prerequisite. See Daimler Trucks North America
LLC v. EPA, 737 F.3d 95, 103 (D.C. Cir. 2013) (“[T]he court
typically vacates rules when an agency ‘entirely fail[s]’ to
provide notice and comment . . . .” (quoting Shell Oil Co. v.
EPA, 950 F.2d 741, 752 (D.C. Cir. 1991))). When an agency
bypasses a fundamental procedural step, the vacatur inquiry
asks not whether the ultimate action could be justified, but
whether the agency could, with further explanation, justify its
decision to skip that procedural step. Otherwise, our cases
                               32
explaining that vacatur is the default response to a fundamental
procedural failure would make little sense.

     Even were we to consider the Corps’s odds of ultimately
approving the easement, our case law still instructs that a
failure to prepare a required EIS should lead us to doubt that
the ultimate action will be approved. In Oglala Sioux Tribe v.
U.S. Nuclear Regulatory Commission, 896 F.3d 520 (D.C. Cir.
2018), we explained that because NEPA is a “purely procedural
statute,” where an agency’s NEPA review suffers from “a
significant deficiency,” refusing to vacate the corresponding
agency action would “vitiate” the statute. Id. at 536 (internal
quotation marks omitted). As we made clear, “[p]art of the
harm NEPA attempts to prevent in requiring an EIS is that,
without one, there may be little if any information about
prospective environmental harms and potential mitigating
measures.” Id. (internal quotation marks omitted). Put another
way, Oglala strongly suggests that where an EIS was required
but not prepared, courts should harbor substantial doubt that
“‘the agency chose correctly’” regarding the substantive action
at issue—in this case, granting the easement. Id. at 538
(quoting Allied-Signal, 988 F.2d at 150–51). The Corps resists
the proposition that Oglala cautions against applying Allied-
Signal in NEPA cases, but that is not the point. The point is that
Oglala’s application of those factors suggests that NEPA
violations are serious notwithstanding an agency’s argument
that it might ultimately be able to justify the challenged action.

    As for vacatur’s consequences, Dakota Access contends
that while the district court “acknowledged the severe
economic disruption that vacatur would cause,” it “wrongly
discounted those severe consequences” and “credit[ed] remote,
unsubstantiated harms.” Intervenor’s Br. 35. But in reviewing
for abuse of discretion, we “consider whether the decision
maker failed to consider a relevant factor, whether he [or she]
                                33
relied on an improper factor, and whether the reasons given
reasonably support the conclusion.” Kickapoo Tribe of Indians
of Kickapoo Reservation in Kansas v. Babbitt, 43 F.3d 1491,
1497 (D.C. Cir. 1995) (alteration in original) (internal
quotation marks omitted). In doing so, we may not “substitute
our judgment for that of the trial court, so we cannot decide the
issue by determining whether we would have reached the same
conclusion.” United States v. Mathis–Gardner, 783 F.3d 1286,
1288 (D.C. Cir. 2015) (citation omitted) (internal quotation
marks omitted). Dakota Access believes that the district court’s
assessment of a shutdown’s economic impacts was far too rosy
and that the court “ignored” a shutdown’s environmental
consequences. But the court considered all important aspects
of the issue and reasonably concluded that the harms were less
severe than the Corps and Dakota Access suggested. In view of
the discretion owed the district court and the seriousness of the
NEPA violation, Dakota Access has given us no basis for
concluding that the district court abused its discretion in
applying the Allied-Signal factors. See National Parks
Conservation Association v. Semonite, 925 F.3d 500, 502
(D.C. Cir. 2019) (“[The district] court is best positioned to . . .
make factual findings[] and determine the remedies necessary
to protect the purpose and integrity of the EIS process.”); Stand
Up for California! v. U.S. Department of Interior, 879 F.3d
1177, 1190 (D.C. Cir. 2018) (“[T]he district court acted well
within its discretion in finding vacatur unnecessary to address
any harm the defect had caused.”).

     In any event, Dakota Access’s assessment of vacatur’s
consequences is undercut significantly by the fact that we agree
that the district court’s shutdown order cannot stand.

     On August 5, 2020, a motions panel of this court ordered
that “to the extent the district court issued an injunction by
ordering Dakota Access LLC to shut down the Dakota Access
                                34
Pipeline and empty it of oil by August 5, 2020, the injunction
be stayed.” August 5 Order at 1. Relying on the Supreme
Court’s decision in Monsanto Co. v. Geertson Seed Farms, 561
U.S. 139 (2010), the panel explained that “[t]he district court
did not make the findings necessary for injunctive relief.”
August 5 Order at 1 (“[B]efore issuing an injunction in a
[NEPA] case, ‘a court must determine that an injunction should
issue under the traditional four-factor test.’” (quoting
Monsanto, 561 U.S. at 158)).

     The Tribes argue that an injunction was unnecessary
because vacatur itself “invalidat[ed] the underlying easement,”
thus requiring the “suspension of pipeline operations pending
compliance with NEPA.” Appellees’ Br. 73–74. That is the
view the district court appeared to adopt, Standing Rock VII,
471 F. Supp. 3d at 88 (requiring, after vacating the pipeline’s
easement, “the oil to stop flowing and the pipeline to be
emptied within 30 days”), and that approach finds some
support in our case law. For instance, in Sierra Club v. FERC,
867 F.3d 1357, 1379 (D.C. Cir. 2017), we vacated a pipeline
authorization due to a NEPA violation and appeared to assume
that vacatur encompassed an end to construction. Likewise in
National Parks, we appeared to accept the parties’ assumption
that vacating Corps-issued construction permits would require
ceasing construction of the challenged electrical towers or
tearing them down. See National Parks, 925 F.3d at 502.

     The Tribes’ approach, however, cannot be squared with
Monsanto, which should caution against reading too far into
our tacit approval of shutdown orders in prior cases. If a district
court could, in every case, effectively enjoin agency action
simply by recharacterizing its injunction as a necessary
consequence of vacatur, that would circumvent the Supreme
Court’s instruction in Monsanto that “a court must determine
that an injunction should issue under the traditional four-factor
                               35
test.” 561 U.S. at 158. In fact, the Tribes have already moved
for a permanent injunction in the district court during the
pendency of this appeal, and that motion is fully briefed.

     Furthermore, Sierra Club and National Parks differ from
this case in a subtle but important way. Those cases involved
challenges to agency authorizations of the very activities the
court assumed would end. Vacating a construction permit in
National Parks, for instance, naturally implied an end to
construction. Here, in contrast, we affirm the vacatur of an
easement authorizing the pipeline to cross federal lands. With
or without oil flowing, the pipeline will remain an
encroachment, leaving the precise consequences of vacatur
uncertain. In fact, the parties have identified no other
instance—and we have found none—in which the sole issue
before a court was whether an easement already in use (rather
than a construction or operating permit) must be vacated on
NEPA grounds. That makes this case quite unusual and cabins
our decision to the facts before us.

     It may well be—though we have no occasion to consider
the matter here—that the law or the Corps’s regulations oblige
the Corps to vindicate its property rights by requiring the
pipeline to cease operation and that the Tribes or others could
seek judicial relief under the APA should the Corps fail to do
so. But how and on what terms the Corps will enforce its
property rights is, absent a properly issued injunction, a matter
for the Corps to consider in the first instance, though we would
expect it to decide promptly. To do otherwise would be to issue
a de facto outgrant without engaging in the NEPA analysis that
the Corps concedes such an action requires. See Oral Arg. Tr.
36:14–15 (“The Corps’[s] regulations contemplate that an
outgrant would require a NEPA analysis.”). Although the
district court was attuned to the discretion owed the Corps, see
Standing Rock VII, 471 F. Supp. 3d at 88 (“Not wishing to
                              36
micromanage the shutdown, [the court] will not prescribe the
method by which DAPL must [make the flow of oil cease].”),
we nonetheless conclude that it could not order the pipeline to
be shut down without, as required by Monsanto, making the
findings necessary for injunctive relief.

                             IV.
     For the foregoing reasons, we affirm the district court’s
order vacating DAPL’s easement and directing the Corps to
prepare an EIS. We reverse to the extent the court’s order
directs that the pipeline be shut down and emptied of oil.

                                                   So ordered.