Volume 1 of 2
                  FOR PUBLICATION
  UNITED STATES COURT OF APPEALS
       FOR THE NINTH CIRCUIT

NATURAL RESOURCES DEFENSE               
COUNCIL, INC.; THE INTERNATIONAL
FUND FOR ANIMAL WELFARE;
CETACEAN SOCIETY INTERNATIONAL;
LEAGUE FOR COASTAL PROTECTION;
OCEAN FUTURES SOCIETY; JEAN-
MICHEL COUSTEAU,
                Plaintiffs-Appellees,
CALIFORNIA COASTAL COMMISSION,
                Intervenor-Appellee,
                  v.                          No. 08-55054
DONALD C. WINTER, Secretary of
the Navy; UNITED STATES                        D.C. No.
                                            CV-07-00335-FMC
DEPARTMENT OF THE NAVY; CARLOS
M. GUTIERREZ, Secretary of the                 OPINION
Department of Commerce;
NATIONAL MARINE FISHERIES
SERVICES; WILLIAM HOGARTH,
Assistant Administrator for
Fisheries of the National
Oceanographic and Atmospheric
Administration; CONRAD C.
LAUTENBACHER, JR., Administrator
of the National Oceanographic and
Atmospheric Administration,
             Defendants-Appellants.
                                        
        Appeal from the United States District Court
            for the Central District of California
      Florence Marie Cooper, District Judge, Presiding

                             2079
2080                   NRDC v. WINTER
                    Argued and Submitted
           February 27, 2008—Pasadena, California

                   Filed February 29, 2008

       Before: Betty B. Fletcher, Dorothy W. Nelson, and
               Stephen Reinhardt, Circuit Judges.

                 Opinion by Judge B. Fletcher
2084                  NRDC v. WINTER


                        COUNSEL

Ronald J. Tenpas (argued), Acting Assistant Attorney Gen-
eral; Michael R. Eitel, Luther L. Hajek, and Allen M. Bra-
bender, Appellate Section, U.S. Department of Justice,
Environment & Natural Resources Division, Washington,
D.C.; Craig D. Jensen and J. Page Turney, Office of General
Counsel, Department of the Navy, for the federal defendants-
appellants.

Joel R. Reynolds, Cara A. Horowitz, Stephen Zak Smith, Nat-
ural Resources Defense Council, Inc., Santa Monica, Califor-
nia; Richard B. Kendall (argued), Gregory A. Fayer, and Josh
B. Gordon, Irell and Manella LLP, Los Angeles, California,
for the plaintiffs-appellees.

Edmund G. Brown, Jr., Janet Gaard, J. Matthew Rodriguez,
Jamee Jordan Patterson (argued), Office of the Attorney Gen-
                            NRDC v. WINTER                             2085
eral of California, for intervenor-appellee California Coastal
Commission.


                               OPINION

B. FLETCHER, Circuit Judge:

   Defendants Secretary of the Navy, Department of the Navy,
Secretary of the Department of Commerce, National Marine
Fisheries Service (NMFS),1 and two Administrators of the
National Oceanographic and Atmospheric Administration
(NOAA) appeal the district court’s January 3, 2008 order, as
modified on January 10, 2008, granting a motion for a prelim-
inary injunction and imposing certain conditions on the com-
pletion of the remaining eight of fourteen large training
exercises scheduled to be conducted by the Navy’s Third
Fleet in the waters off the coast of southern California
between February 2007 and January 2009 (the “SOCAL exer-
cises”).2 The motion was filed by plaintiffs Natural Resources
Defense Council, Inc., International Fund for Animal Wel-
fare, Cetacean Society International, League for Coastal Pro-
tection, Ocean Futures Society, and Jean-Michel Cousteau
(collectively “NRDC” or “plaintiffs”), who are concerned that
the Navy’s use of high-intensity, mid-frequency active sonar
(“MFA sonar”) in the SOCAL exercises will cause serious
harm to various species of marine mammal present in the
southern California waters, and by extension, to plaintiffs
themselves.
  1
     The National Marine Fisheries Service has now been renamed “NOAA
Fisheries.” Because many of the documents refer to the agency by its for-
mer name, it shall be referred to as “the NMFS” throughout this opinion.
   2
     While the district court imposed the injunction when the Navy had yet
to conduct nine training exercises, it issued a temporary partial stay of its
injunction on January 17, 2008, after which the Navy conducted its sixth
exercise. Accordingly, only eight of the fourteen scheduled exercises have
not yet been conducted.
2086                   NRDC v. WINTER
   In granting NRDC’s motion for a preliminary injunction,
the district court found that NRDC had demonstrated probable
success on the merits of its claim that the Navy violated the
National Environmental Policy Act (“NEPA”), 42 U.S.C.
§ 4321 et seq., by failing to prepare an Environmental Impact
Statement (“EIS”). The district court also found that NRDC
had demonstrated probable success on the merits of its claim
that the Navy violated the Coastal Zone Management Act
(“CZMA”), 16 U.S.C. § 1451 et seq., by submitting a consis-
tency determination to the California Coastal Commission
(“CCC”) that did not take into account the planned use of
MFA sonar and by failing to adopt the mitigation measures
the CCC determined were necessary for the SOCAL exercises
to be consistent with the California Coastal Management Pro-
gram (“CCMP”).

   On January 15, 2008, the Council on Environmental Qual-
ity (“CEQ”) purported to approve “alternative arrangements,”
pursuant to 40 C.F.R. § 1506.11, that would permit the Navy
to continue its exercise without first completing an EIS. On
the same day, President George W. Bush, pursuant to 16
U.S.C. § 1456(c)(1)(B), exempted from the requirements of
the CZMA the Navy’s use of MFA sonar in the SOCAL exer-
cises.

   On February 4, 2008, the district court upheld its injunction
on the basis of plaintiffs’ NEPA claim, concluding CEQ’s
action was invalid and therefore not entitled to deference. The
district court also expressed concerns about the constitutional-
ity of the President’s CZMA exemption on the ground that it
appeared to amount to an executive revision of a judicial deci-
sion and thus violated the principle, recognized in Hayburn’s
Case, 2 U.S. (2 Dall.) 408 (1792), that Congress cannot vest
review of the decisions of Article III courts in officials of the
Executive Branch. However, the court declined to decide the
constitutionality of the CZMA exemption because it con-
cluded the preliminary injunction was firmly supported on
                           NRDC v. WINTER                             2087
NEPA grounds.3 The district court also found that plaintiffs
had demonstrated a possibility of irreparable harm and that
the balance of hardships tipped in plaintiffs’ favor. Natural
Res. Def. Council v. Winter, ___ F.Supp. 2d ___, 

(C.D. Cal. Feb. 4, 2008) (“Feb. 4, 2008 Dist. Ct.
Order”).

  For the reasons stated below, we uphold the district court’s
preliminary injunction.

                      I.   Procedural History

   Plaintiffs commenced this action on March 22, 2007. On
August 7, 2007, the district court granted in part NRDC’s
motion for a preliminary injunction and enjoined the Navy
from conducting the then remaining eleven SOCAL exercises.4
After appealing the district court’s preliminary injunction
order,5 the Navy filed an emergency motion with this court for
a stay of the order while its appeal was pending. On August
31, 2007, a divided motions panel granted the Navy’s motion
on the grounds that the district court had failed to consider the
“public interest” in having a trained and effective Navy and
had failed to explain why an unconditional injunction on the
SOCAL exercises, rather than an injunction conditioning the
conduct of those exercises on the adoption of additional miti-
gation measures, was appropriate. NRDC v. Winter, 

(9th Cir. 2007).
  3
     Because the district court did not rule on the likelihood of success of
plaintiffs’ CZMA claim in light of the President’s exemption, we decline
to reach that issue.
   4
     The district court granted NRDC’s motion only in part because it found
that NRDC had not demonstrated probable success on the merits of its
additional claim that the NMFS and the other non-Navy defendants had
failed to prepare an adequate Biological Opinion and Incidental Take
Statement in violation of the Endangered Species Act and the Administra-
tive Procedure Act. NRDC has not appealed that portion of the district
court’s ruling.
   5
     The CCC has intervened as appellee in this action.
2088                   NRDC v. WINTER
   On November 13, 2007, after hearing oral argument, we
filed an order concluding that NRDC had met the necessary
burden of proof to demonstrate that some form of preliminary
injunctive relief was appropriate. Accordingly, we vacated the
stay of the preliminary injunction order effective upon the
Navy’s completion of its fifth SOCAL exercise, which was in
progress at the time of oral argument. However, we also con-
cluded that an injunction conditioning continuation of the
exercises on the Navy’s adoption of narrowly tailored mitiga-
tion measures would be more appropriate than a total injunc-
tion. Accordingly, we remanded the case for the district court
to enter a modified preliminary injunction containing appro-
priate mitigating measures. NRDC v. Winter, 

(9th Cir. 2007).

   On January 3, 2008, the district court, after having received
briefing from the parties and having toured the destroyer USS
Milius at the naval base in San Diego, California, to improve
its understanding of the Navy’s sonar training procedures and
the feasibility of the parties’ proposed mitigation measures,
issued a new preliminary injunction that allowed the Navy to
conduct the remaining SOCAL exercises provided that it
employ certain measures intended to mitigate the impact of
the Navy’s use of MFA sonar on the environment. On January
9, 2008, the Navy applied for a stay pending appeal and
requested relief from the district court by January 14, 2008.

   On January 10, 2008, in response to arguments raised in the
Navy’s stay application, the district court modified the pre-
liminary injunction by narrowing the mitigation measures
contained in the January 3, 2008 order. The Navy filed a
notice of appeal the following day. The district court denied
the Navy’s stay application on January 14, 2008.

  On the evening of January 15, 2008, the Navy filed an
emergency motion with this court requesting vacatur of the
preliminary injunction or, alternatively, a partial stay of the
preliminary injunction pending a decision on its appeal by our
                       NRDC v. WINTER                      2089
court. The Navy’s motion was based in part on two develop-
ments that occurred on the same day that the motion was
filed. First, the President of the United States, pursuant to 16
U.S.C. § 1456(c)(1)(B), exempted from the provisions of the
CZMA the Navy’s use of MFA sonar during the SOCAL
exercises, finding that such use of MFA sonar is “essential to
national security” and in the “paramount interest of the United
States.” Second, the CEQ, finding “emergency circum-
stances,” purported to approve “alternative arrangements” to
accommodate those emergency circumstances, pursuant to 40
C.F.R. § 1506.11. It permitted the Navy to follow the pre-
scribed arrangements to continue its exercises pending com-
pletion of the Navy’s EIS. The Navy subsequently adopted
the alternative arrangements and determined that it would
comply with them. See Decision Memorandum Accepting
Alternative Arrangements for the U.S. Navy’s Southern Cali-
fornia Operating Area Composite Training Unit Exercises and
Joint Task Force Exercises Scheduled To Occur Between
Today and January 2009, 73 Fed. Reg. 4189 (Jan. 24, 2008).

   On January 16, 2008, we remanded the matter to the district
court to consider in the first instance the effect, if any, of
these developments on its preliminary injunction order. On
January 17, 2008, the district court issued a temporary partial
stay of its preliminary injunction order pending the court’s
consideration of the Navy’s ex parte application to vacate the
preliminary injunction. The Navy subsequently conducted its
sixth SOCAL exercise.

   On February 4, 2008, following briefing by the parties and
oral argument, the district court denied the Navy’s application
to vacate the preliminary injunction and lifted the temporary
partial stay. In its published order, the district court held in
relevant part that CEQ’s approval of “alternative arrange-
ments” was invalid because there are no “emergency circum-
stances” within the meaning of 40 C.F.R. § 1506.11. Feb. 4,
2008 Dist. Ct. Order at 13-25. Thus, the district court left in
2090                        NRDC v. WINTER
place the original preliminary injunction. The Navy filed a
notice of appeal two days later.

   On February 8, 2008, we issued a sua sponte order expedit-
ing the appeal of the district court’s order imposing the pre-
liminary injunction. On February 15, 2008, the Navy filed an
emergency motion for a partial stay of the preliminary injunc-
tion pending our consideration of the appeal on the ground
that the injunction would interfere with the Navy’s conduct of
two exercises in March, 2008. On February 19, 2008, we
denied the Navy’s motion for a partial stay in light of our
order expediting the appeal, which set oral argument for
Wednesday, February 27, 2008. We now affirm the district
court’s order imposing the preliminary injunction.

                      II.   Factual Background

      A.   The SOCAL Exercises and the Effect of MFA
                 Sonar on Marine Mammals

   The scheduled SOCAL exercises consist of seven Compos-
ite Training Unit Exercises (“COMPTUEX”), which last three
to four weeks each, and seven Joint Tactical Force Exercises
(“JTFEX”), which last approximately ten days each. The
exercises, which involve the use of multiple surface ships, air-
craft and submarines, are part of the “integrated” training
phase of the Navy’s Fleet Response Training Plan, in which
individual naval units — ships, submarines and aviation
squadrons — learn and demonstrate skills as members of a
strike group.6 Declaration of Captain Martin N. May (“May
  6
   The Navy employs two types of strike groups. A carrier strike group
generally consists of an aircraft carrier and five surface combatant ships.
May decl. ¶ 5. An expeditionary strike group includes surface combatant
ships and an amphibious ready group. 

group starts developing skill sets at the individual ship, subma-
rine, or aviation squadron level by conducting unit-level training. 

As skill levels increase, units coordinate training with other units. 

v. WINTER                             2091
decl.) ¶ 6. In a COMPTUEX exercise a strike group must
demonstrate that it is capable of operating in a “complicated
threat-based scenario environment that simulates real world
situations.” 

JTFEX exercise, which follows the
COMPTUEX exercise and typically includes other Depart-
ment of Defense services and Allied Forces, the focus is on
“mission planning and strategy and on the orchestration of
integrated maneuvers, communication and coordination.” 

Upon completion of the integrated training phase the
Fleet Commander is able to certify that a strike group is ready
for deployment. 

the Navy, the ability to execute anti-
submarine warfare (“ASW”) is critical to a Commander’s cer-
tification of a strike group. 

is the Pacific
Fleet’s top “war-fighting” priority because of the proliferation
of extremely quiet diesel electric submarines throughout the
world.7 

Dec. 14, 2007 Declaration of Rear Admiral
John M. Bird (“Bird. Decl”) ¶ 16. In turn, an important part
of ASW is the use of active sonar, a technology which the
Navy deems absolutely necessary to detect today’s extremely
quiet submarines. The type of active sonar, the use of which

to commencing the “integrated” phase of the Navy’s training plan, the
individual units comprising a strike group must be trained and attain profi-
ciency in the “basic” phase of the training plan. 

During the “inte-
grated” training phase, an individual unit learns and demonstrates skills as
a team member of the strike group. Following completion of the “integrat-
ed” phase, strike groups enter the “sustainment” phase of the training plan.

continues through the strike group’s scheduled deployment
and post-deployment periods and ends with the commencement of the
“maintenance” phase, during which the ships comprising the strike group
undergo maintenance and modernization. 

25.
   7
     We note that neither quiet submarines, nor the use of active sonar to
detect them, are new technologies. According to the Navy’s Environmen-
tal Assessment (“EA”), active sonar was used effectively against German
U-boats during World War II, and during the Cold War technological
developments in active sonar were “critical” for tracking sophisticated
Soviet submarines.
2092                       NRDC v. WINTER
NRDC challenges, is mid frequency active sonar; other cate-
gories of active sonar are low-frequency active sonar and
high-frequency active sonar.8

   Active sonar involves a vessel or other sonar source emit-
ting a loud noise underwater and then listening for whether
the noise comes back to the source, indicating that the noise
may have bounced off the hull of a previously undetected sub-
marine. According to the Navy, active sonar has two impor-
tant advantages over passive sonar, which merely involves
listening for noise made by submarines themselves: active
sonar gives both the bearing and the distance of the target
submarine, while passive sonar gives only the bearing;9 and
active sonar allows the Navy to target submarines that emit
sound at levels below those of the surrounding marine envi-
ronment. Bird decl. ¶ 9. Accordingly, the Navy has concluded
that in certain environments, including shallow coastal waters
where ambient noise levels are high, MFA sonar allows better
detection of quiet submarines than passive sonar. May decl.
¶¶ 9, 14.

   According to the Navy, personnel using MFA sonar must
train with it regularly, under realistic conditions, and in a vari-
ety of situations. May decl. ¶ 10. The Navy therefore trains
with MFA sonar in the ASW exercises that constitute an
important component of the SOCAL exercises.

   The SOCAL exercises are conducted in the Navy’s training
ranges off the coast of southern California (“the Southern Cal-
ifornia Operating Area”). This area is located in biologically
  8
     According to the Navy, low-frequency active sonar is transmitted at
frequencies between 0.1 kHz and 1.0 kHz; mid-frequency active sonar at
frequencies between 1.0 kHz and 10.0 kHz; and high-frequency active
sonar at frequencies greater than 10.0 kHz. Bird decl. ¶ 12; May decl. ¶ 9.
   9
     Active sonar allows a calculation of the distance of a target submarine
by considering the speed at which the sonar sound moves through water
and the time it takes for emitted sonar sound to travel to the target and
back. Bird decl. ¶ 12.
                          NRDC v. WINTER                           2093
diverse waters. At least thirty-seven species of marine mam-
mals are found there, with the most common being various
species of dolphin and whale, as well as the California sea
lion. Nine of those species are listed as threatened or endan-
gered under the Endangered Species Act (“ESA”), 16 U.S.C.
§ 1531 et seq.: the blue whale, fin whale, humpback whale,
Northern Pacific right whale, sei whale, sperm whale, sea
otter, Stellar sea lion, and Guadalupe fur seal. In addition, up
to eight species of beaked whale are found in the Southern
California Operating Area. A study submitted by NRDC clas-
sifies the California coastal waters as a “key area” for beaked
whales because over 25% of all beaked whale species are
found there.10

   The Navy acknowledges in its EA that MFA sonar may
affect both the physiology and behavior of marine mammals.
Exposure to “very high” acoustic energy levels may impair
the functioning of marine mammals’ visual system, vestibular
system and internal organs, and may cause injury to their
lungs and intestines. However, the primary physiological
effects of MFA sonar are on marine mammals’ auditory sys-
tem: very high sound levels may rupture the eardrum or dam-
age small bones in the middle ear, but even exposure to lower
levels of sound may cause permanent or temporary hearing
loss.

   Several studies suggest that active sonar may also cause a
form of decompression sickness (or the “bends”) in marine
mammals by inducing growth of gas bubbles in their blood
stream or tissues, potentially leading to fatal hemorrhaging,
lesions and emboli in the organs. However, the Navy disputes
the conclusions of these studies and it has submitted a decla-
  10
    According to the study, the “key area” in southern California com-
prises the California shelf margins, which the study defines as the area
west of the Californian coast up until the 125.0 degree longitude. Navy
maps in the record show that the Southern California Operating Area falls
largely within this area.
2094                      NRDC v. WINTER
ration by an expert on marine mammal auditory systems stat-
ing that decompression sickness requires super-saturation of
tissue with gas and that such super-saturation has not been
shown to occur in marine mammals. See Declaration of Dr.
Darlene R. Ketten (“Ketten decl.”) ¶¶ 12-16.

   The Navy also acknowledges that the use of MFA sonar
may overtly disrupt the normal behavior of marine mammals
even if it does not affect their physiology. While the Navy
acknowledges that active sonar may cause behavioral
responses such as attempting to avoid the site of sound expo-
sure, swimming erratically, sluggish behavior, tail slapping,
“jaw popping,” and aggressive behavior, those responses were
observed in studies using trained animals held in captivity.11
NOAA concluded in 2006 that studies of marine mammals in
the wild “strongly suggest” that the use of sonar at levels
lower than those found to produce behavioral effects in the
tests of captive animals can result in “profound” behavioral
alterations, including changes in feeding, diving, and social
behavior. In a February 9, 2007 Biological Opinion concern-
ing the SOCAL exercises,12 the NMFS found that acoustic
exposures can impair marine mammals’ foraging ability and
their ability to detect predators or communicate. The NMFS
cited studies finding that noise has caused whales to move
away from their feeding and mating grounds and migration
routes, and to change their calls.
  11
      These behavioral responses were observed in a study to which the EA
cites for the effects of MFA sonar on marine mammals. See J.J. Finneran
and C.E. Schlundt, Effects of Intense Pure Tones on the Behavior of
Trained Odontocetes, Space and Naval Warfare Systems Center, San
Diego, Technical Document (September 2004).
   12
      The NMFS issued the Biological Opinion pursuant to section 7(a)(2)
of the Endangered Species Act (“ESA”), 16 U.S.C. § 1536(a)(2), which
concluded that while the SOCAL exercises might “adversely affect” cer-
tain threatened and endangered species, the exercises were not “likely to
jeopardize the [species’] continued existence.” The NMFS also issued an
Incidental Take Statement under which harm done to animals of the
threatened or endangered species would be excused under the ESA as inci-
dental.
                           NRDC v. WINTER                             2095
   As the record demonstrates, substantial evidence suggests
that beaked whales are particularly vulnerable to MFA sonar.
While it is not settled what causes this vulnerability,13 it is
clear that use of MFA sonar may lead to the stranding of
beaked whales. A 2004 Navy-sponsored study concluded that
“the evidence of sonar causation is . . . completely convincing
and that therefore there is a serious issue of how best to avoid/
minimize future beaching events.” Likewise, the Standing
Working Group on Environmental Concerns of the Interna-
tional Whaling Commission’s Scientific Committee con-
cluded in 2004 that “[t]he weight of accumulated evidence
now associates mid-frequency, military sonar with atypical
beaked whale mass strandings,” and found that “[t]his evi-
dence is very convincing and appears overwhelming.”

   A 2006 study cited as possible explanations for the associa-
tion between MFA sonar and strandings of beaked whales that
(1) beaked whales may swim into shallow waters to avoid the
sonar sound and strand if they are unable to navigate back to
deeper waters, and (2) that behavioral responses to sonar may
lead to tissue damage that in turn leads to stranding. The study
explains that while a stranding need not be fatal, stranded
marine mammals have died from cardiovascular collapse due
to hyperthermia or from the stress associated with the strand-
ing. Several mass strandings of marine mammals—mostly,
though not exclusively, beaked whales—have been associated
with the use of active sonar. Another 2006 study describes a
stranding of twelve beaked whales in Greece in 1996, a
stranding of seventeen marine mammals (including fourteen
beaked whales) in the Bahamas in 2000, and a stranding of
  13
     A 2004 Navy-sponsored study investigated several possible explana-
tions for beaked whales’ vulnerability to MFA sonar, including that
beaked whales have a specialized anatomy, possibly due to their deep div-
ing, which renders them especially sensitive to sound, and that beaked
whales are “especially skittish” such that loud, reverberant acoustic fields
cause “uncontrolled attempts to escape.” The study concluded that the lat-
ter explanation was the most likely to be correct, but that this conclusion
could change “as more research is done on this problem.”
2096                        NRDC v. WINTER
fourteen beaked whales in the Canary Islands in 2002,14 all of
which occurred at the same time and place as the naval use
of MFA sonar. The study also recounts a stranding of three
beaked whales in the Madeira Islands in 2000, which coin-
cided with NATO’s conduct of naval exercises,15 as well as a
stranding of two beaked whales in the Gulf of California,
Mexico, in 2002, which coincided with the conduct of seismic
surveys involving, among other acoustic sources, a multi-
beam, high-frequency sonar. In addition, a 2006 report by the
NMFS recounts that 150 to 200 melon-head whales stranded
in a Hawaiian bay in 2004, at the same time and place as the
Navy’s use of active sonar as part of its biennial Rim of the
Pacific (RIMPAC) exercise.16

   Necropsies of the dead whales involved in the Bahamas,
Canary Islands and Madeira Islands strandings revealed hem-
orrhages in and around the ears, in the cranial spaces, and in
other parts of the body such as the jaw fat, lungs and kidneys.
In a joint report, the Navy and NOAA concluded that the inju-
ries to the whales that stranded in the Bahamas constituted
“some sort of acoustic or impulse trauma,” and that the
Navy’s use of MFA sonar was the “most plausible” source of
that trauma. The International Whaling Commission agreed
   14
      Notably, the study states that after the stranding in the Canary Islands,
local researchers examined past stranding records and found reports of
eight other strandings of beaked whales in the Canaries since 1985, at least
five of which coincided with naval activities offshore.
   15
      According to the study, NATO has been unwilling to provide informa-
tion on its sonar activity during the Madeira Island exercises.
   16
      The RIMPAC exercises have been conducted biennially since 1968 in
the “Hawaiian operating area” and are intended to enhance the communi-
cation and coordination between Pacific Rim armed forces “as a means of
promoting stability in the region to the benefit of all participating nations.”
Bird decl. ¶ 34. Unlike the JTFEX and COMPTUEX exercises, the RIM-
PAC exercises are not part of the “integrated” phase of the Fleet Response
Training Plan; they are focused on command and control among the
nations involved, and not focused on certifying strike groups for deploy-
ment. 

the JTFEX and COMPTUEX exercises, the RIM-
PAC exercises include ASW training that involves the use of MFA sonar.
                      NRDC v. WINTER                     2097
that the hemorrhages in the inner ears and cranial spaces were
consistent with “direct acoustic effects.”

   According to a biologist on whose declaration NRDC
relies, the use of MFA sonar in the Bahamas may also have
had a serious effect on the local population of beaked whales.
See Declaration of Dr. Hal Whitehead. The biologist cites a
study showing that no Cuvier’s beaked whales were sighted
for twenty months following the stranding in the Bahamas,
despite an increased survey effort. 

He also cited
studies showing that of the Cuvier’s beaked whales that had
been photo-identified over a nine-year period, only a few have
been sighted since the stranding in 2000. 

to the stranding of the melon-head whales in
Hawaii, the NMFS issued a report concluding that “[w]hile
causation of this stranding event may never be unequivocally
determined, we consider the active sonar transmissions . . . a
plausible, if not likely, contributing factor in what may have
been a confluence of events.”

   Undoubtedly, many training exercises involving the use of
active sonar occur around the world without marine mammal
strandings being observed or reported. However, a declaration
by a NOAA scientist submitted by the Navy acknowledges
that it is generally poorly understood in which combinations
of physical and biological circumstances such strandings are
likely to occur. Declaration of Brandon L. Southall ¶ 19. The
declaration also acknowledges that whether marine mammal
strandings are observed depends on the extent to which peo-
ple are looking for them. 

Navy’s EA and the Predicted Harm to Marine
      Mammals in the Southern California Waters

  In February 2007, the Navy issued an Environmental
Assessment (“EA”) for the SOCAL exercises pursuant to
NEPA. See 40 C.F.R. § 1501.3. The EA set forth the Navy’s
2098                   NRDC v. WINTER
estimate of how much harm the use of MFA sonar would
inflict on marine mammals, classifying the harm as either
“Level A harassment” or “Level B harassment.”

   Level A harassment is an act that physically injures the
marine mammal. Level A harassment refers to an exposure to
MFA sonar that “injures or has the significant potential to
injure a marine mammal or marine mammal stock in the
wild.” Injury is defined as any destruction or loss of any bio-
logical tissues, and includes permanent hearing loss.

   Level B harassment is an act that disrupts the behavior of
a marine mammal. Level B harassment refers to an exposure
to MFA sonar that “disturbs or is likely to disturb a marine
mammal or marine mammal stock by causing disruption of
natural behavioral patterns including, but not limited to,
migration, surfacing, nursing, feeding, or sheltering to a point
where such behaviors are abandoned or significantly altered.”
Notably, Level B harassment may also, though it need not,
include temporary hearing loss.

  The behavioral effects that result from Level B harassment
may have severe consequences. According to the Biological
Opinion of the NMFS, acoustic exposures can result in the
death of a marine mammal by impairing its foraging or its
ability to detect predators or communicate by increasing stress
or by disrupting important physiological events.

   In its EA, the Navy classified predicted sonar exposures as
either Level A harassments or Level B harassments based on
the sound intensity to which a marine mammal would be
exposed. For cetaceans—which include whales and dolphins
and which are the focus of NRDC’s challenge to the SOCAL
exercises—the Navy applied the following “impact thresh-
olds”: it classified as Level A harassments exposures to sonar
levels of 215 decibels (dB) or greater, as Level B harassments
including temporary hearing loss exposures to sonar levels
between 195 dB and 215 dB, and as Level B harassments not
                           NRDC v. WINTER                             2099
including temporary hearing loss exposures to sonar levels
between 173 dB and 195 dB.17

  Significantly, the Navy acknowledged in the EA that it
does not know whether the above impact thresholds apply to
beaked whales. Recognizing the recent beaked whale strand-
ings and the fact that the exact causes of those strandings are
unknown, the Navy concluded that “separate, meaningful
impact thresholds cannot be derived specifically for beaked
whales.” Put simply, the Navy did not know whether expo-
sure of a beaked whale to an acoustic energy of less than 215
dB might nevertheless cause permanent injury to the whale.
Accordingly, the Navy took a “conservative approach” and
counted all predicted Level B exposures of beaked whales as
non-lethal Level A exposures. Thus, the Navy treated every
predicted exposure of a beaked whale to a sonar level of 173
dB or greater as causing physical injury including permanent
hearing loss.

   In its EA, the Navy estimated that over the course of the
SOCAL exercises, the use of MFA sonar would result in 564
instances of Level A harassment to marine mammals, 548 of
which would be to beaked whales.18 Specifically, the follow-
  17
      The EA expressed the impact thresholds in terms of “energy flux den-
sity level,” which is a measure of the flow of sound energy through an
area, or, more formally, the time integral of the squared pressure divided
by the impedance. Energy flux density is expressed in units of decibels
referenced to the pressure and duration of the sound, e.g., 215 dB re 1 µPa2
-s. For convenience, we express the impact thresholds, and exposures to
MFA sonar generally, only in units of decibels.
   18
      The statements in the EA that “[t]he modeling efforts and harassment
analysis for mid-frequency active sonar estimate that no Level A harass-
ment” of beaked whales will occur is not to the contrary. As explained
above, the Navy acknowledged in the EA that it did not know whether the
impact thresholds it established for cetaceans generally also applied to
beaked whales. Accordingly, the Navy decided that although its harass-
ment model using those impact thresholds did not predict any Level A
harassments to beaked whales, it would nevertheless treat all predicted
Level B harassments to beaked whales as Level A harassments.
2100                       NRDC v. WINTER
ing species would be subjected to Level A harassments:
Cuvier’s beaked whales in 436 instances; Ziphiid beaked
whales in 104 instances; common dolphins in 16 instances;
and Baird’s beaked whales in 8 instances.19

   The Navy also estimated that the use of MFA sonar would
result in 8,160 exposures to Level B harassment with tempo-
rary hearing loss and 161,368 exposures to Level B harass-
ment without hearing loss. Eight marine mammal species,
including one endangered species, would be exposed to over
1,000 incidents of Level B harassment: 145,444 exposures to
common dolphins; 6,460 exposures to Northern Pacific right
whale dolphins; 4,292 exposures to Risso’s dolphins; 4,100
exposures to Pacific white-sided dolphins, 3,252 exposures to
striped dolphins; 1,830 exposures to pygmy sperm whales
(endangered); 1,094 exposures to Pantropical spotted dol-
phins; and 1,092 exposures to bottlenose dolphins.

   In light of the harm that marine mammals are expected to
suffer as a result of the SOCAL exercises, plaintiffs contend
that they and their members living in southern California will
be harmed. For example, plaintiff Jean-Michel Cousteau
alleges that as an environmental enthusiast and film-maker his
ability to enjoy and educate others about the marine environ-
ment in southern California will be impaired if the harmful
effects of MFA sonar on marine mammals are not sufficiently
mitigated. Other plaintiffs make similar allegations.

   The Navy stated in the EA that it “assumed” that its meth-
odology for estimating harm overestimated the effects of
MFA sonar on marine mammals, citing the lack of observed
effects during several past major exercises. However, the EA
also maintained that the methodology used was based on the
  19
    These figures appear in the EA’s Appendix A; the slightly different
figures cited by the parties and the district court appear to be incorrect
because they fail to account for 98 instances of Level A harassment to
Ziphiid whales.
                           NRDC v. WINTER                           2101
“best available science,” and it provided no indication of the
extent to which its methodology overestimated the effects of
MFA sonar. In fact, there is at least some evidence that the
Navy’s methodology may have underestimated the effects of
MFA sonar on marine mammals. NRDC has submitted decla-
rations of several scientists who cite evidence that extraordi-
nary behavior in marine mammals, including stranding by
beaked whales, may be caused by acoustic energy levels
below the Navy’s bottom impact threshold of 173 dB. See
Declaration of Dr. David E. Bain ¶¶ 6-11; Declaration of Dr.
Edward C.M. Parsons ¶ 4; Declaration of Dr. Linda Weilgart
(“Weilgart decl.”) ¶ 9.

   While NRDC has presented no evidence that marine mam-
mals have actually been harmed by the Navy’s use of MFA
sonar in the Southern California Operating Area over the past
forty years,20 the record indicates that because harm to marine
mammals is difficult to detect, except in cases of stranding,
marine mammals may nonetheless be harmed by the Navy’s
use of MFA sonar in the Southern California Operating Area.

   In the Navy’s January 2007 “after action report” following
the completion of the first three SOCAL exercises, the Navy
acknowledged that “it is difficult to assess the potential expo-
  20
    The “after action reports” compiled by the Navy following completion
of COMPTUEX and JTFEX exercises in the Southern California Operat-
ing Area do catalogue a number of marine mammal deaths. For example,
the Navy’s June 2007 “after action report” following the first three
SOCAL exercises notes that during one of the exercises the Navy
observed a floating, badly decomposed whale carcass, and that during two
of the exercises it observed floating pinniped and dolphin carcasses. The
report does not disclose whether necropsies were performed on the car-
casses and it does not discuss whether the deaths might be related to the
use of MFA sonar. Instead, the report proffers only the tentative explana-
tion that “circumstantial evidence” of a link between “potential” algal
toxin in California ocean waters and increased marine mammal mortality
“is not unexpected” and “may” have “contributed” to the fact that pinni-
ped and dolphin carcasses were observed. June 28, 2007 COMPTUEX/
JTFEX Combined After Action Report at 13-14.
2102                      NRDC v. WINTER
sure to sonar for species not observed.” Indeed, the “after
action reports” for the last eight COMPTUEX and JTFEX
exercises in the Southern California Operating Area reveal
that in less than 15% of the instances in which marine mam-
mals were observed, MFA sonar was in fact being used.21
Thus, the Navy’s reports show that relatively few marine
mammals have been observed while MFA sonar was being
used.

   Beaked whales are particularly difficult to observe. Accord-
ing to a 2004 Navy-sponsored report, “[t]heir very low
broaching profile and the limited time they spend at the sur-
face have conspired to make them resistant to easy survey-
ing.” The report notes that beaked whales are “very deep
divers” and spend an estimated 80% of their time at consider-
able depths. According to the EA, Cuvier’s and Mesoplodont
beaked whales make dives of up to 87 minutes.22 An interna-
tional workshop on beaked whales organized by the U.S.
Marine Mammal Commission noted that only 1 in 50 beaked
whales would be detected in naval mitigation surveys using
shipboard visual observation, even assuming ideal observa-
tion conditions.

   Correspondingly, injuries to beaked whales are also diffi-
cult to observe. In a January 2007 memorandum, the NMFS
concluded that “injuries or mortalities . . . would rarely be
documented, due to the remote nature of many [naval activi-
ties] and the low probability that an injured or dead beaked
whale would strand.” While the parties have presented con-
flicting declarations on the issue whether whale carcasses
resurface for some time after they initially sink, compare
  21
     The data in the “after action reports” for the eight COMPTUEX and
JTFEX exercises in the Southern California Operating Area show that of
345 instances of marine mammal sighting, MFA sonar was being used
only in 51 instances, or 14.8% of the time.
  22
     The EA provides no information on the length of the dives of Baird’s
and Ziphiid beaked whales.
                       NRDC v. WINTER                      2103
Weilgart decl. ¶ 7 with Ketten decl. ¶ 17, the likelihood that
a whale carcass would be detected if it does not strand logi-
cally depends on how well the waters are searched for such
carcasses. A 2007 study by NMFS researchers suggests that
the likelihood of detecting dead beaked whales is low, as it
concluded that, given current biological survey efforts, in
90% of beaked whale stocks a decline in population of 50%
over a 15-year period would go undetected as a decline at all.
Non-fatal injuries not leading to stranding would be even
more difficult to detect because no beaked whale carcass
would surface.

   Moreover, it is not clear from the record whether in the past
forty years the waters of southern California have been
exposed to MFA sonar at the same power level and frequency
and for the same duration as they are now. First, the Navy has
provided no information about the frequency with which exer-
cises involving the use of MFA sonar were conducted prior to
1992.

   Second, while the Navy states that “Navy data going back
to 1992 shows that the number of yearly exercises in the last
15 years and amount of [MFA sonar] use in SOCAL waters
was greater in the past than it is now, showing a slight reduc-
tion trend,” Bird decl. ¶ 18, that statement is too vague to
allow conclusions to be drawn from it. The statement says
nothing about the type of exercises or their duration over the
years, and it does not make clear whether “in SOCAL waters”
refers only to sonar use or also to the number of yearly exer-
cises. Further, the statement does not specify whether
“amount of [MFA sonar] use” refers to sonar use in each
exercise or to total sonar use in a year, nor does it make clear
whether that phrase refers to the number of times sonar was
employed or to the aggregate duration of sonar transmission.
The Navy produces no data in the record to clarify its state-
ment.

  Third, while the EA states that an average of seven JTFEX
or COMPTUEX exercises are conducted each year, which is
2104                        NRDC v. WINTER
consistent with fourteen SOCAL exercises to be conducted
over two years, the EA does not state the starting date after
which that average has been maintained.

   Fourth, while the Navy points out that its currently-used
SQS-53 sonar system transmits sonar at the same power
levels and frequencies as the SQS-26 system that the Navy
used in earlier years, it acknowledges that in a new class of
destroyers the SQS-53 system has replaced the SQS-56 sys-
tem. Bird. decl. ¶ 18. As the Biological Opinion of the NMFS
makes clear, the SQS-56 system transmits MFA sonar at a
lower power level and at different frequencies than the SQS-
53 system.23 Thus, the record suggests that with the new class
of destroyers the average MFA sonar transmission may have
increased in power level and changed in frequency. The Navy
does not cite evidence to the contrary.

   Finally, we can draw no conclusion from the statement in
the EA that “output from active sonar systems used in [the
Southern California Operating Area] and throughout the Navy
has remained largely the same for the past 30 years.” The EA
does not explain whether “output” refers to frequency, sound
intensity level, amount of time used during an exercise, or
amount of time used per year.24 Even assuming the statement
  23
      Moreover, the NMFS’ Biological Opinion states that two low-
frequency sonar systems are likely to be employed in the SOCAL exer-
cises, but it does not state in what proportion they have been used in the
past or will be used now. The fact that one of those systems has the same
acoustic capabilities as, but improved processing capabilities over, the
other system—a fact to which the Navy refers in its November 13, 2007
letter to the court—is irrelevant.
   24
      The same lack of clarity afflicts the statement in a Navy declaration
that the Navy’s training activities involving sonar during World War II
“were similar in nature and intensity to those currently analyzed in the
[EA for the SOCAL exercises].” Declaration of Conrad Erkelens ¶ 16. The
statement does not make clear whether “intensity” refers generally to the
frequency or size of the exercises or more specifically to the length of time
during which sonar was used. Even if it refers to the latter, it appears to
refer to the use of sonar per training activity, not per year.
                           NRDC v. WINTER                           2105
refers to the use of the sonar systems during an exercise or
during the year, it is unclear whether it refers to total use in
the Southern California Operating Area or rather to total use
by the Navy in all its training areas combined.

   In any event, the Navy’s estimate that its use of MFA sonar
in the SOCAL exercises will expose marine mammals to 564
instances of Level A harassment and nearly 170,000 instances
of Level B harassment clearly indicates that at least some sub-
stantial harm will likely occur in the Southern California
Operating Area.

  C.    The Mitigation Measures Employed by the Navy
         and Those Imposed by the District Court

   While the Navy adopted a number of mitigation measures
intended to reduce the harm caused by the use of MFA sonar
in the SOCAL exercises, the district court concluded that
those measures were inadequate both to cure the Navy’s
likely NEPA violation and to avoid the possibility of irrepara-
ble harm to NRDC. Accordingly, following our November
13, 2007 remand order, the district court established addi-
tional, narrowly-tailored mitigation measures which the Navy
would have to employ during the remaining SOCAL exer-
cises. To place these mitigation measures in context, we
explain what mitigation measures the Navy has previously
employed and is currently employing in the SOCAL exer-
cises.

   In June 2006, shortly before the Navy was to conduct that
year’s “Rim of the Pacific” exercise off the coast of Hawaii
(the “2006 RIMPAC exercise”), plaintiffs sued the Navy and
the same co-defendants here,25 seeking to enjoin the Navy
from using MFA sonar in that exercise. Following the district
court’s grant of NRDC’s motion for a temporary restraining
  25
    Only the League for Coastal Protection was not a plaintiff in the 2006
action.
2106                       NRDC v. WINTER
order, the parties entered into a settlement agreement that
allowed the Navy to use MFA sonar in the 2006 RIMPAC
exercise but only if it employed certain mitigation measures
in addition to those already imposed by the NMFS in its June
27, 2006 Incidental Harassment Authorization and by the
Department of Defense in its June 30, 2006 National Defense
Exemption (“NDE I”).26

  The mitigation measures the Navy adopted for the 2006
RIMPAC exercise include operating MFA sonar at the lowest
practicable level not to exceed 235 dB except for short peri-
ods to meet tactical training objectives, and using at least one
lookout dedicated to the detection of marine mammals, as
well as three non-dedicated lookouts, on each ship operating
MFA sonar and requiring them to report sightings of marine
mammals.

  The following mitigation measures employed during the
2006 RIMPAC exercise are of particular importance here:

       •   The designation of “safety zones” in which:

            -   the MFA sonar level is reduced by 6 dB
                if a marine mammal is detected within
                1,000 meters of the sonar dome (located
                in the bow of the vessel);27
  26
      The NMFS’s Incidental Harassment Authorization (IHA), issued pur-
suant to 16 U.S.C. § 1371(a)(5), authorized the incidental “taking” of a
small number of marine mammals under the Marine Mammal Protection
Act (MMPA), 16 U.S.C. § 1361 et seq. The Deputy Secretary of Defense
incorporated the mitigation measures imposed by the NMFS into NDE I
issued pursuant to 16 U.S.C. § 1371(f), which exempted from the MMPA
for a period of six months all military readiness activities employing MFA
sonar, including the 2006 RIMPAC exercise. The NDE I imposed differ-
ent mitigation measures for non-RIMPAC exercises during the six month
period, including the prohibition on use of MFA sonar within 12 nautical
miles of a coast.
   27
      Because the decibel is a logarithmic unit of acoustic power (using the
base-10 logarithm), a reduction in sonar level of 6 dB corresponds to a
reduction in sound intensity of approximately 75%, and a reduction in
sonar level of 10 dB corresponds to a reduction in sound intensity of 90%.
See Bird decl. ¶ 29.
                           NRDC v. WINTER                           2107
            -   the MFA sonar level is reduced by 10
                dB if a marine mammal is detected
                within 500 meters of the sonar dome;
                and

            -   the use of MFA sonar is stopped if a
                marine mammal is detected within 200
                meters of the sonar dome.28

       •   In conditions of strong surface ducting—where
           sonar sound carries over a greater distance than
           would otherwise be the case—the safety zones
           will be expanded such that the MFA sonar level
           is reduced by 6 dB if a marine mammal is
           detected within 2,000 meters of the sonar dome
           and by 10 dB if one is detected within 1,000
           meters of the dome, and that the use of MFA
           sonar is stopped if a marine mammal is detected
           within 500 meters of the sonar dome.

       •   In conditions of low visibility—i.e., whenever a
           safety zone is not fully visible—additional detec-
           tion measures will be used, such as infrared or
           enhanced passive acoustic detection. If detection
           of marine mammals is not possible out to the lim-
           its of the safety zone, the sonar level will be
           reduced as if a marine mammal is present imme-
           diately beyond the extent of detection.

       •   With the exception of three specific “choke
           point” exercises, MFA sonar will not be operated
           in constricted channels or canyon-like areas or
           within 25 kilometers of the 200 meter isobath.29
  28
      As one meter equals 1.094 yards, the sizes of the three safety zones
correspond to approximately 1,100 yards, 550 yards, and 220 yards,
respectively.
   29
      An isobath is a line (either imaginary or on a map) joining places
where water has equal depth.
2108                        NRDC v. WINTER
       •   During the three choke point exercises, additional
           mitigation measures will be employed, including
           two hours of pre-exercise monitoring for marine
           mammals.

   Following the 2006 RIMPAC exercise, the Navy issued an
“after action report” in which it reported that it had used MFA
sonar for a total of 472 hours during the 2006 RIMPAC exer-
cise and that the mitigation measures resulted in a loss of 8
hours of MFA sonar use.30 Dec. 7, 2006 Rim of the Pacific
After Action Report at 9. Although no unusual behavior of
marine mammals had been observed, the Navy reported that
of the estimated 256 marine mammals potentially affected by
the 472 hours of MFA sonar use, approximately 100 of them
had been precluded from exposure to MFA sonar as a result
of the mitigation measures.

   In October 2006, in anticipation of the SOCAL exercises,
the Navy submitted a consistency determination to the CCC,
seeking the CCC’s concurrence in the Navy’s determination
that the SOCAL exercises were consistent to the maximum
extent possible with the enforceable policies of the CCMP,
which, for purposes of the CZMA, are contained in the Cali-
fornia Coastal Act. See Cal. Pub. Res. Code § 30008.31 The
CCC disagreed with the Navy that the use of MFA sonar in
the SOCAL exercises would not affect California’s coastal
resources. Accordingly, the CCC conditioned its concurrence
in the Navy’s consistency determination on the Navy adopting
  30
      The Navy asserted in its “after action report” that 8 hours of lost MFA
sonar use translated into a somewhat greater amount of lost exercise time
because once the sonar is turned off, simply turning it back on does not
usually allow the Navy Commander to continue the exercise where it left
off.
   31
      The CZMA requires that a federal agency planning to conduct activi-
ties that may have reasonably foreseeable effects on California’s coastal
resources must apply to the CCC for a determination that those activities
are consistent to the maximum extent practicable with the enforceable pol-
icies of the CCMP. See 16 U.S.C. § 1456(c)(1)(C); 15 C.F.R. § 930.36.
                          NRDC v. WINTER                          2109
fourteen mitigation measures, several of which the Navy had
already employed in the 2006 RIMPAC exercise. The Navy
agreed to adopt only four of the CCC’s measures—all but one
of which do not actually prevent the use of MFA sonar from
harming marine mammals32 —and it refused to adopt the
remaining ten measures:

       •   The measures used in the 2006 RIMPAC exercise
           for reducing sonar levels during conditions of
           low visibility;

       •   The measures used in the 2006 RIMPAC exercise
           in “choke points;”

       •   Increasing the outer safety zone to 2 kilometers
           even in conditions not involving strong surface
           ducting;33

       •   Reducing sonar level by 6 dB during strong sur-
           face ducting conditions;
  32
      Following the CCC’s conditional concurrence, the Navy agreed to
adopt two of the CCC’s mitigation measures: retrieving inert dropped
mine shapes from the water and submitting to the CCC all monitoring
results provided to the NMFS. The mitigation measures set forth in the
EA, which have been standard operating procedure since 2004, appear to
include two other measures proposed by the CCC: requiring passive sonar
operators to monitor for marine mammals and report the detection of any
such mammals; and providing a report to the NMFS following a major
exercise that includes the results of marine mammal monitoring (a mea-
sure already employed in the 2006 RIMPAC exercise). The Navy con-
cedes that the measure concerning dropped mine shapes “has nothing to
do with MFA sonar usage,” Dec. 20, 2007 Declaration of Rear Admiral
John M. Bird ¶ 13, and the two reporting requirements do not mitigate
actual harm to marine mammals but instead assist in the determination of
the impacts of the SOCAL exercises. See Feb. 4, 2008 Dist. Ct. Order at
6 n.7. Only the use of passive sonar to monitor for marine mammals miti-
gates the harm caused by the use of MFA sonar.
   33
      As stated above, the Navy had agreed to employ in the 2006 RIMPAC
exercise a 2,000 meter safety zone in strong surface ducting conditions.
2110                       NRDC v. WINTER
       •   Using two marine mammal observers who have
           received NOAA-approved training for surveil-
           lance during use of MFA sonar;

       •   Requiring aerial monitoring off San Clemente
           Island throughout exercises involving MFA
           sonar;

       •   Monitoring for marine mammals for 30 minutes
           prior to commencing use of MFA sonar;

       •   Avoiding training in areas with known high con-
           centrations of marine mammals; and

       •   Locating and scheduling training outside the
           migration season for the grey whale.34

   In January 2007, the Deputy Secretary of Defense issued,
pursuant to 17 U.S.C. § 1371(f), a second National Defense
Exemption (“NDE II”), which exempted from the require-
ments of the MMPA all the Navy’s military readiness activi-
ties employing MFA sonar for the duration of the SOCAL
exercises. The Deputy Secretary of Defense conditioned the
exemption on the Navy adopting a number of mitigation mea-
sures, which already had been standard operating procedure in
the Navy’s ASW exercises since 2004. As the EA makes
clear, it is those mitigation measures, and only those mea-
sures, which the Navy adopted for the SOCAL exercises.

   While the Navy describes the adopted mitigation measures
as consisting of 29 separate measures, the district court found
that, in effect, they consist of only four basic measures: “(1)
personnel training (providing approved Marine Species
Awareness Training materials for lookouts and commanding
  34
    The nine bullet points reflect the substance of the ten mitigation mea-
sures the Navy refused to adopt; two of the conditions have been consoli-
dated in the fifth bullet point.
                            NRDC v. WINTER                              2111
officers), (2) on-deck lookouts, armed with binoculars or
night vision goggles, to watch for marine mammals, (3) oper-
ating procedures to ensure that any sightings of marine mam-
mals are communicated up the chain of command, so that
MFA sonar is powered down [(i.e., sonar power is reduced)]
when a marine mammal approaches within 1,000 yards, 500
yards, and ‘secured’ (shut-down) at 200 yards,35 and (4) coor-
dination and reporting procedures.” Feb. 4, 2008 Dist. Ct.
Order at 6 n.7. Our review of the EA reveals at most two
additional basic measures: that passive sonar will be used to
detect marine mammals and that Navy aircraft participating in
exercises will conduct surveillance when doing so does not
interfere with safety or the “accomplishment of primary oper-
ational duties.”36 Notably, the measures adopted by the Navy
  35
      We note that the Navy has recently agreed to slightly enlarge its safety
zones by applying safety zones of 1,000, 500 and 200 meters instead of
1,000, 500 and 200 yards. Bird decl. ¶ 58. As a result, those safety zones
are now of equal size as those employed in the 2006 RIMPAC exercise.
However, the Navy has not agreed to employ expanded safety zones in
strong surface ducting conditions.
   36
      The Navy also adopted a mitigation measure requiring “increased vig-
ilance” during major ASW exercises using MFA sonar when “critical con-
ditions” are present: a rapid change in bathymetry in areas of a certain
depth; where three or more vessels are operating MFA sonar in the same
area for six hours or more; where MFA sonar may “cut off” the exit route
for marine mammals from a bay or channel; and the historical presence of
a significant surface duct. However, in its June 2007 “after action report”
the Navy explained that it had assessed the conditions of the Southern Cal-
ifornia Operating Area and concluded that “the requirements stated in [the
aforementioned mitigation measure] do not apply to the physical condi-
tions found in Southern California.” Accordingly, this mitigation measure
is not being employed in the SOCAL exercises and therefore does not, in
fact, mitigate the impact of MFA sonar on marine mammals during those
exercises.
   Similarly, the Navy adopted as a mitigation measure the requirement
that MFA sonar be operated “at the lowest practical level, not to exceed
235 dB, except as required to meet tactical training objectives.” However,
that requirement, also adopted for the 2006 RIMPAC exercise, has no
apparent mitigating effect because (1) it is not clear whether the Navy ever
2112                       NRDC v. WINTER
do not include the ten aforementioned mitigation measures
recommended by the CCC, such as increasing the outer safety
zone to 2 kilometers, monitoring for marine mammals at least
30 minutes before commencing MFA sonar use, and conduct-
ing the SOCAL exercises outside the grey whale’s migratory
season and outside areas with high concentrations of marine
mammals. Moreover, the adopted measures do not include the
2006 RIMPAC measures requiring that MFA sonar not be
used in constricted channels and canyon-like areas or within
25 kilometers of the 200 meter isobath. Nor do they include
the requirement, imposed by NDE I for non-RIMPAC exer-
cises in 2006, that MFA sonar not be used within 12 nautical
miles from the coastline.

   Following our November 13, 2007 remand order, the dis-
trict court set out to determine what narrowly-tailored mitiga-
tion measures should be imposed for the remaining SOCAL
exercises. In the words of the district court, NRDC proposed
“sweeping geographic exclusions” to the Navy’s use of MFA
sonar, including a 25 nautical mile coastal exclusion, locating
exercises to the maximum extent possible in waters deeper
than 1,500 meters, and an exclusion in the Catalina Basin, the
Westfall seamount, and the Cortez and Tanner Banks. Jan. 3,
2008 Dist. Ct. Order at 13 n.6. The Navy also proposed sev-
eral mitigation measures. See 

the parties’ briefs and taking a Navy-
guided tour of the USS Milius, the district court imposed six
mitigation measures in addition to those already required by

assumed higher sonar levels when using its harassment model to predict
harm to marine mammals, (2) the Navy’s interim report on the Bahamas
stranding indicates that sonar levels during the Bahamas exercise did not
exceed 235 dB even without the mitigation measure, and (3) the exception
that the Navy may exceed the 235 dB limit “as required to meet tactical
training objectives” swallows the rule, as tactical training objectives are
the only reason for using MFA sonar to begin with, thus allowing the
Navy to exceed the 235 dB limit whenever it in fact uses MFA sonar.
                           NRDC v. WINTER                             2113
NDE II: (1) the Navy shall suspend use of MFA sonar when
a marine mammal is detected within 2,200 yards from the
sonar source, except where the marine mammal is a dolphin
or a porpoise and it appears that the mammal is intentionally
following the sonar-emitting naval vessel in order to play in
or ride the vessel’s bow wave; (2) the Navy shall reduce the
MFA sonar level by 6 dB when significant surface ducting
conditions are detected;37 (3) the Navy shall not use MFA
sonar within 12 nautical miles from the California coastline;
(4) the Navy shall monitor, including by aircraft, for the pres-
ence of marine mammals for 60 minutes before employing
MFA sonar, shall utilize two dedicated, NOAA- and NMFS-
trained lookouts at all times when MFA sonar is being used,
shall employ passive acoustic monitoring to supplement
visual detection of the presence of marine mammals, and shall
use aircraft participating in the training exercises to monitor
for marine mammals for the duration of the exercises when
MFA sonar is being used; (5) Navy helicopters shall monitor
for marine mammals for 10 minutes before employing active
dipping sonar; and (6) the Navy shall refrain from using MFA
sonar in the Catalina Basin between the Santa Catalina and
San Clemente Islands because ingress and egress to the basin
are restricted and the basin has a high density of marine mam-
mals. See Jan. 10, 2008 Dist. Ct. Order at 1-5.

   The district court rejected many of the geographic exclu-
sions proposed by NRDC in favor of a 2,200-yard safety
zone, accepted the Navy’s representations that the bathymetry
off the shores of southern California presents unique training
opportunities, see Feb. 4, 2008 Dist. Ct. Order at 7, and
declined to limit the use of sonar at night or in conditions of
low visibility despite the Navy’s voluntarily adoption of such
limitations for the 2006 RIMPAC exercise, see Jan. 3, 2008
Dist. Ct. Order at 7-8.
  37
     The parties clarified at oral argument on February 27, 2008, that “sig-
nificant” is the same as “strong,” the term used to modify surface ducting
conditions in the mitigation measures for the 2006 RIMPAC exercise.
2114                     NRDC v. WINTER
   The Navy takes issue only with the first two of the mitiga-
tion measures imposed by the district court, namely the 2,200
yard “shutdown zone” and the “power-down” requirement
during significant surface ducting conditions. Specifically, the
Navy argues that those two mitigation measures tip the bal-
ance of hardships in its favor and are contrary to the public
interest.

   In support of its argument the Navy has submitted declara-
tions by high-ranking officers attesting to the adverse conse-
quences that those measures will have on the Navy’s military
readiness. For example, Vice Admiral Locklear, Commander
of the U.S. Third Fleet, opines that “imposing a 2200-yard
shutdown zone will have crippling implications on Navy’s
ability to conduct realistic pre-deployment ASW training
employing MFA sonar” and will “jeopardize the training and
readiness of U.S. THIRD Fleet [strike groups].” Declaration
of Vice Admiral Samuel J. Locklear. ¶¶ 9, 13. Likewise, Rear
Admiral Bird opines that “[t]raining in surface ducting condi-
tions is critical to effective training.” Bird decl. ¶ 52. In their
classified declarations, Admiral Gary Roughead and Rear
Admiral Ted N. Branch opine that both the 2200-yard shut-
down zone and the power-down requirement in significant
surface ducting conditions will create an unacceptable risk
with respect to the Navy’s ability to certify its strike groups
as combat ready and will thus profoundly affect national
security.

                  III.   Standards of Review

   Our review of a district court’s grant of a preliminary
injunction is “very deferential.” Nat’l Wildlife Fed’n v. Nat’I
Marine Fisheries Serv., 

, 794 (9th Cir. 2005).
We do not reverse the district court unless it “relie[s] on an
erroneous legal premise or abuse[s] its discretion.” Sports
Form, Inc. v. United Press Int’l, Inc., 

, 752 (9th
Cir. 1982) (internal citations omitted). A court abuses its dis-
cretion if it bases its decision on an erroneous legal standard
                        NRDC v. WINTER                        2115
or clearly erroneous findings of fact. Earth Island Inst. v. U.S.
Forest Serv., 

, 1156 (9th Cir. 2006) (“Earth
Island II”).

   A district court may grant a preliminary injunction if one
of two sets of criteria are met. “Under the ‘traditional’ criteria,
a plaintiff must show (1) a strong likelihood of success on the
merits, (2) the possibility of irreparable injury to plaintiff if
preliminary relief is not granted, (3) a balance of hardships
favoring the plaintiff, and (4) advancement of the public inter-
est (in certain cases). Alternatively, a court may grant the
injunction if the plaintiff demonstrates either a combination of
probable success on the merits and the possibility of irrepara-
ble injury or that serious questions are raised and the balance
of hardships tips sharply in his favor.” Freecycle Network,
Inc. v. Oey, 

, 902 (9th Cir. 2007); see also Earth
Island 

.

                        IV.   Discussion

         A.    Likelihood of Success on the Merits

   1.   Effect of CEQ’s Alternative Arrangements for
                    NEPA Compliance

   On January 15, 2008 CEQ purported to approve “alterna-
tive arrangements” for the Navy to continue its use of MFA
sonar while complying with NEPA, reasoning that “emer-
gency circumstances” prevented normal compliance. CEQ’s
authority to grant such relief derives from 40 C.F.R.
§ 1506.11, which provides in full:

    Where emergency circumstances make it necessary
    to take an action with significant environmental
    impact without observing the provisions of these
    regulations, the Federal agency taking the action
    should consult with the Council about alternative
    arrangements. Agencies and the Council will limit
2116                   NRDC v. WINTER
    such arrangements to actions necessary to control the
    immediate impacts of the emergency. Other actions
    remain subject to NEPA review.

40 C.F.R. § 1506.11. CEQ’s letter of explanation to the Navy
stated that the district court’s modified injunction “imposes
training restrictions . . . that continue to create a significant
and unreasonable risk that Strike Groups will not be able to
train and be certified as fully mission capable.” CEQ Letter
to Donald C. Winter at 3. CEQ also stated that “the inability
to train effectively with MFA sonar puts the lives of thou-
sands of Americans directly at risk. . . . Therefore, there are
urgent national security reasons for providing alternative
arrangements under the CEQ regulations.” 

   The Navy then petitioned this court to vacate the district
court’s preliminary injunction, arguing that CEQ’s approval
of “alternative arrangements” deprived NRDC of the “likeli-
hood of success on the merits” of its NEPA claims, thus elim-
inating the legal basis for the injunction. We remanded to the
district court to allow it to consider in the first instance
whether this legal development merited vacatur or a partial
stay of the injunction.

   On remand, the Navy maintained that the CEQ’s “emer-
gency circumstances” determination relieved it of the require-
ment to prepare an EIS prior to commencing the remaining
SOCAL exercises. NRDC argued that CEQ’s action was
beyond the scope of the regulation and otherwise invalid, and
that the preliminary injunction should remain in place. The
district court considered these arguments and concluded that
its preliminary injunction was “not affected by [CEQ’s]
approval of emergency alternative arrangements because there
is no emergency.” Feb. 4, 2008 Dist. Ct. Order at 2. Accord-
ingly, it held that “CEQ’s action is beyond the scope of the
regulation and is invalid[ ]” and that “[t]he Navy is not, there-
fore exempted from compliance” with NEPA and the prelimi-
nary injunction. 

court found that CEQ’s
                            NRDC v. WINTER                             2117
interpretation of “emergency circumstances” to include a
court order entered in the course of pending litigation was not
authorized by 40 C.F.R. §1506.11, because it was contrary to
both the plain meaning of “emergency circumstances” and the
drafters’ original intent. It also found that CEQ’s action was
contrary to the governing statute, NEPA.38

   The Navy makes two basic arguments as to why the district
court erred by failing to vacate the preliminary injunction in
light of CEQ’s approval of “alternative arrangements.” First,
the Navy argues that the district court lacked subject matter
jurisdiction to review CEQ’s approval of alternative arrange-
ments because such approval constitutes a superseding agency
action that removes as moot any basis for an injunction predi-
cated on plaintiffs’ original claims concerning the Navy’s EA.
Second, the Navy argues that, even if the district court could
review CEQ’s action, the court erred by not deferring to
CEQ’s and the Navy’s assessment that “emergency circum-
stances” exist within the meaning of 40 C.F.R. § 1506.11. We
address the Navy’s arguments in turn.

a.     Subject Matter Jurisdiction over NRDC’s Challenge
                      to CEQ’s Action

   [1] The Navy argues that the district court lacked subject
matter jurisdiction to consider the validity of CEQ’s action
because that action allegedly moots the plaintiffs’ original
claims. Specifically, the Navy contends that the adoption of
CEQ’s “alternative arrangements” in a superseding Decision
Memorandum constitutes a new administrative action, which
can only be challenged by a new claim on the merits. See Rat-
  38
     The district court also noted that CEQ had essentially crafted its own,
alternative injunction, which suggested that CEQ, as an executive body,
was effectively “sitting in review of a decision of the judicial branch.” The
district court observed that CEQ’s actions raised “serious constitutional
concerns under the Separation of Powers doctrine,” but it declined, pursu-
ant to the doctrine of constitutional avoidance, to make a finding as to the
constitutionality of its action. See Feb. 4, 2008 Dist. Ct. Order at 24.
2118                       NRDC v. WINTER
tlesnake Coalition v. EPA, 

, 1103-04 (holding
that the district court lacked subject matter jurisdiction over
new final agency action).39 The district court implicitly
rejected this argument by continuing to exercise jurisdiction
over the plaintiffs’ NEPA claim. Its decision to do so is firmly
grounded in the familiar principle that only a valid subsequent
action can render a legal claim moot. See Adarand Construc-
tors, Inc. v. Slater, 

, 222-23 (2000); United
States v. Larson, 

, 1020 (9th Cir. 2002) (“The
stipulation moots [plaintiff’s] challenge to the suppression
ruling only if it is valid.”). Accordingly, the district court did
not rely on an erroneous legal premise or abuse its discretion
in concluding that it had jurisdiction to assess the validity of
the new action in order to determine whether plaintiffs’ origi-
nal claims could survive. See 

.

    [2] Nor did the district court abuse its discretion by leaving
in place the preliminary injunction after determining that
CEQ’s action did not require its vacatur. The Navy’s conten-
tion that the district court issued “an entirely new injunction
. . . based on new, ancillary claims” mischaracterizes the pos-
ture of this case. The Navy challenged the injunction based on
  39
    Notably, none of the cases that the Navy cites in support of this argu-
ment involve a plaintiff’s challenge to the validity of a new agency action.
For example, in Forest Guardians v. U.S. Forest Serv., 

,
1096 (9th Cir. 2003), plaintiffs’ claims were rendered moot because the
superseding and controlling environmental documentation that displaced
the earlier agency action did not rely on the challenged assumptions that
formed the basis of plaintiffs’ claims. In W. Radio Serv. Co. Inc. v. Glick-
man, 

, 974 (9th Cir. 1997), a challenge to a letter postponing
the issuance of certain permits until a fee structure was established was
rendered moot by the subsequent issuance of a fee structure. In Aluminum
Co. of Am. v. Bonneville Power Admin., 

(9th Cir. 1995),
challenges to a 1993 Rule of Decision were moot because augmentations
were being issued under a subsequent Rule of Decision. Finally, in Ore-
gon Natural Res. Council v. Harrell, 

, 1501-02, 1508 (9th
Cir. 1995), the district court ordered a challenged ROD withdrawn
because it was incomplete, and a subsequent challenge to that non-
operative ROD was thus held to be moot.
                       NRDC v. WINTER                       2119
CEQ’s action; NRDC argued only that CEQ’s action did not
change the merits of its NEPA claims. Thus, the district court
here did not “[g]rant[ ] a preliminary injunction based on a
showing that the plaintiffs were likely to succeed in establish-
ing a violation of an ancillary court order, rather than a show-
ing that they were likely to succeed on the merits of any of
their claims.” Alabama v. U.S. Army Corps of Engineers, 

, 1135 (11th Cir. 2005).

    b.   The District Court’s Assessment of Whether
          “Emergency Circumstances” Existed

                        (1)   Deference

   The district court concluded that CEQ’s interpretation of 40
C.F.R. § 1506.11 is not entitled to deference. It reasoned that
under the Administrative Procedure Act (“APA”), 5 U.S.C.
§ 551 et seq., the courts traditionally afford deference to (1)
an agency’s reasonable interpretation of a statute it adminis-
ters “if the statute is silent or ambiguous with respect to the
specific issue . . . ,” citing Chevron, U.S.A., Inc. v. NRDC, 

, 843 (1984), and (2) an agency’s interpretation of its
own regulations unless “an alternative reading is compelled
by the regulation’s plain language or by other indications of
the [agency’s] intent at the time of the regulation’s promulga-
tion,” citing Thomas Jefferson Univ. v. Shalala, 

,
512 (1994), and Bowles v. Seminole Rock & Sand Co., 

, 414 (1945). See Feb. 4, 2008 Dist. Ct. Order at 13-
14.

   NRDC challenged neither the propriety of CEQ’s original
promulgation of 40 C.F.R. § 1506.11 nor the fact that
§ 1506.11 allows alternative arrangements for compliance
with NEPA under genuine emergency circumstances. Instead,
NRDC limited its challenge to CEQ’s application of the regu-
lation to the facts of this case. Accordingly, the district court
considered whether the term “emergency circumstances”
could be afforded so broad an interpretation as to encompass
2120                        NRDC v. WINTER
the Navy’s need to continue its long-planned, routine sonar
training exercises without the mitigation measures imposed
by the district court. The district court concluded that the
plain language of the regulation and the limited indicia of the
agency’s original intent compelled a narrower interpretation
of “emergency circumstances” than the one afforded it by
CEQ. Accordingly, the district court concluded that it did not
owe deference to CEQ’s interpretation of § 1506.11 under
Thomas Jefferson and Seminole Rock. We review that conclu-
sion to determine whether in so doing it relied on an errone-
ous legal premise or abused its discretion; we conclude that
it did neither.

   [3] The district court followed established Supreme Court
precedent in finding that an agency’s interpretation of its own
regulation is not entitled to deference when it is inconsistent
with the regulation itself, conflicts with agency intent at the
time of promulgation, and reaches beyond “the limits imposed
by the statute,” NEPA. See Auer v. Robbins, 

,
461-63 (1997). Next, the district court, after concluding that
the meaning of “emergency circumstances” was clear, applied
the appropriate legal principles that an agency’s interpretation
of its own regulation is entitled to deference “only when the
language of the regulation is ambiguous.” See Christensen v.
Harris County, 

, 588 (2000). Accordingly, the
district court did not abuse its discretion when it determined
not to give deference to CEQ’s overly broad interpretation of
“emergency circumstances.”40
  40
     The district court also held that Skidmore deference is inapplicable
because this case involves an agency’s interpretation of a regulation, not
its informal interpretation of a statute it administers. See Feb. 4, 2008 Dist.
Ct. Order at 19 n.13 (rejecting plaintiffs’ argument that because it does not
arise out of formal rulemaking, CEQ’s interpretation is entitled to little or
no deference under Skidmore v. Swift & Co., 

, 140 (1944)).
While this holding is correct as a matter of law, our determination that the
district court neither abused its discretion nor relied on an erroneous legal
premise in concluding that no deference is owed CEQ’s overly broad
interpretation of “emergency circumstances” renders this issue moot.
                        NRDC v. WINTER                       2121
   (2)   Plain Meaning and Intent of CEQ Regulation

   In finding that no emergency circumstances existed, the
district court reasoned that the “Navy’s current ‘emergency’
is simply a creature of its own making, i.e., its failure to pre-
pare adequate environmental documentation in a timely fash-
ion, via the traditional EIS process or otherwise.” Feb. 4, 2008
Dist. Ct. Order at 17. In short, it was not a sudden unantici-
pated event. The district court supported its conclusion by
noting that the CEQ letter does not specify an “emergency”
other than the district court’s mitigation order itself, which, in
CEQ’s view, creates a “significant and unreasonable risk” that
strike groups will not be able to train and be certified as fully
mission capable. 

   [4] On appeal, the Navy argues that “no matter its gene-
sis[,] . . . the inability to certify its west coast-based strike
groups for deployment to hostile areas overseas during a time
of war” is a pressing national emergency. The Navy cites to
declarations, including one of the Chief of Naval Operations,
that attests to the national security impacts of such a failure
at a time when the United States is currently engaged in war
in two countries. The Navy contends that the district court’s
view that the term “emergency circumstances” as used in the
regulation per se excludes the circumstances presented here is
an impermissible substitution of its judgment for that of mul-
tiple federal agencies (citing Ass’n of Pac. Fisheries v. EPA,

, 810-11 (9th Cir. 1980)). We reject this argu-
ment and hold, for the reasons explained above, that the dis-
trict court did not abuse its discretion in determining that the
plain meaning of “emergency circumstances” precludes an
interpretation so broad as to encompass the Navy’s need to
continue long-planned, routine training exercises without mit-
igation measures ordered by the court.

   [5] There is ample support for the manner in which the dis-
trict court exercised its discretion. The district court properly
relied on dictionary definitions of “emergency” and “emer-
2122                        NRDC v. WINTER
gency circumstances” to support its conclusion that CEQ’s
interpretation is not entitled to deference. See Watson v.
United States, 

, 583 (2007) (noting that terms
are construed consistently with their everyday meaning,
including by reference to the dictionary absent statutory defi-
nition or definitive clue). As the district court observed, the
Oxford English Online Dictionary defines “emergency” as
“[t]he arising, sudden or unexpected occurrence (of a state of
things, an event, etc.).” Oxford English Online Dictionary,
available at http://dictionary.oed.com. Black’s Law Dictio-
nary defines the term “emergency circumstances,” through a
cross-reference to “exigent circumstances,” as “[a] situation
that demands unusual or immediate action and that may allow
people to circumvent usual procedures, as when a neighbor
breaks through a window of a burning house to save someone
inside.” Blacks Law Dictionary 260, 562 (8th ed. 2004)
(emphasis added). The district court did not abuse its discre-
tion in concluding that the circumstances in the present case
fall outside the scope of these definitions because its prelimi-
nary injunction was entirely predictable given the parties’ liti-
gation history. Feb. 4, 2008 Dist. Ct. Order at 15.41

   The Navy urges that the risk to national security created by
the preliminary injunction falls squarely within the legal defi-
nition of “emergency circumstances.” However, the Navy has
been on notice of its possible legal obligations to prepare an
EIS for the SOCAL exercises from the moment it first
   41
      NRDC provides several more dictionary definitions of “emergency,”
all of which include the terms “unexpected” or “unforseen.” The Navy
refers us to another source that defines “emergency” as “a situation
demanding immediate attention.” Random House Dict. of the English
Lang. 636 (2d ed. 1987). We do not adjudicate the meaning of the word
“emergency” here. Rather, we need conclude only that the district court
did not rely on erroneous legal principles or abuse its discretion in reach-
ing its determination as to that term’s plain meaning. Because we are not
“left with the definite and firm conviction that a mistake has been commit-
ted[,]” we leave the district court’s determination intact. Sports Form, 

.
                       NRDC v. WINTER                       2123
planned those exercises. In addition, NRDC filed its com-
plaint almost a year ago, and on August 7, 2007, the district
court held that the Navy was likely to lose on the merits of
NRDC’s claims. We affirmed that ruling in November of
2007. Still, the Navy waited until January 10, 2008, to raise
a cry of “emergency” and request the NEPA and CZMA
waivers it relies on here, in order to continue its routine,
planned training exercises. We find no abuse of discretion in
the district court’s determination that such a series of events
gives rise to a predictable outcome, not an unforeseeable one
demanding “unusual or immediate action.”

   Moreover, the district court’s conclusion finds support in
CEQ’s recent response to Hurricane Katrina. In March 2006,
CEQ approved alternative arrangements to allow the Federal
Emergency Management Agency to respond on an emergency
basis to “[d]amages to the critical physical infrastructure in
the New Orleans Metropolitan Area from the impact of Hurri-
canes Katrina and Rita [that] rendered parts of the city inoper-
able and uninhabitable.” NEPA Alternative Arrangements for
Critical Physical Infrastructure in New Orleans, 71 Fed. Reg.
14712, 14713 (March 23, 2006). The alternative arrangements
explain that generally, such arrangements are made when
“emergency circumstances require taking actions with signifi-
cant environmental impacts and there is not sufficient time to
follow the regular [EIS] process.” 

This
language supports the district court’s view that the words
“emergency circumstances” in 40 C.F.R. § 1506.11 refer to
unexpected, suddenly arising situations that require agency
action in a shorter time frame than would be required to pre-
pare an EIS. By contrast, the Navy’s routine SOCAL exer-
cises were planned well in advance and with “sufficient time
to follow the regular [EIS] process.”

   In concluding that the Navy’s failure to comply timely with
NEPA does not constitute an “emergency circumstance”
within the meaning of 40 C.F.R. § 1506.11, the district court
also relied on the limited drafting history of the regulation. As
2124                        NRDC v. WINTER
the district court pointed out, the initial proposed version of
the regulation required an agency “proposing to take” an
emergency action to consult with CEQ regarding alternative
arrangements.42 See Proposed Implementation of Procedural
Provisions, 43 Fed. Reg. 25230, 25243 (June 9, 1978)
(emphasis added). However, the drafters changed “proposing
to take” to “taking” in order to remove the inference that con-
sultation must necessarily precede agency action because
“such a requirement might be impractical in emergency cir-
cumstances and could defeat the purpose of the section.”
Implementation of Procedural Provisions; Final Regulations,
43 Fed. Reg. 55978, 55988 (Nov. 29, 1978). While the Navy
dismisses this drafting history as reflecting only that “the pro-
posed regulation was broadened to allow actions to be taken
prior to consultation with CEQ,” we find no abuse of discre-
tion in the district court’s contrary conclusion that the regula-
tory history “supports a narrow, rather than a broad
interpretation of the phrase ‘emergency circumstances’ ” and
that the regulatory history reflects CEQ’s intent to use the reg-
ulation to accommodate “sudden unanticipated events” but
not more predictable events such as provisionally unfavorable
litigation results. Feb. 4, 2008 Dist. Ct. Order at 19.

   Accordingly, we conclude that the district court did not rely
on an erroneous legal premise or abuse its discretion in find-
ing that the Navy’s attempt to characterize a federal court
injunction as an “emergency circumstance” is contrary to the
plain meaning of the language and to the intended purpose of
CEQ’s emergency circumstances regulation.43
  42
      In its analysis of the agency’s intent, the district court appropriately
declined to consider the declaration of Nicholas C. Yost, CEQ general
counsel at the time the regulation was drafted, as “an unreliable guide” to
CEQ’s intent, comparing the declaration to “subsequent legislative histo-
ry.” Feb. 4, 2008 Dist. Ct. Order at 18-19 (quoting Chapman v. United
States, 

, 464 n.4 (1991) (internal citations omitted)).
   43
      NRDC makes the additional argument that CEQ’s order goes beyond
the scope of the regulation because the “alternative arrangements,” pre-
                           NRDC v. WINTER                           2125
                       (3)    Prior Decisions

   [6] Prior judicial decisions also support our conclusion that
the district court did not rely on erroneous legal premises or
abuse its discretion in concluding that CEQ’s action is invalid.
In each of the cases sustaining an application of 40 C.F.R.
§ 1506.11, CEQ allowed “alternative arrangements” in
response to unanticipated emergencies that required federal
agencies to respond quickly to new and changing events.

   For example, in Valley Citizens for a Safe Environment v.
Vest, the court upheld “alternative arrangements” which per-
mitted the Air Force to fly C-5A transport planes into and out
of Westover Air Force Base on a twenty-four hour schedule,
despite a previously prepared EIS’s prohibition of such
flights. 

(D. Mass. May 30, 1991). The court
in Valley Citizens denied the plaintiffs’ request for an injunc-
tion until a supplemental EIS was completed because it con-
cluded that the modified flight schedule was essential to
supply military equipment and personnel for Operation Desert
Storm, an emergency response to Iraq’s sudden invasion of
Kuwait that same month. 

The court agreed with
the determination of CEQ and the Air Force that Iraq’s inva-
sion of Kuwait reasonably constituted an emergency “given
the military’s operational and scheduling difficulties and the
hostile and unpredictable nature of the Persian Gulf region.”

   Other cases sustaining CEQ’s application of 40 C.F.R.
§ 1506.11 also support the district court’s narrow interpreta-
tion of the phrase “emergency circumstances.” Courts have

scribed prospectively through January 2009, go beyond those “actions
necessary to control the immediate impacts of the emergency,” even if
there were an emergency. Because we hold that there is no basis for
reversing the district court’s determination that there were no “emergency
circumstances” under 40 C.F.R. § 1506.11, and therefore no basis for
CEQ’s order, we do not reach the merits of this argument.
2126                      NRDC v. WINTER
routinely given deference to CEQ’s finding of “emergency
circumstances” in situations where that finding has been used
to “avert imminent crises outside the agency’s control.” 

see, e.g., Nat’l Audobon Society v. Hester, 

,
405-7 (D.C. Cir. 1986) (giving deference where immediate
action was necessary to prevent the extinction of the Califor-
nia condor); Miccousukee Tribe of Indians of Fla. v. United
States, 

, 1290-91 (S.D. Fla. 2007) (giving
deference where immediate deviation from a water delivery
method was necessary to avoid pending extinction of the
Cape Sable seaside sparrow in the Everglades); NRDC v.
Pena, 

, 50 (D.D.C. 1998) (giving deference
where immediate action was necessary to secure storage of
nuclear materials); Crosby v. Young, 

, 1380,
1386 (E.D. Mich. 1981) (where an immediate response was
required for a city to meet a federal funding deadline for a
development project to counteract potentially dire economic
effects of the closure of a General Motors plant). In looking
to prior decisions, the district court certainly did not apply an
erroneous legal principle. Rather, it found legal authority that
supported its view.

        (4)   CEQ’s Broad Reading of “Emergency
                Circumstances” and NEPA

   The district court also held that CEQ’s broad reading of
“emergency circumstances” here is ultra vires because it sub-
verts NEPA’s directive that agencies perform their NEPA
duties “to the fullest extent possible.”44 See 42 U.S.C. § 4332.
The existence of specific Congressional exemptions to NEPA
informed the district court’s decision not to read the regula-
tion “so broadly as to independently authorize CEQ to do the
same, in the absence of a legitimate ‘emergency.’ ” Feb. 4,
2008 Dist. Ct. Order at 21. Moreover, the court noted that
  44
    Moreover, NEPA also requires agencies to use “all practicable means
and measures” to fulfill their duties under it. 42 U.S.C. § 4331(b).
                           NRDC v. WINTER                            2127
many of the exemptions granted in other cases involved agen-
cies faced with conflicting Congressional mandates. 

   NEPA, the statute authorizing 40 C.F.R. § 1506.11,
requires federal agencies to prepare an EIS for a major federal
action “significantly affecting the quality of the human envi-
ronment,” 42 U.S.C. § 4332(2)(C), or, in the alternative, to
implement mitigation measures to minimize impacts to the
point where an EIS in not required. See, e.g., Nat’l Parks &
Conservation Ass’n v. Babbitt, 

, 734 (9th Cir.
2001). Although CEQ justified its approval of “alternative
arrangements” to satisfy NEPA as grounded in “urgent
national security” concerns, see Jan. 15, 2008 CEQ Letter at
4, the district court noted that, in the absence of a bona fide
emergency, the “alternative arrangements” “operate[ ] to
exempt [the Navy] from the ususal rigors involved in the
preparation of an EIS, which forms the ‘heart’ of NEPA.”
Feb. 4, 2008 Dist. Ct. Order at 21 n.14 (quoting Envtl. Def.
Fund, Inc. v. Andrus, 

, 1374-5 (10th Cir.
1980)).

  Reviewing CEQ’s justification in this light, the district
court concluded that CEQ’s broad reading of “emergency cir-
cumstances” has the effect of reading a “national security” or
“defense” exemption into NEPA, where none exists.45 As the
  45
    Moreover, the district court’s conclusion was grounded in the well-
established fact that there is no “national defense” exception to NEPA. See
San Luis Obispo Mothers for Peace v. Nuclear Regulatory Comm’n, 

, 1035 (9th Cir. 2006); No GWEN Alliance of Lane County, Inc.
v. Aldridge, 

, 1384 (9th Cir. 1988). “ ‘The Navy, just like
any federal agency, must carry out its NEPA mandate to the fullest extent
possible and this mandate includes weighing the environmental costs of
the [project] even though the project has serious security implications.’ ”
San Luis 

(quoting No 

).
   Indeed, Congress has included exemptions for “paramount” national
security concerns in many environmental statutes, but not in NEPA. See,
e.g., Toxic Substances Control Act, 15 U.S.C. § 2621 (compliance waived
if the President determines a requested waiver to be necessary “in the
2128                       NRDC v. WINTER
district court pointed out, Congress knows well how to
exempt planned Defense Department activities from the
requirements of NEPA. See, e.g., Nat’l Defense Auth. Act,
Pub. L. No. 106-398, § 317, 114 Stat. 1654, 1654A-57 (2000)
(specifically exempting Defense Department from preparing
nationwide EIS for low-level flight training). The fact that
Congress has not so exempted the Navy’s exercises in the
Southern California Operating Area further supports the dis-
trict court’s conclusion that 40 C.F.R. § 1506.11 should not be
read to exempt the routine SOCAL exercises from NEPA’s
requirements.

   The district court’s interpretation also comports with well-
established Supreme Court precedent that narrowly interprets
NEPA’s requirement that agencies comply with its provisions
“to the fullest extent possible.” The Supreme Court has made
clear that the “to the fullest extent possible” language was
intended to address only cases in which there is an “irrecon-
cilable and fundamental conflict” between NEPA’s require-

interest of national defense”); Coastal Zone Management Act, 16 U.S.C.
§ 1456(c)(1)(b) (under certain circumstances the President may exempt an
activity that is in the “paramount interest of the United States”); Endan-
gered Species Act, 16 U.S.C. § 1536(j) (exemption granted if the Secre-
tary of Defense finds such exemption necessary “for reasons of national
security”); Clean Water Act, 33 U.S.C. § 1323(a) (the President may
exempt federal effluent source for up to one year if in the “paramount
interest of the United States”); Safe Drinking Water Act, 42 U.S.C. § 300j-
6(a) (the President may exempt federal facility for up to one year if in the
“paramount interest of the United States”); Resource Conservation and
Recovery Act, 42 U.S.C. § 6961(a) (the President may exempt federal
solid waste management facility for up to one year if in the “paramount
interest of the United States”); Clean Air Act, 42 U.S.C. § 7417(b) (the
President may exempt federal emission source for up to one year if in the
“paramount interest of the United States”); Comprehensive Environmental
Response, Compensation and Liability Act, 42 U.S.C. § 9620(j) (the Presi-
dent may issue orders to exempt facilities of the Department of Energy
and the Department of Defense “as may be necessary to protect the
national security interests of the United States . . .”). See Oct. 1, 2007
Brief of Amicus Curiae Law Professors Hope Babcock, et al. at 16 & n.4.
                      NRDC v. WINTER                     2129
ments and the requirements of another statute. See Flint Ridge
Dev. Co. v. Scenic Rivers Ass’n. of OK, 

, 787-88
(1976). Here, as the district court noted, the Navy has never
contended that it could not reconcile the district court’s
injunction with the requirements of NEPA.

   Similarly, NEPA regulations interpret the language “to the
fullest extent possible” to mean that “each agency of the Fed-
eral Government shall comply with that section unless exist-
ing law . . . expressly prohibits or makes compliance
impossible.” 40 C.F.R. § 1500.6. The legislative history of
§ 1500.6 explains that this language “shall not be used by any
Federal agency as a means of avoiding compliance with
[NEPA’s] directives . . . .” 115 Cong. Rec. (Part 29) 39702-
39703 (1969); see also Calvert Cliffs’ Coordinating Comm.
Inc. v. U.S. Atomic Energy Comm’n, 

, 1114
(D.C. Cir. 1971) (“We must stress as forcefully as possible
that this language does not provide an escape hatch for foot-
dragging agencies; it does not make NEPA’s procedural
requirements somehow ‘discretionary.’ . . . Indeed, [the lan-
guage] sets a high standard for the agencies, a standard which
must be rigorously enforced by the reviewing courts.”).

   The Navy asserts that national policy requires that it must
be confident that its west coast-based strike groups are pre-
pared and certified for deployment to hostile areas overseas
during a time of war. However, as the district court noted,
nothing prevented the Navy from preparing an EIS prior to
commencing the SOCAL exercises; indeed, the fact that the
Navy is currently developing an EIS for exercises in the
Southern California Operating Area confirms that it is fully
capable of meeting NEPA’s requirements. See Notice of
Intent To Prepare an Environmental Impact Statement/
Overseas Environmental Impact Statement for the Southern
California Range Complex, 71 Fed. Reg. 76,639 (Dec. 21,
2006).
2130                       NRDC v. WINTER
   Although the Navy argues that “NEPA must give way” so
that it may proceed with its training and certification unhin-
dered by environmental rules, quoting Flint 

, Flint Ridge itself holds that NEPA’s procedural
requirements are not discretionary and do not give way unless
a “clear and unavoidable conflict in statutory authority
exists,” id, here, the district court carefully examined the
record, with which it has longstanding familiarity, and deter-
mined that there was no such conflict in statutory authority,
concluding that conditioning phrases like “consistent with
other essential considerations of national policy,” 42 U.S.C.
§4331(b), and “to the fullest extent possible,” 

not indicate Congressional intent to create a statutory escape
hatch. Nor does any intent appear in the implementing regula-
tions, that would allow the Navy to conduct its exercises
before completing an EIS.46 Feb. 4, 2008 Dist. Ct. Order at
22-23.

   In reaching these conclusions, the district court examined
the various legal rules and applied those that were relevant to
this proceeding. Having done so, it acted well within its dis-
cretion in determining that CEQ’s broad interpretation of
“emergency circumstances” is contrary to the dictates of
NEPA.

                 (5)   Additional Considerations

   We also note that NRDC has raised a serious question as
to whether CEQ acted arbitrarily and capriciously in the pro-
cedure it used to reach its “emergency circumstances” deter-
mination pursuant to 40 C.F.R. § 1506.11. The Navy, arguing
  46
    The Navy argues, in the alternative, that CEQ’s action comports with
NEPA because it actually gives effect to the district court’s conclusion
that an EIS is likely required and provides “alternative arrangements” as
a bridge until the agency completes this EIS. Because we have concluded
that the district court did not abuse its discretion in determining that no
emergency existed and that CEQ had no authority to issue its order, we
need not reach this argument.
                           NRDC v. WINTER                           2131
that the district court’s order created an emergency by com-
promising its ability to effectively train and certify its strike
groups, requested alternative arrangements on January 10,
2008, and submitted evidence supporting that request the fol-
lowing day. Jan. 15, 2008 CEQ Letter at 1. CEQ approved the
Navy’s request four days later, on January 15, 2008. 

intervening time, CEQ held discussions with, and received
briefings from, the Navy and NMFS and reviewed the Navy’s
supporting documents. 

At no point did CEQ request,
nor did the Navy provide it, any of the evidence in the district
court record contrary to the Navy’s position that the chal-
lenged mitigation measures would compromise its ability to
train and certify its strike groups. Thus, CEQ reached its
“emergency circumstances” determination without consider-
ing any of the substantial evidence on which the district court
relied in concluding that the mitigation measures it imposed
would not render the Navy unable to train and certify its strike
groups. Where, as here, the basis for an emergency is alleged
to be the effect of a district court order, entered after careful
review of a full record submitted by both parties, a substantial
question exists as to whether CEQ acted arbitrarily and capri-
ciously when it failed to review the full record, and instead
considered only one side’s views, and on that basis deter-
mined that the court’s order gave rise to “emergency circum-
stances.”47
   47
      CEQ’s action raises a serious question not only under the APA, but
also under the Constitution. The separation of powers doctrine prevents
Congress from vesting review of the decisions of Article III courts in the
Executive Branch. See Plaut v. Spendthrift Farm, Inc., 

, 218-
19 (1995) (explaining that Article III “gives the Federal Judiciary the
power, not merely to rule on cases, but to decide them, subject to review
only by superior courts in the Article III hierarchy”); see also Hayburn’s
Case, 2 U.S. (2 Dall.) 409, 410 (1792). Here, the Navy represented, and
CEQ determined, that “emergency circumstances” existed because the dis-
trict court’s preliminary injunction prevented the Navy from effectively
training and certifying its strike groups for deployment. In making this
determination, CEQ presumably reviewed the same evidence that the
Navy presented to the district court (without, as noted above, the benefit
of NRDC’s evidence) and concluded, despite the district court’s explicit
2132                       NRDC v. WINTER
                           (6)   Conclusion

   [7] For the foregoing reasons, we hold that the district court
did not abuse its discretion or rely on an erroneous legal
premise in determining that CEQ’s broad interpretation of
“emergency circumstances” was not authorized by 40 C.F.R.
§ 1506.11 because it was contrary to the plain meaning of the
regulation and contrary to NEPA and, accordingly, that the
Navy remains subject to the traditional requirements of
NEPA.




factual finding to the contrary, that the imposed mitigation measures
would compromise the Navy’s ability to train and certify its forces. We
find substantial merit in NRDC’s argument that even if the district court’s
factual findings with respect to the effect of its mitigation measures were
erroneous, it was the job of the appellate court—and not the Executive
Branch—to so conclude. However, because the district court declined to
reach this question, we, too, do not consider the constitutional argument
in determining that the district court did not rely on an erroneous legal
premise or abuse its discretion when it held that CEQ’s action was invalid.
NRDC v. WINTER            2133
                 Volume 2 of 2
2134                  NRDC v. WINTER



                2.   NRDC’s NEPA Claim

  We next address the district court’s conclusion that NRDC
has shown probable success on the merits of its claim that the
Navy violated NEPA by failing to prepare an EIS for the
SOCAL exercises.

   In our November 13, 2007 order we concluded that “Plain-
tiffs have shown a strong likelihood of success on the merits
of their claims under [NEPA].” 

.
While that conclusion was based on our review of the record
underlying the district court’s August 7, 2007 preliminary
injunction order, the only subsequent developments are
CEQ’s approval of “alternative arrangements” pursuant to 40
C.F.R. § 1506.11 and the Navy’s concession, by virtue of
seeking such approval, that the SOCAL exercises will have a
“significant environmental impact.” See 40 C.F.R. § 1506.11
(“Where emergency circumstances make it necessary to take
an action with significant environmental impact without
observing the provisions of these regulations, the Federal
agency taking the action should consult with the Council
about alternative arrangements.”) (emphasis added). Although
we elaborate on our reasons, our original conclusion remains
unchanged.
                       NRDC v. WINTER                       2135
                 a.   Statutory Background

   As discussed earlier, NEPA requires a federal agency such
as the Navy to prepare a detailed EIS for all “major Federal
actions significantly affecting the quality of the human envi-
ronment.” 42 U.S.C. § 4332(2)(C). However, if, as here, an
agency’s regulations do not categorically require the prepara-
tion of an EIS, then the agency must first prepare an EA to
determine whether the action will have a significant effect on
the environment. Nat’l Parks & Conservation Ass’n v. Bab-
bitt, 

, 730 (9th Cir. 2001); see 40 C.F.R.
§ 1501.4. If the action will significantly affect the environ-
ment, an EIS must be prepared, while if the project will have
only an insignificant effect, the agency issues a Finding of No
Significant Impact (FONSI). Ocean Advocates v. U.S. Army
Corps of Eng’rs, 

, 864 (9th Cir. 2005); see 40
C.F.R. §§ 1501.3, 1501.4.

   [8] “An EIS must be prepared ‘if substantial questions are
raised as to whether a project . . . may cause significant degra-
dation of some human environmental factor.’ ” Blue Moun-
tains Biodiversity Project v. Blackwood, 

, 1212
(9th Cir. 1998) (quoting Idaho Sporting Congress v. Thomas,

, 1149 (9th Cir. 1998)). Thus, a plaintiff need
not show that significant effects on the environment will in
fact occur; raising “substantial questions whether a project
may have a significant effect” on the environment is enough.
Id.; Idaho 

.

   NEPA’s procedural requirements mandate that an agency
take a “hard look” at the environmental consequences of its
actions. Earth Island 

. NEPA is unique in
that it does not direct or require any particular substantive
action on the part of an agency. Its sole purpose is to require
that the agency be fully informed as to the environmental con-
sequences of its actions, the mitigation measures available,
and the alternatives to its proposed action. Once fully
informed, the agency may make its own final rule or decision.
2136                   NRDC v. WINTER
However, an agency may not avoid preparing an EIS by mak-
ing conclusory assertions that an activity will have only an
insignificant impact on the environment. Ocean 

. If an agency opts not to prepare an EIS, it
must put forth a “convincing statement of reasons” to explain
why a project’s impacts are insignificant. Blue 

(quoting Save the Yaak Comm. v. Block, 

, 717 (9th Cir. 1988)).

   [9] NEPA challenges are reviewed under the APA, which
provides that an agency action may be set aside if it is “arbi-
trary, capricious, an abuse of discretion, or otherwise not in
accordance with law[.]” 5 U.S.C. § 706(2)(A). The agency’s
decisions are “entitled to a presumption of regularity. But that
presumption is not to shield [the agency’s] action from a thor-
ough, probing, in-depth review.” Citizens to Preserve Overton
Park, Inc. v. Volpe, 

, 415 (1971) (citations omit-
ted). Thus, in reviewing an agency’s decision not to prepare
an EIS, a court must “determine whether the agency has taken
a ‘hard look’ at the consequences of its actions, ‘based [its
decision] on a consideration of the relevant factors,’ and pro-
vided a ‘convincing statement of reasons to explain why a
project’s impacts are insignificant.’ ” Native Ecosystems
Council v. U.S. Forest Serv., 

, 1239 (9th Cir.
2005) (quoting Nat’l Parks & Conservation 

).

   Agencies have wide discretion in assessing scientific evi-
dence, but they “must ‘take a hard look at the issues and
respond[ ] to reasonable opposing viewpoints.’ ” Earth Island

(internal quotation omitted, brackets in
original). “ ‘When specialists express conflicting views, an
agency must have discretion to rely on the reasonable opin-
ions of its own experts, even if a court may find contrary
views more persuasive. At the same time, courts must inde-
pendently review the record in order to satisfy themselves that
the agency has made a reasoned decision based on its evalua-
tion of the evidence.’ ” 

v. Or. Nat. Res.
                      NRDC v. WINTER                     2137
Council, 

, 378 (1989)). “If an agency has failed
to make a reasoned decision based on an evaluation of the
evidence, we may properly conclude that an agency has acted
arbitrarily and capriciously.” 

Questions about the Environmental
                 Impact of the Exercises

   The district court found that NRDC had raised substantial
questions as to whether the SOCAL exercises would have a
significant impact on the environment. Jan. 3, 2008 Dist. Ct.
Order at 6-7. Accordingly, the court concluded that NRDC
had demonstrated probable success on the merits of its claim
that the Navy’s failure to prepare an EIS was arbitrary and
capricious and in violation of NEPA and the APA. 

The district court did not rely on an erroneous legal premise
or abuse its discretion in so concluding.

   [10] Initially, we repeat our observation that the Navy, by
seeking approval by CEQ of “alternative arrangements” pur-
suant to 40 C.F.R. § 1506.11, has effectively conceded that
the SOCAL exercises will have a significant impact on the
environment. See 40 C.F.R. § 1506.11. As the text of
§ 1506.11 indicates, the very purpose of the regulation is to
provide for the possibility of alternative arrangements where
emergency circumstances require the taking of an action
“with significant environmental impact” without observing
the requirements of NEPA. See 

that the Navy
sought relief under § 1506.11 is evidence that the Navy recog-
nizes that the SOCAL exercises have a “significant environ-
mental impact.”

   Moreover, the fact that “[t]he Navy is currently evaluating
the environmental impact of MFA sonar training exercises
through its development of the SOCAL Range Complex Envi-
ronmental Impact Statement,” Jan. 15, 2008 CEQ Letter at 2,
confirms that, at the very least, the Navy acknowledges that
substantial questions have been raised as to whether the
2138                       NRDC v. WINTER
SOCAL exercises will have a significant impact on the envi-
ronment. Accordingly, were we not to review the Navy’s EA,
we would have little difficulty concluding that the district
court did not rely on an erroneous legal premise or abuse its
discretion in determining that NRDC has demonstrated proba-
ble success on the merits of its NEPA claim. Our own review
of the EA leads us to the same conclusion.48

   The Navy argues that the district court made a clearly erro-
neous assessment of the evidence of the effect of MFA sonar
on marine mammals in the waters of southern California. Spe-
cifically, the Navy asserts that the court misunderstood the
significance of the EA’s reference to the 548 predicted
instances of Level A harassments of beaked whales. The
Navy explains that it categorized predicted sonar exposures to
beaked whales as Level A harassments not because beaked
whales were expected to suffer such harassments but rather
because such categorization would allow the Navy to analyze
the potential impacts of MFA sonar on each beaked whale
species in greater detail than it would otherwise.

   We find no abuse of discretion in the district court’s rejec-
tion of the Navy’s argument. The Navy’s explanation for its
categorization of predicted sonar exposures to beaked whales
as Level A harassments finds no support in the EA, and coun-
sel for the Navy was unable to explain at oral argument on
November 8, 2007, how classifying exposures as Level A
harassments would allow the Navy to better analyze the
impacts of MFA sonar on various species of beaked whales.
Rather, the EA makes clear that the Navy categorized the
expected exposures to beaked whales as Level A harassments
because it concluded, in light of recent beaked whale strand-
ings linked to the use of MFA sonar, that beaked whales may
  48
     The Navy does not reargue the merits of NRDC’s NEPA claim in its
current appeal brief. We therefore assume that the Navy’s current position
is the same as set forth in its brief filed with our court on September 14,
2007.
                           NRDC v. WINTER                            2139
be more sensitive to sonar exposures than other cetaceans.
Because the lack of data precluded the Navy from establish-
ing separate impact thresholds for beaked whales, the Navy
decided that for beaked whales it would categorize exposures
as Level A harassments that would constitute Level B harass-
ments for other cetaceans.

   The Navy’s decision to do so is supported by ample evi-
dence indicating that beaked whales are particularly vulnera-
ble to MFA sonar. The documented strandings of marine
mammals that have been linked to the use of MFA sonar
overwhelmingly involve beaked whales.49 Indeed, according
to the International Whaling Commission the evidence associ-
ating the use of MFA sonar with beaked whale strandings is
“very convincing” and “appears overwhelming.” Similarly, a
Navy-sponsored study found “completely convincing” the
evidence that MFA sonar had caused strandings of beaked
whales.

   Moreover, to the extent that a paucity of scientific data pre-
vents the Navy from establishing meaningful impact thresh-
olds for beaked whales, that is a reason to conduct further
research and prepare an EIS—not a reason to ignore the data
that does exist and proceed with the SOCAL exercises with-
  49
     We reject the Navy’s argument that the district court erred by relying
on evidence not included in the Navy’s administrative record in reaching
its conclusion that NRDC had demonstrated probable success on the mer-
its. While generally a district court’s review under the APA is limited to
the administrative record before the agency, it may consider evidence
beyond the administrative record in certain situations. See, e.g., Ranchers
Cattlemen Action v. USDA, 

, 1117 (9th Cir. 2007). The dis-
trict court properly considered extra-record evidence here because the
Navy never submitted an administrative record to the district court despite
having had almost a year to do so since NRDC filed its complaint on
March 22, 2007. The Navy has not offered any valid explanation for why
it failed to submit an administrative record. We note that the Navy filed
a fourteen-volume record in the related litigation concerning the 2006
RIMPAC exercise only two days after NRDC filed its complaint in that
case.
2140                      NRDC v. WINTER
out adequate mitigation measures. As we explained in
National Parks:

       An agency must generally prepare an EIS if the envi-
       ronmental effects of a proposed agency action are
       highly uncertain. Preparation of an EIS is mandated
       where uncertainty may be resolved by further collec-
       tion of data or where the collection of such data may
       prevent speculation on potential . . . effects. The pur-
       pose of an EIS is to obviate the need for speculation
       by insuring that available data are gathered and ana-
       lyzed prior to the implementation of the proposed
       action.

Nat’l Parks & Conservation 

(internal
citations and quotations omitted).

   The Navy also contends that the district court misunder-
stood the significance of the nearly 170,000 predicted Level
B harassments. First, the Navy argues that this number is an
overestimate resulting from conservative assumptions in its
model and from the fact that it does not take into account the
mitigation measures the Navy would employ. However, the
Navy stated in the EA that its prediction of the harm to marine
mammals was “consistent with the best available science.”
And while the Navy “assumed” that its prediction was an
overestimate, it acknowledged that the science was “incom-
plete,” which apparently precluded the Navy from even
approximating by how much it had overestimated the harm.
Likewise, the Navy made no attempt to approximate how
many Level B harassments its mitigation measures would pre-
vent.50 Accordingly, we, like the Navy, must rely on the esti-
mate of nearly 170,000 predicted Level B harassments.
  50
    The Navy has estimated that in the 2006 RIMPAC exercise, the
employed mitigation measures prevented approximately 100 marine mam-
mals from being exposed to MFA sonar. Even ignoring the fact that the
mitigation measures employed in the 2006 RIMPAC exercise were more
                          NRDC v. WINTER                           2141
   Second, the Navy argues that most of the predicted Level
B harassments are below the threshold for causing temporary
hearing loss and will induce only temporary behavioral
responses which can be as minor as causing an animal to
avoid the noise source. But according to the Navy’s own defi-
nition of Level B harassment, those temporary behavioral
responses are nevertheless profound, as they cause “disruption
of natural behavioral patterns . . . to a point where such
behaviors are abandoned or significantly altered.”51 As the
NMFS’ Biological Opinion makes clear, such disruption of
natural behavioral patterns can be lethal for marine mammals.

   While the EA also states that it is “highly unlikely” that
Level B harassments would cause disturbance to a point
where behavioral patterns are abandoned or significantly
altered, the EA provides no support for that statement and
fails to explain why those harassments are nevertheless classi-
fied as Level B under the EA’s own definition. The district
court did not abuse its discretion in determining that such a
conclusory statement does not dispense with the requirement
of preparing an EIS. See Ocean 

(“[An agency] cannot avoid preparing an EIS by making con-
clusory assertions that an activity will have only an insignifi-
cant impact on the environment.”).

  Next, the Navy argues that it was “entitled to rely” on the

stringent than those the Navy has agreed to employ in the SOCAL exer-
cises, that estimate would suggest that in the fourteen SOCAL exercises
the mitigation measures would prevent only 1,400 exposures to marine
mammals (14 exercises x 100 prevented exposures). Accordingly, the esti-
mate of 170,000 Level B harassments would hardly be diminished by the
Navy’s current mitigation measures.
   51
      The Navy derived this definition from the MMPA, 16 U.S.C.
§ 1362(18)(B)(ii), which was amended in 2003 to exclude from the defini-
tion those acts that disrupted natural behavioral patterns but not to the
point where the behaviors were abandoned or significantly altered. See
H.R. Rep. No. 99(I), 108th Cong. 1 Sess. 2003 (5/14/03).
2142                      NRDC v. WINTER
conclusion of the NMFS in its Biological Opinion, which the
district court held satisfied the NMFS’ statutory obligation,
see 16 U.S.C. § 1536(a)(2), that the SOCAL exercises were
not likely to jeopardize the continued existence of threatened
or endangered marine mammal species. But the NMFS opined
only on the effects of MFA sonar on six species of marine mam-
mal,52 which do not include several rare or uncommon species
of marine mammal that are expected to be exposed to a signif-
icant number of sonar harassments.53 For example, the EA
predicts 436 Level A harassments of Cuvier’s beaked whales.
According to NOAA, as few as 1,121 Cuvier’s beaked whales
may exist in California, Oregon and Washington combined.
Likewise, the EA predicts 1,092 Level B harassments of bot-
tlenose dolphins, of which only 5,271 may exist in the Cali-
fornia Coastal and Offshore stocks.

   The Navy suggests that the test is whether the continued
existence of marine mammal species is jeopardized. This is
wrong. An agency action can have “significant effects” on the
environment short of threatened extinction. NEPA regulations
promulgated by CEQ provide that “significantly” has two
components: context and intensity. 40 C.F.R. § 1508.27. Con-
text refers to the setting in which the proposed action takes
place, in this case the Southern California Operating Area.
See 

Intensity means “the severity of impact.”

In considering the severity of the potential
  52
      Those six species are the fin whale, humpback whale, sei whale,
sperm whale, and Guadalupe fur seal.
   53
      While in a recent memorandum—written well after the Navy prepared
its EA—the NMFS concluded that the Navy’s own mitigation measures
“will minimize the likelihood of beaked whales being caught in circum-
stances that characterize known strandings of beaked whales,” Jan. 9,
2008 Memorandum from NMFS to the Under Secretary of Commerce for
Oceans and Atmosphere at 3, the NMFS did not conclude that the Navy’s
measures would prevent direct physical injury (such as tissue damage) to
beaked whales. Indeed, the NMFS acknowledged that “the mechanism by
which MFA sonar appears to be injurious to beaked whales is poorly
understood.” 

                       NRDC v. WINTER                      2143
environmental impact, a reviewing agency may consider up to
ten factors to help inform the “significance” of a project,
including the degree to which the effects on the quality of the
human environment are likely to be “highly controversial”
and the degree to which the possible effects on the human
environment are “highly uncertain or involve unique or
unknown risks.” 

(b)(5). We have held
that “one of these factors may be sufficient to require prepara-
tion of an EIS in appropriate circumstances.” Ocean Advo-

; see Ctr. for Biological Diversity v.
Nat’l Highway Traffic Safety Admin., 

, 553 (9th
Cir. 2007); Nat’l Parks & Conservation 

.

   Notably, whether an agency action will jeopardize the con-
tinued existence of an endangered or threatened species is not
among these factors. See 40 C.F.R. § 1508.27(b). However,
the degree to which the action may “adversely affect” an
endangered or threatened species is. See 

While the NMFS’ Biological Opinion concluded that the
SOCAL exercises were not likely to jeopardize the continued
existence of the six endangered species it studied, it neverthe-
less acknowledged that the exercises “may adversely affect”
those species. Thus, while the Navy was not required to disre-
gard the NMFS’ “no jeopardy” opinion, see Envtl Prot. Info.
Ctr. v. U.S. Forest Serv., 

, 1012 (9th Cir. 2006),
the Biological Opinion by its own terms makes clear that the
SOCAL exercises may “significantly” affect the environment,
see Greater Yellowstone Coalition v. Flowers, 

,
1275-76 (10th Cir. 2004) (regarding as not determinative for
NEPA purposes the Fish and Wildlife Service’s “no jeopardy”
opinion as to bald eagles).

   The Navy further argues that its finding of no significant
impact was not arbitrary and capricious because no sonar-
inflicted injuries have been observed in the Southern Califor-
nia Operating Area in almost forty years of MFA sonar use
by the Navy. But as explained above, that fact has limited
2144                   NRDC v. WINTER
probative value in establishing whether marine mammals will
in fact be harmed by the Navy’s use of MFA sonar. Exposure
to MFA sonar may physically and behaviorally harm marine
mammals even if it does not cause them to strand. And absent
stranding, such harm is difficult to observe. That is particu-
larly true for the beaked whale—the marine mammal most
vulnerable to MFA sonar—in which a population decline of
50% over 15 years would go undetected as a decline at all in
90% of the beaked whale stocks.

   Finally, the Navy claims that the correlation between the
past marine mammal strandings and MFA sonar is irrelevant
because the combination of environmental conditions “found
at the locations of documented marine mammal stranding
incidents” is not present in the Southern California Operating
Area. But contrary to this claim, the EA indicates that the
Navy has studied only the stranding in the Bahamas in 2000
and not any of the other stranding events.

   Moreover, while the Navy’s study of the Bahamas strand-
ing identified a combination of factors that contributed to the
whales’ injury—the presence of a strong surface duct, unusual
water bathymetry, intensive use of multiple sonar units over
an extended period of time, a constricted channel with limited
egress, and the presence of beaked whales that appear to be
sensitive to the frequencies produced by these sonars—the EA
provided no supporting data or analysis for its conclusion that
this combination of factors does not exist in the Southern Cal-
ifornia Operating Area. Thus, we find no abuse of discretion
in the district court’s conclusion that the Navy has failed to
provide the support for its conclusion that NEPA requires. See
40 C.F.R. § 1502.24.

  In addition, the Navy’s Bahamas study acknowledged that
combinations of factors different from the one present in the
Bahamas “may be more or less likely to cause strandings”
(emphasis added). Thus, even if the combination of factors
present in the Bahamas in fact does not exist in the Southern
                       NRDC v. WINTER                     2145
California Operating Area, the combination of factors that
does exist may be even more likely to cause injury to marine
mammals. The EA does not explore that possibility.

   [11] In sum, the district court did not abuse its discretion
in concluding that NRDC raised substantial questions as to
whether the SOCAL exercises would have a significant
impact on the environment. All of the reasons stated in the EA
for why the Navy believed the SOCAL exercises would not
have the deleterious effect that the Navy’s own model pre-
dicted were cursory, unsupported by cited evidence, or uncon-
vincing. Thus, we find ample support for the district court’s
conclusion that the Navy has not “articulate[d] a rational con-
nection between the facts found and the conclusion[ ]
reached.” Earth Island 

(quoting Mid-
water Trawlers Co-op v. Envtl. Def. Ctr., 

, 716
(9th Cir. 2002)).

           c.   The Navy’s Mitigation Measures

   The district court also concluded that NRDC had demon-
strated probable success on the merits of its claim that the
Navy’s mitigation measures were inadequate to obviate the
need for preparing an EIS. Again, we find no reliance on an
erroneous legal premise and no abuse of discretion in the dis-
trict court’s conclusion.

   [12] The Navy correctly points out that “[a]n agency’s
decision to forego issuing an EIS may be justified in some cir-
cumstances by the adoption of [mitigation] measures” and
that those measures, if significant, “need not completely com-
pensate for adverse environmental impacts.” Nat’l Parks &
Conservation 

(citations and internal
quotation marks omitted). However, we have also held that a
“perfunctory description” or “mere listing of mitigation mea-
sures, without supporting analytical data,” is insufficient to
support a finding of no significant impact. Okanogan High-
lands Alliance v. Williams, 

, 473 (9th Cir. 2000)
2146                      NRDC v. WINTER
(citations and internal quotation marks omitted). We find no
reversible error in the district court’s conclusion that the
Navy’s list of proposed mitigation measures was precisely
such a perfunctory description devoid of supporting data.

   [13] The explanation contained in the EA as to why the
mitigation measures are effective is contained in four short
bullet points, stating that whales and dolphins spend extended
periods of time on the surface, have relatively short dive peri-
ods, tend to move in large groups (pods), and frequently come
to the surface and have a high level of activity there. Three of
those bullet points in effect state the same thing, namely that
whales and dolphins spend little time under water. This expla-
nation is inadequate for several reasons.

   [14] First, the Navy’s explanation overlooks the fact that
beaked whales spend much of their time under water, surface
infrequently, and are generally difficult to detect. A study by
NMFS scientists observed that “beaked whales are always dif-
ficult to see when they are on the surface, spend most of their
time below the surface, and are found at low densities over
large areas.” Likewise, NRDC submitted a declaration by a
biologist who opines that visual monitoring by ship-based
lookouts would result in the detection of only 2% of beaked
whales in the Southern California Operating Area, in part
because of the speed at which Navy vessels travel. Declara-
tion of Dr. Robin William Baird ¶ 6.

   [15] Second, the Navy’s explanation fails to address the
effectiveness of the Navy’s safety zones—the only measure
that directly reduces exposure of marine mammals to MFA
sonar. Specifically, the EA fails to explain why a safety zone
of only 1,000 yards is adequate, why reducing the sonar level
by 6dB and 10dB at the 1,000-yard and 500-yard marks,
respectively, is adequate, and why it is effective to halt MFA
sonar transmission altogether only at the 200-yard mark.54 The
  54
    As stated earlier, the Navy has recently agreed to adopt safety zones
with radii of 1,000, 500 and 200 meters instead of 1,000, 500 and 200
                           NRDC v. WINTER                            2147
Navy’s explanation also does not relate to the effectiveness of
the measure requiring passive sonar to be used to detect
sounds made by marine mammals.

  While the Navy claims in the EA that it is “very likely” that
lookouts would detect a group of common dolphins because
of “frequent surfacing” and group sizes of “over a thousand
animals,” it notably makes no such claims about Cuvier’s
beaked whales, Mesoplodont beaked whales or Ziphiid
beaked whales. Indeed, the EA acknowledged that Cuvier’s
beaked whales and Mesoplodont beaked whales exhibit a
range of dives lasting up to 87 minutes.

   Moreover, while the EA claims that it is “very likely” that
Baird’s beaked whales will be detected by lookouts, it states
nothing about the frequency with which those whales surface.
While it may be that, as the EA states, beaked whales are
large in size and travel in groups of between nine and thirteen
animals, those facts hardly prove the effectiveness of visual
surveillance measures considering that beaked whales gener-
ally come to the surface infrequently.

   We find further support for the district court’s conclusion
that the Navy’s mitigation measures did not obviate the need
to prepare an EIS in the fact that, as explained above, the
Navy refused to adopt several of the more aggressive mitiga-
tion measures recommended by the CCC, employed in the
2006 RIMPAC exercise, or imposed by the Department of

yards. However, the radius of the Navy’s outer safety zone is still half of
the radius recommended by the CCC, resulting in a safety zone that is
75% smaller than the one found necessary by the CCC.
   Notably, NRDC has submitted declarations by scientists who state that
sonar levels even below the Navy’s lowest impact threshold of 173 dB
may be fatal and that sonar sound can travel up to hundreds of miles under
water, which suggests that the Navy’s significantly smaller safety zones
are inadequate. See Parsons decl. ¶ 13, Weilgart decl. ¶ 10.
2148                       NRDC v. WINTER
Defense for non-RIMPAC exercises in 2006. Specifically, the
Navy refused to:

       •   Expand the outer safety zone to 2 kilometers;55

       •   Reduce sonar power in conditions of low visibil-
           ity or strong surface ducting;

       •   Avoid training in areas known to have high con-
           centrations of marine mammals or during the
           gray whale’s migratory season;

       •   Monitor for 30 minutes prior to commencement
           of MFA sonar use;

       •   Restrict operation of MFA sonar within 25 kilo-
           meters from the 200-meter isobath;

       •   Restrict operation of MFA sonar within 12 nauti-
           cal miles from the coast;56 and

       •   Restrict operation of MFA sonar in choke points,
           constricted channels, or canyon-like areas.57
  55
      The Navy did not discuss this measure in its EA, let alone explain why
it would not be effective. In its “after action report” following the 2006
RIMPAC exercise, the Navy disposed of this mitigation measure simply
by declaring it “not prudent” and “without scientific merit.”
   56
      The Navy asserts on appeal that the SOCAL exercises will occur more
than twelve nautical miles from the mainland coastline, but concedes that
the exercises will occur within twelve nautical miles of Clemente Island,
which falls within the Southern California Operating Area. In its January
3, 2008 preliminary injunction order the district court required the Navy
to refrain from using MFA sonar within 12 nautical miles from the Cali-
fornia coastline. Jan. 3, 2008 Dist. Ct. Order at 14.
   57
      The parties disagree as to whether there are any choke points in the
Southern California Operating Area. In its January 3, 2008 preliminary
injunction order the district court found that the Catalina Basin poses the
same concerns as a choke point: ingress and egress to the basin are
restricted and the area has a high density of marine mammals. Jan. 3, 2008
Dist. Ct. Order at 17. Accordingly, the district court ordered the Navy to
refrain from using MFA sonar in the Catalina Basin. 

                           NRDC v. WINTER                            2149
   Notably, as to most of these measures the Navy does not
contest that they would be effective. While the Navy claims
that some of the measures would adversely affect its ability to
achieve the objectives of the exercises, that does not render
the measures the Navy has adopted adequate to avoid the
need for preparing an EIS. Indeed, the Navy states in its “after
action report” following the first three SOCAL exercises that
in future exercises it intends to incorporate data collection
necessary to address the question of how many marine mam-
mals not observed by lookouts may have been exposed to
dangerous sonar levels, and will integrate additional monitor-
ing tools and techniques. While the Navy’s intent is com-
mendable, it implicitly acknowledges that its current
mitigation and data collection efforts are less than adequate.

   [16] We conclude that the district court did not abuse its
discretion in determining that the Navy’s cursory explanation
in the EA for why its mitigation measures are effective does
not demonstrate that those measures “constitute an adequate
buffer against the negative impacts” that may result from the
SOCAL exercises. See Nat’l Parks & Conservation 

. Accordingly, we uphold the district court’s
conclusion that the Navy’s reliance on its incomplete mitiga-
tion plan in deciding not to prepare an EIS was likely arbi-
trary and capricious and affirm its determination that NRDC
has demonstrated probable success on the merits of its NEPA
claims. Cf. Wetlands Action Network v. U.S. Army Corps of
Eng’rs, 

, 1112 (9th Cir. 2000).58
  58
     The district court also concluded that NRDC had demonstrated proba-
ble success on the merits of its claims that the Navy violated NEPA by
preparing an EA that failed to adequately consider reasonable alternatives
to its proposed action, see, e.g., Bob Marshall Alliance v. Hodel, 

, 1228 (9th Cir. 1998) (explaining that under NEPA federal agencies
must sufficiently study, develop, and describe alternatives as part of the
“environmental decisionmaking process”), as well as the cumulative
impacts of its actions, see, e.g., Klamath-Siskiyou Wildlands Ctr. v.
Bureau of Land Mgmt., 

, 994 (9th Cir. 2004) (explaining that
an EA must provide a “useful analysis of the cumulative impacts of past,
2150                        NRDC v. WINTER
             B.    Possibility of Irreparable Injury

   In our November 13, 2007 order we affirmed the district
court’s conclusion that NRDC had met its burden of demon-
strating the possibility of irreparable injury. 

. We now explain that decision.

   [17] As the district court noted, “[w]here, as here, plaintiffs
demonstrate a likelihood of prevailing on the merits of their
claims, injunctive relief is appropriate where there is a ‘possi-
bility of irreparable harm.’ ” Jan. 3, 2008 Dist. Ct. Order at 12
(quoting Faith Ctr. Church Evangelistic Ministries v. Glover,

, 906 (9th Cir. 2007)). NRDC must show the
possibility of irreparable harm to its membership. See Friends
of the Earth, Inc. v. Laidlaw Environmental Services (TOC),
Inc., 

, 180-81 (2000). After analyzing the “nu-
merous scientific studies, declarations, reports, and other evi-
dence submitted,” the district court concluded that NRDC had
established “to a near certainty” that use of MFA sonar in the
SOCAL exercises will cause irreparable harm to the environ-
ment and to NRDC’s membership. Jan. 3, 2008 Dist. Ct.
Order at 12.

   [18] The Navy argues that the harm resulting to the envi-
ronment from the use of MFA sonar in the SOCAL exercises
is merely “speculative.” But the Navy’s own EA proves other-
wise. The EA estimates that the use of MFA sonar in the
SOCAL exercises will result in 564 instances of physical
injury including permanent hearing loss (Level A harassment)
and nearly 170,000 behavioral disturbances (Level B harass-
ment), more than 8,000 of which would also involve tempo-

present, and future projects”). Jan. 3, 2008 Dist. Ct. Order at 8-10. We
decline to address this aspect of the district court’s order as we have
already concluded that the court did not abuse its discretion in finding that
NRDC has demonstrated probable success on the merits of its other NEPA
claims.
                       NRDC v. WINTER                       2151
rary hearing loss. As explained above, while the Navy
protests that these figures are overestimates resulting from its
conservative approach, the EA makes clear that the figures are
“consistent with the best available science.” Indeed, the
Navy’s failure to suggest by how much its figures overesti-
mate the actual harm to marine mammals confirms that the
EA’s figures are the best available estimates. Those estimates,
in turn, contradict the Navy’s suggestion that the harm caused
by MFA sonar in the SOCAL exercises is merely speculative.

   Moreover, while the record contains no evidence that
marine mammals have been harmed by the use of MFA sonar
in the Southern California Operating Area, the scientific con-
sensus that MFA sonar may cause injury and death to marine
mammals combined with the evidence that such injury, absent
a stranding, is difficult to detect—especially in the case of the
vulnerable beaked whale—further disproves the suggestion
that the harm caused by MFA sonar in the SOCAL exercises
is merely speculative.

   The Navy also argues that its use of MFA sonar will cause
only temporary harm to marine mammals and therefore will
not result in irreparable injury. But the EA again undermines
the Navy’s argument. The EA’s estimate that 564 instances of
Level A harassment will occur demonstrates that the use of
MFA sonar will also cause permanent harm to marine mam-
mals. Likewise, the EA’s own definition of Level B harass-
ment indicates that the nearly 170,000 estimated instances of
such harassment may result in the outright abandonment of
important behaviors by marine mammals.

   Finally, the Navy argues that NRDC failed to meet its bur-
den of demonstrating that marine mammals “will suffer irrep-
arable injury at the species or stock-level.” For two reasons,
the Navy has not shown that the district court relied on an
erroneous legal premise or abused its discretion in rejecting
this argument.
2152                   NRDC v. WINTER
   First, NRDC had only the burden of demonstrating the
“possibility of irreparable injury,” Freecycle 

, not that irreparable injury “will” necessarily
occur. See Earth Island 

(holding that the
district court erred in requiring that the plaintiff show a “sig-
nificant threat of irreparable injury” because that standard
imposes a higher burden of proof than the “mere possibility
of irreparable harm” standard).

   Second, the Navy has cited no support for the proposition
that NRDC was required to demonstrate the possibility of
irreparable injury at the species or stock-level. While the
Navy relies on Water Keeper Alliance v. Dep’t of Defense,

(1st Cir. 2001), the court in that case only con-
cluded that the district court did not abuse its discretion in
concluding that the “death of even a single member of an
endangered species” would be an insufficient basis for the
plaintiffs to demonstrate the possibility of irreparable injury.
See 

Here, the district court found that, according to
the Navy’s EA, the SOCAL exercises “will cause widespread
harm to nearly thirty species of marine mammals, including
five species of endangered species, and may cause permanent
injury and death.” January 3, 2008 Dist. Ct. Order at 12.

   In any event, even if NRDC were required to demonstrate
the possibility of irreparable injury at the species or stock-
level, it would have satisfied that requirement here. As dis-
cussed earlier, the EA predicts 436 Level A harassments of
Cuvier’s beaked whales, of which, according to NOAA, as
few as 1,121 may exist in California, Oregon and Washington
combined. Similarly, the EA predicts 1,092 Level B harass-
ments of bottlenose dolphins, of which only 5,271 may exist
in the California Coastal and Offshore stocks.

   [19] Accordingly, we hold that the district court did not
rely on an erroneous legal premise or abuse its discretion in
concluding that NRDC has demonstrated the possibility of
irreparable injury.
                             NRDC v. WINTER                                2153
                      C.    Balance of Hardships

  [20] Having determined that the district court neither relied
on erroneous legal premises nor abused its discretion in deter-
mining that NRDC has demonstrated a strong likelihood of
success on the merits of its NEPA claim and the possibility
of irreparable injury, we turn to the “balance of hardships.”
See Freecycle 

. The Supreme Court
has held that environmental injury, in addition to often being
permanent or of long duration, can seldom, by its nature, “be
adequately remedied by money damages”; therefore, “[i]f
such injury is sufficiently likely, . . . the balance of harms will
usually favor the issuance of an injunction to protect the envi-
ronment.” Amoco Prod. Co. v. Vill. of Gambell, AK, 

, 545 (1987).

   The Navy maintains, and the district court did not contest,
that its ability to train and certify its west-coast strike groups
for combat deployment is critical.59 The Navy argues that its
  59
     For example, Captain Martin M. May states that “[m]odern, quiet sub-
marines . . . pose the primary threat to the littorals, control of strategic
maritime choke points, transit through international straits, and protection
of sea lines of communication vital to international commerce.” May decl.
¶ 19. These submarines are “nearly undetectable to U.S. and allied naval
forces without the use of [MFA] sonar.” 

20. Captain May also
asserts that “[s]onar operators and crews must train regularly and fre-
quently to develop the skills necessary to master the art and processes of
identifying submarines in the complex subsurface environment[,]” and
that computer simulations are not adequate for teaching these skills. 

22. Captain May maintains that if sonar use is enjoined, the Navy would
be unable to gauge a fleet’s ability to use active sonar. 

26.
   Captain May also contends that “[t]he ranges that comprise the South-
ern California Operating Area make up a unique area in which strike
groups can meet all required training objectives at the same time. 

25. Another significant factor is that many of the operating areas have
been surveyed and closely mirror the prospective operating environments
in many of the world’s ‘hot’ spots where U.S. Naval forces may be
required to fight.” 

our own littorals . . . also build[s] profi-
ciency and experience in our own waters should the enemy attempt to
interdict U.S. forces deploying to the area of conflict.” 

v. WINTER
ability to engage in these critical activities is severely
degraded by: (1) the requirement that it shut down its use of
MFA sonar when a marine mammal is detected within 2,200
yards of a sonar-emitting source; and (2) the requirement that
it power down its sonar use by 6 dB (75%) when significant
surface ducting conditions are detected. Without proper train-
ing and certification, the Navy asserts that it will not be able
to deploy a sufficient number of combat-ready forces to pro-
vide for the national defense.

   The district court rejected the Navy’s argument that the bal-
ance of hardships tipped in its favor. It concluded that
although “the imposition of these mitigation measures will
require the Navy to alter and adapt the way it conducts anti-
submarine warfare training—a substantial challenge,” the
measures would not preclude the Navy from effectively train-
ing and certifying forces for deployment to combat zones in
the western Pacific and the Middle East.60 Feb. 4, 2008 Dist.
Ct. Order.

   The gravity of the Navy’s asserted hardship requires that
we review the district court’s determination, as well as the
affidavits submitted by the Navy, with the utmost care. We
have done so here and, for the reasons set forth below, we
conclude that the district court did not abuse its discretion in
reaching its determination.

  We note that any negative impact on the Navy’s ability to
successfully conduct its exercises under the challenged miti-
  60
     The district court crafted its mitigation order after carefully weighing
evidence submitted by the parties over a period of “longstanding involve-
ment” with the matters at issue. It took into account the Navy’s need for
training in certain bathymetry and under certain conditions, and declined
to impose several of plaintiffs’ proposed “sweeping geographic exclu-
sions” which would have precluded the Navy from training with MFA
sonar within 25 nautical miles of the coast, in waters shallower than 1,500
meters to the maximum extent possible, and in the Westfall seamount and
the Cortez and Tanner Banks. Jan. 3, 2008 Dist. Ct. Order at 13 n.6.
                            NRDC v. WINTER                               2155
gation measures is necessarily speculative because the Navy
has never before employed these measures in the context of
MFA sonar training.61 The speculative nature of the Navy’s
asserted harm is reflected in its own characterization of the
hardship; the Navy does not claim that the challenged mea-
sures will categorically preclude effective training and certifi-
cation, but rather that they will give rise to an “unacceptable
risk” of such a result. Our task, then, is to determine, as best
we are able from the record before us, whether the district
court erred by giving insufficient weight to the Navy’s
asserted hardship or by resting its conclusions on clearly erro-
neous factual findings.

   In support of its assessment that the challenged measures
will significantly impair its training exercises, the Navy offers
the declarations of various high-ranking officers. For exam-
ple, Admiral John Locklear explains that the 2,200 yard safety
zone will in “[his] opinion . . . have crippling implications on
Navy’s ability to conduct realistic pre-deployment [ASW]
training employing MFA sonar” and “will significantly
impact ASW training.” Locklear decl. ¶¶ 9, 13. The Chief of
Naval Operations refers to an unacceptable risk to strike
group certification posed by both the safety zone and surface
ducting measures. The judgment of these naval officers and,
in particular, that of the Chief of Naval Operations, who is
charged with the statutory responsibility under 10 U.S.C.
   61
      The Navy has not represented that, if the challenged measures are
upheld, it will cease its training exercises. Further, the Navy acknowl-
edged at oral argument on February 27, 2008, that it can certify strike
groups despite the inability to train in surface ducting conditions. We
therefore proceed under the assumption that the exercises will continue to
take place, thereby preserving the possibility of successful training and
certification of strike groups. Thus, the district court did not err in failing
to weigh the hardship to the Navy, and the public, that would result if the
Navy stopped training altogether. Rather, the district court did not abuse
its discretion and properly considered the hardship that would result if the
Navy were required to abide by the challenged mitigation measures during
its remaining exercises in 2008 and 2009.
2156                       NRDC v. WINTER
§ 5062 for organizing, training, and equipping the Navy, is
entitled to substantial deference. See Khalsa v. Weinberger,

, 1400 n.4 (9th Cir. 1985) (“The degree of def-
erence due to factual assertions by the military is proportion-
ate to the need for the application of military experience,
judgment, and expertise in evaluating the assertion.”).

   Nevertheless, a court’s deference is not absolute, even
when a government agency claims a national security interest.
See, e.g., Campbell v. U.S. Dep’t of Justice, 

, 30
(D.C. Cir. 1998) (“[D]eference is not equivalent to acquies-
cence . . . .”). The district court therefore did not abuse its dis-
cretion when it considered the Navy’s declarations along with
the evidence contained in the record as a whole. This evi-
dence, much of it submitted by the Navy itself, supports the
district court’s conclusion that the challenged mitigation mea-
sures will not likely compromise the Navy’s ability to effec-
tively train and certify its west-coast strike groups. We
address the evidence with respect to each of the challenged
mitigation measures in turn.

             1.    The 2,200 Yard Shutdown Zone

   We first consider the requirement that the Navy shut down
its use of MFA sonar when a marine mammal is detected
within 2,200 yards of a sonar-emitting source.62 Upon a care-
ful review of the record, we find no clear error in the district
court’s factual finding that this measure would not compro-
mise the Navy’s ability to train and certify its strike groups.
  62
     Although our 

addresses the evidence contained in the
record with respect to the likely impact that the increase in the safety
zone’s size will have on training and certification, the Navy also contests
this measure’s application to all sonar sources, including helicopter dip-
ping sonar and sonobuoys which emit less powerful sonar waves. See, e.g.,
Locklear decl. ¶ 11. Despite the fact that the dipping sonar and sonobuoys
have lower energy source levels, they still operate at levels above those
shown to pose a danger to marine mammals. See, e.g., Parsons decl. ¶ 13.
Accordingly, the district court’s decision to include these sonar sources in
its mitigation measure was not an abuse of discretion.
                            NRDC v. WINTER                             2157
   [21] The “after action reports” compiled by the Navy fol-
lowing eight prior COMPTUEX and JTFEX exercises in the
Southern California Operating Area undermine the Navy’s
assessment of the significance of the hardship that this mitiga-
tion measure would impose.63 The reports contain several rel-
evant data points. First, the data reveal a relatively low
likelihood that a marine mammal will be sighted during a time
when the Navy’s MFA sonar is in use, thereby triggering
implementation of this mitigation measure. During two of the
eight exercises, no marine mammals were sighted while MFA
sonar was in use. In the remaining exercises, sightings rarely
occurred while MFA sonar was in use.64 In fact, over the
course of eight exercises lasting one or more weeks each, the
Navy observed marine mammals only 51 times while using
MFA sonar, which represents less than 15% of all marine
mammal observations (345) during those exercises. Second,
the data show that the Navy shut down its MFA sonar twenty-
seven times over the course of its eight prior exercises. If the
district court’s mitigation measure had been in place, the
Navy would have had to shut down at most an additional
twenty-one times: an increase of only two to three shutdowns
per exercise.65 Moreover, of the twenty-one additional times
  63
    During these exercises, the Navy’s preferred 1,000 yard safety zone
was implemented requiring a 6 dB power-down when a marine mammal
was detected within 500-1,000 yards, an additional 4 dB power-down
within 200 and 500 yards, and a mandatory shutdown within 200 yards.
   We note that the first four of the eight exercises for which the record
contains “after action reports” were not part of the current SOCAL exer-
cises.
   64
      The low likelihood of a sighting occurring while MFA sonar is being
used does not mean it is rare for marine mammals to be exposed to dan-
gerous noise levels. Rather, two facts demonstrate that the number of
sightings does not equate to the number of mammals affected by an MFA
sonar event: (1) the presence of marine mammals is difficult to detect by
“sight” because many spend significant amounts of time submerged
underwater; and (2) sound travels long distances in water (in some cases
up to many hundreds of miles) creating the potential for adverse effects
beyond the range of sight.
   65
      The Navy argues that this mitigation measure will result in a five-fold
increase in the number of times it is required to shut down during training
2158                        NRDC v. WINTER
that the Navy would have been required to shut down, in
eleven of these instances, the Navy powered down its sonar.
During these power-downs, the Navy’s detection capability
was significantly reduced. See, e.g., Bird decl. ¶ 49. Accord-
ingly, if the district court’s injunction had been in place, the
Navy would have had to compromise detection capability
only approximately one more time per exercise. Third, the
“after action reports” do not establish any serious, negative
effects on operational impact from the smaller safety zone
imposed in the eight exercises.66 Because the broader safety
zone likely will not require significantly more shutdowns, the

exercises. The Navy arrives at this higher number by excluding from its
calculation shutdowns that occurred beyond the 200 yard mandatory shut-
down zone. But this calculation fails to account for the Navy’s actual prac-
tice of shutting down sonar in a significant number of cases where marine
mammals were detected beyond 200 yards, including at distances as far as
3,100, 4,000, and 6,000 yards. In fact, the “after action reports” reveal that
of the twenty-seven times the Navy shut down MFA sonar, it did so fifteen
times when the observed marine mammals were outside the 200 yard man-
datory shutdown zone and four times when the marine mammals were at
an “unknown” distance.
   There is no dispute that the Navy continued to certify its strike groups
throughout these exercises. The Navy defends its exclusion of any non-
mandatory shutdowns in its count by claiming that any shutdowns that
occurred beyond the 200 yard range “likely occurred during tactically
insignificant times.” Locklear decl. ¶ 11. The record fails to support this
claim. The “after action reports” do not distinguish between shutdown
events in evaluating training impacts. Indeed, the very same language is
used to describe the loss of detection opportunities during all of the exer-
cises, without regard to whether the shutdowns occurred within or beyond
200 yards.
   66
      Indeed, the “after action reports” contain only one instance in which
the Navy comments that its operation was actually affected by a safety
zone mitigation measure. See After Action Report for COMPTUEX 07-01
(19 November-19 December 2006) at 6. All the other reports state that the
impact of the safety zone measure was “not determinable in the reactions
of the particular units” and only speculate that “the proximity of a subma-
rine in the vicinity meant there was a potential submarine detection oppor-
tunity missed by the exercise participants.”
                            NRDC v. WINTER                              2159
district court’s conclusion that this mitigation measure does
not pose a significant risk of rendering the Navy’s training
exercises ineffective is well-supported. We cannot, on this
record, find that the district court abused its discretion.

   In addition to data drawn from “after action reports,” the
record contains other evidence of the feasibility of mandatory
shutdown zones of this size, and even greater, during naval
training exercises. Indeed, the size of the district court’s
imposed shutdown zone was based on the CCC’s proposed
mitigation measure, which in turn was drawn from the Navy’s
own imposition of a 2,000 meter shutdown requirement when
it uses low-frequency active sonar.67 While the record does
not indicate whether low-frequency active sonar has the same
effect on marine mammals as MFA sonar, the Navy offers no
explanation as to why a 2,000 meter safety zone is feasible
during its operation of low-frequency active sonar but not
during its use of MFA sonar.68

  67
    While we recognize that each Navy has unique operating require-
ments, the record shows that NATO imposes a 2,000 meter shutdown zone
when a marine mammal is detected—the same zone that the district
court’s preliminary injunction requires. The Australian Navy goes farther,
mandating a shutdown of activities if a marine mammal is detected within
4,000 yards of a sonar-emitting vessel.
  68
    To the contrary, in its initial submissions to the court, the Navy repre-
sented that “[p]rior to [the district court’s] requirement, the maximum
mandatory shutdown zone the Navy ever employed was 200 meters.” Jan.
15 Emergency Motion at 16 (emphasis added). This representation is
plainly contradicted by the record. In its most recent brief to the court, the
Navy clarified that, “Prior to this requirement, the maximum mandatory
shutdown zone ever employed for MFA sonar was 200 meters.” However,
the Navy still does not discuss nor defend its ability to implement a 2,000
meter safety zone in its low-frequency, but not in its mid-frequency, sonar
activities.
2160                   NRDC v. WINTER
   2.   The Requirement to Power-Down in Significant
               Surface Ducting Conditions

   [22] We next consider the requirement that the Navy
power-down its sonar use by 6 dB when significant surface
ducting conditions are detected. Although the Navy stresses
the importance of training in surface ducting conditions, it
admits—and the record confirms—that such conditions occur
relatively rarely in the southern California waters in which the
Navy has chosen to conduct its exercises. Indeed, the “after
action reports” from JTFEX and COMPTUEX exercises con-
ducted in the Southern California Operating Area in 2006
show that significant surface ducting conditions were not
detected during any of those exercises. Yet despite the strike
groups’ inability to train under such conditions, the Navy cer-
tified them. The record, then, undermines the Navy’s conten-
tion that the district court’s mitigation measure regarding
surface ducting conditions will significantly impact its ability
to certify strike groups.

   [23] Thus, although the actual effect of the challenged miti-
gation measures on the exercises at issue is necessarily specu-
lative, data from past Navy exercises and practices supports
the district court’s conclusion that the imposition of these
measures is not likely to prevent effective training and certifi-
cation of strike groups. After a thorough review of this record,
we are not “left with the definite and firm conviction that a
mistake has been committed.” Sports Form, 

. To the contrary, there is significant evidence of the
Navy’s ability to successfully train and certify its strike
groups under the conditions imposed by the district court. We
therefore conclude that the district court did not abuse its dis-
cretion in ordering the Navy to comply with the challenged
mitigation measures.

                        3.   Balancing

  As explained earlier, the scientific studies, declarations and
reports in the record confirm the district court’s determination
                            NRDC v. WINTER                              2161
that irreparable harm to marine mammals will almost cer-
tainly result should the Navy be permitted to conduct its
remaining exercises without appropriate mitigation measures.
See, e.g., Bain decl. ¶ 14 (explaining that “the monitoring and
mitigation adopted by the Navy is insufficient to detect, much
less prevent, marine mammal injury and mortality”); Declara-
tion of Dr. Thomas A. Jefferson decl. ¶ 4 (describing the link
between military sonar and the stranding and deaths of beaked
whales and other cetaceans). As the district court observed,
the 2,200 yard shutdown zone might protect marine mammals
from only “the harshest of sonar-related consequences.” Jan.
3, 2008 Dist. Ct. Order at 15.

   Further, as the district court noted, the exercises in southern
California are only a subset of the Navy’s training activities
involving active sonar.69 

NRDC submitted evi-
dence that the Navy uses active sonar in hundreds of exercises
each year throughout the world. The evidence linking several
whale strandings to the Navy’s use of active sonar in training
exercises around the world further confirms that the Navy
trains in its use of active sonar in many different areas.

   [24] While we are mindful of the importance of protecting
national security, courts have often held, in the face of asser-
tions of potential harm to military readiness, that the armed
forces must take precautionary measures to comply with the
law during its training. See, e.g., NRDC v. Evans, 

, 1143 (N.D. Cal. 2003) (“A tailored injunc-
tion reconciles the very compelling interests on both sides of
  69
    Indeed, the EA shows that the Navy considered as alternatives con-
ducting the exercises in other locations, including Alaska and Hawaii,
reducing the number of exercises, and using exercise simulation. While
the Navy’s EA provides reasonably detailed justifications for why the
Southern California Operating Area is uniquely suited to these exercises,
and demonstrates that the Navy would suffer a certain hardship if the con-
sidered alternatives were employed instead, the EA nonetheless shows the
Navy is still able to conduct its exercises in alternate locations, in reduced
number, or through simulation.
2162                   NRDC v. WINTER
this case, by enabling the Navy to continue to train with and
test [low-frequency active] sonar as it needs to do, while tak-
ing some additional measures to better protect against harm to
marine life.”); Makua v. Rumsfeld, 

, 1221
(D. Haw. 2001) (“Although the court recognizes the impor-
tance of national security and live-fire training, the potential
harm to the Army resulting from a brief preliminary injunc-
tion will not be significant.”). As in those cases, the district
court here carefully balanced the significant interests and
hardships at stake to ensure that the Navy could continue to
train without causing undue harm to the environment. We
review that balance to determine whether it rests on clearly
erroneous findings of fact. Having concluded that it does not,
we determine that the district court did not abuse its discretion
and therefore do not disturb its carefully considered injunc-
tion.

   We recognize that although the record indicates that the
Navy will be able to continue to train and certify strike groups
effectively despite the two challenged mitigation measures,
there remains the possibility that, when they are actually
implemented, it will be unable to do so. In light of the hard-
ship that the Navy and the public would suffer should the
imposed measures actually result in an inability to train and
certify sufficient naval forces to provide for the national
defense, we conclude that, in the unlikely event that such a
situation arises, the Navy may return to the district court to
request relief on an emergency basis.

         D.   Advancement of the Public Interest

   There are two dimensions to the public interest in this case.
The public has an interest both in national security and in pro-
tection of the marine environment. The public interest with
respect to national security is the same as that discussed in our
consideration of the hardship the Navy would suffer if it were
unable to effectively train and certify its strike groups. The
public interest with respect to protection of the marine envi-
                        NRDC v. WINTER                       2163
ronment is the same as that discussed in our consideration of
the irreparable injury NRDC would suffer if the SOCAL exer-
cises were carried out in the absence of appropriate mitigation
measures. As our discussion makes clear, we conclude that
the district court did not rely on an erroneous legal premise
or abuse its discretion in analyzing either of these interests.
Accordingly, there is no need for any additional discussion of
the public interest.

                       V.   Conclusion

   The district court concluded that plaintiffs have met the
necessary burden of proof to demonstrate that preliminary
injunctive relief is appropriate. It held that plaintiffs have
shown a strong likelihood of success on the merits, as well as
the possibility of irreparable injury if relief is not granted. It
also held that plaintiffs have shown that the balance of hard-
ships tips in their favor in light of the preliminary injunction’s
narrowly-tailored mitigation measures which provide that the
Navy’s SOCAL exercises may proceed as planned if con-
ducted under circumstances that provide satisfactory safe-
guards for the protection of the environment. Finally, it held
that the public interest is advanced by a preliminary injunc-
tion that imposes adequate mitigation measures. In reaching
these conclusions, the district court neither relied on errone-
ous legal premises nor abused its discretion. We therefore
affirm the district court’s preliminary injunction.

  AFFIRMED.