San Luis & Delta-Mendota Water v. Natural Resources Defense Coun

FOR PUBLICATION

UNITED STATES COURT OF APPEALS
FOR THE NINTH CIRCUIT

SAN LUIS & DELTA-MENDOTA No. 11-15871
WATER AUTHORITY; WESTLANDS
WATER DISTRICT; STEWART & D.C. No.
JASPER ORCHARDS; ARROYO FARMS, 1:09-cv-00407-
LLC; KING PISTACHIO GROVE; OWW-DLB
STATE WATER CONTRACTORS;
METROPOLITAN WATER DISTRICT OF
SOUTHERN CALIFORNIA; COALITION
FOR A SUSTAINABLE DELTA; KERN
COUNTY WATER AGENCY; FAMILY
FARM ALLIANCE,
Plaintiffs-Appellees,

CALIFORNIA DEPARTMENT OF
WATER RESOURCES,
Intervenor-Plaintiff-Appellee,

v.

SALLY JEWELL, as Secretary of the
Department of the Interior; U.S.
DEPARTMENT OF THE INTERIOR; U.S.
FISH & WILDLIFE SERVICE; DANIEL
M. ASHE, as Director of the U.S.
Fish and Wildlife Service; REN
LOHOEFENER, as Regional Director
of the U.S. Fish and Wildlife
Service, Pacific Southwest Region,
U.S. Department of the Interior;
2 SAN LUIS V. JEWELL

UNITED STATES BUREAU OF
RECLAMATION; MICHAEL L.
CONNOR, as Commissioner of the
U.S. Bureau of Reclamation, U.S.
Department of the Interior; DAVID
MURILLO, as Director of the U.S.
Bureau of Reclamation, Mid-Pacific
Region, U.S. Department of the
Interior; MARK COWIN, Director,
California Department of Water
Resources; UNITED STATES
DEPARTMENT OF JUSTICE; U.S.
ENVIRONMENTAL PROTECTION
AGENCY; GINA MCCARTHY, in her
official capacity as Administrator of
the Environmental Protection
Agency; U.S. DEPARTMENT OF
TRANSPORTATION; ANTHONY FOXX,
in his official capacity as Secretary
of Transportation; MARITIME
ADMINISTRATION; PAUL N.
JAENICHEN, SR., in his official
capacity as Acting Maritime
Administrator; U.S. DEPARTMENT OF
HOMELAND SECURITY; JEH
JOHNSON, in his official capacity as
Secretary of Homeland Security;
FEDERAL EMERGENCY
MANAGEMENT AGENCY; WILLIAM
CRAIG FUGATE, in his official
capacity as Administrator of the
Federal Emergency Management
Agency; UNITED STATES ARMY
 SAN LUIS V. JEWELL 3

CORPS OF ENGINEERS; THOMAS P.
BOSTICK, Commanding General and
Chief of Engineers, United States
Army Corps of Engineers,
Defendants,

and

NATURAL RESOURCES DEFENSE
COUNCIL; THE BAY INSTITUTE,
Intervenor-Defendants-Appellants.

SAN LUIS & DELTA-MENDOTA No. 11-16617
WATER AUTHORITY; WESTLANDS
WATER DISTRICT; STEWART & D.C. No.
JASPER ORCHARDS; ARROYO FARMS, 1:09-cv-00407-
LLC; KING PISTACHIO GROVE; OWW-DLB
METROPOLITAN WATER DISTRICT OF
SOUTHERN CALIFORNIA; COALITION
FOR A SUSTAINABLE DELTA; KERN
COUNTY WATER AGENCY; FAMILY
FARM ALLIANCE,
Plaintiffs,

CALIFORNIA DEPARTMENT OF
WATER RESOURCES,
Intervenor-Plaintiff,

and

STATE WATER CONTRACTORS,
Plaintiff-Appellant,
4 SAN LUIS V. JEWELL

v.

SALLY JEWELL, as Secretary of the
Department of the Interior; U.S. FISH
& WILDLIFE SERVICE; DANIEL M.
ASHE, as Director of the U.S. Fish
and Wildlife Service; REN
LOHOEFENER, as Regional Director
of the U.S. Fish and Wildlife
Service, Pacific Southwest Region,
U.S. Department of the Interior;
UNITED STATES BUREAU OF
RECLAMATION; MICHAEL L.
CONNOR, as Commissioner of the
U.S. Bureau of Reclamation, U.S.
Department of the Interior; DAVID
MURILLO, as Director of the U.S.
Bureau of Reclamation, Mid-Pacific
Region, U.S. Department of the
Interior; UNITED STATES
DEPARTMENT OF JUSTICE; U.S.
ENVIRONMENTAL PROTECTION
AGENCY; GINA MCCARTHY, in her
official capacity as Administrator of
the Environmental Protection
Agency; U.S. DEPARTMENT OF
TRANSPORTATION; ANTHONY FOXX,
in his official capacity as Secretary
of Transportation; MARITIME
ADMINISTRATION; PAUL N.
JAENICHEN, SR., in his official
capacity as Acting Deputy Maritime
Administrator; U.S. DEPARTMENT OF
 SAN LUIS V. JEWELL 5

HOMELAND SECURITY; JEH
JOHNSON, in his official capacity as
Secretary of Homeland Security;
FEDERAL EMERGENCY
MANAGEMENT AGENCY; WILLIAM
CRAIG FUGATE, in his official
capacity as Administrator of the
Federal Emergency Management
Agency; UNITED STATES ARMY
CORPS OF ENGINEERS; THOMAS P.
BOSTICK, Commanding General and
Chief of Engineers, United States
Army Corps of Engineers; U.S.
DEPARTMENT OF THE INTERIOR,
Defendants-Appellees,

NATURAL RESOURCES DEFENSE
COUNCIL; THE BAY INSTITUTE,
Intervenor-Defendants-Appellees.

SAN LUIS & DELTA-MENDOTA No. 11-16621
WATER AUTHORITY; WESTLANDS
WATER DISTRICT; STEWART & D.C. No.
JASPER ORCHARDS; ARROYO FARMS, 1:09-cv-00407-
LLC; KING PISTACHIO GROVE; OWW-DLB
COALITION FOR A SUSTAINABLE
DELTA; KERN COUNTY WATER
AGENCY; FAMILY FARM ALLIANCE;
STATE WATER CONTRACTORS,
Plaintiffs,
6 SAN LUIS V. JEWELL

CALIFORNIA DEPARTMENT OF
WATER RESOURCES,
Intervenor-Plaintiff,

and

METROPOLITAN WATER DISTRICT OF
SOUTHERN CALIFORNIA,
Plaintiff-Appellant,

v.

SALLY JEWELL, as Secretary of the
Department of the Interior; U.S.
DEPARTMENT OF THE INTERIOR; U.S.
FISH & WILDLIFE SERVICE; DANIEL
M. ASHE, as Director of the U.S.
Fish and Wildlife Service; REN
LOHOEFENER, as Regional Director
of the U.S. Fish and Wildlife
Service, Pacific Southwest Region,
U.S. Department of the Interior;
UNITED STATES BUREAU OF
RECLAMATION; MICHAEL L.
CONNOR, as Commissioner of the
U.S. Bureau of Reclamation, U.S.
Department of the Interior; DAVID
MURILLO, as Director of the U.S.
Bureau of Reclamation, Mid-Pacific
Region, U.S. Department of the
Interior; UNITED STATES
DEPARTMENT OF JUSTICE; U.S.
ENVIRONMENTAL PROTECTION
 SAN LUIS V. JEWELL 7

AGENCY; GINA MCCARTHY, in her
official capacity as Administrator of
the Environmental Protection
Agency; U.S. DEPARTMENT OF
TRANSPORTATION; ANTHONY FOXX,
in his official capacity as Secretary
of Transportation; MARITIME
ADMINISTRATION; PAUL N.
JAENICHEN, SR., in his official
capacity as Acting Deputy Maritime
Administrator; U.S. DEPARTMENT OF
HOMELAND SECURITY; JEH
JOHNSON, in his official capacity as
Secretary of Homeland Security;
FEDERAL EMERGENCY
MANAGEMENT AGENCY; WILLIAM
CRAIG FUGATE, in his official
capacity as Administrator of the
Federal Emergency Management
Agency; UNITED STATES ARMY
CORPS OF ENGINEERS; THOMAS P.
BOSTICK, Commanding General and
Chief of Engineers, United States
Army Corps of Engineers,
Defendants-Appellees,

NATURAL RESOURCES DEFENSE
COUNCIL; THE BAY INSTITUTE,
Intervenor-Defendants-Appellees.
8 SAN LUIS V. JEWELL

SAN LUIS & DELTA-MENDOTA No. 11-16623
WATER AUTHORITY; WESTLANDS
WATER DISTRICT; STEWART & D.C. No.
JASPER ORCHARDS; ARROYO FARMS, 1:09-cv-00407-
LLC; KING PISTACHIO GROVE; OWW-DLB
STATE WATER CONTRACTORS;
METROPOLITAN WATER DISTRICT OF
SOUTHERN CALIFORNIA; COALITION
FOR A SUSTAINABLE DELTA; KERN
COUNTY WATER AGENCY; FAMILY
FARM ALLIANCE,
Plaintiffs-Appellees,

CALIFORNIA DEPARTMENT OF
WATER RESOURCES,
Intervenor-Plaintiff-Appellee,

v.

SALLY JEWELL, as Secretary of the
Department of the Interior; U.S.
DEPARTMENT OF THE INTERIOR; U.S.
FISH & WILDLIFE SERVICE; DANIEL
M. ASHE, as Director of the U.S.
Fish and Wildlife Service; REN
LOHOEFENER, as Regional Director
of the U.S. Fish and Wildlife
Service, Pacific Southwest Region,
U.S. Department of the Interior;
UNITED STATES BUREAU OF
RECLAMATION; MICHAEL L.
CONNOR, as Commissioner of the
U.S. Bureau of Reclamation, U.S.
 SAN LUIS V. JEWELL 9

Department of the Interior; DAVID
MURILLO, as Director of the U.S.
Bureau of Reclamation, Mid-Pacific
Region, U.S. Department of the
Interior; UNITED STATES
DEPARTMENT OF JUSTICE; U.S.
ENVIRONMENTAL PROTECTION
AGENCY; GINA MCCARTHY, in her
official capacity as Administrator of
the Environmental Protection
Agency; U.S. DEPARTMENT OF
TRANSPORTATION; ANTHONY FOXX,
in his official capacity as Secretary
of Transportation; MARITIME
ADMINISTRATION; PAUL N.
JAENICHEN, SR., in his official
capacity as Acting Deputy Maritime
Administrator; U.S. DEPARTMENT OF
HOMELAND SECURITY; JEH
JOHNSON, in his official capacity as
Secretary of Homeland Security;
FEDERAL EMERGENCY
MANAGEMENT AGENCY; WILLIAM
CRAIG FUGATE, in his official
capacity as Administrator of the
Federal Emergency Management
Agency; UNITED STATES ARMY
CORPS OF ENGINEERS; THOMAS P.
BOSTICK, Commanding General and
Chief of Engineers, United States
Army Corps of Engineers,
Defendants-Appellants,
10 SAN LUIS V. JEWELL

and

NATURAL RESOURCES DEFENSE
COUNCIL; THE BAY INSTITUTE,
Intervenor-Defendants.

SAN LUIS & DELTA-MENDOTA No. 11-16624
WATER AUTHORITY; WESTLANDS
WATER DISTRICT, D.C. No.
Plaintiffs-Appellants, 1:09-cv-00407-
OWW-DLB
and

STEWART & JASPER ORCHARDS;
ARROYO FARMS, LLC; KING
PISTACHIO GROVE; COALITION FOR A
SUSTAINABLE DELTA; KERN
COUNTY WATER AGENCY; FAMILY
FARM ALLIANCE; STATE WATER
CONTRACTORS; METROPOLITAN
WATER DISTRICT OF SOUTHERN
CALIFORNIA,
Plaintiffs,

CALIFORNIA DEPARTMENT OF
WATER RESOURCES,
Intervenor-Plaintiff,

v.

SALLY JEWELL, as Secretary of the
Department of the Interior; U.S.
 SAN LUIS V. JEWELL 11

DEPARTMENT OF THE INTERIOR; U.S.
FISH & WILDLIFE SERVICE; DANIEL
M. ASHE, as Director of the U.S.
Fish and Wildlife Service; REN
LOHOEFENER, as Regional Director
of the U.S. Fish and Wildlife
Service, Pacific Southwest Region,
U.S. Department of the Interior;
UNITED STATES BUREAU OF
RECLAMATION; MICHAEL L.
CONNOR, as Commissioner of the
U.S. Bureau of Reclamation, U.S.
Department of the Interior; DAVID
MURILLO, as Director of the U.S.
Bureau of Reclamation, Mid-Pacific
Region, U.S. Department of the
Interior; UNITED STATES
DEPARTMENT OF JUSTICE; U.S.
ENVIRONMENTAL PROTECTION
AGENCY; GINA MCCARTHY, in her
official capacity as Administrator of
the Environmental Protection
Agency; U.S. DEPARTMENT OF
TRANSPORTATION; ANTHONY FOXX,
in his official capacity as Secretary
of Transportation; MARITIME
ADMINISTRATION; PAUL N.
JAENICHEN, SR., in his official
capacity as Acting Deputy Maritime
Administrator; U.S. DEPARTMENT OF
HOMELAND SECURITY; JEH
JOHNSON, in his official capacity as
Secretary of Homeland Security;
12 SAN LUIS V. JEWELL

FEDERAL EMERGENCY
MANAGEMENT AGENCY; WILLIAM
CRAIG FUGATE, in his official
capacity as Administrator of the
Federal Emergency Management
Agency; UNITED STATES ARMY
CORPS OF ENGINEERS; THOMAS P.
BOSTICK, Commanding General and
Chief of Engineers, United States
Army Corps of Engineers,
Defendants-Appellees,

NATURAL RESOURCES DEFENSE
COUNCIL; THE BAY INSTITUTE,
Intervenor-Defendants-Appellees.

STATE WATER CONTRACTORS, No. 11-16660
Plaintiff-Appellant,
D.C. No.
v. 1:09-cv-00422-
OWW-GSA
SALLY JEWELL, as Secretary of the
Department of the Interior; UNITED
STATES DEPARTMENT OF JUSTICE;
DANIEL M. ASHE, as Acting Director
of the U.S. Fish and Wildlife
Service; U.S. FISH & WILDLIFE
SERVICE; MARK COWIN, Director,
California Department of Water
Resources; CALIFORNIA
 SAN LUIS V. JEWELL 13

DEPARTMENT OF WATER
RESOURCES,
Defendants-Appellees.

METROPOLITAN WATER DISTRICT OF No. 11-16662
SOUTHERN CALIFORNIA,
Plaintiff-Appellant, D.C. No.
1:09-cv-00631-
v. OWW-DLB

U.S. FISH & WILDLIFE SERVICE;
SALLY JEWELL, Secretary of the OPINION
Department of the Interior; DANIEL
M. ASHE, Acting Director of the
U.S. Fish and Wildlife Service;
UNITED STATES BUREAU OF
RECLAMATION; J. WILLIAM
MCDONALD; CALIFORNIA
DEPARTMENT OF WATER
RESOURCES; MARK COWIN, Director,
California Department of Water
Resources,
Defendants-Appellees.

Appeal from the United States District Court
for the Eastern District of California
Oliver W. Wanger, Senior District Judge, Presiding

Argued and Submitted
September 10, 2012—Las Vegas, Nevada
14 SAN LUIS V. JEWELL

Filed March 13, 2014

Before: Morris S. Arnold,* Johnnie B. Rawlinson,
and Jay S. Bybee, Circuit Judges.

Opinion by Judge Bybee;
Partial Concurrence and Partial Dissent by Judge Arnold;
Partial Concurrence and Partial Dissent by Judge
Rawlinson

*
The Honorable Morris S. Arnold, Senior Circuit Judge for the U.S.
Court of Appeals for the Eighth Circuit, sitting by designation.
 SAN LUIS V. JEWELL 15

SUMMARY**

Environmental Law

The panel reversed in part and affirmed in part the district
court’s judgment invalidating a 2008 biological opinion by
the U.S. Fish and Wildlife Service that concluded that the
Central Valley and State Water Projects jeopardized the
continued existence of the delta smelt and its habitat.

The Central Valley Project and the State Water Project,
operated respectively by the U.S. Bureau of Reclamation and
the State of California, supply water originating in northern
California to agricultural and domestic consumers in central
and southern California. The source of the water—the
estuary at the confluence of the San Francisco Bay and the
Sacramento-San Joaquin Delta—is the lone habitat for the
delta smelt, a threatened species under the Endangered
Species Act (“ESA”). After the Bureau of Reclamation
requested a biological opinion (“BiOp”), the U.S. Fish and
Wildlife Service (“FWS”) concluded that the Central Valley
operations would threaten the delta smelt and, as required by
the ESA, proposed alternatives to ameliorate the effect on the
smelt, including reducing the water exported to southern
California. The plaintiffs-appellees—various water districts,
water contractors, and agricultural consumers—brought suit
under the Administrative Procedure Act against various
federal defendants. The district court concluded that the 2008
BiOp was arbitrary and capricious.

**
This summary constitutes no part of the opinion of the court. It has
been prepared by court staff for the convenience of the reader.
16 SAN LUIS V. JEWELL

Concerning the scope of the record, the panel held that the
district court overstepped its bounds in admitting additional
declarations from the parties’ experts. The panel held that it
would consider the BiOp and evidence submitted by the
parties that the FWS considered in making its decision, and
the testimony of the four experts the district court appointed
pursuant to Federal Rule of Evidence 706.

Concerning the merits, the panel held that the 2008
BiOp’s reliance on raw salvage figures to set the upper and
lower Old and Middle Rivers flow limits was not arbitrary
and capricious. The panel also held that the 2008 BiOp’s
determination of X2 (the point in the Bay-Delta at which the
salinity is less than two parts per thousand) was not arbitrary
and capricious. The panel further held that the BiOp’s
incidental take statement was not arbitrary and capricious
because it included adequate explanation and support for its
determinations. The panel also held the record supported the
BiOp’s conclusions regarding the indirect effects of project
operations. The panel disagreed with the district court’s
determination that the FWS’s own regulations and the
Administrative Procedure Act required the FWS to explain
that the reasonable and prudent alternatives satisfied 50
C.F.R. § 402.02’s non-jeopardy factors. The panel held that
the FWS’s consideration of these factors could be reasonably
discerned from the record to satisfy any explanation
requirements.

Concerning the cross appeal, the panel held that the FWS
did not violate the ESA by not separating the discretionary
from nondiscretionary actions when it set the environmental
baseline. The panel also held that the Bureau of Reclamation
did not violate the ESA by accepting the 2008 BiOp. The
panel affirmed the district court’s judgment with respect to
 SAN LUIS V. JEWELL 17

the National Environmental Policy Act (“NEPA”) claims, and
held: NEPA does not require the FWS to prepare an
Environmental Impact Statement in conjunction with the
issuance of the BiOp; and the Bureau of Reclamation’s
provisional adoption and implementation of the BiOp
triggered its obligation to comply with NEPA. The panel
affirmed the district court’s order remanding to the Bureau of
Reclamation so that it can complete an Environmental Impact
Statement evaluating the effects of its adoption and
implementation of the BiOp.

Eighth Circuit Judge Arnold dissented from Parts III,
IV.A., IV.B, IV.E, and V.B. of the majority opinion, and
concurred in the rest. Judge Arnold would uphold the district
court’s limited admission of evidence outside the
administrative record as relevant to the Old and Middle River
flow limits and determination of X2, and agreed with the
district court that the FWS’s determination as to the flow
prescription and X2 was arbitrary and capricious. Judge
Arnold disagreed with the basis of the district court’s
conclusion that the non-jeopardy elements must be addressed
in the BiOp or administrative record, but would affirm on the
issue. Finally, Judge Arnold believes the district court should
have found the Bureau of Reclamation independently liable
under the ESA for relying on a legally flawed BiOp.

Judge Rawlinson concurred in the bulk of the majority
opinion, but dissented from Part V.C.2. Judge Rawlinson
disagreed only with the rationale and conclusion that the
Bureau of Reclamation’s adoption and implementation of the
BiOp triggered its obligation to comply with NEPA by
preparing an Environmental Impact Statement that is
generally required under the ESA.
18 SAN LUIS V. JEWELL

COUNSEL

Robert H. Oakley (argued), Ethan Carson Eddy, and Charles
R. Scott, United States Department of Justice, Environment
& Natural Resources Division, Washington, D.C.; Ignacia S.
Moreno, United States Department of Justice, Assistant
Attorney General, Washington, D.C.; Jim Monroe,
Department of the Interior, Office of the Solicitor,
Sacramento, California, for Federal Defendants-Appellants.

Katherine Poole (argued) and Doug Obegi, Natural Resources
Defense Council, San Francisco, California; Trent W. Orr and
George M. Torgun, Earthjustice, San Francisco, California,
for Defendant-Intervenors-Appellants.

Daniel J. O’Hanlon (argued), Hanspeter Walter, and Rebecca
R. Akroyd, Kronick, Moskovitz, Tiedemann & Girard,
Sacramento, California; Craig Manson, General Counsel,
Westlands Water District, Fresno, California; Steven O. Sims,
Michelle C. Kales, and Geoffrey M. Williamson, Brownstein
Hyatt Farber Schreck LLP, Denver, Colorado, for Plaintiffs-
Appellees-Cross-Appellants San Luis & Delta-Mendota
Water Authority and Westlands Water District.

Gregory K. Wilkinson (argued), Steven M. Anderson,
Melissa R. Cushman, and Steven G. Martin, Best Best &
Krieger LLP, Riverside, California, for Plaintiffs-Appellees
State Water Contractors.

Marcia L. Scully, Interim General Counsel; Linus
Masouredis, Chief Deputy General Counsel, The
Metropolitan Water District of Southern California,
Sacramento, California; Christopher J. Carr, William N.
Sloan, and Travis Brandon, Morrison & Foerster LLP, San
 SAN LUIS V. JEWELL 19

Francisco, California, for Plaintiff-Appellee-Cross-Appellant
The Metropolitan Water District of Southern California.

Clifford T. Lee (argued), Deputy Attorney General; Kamala
D. Harris, Attorney General of California; Kathleen A.
Kenealy, Senior Assistant Attorney General; Robert W.
Byrne, Supervising Deputy Attorney General; Cecilia L.
Dennis, and Gary Alexander, Deputy Attorneys General, San
Francisco, California, for Plaintiff-Intervenor-Appellee,
California Department of Water Resources.
20 SAN LUIS V. JEWELL

TABLE OF CONTENTS

I. FACTS AND PROCEEDINGS BELOW. . . . . . . . . . . 26

A. Background. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

1. The Sacramento-San Joaquin Delta. . . . . . . . . 26

2. The delta smelt.. . . . . . . . . . . . . . . . . . . . . . . . 31

B. Proceedings Leading To The Present Controversy
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

1. The FWS’s 2008 Biological Opinion. . . . . . . 33

2. The present case.. . . . . . . . . . . . . . . . . . . . . . . 39

II. STANDARDS OF REVIEW. . . . . . . . . . . . . . . . . . . . 43

III. THE SCOPE OF THE RECORD. . . . . . . . . . . . . . . . 45

IV. MERITS OF THE BiOp. . . . . . . . . . . . . . . . . . . . . . . 50

A. The 2008 BiOp’s Reliance on Raw Salvage Figures to
Set the Upper and Lower OMR Flow Limits Was Not
Arbitrary and Capricious.. . . . . . . . . . . . . . . . . . . 53

1. The FWS’s choice of a more conservative model
to calculate flow limits in Figures B-13 and B-14
was supported by substantial evidence. . . . . . 56

2. The BiOp’s determination of OMR flow limits
was influenced by more than Figures B-13 and B-
14. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
 SAN LUIS V. JEWELL 21

3. The OMR flow limits exist as one part in a
dynamic monitoring system that accounts for the
smelt population as a whole.. . . . . . . . . . . . . . 71

B. The 2008 BiOp’s Determination of X2 Was Not
Arbitrary and Capricious.. . . . . . . . . . . . . . . . . . . 74

1. The FWS was not arbitrary and capricious in
comparing DAYFLOW to CALSIM II. . . . . . 75

2. The BiOp sufficiently explained the fall X2
locations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

C. The BiOp’s Incidental Take Statement Is Not Flawed
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

1. The ITS reasonably uses different data sets for
adult and juvenile take limits. . . . . . . . . . . . . . 93

2. The FWS reasonably uses an average cumulative
salvage index. . . . . . . . . . . . . . . . . . . . . . . . . . 94

D. The Record Supports the BiOp’s Conclusions
Regarding the Indirect Effects of Project Operations
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

1. Project operations indirectly affect smelt food
supply. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

2. Project operations indirectly affect the smelt
through water contamination. . . . . . . . . . . . . 101
22 SAN LUIS V. JEWELL

3. Project operations indirectly affect the smelt
through the “other stressors” of predation,
macrophytes, and microcystis. . . . . . . . . . . . 103
a. Predation. . . . . . . . . . . . . . . . . . . . . . . . . . 104
b. Aquatic Macrophytes. . . . . . . . . . . . . . . . 107
c. Microcystis. . . . . . . . . . . . . . . . . . . . . . . . 109

E. The FWS Is Not Required to Support the “Non-
Jeopardy” Elements of its RPA. . . . . . . . . . . . . . 110

V. CROSS-APPEAL.. . . . . . . . . . . . . . . . . . . . . . . . . . . 119

A. Segregating Discretionary From Nondiscretionary
Actions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

B. Reclamation Did Not Violate the ESA by Accepting
the 2008 BiOp. . . . . . . . . . . . . . . . . . . . . . . . . . . 124

C. Application of NEPA to the FWS and Reclamation
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

1. Application of NEPA to the FWS. . . . . . . . . 127

2. Application of NEPA to Reclamation. . . . . . 133

VI. CONCLUSION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

GLOSSARY OF TERMS. . . . . . . . . . . . . . . . . . . . . . . . 154

PARTIAL DISSENT BY JUDGE ARNOLD. . . . . . . . . 155

PARTIAL DISSENT BY JUDGE RAWLINSON. . . . . 161
 SAN LUIS V. JEWELL 23

OPINION

BYBEE, Circuit Judge, with whom ARNOLD, Circuit Judge,
joins as to Parts I, II, IV.C, IV.D, V.A, and V.C, and with
whom RAWLINSON, Circuit Judge, joins except as to Part
V.C.2:

As the district court aptly put it, these cases arise from the
“continuing war over protection of the delta smelt.” San Luis
& Delta-Mendota Water Auth. v. Salazar, 760 F. Supp. 2d
855, 863 (E.D. Cal. 2010). We are joined to the fray. The
district court invalidated a biological opinion by the Fish and
Wildlife Service that concluded that the Central Valley and
State Water Projects jeopardize the continued existence of a
three-inch fish and its habitat. We reverse in part and affirm
in part.

The Central Valley Project and the State Water Project,
operated respectively by the Bureau of Reclamation
(Reclamation)1 and the State of California, are perhaps the
two largest and most important water projects in the United
States. These combined projects supply water originating in
northern California to more than 20,000,000 agricultural and
domestic consumers in central and southern California. The
source of this water, the estuary at the confluence of the San
Francisco Bay and Sacramento-San Joaquin Delta (Bay-
Delta), is also the lone habitat for the delta smelt, a threatened
species under the Endangered Species Act. 16 U.S.C. § 1531
et seq.

1
See the Glossary of Terms at the end of this opinion for a complete list
of abbreviations and acronyms.
24 SAN LUIS V. JEWELL

In 2008, Reclamation requested a biological opinion
(BiOp) from the U.S. Fish and Wildlife Service (FWS), in
accord with the Endangered Species Act (ESA), on whether
its continued operations would jeopardize the smelt. In a
more than 400-page opinion—described by the FWS as the
most complex biological opinion ever prepared—the FWS
concluded that the Central Valley operations would threaten
the delta smelt and, as required by the Endangered Species
Act, proposed “reasonable and prudent alternatives” that
Reclamation should take to ameliorate the effect on the smelt.
The alternatives recommended by the FWS would reduce the
water exported from northern California to southern
California through the Central Valley and State Water
Projects. Reclamation has notified the FWS that it intends to
operate the Projects in compliance with the biological
opinion.

The plaintiffs-appellees—various water districts, water
contractors, and agricultural consumers2—brought suit under
the Administrative Procedure Act against various federal
defendants, including Reclamation, the FWS, and the
Secretary of the Interior, to prevent the federal defendants
from implementing the biological opinion and its proposed
alternatives. The district court, in a lengthy and
comprehensive opinion, was deeply critical of the biological
opinion and concluded that it was arbitrary and capricious.
The court accused the FWS of repeatedly “ignoring [the] best
science available” to reach a “results-driven choice.” 760 F.
Supp. 2d at 940, and “show[ing] no inclination to fully and

2
The plaintiffs-appellees include the San Luis & Delta-Mendota Water
Authority, the Westlands Water District, Stewart & Jasper Orchards, the
California Department of Water Resources, and the Metropolitan Water
District of Southern California.
 SAN LUIS V. JEWELL 25

honestly address water supply needs beyond the species,”
even as it “interdict[s] the water supply for domestic human
consumption and agricultural use for over twenty million
people who depend on the Projects for their water supply,” id.
at 956–57 (quoting the FWS).

We are acutely aware of the consequences of this
proceeding. As a court, however, we are limited in our
review of matters within the expertise of an agency. We may
review the FWS’s biological opinion and Reclamation’s
implementation for arbitrariness, caprice, or actions otherwise
not in accordance with law. 5 U.S.C. § 706(2)(A). Although
the FWS must employ “the best scientific and commercial
data available,” 16 U.S.C. § 1536(a)(2), it is “not required to
support its finding that a significant risk exists with anything
approaching scientific certainty,” Indus. Union Dep’t v. Am.
Petroleum Inst., 448 U.S. 607, 656 (1980) (plurality opinion).
And, “[w]hen examining this kind of scientific determination
. . . a reviewing court must generally be at its most
deferential.” Baltimore Gas & Elec. Co. v. NRDC, 462 U.S.
87, 103 (1983). For the reasons explained below, we
conclude that the district court failed to observe these
standards and we reverse its judgment.

We recognize the enormous practical implications of this
decision. But the consequences were prescribed when
Congress determined that “these species of fish, wildlife, and
plants are of esthetic, ecological, educational, historical,
recreational, and scientific value to the Nation and its
people.” 16 U.S.C. § 1531(a)(3). As the Supreme Court
observed in Tennessee Valley Authority v. Hill: “It may seem
curious to some that the survival of a relatively small number
of three-inch fish . . . would require the permanent halting of
a virtually completed dam,” but “the explicit provisions of the
26 SAN LUIS V. JEWELL

Endangered Species Act require precisely that result.”
437 U.S. 153, 172–73 (1978). Such species have been
“afforded the highest of priorities,” by Congress, even if it
means “the sacrifice of the anticipated benefits of the project
and of many millions of dollars in public funds.” Id. at 174
(footnote omitted). The law prohibits us from making “such
fine utilitarian calculations” to balance the smelt’s interests
against the interests of the citizens of California. Id. at 187.
Consequently, any other “[r]esolution of these fundamental
policy questions” about the allocation of water resources in
California “lies . . . with Congress and the agencies to which
Congress has delegated authority, as well as with state
legislatures and, ultimately, the populace as a whole.”
Baltimore Gas & Elec., 462 U.S. at 97.

I. FACTS AND PROCEEDINGS BELOW

A. Background

1. The Sacramento-San Joaquin Delta

“The history of California water development and
distribution is a story of supply and demand. California’s
critical water problem is not a lack of water but uneven
distribution of water resources.” United States v. State Water
Res. Control Bd., 182 Cal. App.3d 82, 98 (Cal. Ct. App.
1986). California’s Central Valley comprises some of the
most productive farmland in the world. Extending 450 miles
from north to south, and averaging 100 miles wide east to
west, the Central Valley includes two principal rivers: The
Sacramento River begins in the northern part of the valley,
runs south past Sacramento, and is fed by the Feather and
American Rivers. The San Joaquin River begins in the Sierra
Nevadas, northeast of Fresno, runs west and northwest into
 SAN LUIS V. JEWELL 27

the Central Valley, and is fed by smaller rivers, including the
Calaveras, Chowchilla, Fresno, Kings, Merced, Mokelumne,
Stanislaus, and Tuolumne Rivers. The two rivers converge in
the Sacramento-San Joaquin Delta and form an estuary that
joins Suisun Bay, San Francisco Bay, and the Pacific Ocean.
Although over 70 percent of California’s water originates
north of Sacramento, more than 70 percent of the state’s
demand is in the south. The water from this region supplies
irrigation for seven million acres of agriculture and more than
twenty million people, nearly half of California’s residents.
See Where Does California’s Water Come From?,
Aquafornia, The California Water News Blog, Aug. 13, 2008,
9:29 a.m., http://www.aquafornia.com/index.php/where-does-
californias-water-come-from/. See generally Dugan v. Rank,
372 U.S. 609, 612–13 (1963); United States v. Gerlach Live
Stock Co., 339 U.S. 725, 728–29 (1950); San Luis & Delta-
Mendota Water Auth. v. United States, 672 F.3d 676, 681–83
(9th Cir. 2012); Westlands Water Dist. v. United States,
337 F.3d 1092, 1095–96 (9th Cir. 2003); In re Bay-Delta
Programmatic Envtl. Impact Report Coordinated
Proceedings, 184 P.3d 709, 715–17 (Cal. 2008); State Water
Res. Control Bd., 182 Cal. App.3d at 97–100.

In an effort to manage the increasing and conflicting
demands placed on the water flowing through the
Sacramento-San Joaquin Delta region, California and the
United States have embarked on two massive projects. First,
in 1933, California proposed the Central Valley Project
(CVP), a plan to transfer water from the Sacramento River to
water-deficient areas in the San Joaquin Valley and from the
San Joaquin River to the southern regions of the Central
Valley. State Water Res. Control Bd., 182 Cal. App.3d at
98–100. Reclamation took over the project in 1935, and it is
now “the largest federal water management project in the
28 SAN LUIS V. JEWELL

United States.” Central Delta Water Agency v. United States,
306 F.3d 938, 943 (9th Cir. 2002). The CVP consists of a
series of dams, including Shasta, Folsom, and Nimbus Dams;
21 reservoirs; 11 hydropower plants; and 500 miles of canals
and aqueducts. In re Bay-Delta, 184 P.3d at 716 n.1. In
1992, Congress adopted the Central Valley Project
Improvement Act (CVPIA), Pub. L. No. 102-575, 106 Stat.
4706, which Congress described as designed “to achieve a
reasonable balance among competing demands for use of
Central Valley Project water, including the requirements of
fish and wildlife, agricultural, municipal and industrial and
power contractors.” CVPIA, § 3402(f), 106 Stat. at 4706.

In 1951, California approved what is known as the State
Water Project (SWP), the largest state-built water project in
the United States. San Luis & Delta-Mendota Water Auth.,
672 F.3d at 683. Managed by the California Department of
Water Resources (DWR), “[t]he SWP serves the domestic
water needs of approximately two-thirds of all Californians,”
principally in Southern California. In re Bay-Delta, 184 P.3d
at 716. SWP consists of “21 dams and reservoirs, . . . five
power plants, 16 pumping plants, and 662 miles of aqueduct.”
Id. at 716 n.2. In 1994, eight state agencies and 10 federal
agencies formed the CALFED Bay-Delta Program
(CALFED) to address comprehensively the challenges of
managing the Bay-Delta estuary. Id. at 717.

The CVP and SWP each operate a major station for
pumping water from the Bay-Delta to canals and aqueducts
that will carry the water to the south. Both plants are located
near Tracy, California, and together they reverse the natural
flow of the southern part of the Bay-Delta through two
distributaries of the San Joaquin, Old, and Middle Rivers,
referred to as “OMR.” San Luis & Delta-Mendota, 760 F.
 SAN LUIS V. JEWELL 29

Supp. 2d at 863. The CVP operates the Jones Pumping Plant,
capable of diverting 4,600 cubic feet per second (cfs).
Nearby, the SWP operates the Harvey O. Banks Pumping
Plant, with a capacity of 10,300 cfs, although it generally
operates at or below 6,680 cfs. BiOp at 82, 108, 159–60.
The plants have been constructed with louvers that allow
water to pass through into the pumping plant, but will prevent
most fish from entering the plants. The process of the fish
entering the plants, known as entrainment, traps some 52
different species of fish. BiOp at 67. The salvaged fish are
hauled in trucks injected with oxygen and released at sites on
the Sacramento and San Joaquin Rivers. BiOp at 67, 145.
Over a recent 15-year period, more than 110 million fish were
salvaged from the Jones and Banks facility. BiOp at 160.
This number, however, greatly underestimates the number of
fish actually entrained at the facilities, because fish less than
30 mm (1.2 inches) are not efficiently collected at the
louvers. BiOp at 160–61. Smaller fish, especially those in the
juvenile or larval stage, are killed in the pumps. BiOp at 210.
Those that are salvaged frequently do not survive the salvage
process. BiOp at 338.

The Colorado River and the SWP have historically been
the major supply of water for southern California. As the
result of an interstate agreement signed in 2003, California
will receive less water from the Colorado River.
Quantification Settlement Agreement, San Diego Water
Authority, http://www.sdcwa.org/quantification-settlement-
agreement (last visited July 29, 2013). As a consequence,
southern California has sought more water from SWP. BiOp
at 89–90. Increased demand for water from SWP has had a
predictable effect on the water flowing into the Bay-Delta.
As more water is diverted from the rivers that feed the Delta
into the southern Central Valley, the salinity of the Delta and
30 SAN LUIS V. JEWELL

its estuaries increases along with the threat to the species that
thrive there.

The Delta generally describes a large
lowland area with a labyrinth of natural
channels in and around the confluence of the
Sacramento and San Joaquin Rivers. . . .

The major factor affecting water quality in
the Delta is saltwater intrusion. Delta Lands,
situated at or below sea level, are constantly
subject to ocean tidal action. Salt water
entering from San Francisco Bay extends well
into the Delta, and intrusion of the saline tidal
waters is checked only by the natural barrier
formed by fresh water flowing out from the
Delta.

State Water Resources Control Bd., 182 Cal. App.3d at 107.
Since the 1970s, Reclamation and a raft of state agencies
have worked to mitigate the effects of increased water salinity
on Suisun Bay resulting from the upstream diversion of water
that would otherwise naturally flow through Suisun. BiOp at
112–13. Salinity levels in Suisun Bay are highly sensitive to
diversion from the Delta.

Two related standards are used to describe the salinity of
the Bay-Delta. The first is the Low Salinity Zone or LSZ.
BiOp at 147. The LSZ is the transition point between the
freshwater of the inland rivers and brackish water flowing
eastward from San Francisco Bay and the Pacific Ocean, and
includes water ranging in salinity from 0.5 parts per thousand
to six parts per thousand. BiOp at 191. The second is
referred to as X2. X2 represents the point in the Bay-Delta at
 SAN LUIS V. JEWELL 31

which the salinity is less than two parts per thousand. See
Westlands Water Dist. v. Dep’t of Interior, 376 F.3d 853, 876
(9th Cir. 2004). The LSZ, which encompasses a larger region
of the Bay-Delta, is generally centered around X2. Together,
these regions are largely determined by Bay-Delta outflow,
which is the difference between the inflow and the water
exported. The agencies use X2 as a marker for the LSZ as
well as a habitat indicator for fish and as a regulatory
standard. BiOp at 149–50, 236; San Luis & Delta Mendota,
760 F. Supp. 2d at 864 & nn.4–5. They express the location
of X2 as its distance in kilometers east of the Golden Gate
Bridge, Westlands Water Dist., 376 F.3d at 876.

2. The delta smelt

The delta smelt (Hypomesus transpacificus) is a small,
two-to-three inch species of fish endemic to the San
Francisco Bay/Sacramento-San Joaquin Delta Estuary. BiOp
at 140–41. Once an abundant species in the Bay-Delta
ecosystem, the delta smelt is now in imminent danger of
extinction. In March 1993, the species was listed as
threatened under the ESA, and the FWS designated the Bay-
Delta system a critical habitat for the delta smelt in 1994.3
50 C.F.R. § 17.11; BiOp at 140. Yet, over the past decade,
the delta smelt population has been decimated even relative
to these depleted levels, with a measured decline since 2000

3
In 1990, the National Marine Fisheries Service listed the Sacramento
River winter-run chinook salmon as a threatened species under the
Endangered Species Act. 50 C.F.R. § 227.4; Orff v. United States,
545 U.S. 596, 599 (2005); O’Neill v. United States, 50 F.3d 677, 681 (9th
Cir. 1995). This case does not involve the salmon.
32 SAN LUIS V. JEWELL

of up to three orders of magnitude below historic lows.4 San
Luis & Delta-Mendota, 760 F. Supp. 2d at 866. As a
consequence, the FWS announced in 2010 that reclassifying
the delta smelt from a threatened to an endangered species
was warranted but precluded by higher priority listings.5 Id.

The ESA provides “both substantive and procedural
provisions designed to protect endangered species and their
habitat.” Am. Rivers v. Nat’l Marine Fisheries Serv.,
126 F.3d 1118, 1121 (9th Cir. 1997).6 One such protection,
§ 7(a)(2) of the ESA, requires federal agencies to “insure that
any action authorized, funded, or carried out by such agency
. . . is not likely to jeopardize the continued existence of any
endangered species or threatened species or result in the
destruction or adverse modification of habitat of such
species.” 16 U.S.C. § 1536(a)(2). Should the agency find that
its proposed action may affect a listed species or critical
habitat, it must formally or informally consult with the
Secretary of the Interior, or his or her delegee. 16 U.S.C.
§ 1536(a)(4); 50 C.F.R. § 402.14(a); see Am. Rivers, 126 F.3d
at 1122. If no effect is found, consultation is not required.

4
The 2008 delta smelt population was estimated at 1.5% of the 1980
level, 75 Fed. Reg. 17667 (April 7, 2010), and 2009 levels were estimated
to be the lowest on record. San Luis & Delta-Mendota, 760 F. Supp. 2d
at 866.
5
The delta smelt was assigned a listing priority of 2 on a scale from 1
to 12, with 1 being the highest priority. San Luis & Delta-Mendota, 760
F. Supp. 2d at 866–67 n.6.
6
Among other things, the CVPIA instructs the Secretary of the Interior
to “operate the Central Valley Project to meet all obligations under State
and Federal law, including but not limited to the Federal Endangered
Species Act, 16 U.S.C. § 1531 et seq.” CVPIA, § 3406(b), 106 Stat. at
4714.
 SAN LUIS V. JEWELL 33

50 C.F.R. § 402.14. Formal consultation is required when the
acting agency or consulting agency determines that the
proposed action is likely to adversely affect a listed species or
critical habitat. 50 C.F.R. §§ 402.13, 402.14. Formal
consultation requires the consulting agency, here the FWS, to
issue a biological opinion stating whether the proposed action
is likely to jeopardize such species or habitat. 16 U.S.C.
§ 1536(b); 50 C.F.R. § 402.14. Should the action jeopardize
the species or habitat, the consulting agency must suggest any
“reasonable and prudent alternatives” (RPA) that would allow
the projects to continue operation without causing jeopardy
to the species or adverse modification to its critical habitat.
16 U.S.C. § 1536(b)(3)(A). Once it receives the BiOp, the
acting agency “shall determine whether and in what manner
to proceed with the action in light of its section 7 obligations
and the [FWS’s] biological opinion.” 50 C.F.R. § 402.15(a).
If, after consultation, the agency determines that it cannot
comply with § 7(a)(2), it may apply for an exemption, which
can only be authorized by the Endangered Species
Committee, an ad hoc panel composed of executive branch
members and at least one appointee from the state in which
the project is to occur. 16 U.S.C. § 1536(e); 50 C.F.R.
§§ 402.15(c), 451.

B. Proceedings Leading To The Present Controversy

1. The FWS’s 2008 Biological Opinion

Reclamation sought a biological opinion from the FWS as
part of its continued long-term operation of the CVP and its
coordinated operations with state agencies of the SWP. BiOp
34 SAN LUIS V. JEWELL

at 1, 8.7 Following § 7(a)(2) review and a subsequent formal
consultation, the FWS issued a biological opinion in 2005
(2005 BiOp). The 2005 BiOp found that the proposed
coordinated operations of the CVP and SWP would not have
an adverse effect on the continued existence and recovery of
the delta smelt and its critical habitat. San Luis & Delta-
Mendota, 760 F. Supp. 2d at 863. The Natural Resources
Defense Council—defendants-intervenors-appellants in the
present case—challenged the FWS’s conclusion in U.S.
District Court for the Eastern District of California, and the
court found the 2005 BiOp arbitrary and capricious.
Kempthorne, 506 F. Supp. 2d at 387. After conducting an
extensive evidentiary hearing, the district court issued an
interim remedial order and findings of fact and conclusions
of law, which covered, among other things, the effects on
delta smelt of negative flows in OMR. See San Luis & Delta-
Mendota, 760 F. Supp. 2d at 864. The district court ordered
Reclamation and DWR to implement a winter “pulse flow”
in OMR of no more negative than !2,000 cfs, and to “operate
the CVP and SWP to achieve a daily average net upstream
(reverse) flow in the OMR not to exceed !5,000 cfs on a
seven-day running average during a defined period in the
spring.”8 Id.; Int. Rem. Order at 5–7. The district court also

7
The precipitating event for Reclamation seeking a biological opinion
was a 2004 Operating Criteria and Plan (OCAP), reflecting changes in
facilities, water delivery requirements, and regulatory restrictions and
increasing coordination between federal and state agencies. NRDC v.
Kempthorne, 506 F. Supp. 2d 322, 330 (E.D. Cal. 2007).
8
When water is diverted from the Bay-Delta contrary to its natural
course, the amount of water exported is usually expressed as a negative
number because the pumping plants reverse OMR flow. For example, if
the Banks facility is pumping 2,500 cubic feet per second to the California
Aqueduct, it is noted as !2,500 cfs.
 SAN LUIS V. JEWELL 35

ordered the FWS to complete a new BiOp in just nine
months, a deadline that it would ultimately extend to one
year. San Luis & Delta-Mendota, 760 F. Supp. 2d at 865; Int.
Rem. Order at 2.

The FWS issued a new delta smelt biological opinion on
the deadline, December 15, 2008 (2008 BiOp or BiOp).9
BiOp at 1–396. In stark contrast to the 2005 BiOp, the 2008
BiOp concluded that the “coordinated operations of the CVP
and SWP, as proposed, are likely to jeopardize the continued
existence of the delta smelt” and “adversely modify delta
smelt critical habitat.” BiOp at 276–78. With respect to the
delta smelt, the FWS entered five findings of fact:
(1) “Diversions of water from the Delta have increased since
1967 when the SWP began operation in conjunction with the
CVP.” BiOp at 276. The CVP/SWP operations have
entrained smelt, including adults, larvae, and juveniles, at the
Banks and Jones facilities; reduced smelt habitat; and reduced
the Delta outflows, altering the location of the LSZ; (2) “The
delta smelt is currently at its lowest level of abundance since
monitoring began in 1967”; (3) “Under the proposed
CVP/SWP operations, inflows to the Delta are likely to be
further reduced, as water demands upstream of the Delta
increase, most notably on the American River.” BiOp at 276.
This is likely to “cause increased relative entrainment of adult
delta smelt in the winter and spring, and of larval and juvenile

9
In preparing this BiOp, the FWS subjected it to an Internal Peer
Review Team, consisting of experts in the development of complex
biological opinions under the ESA from throughout the FWS, comments
from a team of delta smelt experts from within the FWS and other related
agencies, the review of PBS&J, an environmental consulting firm, as well
as feedback from Reclamation and the DWR. See BiOp at i–vi. In
response to the significant feedback it received, the FWS made substantial
changes to the BiOp. See BiOp at ii.
36 SAN LUIS V. JEWELL

delta smelt in the spring”; (4) “Other baseline stressors will
continue to adversely affect the delta smelt, such as
contaminants, microcystis, aquatic macrophytes, and invasive
species”; (5) To recover, delta smelt will need a more
abundant adult population, an increase in the quality and
quantity of spawning, rearing, and migratory habitat, a
reduction in contaminants and pollutants, a reduction in
exposure to disease and toxic algal blooms, and a reduction
in entrainment at water-diversion facilities in the Bay-Delta.
BiOp at 276–77. With respect to delta smelt critical habitat,
the FWS found that “past and present operations of the
CVP/SWP have degraded these habitat elements” such that
they are “insufficient to support successful delta smelt
recruitment at levels that will provide for the species’
conservation.” BiOp at 278.

The FWS recommended five components and listed six
separate actions as “reasonable and prudent alternatives”
(RPA):

RPA Component 1 (Actions 1 and 2). Component 1
protects the adult delta smelt life stage by controlling OMR
flows during the vulnerable December to March period. It
has two proposed actions. Action 1 is “designed to protect
upmigrating delta smelt” and describes the two periods when
delta smelt are most vulnerable to entrainment: in December
and when the first flush appears. BiOp at 280–81. Action 1
therefore proposes limiting the negative flows at OMR based
on a “daily salvage index.” Id. In effect, this means that
when the “daily salvage index” reaches a critical point (“the
risk threshhold”), the Projects have to reduce their diversion
for 14 days. During that period, OMR flows can be “no more
negative than !2,000 cfs” for a 14-day running average and
“no more negative than -2,500 cfs” for a 5-day running
 SAN LUIS V. JEWELL 37

average. BiOp at 281, 329. Action 2 follows from Action 1
but covers the period from December through March, when
pre-spawning adult delta smelt are vulnerable to entrainment.
BiOp at 352. During that period, OMR flows can be no more
negative than !5,000 cfs, although the FWS expected that
flows would generally be in the range of !2,000 cfs to !3,500
cfs.

RPA Component 2 (Action 3). Component 2 protects
larval and juvenile delta smelt by limiting OMR flows
following the completion of Component 1 when the Bay-
Delta water temperatures reach 12EC, or when a spent female
smelt is detected in trawls at Jones or Banks or is found in the
salvage facilities. Action 3 requires the CVP/SWP projects
to maintain their average OMR flows between !1,250 and
!5,000 cfs until June 30, or until the mean water temperature
reaches a target level, whichever occurs earlier. BiOp at 290,
357–58.

RPA Component 3 (Action 4). Component 3 improves
smelt habitat by increasing Bay-Delta outflow during the fall.
Action 4 requires that in September and October, in years
when the precipitation and runoff is defined as “wet or above
normal,” Reclamation and DWR must provide sufficient
Delta outflow to maintain X2 no more eastward than 74 km
from the Golden Gate in wet years and 81 km in above-
normal years.10 BiOp at 282–83, 369.

10
In Action 4, the FWS noted that “there may be other ways to achieve
the biological goals of this action,” and that it would evaluate alternatives
and modify Action 4 “consistent with the intention of this action.” The
FWS referred to this process as an “adaptive management process.” BiOp
at 283.
38 SAN LUIS V. JEWELL

RPA Component 4 (Action 611). Component 4 restores
habitat by establishing a program to create or restore
intertidal and associated subtidal habitat to the Bay-Delta and
Suisun Marsh. Action 6 requires DWR to create or restore at
least 8,000 acres in the Delta and Suisun Marsh. BiOp at 283.

RPA Component 5. Component 5 monitors and reports on
the implementation, success, and possible improvements of
Components 1–4.12

Finally, the FWS issued an “incidental take statement”
(ITS) in accord with 50 C.F.R. § 402.02. For purposes of the
ITS, the FWS presumed that its reasonable and prudent
alternatives would be implemented. Based on that premise,
the FWS found that, as a result of CVP/SWP operations, there
would be a take of the delta smelt, and that although the
extent of the take would be difficult to estimate, smelt
entrainment would be minimized when OMR flows were
regulated according to the FWS’s proposed RPA. BiOp at
285–86. As a consequence, the FWS concluded that “this
level of anticipated take is not likely to result in jeopardy to
the species or destruction or adverse modification of critical
habitat when the RPA is implemented.” BiOp at 293.

11
For Action 5 see, infra, Note 12.
12
Component 5 did not have a separate action item, and Action 5 was
not formally associated with any of the RPA Components. Action 5 was
specific to the installation of a physical barrier on the Old River that
affected entrainment of larval and juvenile delta smelt. BiOp at 377–78.
 SAN LUIS V. JEWELL 39

2. The present case

The first of six complaints challenging the FWS’s 2008
BiOp was filed in March 2009. San Luis & Delta-Mendota,
760 F. Supp. 2d at 865. “Plaintiffs moved for a preliminary
injunction . . . to prevent Reclamation from implementing
Component 2 of the RPA, alleging that FWS violated the
National Environmental Policy Act (“NEPA”) and the ESA.”
Id. The district court granted the motion in part, finding that
plaintiffs-appellees were likely to succeed on the merits of
their NEPA claim, and requiring the FWS to make specific
written findings to justify weekly decisions regarding OMR
flow restrictions. Id.

Plaintiffs-appellees sought a preliminary injunction
against the implementation of RPA Component 3. Id.
Following an evidentiary hearing,13 the district court issued
a preliminary injunction confirming that plaintiffs-appellees
had succeeded on their NEPA claims and finding that

13
At the evidentiary hearing, the district court appointed four experts
pursuant to Federal Rule of Evidence 706, which permits the court to
“appoint any expert that the parties agree on and any of its own choosing.”
Fed. R. Evid. 706(a). The court appointed two professors from the
University of Washington, Dr. Andre Punt and Dr. Thomas Quinn, to
advise the court on the complex technical and scientific matters. The
court also appointed two additional experts: a Reclamation employee
knowledgeable on the CVP, and a DWR employee knowledgeable in the
SWP.

In addition to these court-appointed experts, the district court
permitted substantial declarations from experts selected (but not agreed to)
by the parties. In its written decision, the district court relied extensively
on opinions and evidence submitted by its own experts and by the parties’
experts.
40 SAN LUIS V. JEWELL

plaintiffs-appellees were likely to succeed on the merits of
their ESA claim. Id.

In December 2010, the district court entered final
judgment on the primary claims in a 115-page opinion. Id. at
967–70. Although the FWS’s 2008 BiOp reached antipodal
conclusions to the 2005 BiOp—which reached a no-jeopardy
conclusion and was found arbitrary and capricious—the
district court once again found the 2008 BiOp to be arbitrary
and capricious under the ESA and the APA and remanded the
BiOp, its RPA, and Reclamation’s provisional acceptance of
the RPA to the agency. San Luis & Delta-Mendota, 760 F.
Supp. 2d at 855, 970. The court’s remand required the
completion of yet a third BiOp analyzing the impact of CVP
and SWP operations on the delta smelt. San Luis & Delta-
Mendota, 760 F. Supp. 2d at 870. In March 2011, the district
court entered final judgment on all remaining claims.

Although the district court accepted the BiOp’s central
conclusion that “Project operations are likely to jeopardize
the continued existence and/or adversely modify the critical
habitat of the delta smelt,” id. at 969, the district court
determined that there were a number of specific flaws with
the BiOp, id. at 967–70. We will briefly set forth the district
court’s principal objections here—which are highly technical
and somewhat obtuse out of context—and explain them in
more detail in the discussion section.

First, the district court found the BiOp’s reliance on
analyses using raw salvage figures—e.g., those calculations
that incorporated the absolute “raw” number of smelt
entrained in pumping stations, as opposed to the smelt
entrained as a percentage of the total population—to be
arbitrary and capricious and not the result of the best
 SAN LUIS V. JEWELL 41

available science. Id. at 968. These calculations significantly
influenced the upper and lower OMR flow limits in Actions
1, 2, and 3.

Second, the district court found that the BiOp’s use of two
different models, CALSIM II and DAYFLOW, that predict
the location of X2, introduced bias requiring a corrective
calibration or, at the very least, explanation. Id. The district
court also found that the bias produced by the comparison of
CALSIM II to DAYFLOW tainted the BiOp’s justification
for Action 4, which involves the management of X2. In
addition, the district court found that the BiOp did not
sufficiently explain why it is essential in Component 3
(Action 4) to maintain X2 at the specific locations of 74 km
upstream from the Golden Gate Bridge following “wet years”
and 81 km following “above-normal years.” Id. at 969.

Third, the district court found that the BiOp did not
sufficiently explain why different data sets were used to
calculate the incidental take limit for juvenile and for adult
smelt, and why these limits were calculated using an average
of previous years’ smelt salvage (which would be expected to
be exceeded in 50% of all future years).

Fourth, the district court found the BiOp did not
adequately support its conclusions that Project operations are
reasonably certain to indirectly affect the delta smelt by
limiting delta smelt food supply, by increasing harmful
pollution and contaminants, and by increasing the detrimental
impact of the “other stressors” of predation, macrophytes, and
microcystis on delta smelt. Id.

Fifth, the district court held that the BiOp failed to
analyze economic feasibility, consistency with the purpose of
42 SAN LUIS V. JEWELL

the action, and consistency with the action agency’s authority,
as required by § 402.02. See 5 U.S.C. § 551 et seq; 50 C.F.R.
§ 402.02. Id. at 969–70.

Federal Defendants and the NRDC have timely appealed,
urging reversal of the district court’s remand. We address
their claims in Part IV.

The San Luis & Delta-Mendota Water Authority and
other appellees timely cross-appealed, arguing that the district
court did not go far enough. They raise three claims: First,
that the FWS violated the ESA by not separating out non-
discretionary actions from discretionary actions in setting an
environmental baseline; second, that Reclamation acted
arbitrarily and capriciously in adopting the flawed BiOp; and,
third, that the FWS failed to conduct the review required by
NEPA.14 We address their claims in Part V.

This tortured procedural history has extended over seven
years, and has led to five fully consolidated suits and one
partially consolidated suit brought by various groups who use
water supplied by the CVP and SWP, as well as to the
completion of two extensively researched BiOps—with a
third currently in progress. All the while, the delta smelt has
teetered on the brink of extinction.

We agree with Federal Defendants and the NRDC and
reverse the district court's remand of the BiOp. We affirm the

14
Federal Defendants and the NRDC also appeal the district court’s
determination that Reclamation failed to conduct review required by
NEPA. We address this direct appeal together with appellees’ cross
appeal.
 SAN LUIS V. JEWELL 43

district court’s judgment that Reclamation failed to comply
with NEPA before implementing the FWS’s BiOp.

II. STANDARDS OF REVIEW

We review the district court’s ruling on summary
judgment de novo. McFarland v. Kempthorne, 545 F.3d
1106, 1110 (9th Cir. 2008).

Neither the ESA nor NEPA supply a separate standard for
our review, so we review claims under these Acts under the
standards of the APA. Bennett v. Spear, 520 U.S. 154, 174
(1997); Oregon Natural Desert Ass’n v. Bureau of Land
Mgmt., 625 F.3d 1092, 1109 (9th Cir. 2010); Pyramid Lake
Paiute Tribe of Indians v. U.S. Dept. of Navy, 898 F.2d 1410,
1414 (9th Cir. 1990). Section 706(2) of the APA provides
that an agency action must be upheld on review unless it is
“arbitrary, capricious, an abuse of discretion, or otherwise not
in accordance with law.” 5 U.S.C. § 706(2)(A). As a
reviewing court, we “must consider whether the decision was
based on a consideration of the relevant factors and whether
there has been a clear error of judgment.” Citizens to
Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 416
(1971), abrogated in part on other grounds as recognized in
Califano v. Sanders, 430 U.S. 99, 105 (1977). Although our
inquiry must be thorough, the standard of review is highly
deferential; the agency’s decision is “entitled to a
presumption of regularity,” and we may not substitute our
judgment for that of the agency. Id. at 415–16. Where the
agency has relied on “relevant evidence [such that] a
reasonable mind might accept as adequate to support a
conclusion,” its decision is supported by “substantial
evidence.” Bear Lake Watch, Inc. v. FERC, 324 F.3d 1071,
1076 (9th Cir. 2003). Even “[i]f the evidence is susceptible
44 SAN LUIS V. JEWELL

of more than one rational interpretation, [the court] must
uphold [the agency’s] findings.” Id.

Under the ESA, the agency must base its actions on
evidence supported by “the best scientific and commercial
data available.” 50 C.F.R. § 402.14(g)(8); 16 U.S.C.
§ 1536(a)(2). The determination of what constitutes the “best
scientific data available” belongs to the agency’s “special
expertise . . . . When examining this kind of scientific
determination, as opposed to simple findings of fact, a
reviewing court must generally be at its most deferential.”
Baltimore Gas & Elec. Co., 462 U.S. at 103. “Absent superior
data[,] occasional imperfections do not violate” the ESA best
available standard. Kern Cnty. Farm Bureau v. Allen,
450 F.3d 1072, 1080–81 (9th Cir. 2006).

“The best available data requirement ‘merely prohibits
[an agency] from disregarding available scientific evidence
that is in some way better than the evidence [it] relies on.’”
Kern Cnty. Farm Bureau, 450 F.3d at 1080 (emphasis added)
(quoting Sw. Ctr. for Biological Diversity v. Babbitt, 215 F.3d
58, 60 (D.C. Cir. 2000)). “Essentially, FWS ‘cannot ignore
available biological information.’” Id. at 1080–81 (quoting
Connor v. Burford, 848 F.2d 1441, 1454 (9th Cir. 1988)).
Thus, “insufficient . . . [or] incomplete information . . . does
not excuse [an agency’s] failure to comply with the statutory
requirement of a comprehensive biological opinion using the
best information available” where there was some additional
superior information available. Conner, 848 F.2d at 1454–55.
On the other hand, where the information is not readily
available, we cannot insist on perfection: “[T]he ‘best
scientific . . . data available,’” does not mean “the best
scientific data possible.” Building Indus. Ass’n v. Norton,
247 F.3d 1241, 1246 (D.C. Cir. 2001).
 SAN LUIS V. JEWELL 45

III. THE SCOPE OF THE RECORD

Our review is limited to “the administrative record
already in existence, not some new record made initially in
the reviewing court.” Camp v. Pitts, 411 U.S. 138, 142
(1973).

If the record before the agency does not
support the agency action, if the agency has
not considered all relevant factors, or if the
reviewing court simply cannot evaluate the
challenged agency action on the basis of the
record before it, the proper course, except in
rare circumstances, is to remand to the agency
for additional investigation or explanation.
The reviewing court is not generally
empowered to conduct a de novo inquiry into
the matter being reviewed and to reach its
own conclusions based on such an inquiry. . . .

The factfinding capacity of the district
court is thus typically unnecessary to judicial
review of agency decisionmaking.

Florida Power & Light Co. v. Lorion, 470 U.S. 729, 744
(1985). See also Citizens to Preserve Overton Park, 401 U.S.
at 419–20.

There is a danger when a reviewing court goes beyond the
record before the agency. “When a reviewing court considers
evidence that was not before the agency, it inevitably leads
the reviewing court to substitute its judgment for that of the
agency.” Asarco, Inc. v. EPA, 616 F.2d 1153, 1160 (9th Cir.
1980). See Sw. Ctr. for Biological Diversity v. U.S. Forest
46 SAN LUIS V. JEWELL

Serv., 100 F.3d 1443, 1450 (9th Cir. 1996) (“Judicial review
of an agency decision typically focuses on the administrative
record in existence at the time of the decision and does not
encompass any part of the record that is made initially in the
reviewing court.”). Accordingly, we do not review “the
evidence to determine the correctness or wisdom of the
agency’s decision . . . even if the court has also examined the
administrative record.” Asarco, 616 F.2d at 1160. If the
reviewing court cannot find substantial evidence in the
record, it should “not compensate for the agency’s dereliction
by undertaking its own inquiry into the merits,” id., but
should remand to the agency for further proceedings, see INS
v. Ventura, 537 U.S. 12, 16 (2002).

“We have, however, crafted narrow exceptions to this
general rule.” Lands Council v. Powell, 395 F.3d 1019, 1030
(9th Cir. 2005). As we have explained,

[w]e allow expansion of the administrative
record in four narrowly construed
circumstances: (1) supplementation is
necessary to determine if the agency has
considered all factors and explained its
decision; (2) the agency relied on documents
not in the record; (3) supplementation is
needed to explain technical terms or complex
subjects; or (4) plaintiffs have shown bad faith
on the part of the agency.

Fence Creek Cattle Co. v. U.S. Forest Serv., 602 F.3d 1125,
1131 (9th Cir. 2010). Keeping in mind the Supreme Court’s
concerns with reviewing court factfinding, we have
approached these exceptions with caution, lest “the exception
 SAN LUIS V. JEWELL 47

. . . undermine the general rule.” Lands Council, 395 F.3d at
1030.

We have serious concerns that the district court failed to
observe these rules. First, the district court appointed four
experts to aid it in understanding the technical and scientific
aspects of the BiOp. These experts were appointed consistent
with Federal Rule of Evidence 706. See supra Note 13.
Having read the BiOp, we are sympathetic to the district
court’s need for a scientific interpreter. No party has objected
on appeal to the district court’s appointments, and we can see
no reasonable objection to the use of experts to explain the
highly technical material in the BiOp.

The district court, however, did not limit itself to the
court-appointed experts. Over the vigorous objection of the
appellants, the court admitted multiple declarations from
multiple experts hired by the appellees, even though the
party-appointed experts addressed many of the same issues
being addressed by the court-appointed experts. By the
government’s count—which we have not sought to verify,
although we have examined the declarations in the
record—the district court admitted more than forty expert
declarations from the appellees; once the court denied the
appellant’s motion to exclude the declarations, the appellants
submitted their own experts’ declarations. Yet, we cannot see
what the parties’ experts added that the court-appointed
experts could not have reasonably provided to the district
court. See Animal Def. Council v. Hodel, 840 F.2d 1432,
1437 (9th Cir. 1988) (plaintiff failed to show why the district
court needed to go outside the administrative record). The
effect of this was, predictably, to create a battle of the
experts. Moreover, it gave the proceedings in the district
court the appearance that the administrative record was open
48 SAN LUIS V. JEWELL

and that the proceedings were a forum for debating the merits
of the BiOp. See Ariz. Cattle Growers’ Ass’n v. U.S. Fish &
Wildlife, 273 F.3d 1229, 1245 (9th Cir. 2001) (“Considering
evidence outside the record would render the extraordinarily
complex consultation process . . . meaningless.”). Just as we
will not allow the agency to supply post-hoc rationalizations
for its actions, so “post-decision information . . . may not be
advanced as a new rationalization either for sustaining or
attacking an agency’s decision.” Sw. Ctr. for Biological
Diversity, 100 F.3d at 1450. As the Court cautioned, “[w]hen
specialists express conflicting views, an agency must have
discretion to rely on the reasonable opinions of its own
qualified experts even if, as an original matter, a court might
find contrary views more persuasive.” Marsh v. Ore. Natural
Res. Def. Council, 490 U.S. 360, 378 (1989).

Even a quick review of the district court’s opinion shows
that the appearance of an open record was the reality. The
district court relied extensively on the declarations of the
parties’ experts-as-advocates as the basis for rejecting the
BiOp. See, e.g., San Luis & Delta-Mendota, 760 F. Supp. 2d
at 877–79, 881–84, 889–90, 894–97, 903–07, 912, 922. In
places, the district court pits the experts against each other
and resolves their contrary positions as a matter of scientific
fact. E.g., San Luis & Delta-Mendota, 760 F. Supp. 2d at
884, 904–07. In effect, the district court opened the BiOp to
a post-hoc notice-and-comment proceeding involving the
parties’ experts, and then judged the BiOp against the
comments received. The ESA consultation process is not a
rulemaking proceeding, but a request from one agency for the
expertise of a second agency. Although we may review each
of those proceedings under the APA, the agency’s obligations
under each is slightly different, and we must account for that
difference in our review.
 SAN LUIS V. JEWELL 49

Our opinion in Asarco v. EPA addressed the question of
using experts in a technical case and is on point here. In that
case, the district court had a technical matter before
it—involving a 1,000 foot stack at a smelter—and permitted
the parties to call their experts. The court held four days of
hearings in which at least ten witnesses testified. We held
that some of the testimony, mainly from experts who
provided “background material” on the smelter’s operations,
was appropriate; the rest of it was not appropriate, even if it
came from experts and “served some marginal purpose”:

Most of the expert testimony, however, should
not have been admitted or at least not have
been considered for the purpose of judging the
wisdom of the EPA’s stack-testing
requirement. This technical testimony was
plainly elicited for the purpose of determining
the scientific merit of the EPA’s decision.
Although the testimony may have served
some marginal purpose in allowing the district
court to evaluate the EPA’s court of inquiry,
we can only conclude that the extent of the
scientific inquiry undertaken at trial
necessarily led the district court to substitute
its judgement for that of the agency.

616 F.2d 1155, 1160–61 (9th Cir. 1980).

The district court similarly overstepped its bounds here.
Because we review the court’s judgment de novo, however,
we can confine our own scope of review to the administrative
record, plus that evidence that satisfies the standards we have
set forth here. See Dow Agrosciences v. Nat’l Marine
Fisheries Serv., 707 F.3d 462, 469 (4th Cir. 2013) (confining
50 SAN LUIS V. JEWELL

review to the administrative record and disregarding an
affidavit submitted to the district court). We will consider the
BiOp and any other evidence in the record submitted by the
parties that the FWS considered in making its decision. We
will also consider the testimony of the four experts the district
court appointed pursuant to Rule 706, as to whom there is no
objection.

IV. MERITS OF THE BiOp

Before we consider the challenges to the BiOp, we have
some preliminary observations. First, the BiOp is a bit of a
mess. And not just a little bit of a mess, but, at more than 400
pages, a big bit of a mess. And the FWS knew it. In
November 2008, shortly before the FWS submitted its final
draft, the California Department of Water Resources
commented on a portion of the FWS’s draft. Under the title
“The Document is Confusing and Disorganized,” the DWR
advised that “the document is especially unrefined. The text
is often out of logical order . . . , many actions are fairly
vague, and the document is unpolished.” Some portions,
DWR wrote, were “largely unintelligible.” The FWS also
submitted its draft to a peer review panel, coordinated by an
outside engineering firm. One of the questions the FWS
asked the panel was whether the BiOp was “organized in a
manner that is clear, concise and complete (i.e., is it
understandable)?” The panel pulled no punches and
responded as follows: “The Panel’s response to this question
is ‘no.’ The version of the [Effects Analysis] provided to the
Panel was a draft and had not been adequately edited for
general organization, consistency across sections in how
analyses were described and reported, and for redundancies.”
The bottom line was that “most readers would have a difficult
time.”
 SAN LUIS V. JEWELL 51

Both the California DWR and the peer review panel
commented on the FWS’s imposing schedule for issuing the
BiOp. DWR observed that the “FWS was under an
exceptionally difficult deadline to get this document in.” In
a comment letter sent from the DWR to Reclamation just two
weeks before the deadline, the DWR reflected on the
pressures the FWS labored under: “Our concerns with the
Draft [BiOp] are extensive but are correctable before the
December 15 court ordered [deadline] for its finalization. We
are willing to work closely with the FWS to address these
concerns quickly. . . .” And the peer review panel
complained that its own “review was conducted in a four-day
period under a tight schedule.” See Ass’n of Pac. Fisheries v.
EPA, 615 F.2d 794, 811 (9th Cir. 1980) (“The Agency itself
recognized that its data collection was not as thorough as it
otherwise would have been: ‘The time constraints imposed by
the statutory deadlines precluded the Agency from
conducting an exhaustive sampling program.’”). This
challenging deadline was not the fault of the agency, but was
set by the same district court that would later hold that the
FWS’s rushed BiOp was arbitrary and capricious. See San
Luis & Delta-Mendota, 506 F. Supp. at 322 (ordering the
FWS to produce a revised BiOp in just nine months—a
deadline that would later be extended by three months to give
the FWS just one year to produce a new BiOp).15

15
One of the ironies of the district court’s deadlines is that the FWS had
less time to produce its opinion than either the district court or we will
have had to review it: The 2008 BiOp was issued in December 2008. The
district court’s summary judgment decision issued almost two years to the
day, in 2010. This opinion issues nearly three years later.

The concerns in adhering to deadlines is familiar to courts who must
occasionally rush to judgment. Justice Powell once bemoaned the
problems arising from deciding cases “late in the Term” when the Court
52 SAN LUIS V. JEWELL

We concur in these assessments. The BiOp is a jumble of
disjointed facts and analyses. It appears to be the result of
exactly what we would imagine happens when an agency is
ordered to produce an important opinion on an extremely
complicated and technical subject matter covering multiple
federal and state agencies and affecting millions of acres of
land and tens of millions of people. We expect that the
document was patched together from prior documents,
assembled quickly by individuals working independent of
each other, and not edited for readability, redundancies, and
flow. It is a ponderous, chaotic document, overwhelming in
size, and without the kinds of signposts and roadmaps that
even trained, intelligent readers need in order to follow the
agency’s reasoning. We wonder whether anyone was
ultimately well-served by the imposition of tight deadlines in
a matter of such consequence.16 Deadlines become a
substantive constraint on what an agency can reasonably do.
In this case, the FWS not only had to write and compile the
report—a substantial task in and of itself—but was under
pressure to, among other things, produce a reliable population
estimate of the delta smelt. Kempthorne, 506 F. Supp. 2d at
373 (faulting the 2005 BiOp for failing to produce such an
estimate, and noting that “[t]he viability of Delta smelt has
been under scrutiny for over ten years. No party has shown

was precluded by its own deadlines from “an opportunity for more
thorough consideration of the basic principles at risk.” Robbins v.
California, 453 U.S. 420, 435–36 (1981) (Powell, J., concurring in the
judgment). Robbins was overruled the following Term in United States
v. Ross, 456 U.S. 798, 824 (1982).
16
We recognize that the ESA itself imposes deadlines for a consulting
agency to produce a BiOp, but those deadlines may be extended by
consent of the requesting agency or the concerned applicant. See
16 U.S.C. § 1536(b)(1), (2). Here, the district court fixed the deadline.
 SAN LUIS V. JEWELL 53

that producing a reliable population estimate is scientifically
unfeasible”). Such scientific tasks may not be as well suited
to deadlines as producing written copy; the final product will
necessarily reflect the time allotted to the agency. The FWS
is currently preparing its third BiOp, again under the orders
of the district court and on the district court’s deadlines.
Although we ultimately conclude that we can discern the
agency’s reasoning and that the FWS’s 2008 BiOp is
adequately supported by the record and not arbitrary and
capricious, we also recognize that Reclamation has
continuing responsibilities under CVP and SWP and that this
is likely not the last BiOp that the FWS will issue with
respect to the delta smelt, nor is this the last legal challenge
that we will hear. Future analyses should be given the time
and attention that these serious issues deserve.

A. The 2008 BiOp’s Reliance on Raw Salvage Figures to Set
the Upper and Lower OMR Flow Limits Was Not
Arbitrary and Capricious

Under normal pumping operations, enough water is
pumped from the OMR that the river’s flow reverses. BiOp
at 159–60. As vast quantities of water are pumped from the
river, smelt and other fish are entrained in the pumps, where
they are captured and counted in fish salvage facilities. Id.
Not all smelt are salvaged, however, as juvenile smelt smaller
than 20 mm (0.79 inches), smelt still in the larval stage, and
some percentage of adult delta smelt, are killed by the pumps.
BiOp at 338. The prospects for salvaged smelt are also grim,
as smelt rarely survive the salvage process, id.; up to sixty
percent of the smelt population is lost each year at the
pumping plants, BiOp at 210. Unsurprisingly, the 2008 BiOp
found that as the OMR flows became increasingly negative,
the entrainment risk and accompanying population loss
54 SAN LUIS V. JEWELL

increased accordingly, thereby threatening the smelt
population’s continuing viability. BiOp at 163.

To mitigate this effect, the BiOp RPA imposed pumping
limits expressed in terms of negative OMR flows,17 ranging
from !1,250 cubic feet per second (cfs) to !5,000 cfs. BiOp
at 280. Which limit is applicable is determined by the
location of the smelt population, water turbidity,18 and a range
of other factors. BiOp at 280. In determining the OMR flow
limits, the FWS relied, in part, on the number of delta smelt
salvaged from the fish screening facilities. Although it
acknowledged that the number of smelt salvaged only
represents “a small percentage of the actual number
entrained,” and “is not a good estimate of actual delta smelt
mortality through entrainment,” BiOp at 338, the FWS relied
on this information because “[d]ata on the salvage of delta
smelt is typically used to provide an index of entrainment into
the diversion pumps,” BiOp at 145.

The OMR flow limit has a great practical significance, not
merely to the delta smelt but to Californians, as it represents
the ultimate limit on the amount of water available to sustain
California’s millions of urban and agricultural users.

Appellees challenge the !5,000 cfs upper limit on OMR
flows, claiming that the FWS based its calculation of the flow
limit solely on information contained in two figures that show

17
Actions 1 and 2 of the RPA protect adult delta smelt migrating
upstream in winter months to spawn, and Action 3 protects larval and
juvenile delta smelt in the spring after they hatch. BiOp at 280, 282.
18
Water turbidity is the measurement of how much sediment or foreign
particles are suspended in the water.
 SAN LUIS V. JEWELL 55

a significant increase in the number of smelt salvaged at
approximately !5,000 cfs: Figures B-13 and B-14. See BiOp
at 347, 350. These figures, appellees assert, failed to justify
the !5,000 cfs limit because the number of delta smelt
salvaged (raw salvage) was not adjusted for the smelt’s
estimated population for that year (normalized). Appellees’
objection to the BiOp’s use of raw salvage data, rather than
normalized data, in calculating appropriate OMR flow limits
is essentially an omitted variable bias argument; appellees
assert that the number of smelt salvaged in a year is highly
influenced by the total population of smelt that year, and that
therefore the BiOp’s failure to account for this relation
renders its calculated flow limits unreliable. Put another way,
according to appellees, when there are more smelt in the Bay-
Delta, more smelt are salvaged. Any apparent relationship
between OMR flows and smelt salvage, therefore, may
actually be a relationship between smelt population size and
smelt salvage. Failing to account for this will skew, if not
invalidate entirely, the analysis.

The district court agreed, finding that “[t]he use of raw
salvage data, as opposed to salvage data scaled to population
size, is problematic because raw salvage figures do not
account for the size (or relative size) of the smelt population.”
San Luis & Delta-Mendota, 760 F. Supp. 2d at 889. The
district court further found that “the use of normalized
salvage data rather than gross salvage data is the standard
accepted scientific methodology.” Id. It concluded that the
FWS’s use of raw salvage data in Figures B-13 and B-14 to
determine restrictions on OMR flows was “scientifically
unacceptable.” Id. at 891. Because the FWS failed to use the
best available scientific data, its !5,000 cfs flow limit in
Actions 1, 2 and 3 was arbitrary and capricious. Id.; see id. at
891–94.
56 SAN LUIS V. JEWELL

These objections, however, suffer from several problems.
First, the FWS appropriately relied on Figures B-13 and B-14
to justify its !5,000 cfs flow limits. In Figures B-13 and B-
14, the FWS acknowledged the uncertainty inherent to
modeling the relation between OMR flows and smelt and
chose a conservative model, a choice that is within the FWS’s
discretion to make. See Nw. Coal. for Alts. to Pesticides v.
EPA, 544 F.3d 1043, 1050 (9th Cir. 2008). Second, the
!5,000 cfs flow limit prescribed in the RPA was not solely
determined from Figures B-13 and B-14, and therefore flaws
in those figures do not necessarily doom the BiOp’s
conclusions. Third, the BiOp’s OMR flow limits work in
tandem with the incidental take statement (ITS), which
accounts for population-level impacts.

That the FWS could have done more in determining OMR
flow limits is uncontroverted. This, however, is not to say
that the FWS acted arbitrarily and capriciously; we hold,
contrary to the district court, that the FWS’s OMR flow limits
are supported by substantial evidence.

1. The FWS’s choice of a more conservative model to
calculate flow limits in Figures B-13 and B-14 was
supported by substantial evidence

Figure B-13 in the BiOp19 is a graph showing the
relationship between salvaged adult delta smelt and the OMR
flow, measured in cfs. The graph shows a positive correlation
between salvaged smelt and the reverse flow of the river.
That is, the greater the water pumped through the Jones and
Banks pumping stations, the greater the count of smelt

19
Figures B-13 and B-14 are reproduced in the district court’s published
opinion. Kempthorne, 506 F. Supp. 2d at 886–87.
 SAN LUIS V. JEWELL 57

salvaged at those stations. The FWS noted that the graph is
upward sloping and linear in the lower half of the curve. At
about !5,000 cfs there is a “break” in the data,20 and for flows
more negative than !5,000 cfs (meaning more water exported
from the Bay-Delta), the upward slope increases at an
increasing rate. The FWS sought to verify whether the break
in the data was actual. It conducted additional analyses of the
data to verify that there was not a natural break at any other
point and that any error in the OMR flow rates or salvage
could not have caused the break. BiOp at 349–51. The FWS
concluded that with “flows more negative than !1683,
salvage increased.” BiOp at 351.

The FWS stated that “[a] major assumption of this
analysis is that as the population of Delta smelt declined, the
number of fish at risk of entrainment remained constant.”
BiOp at 349. What B-13 did was compare actual salvage
numbers with OMR flow. What the FWS did not do—and
what the appellees and the district court claim the FWS
should have done—was prepare an additional figure in which
it compared “normalized” salvage numbers with OMR flows.
Normalized salvage would be the measure of the salvaged
smelt divided by the total population of smelt,21 effectively

20
The break in the data in B-13 appears to span about !3,000 cfs to just
under !6,000 cfs. BiOp at 348. The FWS did not attempt to quantify the
break; it simply observed that there “appears to be a ‘break’ in the dataset
at approximately !5,000 OMR.” BiOp at 347.
21
Part of the district court’s frustration may be that, to date, no one has
been able to produce a reliable population estimate for the delta smelt. In
2007 when the district court sent the 2005 BiOp back to the FWS, it
commented that
58 SAN LUIS V. JEWELL

yielding figures showing what percentage of smelt each year
were salvaged at the pumps.

In fact, the FWS itself had stated that it could verify its
conclusion “by normalizing the salvage data by the estimated
population size based on the [Fall Midwater Trawl] data.”
BiOp at 349. The peer review had similarly suggested
normalizing the data:

The Panel suggests that the use of predicted
salvage of adult smelt should be normalized
for population size. . . . One way to normalize

The viability of Delta smelt has been under
scrutiny for over ten years. No party has shown that
producing a reliable population estimate is scientifically
unfeasible. . . . Without population estimates, it is
arbitrary for the agency to conclude that project
operations will not result in jeopardy simply because
the projects will take relatively fewer smelt than they
did in the past, in the fact of the undisputed fact that the
smelt population has been declining steadily in recent
years. Failing to incorporate any information about
smelt population abundance into the setting of the take
limits is a fundamental failure rendering the BiOp
arbitrary and capricious.

Kempthorne, 506 F. Supp. 2d at 373. Despite the court’s warning, the
FWS did not conduct a population study, nor did it explain why one could
not be conducted. Yet, the FWS, its peer reviewers, and the district
court’s experts suggested that there were proxies for population (such as
the FMWT count) available, even if a strict population count was not.

Sometimes we have to read the shadows to discern the reality behind
it: Nearly twenty years after the smelt were declared endangered, we know
the smelt population is continuing to decline and is imperiled, but still no
one knows how many there are. It must tell us something about the
difficulties that inhere in trying to count migrating, two-inch fish.
 SAN LUIS V. JEWELL 59

salvage for population size is to divide by the
previous fall Midwater Trawl (MWT) index.
A similar regression model to the one fitted to
salvage would relate the normalized salvage
to Old and Middle River (OMR) flows. . . .
Expressing salvage as a normalized index may
help remove some of the confounding of the
temporal trends during the baseline period
....

As the district court’s experts acknowledged, the FWS faces
significant practical challenges in setting OMR flow rates to
minimize delta smelt entrainment. For example, day-to-day
variations in OMR flows and “noise” in smelt sampling22
used to establish abundance and distribution of the delta smelt
are significant confounding factors in determining
appropriate OMR flow rates, as is the distribution of the delta

22
“Noise” is a statistical term that refers to the unexplained randomness
or variation that is found in a sample. It is of particular concern when
statistical samples are small. As Dr. Quinn explained:

[W]hat are the uncertainties in the population estimates
themselves, and might there be shifting levels of
accuracy as population levels change? . . . [It] is
certainly true [that population growth rate is an
appropriate and reliable measure of the population
increases and decreases from year to year], if it is
known without error but what are the assumptions
about sampling? That is, as smelt become increasingly
scarce, does their overall distribution become ‘thinner
all around’ or is it ‘patchy’, and how might such
changes influence the reliability of data from different
surveys of abundance? In general, ‘noisy’ data make it
more difficult to detect underlying patterns, even if the
patterns are genuine.
60 SAN LUIS V. JEWELL

smelt population in relation to the pumps. BiOp at 165, 331,
353–55. A lack of real-time information and variations
inherent to environmental systems make precision virtually
impossible. BiOp at 165, 331, 353–55. Yet, as even the
district court recognized, population numbers of the delta
smelt are perilously low, San Luis & Delta-Mendota, 760 F.
Supp. 2d at 866, and entrainment by the pumping plants has
a “sporadically significant influence on population
dynamics,” id. at 877 (emphasis added).

In such circumstances, the FWS’s decision to use raw
salvage data rather than population-adjusted salvage data
reasonably protects the delta smelt population without regard
to year-by-year fluctuation in population size. The BiOp
notes that this decision was motivated by a concern for the
absolute number of smelt entrained in the pumps, not the
relative number of smelt: “The current population cannot
tolerate direct mortality through adult entrainment at levels
approaching even ‘moderate’ take as observed through the
historic record of recent decades.” BiOp at 287. Thus, the
RPA is designed to “reduce entrainment of pre-spawning
adult delta smelt during December to March” and to
“[m]inimize the number of larval delta smelt entrained at the
facilities” by controlling OMR flow from March to June.
BiOp at 280, 357. The analytical approach preferred by
appellees and the district court is best gauged to measure the
number of smelt entrained at the pumps relative to the
population size. This may be a more accurate reflection of
the relative impact of OMR flows on the smelt population,
but it is not tailored to protect the maximum absolute number
of individual smelt, as the BiOp’s approach is; the process of
adjusting raw salvage for the smelt population size results in
 SAN LUIS V. JEWELL 61

normalized numbers, but it does so at the potential cost of
minimizing the impact of each individual smelt lost.23

Our deference to agency determinations is at its greatest
when that agency is choosing between various scientific
models, as the FWS did in the present instance. See Nw.
Coal. for Alts. to Pesticides, 544 F.3d at 1050. Facing great
measurement uncertainty and a smelt population whose
existence is threatened, the FWS chose to be conservative in
setting the flow limits in Actions 1, 2, and 3. This choice was
well within its discretion; the Supreme Court has held that an
agency may choose to “counteract the uncertainties” inherent
in its scientific analyses by “overestimat[ing]” known
parameters without being unreasonable, Baltimore Gas &
Elec. Co., 462 U.S. at 103, and we have upheld an agency’s
reliance on models that “yield conservative data because the

23
Appellee-Metropolitan Water District uses “one of the data points
located at about !5,000 cfs on Figure B-13” for the year 2000 as an
example. An unusually high smelt salvage was observed in 2000, but, as
Appellee indicates, the smelt population was also higher than usual in that
year. Appellee argues that because “[i]t is to be expected that more fish
would be salvaged in a year in which the population was extremely large,”
the raw salvage number should be normalized for total smelt population.
Accepting appellee’s argument that salvage rates should be normalized,
the year 2000 would have actually represented a below average
(normalized) salvage. Yet, it is undisputed that an extraordinary number
of smelt were salvaged in that year.

A normalized analysis of smelt salvage counts the year 2000 as a
below-average year, while an analysis of raw smelt salvage counts the
year 2000 as an above-average year. The FWS’s choice of analysis
influences whether it is the OMR flow’s relative or absolute impact on
smelt population that is prioritized. Thus, the quality of the statistical
method is not the only relevant factor at play, as the district court
erroneously concluded: the BiOp’s choice of one model over the other
implicates significantly differing management policies.
62 SAN LUIS V. JEWELL

models incorporate the higher of [known potential values] in
assessing the overall risk,” Nw. Coal. for Alts to Pesticides,
544 F.3d at 1050. Likewise, we give the FWS great deference
in its choice of scientific tools, and, in these circumstances,
hold that the FWS did not act arbitrarily or capriciously in
choosing an analytical tool that resulted in greater protections
for the imperiled smelt population.24

24
The district court relied substantially on the testimony of its experts,
Dr. Andre Punt and Dr. Thomas Quinn, in concluding that the BiOp’s
reliance on Figures B-13 and B-14 was “scientifically unacceptable.” San
Luis & Delta-Mendota, 760 F. Supp. 2d at 891. But these experts present
a more nuanced view of the FWS’s use of Figures B-13 and B-14. The
results confirm what the FWS had already said: that additional analyses
using normalized data from B-13 could have informed the FWS’s
conclusions. The experts noted that B-13 was a proper measure, although
not the only, and perhaps not the preferable, measure. The experts also
testified as to the need for the FWS to set some parameters on OMR
flows, and the difficulty in figuring out precisely where the parameters
should be.

In response to the district court’s question whether it was
“unreasonable for FWS to rely in part on the information represented in
figure B-13,” Dr. Quinn answered that he did “not regard it as
unreasonable for FWS to have relied in part on this figure and the data
behind it.” But he cautioned that “[t]o rely entirely on it would, however,
have neglected the complexities of the issue . . . . Both the number of fish
salvaged and the proportion salvaged . . . are relevant, in my view, as are
other kinds of information.” Dr. Punt had a similar reaction: “it was
unreasonable . . . to have only relied on the information in Figures B-13
and B-14 rather than on an analysis in which salvage is expressed relative
to population size.” (emphasis added).

Moreover, although both experts had been critical of the FWS’s
failure to run the additional numbers, both were cautious about what the
normalized data might have shown, whether any hard conclusions could
be drawn from any data set, and even whether normalized data would be
preferable to the non-normalized data produced. Dr. Quinn acknowledged
that even if the FWS had normalized the data in B-13, “plotting flow
 SAN LUIS V. JEWELL 63

against a salvage index might not fully capture the risk to the population.”
With respect to what B-13 showed, Dr. Quinn cautioned that it was
“unwise to overestimate the precision” of the data. He was not convinced
that there was a “break” in the data, “though [his] basis for saying so [was]
more intuitive than statistical.” He “emphasize[d] that any point value has
a measure of arbitrariness. If !5000, then why not !4900 cfs? Given the
many sources of variation in the data, it strikes me as necessary to set
limits even though there may not be strong statistical basis for a particular
figure rather than a slightly different one.” He concluded that although
“the validity of [FWS’s] specific flow regimes [was] undermined by the
incomplete analyses that were done on the available data,” it was
“appropriate for the FWS to have some leeway in making decision and
setting limits in their efforts to protect and recover listed species.” Dr.
Punt went one step further, recognizing the advantage of non-normalized
data: B-13, as it is produced in the BiOp, would be justified if “any
entrainment, no matter how small relative to the total population size, has
long-term consequences for the population size of delta smelt.”

These responses from the experts question the FWS’s failure to
expand its analysis and to normalize the data underlying B-13. Had it
done so, the experts concluded, the FWS would have had a more complete
sense of the relationship between OMR flow and salvage, although it is far
from clear that that study would have affected the FWS’s conclusions.
What the experts’ testimonies do not support is the district court’s
overstated conclusion that the FWS’s “use of raw salvage in the analyses
depicted in Figures B-13 and B-14 is scientifically unacceptable” or that
“such metrics are meaningless as management tools.” San Luis & Delta-
Mendota, 760 F. Supp. 2d at 891. Were we only evaluating the experts’
opinion on the BiOp, we would face a difficult question as to the
continuing validity of this aspect of the BiOp: it is clear that these two
experts believed the BiOp to have fallen short in this analysis, although as
Dr. Punt indicates, persuasive justifications exist for the BiOp’s reliance
on non-normalized data. It is less clear, however, that the BiOp—even as
seen through the eyes of Drs. Punt and Quinn—would be rendered
arbitrary and capricious by a sole reliance on Figures B-13 and B-14. We
need not reach this question, as we accept the opinions of the district
court’s experts only insofar as they are persuasive and informative,
and—as will be subsequently described—we independently conclude that
64 SAN LUIS V. JEWELL

2. The BiOp’s determination of OMR flow limits was
influenced by more than Figures B-13 and B-14

The district court’s experts concluded that it was the
BiOp’s apparent exclusive reliance on Figures B-13 and B-14
that was problematic: as they acknowledged, the figures
could be useful in tandem with other analyses and data. The
district court therefore based its invalidation of the !5,000 cfs
OMR flow limit on its finding that the limit “depend[s] so
heavily” on Figures B-13 and B-14. San Luis & Delta-
Mendota, 760 F. Supp. 2d at 968. After conducting an
independent review of the record, however, we hold that the
BiOp’s determination of OMR flow limits was sufficiently
influenced by several other population-level analyses in
addition to Figures B-13 and B-14. Together with Figures B-
13 and B-14, these analyses provide substantial evidence that
supports the BiOp’s !5,000 cfs OMR flow limit.

First, the BiOp expressly notes that it employed multiple
analyses in determining the OMR flow limits, stating with
regards to Actions 1 and 2 that “recent analyses indicate that
cumulative adult entrainment and salvage are lower when
OMR flows are no more negative than !5,000 cfs in the
December through March period.” BiOp at 281 (emphasis
added). Although the BiOp fails to specify exactly which
studies it relies on,25 at least one of these studies is

the BiOp does not exclusively rely on Figures B-13 and B-14 in
determining the OMR flow limits.
25
The BiOp should have been more explicit in describing exactly which
studies it used in its analysis. Had the BiOp been similarly vague
throughout its analysis, its lack of specificity may have presented a
problem. Here, however, the BiOp has provided sufficient support for its
conclusions.
 SAN LUIS V. JEWELL 65

contextually apparent: a 2008 study by Wim J. Kimmerer,
Losses of Sacramento River Chinook Salmon and Delta Smelt
to Entrainment, San Francisco Estuary and Watershed
Science (hereinafter “Kimmerer 2008”). Kimmerer’s study
was referred to throughout the BiOp, and his data was
reproduced in a chart estimating adult entrainment. BiOp at
281. The Kimmerer 2008 study quantified the relationship
between population losses and OMR flows between 1995 and
2006, concluding—consistent with Figure B-14—that as
OMR flow becomes more negative, delta smelt population
losses increase. Moreover, in Figure E-4, the FWS read
Kimmerer 2008 to show that when OMR flows are more
negative than !5,000 cfs—the “break” observed in Figure B-
13—population loss typically exceeds 10 percent. BiOp at
250 (Figure E-4). Figure E-4 shows a strong correlation
between flows more negative than !5,000 cfs and high smelt
population losses, supporting Figure B-13’s “break.” See
BiOp at 250; see also BiOp at 213 (discussing Figure E-4).26

26
Even if Figure E-4’s correlation were less apparent, not all analyses
will yield clear “change points,” and therefore the fact that none is
immediately apparent in a study does not mean that the study cannot
support an agency’s specifically defined limits; such a holding would
effectively prohibit an agency from setting exact limits in any
circumstance in which a clear change point did not emerge from the data.
In the present situation, Kimmerer 2008 and Figure E-4 present a
reasonable justification for a !5,000 cfs limit, which does appear to be an
approximate point of some significance. That Kimmerer 2008 may also
provide a reasonable justification for a !4,999 or a !5,001 limit does not
make the FWS’s policy choice arbitrary and capricious, especially given
the significant deference owed to the agency in our review. To do so
smacks more of strict scrutiny than arbitrary and capricious review under
the APA and ESA.

Even were the range of the limit justified by Kimmerer 2008
substantially greater than this, and it most likely is, it is not appropriate for
66 SAN LUIS V. JEWELL

Second, the BiOp was more explicit in describing its
bases for determining the !5,000 cfs limit for Action 3,
stating that “[t]he OMR flows associated with the
protectiveness criteria defined above have been derived from
particle tracking modeling27 with the input assumptions
defined below.” BiOp at 360. Because Action 3 protects
larval and juvenile delta smelt specifically, which are
generally smaller than 20 mm (0.79 inches)—too small for
salvage facility tracking—particle tracking modeling was
necessary to estimate Action 3 entrainment. BiOp at 282.
The district court appears to have misunderstood the
distinction between the analyses supporting the Action 1 and
2 limits and those supporting the Action 3 limits. San Luis &
Delta-Mendota, 760 F. Supp. 2d at 922 (finding that the FWS
derived the !5,000 cfs limit for Action 3 from the district
court’s previous order, from a non-linear DWR model, and
from Figure B-13, failing to mention the particle tracking
modeling). The particle tracking modeling results are
reflected in the BiOp’s figure B-16, and demonstrate a low
risk of entrainment as long as OMR flows remain below
!2,000 cfs, increasing to a 20 percent risk at !3,500 cfs.
BiOp at 366–67. The high entrainment at more negative
flows would potentially place at risk 80 percent of all
larval/juvenile smelt over the period of the approximately
four months that Action 3 is under effect. BiOp at 366. As
the OMR flow approaches !5,000 cfs, the modeling predicts
that smelt entrainment will near 100 percent at multiple

a reviewing court to second guess the agency’s scientific judgments in
such matters where, as is the case here, the agency’s determination
appears to fall within the data-justified range.
27
Particle tracking modeling simulates larval fish by inserting neutrally
buoyant particles into a model domain with flow conditions.
 SAN LUIS V. JEWELL 67

stations. BiOp at 367. Given such findings, the FWS did not
act unreasonably in setting the OMR flow limits; the particle
tracking modeling completed for Action 3 provides
independent support for the OMR limits.

Third, the DWR recommended to the FWS that it set the
parameters at !5,000 cfs. In its informal comments to the
FWS, it recommended that “[t]he justification for using a
range of !5000 to !1250 cfs OMR rather than the !5000 cfs
the FWS, USBR, and DWR proposed to the District Court
last summer should be more clearly explained.” In formal
comments submitted to Reclamation just two weeks before
the BiOp was due, the DWR criticized the RPA Component
1 (Action 2) because it “call[ed] for much more restrictive
Old and Middle River Flows than supported by the best
available data to reasonably minimize entrainment of adult
delta smelt.” As the DWR noted, the “proposed flows range
from !1,250 cfs to !5,000 cfs. The current set flow of
!5,000 cfs should be used except in specific circumstances
relating to fish survey and salvage data, because it is better
supported by the available scientific information.” (emphasis
added) The DWR added that

[i]n DWR’s October 16 submittal we
presented the monthly analysis of Old and
Middle River (OMR) flows and salvage of
adult delta smelt that was also provided to
Judge Wanger in 2007. This analysis appears
to have been ignored by the FWS even though
the analysis is more highly predictive of adult
delta smelt salvage than any of the analysis
presented by the FWS in the Draft BO. It
clearly shows that when the monthly OMR
68 SAN LUIS V. JEWELL

flows are more negative than about !5,000
cfs., the risk of salvage increases dramatically.

....

Given the lack of clear relationship between
salvage and population effects, any constraints
on salvage needs to focus on avoiding peak
entrainment events rather than attempting to
eliminate salvage. Therefore, we again
suggest that this analysis be included in the
BO and that the !5,000 cfs OMR flows on a
14 day average period be used except in rare
circumstances where the data indicates that
less negative flows are needed to protect
against peak entrainment events.

(emphasis added).

Fourth, in its comments on an early draft BiOp, the
independent peer review also urged consideration of the
relationship between salvage and OMR flows: “[The
relationship between OMR flows and salvage] is a sound and
valuable way to set targets to reduce entrainment. The
USFWS also presents a reasonable regression analysis to
determine the break-point in the OMR-salvage relationship
. . . . The breakpoints determined by these analyses were used
to justify the selection of target OMR flows.”28 It was not
arbitrary and capricious for the FWS to rely on DWR’s own

28
At the same time, the DWR criticized the draft because “the analysis
appeared to be well done but was poorly described and largely
undocumented.”
 SAN LUIS V. JEWELL 69

expertise and views on the appropriate flow limits, even if it
did not accept DWR’s preferred recommendation.

Finally, we have one last observation in this area. When
the district court remanded the 2005 BiOp as arbitrary and
capricious—a BiOp that concluded that the operations of
CVP and SWP would not adversely affect the delta
smelt—the district court ordered Reclamation and DWR to
“operate the CVP and SWP to achieve a daily average net
upstream (reverse) flow in the OMR not to exceed [!]5,000
cfs on a seven-day running average.” San Luis & Delta-
Mendota, 760 F. Supp. 2d at 863–64. During the drafting of
the BiOp, an interagency delta smelt team (known as the
“Delta Smelt OCAP BO Technical Team”) “discussed the
merits of using !5,000 OMR per Wanger Order rather than
!3,500 OMR as recommended in the proposed draft action.”
Notes of the meeting state:

The !5,000 OMR cap was established by
Wanger. The Team discussed the biological
needs of the smelt to remove jeopardy and
adverse modification. What might have
worked in the year 2002 might not work
correctly in 2009 because the population has
crashed. It was suggested that scaling the
amount of protection to the fall mid-water
trawl (FMWT) (i.e. an estimate of abundance)
would be helpful.

....

If !5,000 OMR is the cap to protect adults,
then the cap for Action #3 should be less than
that, because the juvenile behave more like
70 SAN LUIS V. JEWELL

particles. . . . OMR rates directly relate to
zone of entrainment. If fish are really out of
the central Delta then perhaps !5,000 OMR
might provide sufficient protection. However,
it is extremely difficult to determine when fish
are out of South/Central Delta.

The district court criticized the FWS for relying on “a
provisional court order, entered as a remedial stopgap
measure pending comprehensive scientific analysis.” San
Luis & Delta-Mendota, 760 F. Supp. 2d at 896. We
understand the district court’s concern and agree with the
court’s general proposition that evidence used to support a
“stopgap measure” is not, without more, substantial evidence
supporting a final measure. But we also think the principle is
not so easily applied in this case. As the district court noted,
its 2007 order—in which the !5,000 cfs figure was
found—was based on an evidentiary hearing in which two
studies were introduced that considered the relationship
between OMR flows and delta smelt salvage. Id. According
to the district court, one of these studies was the basis for B-
13.

The FWS can hardly be faulted for thinking that the
district court’s acceptance of those studies and the issuing of
an order with real-world consequences for people and smelt
might present at least a prima facie case for the !5,000 cfs
figure. We understand and agree with the district court that
if the district court’s interim order was the sole basis for the
FWS’s BiOp that it would not constitute substantial evidence.
But, in 2008, the FWS was at least entitled to rely on the
studies the district court had accepted (albeit on an interim
basis) in 2007. The interagency team’s infelicitous reference
to the “Wanger Order” was surely a shorthand for the
 SAN LUIS V. JEWELL 71

evidence on which the district court based its interim order,
and not just the order itself. Additionally, we can take notice
of the fact that by 2008, when the BiOp was issued, the FWS
had a year’s experience living under the district court’s order,
as the members of the interagency delta smelt team
undoubtedly knew. Again, if the “Wanger Order” was the
sole evidence in the record for the FWS’s RPA Component
2 (Action 3), we would not hesitate to find that the agency
had not relied on the “best scientific data available,” but
where the !5,000 cfs figure in the Wanger Order was one
more data point for the agency, we cannot find that the
agency’s reference is irrelevant or improper.

3. The OMR flow limits exist as one part in a dynamic
monitoring system that accounts for the smelt
population as a whole29

As we have described, the FWS’s task of monitoring
OMR flow and smelt population is a daunting one. The BiOp
accounts for these challenges in a number of ways, including
choosing conservative models so as to best meet its ESA
obligations. One other way in which the BiOp addresses
these practical difficulties is by integrating its various
protections; the OMR flow limits exist as but one part of a
complicated dynamic system. Another significant part of this
system is a limit on the total allowable take of delta smelt: the
incidental take limit (ITS). BiOp at 387.

29
We do not believe the BiOp’s OMR flow limits to be arbitrary and
capricious for several alternative reasons previously described. But, even
had the FWS relied entirely on Figures B-13 and B-14 in setting the OMR
flow limits—it did not—and even if such total reliance would otherwise
be arbitrary and capricious—it would not—we hold that the BiOp’s use
of whole population numbers in the ITS fully supports its OMR flow
limits.
72 SAN LUIS V. JEWELL

The ITS influences OMR flow levels in two primary
ways. First, the ITS is used to establish a “Concern Level”
estimate to “help guide implementation of the RPA.” Id. By
“indicat[ing when] salvage levels approach[] the take
threshold,” the Concern Level acts “as an indicator that
operations need to be more constrained to avoid exceeding
the incidental take.” Id. If the Concern Level is actually
reached, a meeting of the Smelt Working Group—a group
comprised of interagency biologists who monitor delta smelt
conditions and recommend OMR flow levels—is triggered.
Id. Second, the ITS “functions as an action that influences
operations under the RPA.” San Luis & Delta-Mendota,
760 F. Supp. 2d at 929. Pertinent to the present discussion,
the ITS is used to set actual flow levels under the RPA in real
time. BiOp at 352, 354, 357. The real-time actual flow levels
are set and adjusted based on the input of the Smelt Working
Group and Project operators—a group informed by the ITS.
BiOp at 352, 354, 357.

Thus, the actual OMR flow is highly influenced by the
ITS, as these two provisions work together as part of the
complex dynamic system established by the BiOp. This is
relevant to the present discussion: even while the OMR flow
limits, viewed in isolation, may not account for total smelt
population in a manner acceptable to the district court, the
ITS—and therefore the actual OMR flows—takes the total
smelt population into account.

The ITS establishes take limits that vary each year based
on the preceding Fall Midwater Trawl index (FMWT), an
abundance proxy for the delta smelt’s population size—and
the proxy for population suggested both by the FWS and the
court-appointed experts as they considered normalizing the
data in B-13. BiOp at 287, 383–86. The ITS is calculated by
 SAN LUIS V. JEWELL 73

projecting future salvage from historic salvage in years with
comparable flows to the RPA, BiOp at 384, and then scaling
that number to overall abundance using the prior year’s
FMWT, BiOp at 385. This procedure “yields a discrete value
for take as salvage so that the adaptive process can operate
relative to an estimate of the absolute number of fish extant
in the system.” San Luis & Delta-Mendota, 760 F. Supp. 2d
at 926–27. As such, the take limits are adjusted to reflect the
best estimate of smelt’s existing population. BiOp at 354
(noting both salvage and population size as “important
variables” in setting flow). Thus, in making real-time
operational recommendations to implement the RPA, the
BiOp relies on both raw salvage numbers and the whole smelt
population. BiOp at 352.

In conclusion, we agree that the FWS should have at least
prepared a graph similar to B-13 based on normalized data or
explained why it could not. Nevertheless, overall, its use of
OMR flows is supported by substantial evidence in the
record. As convoluted as the BiOp is, we can discern the path
the agency took to arrive at the !5,000 cfs figure used in RPA
Component 1 (Actions 1 and 2) and RPA Component 2
(Action 3). See Bowman Transp., Inc. v. Arkansas-Best
Freight Sys., Inc., 419 U.S. 281, 286 (1974) (“[W]e will
uphold a decision of less than ideal clarity if the agency’s
path may reasonably be discerned.”). In the sense explained
by Dr. Quinn, the BiOp’s choice of !5,000 cfs was arbitrary;
that is, the FWS had to choose some number from a broad
range—perhaps in the !4,000 cfs to !6,000 cfs range—and
no graphs or charts were likely to give the FWS a precise
number. In this sense, !5,000 cfs is arbitrary because no
more precise number can be identified, and a number from
the range must be selected; !5,000 is thus an arbitrary
number because the FWS could also have chosen !4,999 or
74 SAN LUIS V. JEWELL

!5,001 or some other number within the range. That !5,000
is arbitrary in this sense does not make the choice of !5,000
arbitrary in the sense captured by the APA; its decision was
not the result of arbitrariness or caprice. The APA does not
demand strict scrutiny. The FWS is charged with protecting
the delta smelt, and it chose a reasonable figure.

B. The 2008 BiOp’s Determination of X2 Was Not Arbitrary
and Capricious

The FWS found that Reclamation and DWR’s proposed
operations “are likely to negatively affect the abundance of
delta smelt” by “substantially decreas[ing] the amount of
suitable abiotic habitat for delta smelt.” BiOp at 236–37. To
address the loss of habitat, the FWS proposed in RPA
Component 3 (Action 4) that in September and October, in
years when the preceding precipitation and runoff period was
defined as “wet or above normal,” Reclamation and the DWR
must provide sufficient Delta outflow to maintain monthly
average X2 no more eastward than 74 km from the Golden
Gate in wet years and 81 km in “above normal” immediate
water years. BiOp at 282, 369. The FWS had previously
found that the amount and quality of spawning habitat
available to delta smelt is linked to the location of X2. BiOp
at 239–40. As we previously discussed, X2 represents the
point in the Bay-Delta estuary where the salinity is two parts
per thousand, and is the center point of the LSZ, which is
considered suitable spawning habitat for the smelt. X2, in
turn, depends on delta outflow, which is largely determined
by the difference between the total inflow from the
Sacramento and San Joaquin Rivers and the total amount of
water exported through the Banks and Jones pumping
 SAN LUIS V. JEWELL 75

stations, which changes both annually and seasonally.30 BiOp
at 236. As the BiOp found, “CVP/SWP operations control
the position of X2 and therefore are a primary driver of delta
smelt habitat suitability.” BiOp at 234. Because the location
of X2 directly affects how much water can be exported to
southern California for agricultural and domestic purposes,
the determination of where X2 is located was critical to the
parties.

The district court found the BiOp was arbitrary and
capricious with respect to the location of X2 on two grounds.
First, the district court objected to the data the FWS used to
locate X2. San Luis & Delta-Mendota, 760 F. Supp. 2d at
903–09. Second, it found that the actual location of X2
lacked support in the record. Id. at 910–13, 922–23. We will
address each point in turn.

1. The FWS was not arbitrary and capricious in
comparing DAYFLOW to CALSIM II

The FWS used two sophisticated models to determine the
impacts of past, present, and future operations of the Project
on X2. The FWS chose a computer model called
DAYFLOW, developed by DWR, to measure the historical
environmental baseline, and a computer simulation model
known as CALSIM II, developed jointly by DWR and
Reclamation, to measure future operations. Id. at 896–98,
907. These models were used in a number of ways in the
BiOp, and were important to the BiOp’s determination of the

30
It is more correct to think of X2 as a range of points rather than a
single, fixed point in the estuary. As the BiOp points out, “X2 is strongly
influenced by tidal cycles, moving twice daily up and downstream 6–10
km for its average daily location.” BiOp at 372.
76 SAN LUIS V. JEWELL

impacts of Project operation on the location of X2. BiOp at
145.31

The DAYFLOW computer model uses past river flow,
export pumping, precipitation, and estimated agricultural
diversions from 1967–2007 to estimate the outflow from the
Bay-Delta to the San Francisco Bay. This data was used by
the BiOp as a “historic” baseline for X2. BiOp at 207.
CALSIM II is a computer simulation model that uses Central
Valley hydrology from 1922–2003 to simulate Project
operations. Despite the fact that environmental regulation
and non-Project water demands have historically changed, the
CALSIM II model assumes that these factors are fixed in its
modeling scenarios. BiOp at 207. This is in contrast with the
DAYFLOW model, which, because it relies largely on actual
reported data, does account for changes in regulations that
impact water demand. DAYFLOW and CALSIM II also
differ in other ways: for example, DAYFLOW uses historical
data to provide daily simulations of operation whereas
CALSIM II models Project operations on a monthly basis,
and DAYFLOW calculates X2 location using a mathematical
method known as the “KM” method whereas CALSIM II
uses the “ANN” mathematical method.

The BiOp, in analyzing the predicted location of X2,
estimated that median X2 would move 10 to 15 percent
farther upstream under the proposed action relative to the
historic median X2 baseline. BiOp at 265; see also BiOp at
235. Appellees assert, however, that it is impossible to
discern whether “this change was due to continued project
operations into the future or whether the change was due to

31
Action 4, which involves the management of X2, was also
significantly influenced by these models.
 SAN LUIS V. JEWELL 77

the modeling differences between the ‘historic’ Dayflow-
derived baseline and the Calsim II studies.” Appellees argue
that a CALSIM II to CALSIM II comparison would have
been preferable. The district court agreed, concluding that
although the FWS had discretion to use a historical baseline
model such as DAYFLOW, the FWS abused its discretion
when it compared the two different models without
discussing or accounting for the resulting bias. See San Luis
& Delta-Mendota, 760 F. Supp. 2d at 902–05. Although the
district court acknowledged that “no superior set of models
have been identified,” it held that a calibration problem
created by using DAYFLOW with CALSIM II made the
FWS’s choice of methodology arbitrary and capricious and
required correction or explanation. Id. at 877, 899, 909. We
recognize that the CALSIM II to DAYFLOW comparison is
not without its limitations, but hold that the FWS’s decision
to use these two models together, even without further
calibration, was not arbitrary and capricious.32

Because the FWS’s decision to use the DAYFLOW and
CALSIM II models together is a “scientific determination,”
Baltimore Gas & Elec. Co., 462 U.S. at 103, that “requires a

32
It is not clear from the record that the district court was correct in
concluding that the FWS’s reliance on a comparison of CALSIM II and
DAYFLOW models taints the BiOp’s conclusions regarding the impacts
of Project operations on the location of X2, or regarding Action 4. See
San Luis & Delta-Mendota, 760 F. Supp. 2d at 922–23. The FWS has
indicated other possible bases for its conclusions. For example, the FWS
refers to field sampling data, and not modeling results, as providing the
basis for its conclusions about X2 conditions, at least in the fall. The BiOp
also relies on historical data to demonstrate the upstream shift of fall X2
over time, BiOp at 237, 264, as well as the extent to which the Project
operations influenced the upstream shift in fall X2, BiOp at 179–82, 236,
270. See also Note 33 infra.
78 SAN LUIS V. JEWELL

high level of technical expertise,” Marsh, 490 U.S. at 377, we
must be at our most deferential in reviewing this provision of
the BiOp, Baltimore Gas & Elec. Co., 462 U.S. at 103.
Therefore, the question, under the ESA is not whether the
FWS should have conducted independent studies in lieu of
DAYFLOW or CALSIM II, Sw. Ctr. for Biological Diversity,
215 F.3d at 60 (“[T]he Secretary has no obligation to conduct
independent studies.”), but whether these models represent
the “best scientific and commercial data [currently]
available,” 16 U.S.C. § 1536(a)(2); Inland Empire Public
Lands Council v. U.S. Forest Serv., 88 F.3d 754, 762 (9th Cir.
1996). Because we agree with the district court’s conclusion
that “no superior set of models have been identified,” San
Luis & Delta-Mendota, 760 F. Supp. 2d at 909, the only
question is whether the FWS’s failure to calibrate the two
models renders conclusions drawn from their evidence not
merely “weak,” but “arbitrary and capricious.” Greenpeace
Action v. Franklin, 14 F.3d 1324, 1336 (9th Cir. 1992)
(noting that “the fact that the evidence [that an agency relies
on] is ‘weak’” is not dispositive); see also Bldg. Indus. Ass’n
of Superior Cal. v. Norton, 247 F.3d 1241, 1246–47 (D.C.
Cir. 2001) (holding that the fact that the “studies the Service
relied on were imperfect . . . alone is insufficient to
undermine those authorities’ status as the ‘best scientific . . .
data available’”).

The FWS explained why it chose to use “a combination
of available tools and data.”33 San Luis & Delta-Mendota,

33
The BiOp makes clear that the FWS did not rely exclusively on the
CALSIM II/DAYFLOW comparison, but looked to previous government
studies, and several peer-reviewed scientific studies:
 SAN LUIS V. JEWELL 79

760 F. Supp. 2d at 912; BiOp at 204–05. The BiOp explained
that it used the DAYFLOW model as a baseline, rather than
a different CALSIM II simulation, because when it compared
a CALSIM II simulation of current Project operations to a
CALSIM II simulation of past Project operations, it found
that the two were “nearly identical,” despite the fact that past
and current Project operations are significantly different.
BiOp at 204–05. According to the BiOp, “changes were
expected” between the two CALSIM II studies. BiOp at 204.
Because that comparison did not yield significant differences,
as the FWS anticipated, the BiOp concluded that “the
CALSIM monthly simulation model does not capture a
precise Delta operation.” BiOp at 204. These “inaccuracies
in CALSIM [lead the FWS] to use actual data to develop an
empirical baseline.” BiOp at 206. The FWS chose “the
DAYFLOW database . . . and OMR data obtained from
USGS.” BiOp at 206. The BiOp also noted that the
CALSIM II simulation provided “an imperfect representation
of the pre-POD [pelagic organism decline]” environmental
situation and that supplemental analysis was needed “to
compensate for this modeling limitation.” BiOp at 205. In

This analysis of the effects of proposed CVP and SWP
operations on the delta smelt and its critical habitat uses
a combination of available tools and data, including the
CALSIM II model outputs provided in the appendices
of Reclamations’ 2008 biological assessment, historical
hydrologic data provided in the DAYFLOW database,
statistical summaries derived from 936 unique 90-day
particle tracking simulations published by Kimmerer
and Nobriga (2008), and statistical summaries and
derivative analyses of hydrodynamic and fishers data
published by Feyrer et al. (2007), Kimmerer (2008),
and Grimaldo et al. (accepted manuscript).

BiOp at 204.
80 SAN LUIS V. JEWELL

other words, when the FWS used CALSIM II as both a
baseline and as a predictor, it appeared to yield inaccurate
results, so the FWS used the next best available baseline: the
DAYFLOW model. Thus, the FWS explained: the
“[CALSIM II] Study 7.0 was the model run that Reclamation
and DWR thought best represented current operations, and
was thus intended as a ‘current baseline.’ However, due to
limitations of CALSIM II to accurately model actual
operations, [the FWS] also used the 1967–2007 DAYFLOW
summaries . . . to compare against CALSIM II outputs.” BiOp
at 207.

The district court acknowledged that “[t]he theoretical
problems with using a Calsim II to Calsim II comparison
were manifest,” yet it found that the “FWS’s decision to use
a Calsim II to Dayflow comparison . . . without attempting to
calibrate the two models . . . was arbitrary and capricious and
ignored the best available science showing that a bias was
present.” San Luis & Delta-Mendota, 760 F. Supp. 2d at
906–07. The evidence for the modeling bias in the CALSIM
II/DAYFLOW comparison was not clearly identified in the
comments sent to the FWS before it issued the BiOp, and the
comments largely recommended correcting the bias by
comparing one CALSIM II model to another CALSIM II
model—precisely the comparison the FWS found flawed and
inaccurate. For example, San Luis & Delta-Mendota Water
Authority, one of the plaintiffs-appellees here, submitted
extensive and thoughtful comments on a portion of the draft
BiOp and peer review analysis. The water authority
expressed its concern that it was “methodologically
inappropriate to compare historical data to simulated data.
Simulated data must be compared to simulated data.” The
DWR advised the FWS of some of the same problems. In its
comments on a portion of the draft BiOp, the DWR pointed
 SAN LUIS V. JEWELL 81

out that the model made assumptions that might not be borne
out in the historical data and that “[g]reat caution should be
taken when comparing actual data to modeled data.” Like the
water authority, it too recommended that CALSIM II
modeling “be used to compare one set of model runs to
another.” In later comments, the DWR repeated its concern
that the BiOp “compared model runs of future operation with
historical conditions” and suggested that “[i]mpact analyses
often compare only model scenarios to avoid these . . .
problems.” The independent peer review panel also
expressed some concern with comparing the historical
baseline with the CALSIM II simulated results. It pointed out
that the “large difference between [CALSIM II] results and
the historical baseline conditions defined with data can
confuse the comparisons of metrics . . . between a simulated
study and historical baseline.” The panel suggested that
“[i]deally, a model-simulated baseline should be available
that is consistent with the historical data . . . . It is unfortunate
that model-generated baselines with a high degree of
reliability were not made available for this analysis.”

The post-hoc views of the court’s experts reflect some
similar concerns. Dr. Quinn agreed that a comparison
between models with identical databases and assumptions
was preferable, but advised that “[a]s long as we bear in mind
the fact that these are two very different models, I do not see
why we cannot compare them.” After discussing some of the
data provided to him, he concluded that

the outputs from the two models cannot be
used interchangeably for estimating either X2
(in km) or flow (in cfs). This does not reflect
any criticism of either model. Their inner
workings are apparently quite different, as are
82 SAN LUIS V. JEWELL

their fundamental purposes, as explained to
us. However any comparisons between them
must explicitly account for the differences.

Similarly, Dr. Punt stated that

[i]n principle, there is nothing wrong with
fitting a model using a set of OMR/X2 values
from one model and making predictions using
OMR/X2 values which are based on the
output from a different model, as long as the
two sets of values are calibrated . . . .

Dr. Punt noted that the FWS had articulated its assumptions
and explained why he thought the FWS should have
performed additional calibration:

It is recognized in the record that the modeled
X2 does not reflect the “historical” X2 (BiOp
Figure E-28), and the BiOp does compare
historical and CALSIM-predicted X2 values
by month (Figure E-26). However, the BiOp
does not make this comparison for
comparable years. Failure to attempt
examination of whether it is necessary to
calibrate the historical data and the CALSIM
output would not normally be considered
appropriate scientific practice in the field.

(internal citations omitted).

We recognize that the FWS’s decision to compare its
chosen baseline, produced by DAYFLOW, to its future
projection, produced by CALSIM II, was not perfect—as
 SAN LUIS V. JEWELL 83

everyone has acknowledged. We nevertheless hold that this
comparison was not arbitrary and capricious. The BiOp
explained its assumptions and explained why it rejected the
CALSIM-to-CALSIM comparison suggested by the DWR
and others who reviewed the draft BiOp. The fact that the
FWS chose one flawed model over another flawed model is
the kind of judgment to which we must defer. As we said in
Environmental Defense Center, Inc. v. EPA: “We defer to an
agency decision not to invest the resources necessary to
conduct the perfect study, and we defer to a decision to use
available data unless there is no rational relationship between
the means [the FWS] use[d] to account for any imperfections
in its data and the situation to which those means are
applied.” 344 F.3d 832, 872 (9th Cir. 2003) (citations
omitted). “The existence of a flaw . . . does not require us to
hold that the agency’s use of the model was arbitrary.” Am.
Iron & Steel Inst. v. EPA, 115 F.3d 979, 1005 (D.C. Cir.
1997) (per curiam). Rather, we will “reject an agency’s
choice of a scientific model ‘only when the model bears no
rational relationship to the characteristics of the data to which
it is applied.’” Nat’l Wildlife Fed’n v. EPA, 286 F.3d 554, 565
(D.C. Cir. 2002) (quoting Appalachian Power Co. v. EPA,
135 F.3d 791, 802 (D.C. Cir. 1998)); accord Envtl. Def. Ctr.,
344 F.3d at 872. That is not the case here. The district
court’s determination to the contrary was heavily influenced
by experts supplied by the parties which, for the reasons that
we have explained, was inappropriate. The record is less
certain than the district court was willing to admit.34 That

34
The district court refers to testimony from the parties’ experts in terms
that overstate what the record will bear. See, e.g., San Luis & Delta-
Mendota, 760 F. Supp. 2d at 909 (“undisputed expert testimony”), id.
(“[A]ll experts in this case agree”), id. at 912 (“[u]ndisputed expert
84 SAN LUIS V. JEWELL

appellees were able to produce post-hoc theories alluding to
the possibility of bias from such a comparison does not
invalidate the FWS’s choices where there is no indication that
a failure to calibrate would do more than add an uncertainty
factor to the results, as well as no indication that a de-biasing
calibration is technically feasible. Under our deferential
standard of review, we therefore hold that the FWS’s choice
and use of models was based on the best available science.

Ideally, the FWS would have thoroughly discussed its
reasoning with regard to possible issues arising from the use
of DAYFLOW with CALSIM II. But, the fact that the
FWS’s explanation for its choices does not fully address
every possible issue that flows from that choice does not
render the FWS’s determination unreasonable or
unsupported. We do not require agencies to analyze every
potential consequence of every choice they make; to do so
would put an impossible burden on agencies. Rather, “we
review all agency choices with respect to models,
methodologies, and weighing scientific evidence” to ensure
that the agency’s “choices [are] supported by reasoned
analysis.” Ecology Ctr. v. Castaneda, 574 F.3d 652, 665 (9th
Cir. 2009). Particularly in an area as “unwieldy and science-
driven” as this, the FWS’s statistical modeling “does not
easily lend itself to judicial review.” Appalachian Power,
135 F.3d at 802; see id. (“Statistical analysis is perhaps the
prime example of those areas of technical wilderness into
which judicial expeditions are best limited to ascertaining the
lay of the land.”). “[T]hat some or many [experts] would
disapprove [of the FWS’s] approach does not answer the
question presented to us. In reviewing [the FWS’s BiOp], we

testimony”), id. at 922 (“the Calsim II to Dayflow comparison has the
potential to introduce significant, if not overwhelming, bias”).
 SAN LUIS V. JEWELL 85

do not sit as a panel of referees on a professional [scientific]
journal, but as a panel of generalist judges obliged to defer to
a reasonable judgment by an agency acting pursuant to
congressionally delegated authority.” City of L.A. v. Dep’t of
Transp., 165 F.3d 972, 977 (D.C. Cir. 1999). In this instance,
the FWS has provided a reasoned analysis explaining why the
DAYFLOW model was used as the baseline.

Consequently, we hold that the FWS did not act
arbitrarily and capriciously in relying on the CALSIM II and
DAYFLOW models in evaluating the impacts of Project
operations on the location of X2, or in justifying Action 4.

2. The BiOp sufficiently explained the fall X2 locations

The district court held that the BiOp “fails to explain why
it is essential to maintain X2 at 74 km and 81 km,
respectively, as opposed to any other specific location.” San
Luis & Delta-Mendota, 760 F. Supp. 2d at 923. Specifically,
the court found the “Federal Defendants have not identified
any record evidence that provides such an explanation. This
total lack of explanation violates the APA[].” Id. at 922.
With respect to the latter finding, the record is contrary to the
district court’s finding. There is record support for the
BiOp’s proposal. Whether the record will “explain why it is
essential to maintain X2 at 74 km and 81 km . . . as opposed
to any other specific location” is a different question. In fact,
it is the wrong question.

As we have previously explained, as the combined
pumping operations of the SWP/CVP remove hundreds of
thousands of gallons of fresh water from the Bay-Delta,
X2—the salinity-defined location of the smelt’s primary
spawning habitat—shifts eastward towards the delta. BiOp at
86 SAN LUIS V. JEWELL

373. As the ocean’s salty influence encroaches further
upstream, it mimics drought conditions in the Bay-Delta,
regardless of the previous season’s precipitation. BiOp at
271, 273–74 (plotting spring and fall X2 locations). Among
other things, this has resulted in an increasing divergence
between spring and fall X2 locations. The BiOp determined
that the “long-term upstream shift in X2 during fall has
caused a long-term decrease in habitat area availability for
delta smelt,” and it set forth an adaptive management
program to minimize the effects of Project pumping on X2:
RPA Action 4. BiOp at 374. Action 4 targets the fall
location of X2 during “wet” and “above normal” years: “the
years in which project operations have most significantly
adversely affected fall [X2 location].” BiOp at 373.
Specifically, Action 4 requires that fall X2 be maintained at
a location no greater than 74 km upstream from the Golden
Gate Bridge following “wet” water years, and no greater than
81 km upstream following “above normal” water years.
BiOp at 282–83.

The BiOp includes a number of different explanations,
both specific and general, for RPA Component 3 (Action 4)’s
regulation of X2 location. First, as discussed in the previous
section, the FWS used CALSIM II and DAYFLOW to give
it a picture of where X2 was located. Those models predicted
different values for the location of X2 based on differing
assumptions. DAYFLOW relied on historical data, while
CALSIM II predicted where X2 might be located based on
future Project operations. The FWS found that the two
models offered different estimates of X2 location, and that, in
general, “CALSIM II modeled scenarios were 10–15 percent
further upstream than actual historic X2.” BiOp at 235.
Thus, “[m]edian historic fall X2 was 79 km, while median
values for the CALSIM II modeled scenarios ranged from 87
 SAN LUIS V. JEWELL 87

to 91 km. The CALSIM II modeled scenarios all had an
upper range of X2 at about 90 km.” BiOp at 235. The FWS
generated numerous figures and charts with information taken
from these models, many of which graphed X2 location as a
function of another variable. See, e.g., BiOp at 260–61, 265,
270, 281, 282, 283, 284, 285. The FWS’s graphing range for
X2 location was 60–95 km, with most data points between
65–85 km, and a significant cluster between 70–85 km. BiOp
at 270.

Second, the data from the models was consistent with the
FWS’s written finding that “[d]uring the past 40 years,
monthly average X2 has varied from as far downstream as
San Pablo Bay (45 km) to as far upstream as Rio Vista on the
Sacramento River (95 km). . . . In general, delta smelt habitat
quality and surface area are greater when X2 is located in
Suisun Bay.” BiOp at 191. Elsewhere in the BiOp, the FWS
found based on an outside scientific study that “prior to
spawning entrainment vulnerability of adult delta smelt
increased at the SWP and CVP when X2 was upstream of
80km.” BiOp at 219.

Third, the BiOp pointed out that X2 varies in response to
a number of factors, both natural and man-made. The natural
factors include the tides and the spring inflow. The BiOp
found that “X2 is strongly influenced by tidal cycles, moving
twice daily up and downstream 6–10 km from its average
daily location.” BiOp at 372. The spring runoff affects X2 as
well. The BiOp found that “very high spring outflows have
always pushed X2 far downstream resulting in delta smelt
distributions distant from the influence of Banks and Jones.”
BiOp at 221. Aside from CVP/SWP’s operations, there are
other man-made factors to consider. For example, the Suisun
Marsh Salinity Control Gates have historically been
88 SAN LUIS V. JEWELL

operational for anywhere from 10 to 120 days a year. BiOp at
218. Operation of the salinity control gates may shift X2 as
far as 3 km upstream. BiOp at 218; see also BiOp at 241,
243–44. The BiOp noted that the 3 km upstream shift was
preferable to the “10–20 km shifts that have occurred for up
to 120 or more days per year during late summer through
early winter due to South Delta diversions.” BiOp at 219.

Fourth, the BiOp also points out that not all habitats in the
Bay-Delta are equally suited to the smelt, so that as X2
location shifts, the smelt may be affected by other factors as
well. For example, as X2 location shifts, the delta smelt
encounter changes in agricultural runoff; changes in turbidity,
which affects smelt feeding and predation; and exposure to
predators. E.g., BiOp at 372 (“the daily fluctuation in X2
around an upstream point such as Brown’s Island confines the
[smelt] population to narrow channels, where delta smelt may
be exposed to more stressors (e.g., agricultural diversions,
predation) relative to a downstream X2”); BiOp at 238
(“[T]he eastward movement of X2 will shift the distribution
of delta smelt upstream, and provide environmental
conditions for nonnative fishes that thrive in stable
conditions.” (internal citation omitted)). Thus, the BiOp
points out that although CVP/SWP operations remain the
most important factor affecting smelt habitat, BiOp at
177–79, “there is no single driver of delta smelt population
dynamics,” BiOp at 238; see also BiOp at 202 (referring to
“the multitude of factors that affect delta smelt population
dynamics”). The smelt habitat is a complex and dynamic
system.

Fifth, because X2 varies not only with CVP/SWP’s
operations, but with natural phenomena beyond the control of
the agencies—such as the tides, the seasons, and the annual
 SAN LUIS V. JEWELL 89

rainfall—the FWS had a range of choices for X2. Its goal
was to “move the habitat away from Delta impacts and into
broader open waters west of Sherman Island.”35 BiOp at 373.
Rather than trying to define a single point for X2, the FWS
chose a range, and refined the range by whether the year was
“wet” or “above normal.” We think there was ample evidence
in the record to support the choice of X2 between 74 and 81
km, depending on the season. The district court’s finding that
there was a “total lack of explanation” is belied by the
record.36

It appears the district court focused on whether there was
support in the record for the FWS’s choice of 74 km, as
opposed to 73 km or 75 km, hence its decision to question
whether the FWS had adequately explained “why it is
essential to maintain X2 at 74km and 81km . . . as opposed to
any other specific location.” San Luis & Delta-Mendota,
760 F. Supp. 2d at 923. This was not the district court’s
responsibility. The ESA provides that Reclamation had to
seek the FWS’s opinion to ensure that its operations were
“not likely to jeopardize the continued existence of any
endangered species or threatened species or result in the

35
Sherman Island lies at the confluence of the Sacramento and San
Joaquin Rivers. The 74–81 km range selected by the FWS lies west of
Sherman Island.
36
The district court’s conclusion is especially puzzling because it
accepted a similar FWS finding in its prior decision. See Kempthorne,
506 F. Supp. 2d at 335, 380 (noting that a survey of delta smelt abundance
“increases dramatically whenever X2 is located between Chipps and Roe
islands,” and that “[w]hen X2 is located upstream of Chipps Island, smelt
are vulnerable to entrainment and are located in an area that is not ideal
for feeding or protection”) (citing the administrative record). See also
BiOp at 114, 132.
90 SAN LUIS V. JEWELL

destruction or adverse modification of habitat of such
species.” 16 U.S.C. § 1536(a)(2). It is the FWS’s
responsibility to set forth its opinion and “those reasonable
and prudent alternatives which [the Secretary of the Interior]
believes would not violate [§ 1536(a)(2)] and can be taken by
[Reclamation].” Id. § 1536(b)(3)(A). A “reasonable and
prudent alternative” is a flexible standard for the consulting
agency; it is not the equivalent of the “least restrictive
alternative,” which is the way the district court treated the
inquiry. Sw. Ctr. for Biological Diversity, 143 F.3d at 523.
Under the ESA, the consulting agency is “not required to
explain why he chose one RPA over another,” nor is it
“required to pick the best alternative or the one that would
most effectively protect the [species] from jeopardy.” Id.
Rather, the FWS “need only have adopted a final RPA which
complied with the jeopardy standard and which could be
implemented by the agency.” Id. Accordingly, it was error
for the district court to hold that the FWS must “explain why
it is essential to maintain X2 at 74 km and 81 km . . . as
opposed to any other specific location.” San Luis & Delta-
Mendota, 760 F. Supp. 2d at 923.

Even if we thought the ESA demanded the kind of
precision insisted on by the district court, we believe there is
an explanation in the record for the FWS’s choice of 74 km
and 81 km as its seasonal parameters. And the explanation is
both logical and simple: At 74 km there is a monitoring
station for the Bay-Delta at Chipps Island; at 81 km there is
a monitoring station at Collinsville. BiOp at 114, 132. In its
comments to the FWS, the DWR questioned the feasibility of
locating fall X2, “especially if the spring X2 location is
 SAN LUIS V. JEWELL 91

significantly west of Port Chicago.”37 The DWR was
concerned with how the agencies would measure compliance
with any standard selected by the FWS and pointed out that
“it would be difficult to measure an X2 at 85 km, whereas it
would be much easier to measure at Collinsville (81 km) or
Emmaton (92 km).” Where X2 represents a range of choices,
choosing an X2 that can be measured and enforced is a
perfectly rational response. See Sw. Ctr. for Biological
Diversity, 143 F.3d at 523.

We recognize that the stakes are exceedingly high in this
case, but we conclude that the FWS’s choice, given the
record, represented a reasonable and prudent alternative.

C. The BiOp’s Incidental Take Statement Is Not Flawed

When the implementation of an RPA will cause
“incidental take” of a species, defined as “takings that result
from, but are not the purpose of, carrying out an otherwise
lawful activity conducted by the Federal agency or
applicant,” 50 C.F.R. § 402.02, the agency is required to
provide an incidental take statement (ITS) that “[s]pecifies
the impact, i.e., the amount or extent, of such incidental
taking on the species,” 50 C.F.R. § 402.14(i)(1)(i); see
16 U.S.C. § 1536(b)(4)(C)(i). The FWS determined that
“take of the delta smelt is likely to occur in the form of kill,
capture (via salvage), wound, harm, and harass as a result of
CVP/SWP operations within the action area,” BiOp at 286,
and prepared an ITS, BiOp at 285–310. In the ITS, the FWS
provided separate take limits for juvenile and adult smelt,
noting that “individuals of the larval/juvenile lifestage are

37
The FWS did not set its fall X2 target at Port Chicago, which is west
of the 74 km marker proposed by the FWS.
92 SAN LUIS V. JEWELL

less demographically significant than adults.” BiOp at 289.
The FWS used different data sets in determining these
separate limits. As the ITS notes, “[t]he mean values from
2005–2008 were used as an estimate of [juvenile] take under
the RPA,” BiOp at 289, while “[t]he average[] value for
[water years] 2006 to 2008" was used to calculate pre-
spawning adult delta smelt, BiOp at 287. In other words, the
year 2005 was used in calculating the juvenile, but not the
adult, take limit. BiOp at 287, 289. The ITS also
acknowledges its limitations: there are numerous
uncertainties inherent in the measurement of smelt take, and
“salvage data (our most definitive measurement endpoint)
reflects only a portion of the total mortality associated with
entrainment.” BiOp at 383.

The district court concluded that the BiOp’s ITS was
arbitrary or capricious for two reasons: First, the district court
found that the FWS failed to explain adequately its decision
to use 2005–2008 salvage data when setting the incidental
take limit for juvenile smelt, but to only use 2006–2008
salvage data when setting the incidental take limit for adult
smelt. San Luis & Delta-Mendota, 760 F. Supp. 2d at 925–28.
Second, the district court found that the FWS failed to explain
its decision to rely on an average cumulative salvage index
when it set a “Concern Level,” which, when triggered,
requires the Smelt Working Group to make “an immediate
specific recommendation to the [FWS].” BiOp at 387; see
San Luis & Delta-Mendota, 760 F. Supp. 2d at 928–29. We
hold that the ITS is not arbitrary and capricious because it
includes adequate explanation and support for its
determinations.
 SAN LUIS V. JEWELL 93

1. The ITS reasonably uses different data sets for adult
and juvenile take limits

First, the district court objected to the FWS’s choice of
data. The ITS’s use of separate data sets to calculate the
separate incidental take limits for juvenile and adult smelt is
adequately explained in the ITS. In discussing the use of
“[t]he mean values from 2005–2008” to calculate juvenile
take limits, the ITS explains that “[t]he reason for selecting
this span of years is that the apparent abundance of delta
smelt since 2005 as indexed by the 20-mm Survey and the
[Summer Townet Survey] is the lowest on record.” BiOp at
289. Accordingly, “[i]t was necessary to separate out this
abundance variable, but also to account for other poorly
understood factors relating salvage to OMR, distribution, and
the extant conditions.” BiOp at 289. As noted elsewhere in
the BiOp, “[i]n contrast to adult delta smelt, there is no well
established index of larval and juvenile abundance to reliably
scale the take of this lifestage to abundance. . . . This should
be kept in mind . . . .” BiOp at 389. Therefore, faced with a
greater degree of uncertainty in calculating juvenile incidental
take, the FWS chose, conservatively, to incorporate an
additional year of data—a year in which smelt abundance was
“the lowest on record.”38 BiOp at 389–90. The BiOp is clear
that it used 2006–2008 data for adults because “these years
within the historic dataset best approximate expected
salvage,” BiOp at 287, and 2005–2008 data for juveniles
because juvenile smelt “are less demographically significant
than adults,” BiOp at 289, and therefore a larger data set

38
This determination came on the heels of the district court’s
invalidation of the 2005 BiOp’s ITS based on its failure to consider
record-low population abundance in setting take limits. See San Luis &
Delta-Mendota, 760 F. Supp. 2d at 918.
94 SAN LUIS V. JEWELL

would be preferable.39 Such a decision to use a more
conservative data set, when necessary, is exactly the sort that
we afford agencies discretion to make. See, e.g., Fishermen’s
Finest, Inc. v. Locke, 593 F.3d 886, 893, 896–97 (9th Cir.
2010) (upholding agency’s decision to rely on some data
while disregarding other data).

2. The FWS reasonably uses an average cumulative
salvage index

Second, the district court objected to the FWS’s decision
to use an average cumulative salvage index to create its
“Concern Level.” BiOp at 387, 389–90. The Concern Level
“indicate[s] salvage levels approaching the take threshhold.”
BiOp at 387; see also BiOp at 289 (noting that the estimated
number represents a concern level where “entrainment has
reached high enough numbers to indicate the need for more
protective OMR restrictions.”). When the Concern Level is
reached, two actions follow: The Smelt Working Group must
convene and make “an immediate specific recommendation
to the [FWS],” and OMR flows may have to be adjusted “to
a more restrictive level.” BiOp at 387. The Concern Level
does not trigger automatic restrictions in OMR flows. It
“may require” a reduction “unless available data indicate
some greater level of exports is possible without increasing
entrainment.” BiOp at 387 (emphasis added).

The FWS’s decision to use an average in setting the
incidental take limits is, like its choice of data sets, a choice

39
For similar reasons we disagree with the district court that the FWS
must explain why it decided to include 2006 when it calculated
larval/juvenile abundance. See San Luis & Delta-Mendota, 760 F. Supp.
2d at 917–18.
 SAN LUIS V. JEWELL 95

entitled to substantial difference. See The Lands Council v.
McNair, 537 F.3d 981, 991 (9th Cir. 2008) (noting that it is
“well-established law” that courts owe deference “to agencies
and their methodological choices”). As appellees note, the
ITS’s “use of averaging . . . provides an ‘estimate’ of
expected take that, based on the historical record of salvage,
when applied[,] would likely be exceeded in many years.” In
other words, appellees object to the fact that the ITS’s choice
of methodology results in too restrictive of a management
regime. The district court agreed, finding that “[t]he record
does not explain why an ‘averaging’ approach was used.” San
Luis & Delta-Mendota, 760 F. Supp. 2d at 929 (“Based on
known adverse water supply consequences of operating the
Projects in a ‘constrained’ manner, it is inexplicable that the
FWS did not provide a clear and rational explanation of how
the ITS is set.”).

We hold that the record offers an adequate explanation.
As with its decision to use an additional year of data when
calculating the juvenile take limit, the ITS’s use of an
averaging methodology “counteract[s] the uncertainties”
inherent in its analyses by “overestimat[ing]” known
parameters. See Baltimore Gas & Elec. Co., 462 U.S. at 103.
In the context of the ITS’s repeated recognition of the
“difficult[y] [in] definitively project[ing]” the “specific level
of take of adult delta smelt at the CVP/SWP pumping
facilities . . . due to inherent uncertainties,” BiOp at 383, the
FWS’s use of an averaging methodology—that, by its nature,
yields conservative limits that otherwise would have been
exceeded in eleven of the past sixteen years, San Luis &
Delta-Mendota, 760 F. Supp. 2d at 929—is a discernable, and
justified path. See Bowman Transp., 419 U.S. at 285–86; see
also Nw. Coal. for Alternatives to Pesticides, 544 F.3d at
1050 (upholding an agency’s discernable reliance on models
96 SAN LUIS V. JEWELL

that “yield conservative data [where] the models
incorporate[d] the higher of [the potential values] in assessing
the overall risk”). Moreover, the establishment of a Concern
Level is an enforcement norm, and thus a policy decision.
“[S]election of an action level is primarily a legislative policy
decision that we will uphold so long as it was reasonably
drawn from the record.” Public Citizen Health Res. Group v.
Dep’t of Labor, 557 F.3d 165, 184 (3d Cir. 2009).
Ultimately, the appellees’ objection is not really to the record
support for the Concern Level, but to the Concern Level
itself. The choice of the Concern Level is quintessentially
one within the discretion of the agency, and we have no basis
for disturbing it here.

D. The Record Supports the BiOp’s Conclusions Regarding
the Indirect Effects of Project Operations

The FWS is required to take into account both the “direct
and indirect effects of an action on the species or critical
habitat” when determining whether an action is likely to
cause jeopardy. 50 C.F.R. § 402.02; see also 16 U.S.C.
§ 1536(a)(2). “Indirect effects” are “those that are caused by
the proposed action and are later in time, but still are
reasonably certain to occur.” 50 C.F.R. § 402.02. Here, the
BiOp identified a number of indirect effects on the delta
smelt from Project operations. Among other things, it
concluded that Project operations were reasonably certain to
(1) limit delta smelt food supply, and (2) increase harmful
pollution and contaminants. It also noted (3) the harmful
indirect effects that would likely spring from three “other
stressors”: predation, aquatic macrophytes, and microcystis.
See BiOp at 182–88, 202.
 SAN LUIS V. JEWELL 97

The BiOp found that a “multitude of factors . . . affect
delta smelt population dynamics including predation,
contaminants, introduced species, entrainment, habitat
suitability, food supply, aquatic macrophytes, and
microcystis.” BiOp at 202. It concluded that “[t]he extent to
which these factors adversely affect delta smelt is related to
hydrodynamic conditions in the Delta, which in turn are
controlled to a large extent by CVP and SWP operations.”
BiOp at 202. It noted that there were other sources of water
diversion that affect smelt through entrainment, but that, even
when “taken together,” they did not “approach[] the influence
of the Banks and Jones export facilities.” BiOp at 202.
Although the BiOp candidly assessed that there was “no
single primary driver of delta smelt population dynamics”
and that there were “non-CVP/SWP factors,” it ultimately
concluded that “the CVP and SWP are a primary driver of
delta smelt abiotic and biotic habitat suitability, health, and
mortality” and that CVP/SWP operations “have also played
an indirect role in the delta smelt’s decline by creating an
altered environment in the Delta that has fostered the
establishment of non-indigenous species and exacerbates
these and other stressors that are adversely impacting delta
smelt.” BiOp at 202–03.

The district court was not persuaded, finding that “[t]he
record does not support the BiOp’s conclusion that food web
and pollutants/contaminant impacts are indirect effects of
Project operations.” San Luis & Delta-Mendota, 760 F. Supp.
2d at 969. The district court also separately noted that “the
BiOp’s conclusions about the causal connections between
Project operations and ‘other stressors’ are ambiguous,” and
thought the effects were “unsupported by record evidence
and/or explanation” Id. We address each of these indirect
effects in turn, and conclude that the BiOp analysis was
98 SAN LUIS V. JEWELL

sufficiently clear and thorough so as not to be arbitrary and
capricious, and that it was based on the best available science.
See Kern, 450 F.3d at 1080–81.

1. Project operations indirectly affect smelt food supply

First, the BiOp explores the likely impact of Project
operations on smelt food supply. The BiOp describes
“entrainment of Pseudodiaptomus forbesi (P. forbesi), the
primary prey of delta smelt,” as a primary effect that “will
adversely affect delta smelt.”40 BiOp at 203–04. Plainly
stated, delta smelt appear severely food limited much of the
time. P. forbesi is the smelt’s principal food source from
summer to early fall, and so any Project impact on the
availability of P. forbesi represents a threat to the smelt. See
BiOp 228, 380. The BiOp concludes that such a threat is
present: high water exports reduce flows that would otherwise
transport P. forbesi into the delta smelt’s habitat, thereby
contributing to smelt mortality and population declines. BiOp
at 184–85, 228. Moreover, as water is pumped from the
south Bay-Delta during June through September, P. forbesi
are entrained in pumping stations, thereby reducing the
overall availability of P. forbesi in the delta. BiOp at 228.
Because “statistical evidence suggest[s] that the co-
occurrence of delta smelt and [prey such as P. forbesi] has a
strong . . . influence on the survival of young delta smelt from
summer to fall,” BiOp at 228, the BiOp concluded that the
Project’s effects on the P. forbesi population indirectly
threaten the viability of delta smelt.

40
The FWS proposed to address this through Action 6, which requires
the creation or restoration of 8,000 acres of habitat in the Delta and in
Suisun Marsh. BiOp at 381.
 SAN LUIS V. JEWELL 99

We consider whether this analysis is supported by the best
available science, and whether the FWS acted arbitrarily and
capriciously in concluding, to a “reasonable certainty,” that
Project operations indirectly affect delta smelt through its
food supply. See 50 C.F.R. § 402.02. As both the district
court and the peer review panel indicated, the FWS’s analysis
suffers from some methodological limitations; some of the
analyses underlying the agency’s conclusions involved
extrapolating from a few sampling sites, and therefore may
have overstated the scarcity of P. forbesi. The district court
seized on the peer review panel’s comments, which pointed
out: “Rather than correct this problem, FWS’s response was
to abandon the quantitative analysis, choosing to advance the
same, potentially flawed conclusion in a more subjective,
qualitative analysis. This conduct suggests another unlawful,
results-driven choice, ignoring best available science.” San
Luis & Delta-Mendota, 760 F. Supp. 2d at 940. We think the
district court’s criticism is incorrect.

The independent peer review panel commented directly
on the FWS’s analysis of the impact of CVP/SWP on P.
forbesi and the BiOp’s conclusion that the abundance of P.
forbesi “may vary inversely with export flow . . . and directly
with outflow.” It stated: “The [p]anel agrees with this
conceptual model and with the justification of its elements,
which are well-supported.” The panel offered technical
suggestions and then commented, “the figures meant to
support this analysis are not convincing.” The panel then
“suggest[ed] that this analysis be redone with the above
considerations in mind. If [the] revised analysis does not
show a substantial (not necessarily statistically significant)
pattern, the analysis should be mentioned but the results
dropped as quantitative metric from the EA.” (emphasis
added). The FWS did exactly as the panel recommended. It
100 SAN LUIS V. JEWELL

omitted the statistical analysis as justification for its
conclusion.

We cannot see the error in the FWS following the
recommendation of the peer review panel. Moreover, the
FWS had other reasons, explained in the record, for
concluding that CVP/SWP operations had an effect on P.
forbesi and that the abundance of P. forbesi indirectly affects
the delta smelt. See BiOp at 184–85, 228. That it omitted a
statistical study because it did not have sufficient data to
justify it does not strike us as a “results-driven choice,” but as
responsible science. Nothing in the ESA compelled the FWS
to conduct the particular study the peer review panel thought
inadequately supported by the data, and nothing in the peer
review panel’s comments even hints that the statistics
suggested a contrary conclusion—there is no evidence here
to suggest that the FWS, in conspiracy with the peer review
panel, was trying to hide evidence. Even if we thought that
a “rigorous, large-scale study . . . would be preferable,” we
have no authority to compel one: “in the absence of such a
study,” even “credible anecdotal evidence” can “represent[]
the best scientific . . . data available.” Nw. Ecosystem
Alliance v. FWS, 475 F.3d 1136, 1147 (9th Cir. 2007)
(internal quotation marks omitted).

Like the peer review panel in question, we conclude that
although the FWS’s analysis here is not without its
limitations, it is “based upon the best available science.” See
Nw. Ecosystem Alliance, 475 F.3d at 1147. Similarly, we are
persuaded that the BiOp’s conclusion that Project operations
indirectly affect the delta smelt by impacting the smelt food
supply was “well supported.” We need not independently
conclude that this conclusion is well supported to a
“reasonable certainty.” It is sufficient that we conclude that
 SAN LUIS V. JEWELL 101

the BiOp’s conclusion was sufficiently supported such that
the FWS did not arbitrarily and capriciously find this to be an
indirect effect to a reasonable certainty.

2. Project operations indirectly affect the smelt through
water contamination

Second, the BiOp explores the Project’s impact on water
contamination and therefore, indirectly, on smelt population
viability. The BiOp first concludes that water contamination
poses a threat to the delta smelt. Smelt throughout the Delta
are exposed to various pesticides and contaminants, which
“may affect embryo survival or inhibit prey production.” The
BiOp singles out ammonia released from a waste processing
facility in Sacramento and pesticides from agricultural
operations. BiOp at 153, 187, 237. The BiOp observes that
“concern over contaminants in the Delta is not new” and
refers to mercury, selenium, pesticides, herbicides,
ammonium concentrations, and undiluted drainwater. BiOp
at 187. Reclamation observes that the “delta smelt are highly
sensitive to high levels of ammonia” and that such
contaminants are detrimental to the health of the smelt
population because they render smelt susceptible to disease.
BiOp at 187–88.

The BiOp concludes that Project operations will
dangerously increase the impact of contaminants on the
smelt. One reason is topographical: as Project operations
constrain the smelt habitat to smaller rivers and estuaries, the
smelt’s overall exposure to these contaminants, which result
primarily from land runoff, increases. See BiOp at 153.
Relatedly, when Project operations reduce overall suitable
habitat, the impact of contaminants on the smelt in the
remaining habitat become intensified. The BiOp analyzes
102 SAN LUIS V. JEWELL

this latter effect in the context of recognizing the importance
of fall X2 to delta smelt viability: The FWS described two
“[p]otential mechanisms for the observed effect,” one of
which was that “a more confined distribution may increase
the impact of stochastic events that increase mortality rates of
delta smelt . . . includ[ing] . . . anthropogenic effects such as
contaminants.” BiOp at 234 (citing Sommer et al. 2007). One
other harmful effect of Project operations, the FWS notes,
comes from the flow created by Project pumping: Project-
related flow “increase[s] exposure to many pesticides [during
spawning].” BiOp at 153. This is because flows can
“mobilize contaminants.” BiOp at 240.

The district court concluded that “[i]t is not clear how the
BiOp or any other document in the record links the impacts
of contaminants to Project Operations.” San Luis & Delta-
Mendota, 760 F. Supp. 2d at 942. This criticism is not well-
founded. Although the FWS recognized that “contaminant
loading and its ecosystem effects within the Delta are not well
understood,” BiOp at 186, the fact that science must advance
further before the complicated ecosystem interactions in the
Bay-Delta are fully understood does not necessarily mean that
the FWS failed to rely on the best available science, or that it
arbitrarily and capriciously concluded that there was a
reasonable certainty that Project operations will indirectly
affect smelt through water contamination. Appellees would
presumably have us hold that it is impossible for an agency
to simultaneously recognize that some characteristics of an
indirect effect “are not well understood” or are “highly
uncertain,” and that it is reasonably certain that those indirect
effects are harmful and will result from the actions at issue.
We decline to do so. We are confident that if we returned the
BiOp to the agency, we could help the agency improve it by
“point[ing] out errors and missing information” and
 SAN LUIS V. JEWELL 103

“insist[ing] on additional detail.” Churchill Cnty. v. Norton,
276 F.3d 1060, 1081 (9th Cir. 2001). But “[t]hat is not our
role, of course.” Id. Instead, we hold that the BiOp has
sufficiently explained the harmful relation between Project
operations, contaminants, and delta smelt such that the FWS
did not arbitrarily and capriciously conclude to a reasonable
certainty that Project operations contribute to harmful
contaminant-related indirect effects.

3. Project operations indirectly affect the smelt through
the “other stressors” of predation, macrophytes, and
microcystis

As part of its analysis of the impacts of Project operations
on the smelt, the BiOp discusses the effect of “other
stressors”: “the multitude of factors that affect delta smelt
population dynamics including predation, . . . aquatic
macrophytes, and microcystis.” BiOp at 202. Because “[t]he
extent to which these factors adversely affect delta smelt is
related to hydrodynamic conditions in the Delta, which in
turn are controlled to a large extent by CVP and SWP
operations,” BiOp at 202, the BiOp’s analysis considers these
“other stressors” when determining “the effects of proposed
CVP/SWP operations on delta smelt,” BiOp at 203. We hold
first that the BiOp’s conclusion that “hydrodynamic
conditions driven or influenced by CVP/SWP operations . . .
influence the dynamics of delta smelt interaction with these
other stressors” was sufficiently supported in the record.
BiOp at 202. The BiOp explicitly relied on outside scientific
studies in concluding that Project operations “affect[] or
control[]” some of these other stressors impacting delta smelt
abundance. BiOp at 203. Second, having determined that
Project operations likely affect the impact of “other stressors”
104 SAN LUIS V. JEWELL

on the smelt, the BiOp considers the harmful effects of these
other stressors on the smelt population.

The district court faulted the FWS for failing to consider
“available information,” San Luis & Delta-Mendota, 760 F.
Supp. 2d at 934–35, or “make a rational connection between
the facts in the record and its conclusions,” id. at 936. We
disagree, and find the BiOp sufficiently thorough and hold
that the FWS did not arbitrarily and capriciously reach this
conclusion.

a. Predation

One stressor to the smelt is predation. As the BiOp
acknowledged, there is much here that is unknown. It is
known that in the 1960s, when the delta smelt were more
plentiful, they were prey for striped bass, black crappie, and
white catfish. BiOp at 183. The BiOp observed that “[i]t is
unknown whether incidental predation by striped bass (and
other lesser predators) represents a substantial source of
mortality for delta smelt,” in part because the scarcity of
smelt means that smelt have recently gone undetected in
recent studies of predator stomach contents. BiOp at 183.
The BiOp speculated that “[d]elta smelt may experience high
predation mortality around water diversion where smelt are
entrained and predators aggregate,” citing to an outside study
showing that smelt eggs and larvae were prey for inland
silversides. BiOp at 183. The BiOp also recognized the risks
presented from other potential predators of smelt eggs and
 SAN LUIS V. JEWELL 105

larvae in such areas, including yellowfin goby, centrarchids,
and Chinook salmon.41 BiOp at 183.

After concluding that predation poses a harmful indirect
effect to delta smelt population viability, the BiOp discusses
the relation between Project operations and smelt predation.
By shifting X2 further upstream, Project operations move
smelt habitat to include more of these littoral areas (areas of
the Bay-Delta close to shore), thereby increasing smelt
exposure to predators. BiOp at 153. As the location of X2
shifts towards new Bay-Delta regions that are rife with
predators the smelt follow. Moreover, the BiOp notes that
“[t]he Delta-wide increase in water transparency may have
intensified predation pressures on delta smelt,” citing the
“[w]ide documenta[tion],” including several specific studies,
indicating that water clarity significantly influences predation
of pelagic fishes, including many smelt species. BiOp at 183.

The district court found this analysis lacking, specifically
citing the FWS’s failure to explain whether striped bass
predation on the smelt “should be considered significant.”
San Luis & Delta-Mendota, 760 F. Supp. 2d at 934. Here, the
district court was very specific: It faulted the BiOp for failing
to “include any estimates of the effect of predation on the
delta smelt population” when “[s]uch information was
available” and was “decidedly contrary to BiOp findings.”
Id. In support, the district court referenced a 1999 California

41
The BiOp considered whether delta smelt are affected by competition
from other fish, and cited one study suggesting that they were. The BiOp
did not rely on competition as a stressor, however, because “there is no
empirical evidence to support the conclusion that competition between
these species is a factor that influences the abundance of delta smelt in the
wild.” BiOp at 183.
106 SAN LUIS V. JEWELL

Department of Fish & Game report submitted to the FWS as
part of an incidental take permit. Id. (citing California Dep’t
of Fish and Game, Conservation Plan for The California
Department of Fish and Game Striped Bass Management
Program (November 12, 1999)). We are not sure what the
Fish & Game report adds, and we cannot see that the report
is “decidedly contrary” to the FWS’s conclusions.

The Fish & Game report stated that “the best available
information yields imprecise, loosely constrained estimates
of striped bass predation on delta smelt.” It said that there
was considerable overlap between the delta smelt and striped
bass, but that “striped bass rarely ate delta smelt,” likely
because “delta smelt are surface oriented while striped bass
tend to forage near the bottom.” Fish & Game estimated that
striped bass were responsible for “an estimated annual
consumption of about 5.3% of the delta smelt population.”
We do not see the conflict between the Fish & Game report
and the BiOp. The BiOp concluded that “[i]t is unknown
whether incidental predation by striped bass . . . represents a
substantial source of mortality for delta smelt.” BiOp at 183.
The Fish & Game report is consistent with the FWS’s
conclusion. If anything, the Fish & Game report might
suggest that further study is required to see if the striped bass
is a substantial source of mortality for the smelt. Although
the BiOp cited post-1999 studies showing that the striped
bass was not a significant predator of the delta smelt, if Fish
& Game is correct (and current) in its estimates, a predator
responsible for a 5.3 percent mortality rate of an endangered
species might be significant. In any event, we fail to see its
significance with respect to the conclusions of the BiOp. The
BiOp concluded that predators—primarily fish other than the
striped bass—were a potential threat to the delta smelt. More
 SAN LUIS V. JEWELL 107

broadly, we decline to review with a fine-toothed comb the
studies on which the FWS relied in reaching its conclusions.

b. Aquatic Macrophytes

Another stressor noted in the BiOp is aquatic
macrophytes. Macrophytes—aquatic plants that grow in or
near water—have extensively colonized the interior Delta
over the past two decades. BiOp at 182. As the BiOp notes,
research suggests that these macrophytes have “altered fish
community dynamics in the Delta, including increasing
habitat for centrarchid fishes including largemouth bass,
reducing habitat for native fishes, and supporting a food web
pathway for centrarchids and other littoral fishes.” BiOp at
182 (citations to scientific studies omitted). These effects
impact smelt both directly and indirectly. Submerged aquatic
vegetation can overwhelm littoral habitats, such as inter-tidal
shoals and beaches, that serve as delta smelt spawning
locations, thereby rendering them unsuitable for spawning.
BiOp at 182. Moreover, macrophytes trap suspended
sediment and therefore reduce water turbidity, which has
contributed to a decrease in both juvenile and adult smelt
habitat while increasing the available habitat for fish that prey
on smelt. BiOp at 182–83 (citing Feyrer et al. 2007 and
Nobriga et al. 2008 in support). This decreased turbidity also
may facilitate the predation of delta smelt while hampering
the smelt’s own feeding, thereby further harming smelt
population viability. BiOp at 183.

As the BiOp notes, hydrologic conditions and water
temperature play a significant role in macrophyte
colonization of the Delta. BiOp at 182. The FWS concludes
that it is “likely” that Project operations’ impact on Bay-Delta
hydrologic conditions and reduction of seasonal flushing
108 SAN LUIS V. JEWELL

flows exacerbate the spread of macrophytes in the Bay-Delta.
BiOp at 277. Flushing flows are known to lead to “abrupt
changes in flow and turbidity.” BiOp at 146. Indeed, the
FWS found that “[a]vailable information is inconclusive
regarding the extent, magnitude and pathways by which delta
smelt may be affected by these stressors independent of
CVP/SWP operations.” BiOp at 277.

The district court was also not persuaded by this analysis,
finding that “[a]lthough a connection [between Project
operations and macrophytes] may exist, the record does not
reflect any discussion, nor have the parties pointed to any
study, connecting ‘seasonal flushing flows . . . the natural
frequency of upstream and downstream movement of the
LSZ, and lengthen[ed] upstream shifts of the LSZ’ to the
presence of any aquatic macrophyte.” San Luis & Delta-
Mendota, 760 F. Supp. 2d at 935–36. Again, we disagree.
The BiOp set forth several plausible explanations for how
Project operations will increase the detrimental impact of
macrophytes on delta smelt viability, and cited studies in
support. See BiOp at 146, 182–83, 277. That the BiOp did
not, as the district court requests, point to a study directly
addressing the Project’s effect on Bay-Delta macrophytes
does not render the FWS’s conclusions unreasonable or
unsupported: The FWS has drawn rational conclusions from
the best available science, and, consequently, we hold that the
BiOp’s determination that it is reasonably certain that
macrophytes will indirectly affect delta smelt is not arbitrary
and capricious. It is not our job to task the FWS with filling
the gaps in the scientific evidence. We must respect the
agency’s judgment even “in the face of uncertainty.” Ariz.
Cattle Growers Ass’n v. Salazar, 606 F.3d 1160, 1164 (9th
Cir. 2010).
 SAN LUIS V. JEWELL 109

c. Microcystis

Microcystis aeruginosa is a cyanobacterium that produces
toxins throughout its life cycle, with toxin concentrations
sharply increasing when the bacteria population dies, usually
in September or October. BiOp at 372. These high toxin
levels present a threat to the delta smelt and, as the BiOp
recognizes, high microcystis toxin levels have been associated
with low delta smelt abundances. BiOp at 372. Microcystis
can directly “pose animal and human health risks if contacted
or ingested directly,” although it does not appear that current
concentrations are sufficiently severe to threaten smelt. BiOp
at 186. Rather, microcystis’s primary threat to the smelt is
indirect, as “it appears that M. aeruginosa is toxic to
copepods that delta smelt eat.” BiOp at 186 (citing an outside
scientific study). There is also concern that microcystis
“could out-compete diatoms[, a rich food source for
zooplankton,] for light and nutrients.” BiOp at 186. As the
BiOp notes, however, more studies are needed, and, in fact,
“are underway to determine if zooplankton production is
compromised during M. aeruginosa blooms to an extent that
is likely to adversely affect delta smelt.” BiOp at 186.

As the BiOp also discusses, CVP/SWP operations are
likely to increase the harmful impact of microcystis on delta
smelt because “[l]ow flow conditions are among the factors
associated with Microcystis blooms.” BiOp at 372. By
reducing flows, Project operations would cause “larval and
juvenile delta smelt . . . [to] remain in the Central and South
Delta, where they could . . . succumb to predation or
microcystis blooms.” BiOp at 224. Overall, Microcystis
“reduce[s] habitat suitability.” BiOp at 373.
110 SAN LUIS V. JEWELL

The district court found that the BiOp “makes no
connection whatsoever between microcystis . . . and
continued CVP and SWP operation” and that “[g]iven that the
impacts of regulating Project Operations are so consequential,
such unsupported attributions (a result in search of a
rationale) are unconscionable.” San Luis & Delta-Mendota,
760 F. Supp. 2d at 936. Again, we disagree. The FWS has
proposed several plausible evidence-based hypotheses
explaining the harmful Project-related impact of
microcystis on delta smelt. The BiOp candidly acknowledges
that additional studies in this area are underway. That
CVP/SWP operations are not the only dynamic force acting
on the Bay-Delta does not render the BiOp’s recognition of
the inherent uncertainty associated with the highly
interdependent ecosystem unreasonable. We should not deter
agencies from recognizing the limitations of either science or
their own knowledge. In this instance, the evidence linking
Project operations, Bay-Delta hydrologic conditions, and
microcystis harms is sufficient that we hold that the FWS’s
microcystis conclusions were not arbitrary and capricious.

E. The FWS Is Not Required to Support the “Non-Jeopardy”
Elements of its RPA

When the Secretary determines that an agency action will
cause jeopardy to, or an adverse habitat modification of, an
endangered or threatened species, the Secretary “shall suggest
those reasonable and prudent alternatives which he believes
would not [jeopardize the species or adversely modify its
habitat] and can be taken by the Federal agency or applicant
in implementing the agency action.” 16 U.S.C.
§ 1536(b)(3)(A); see also 50 C.F.R. §§ 402.14(h)(3),
402.14(g)(5). The FWS’s regulations further explain its duty
 SAN LUIS V. JEWELL 111

under the ESA. The regulations define “reasonable and
prudent alternatives”—RPAs—as

alternative actions identified during formal
consultation [1] that can be implemented in a
manner consistent with the intended purpose
of the action, [2] that can be implemented
consistent with the scope of the Federal
agency’s legal authority and jurisdiction, [3]
that is economically and technologically
feasible, and [4] that the Director believes
would avoid the likelihood of jeopardizing the
continued existence of listed species or
resulting in the destruction or adverse
modification of critical habitat.

50 C.F.R. § 402.02. Element [4] in § 402.02 is commonly
referred to as the “jeopardy” factor; elements[1] through [3]
are referred to as the “non-jeopardy” factors. The FWS’s
Consultation Handbook explains further: “If the services
conclude that certain alternatives are available that would
avoid jeopardy and adverse modification, but such
alternatives fail to meet one of the other three elements in the
definition of ‘reasonable and prudent alternative,’ the
Services should document the alternative in the biological
opinion to show it was considered during the formal
consultation process.” U.S. Fish & Wildlife Serv. & Nat’l
Marine Fisheries Serv., ESA at 4–41 (March 1998), Section
7 Consultation Handbook, available at http://www.fws.gov/
endangered/esa-library/pdf/CH4.pdf (last visited July 27,
2013) (second emphasis added). Thus, according to the
Consultation Handbook, if a draft alternative fails to meet one
of the non-jeopardy “elements” of a valid RPA, the Service
should provide documentation to show that it considered
112 SAN LUIS V. JEWELL

alternatives during consultation. Id. We have previously
afforded Skidmore deference to the FWS’s Consultation
Handbook. See Ariz. Cattle Growers’ Ass’n, 606 F.3d at
1165 (concluding that “[t]he definition in the handbook
appears to be the result of the agency's considered judgment
and . . . we are persuaded [that it is a reasonable one entitled
to deference]”); see also Skidmore v. Swift & Co., 323 U.S.
134 (1944).

Referring to the non-jeopardy factors, the district court
found that the FWS “has articulated absolutely no connection
between the facts in the record and the required conclusion
that the RPA is (1) consistent with the purpose of the
underlying action; (2) consistent with the action agency’s
authority; and (3) economically and technologically feasible.”
San Luis & Delta-Mendota, 760 F. Supp. 2d at 956–57. The
court continued:

the APA requires, and the public is entitled
under the law to receive, some exposition in
the record of why the agency concluded (if it
did so at all) that all four regulatory
requirements for a valid RPA were satisfied.
The RPA Actions manifestly interdict the
water supply for domestic human
consumption and agricultural use for over
twenty million people who depend on the
Projects for their water supply. “Trust us” is
not acceptable. FWS has shown no
inclination to fully and honestly address water
supply needs beyond the species despite the
fact that its own regulation requires such
consideration.
 SAN LUIS V. JEWELL 113

San Luis & Delta-Mendota, 760 F. Supp. 2d at 957. Because
the FWS had failed to explain why it chose its RPAs, “to the
exclusion of implementing less harmful alternatives,” the
district court remanded to the FWS. Id. Put more simply, the
district court found that both the FWS’s regulation and the
APA required the FWS to engage in a record exposition of
the non-jeopardy factors, and that the FWS did not do so.

We disagree both with the district court’s legal analysis
and with its reading of the record.

First, contrary to the district court’s conclusion, the
FWS’s “own regulation” does not require the FWS to address
the non-jeopardy factors. San Luis & Delta-Mendota, 760 F.
Supp. 2d at 957. Nothing in § 402.02 obligates the FWS to
address the non-jeopardy factors when it proposes RPAs.
Section 402.02 is a definitional section; it is defining what
constitutes an RPA, not setting out hoops that the FWS must
jump through. See 50 C.F.R. § 402.02; see also id. at
§ 402.14(g)(5) (the FWS shall “discuss [with an agency] . . .
the availability of reasonable and prudent alternatives”);
302.02(h)(3) (“A ‘jeopardy’ biological opinion shall include
reasonable and prudent alternatives if any.”). Moreover, the
Consultation Handbook implies that no such discussion is
necessary. As the Handbook notes:

[Although] it is imperative that the opinion
contain a thorough explanation of how each
component of the [reasonable and prudent]
alternative is essential to avoid jeopardy
and/or adverse modification[,] . . . [i]f the
Services conclude that certain alternatives are
available that would avoid jeopardy and
adverse modification, but such alternatives
114 SAN LUIS V. JEWELL

fail to meet one of the other three elements in
the definition of ‘reasonable and prudent
alternative,’ the Services should document the
alternative in the biological opinion to show it
was considered during the formal consultation
process.

Section 7 Consultation Handbook, available at
http://www.fws.gov/endangered/esa-library/pdf/CH4.pdf (last
visited July 1, 2013); see also 50 C.F.R. § 402.14(h)(3) (“If
the Service is unable to develop such [reasonable and
prudent] alternatives, it [must] indicate that to the best of its
knowledge there are no reasonable and prudent
alternatives.”). In other words, a “thorough” documentation
of jeopardy/adverse modification in the BiOp is always
required, whereas documentation of the non-jeopardy factors
is only required when the RPA fails to meet a non-jeopardy
factor.

We fail to see anywhere that the FWS has required itself
to provide an explanation of the non-jeopardy factors when
it lays out an RPA. We may not “impose on the agency [our]
own notion of which procedures are ‘best’ or most likely to
further some vague, undefined public good. Nor may we
impose procedural requirements [not] explicitly enumerated
in the pertinent statutes.” McNair, 537 F.3d at 993 (internal
quotation marks and citations omitted).42

42
We note that the Fourth Circuit recently remanded a BiOp to the FWS
for failure to evaluate an RPA for its economic and technological
feasibility. Dow AgroSciences, 707 F.3d at 474–75. We do not read Dow
to require the FWS to address economic and technological feasibility as
a procedural matter. As we read Dow, the court was concerned that the
FWS had imposed an especially onerous requirement without any thought
for whether it was feasible. Id. at 475 (RPA would prohibit pesticide
 SAN LUIS V. JEWELL 115

Second, the APA does not, as the district court held,
require the FWS to address the non-jeopardy factors in this
case. See San Luis & Delta-Mendota, 760 F. Supp. 2d at 957.
Under the APA, the Supreme Court has held that agency
decisions that “entirely fail[] to consider an important aspect
of the problem” are arbitrary and capricious. Motor Vehicle
Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto Ins. Co.,
463 U.S. 29, 43 (1983). We have held that whether an RPA
will prevent jeopardy or adverse modification of critical
habitat is “an important aspect of the problem.” See, e.g.,
Wild Fish Conservancy v. Salazar, 628 F.3d 513, 522–23 (9th
Cir. 2010) (finding a BiOp that failed to explain how the RPA
avoided jeopardy arbitrary and capricious); Pacific Coast
Fed’n of Fishermen’s Ass’ns v. U.S. Bureau of Reclamation,
426 F.3d 1082, 1091 (9th Cir. 2005) (same). But the
jeopardy factor in the RPA is independently demanded by the
ESA itself. Section 1536(a)(2) requires that each federal
agency shall “insure that any action . . . is not likely to
jeopardize” the species or its habitat. 16 U.S.C. § 1536(a)(2).
This includes the FWS, which must warrant that its RPA
“would not violate [§ 1536(a)(2)].” 16 U.S.C.
§ 1536(b)(3)(a). That is, the FWS, in the course of proposing
an RPA, must insure that the RPA does not jeopardize the
species or its habitat. We can find no similar requirement in
the ESA that the FWS address the remaining three non-
jeopardy factors. If the ESA does not require it, we are
extremely reluctant to read such a requirement into the APA.

applications “within 500 feet (for ground applications) and 1,000 feet (for
aerial applications) of any waterway that is connected, directly or
indirectly, at any time of the year, to any water body in which salmonids
might be found at some point.” (emphases in original)).
116 SAN LUIS V. JEWELL

Moreover, we are persuaded that the district court misread
what the economic feasibility factor addresses. The court
faulted the FWS for not accounting for the cost of
“interdict[ing] the water supply for domestic human
consumption and agricultural use for over twenty million
people who depend on the Projects for their water supply.”
San Luis & Delta-Mendota, 760 F. Supp. 2d at 957. This
misreads the ESA and its implementing regulations. Section
402.02 is only concerned with the economic and
technological feasibility of the RPA. That is, the FWS must
consider whether its proposed alternative is financially and
technologically possible. Those two considerations—
economics and technology—are constraints on what
measures the FWS can recommend to the agency as an
alternative to ceasing the activity entirely. To put it into
perspective in this case: Reclamation has consulted with the
FWS because it has legitimate concerns whether its continued
CVP activities may jeopardize the smelt or its habitat. When
the FWS concludes that Reclamation’s continued activities
will jeopardize the smelt then, presumptively, Reclamation
may not take or continue such activities. See Nat’l Assn of
Home Builders v. Defenders of Wildlife, 551 U.S. 644, 652
(2007) (“Following the issuance of a ‘jeopardy’ opinion, the
agency must either terminate the action, implement the
proposed alternative, or seek an exemption from the Cabinet-
Level Endangered Species Committee . . . .”). In this case, of
course, terminating Reclamations’ CVP-related activities is
unthinkable. The whole point of the “reasonable and prudent
alternative” is for the FWS to suggest what Reclamation can
do to avoid such a result. The regulation identifies
“economic and technological feasibility” as factors because
these go to whether the RPA “can be taken by the Federal
agency . . in implementing the agency action,” 16 U.S.C.
§ 1536(b)(3)(A) (emphasis added), not to whether restricting
 SAN LUIS V. JEWELL 117

CVP activities will affect its consumers.43 The “economic
and technological feasibility” factor does not address the
downstream economic impacts of Reclamation being unable
to continue its CVP operations as it has done in the past. As
important and consequential as the question is, the FWS is
not responsible for balancing the life of the delta smelt
against the impact of restrictions on CVP/SWP operations.
That balance has already been struck by Congress in the ESA
and the Central Valley Project Improvement Act. See CVPIA
§ 3406(b), Pub. L. No. 102-575, 106 Stat. 4600, 4714 (stating
that the Secretary of the Interior is to “operate the Central
Valley Project to meet all obligations under State and Federal
law, including but not limited to the Federal Endangered
Species Act, 16 U.S.C. § 1531 et seq”); Tenn. Valley Auth.,
437 U.S. at 185 (holding that the ESA reflects “a conscious
decision by Congress to give endangered species priority over
the ‘primary missions’ of federal agencies”). Accordingly,
the FWS’s duty is to opine on the viability of the smelt and
“to halt and reverse the trend toward species extinction,
whatever the cost.” Id. at 184 (emphasis added).

Even if the APA did require the FWS to consider the non-
jeopardy factors, the record shows that the FWS has
sufficiently considered them. See Motor Vehicle Mfrs. Ass’n
of U.S., Inc, 463 U.S. at 43 (holding that insufficient
consideration for purposes of APA arbitrary and capricious
review is an “entire[] fail[ure] to consider”). Although the
FWS’s consideration of the non-jeopardy factors could
certainly have been even more exhaustive, or stated more
expressly, its determination that the RPA satisfied the non-

43
Neither the parties nor the district court argue that the RPAs
themselves (and their proposed Actions) are not economically and
technologically feasible.
118 SAN LUIS V. JEWELL

jeopardy factors “may be reasonably discerned” from the
record, and therefore should be upheld. Id. (holding that even
“a decision of less than ideal clarity” should be upheld in
such circumstances). Application of the non-jeopardy factors
in this case is really quite straightforward. See 50 C.F.R.
§ 402.02. First, the record shows that the RPA is consistent
with the purpose of the underlying action. The document that
was prepared as a basis for consultation, Reclamation’s
biological assessment (BA), identified the purpose of this
action to be “operat[ing] the [Projects] to divert, store, re-
divert, and convey CVP and SWP . . . water consistent with
applicable law.” The RPAs—which largely deal with
regulating the water that CVP/SWP export from the Delta-
Bay—do not require any major changes in the way
Reclamation runs its operations. Second, and similarly,
the record indicates that the RPA can be implemented
“consistent with the scope of the Federal agency’s legal
authority.” Both the BA and the BiOp discuss the extent of
Reclamation’s authority. See, e.g., BiOp at 21–25
(Reclamation’s obligations under its Coordinated Operations
Agreement with DWR). Finally, there is support in the
record for the FWS’s conclusion that the RPA is both
technologically and economically feasible. We think this is
nearly self-evident. The RPA closely resembles measures in
the interim remedial order, the feasibility of which was
proven in its mid-December 2007 through December 2008
implementation. BiOp at 327–28. Additionally, RPA
incorporates feasibility-related comments that were made on
a draft RPA from Reclamation and DWR. Again, the RPAs
propose regulatory changes in what Reclamation does on a
day-to-day basis, but the RPAs do not require major changes
affecting Reclamation’s ability—financially or
technologically—to comply with the RPAs.
 SAN LUIS V. JEWELL 119

In sum, we disagree with the district court’s determination
that the FWS’s own regulation and the APA require the FWS
to explain that the RPA satisfies § 402.02 non-jeopardy
factors. Alternatively, we hold that the FWS’s consideration
of these factors may be reasonably discerned from the record
to satisfy any explanation requirements.44

V. CROSS-APPEAL

The appellees raise three claims of error in the district
court: that the FWS violated the ESA by not separating
discretionary from nondiscretionary actions when it set the
environmental baseline; that Reclamation acted arbitrarily
and capriciously when it accepted the BiOp; and that the
FWS and Reclamation failed to prepare an environmental
impact statement, as required by NEPA. We consider each in
turn.

A. Segregating Discretionary From Nondiscretionary
Actions

When determining whether an agency’s operations are
likely to jeopardize the continued existence of a listed species
or result in the destruction or adverse modification of a
critical habitat, the FWS must “[e]valuate the effects of the
action and cumulative effects on the listed species or critical

44
We also hold that the FWS need not explain why it chose the RPA
measures over “less harmful alternatives.” San Luis & Delta-Mendota,
760 F. Supp. 2d at 957. In Southwest Center for Biological Diversity, we
held that “under the ESA, the Secretary was not required to explain why
he chose one RPA over another, or to justify his decision based solely on
apolitical factors.” 143 F.3d at 523. Consequently, we do not agree with
the district court that such a failure demands remand to the agency. San
Luis & Delta-Mendota, 760 F. Supp. 2d at 957.
120 SAN LUIS V. JEWELL

habitat.” 50 C.F.R. § 402.14(g)(3). This evaluation requires
determining an “environmental baseline,” which “does not
include the effects of the action under review,” and then
adding the direct and indirect effects of the proposed federal
action to determine whether that action will jeopardize a
listed species. See 50 C.F.R. § 402.02; Section 7 Consultation
Handbook, available at http://www.fws.gov/endangered/esa-
library/pdf/CH4.pdf; see also Interangency Cooperation—
Endangered Species Act of 1973 as Amended; Final Rule, 51
Fed. Reg. 19,926, 19,932 (June 3, 1986) (noting that the
environmental baseline “serve[s] as the baseline for
determining the effects of the action on the species or critical
habitat”).

In Home Builders, the Supreme Court held that the
consultation requirement in ESA § 7(a)(2), 16 U.S.C.
§ 1536(a)(2), does not “impliedly repeal[] nondiscretionary
statutory mandates, even when they might result in some
agency action.” Nat’l Ass’n of Home Builders, 551 U.S. at
665. In that case, the Clean Water Act required the EPA to
transfer certain permitting powers to state authorities if nine
criteria were satisfied. 33 U.S.C. § 1342(b). In accordance
with the ESA § 7(a)(2), the EPA consulted with the FWS.
The Service found that the transfer would not have any direct
effect on any species in Arizona listed under the ESA. The
FWS, however, was concerned that once EPA transferred
permitting authority to Arizona, the ESA would no longer
apply to permitting decisions, and Arizona could issue
permits without regard to listed species. The FWS concluded
that the transfer might have indirect effects on the species,
although no decision or action of EPA was in question, other
than the transfer of permitting authority. Id. at 653–54. The
Court held that the ESA did not function as an additional
constraint on the EPA’s duty to transfer permitting authority
 SAN LUIS V. JEWELL 121

to Arizona: “[Section] 7(a)(2)’s no-jeopardy duty covers only
discretionary agency actions and does not attach to actions
. . . that an agency is required by statute to undertake once
certain specified triggering events have occurred.” Id. at 669.

Relying on Home Builders, San Luis & Delta-Mendota
Water Authority asserts that “[a]s part of this [§ 7(a)(2)]
analysis, the FWS must, among other things, distinguish
between the discretionary and nondiscretionary actions of an
operation, so that only the discretionary actions are
considered as effects of the agency action.” This argument
was considered and rejected by the district court. See San
Luis & Delta-Mendota, 760 F. Supp. 2d at 947–48. The
district court reasoned that “Home Builders addressed
whether the section 7 consultation obligation attaches to a
particular agency at all.” San Luis & Delta-Mendota, 760 F.
Supp. 2d at 948. It did not, however, “address whether, once
section 7 consultation is triggered, the jeopardy analysis must
separately identify and segregate discretionary from non-
discretionary actions, relegating the non-discretionary actions
to the environmental baseline.” Id. (emphasis removed).

We agree with the district court’s analysis that Home
Builders does not require the agency to segregate
discretionary from non-discretionary actions when it
considers the environmental baseline. Home Builders dealt
only with whether § 7(a)(2) applies. The real question after
Home Builders is what counts as a non-discretionary action,
to which § 7(a)(2) does not apply.

We addressed this question in National Wildlife
Federation v. National Marine Fisheries Service, 524 F.3d
917 (9th Cir. 2008) (“NWF”). In NWF, the National Marine
Fisheries Service (NMFS) had issued a BiOp on the impacts
122 SAN LUIS V. JEWELL

of the operation of the Federal Columbia River Power System
(FCRPS) on a listed species. Id. at 921.

The 2004 BiOp’s jeopardy analysis included
in the environmental baseline for the proposed
action the existing FCRPS, various
supposedly nondiscretionary dam operations,
and all past and present impacts from
discretionary operations. . . . NMFS also
found, though, that certain aspects of FCRPS
operations—such as operations relating to
irrigation, flood control, and power
generation—were nondiscretionary, given the
dams’ existence, and that those aspects
should not be considered part of the action
under ESA review.

Id. at 926 (emphasis added). The agency “segregated its
analysis, first evaluating whether the proposed agency
action—consisting of only the proposed discretionary
operation of the FCRPS—would have an appreciable net
effect on a species.” Id. The agency did this “instead of
assessing whether the listed fish would be jeopardized by the
aggregate of the proposed agency action, the environmental
baseline, cumulative effects, and current status of the
species.” Id.

This approach, we held, was incorrect: “NMFS may not
avoid determining the limits of the action agencies’ discretion
by using a reference operation to sweep so-called
‘nondiscretionary’ operations into the environmental
baseline, thereby excluding them from the requisite ESA
jeopardy analysis.” Id. at 929. We distinguished NWF from
Home Builders on the basis of the specificity of the mandate
 SAN LUIS V. JEWELL 123

in question. “[I]n the present case Congress has imposed
broad mandates, rather than directing the agency to take
specific actions, and the agencies are perfectly capable of
simultaneously obeying Section 7 and those mandates.” Id.
at 928. We repeated that “in contrast [to Home Builders,]
Congress has imposed broad mandates which do not direct
agencies to perform any specific nondiscretionary actions, but
rather, are better characterized as directing the agencies to
achieve particular goals.” Id. at 928. Thus, “while the goals
themselves may be mandatory, the agencies retain
considerable discretion in choosing what specific actions to
take in order to implement them.” Id. at 929. “‘[A]n agency
cannot escape its obligation to comply with the ESA merely
because it is bound to comply with another statute that has
consistent, complementary objectives.’” Id. (quoting
Washington Toxics Coal. v. EPA, 413 F.3d 1024, 1032 (9th
Cir. 2005)).

The Water Authority has not pointed us to any statutory
obligation that Congress has imposed on Reclamation that is
both mandatory and inconsistent with its obligations under
the ESA.45 Like the FCRPS in NWF, Reclamation has a very
broad mandate. Moreover, Congress has stated, as clearly as
it can, that Reclamation is to administer its obligations to the
CVP consistent with the mandates of the ESA. CVPIA,
§ 3406(b), 106 Stat. at 4714 (stating that the Secretary of the
Interior is to “operate the Central Valley Project to meet all
obligations under State and Federal law, including but not

45
The Water Authority has pointed us to water contracts between
Reclamation and wildlife refuge contractors, water exchange contracts
with senior water rights holders, and a decision of the California State
Water Resources Control Board. These do not approach the statutory
mandate that the Court found EPA was under in Home Builders.
124 SAN LUIS V. JEWELL

limited to the Federal Endangered Species Act, 16 U.S.C.
§ 1531 et seq.”).

B. Reclamation Did Not Violate the ESA by Accepting the
2008 BiOp

The Water Authority argues that Reclamation committed
an independent violation of the ESA by relying on the BiOp.
See Defenders of Wildlife, 420 F.3d at 976 (noting that
“[a]rbitrarily and capriciously relying on a faulty Biological
Opinion violates [the action agency’s independent and
substantive] duty” (emphasis added)). Because we do not
believe the 2008 BiOp to be arbitrary and capricious, we join
the district court in declining to find that Reclamation’s
reliance on the BiOp was arbitrary and capricious. See San
Luis & Delta-Mendota, 760 F. Supp. 2d at 966–67.

C. Application of NEPA to the FWS and Reclamation

NEPA requires that “to the fullest extent possible . . . all
agencies of the Federal Government shall” complete an
environmental impact statement (EIS) in connection with
“every recommendation or report on proposals for legislation
and other major Federal actions significantly affecting the
quality of the human environment.” 42 U.S.C. § 4332(2)(C).
The agency may begin by preparing an environmental
assessment (EA). 40 C.F.R. §§ 1501.4(a)–(c); 1508.9. It may
then issue a finding of no significant impact (FONSI) rather
than completing an EIS if the EA reveals that the action in
question “will not have a significant effect on the human
environment.” 40 C.F.R. §§ 1501.4(e); 1508.13.

In this case, neither the FWS nor Reclamation prepared an
EA or an EIS. We are thus confronted with two questions
 SAN LUIS V. JEWELL 125

concerning the scope of NEPA’s requirement that agencies
produce an EA and, if necessary, an EIS. First, was the
FWS’s issuance of the BiOp a “major Federal action[]
significantly affecting the quality of the human environment”
that imposed on the FWS an obligation to comply with
NEPA? And second, was Reclamation’s provisional adoption
and implementation of the BiOp a “major Federal action[]
significantly affecting the quality of the human environment”
that imposed on Reclamation an obligation to comply with
NEPA?46

Before the district court, the plaintiffs initially alleged that
the FWS violated NEPA by failing to prepare an EIS when
issuing the 2008 BiOp. In their first amended complaint, the
San Luis plaintiffs further asserted that Reclamation violated
NEPA by accepting and implementing the BiOp without
completing an EIS. In three of the five consolidated cases,
the plaintiffs moved for summary judgment on their claims
that the FWS and Reclamation failed to comply with NEPA.
The defendant-intervenors, including NRDC, filed a cross-
motion for summary judgment, arguing that the agencies
were not obligated to adhere to NEPA in issuing or
implementing the BiOp. In November 2009, the district court
issued its decision on the parties’ cross-motions for summary
judgment on the NEPA issues. San Luis & Delta-Mendota
Water Auth. v. Salazar, 686 F. Supp. 2d 1026 (E.D. Cal.
2009). The district court observed that “[i]t is a close call

46
We review an agency’s decision that it need not prepare an EIS for
“reasonableness.” See Northcoast Envtl. Ctr. v. Glickman, 135 F.3d 660,
667 (9th Cir. 1998) (“Here, we have a threshold question of NEPA
applicability. The Secretaries have not prepared an EIS or EA . . .
contending that NEPA does not apply . . . . We hold the less deferential
standard of ‘reasonableness’ applies to threshold agency decisions that
certain activities are not subject to NEPA’s procedures.”).
126 SAN LUIS V. JEWELL

whether FWS’s issuance of the BiOp and its RPA under these
circumstances” requires the preparation of an EIS pursuant to
NEPA. Id. at 1044. It concluded that “[t]his call need not be
made, because Reclamation, the agency with the ultimate
authority to implement the RPA, is now joined as a party,
whose actions must be evaluated under NEPA.” Id. The
court granted summary judgment in favor of the plaintiffs on
their claim against Reclamation, concluding “that
Reclamation violated NEPA by failing to perform any NEPA
analysis prior to provisionally adopting and implementing the
2008 BiOp and its RPA.” Id. at 1051.

In a subsequent motion for summary judgment, the
plaintiffs again asserted that the FWS violated NEPA by not
issuing an EIS along with the BiOp. The district court
remarked that “[t]his was an attempt to re-argue and re-frame
arguments previously decided” because the district court’s
“prior NEPA rulings determined that Reclamation bears the
NEPA responsibility in this case.” San Luis & Delta-
Mendota, 760 F. Supp. 2d at 965. The defendants filed an
opposition and cross-motion for summary judgment, urging
the district court to reaffirm that the FWS did not need to
comply with NEPA in issuing the BiOp. The district court
denied the plaintiffs’ motion and granted the defendants’
motion with respect to the NEPA claims against the FWS. Id.
at 966.

On appeal, the plaintiffs argue that both the FWS and
Reclamation must comply with NEPA. The federal
defendants contend that the FWS need not comply with
NEPA because Reclamation will complete an EIS. And
NRDC takes the position that neither the issuance of the BiOp
by the FWS nor the acceptance and implementation of the
BiOp by Reclamation triggers obligations under NEPA.
 SAN LUIS V. JEWELL 127

We affirm the district court’s judgment with respect to the
NEPA claims. First, we hold that, under these circumstances,
NEPA does not require the FWS to prepare an EIS in
conjunction with the issuance of the BiOp. Second, we hold
that Reclamation’s provisional adoption and implementation
of the BiOp triggered its obligation to comply with NEPA.
We therefore affirm the district court’s order remanding to
Reclamation so that it can complete an EIS evaluating the
effects of its adoption and implementation of the BiOp.47

1. Application of NEPA to the FWS

We first consider whether the FWS’s issuance of the
BiOp was a “major Federal action[] significantly affecting the
quality of the human environment” such that the FWS was
obligated to complete an EIS. A “[m]ajor federal action
includes actions with effects that may be major and which are
potentially subject to Federal control and responsibility.”
40 C.F.R. § 1508.18. The regulations offer several categories
of major federal actions, including “[a]doption of formal
plans, such as official documents prepared or approved by
federal agencies which guide or prescribe alternative uses of
Federal resources, upon which future agency actions will be
based” and “[a]pproval of specific projects, such as

47
Reclamation provided notice of its intent to prepare an EIS on March
28, 2012. See Remanded Biological Opinions on the Coordinated Long-
Term Operation of the Central Valley Project and State Water Project:
Notice of Intent to Prepare an Environmental Impact Statement and Notice
of Scoping Meetings, 77 Fed. Reg. 18858-02 (March 28, 2012). The
district court continues to actively manage Reclamation’s deadline for
completing the EIS process. See Memorandum Decision and Order
Regarding Motion to Extend Remand Schedule, Consolidated Delta Smelt
Cases, No. 1:09-cv-00407(E.D. Cal. Apr. 9, 2013), ECF No. 1106.
128 SAN LUIS V. JEWELL

construction or management activities located in a defined
geographic area.” Id. at § 1508.18(b)(2), (4).

The federal defendants argue that the FWS, in its capacity
as a consulting agency under Section 7 of the ESA, is merely
offering its opinions and suggestions to Reclamation, which,
as the action agency, ultimately decides whether to adopt or
approve the plan. This view is well supported by the statute,
regulations, and our case law. Section 7(b) explains that the
FWS “shall provide to the Federal agency [e.g., Reclamation]
. . . a written statement setting forth the Secretary’s opinion,
and a summary of the information on which the opinion is
based, detailing how the agency action affects the species or
its critical habitat.” 16 U.S.C. § 1536(b)(3)(A) (emphasis
added). “If jeopardy or adverse modification is found, the
Secretary shall suggest those reasonable and prudent
alternatives which he believes would not violated subsection
(a)(2) of this section and can be taken by the Federal agency
[e.g., Reclamation] . . . in implementing the agency action.”
Id. (emphasis added). We would not ordinarily consider an
“opinion” or “suggest[ion]” a “major Federal action[].” The
regulations further provide that “[f]ollowing the issuance of
a biological opinion, the Federal agency [e.g., Reclamation]
shall determine whether and in what manner to proceed with
the action in light of its section 7 obligations and the
Service’s biological opinion.” 50 C.F.R. § 402.15(a). Our
cases confirm that an action agency like Reclamation has
some discretion to deviate from the BiOp and its RPAs. See
Pyramid Lake Paiute Tribe of Indians v. U.S. Dep’t of Navy,
898 F.2d 1410, 1418 (9th Cir. 1990) (“We have recognized
that the Secretary is to be afforded some discretion in
ascertaining how best to fulfill the mandate to conserve under
section 7(a)(1) [of the ESA] . . . . For example, [an action]
agency is given discretion to decide whether to implement
 SAN LUIS V. JEWELL 129

conservation recommendations put forth by the FWS.”);
Tribal Vill. of Akutan v. Hodel, 869 F.2d 1185, 1193 (9th Cir.
1988) (“The agency is not required to adopt the alternatives
suggested in the biological opinion . . . . [The Secretary]
satisfied section 7(a)(2) if he took alternative, reasonably
adequate steps to insure the continued existence of any
endangered or threatened species.”).

We are mindful of the fact that “while the Service’s
Biological Opinion theoretically serves an ‘advisory
function,’ in reality it has a powerful coercive effect on the
action agency.” Bennett v. Spear, 520 U.S. 154, 169 (1997)
(citation omitted). “The action agency is technically free to
disregard the Biological Opinion and proceed with its
proposed action, but it does so at its own peril (and that of its
employees), for ‘any person’ who knowingly ‘takes’ an
endangered or threatened species is subject to substantial civil
and criminal penalties, including imprisonment.” Id. at 170.48
But the “powerful coercive effect” of a BiOp on an action
agency like Reclamation does not render it akin to the
“[a]doption of formal plans” or “[a]pproval of specific
projects,” which tend to trigger NEPA’s requirements.
40 C.F.R. 1508.18(b)(2), (4). Unlike Reclamation, the FWS
is not responsible for, and will not implement, the RPAs.
And even if Reclamation felt compelled to implement the
FWS’s proposal, we must bear in mind that Reclamation will

48
In Bennett, “The question for decision [was] whether the petitioners
. . . [had] standing to seek judicial review of the biological opinion under
the citizen-suit provision of the ESA . . . and the Administrative Procedure
Act.” Id. at 157. The case did not raise or resolve any questions
concerning the application of NEPA to agency action under Section 7 of
the ESA, which is at issue in this case.
130 SAN LUIS V. JEWELL

complete an EIS evaluating the effects of implementing the
BiOp. See infra section IV.B.2.

The fact that Reclamation, and not the FWS, bears
responsibility for implementing the BiOp—or an alternative
that complies with Section 7’s mandate—distinguishes this
case from Ramsey v. Kantor, 96 F.3d 343 (9th Cir. 1996),
where we held that the agency issuing a BiOp and ITS was
required to comply with NEPA. In Ramsey, the National
Marine Fisheries Service (NMFS) produced a BiOp and ITS
in connection with the Columbia River Fish Management
Plan, which is a “unique, judicially created, federal-state-
tribal compact” that “apportions the fishing rights to the state
and tribal members.” Id. at 438. Notably, “The states then
enact regulations governing fishing in the Columbia River,
although they must do so in compliance with the terms of the
Columbia River Fish Management Plan.” Id. After the
NMFS completed its BiOp and ITS, the states of Washington
and Oregon issued regulations, which “would be illegal, if not
for that [incidental take] statement,” permitting a specified
amount of salmon fishing in the Columbia River. Id. at 444.
The Ramsey court “conclude[d] that the incidental take
statement in this case is functionally equivalent to a permit.”
Id. Our cases had already established that “if a federal permit
is a prerequisite for a project with adverse impact on the
environment, issuance of that permit does constitute major
federal action and the federal agency involved must conduct
an EA and possibly an EIS before granting it.” Id. (citing
Jones v. Gordon, 792 F.2d 821, 827–29 (9th Cir. 1986); Port
of Astoria v. Hodel, 595 F.2d 467, 478–79 (9th Cir. 1979)).
For this reason, the court held “that the issuance of [the
incidental take] statement constitutes major federal action for
purposes of NEPA.” Ramsey, 96 F.3d at 444.
 SAN LUIS V. JEWELL 131

In Ramsey, the states of Washington and Oregon occupied
the position typically inhabited by a federal action agency
like Reclamation because the BiOp and ITS were issued as
part of a federal-state-tribal compact. Because NEPA applies
only to “federal actions,” 42 U.S.C. § 4332(2)(C), in that case
there was no downstream federal agency to complete an EIS.
If the consulting agency, the NMFS, did not comply with the
EIS requirement in Ramsey, then the action would have
evaded NEPA review altogether even though the action was,
in substance, identical to the process for issuing a permit,
which would require the issuing agency to prepare an EIS.

Here, there is no comparable need to require the FWS to
prepare an EIS because Reclamation stands ready to do so.
We have held that an agency need not complete an EIS where
another agency will authorize or implement the action that
triggers NEPA. In Sierra Club v. FERC, 754 F.2d 1506 (9th
Cir. 1985), the Federal Energy Regulatory Commission
issued a preliminary permit to construct a hydroelectric
power plant. Id. at 1508. The plaintiffs argued that the
Commission should have conducted an EIS before issuing the
permit. Id. at 1509. We disagreed, explaining that the
applicants “can only enter federal land and conduct ground-
breaking activities after obtaining Forest Service and BLM
special use permits. Thus, these agencies, not the
Commission, will be responsible for evaluating the
environmental impact of activities authorized by their special
use permits.” Id.; see also Conner v. Burford, 848 F.2d 1441
(9th Cir. 1988) (holding that the sale of particular oil and gas
leases did not require the Forest Service to complete an EIS
because “absent further governmental approval, the [ ] leases
absolutely prohibit surface-disturbing activity.” (emphasis
added)).
132 SAN LUIS V. JEWELL

We see no reason to require a consulting agency like the
FWS to complete an EIS when an action agency like
Reclamation will either (1) prepare an EIS when it
implements FWS’s proposal or (2) reject FWS’s proposal and
prepare an EIS on whatever alternative it implements. We
have condemned efforts to use NEPA as an “obstructionist
tactic.” See Drakes Bay Oyster Co. v. Jewell, 729 F.3d 967,
984 (9th Cir. 2013); Douglas Cnty., 48 F.3d at 1508. Of
course the fact that completing an EIS might be time
consuming or costly does not excuse an agency from
complying with NEPA; that is a balance struck by Congress,
not the courts. But the statute requires completion of an EIS
in connection with all “major Federal actions.” 42 U.S.C.
§ 4332(2)(C) (emphasis added). It does not require
completion of an EIS by all agencies that make
recommendations with respect to an action or participate in
formulating a proposal for action. As the district court
observed, “it makes little sense to have two agencies prepare
separate NEPA documents for the same agency action.” San
Luis & Delta-Mendota, 686 F. Supp. 2d at 1042.

We hold that, under these circumstances, the FWS was
not required to comply with NEPA in issuing the BiOp.
There are situations where a consulting agency like the FWS
is required to complete an EIS in conjunction with the
preparation of a BiOp under Section 7 of the ESA. We have
already identified one such scenario in Ramsey. But neither
the statute nor our case law supports the proposition that the
FWS’s production of a BiOp constitutes a “major Federal
action[]” when its implementation is contingent on
Reclamation’s adoption of the BiOp, which is an action that
will trigger Reclamation’s obligation to complete an EIS.
 SAN LUIS V. JEWELL 133

2. Application of NEPA to Reclamation

Having explained why the FWS was not required to
produce an EIS when it issued the BiOp, we now address why
the district court correctly concluded that Reclamation’s
adoption and implementation of the BiOp requires the
preparation of an EIS. The federal defendants do not contest
the district court’s decision that Reclamation should have
completed an EA and, if necessary, an EIS in conjunction
with its implementation of the BiOp. But NRDC appeals the
district court’s order granting summary judgment in favor of
the plaintiffs on their claim that Reclamation violated
NEPA.49

49
Because the federal defendants have not appealed the district court’s
decision that Reclamation’s adoption of the BiOp is subject to NEPA, we
must first confirm that NRDC has standing to challenge the decision. See
Diamond v. Charles, 476 U.S. 54, 68 (1986) (“[A]n intervenor’s right to
continue a suit in the absence of the party on whose side intervention was
permitted is contingent upon a showing by the intervenor that he fulfills
the requirements of Art. III.”).

We hold that NRDC has standing to appeal the district court’s
decision. “To determine whether an intervenor may appeal from a
decision not being appealed by one of the parties in the district court, the
test is whether the intervenor’s interests have been adversely affected by
the judgment.” Didrickson v. U.S. Dep’t of the Interior, 982 F.2d 1332,
1338 (9th Cir. 1992). “To invoke this court’s jurisdiction on the basis of
an injury related to the judgment, Intervenors must establish that the
district court’s judgment causes their members a concrete and
particularized injury that is actual or imminent and is likely to be redressed
by a favorable decision.” W. Watersheds Project v. Kraayenbrink,
632 F.3d 472, 482 (9th Cir. 2010). “[A] credible threat of harm is
sufficient to constitute actual injury for standing purposes, whether or not
a statutory violation has occurred.” Cent. Delta Water Agency v. United
States, 306 F.3d 938, 950.
134 SAN LUIS V. JEWELL

We begin by noting that we agree with the district court’s
conclusion that Reclamation’s implementation of the BiOp is
a “major Federal action[] significantly affecting the quality of
the human environment,” even though NRDC does not
directly challenge this portion of the court’s decision.
42 U.S.C. § 4332(2)(C), First, implementation of the BiOp
is a “major Federal action.” We have held “that where a
proposed federal action would not change the status quo, an

NRDC has demonstrated, for Article III standing purposes, that the
district court’s judgment requiring Reclamation to complete an EIS poses
“a credible threat of harm” to the delta smelt. The 2008 BiOp—which we
hold today is not arbitrary and capricious—concluded that project
operations jeopardized the delta smelt. There is, therefore, “a credible
threat of harm” to the delta smelt if project operations continue. Although
the district court remanded without vacatur, the plaintiffs can potentially
use the judgment that Reclamation violated NEPA to enjoin
implementation of the RPAs so long as doing so does not violate Section
7 of the ESA. See In re Consolidated Delta Smelt Cases, 812 F. Supp. 2d
1133, 1198 (E.D. Cal. 2011) (judgment vacated by San Luis & Delta-
Mendota Water Auth. v. Salazar, No. 11-17143 (9th Cir. Aug. 23, 2012))
(“A court may not issue an injunction under NEPA that would cause a
violation of other statutory requirements, such as those found in Section
7 of the ESA. . . . However, where the evidence indicates that the ESA
will not be violated by injunctive relief issued under NEPA, the presence
of a NEPA claim permits consideration of economic harm evidence.”).
The fact that a court cannot grant an injunction based on the judgment that
Reclamation violated NEPA if doing so would cause a violation of Section
7 of the ESA does not preclude NRDC from having standing to appeal the
judgment because “a credible threat of harm is sufficient to constitute
actual injury for standing purposes, whether or not a statutory violation
has occurred.” Cent. Delta Water Agency, 306 F.3d at 950 (emphasis
added). The judgment thus introduces some probabilistic chance of
environmental harm short of a statutory violation of Section 7 of ESA,
which is sufficient for Article III standing. See id. at 948 (“‘Threatened
environmental harm is by nature probabilistic.’” (quoting Friends of the
Earth, Inc. v. Gaston Copper Recycling Corp., 204 F.3d 149, 160 (4th Cir.
2000) (en banc)).
 SAN LUIS V. JEWELL 135

EIS is not necessary.” Upper Snake River Chapter of Trout
Unlimited v. Hodel, 921 F.2d 232, 235 (9th Cir. 1990); see
also Burbank Anti-Noise Group v. Goldschmidt, 623 F.2d
115, 116 (9th Cir. 1980) (“An EIS need not discuss the
environmental effects of mere continued operation of a
facility.”). In Upper Snake River, the court held that
Reclamation’s decision to reduce the flow of water from a
dam to 1,000 cfs was not a major federal action. Upper Snake
River, 921 F.2d at 233. Reclamation had already set the flow
rate below 1,000 cfs during 4.75% of the total days that the
dam had been in operation, which led the court to conclude
that “[w]hat they did in prior years and what they were doing
during the period under consideration were no more than the
routine managerial actions regularly carried on from the
outset without change . . . they are doing nothing new, nor
more extensive, nor other than that contemplated when the
project was first operational.” Id. at 234–35. The district
court correctly distinguished Upper Snake River by observing
that the BiOp does not merely involve a “routine” adjustment
to the operation of the project. Rather, it “can be determined
from the face of the BiOp and uncontroverted analyses of
public data . . . [that] the Projects’ water delivery operations
must be materially changed to restrict project water flows to
protect the smelt.” San Luis & Delta-Mendota, 686 F. Supp.
2d at 1049.

Second, Reclamation’s implementation of the BiOp
“significantly affect[s] the human environment.” We have
held that “[a]n agency is required to prepare an EIS where
there are substantial questions about whether a project may
cause significant degradation of the human environment.”
Native Ecosystems Council v. U.S. Forest Serv., 428 F.3d
1233, 1239 (9th Cir. 2005) (emphasis in original). The
district court correctly concluded that “dispositive
136 SAN LUIS V. JEWELL

conclusions can be made without looking to the
[administrative record].” San Luis & Delta-Mendota, 686 F.
Supp. 2d at 1050. For example, the federal defendants’
answer states that “Defendants aver that ‘reductions in
exports from the Delta’ may ‘place greater demands upon
alternative sources of water, including groundwater.’” Id.
The district court correctly concluded that Reclamation’s
implementation of the BiOp is a “major Federal action[]
significantly affecting the quality of the human environment.”
42 U.S.C. § 4332(2)(C).

NRDC’s argument on appeal is that, under these
circumstances, NEPA conflicts with the ESA’s goal of
preserving listed species like the delta smelt by imposing an
additional procedural requirement on Reclamation. For this
reason, NRDC contends that NEPA does not apply to
Reclamation’s adoption and implementation of the BiOp,
even if the action is a “major Federal action[] significantly
affecting the quality of the human environment” that would
ordinarily trigger NEPA review. We therefore consider
whether we should set aside the EIS requirement when an
agency implements a BiOp and RPAs designed to ensure that
its action “is not likely to jeopardize the continued existence
of any endangered species or threatened species or result in
the destruction or adverse modification of habitat of such
species.” 16 U.S.C. § 1536(a)(2). We find no basis in the
statute or our case law for excusing Reclamation from its
NEPA obligations under these circumstances.

On its face, the statute does not permit case-by-case
exceptions that assess how NEPA interacts with the
substantive statute at issue. It simply requires that “to the
fullest extent possible . . . all agencies of the Federal
Government shall” complete an EA and, if necessary, an EIS
 SAN LUIS V. JEWELL 137

for all “major Federal actions significantly affecting the
quality of the human environment.” 42 U.S.C. § 4332(2)(C).
Congress has expressly provided that NEPA does not apply
to certain statutory schemes. For example, “No action taken
under the Clean Air Act shall be deemed a major Federal
action significantly affecting the quality of the human
environment within the meaning of the National
Environmental Policy Act of 1969.” 15 U.S.C. § 793(c)(1).
In other words, no action taken under the Clean Air Act will
trigger NEPA’s requirement that the agency produce an EIS.
See also 42 U.S.C. § 8473 (stating that the EIS requirement
does not apply to certain exemptions for electric
powerplants); 42 U.S.C. § 5159 (providing that actions
restoring particular facilities to their condition prior to a
disaster or emergency are exempt from the EIS requirement).
Most notably, Section 7 of the ESA—the provision at issue
here—carves out a narrow exception to the EIS requirement.
The ESA authorizes the formation of an “Endangered Species
Committee” that is empowered to grant exemptions from the
general prohibition on agency actions that “jeopardize the
continued existence” of listed species or “result in the
destruction or adverse modification of [their] habitat.”
16 U.S.C. § 1536(e). The statute specifically provides that a
decision by the Endangered Species Committee to exempt an
agency action from the ESA’s prohibitions is not subject to
NEPA if the agency already completed an EIS concerning the
effects of the action. Id. at § 1536(k).

This could very well be the end of our inquiry. Congress
has repeatedly demonstrated that it knows how to exempt
particular substantive statutes from the EIS requirement when
it wishes to do so. Moreover, Congress has expressly
exempted a particular subset of actions under Section 7 of the
ESA—decisions by the Endangered Species Committee
138 SAN LUIS V. JEWELL

where an EIS was already completed by the action agency.
But Congress did not exempt all efforts to avoid jeopardizing
the survival of a listed species from the EIS requirement. In
fact, Congress’s decision to exempt certain decisions by the
Endangered Species Committee from the EIS requirement
reaffirms that NEPA applies to other actions under Section 7
of the ESA, including Reclamation’s implementation of the
BiOp.

There is additional statutory evidence that Reclamation’s
adoption and implementation of the BiOp triggers its
obligations under NEPA. Section 7 of the ESA provides that
a biological assessment “may be undertaken as part of a
Federal agency’s compliance with the requirements of section
102 of the National Environmental Policy Act of 1969
(42 U.S.C. § 4332),” which is the section that governs the
preparation of an EIS. 16 U.S.C. § 1536(c)(1). This is
evidence that Congress specifically contemplated that an
action agency discharging its duties under Section 7 of the
ESA would also comply with NEPA by completing an EA
and, if necessary, an EIS. The regulations also acknowledge
that the agencies are expected to concurrently comply with
both Section 7 of the ESA and NEPA. See 50 C.F.R.
§ 402.06 (“Consultation, conference, and biological
assessment procedures under section 7 may be consolidated
with interagency cooperation procedures required by other
statutes, such as the National Environmental Policy Act
(NEPA).”).

NRDC does not cite a single case where any court has
held that an action agency’s obligations under Section 7 of
the ESA excuse it from complying with NEPA. There are,
however, a number of cases holding that other substantive
statutes are exempt from the EIS requirement, even though
 SAN LUIS V. JEWELL 139

Congress has not expressly provided an exemption. NRDC
would have us extend the rationale of those cases to these
circumstances.

We have recognized two circumstances where an agency
need not complete an EIS despite an absence of an express
statutory exemption. First, an agency is excused from
complying with NEPA where doing so “would create an
irreconcilable and fundamental conflict” with the substantive
statute at issue. Flint Ridge Dev. Co. v. Scenic Rivers Ass’n
of Okla., 426 U.S. 776, 788 (1976). Second, we have
identified a limited number of instances where a substantive
statute has “displaced” NEPA’s requirements, even though
there is not “an irreconcilable” conflict between the
substantive statute and the EIS requirement. See Douglas
Cnty., 48 F.3d at 1502 (“[The plaintiff] argues that without
this ‘irreconcilable’ statutory conflict NEPA must apply. We
disagree, and . . . we hold that NEPA does not apply to the
designation of a critical habitat.”).

First, in Flint Ridge the Supreme Court held that the EIS
requirement did not apply because requiring the agency to
prepare an EIS “would create an irreconcilable and
fundamental conflict with the Secretary’s duties under the
[substantive statute at issue].” Flint Ridge, 426 U.S. at 788.
There, the substantive statute provided that a document filed
with the agency would automatically become effective in
thirty days under certain circumstances. Id. at 788. The
Court explained that “[i]t is inconceivable that an
environmental impact statement could, in 30 days, be drafted,
circulated, commented upon, and then reviewed and revised
in light of the comments.” Id. at 789. But in Jones v.
Gordon, 792 F.2d 821 (9th Cir. 1986), we observed that
“Flint Ridge applies only when a conflict is ‘clear and
140 SAN LUIS V. JEWELL

unavoidable’ and ‘irreconcilable and fundamental.’” Id. at
826. The Jones court explained that, unlike in Flint Ridge,
the agency “could withhold publication long enough to
comply with any NEPA requirement for preparation of an
environmental impact statement.” Id.

There is no “irreconcilable and fundamental conflict”
between NEPA and Section 7 of the ESA. Although the
statute sets out a timetable for the consultation process, it is
flexible enough to accommodate the preparation of an EIS.
See 16 U.S.C. § 1536(b)(1)(A) (“Consultation under
subsection (a)(2) of this section with respect to any agency
action shall be concluded within the 90-day period beginning
on the date on which initiated or, subject to paragraph (B),
within such other period of time as is mutually agreeable to
the Secretary and federal agency.” (emphasis added));
16 U.S.C. § 1536(b)(1)(B) (“The Secretary and the Federal
agency may mutually agree to extend a consultation period
established under the preceding sentence if the Secretary,
before the close of such period, obtains the consent of the
application to the extension.”); see also Westlands Water
Dist. v. United States Dep’t of Interior, 850 F. Supp. 1388,
1423 (E.D. Cal. 1994) (“Section 7 of the ESA gives agencies
control over the time within which consultation is to be
concluded . . . . [And] ESA § 7 provides for the inclusion of
‘applicants’ within the consultation process, which
demonstrates access to the ESA process by interested parties
. . . . Neither timing nor secrecy concerns bar the ability to
comply with NEPA.”).

Second, we have held that an agency action might be
exempt from NEPA even “without this ‘irreconcilable’
statutory conflict” identified in Flint Ridge. Douglas Cnty.,
48 F.3d at 1502; see also Drakes Bay Oyster, 729 F.3d at 984;
 SAN LUIS V. JEWELL 141

Merrell, 807 F.2d at 778. In Douglas County, we held that
the Secretary of the Interior need not complete an EIS when
designating the critical habitat of a listed species pursuant to
Section 4 of the ESA. Douglas Cnty., 48 F.3d at 1507.50 The
Douglas County court concluded that “Congress intended to
displace” NEPA’s procedures when authorizing the agency to
designate critical habitat under Section 4 of the ESA. Id. at
1504 n.10. But none of the factors relied on by the Douglas
County court in reaching this decision apply with the same
force where, as here, the agency action at issue is the
implementation of a BiOp under Section 7 of the ESA.

The Douglas County court reasoned that the process for
designating critical habitat under Section 4 of the ESA
effectively accomplished all of NEPA’s goals without
requiring an EIS, thereby “mak[ing] the NEPA procedure
seem ‘superfluous.’” Id. at 1503. Section 4 of the ESA
compels “the Secretary [to] consider impacts that concern
NEPA, to the extent that the critical habitat designation has
a positive environmental effect on the species in question.”
Id. Furthermore, “The critical designation process also
provides for public notice, another goal of NEPA.” Id.

But the same cannot be said for Section 7 of the ESA,
which is at issue in this case. In Save the Yaak Committee v.
Block, 840 F.2d 714 (9th Cir. 1988), we explained the
difference between a biological assessment (BA) produced

50
The Tenth Circuit subsequently disagreed with the result that we
reached in Douglas County, creating a circuit split concerning whether
NEPA applies to the designation of critical habitat under Section 4 of the
ESA. See Catron Cnty. Bd. of Comm’rs v. U.S. Fish & Wildlife Serv.,
75 F.3d 1429, 1436 (10th Cir. 1996); see also Cape Hatteras Access Pres.
Alliance v. U.S. Dep’t of Interior, 344 F. Supp. 2d 108, 133–36 (D.D.C.
2004).
142 SAN LUIS V. JEWELL

pursuant to Section 7 of the ESA, and an EA or EIS prepared
in accordance with NEPA. The Save the Yaak court
considered an argument that “even if the EA was inadequate,
it was supplemented by the biological assessment (BA)”
completed under Section 7 of the ESA. Id. at 718. We
rejected this reasoning, explaining that “[w]hile a BA
analyzes the impact of a proposed action upon endangered
species, an EA analyzes the impact of the proposed action on
all facets of the environment. Thus, if only a BA is prepared
there may be gaps in the agency’s environmental analysis.”
Id.

Courts have offered several other examples of the
differences between the Section 7 process and the one
prescribed by NEPA. For instance, “the ESA’s Section 7
consultation process fails to provide for public comment in
the same way that NEPA does.” Fund for Animals v. Hall,
448 F. Supp. 2d 127, 136 (D.D.C. 2006). This is particularly
important because “[p]ublication of an EIS, both in draft and
final form, also serves a larger informational role. It gives the
public the assurance that the agency ‘has indeed considered
environmental concerns in the decisionmaking process,’ and,
perhaps more significantly, provides a springboard for public
comment.” Robertson v. Methow Valley Citizens Council,
490 U.S. 332, 350 (1989) (quoting Baltimore Gas & Elec.
Co., 462 U.S. at 97) (internal citation omitted). Additionally,
“the ESA only requires agencies to consider the cumulative
impacts of non-federal actions, while NEPA requires agencies
to consider the cumulative impacts of all actions.” Fund for
Animals, 448 F. Supp. 2d at 136; see also 50 C.F.R. § 402.02
(describing the cumulative effects analysis under the ESA);
40 C.F.R. § 1508.7 (describing the cumulative impacts
analysis under NEPA). We cannot say that Section 7 of the
ESA renders NEPA “superfluous” when the statutes evaluate
 SAN LUIS V. JEWELL 143

different types of environmental impacts through processes
that involve varying degrees of public participation.

In Merrell, we held that the EIS requirement does not
apply to FIFRA’s scheme for registering pesticides because
the two processes are markedly different. See Merrell,
807 F.2d at 778 (“The differences between FIFRA’s
registration procedure and NEPA’s requirements indicate that
Congress did not intend NEPA to apply.”). Although both
Merrell and Douglas County conclude that NEPA does not
apply to a particular substantive statute, they do so for
opposite reasons. Douglas County holds that Section 4 of the
ESA renders NEPA superfluous because the processes are
sufficiently similar, while Merrell holds that FIFRA renders
NEPA superfluous because the processes are sufficiently
different. See Douglas Cnty., 48 F.3d at 1503 (“[T]he
Secretary will consider impacts that concern NEPA, to the
extent that the critical habitat designation has a positive
environmental effect on the species in question. The critical
designation process also provides for public notice, another
goal of NEPA.”); Merrell, 807 F.2d at 779 (“[W]hen
Congress revised FIFRA in 1972, it designed a registration
procedure with public notice and public participation
provisions that differ materially from those that NEPA would
require.”).

Although we have already acknowledged the differences
between Section 7 of the ESA and NEPA, we do not think
that the distinctions are as pronounced as those in Merrell,
where the court concluded that “[t]o apply NEPA to FIFRA’s
registration process would sabotage the delicate machinery
that Congress designed to register new pesticides.” Merrell,
807 F.2d at 779. As we have observed, Congress specifically
contemplated that an agency could comply with NEPA while
144 SAN LUIS V. JEWELL

discharging its duties under Section 7 of the ESA. See
16 U.S.C. § 1536(c)(1) (explaining that the biological
assessment required by Section 7 of the ESA “may be
undertaken as part of a Federal agency’s compliance with the
requirements of section 102 of the National Environmental
Policy Act of 1969 (42 U.S.C. § 4332)”). In this very case,
the federal agencies acknowledge that Reclamation will
complete an EIS in conjunction with its adoption and
implementation of the BiOp, which undercuts the notion that
the two processes are incompatible. Under these
circumstances, we cannot conclude that the process set out by
Section 7 of the ESA clashes with NEPA to such an extent
that requiring Reclamation to produce an EIS “would
sabotage the delicate machinery that Congress designed.”
Merrell, 807 F.2d at 779. Instead, we find that Section 7 of
the ESA fits within the broad swath of statutes that coexist
with NEPA.51

51
A number of other circuits have held that an agency need not produce
an EIS where the substantive statute at issue offers a procedure that is the
“functional equivalent” of the EIS process. See, e.g., State of Ala. ex rel.
Siegleman v. EPA, 911 F.2d 499, 504 (11th Cir. 1990); Limerick
Ecological Action, Inc. v. U.S. Nuclear Regulatory Comm’n., 869 F.2d
719, 729 n.7 (3d Cir. 1989); Izaak Walton League of Am. v. Marsh,
655 F.2d 346, 367 n.51 (D.C. Cir. 1981).

We have been skeptical of the “functional equivalent” approach and
have not used this language in our cases. See Douglas Cnty., 48 F.3d at
1504 n.10 (“Courts have used a ‘functional equivalent’ test to exempt
agency action from NEPA requirements . . . . The defendants here do not
advance the functional equivalent argument, so we do not address it. The
[plaintiff] would have us believe that the ‘displacement’ argument
defendants make is the same as the ‘functional equivalent’ test. We do not
agree. The ‘displacement’ argument asserts that Congress intended to
displace one procedure with another. The ‘functional equivalent’
argument is that one process requires the same steps as another.”);
Merrell, 807 F.2d at 781 (“While we hesitate to adopt the ‘functional
 SAN LUIS V. JEWELL 145

The Douglas County court also noted that Congress
acquiesced to decisions by the agencies and courts that
compliance with NEPA was not required when designating
critical under habitat Section 4 of the ESA. We noted that in
1983 the Secretary of the Interior stopped preparing EAs and
EISs before designating critical habitat, yet Congress did not
address the agency’s interpretation of the statute when it
amended the ESA in 1988. Douglas Cnty., 48 F.3d at 1504.
The Sixth Circuit had also suggested in dicta that it might not
be necessary to prepare an EIS before designating critical
habitat before the 1988 amendments to the ESA. Id. (citing
Pac. Legal Found. v. Andrus, 657 F.2d 829, 835 (6th Cir.
1981)). But here, neither the agencies nor the courts have
interpreted Section 7 of the ESA to permit noncompliance
with NEPA. As noted, the relevant regulations indicate that
the action agency will complete an EIS while carrying out its
duties under Section 7, and the federal defendants in this case
have assumed that Reclamation will complete an EIS
evaluating the effects of implementing the BiOp. See
50 C.F.R. § 402.06 (“Consultation, conference, and biological
assessment procedures under section 7 may be consolidated

equivalence’ rationale, we are confident that Congress did not intend
NEPA to apply to FIFRA registrations.”).

NRDC relies on Douglas County and Merrell, which are
“displacement” cases rather than “functional equivalent” cases. To our
knowledge, none of the circuits that have adopted the “functional
equivalent” test have held that the procedures set out by Section 7 of the
ESA are equivalent to the EIS requirement. Although NRDC does not
urge us to adopt the “functional equivalent” approach, we note that the
factors considered in the preceding paragraphs are the same ones that we
would address under that analysis. Regardless of the language used to
conduct the analysis, the statutes and regulations reveal that Section 7 of
the ESA and NEPA involve different processes that measure different
kinds of environmental impacts.
146 SAN LUIS V. JEWELL

with interagency cooperation procedures required by other
statutes, such as the National Environmental Policy Act
(NEPA).”). And, to our knowledge, there are no cases stating
that an agency need not comply with NEPA because of its
obligations under Section 7 of the ESA.

Next, the Douglas County court stated that there is no
reason to prepare an EIS “when the action at issue does not
alter the natural, untouched physical environment at all.”
Douglas Cnty., 48 F.3d at 1505. We explained that “the
purpose of NEPA is to protect the physical environment, and
the purpose of preparing an EIS is to alert agencies and the
public to potential adverse consequences to the land, sea or
air.” Id. Therefore, the designation of critical habitat, which
results in the land at issue being left alone, does not require
an EIS. Here, the BiOp does far more than leave nature
alone. Humans have dramatically altered the Bay-Delta
ecosystem. Even if we assume that implementing the BiOp
is a step toward returning the ecosystem to its natural state,
there is no doubt that project operations will continue to alter
the physical environment, albeit in different ways. We have
already interpreted this portion of Douglas County quite
narrowly. See Kootenai Tribe of Idaho v. Veneman, 313 F.3d
1094 (9th Cir. 2002) (abrogated on other grounds by
Wilderness Soc’y v. U.S. Forest Serv., 630 F.3d 1173 (9th
Cir. 2011)). In Kootenai Tribe, the agency announced a
“Roadless Rule” that prohibited construction of new roads in
certain areas. Kootenai Tribe, 313 F.3d at 1105. We
considered an argument that NEPA should not apply to the
agency’s action in light of the Douglas County court’s
statement that “an EIS is not required to leave nature alone.”
Id. at 1114 (quoting Douglas Cnty., 48 F.3d at 1505). We
held that “[b]ecause human intervention, in the form of forest
management, has been part of the fabric of our national
 SAN LUIS V. JEWELL 147

forests for so long, we conclude that, in the context of this
unusual case, the reduction in human intervention that would
result from the Roadless Rule actually does alter the
environmental status quo . . . . The Forest Service’s Roadless
initiative thus required an EIS under NEPA.” Id. at 1115. An
action to lessen one form of pressure on the natural
environment, such as Reclamation’s implementation of the
BiOp, is distinguishable from a decision to continue to leave
a portion of nature untouched altogether.

At its broadest point, our opinion in Douglas County
implied that the agency’s designation of critical habitat did
not trigger NEPA review because it was an environmental
preservation effort. We explained that “[b]y designating
critical habitats for endangered or threatened species, the
Secretary ‘is working to preserve the environment and
prevent the irretrievable loss of a natural resource.’ Thus the
action of the Secretary in designating a critical habitat
furthers the purpose of NEPA. Requiring the EPA to file an
EIS ‘would only hinder its efforts at attaining the goal of
improving the environment.’” Douglas Cnty., 48 F.3d at
1506. We recently applied this principle in Drakes Bay
Oyster, where we wrote that “[t]he Secretary’s decision is
essentially an environmental conservation effort, which has
not triggered NEPA in the past.” Drakes Bay Oyster,
729 F.3d at 984 (citing Douglas Cnty., 48 F.3d at 1505–06).

We do not read either Douglas County or Drakes Bay
Oyster to stand for the proposition that efforts to preserve the
natural environment are per se exempt from NEPA.52 As

52
We recently observed that our court has yet to hold that an agency that
has already produced an EA need not produce an EIS when the action in
question will only have beneficial impacts on the environment. See
148 SAN LUIS V. JEWELL

noted, the Douglas County court relied on the observation
that designating critical habitat under Section 4 of the ESA is
an “action[] that do[es] nothing to alter the natural physical
environment.” Douglas Cnty., 48 F.3d at 1505. Similarly, in
Drakes Bay Oyster, the only purported “adverse
environmental consequences” of designating the area in
question as a wilderness were “short-term harms, such as
noise associated with heavy machinery needed to remove
Drakes Bay’s structures” in order to return the area to its
natural state. Drakes Bay Oyster, 729 F.3d at 984. We noted
that “such relatively minor harms do not by themselves
‘significantly affect[]’ the environment in such a way as to

Humane Soc’y of U.S. v. Locke, 626 F.3d 1040, 1056 (9th Cir. 2010) (“As
a threshold matter, plaintiffs’ argument appears to raise an issue of first
impression in this circuit: whether NEPA requires an agency to prepare an
EIS when an action has a significant beneficial impact but not significant
adverse impact on the environment.”); see also id. at 1046 (“[T]o comply
with NEPA, NMFS prepared an environmental assessment . . . . The final
environmental assessment resulted in a finding of no significant impact
under NEPA.”). In Humane Society, we did not resolve the question
whether an agency that has produced an EA showing significant beneficial
environmental impacts and no adverse environmental impacts must still
complete an EIS. See id. at 1056.

In order to hold that Reclamation need not comply with NEPA, we
would need to take two substantial steps forward because, in this case,
Reclamation did not even complete an EA. We see no basis for holding
that an agency can avoid NEPA review altogether when it believes that an
agency action will have beneficial impacts on the environment when we
have not even excused an agency from producing an EIS when its EA
shows that its action will have exclusively beneficial impacts on the
environment. In other words, even if we had some basis for assuming that
Reclamation’s implementation of the BiOp would have exclusively
beneficial impacts on the environment, we would still lack a firm
foundation for holding that Reclamation need not prepare an EA and, if
necessary, an EIS.
 SAN LUIS V. JEWELL 149

implicate NEPA.” Id. But here, NRDC does not even contest
the district court’s conclusion that implementation of the
BiOp “significantly affect[s] the human environment.” See
San Luis & Delta-Mendota, 686 F. Supp. 2d at 1050.
Whatever effects implementing the BiOp might have on the
human environment, it is apparent that they are more complex
and wide-ranging than the removal of a few buildings in
Drakes Bay Oyster.

At this point, we can only speculate about what kind of
significant effects will eventually result from implementation
of the BiOp because Reclamation has not yet completed its
EIS. But it is beyond dispute that Reclamation’s
implementation of the BiOp has important effects on human
interaction with the natural environment. We know that
millions of people and vast areas of some of America’s most
productive farmland will be impacted by Reclamation’s
actions. Those impacts were not the focus of the BiOp. In
sum, we cannot reach an informed decision about the extent
to which implementation of the BiOp is an environmental
preservation action in the vein of Douglas County and Drakes
Bay Oyster because we do not know how the action will
impact the broader natural environment. We find no basis for
exempting Reclamation from the EIS requirement. See
Methow Valley Citizens Council, 490 U.S. at 349 (“NEPA
ensures that important effects will not be overlooked or
underestimated only to be discovered after resources have
been committed or the die otherwise cast.”). We recognize
that the preparation of an EIS will not alter Reclamation’s
obligations under the ESA. But the EIS may well inform
Reclamation of the overall costs—including the human
costs—of furthering the ESA. So informed, Reclamation has
150 SAN LUIS V. JEWELL

the option of seeking an exemption from the ESA from the
Endangered Species Committee. See 16 U.S.C. § 1536(e).53

Finally, NRDC argues that Reclamation’s issuance of the
2004 OCAP is the “major Federal action[]” that should have
been subject to NEPA review instead of its implementation

53
The Endangered Species Committee is the exclusive avenue through
which an applicant may obtain an exemption from the ESA’s substantive
prohibitions. See 16 U.S.C. §1536(a)(2) (“Each federal agency shall . . .
insure that any action authorized, funded, or carried out by such agency
. . . is not likely to jeopardize the continued existence of any endangered
species or threatened species or result in the destruction or adverse
modification of habitat of such species . . . unless such agency has been
granted an exemption for such action by the Committee pursuant to
[16 U.S.C. § 1536(h)].”); see also Portland Audubon Soc’y v. Endangered
Species Comm., 984 F.2d 1534, 1537 (9th Cir. 1993) (“The Committee
was created by the Endangered Species Act for the sole purpose of making
final decisions on applications for exemptions from the Act . . . . Because
it is the ultimate arbiter of the fate of an endangered species, the
Committee is known as “The God Squad.”). The Committee has only
convened on a handful of occasions and has only granted two exemptions.
See Portland Audubon Soc’y, 984 F.2d at 1537. Commentators have
discussed whether action from the Committee would be appropriate under
these circumstances. See generally Eric M. Yuknis, Note, Would a “God
Squad” Exemption under the Endangered Species Act Solve the California
Water Crisis?, 38 B.C. Envtl. Aff. L. Rev. 567 (2011).

For our purposes, it is worth noting that the possibility of review by
the Endangered Species Committee, however unlikely it may be, renders
the preparation of an EIS more than a mere academic exercise in cases
involving Section 7 of the ESA. Although the agency cannot ignore its
obligations under the ESA because of the impacts on other aspects of the
human environment, the Committee might wish to consider these factors
in making an exemption decision. And, as previously noted, the statute
contemplates that the agency will have completed an EIS by providing
that the Committee itself need not do so if the agency has already prepared
an EIS with respect to the action subject to the Committee’s review. See
16 U.S.C. § 1536(k).
 SAN LUIS V. JEWELL 151

of the 2008 BiOp. In a different case, NRDC alleged that
Reclamation’s “approval and implementation” of the 2004
OCAP triggered its obligation to complete an EIS. See Pac.
Coast Fed’n of Fishermen’s Ass’n/Inst. for Fisheries Res. v.
Gutierrez, No. 1:06-cv-00245, 2007 WL 1752289, at *4 (E.D.
Cal. June 15, 2007). In Gutierrez, the district court held that
the 2004 OCAP was not subject to NEPA because its was not
a “final agency action.” See id. at *12–13 (“[The OCAP]
do[es] not implement any actions or inactions. They are
informational. If any proposed changes are initiated that will
have the requisite effect on the environment, such changes
will be agency action subject to NEPA review. The purpose
of the OCAP is ‘to serve as a baseline description of the
facilities and operating environment of the CVP and SWP.’”).

NRDC contends that Gutierrez was wrongly decided and
that the district court should have required Reclamation to
complete an EIS on the 2004 OCAP. Yet NRDC and the
other plaintiffs in Gutierrez did not appeal the district court’s
decision concerning its NEPA claims. Not only is NRDC
collaterally estopped from relitigating the decision in
Gutierrez, but the issue is also not pertinent to our holding
that Reclamation’s implementation of the BiOp requires the
agency to prepare an EIS. Even if Reclamation’s
implementation of the 2004 OCAP was a final decision that
changed the status quo of the project operations in a way that
significantly affected the environment, that does not mean
that Reclamation’s implementation of the 2008 BiOp did not
also change the status quo in a way that significantly alters
the environment, thereby requiring an additional EIS.

NRDC bolsters its contention that NEPA should apply to
the 2004 OCAP in lieu of the 2008 BiOp by referencing
Section 7 of the ESA and its regulations. As noted above, the
152 SAN LUIS V. JEWELL

statute and its regulations explain that an action agency like
Reclamation can coordinate the preparation of its biological
assessment with its obligations under NEPA. See 16 U.S.C.
§ 1536(c)(1); 50 C.F.R. § 402.06. NRDC reasons that these
provisions suggest that NEPA review of the BiOp is not
required because Reclamation could not complete its EIS
until after its Section 7 consultation with the FWS rather than
at the same time as its consultation obligations. But even if
we fully credit this line of reasoning, it does not affect our
conclusion that NEPA applies to Reclamation’s
implementation of the BiOp. We need not locate affirmative
and unequivocal confirmation in every substantive statute that
a particular agency action requires NEPA review; most
substantive statutes never mention NEPA at all. They do not
need to because NEPA itself provides that “to the fullest
extent possible . . . all agencies of the Federal Government
shall” complete an EA and, if necessary, an EIS for all “major
Federal actions significantly affecting the quality of the
human environment.” 42 U.S.C. § 4332(2)(C). The fact that
Section 7 of the ESA expressly mentions the EIS requirement
supports, rather than rebuts, the notion that NEPA applies to
the action at issue here.

Our starting point was that NEPA applies to
Reclamation’s implementation of the BiOp because it is a
“major Federal action[] significantly affecting the quality of
the human environment.” We acknowledge that we have
previously held that the EIS requirement does not apply to
particular agency actions even in the absence of an express
statutory exemption. See Drakes Bay Oyster, 729 F.3d at 984;
Douglas Cnty., 48 F.3d at 1507; Merrell, 807 F.2d at 781.
But the factors identified in those cases are simply not present
here. We are cognizant of our commitment to avoid
“mak[ing] NEPA more of an ‘obstructionist tactic’ to prevent
 SAN LUIS V. JEWELL 153

environmental protection than it may already have become.”
Douglas Cnty., 48 F.3d at 1508; see also Drakes Bay Oyster,
729 F.3d at 984. But, as noted, the district court remanded
the BiOp and RPAs without vacatur, and it continues to
actively manage Reclamation’s deadline for completing the
EIS process. We conclude that Reclamation is obligated to
comply with NEPA, and we affirm the judgment of the
district court with respect to the NEPA claims.

VI. CONCLUSION

For the forgoing reasons, the judgment of the district
court is reversed in part and affirmed in part. The matter is
remanded to the district court for further proceedings
consistent with this opinion. Each party shall bear its own
costs on appeal.

REVERSED IN PART AND AFFIRMED IN PART.
154 SAN LUIS V. JEWELL

GLOSSARY OF TERMS

APA Administrative Procedure Act

BA Biological Assessment

Bay-Delta San Francisco Bay and Sacramento-San
Joaquin Delta

BiOp 2008 biological opinion

CALFED CALFED Bay-Delta Program

cfs cubic feet per second

CVP Central Valley Project

CVPIA Central Valley Project Improvement Act

DWR California Department of Water
Resources

EIS Environmental Impact Statement

ESA Endangered Species Act

FCRPS Federal Columbia River Power System

FIFRA Federal Insecticide, Fungicide, and
Rodenticide Act

FMWT Fall Midwater Trawl index

FWS U.S. Fish and Wildlife Service
 SAN LUIS V. JEWELL 155

ITS incidental take statement

NMFS National Marine Fisheries Service

NEPA National Environmental Protection Act

OCAP 2004 Operating Criteria and Plan

OMR Old and Middle Rivers

RPA reasonable and prudent alternatives

Reclamation U.S. Bureau of Reclamation

SWP State Water Project

ARNOLD, Circuit Judge, concurring in part and dissenting in
part:

I respectfully dissent from Parts III, IV.A, IV.B, IV.E, and
V.B of the court’s opinion and concur in the rest of it. I
address the issue dealt with in Part III as it arises in
considering the merits of the challenges to the BiOp.

1. I do not believe that the district court erred in holding
that the BiOp’s OMR flow limits were set arbitrarily and
capriciously. First of all, I discern no error in admitting a
portion of the declaration from Dr. Richard Deriso, who holds
advanced degrees in mathematics and biomathematics,
discussing the use of raw salvage data to justify the flow
prescription. A decision to include evidence that is outside
the administrative record is reviewed for abuse of discretion,
156 SAN LUIS V. JEWELL

see Lands Council, 395 F.3d at 1030 n.11, and admitting this
evidence fell within one of the narrow exceptions to the
general rule against extra-record evidence, because it was
necessary to explain technical terms or complex subject
matter, see Nw. Envtl. Advocates v. Nat’l Marine Fisheries
Serv., 460 F.3d 1125, 1145 (9th Cir. 2006). Furthermore, Dr.
Deriso’s declarations were consistent with advice offered by
independent peer reviewers and draft notes of a delta smelt
evaluation team at FWS assembled before the final BiOp
issued, and with the testimony of Rule 706 experts Dr. Punt
and Dr. Quinn, who recognized Dr. Deriso’s declarations and
stated that the validity of the flow regimes specified in the
BiOp was undermined by its incomplete analysis. Nor was
there any “battle of the experts” here, as the court maintains,
because the responses to Dr. Deriso’s declarations from
FWS’s mathematical statistician, Dr. Ken Newman, were
mostly vague, and he generally agreed with Dr. Deriso that
salvage should be scaled by some measure of population
abundance.

As for the merits of this issue, Appellants do not contend
that the use of raw salvage data was scientifically acceptable;
they maintain instead that the flow prescription also relied on
and was supported by other information. Based on my
review of this information, however, the BiOp did not
connect it to flow limits at all, or there was no explanation for
why it yielded the flow prescription that the BiOp specified.
As to FWS’s use of normalized data in the ITS, I am not
convinced that this is relevant to whether it was scientifically
sound for FWS to use only raw salvage data to set the flow
prescription. While certain DWR comments support the flow
prescription, the parties do not dispute that these comments
arose from the district court’s previous remedial imposition
of such a prescription, which, as the district court noted,
 SAN LUIS V. JEWELL 157

occurred before the court became aware that using raw
salvage data was not accepted scientific methodology.
Because FWS based its flow prescription solely on the
unexplained use of raw salvage data, I believe that its
expertise in methodological matters is not entitled to
deference, since that use was not rationally connected to the
best available science, see W. Watersheds Project v.
Kraayenbrink, 632 F.3d 472, 493 (9th Cir. 2011); and
because FWS did not consider all relevant factors or
articulate a rational connection between the facts found and
the choices made, I agree with the district court that its
determination as to the flow prescription was arbitrary and
capricious, see id. at 496.

2. I also concur in the district court’s conclusion that the
BiOp’s determination of X2 was arbitrary and capricious.
First of all, there was no abuse of discretion in the district
court’s decision to admit the declarations of Aaron Miller, a
DWR technical engineer who worked closely with
Reclamation to develop Calsim II, which were relevant to the
Calsim II-Dayflow comparison. See Lands Council, 395 F.3d
at 1030 n.11. No battle of the experts was created by doing
so: Miller’s assessment of the validity of the Calsim II-
Dayflow comparison was consistent with the testimony of
Rule 706 experts Drs. Quinn and Punt, who recognized
Miller’s assessment; and the declarations of FWS hydrologist
Derek Hilts—whose credentials and experience are similar to
Miller’s, and who helped draft the BiOp—were mostly
unresponsive to Miller’s declarations on the several sources
of bias that the comparison introduced. In my view, the
district court relied on this extra-record evidence simply to
determine whether FWS had considered all relevant factors,
here, the sources of bias, see Nw. Envtl. Advocates, 460 F.3d
at 1145, before relying on the comparison to analyze the
158 SAN LUIS V. JEWELL

effects of proposed Projects operations on smelt and its
habitat, including X2’s location. Doing so was well within
the court’s role. Because highly technical matters were
involved, it was difficult to determine if FWS considered all
relevant factors without looking outside the record to see
what matters should have considered, but were not. See
Inland Empire Pub. Lands Council v. U.S. Forest Serv.,
88 F.3d 754, 760 n.5 (9th Cir. 1996). The district court could
not properly discharge its duty to engage in “‘substantial
inquiry’” by simply taking FWS’s word that it had considered
all relevant matters. See Asarco, 616 F.2d at 1160.

FWS’s choice to use the Calsim II-Dayflow comparison
was unsupported by the requisite reasoned analysis. See
Ecology Ctr. v. Castaneda, 574 F.3d 652, 665 (9th Cir. 2009).
Comments received from DWR and other entities—which
were echoed by the Rule 706 experts’ testimony—alerted
FWS to the several sources of bias, yet the only explanation
in the BiOp for using the comparison was that a Calsim II to
Calsim II comparison that FWS had conducted did not show
differences that were expected. As the district court noted,
the record did not reflect that FWS considered, much less
recognized, the sources of bias. Furthermore, there are
significant differences between the two models, including
how X2 positions are determined. FWS would have to
address these significant differences in some way to obtain
information on which it could reasonably rely to base the
BiOp’s conclusions, including Action 4 on the management
of X2’s location. FWS was required to provide some
evidence supporting its conclusions to ensure that no clear
error of judgment rendered its actions arbitrary and
capricious. See League of Wilderness Defenders-Blue
Mountains Biodiversity Project v. U.S. Forest Serv., 549 F.3d
1211, 1218 (9th Cir. 2008). Appellants do not dispute that
 SAN LUIS V. JEWELL 159

the sources of bias existed, or that the biases were significant
or material; and the clear purpose of requiring FWS to use the
best scientific evidence available is to ensure that the ESA is
not implemented haphazardly or based on surmise or
speculation. See Pac. Coast Fed’n of Fishermen’s Ass’ns,
426 F.3d at 1094–95.

Finally, I agree with the district court that FWS did not
sufficiently explain why 74 km and 81 km were selected as
critical points for X2 to preserve smelt habitat. The district
court was not therefore required to give FWS deference on
this matter, as FWS’s reasoning could not be reasonably
inferred from the record. See San Luis & Delta-Mendota
Water Auth., 672 F.3d at 700. As to the other justifying
evidence, it either does not relate to the choice of critical
points or it is undisputed that it was offered for the first time
in post-judgment proceedings. I am reluctant to pass over the
absence of an adequate explanation in the administrative
record by relying on Appellants’ post-hoc rationalizations.
See Humane Soc’y of United States v. Locke, 626 F.3d 1040,
1049–50 (9th Cir. 2010).

3. I find no authority requiring FWS to address
specifically and analyze, in the BiOp or administrative record,
the question of whether the RPA meets the non-jeopardy
elements. I also question the district court’s reliance on
Greenpeace v. Nat’l Marine Fisheries Serv., 55 F. Supp. 2d
1248 (W.D. Wash. 1999), for the proposition that there must
be some explanation in the administrative record as to why
FWS concluded that all four elements for a valid RPA were
satisfied. Nonetheless, I would affirm the district court on
this issue, because the record belies Appellants’ contention
that DWR and Reclamation raised no concerns about the non-
jeopardy elements. The record shows that concerns were
160 SAN LUIS V. JEWELL

raised relating to RPA feasability and its relationship to the
action’s intended purpose (providing water for various uses),
and possibly to DWR’s and Reclamation’s authority to
implement the RPA. Thus, under FWS’s own interpretation
of § 402.02, it was required to consider and address these
elements specifically in the instant BiOp or administrative
record.

4. Finally, as the action agency, Reclamation could not
rely solely on FWS’s BiOp to establish conclusively its
compliance with its substantive obligations under ESA § 7,
because it could not delegate its responsibility to see that its
actions would not jeopardize smelt, see Pyramid Lake,
898 F.2d at 1415; and as the action agency it could not
blindly adopt FWS’s conclusions because it is ultimately
responsible for ESA compliance, see City of Tacoma, Wash.
v. Fed. Energy Regulatory Comm’n, 460 F.3d 53, 76 (D.C.
Cir. 2006). I agree with the district court that Reclamation
would be subject to independent ESA liability if it possessed
new information not considered by FWS which challenged
the BiOp’s conclusions, see Pyramid Lake, 898 F.3d at 1415,
and that there is no indication that this occurred here. But the
district court failed to consider another basis for finding an
action agency independently liable, namely, reliance on a
legally flawed BiOp. Discerning such flaws involves no
technical or scientific expertise, so failure to do so may result
in action based on reasoning not in accordance with the law
thus rendering the action arbitrary and capricious. See Wild
Fish Conservancy v. Salazar, 628 F.3d 513, 532 (9th Cir.
2010). The district court’s legal conclusions necessarily
arose from fact-finding, but the court clearly, and I believe
correctly, concluded that FWS had not used the best available
science or considered relevant factors, and had acted
arbitrarily and capriciously, because, among other things, it
 SAN LUIS V. JEWELL 161

relied on the Calsim II-Dayflow comparison, and did not use
normalized salvage data to set the flow prescription. The
district court therefore should have found Reclamation
independently liable under ESA § 7 for accepting a legally
flawed BiOp and immediately beginning implementation of
the RPA by modifying operations.

In sum, I find no abuse of discretion in the district court’s
limited admission of evidence outside the administrative
record as relevant to the OMR flow limits and the
determination of X2, including the use of the Calsim II-
Dayflow comparison. I believe that in determining whether
FWS’s decisions on these matters in the BiOp were arbitrary,
capricious, or otherwise not in accordance with the law, the
district court’s analysis was thorough and well-reasoned.
While I disagree with the basis for the district court’s
conclusion that the non-jeopardy elements must be addressed
in the BiOp or administrative record, I nonetheless believe
that affirmance is warranted on this issue. Finally, I believe
the district court should have found Reclamation
independently liable under the ESA for relying on a legally
flawed BiOp.

RAWLINSON, Circuit Judge, concurring in part and
dissenting in part:

I concur in the bulk of the majority opinion. I disagree
only with the rationale and conclusion that the Bureau of
Reclamation’s adoption and implementation of the Biological
Opinion triggered its obligation to comply with the National
Environmental Policy Act (NEPA) by preparing an
162 SAN LUIS V. JEWELL

Environmental Impact Statement that is generally required
under the Endangered Species Act.

It is important to keep in mind that the Bureau of
Reclamation adopted and implemented the detailed
Biological Opinion in this case to alleviate harm to threatened
and endangered species and/or their critical habitat caused by
operation of the federally operated Central Valley Water
Project and the state operated State Water Project.

As an initial matter, I do not agree that the adoption of
“reasonable and prudent alternatives” to alleviate harm to
threatened and endangered species and/or their critical habitat
from ongoing operations of the water projects constituted a
“major Federal action” triggering the requirement of an
Environmental Impact Statement. See 42 U.S.C.
§ 4332(2)(C) (requiring the preparation of an Environmental
Impact Statement for “major Federal actions”). I am
persuaded toward that view by our rationale in two cases with
facts similar to those in this case. The first is Upper Snake
River Chapter of Trout Unlimited v. Hodel, 921 F.2d 232 (9th
Cir. 1990). In Upper Snake River, the panel posed the
following question: “Did the District Court err in concluding
that the National Environmental Policy Act) (“NEPA”) did
not require the Bureau of Reclamation (“Bureau”) to prepare
environmental impact statements (“EIS”) before periodically
adjusting the flow of water from the Palisades Dam?” The
panel responded: “We are clear that the answer is ‘No.’” Id.
at 233.

To apply the reasoning of Upper Snake River, we need
only change the name of the dam in the question posed. To
me, the answer is equally clear. The Bureau of Reclamation
was not required to prepare an Environmental Impact
 SAN LUIS V. JEWELL 163

Statement when it adopted the reasonable and prudent
alternatives from the Biological Opinion to set flow limits for
water controlled by the Central Valley water project.

Much like the water supply in the Central Valley, the
amount of water in the Snake River “fluctuates considerably
from year to year, depending on the amount of snow pack in
the mountains. . . .” Id. Water from the river is captured in
reservoirs and the water flow is controlled and regulated by
dams. See id. The Plaintiffs challenged the Bureau of
Reclamation’s decision to reduce the flow rate below 1,000
cubic feet per second in times of less precipitation. See id. at
234. In rejecting the contention that NEPA was applicable,
the district court held:

In the case of the Palisades Dam, the
fluctuating flows are routine actions which are
contingent upon Mother Nature for snow-
pack, runoff, precipitation and carryover. As
part of its routine and ongoing operations, the
Bureau of Reclamation fluctuates the flows
depending upon weather conditions past and
future. Overall, the Court views the
fluctuation of flows below Palisades as
“ongoing operations,” which do not have to
comply with . . . NEPA.

Id. (alterations omitted) (emphasis added).

Once again and without equivocation, the panel
concluded: “[W]e are clear that the district court’s
conclusion was correct.” Id. The panel explicitly determined
that water flow fluctuation was not a “major Federal action”
under NEPA. Id. As with Central Valley operations, the
164 SAN LUIS V. JEWELL

Palisades project was completed well before the effective date
of NEPA. See id. According to the panel, post-construction
fluctuation of water flow constituted routine operation of the
dam rather than a major action triggering NEPA
requirements. See id. at 234–35.

The panel relied heavily on the reasoning in Trinity
County v. Andrus, 438 F. Supp. 1368 (E.D. Cal. 1977). In
Trinity County, Plaintiffs sought an injunction prohibiting the
Bureau of Reclamation from lowering the water level in the
Central Valley water project during a drought year. See id. at
1371. The district court framed the issue as “not whether the
actions are of sufficient magnitude to require the preparation
of an [Environmental Impact Statement], but rather whether
NEPA was intended to apply at all to the continuing
operations of completed facilities. . . .” Id. at 1388 (citation
omitted).

The district court distinguished water flow determinations
from cases “when a project takes place in incremental stages
of major proportions or when a revision or expansion of the
original facilities is contemplated . . .” Id. (citations omitted).
The district court in Trinity County also noted that the Bureau
of Reclamation had not enlarged its diversion capacity, or
revised its procedures or standards for release of water or
draws from reservoirs. See id. at 1388–89. Under these
circumstances, the requirements of NEPA simply were not
triggered. See id. Taking its cue from Trinity County, the
panel in Upper Snake River reasoned that:

The Federal defendants in this case had
been operating the dam for upwards of ten
years before the effective date of [NEPA].
During that period, they have from time to
 SAN LUIS V. JEWELL 165

time and depending on the river’s flow level,
adjusted up or down the volume of water
released from the Dam. What they did in
prior years and what they were doing during
the period under consideration were no more
than the routine managerial actions regularly
carried on from the outset without change.
They are simply operating the facility in the
manner intended. In short, they are doing
nothing new, nor more extensive, nor other
than that contemplated when the project was
first operational. . . .

921 F.2d at 235.

Because the Bureau of Reclamation similarly continued
to do nothing more than “monitor[ ] and control . . . the flow
rate to ensure that the most practicable conservation of water
is achieved,” id. at 235–36, the panel concluded that no
Environmental Impact Statement was required. See id. at
236; see also Grand Canyon Trust v. United States Bureau of
Reclamation, 691 F.3d 1008, 1021–22 (9th Cir. 2012), as
amended (citing Upper Snake River and similarly concluding
that execution of annual operating plans regulating
fluctuations in water flow was “not a major federal action
requiring compliance with NEPA”); Burbank Anti-Noise
Group v. Goldschmidt, 623 F.2d 115, 116 (9th Cir. 1980)
(“An EIS need not discuss the environmental effects of mere
continued operation of a facility. . . .”) (citation omitted).

The second case with similar facts is Douglas County v.
Babbitt, 48 F.3d 1495 (9th Cir. 1995). In Douglas County,
the panel considered whether the designation of critical
habitat under the Endangered Species Act required
166 SAN LUIS V. JEWELL

compliance with NEPA. See id. at 1497. In the process of
deciding that issue, the panel focused on the difference
between the requirements and purpose of NEPA and the
requirements and purpose of the Endangered Species Act.
“The [Endangered Species Act] furthers the goals of NEPA
without demanding an [Environmental Impact Statement]. . .”
Id. at 1506 (emphasis added). Indeed, “[t]he [Endangered
Species Act] is a substantive statute whose goal is to prevent
extinction . . . By designating critical habitats . . . the
Secretary is working to preserve the environment . . .” Id. at
1506 (citations and internal quotation marks omitted). The
panel concluded that the designation of critical habitat
“further[ed] the purpose of NEPA [and] [r]equiring the
[agency] to file an EIS would only hinder its efforts at
attaining the goal of improving the environment.” Id.
(citation and internal quotation marks omitted).

In contrast to the substantive nature of the Endangered
Species Act, “NEPA is essentially a procedural statute
designed to insure that environmental issues are given proper
consideration in the decisionmaking process. . . .” Trustees
for Alaska v. Hodel, 806 F.2d 1378, 1382 (9th Cir. 1986)
(citation omitted). “NEPA does not mandate particular
results, but simply provides the necessary process to ensure
that federal agencies take a hard look at the environmental
consequences of their action.” Kootenai Tribe of Idaho v.
Veneman, 313 F.3d 1094, 1115–16 (9th Cir. 2002), overruled
in part on other grounds, Wilderness Soc’y v. United States
Forest Serv., 630 F.3d 1173, 1178–79 (9th Cir. 2011)
(citation and internal quotation marks omitted). “The purpose
of NEPA is to provide a mechanism to enhance or improve
the environment and prevent further irreparable damage.”
Drakes Bay Oyster Co. v. Jewell, No. 13-15227, 2014 WL
114699 at *12 (9th Cir. Jan. 14, 2014), as amended (citation
 SAN LUIS V. JEWELL 167

and internal quotation marks omitted). An Environmental
Impact Statement implements the NEPA regulations by
“rigorously explor[ing] and objectively evaluat[ing] all
reasonable alternatives to a proposed action . . .” Southeast
Alaska Conserv. Council v. Federal Highway Admin.,
649 F.3d 1050, 1056 (9th Cir. 2011) (quoting 40 C.F.R.
§ 1502.14(a)). An agency does not violate NEPA by
declining to re-examine alternatives that have previously been
evaluated. See, e.g., Honolulu Traffic.com v. Federal Transit
Admin., No. 13-15277, 2014 WL 607320 at *6 (9th Cir.
February 18, 2014).

It is undisputed that the Biological Opinion “rigorously
explore[d] and objectively evaluate[d] all reasonable
alternatives” over hundreds of pages. 40 C.F.R. § 1502.14(a).
Indeed, the express purpose of the Biological Opinion was to
develop reasonable and prudent alternatives to the existing
activities that were harmful to the environment. See
50 C.F.R. § 402.14 (setting forth formal consultation
procedures to ensure that the impact of proposed actions on
listed species and their critical habitat are fully considered);
see also Save the Yak Committee v. Block, 840 F.2d 714, 718
(9th Cir. 1988), as amended (noting the NEPA requirement
that an environmental assessment “include brief discussions
of the environmental impacts of the proposed action and
alternatives”) (emphasis added).

As with the habitat listing in Douglas County, the
Biological Opinion in this case served the same function as an
Environmental Impact Statement. See 48 F.3d at 1498 (“The
EIS is a procedural obligation designed to assure that
agencies give proper consideration to the environmental
consequences of their actions. . . .”) (citation and internal
quotation marks omitted). Because the NEPA requirements
168 SAN LUIS V. JEWELL

are designed to “provide a mechanism to enhance or improve
the environment and prevent further irreparable damage” to
the environment, no Environmental Impact Statement is
needed “for federal actions that conserve the
environment. . . .” Id. at 1505 (citation and footnote reference
omitted). In this circumstance, as in Douglas County, “the
NEPA procedure seem[s] superfluous.” Id. at 1503 (internal
quotation marks omitted). As the panel recognized in Drakes
Bay, if the federal decision “is essentially an environmental
conservation effort,” NEPA obligations are not triggered.
2014 WL 114699 at *12. The panel explained that the
Endangered Species Act “furthers the goals of NEPA without
demanding an EIS. . . .” Id. (quoting Douglas County,
48 F.3d at 1505, 1506) (emphasis in Drakes Bay). We should
be similarly “reluctant to make NEPA more of an
obstructionist tactic to prevent environmental protection than
it may already have become.” Id. at *13 (quoting Douglas
County, 48 F.3d at 1508). At this point, imposing an overlay
of NEPA requirements “would only hinder [the Bureau of
Reclamation’s] efforts at attaining the goal of improving the
environment.” Douglas County, 48 F.3d at 1506 (citation
omitted). See Majority Opinion, p. 149 ([T]he preparation of
an EIS will not alter Reclamation’s obligations under the
ESA. . . .”). There is no need to require the Bureau of
Reclamation to engage in scientific busywork. “This case
simply does not present the type of situation NEPA was
intended to address.” Burbank Anti-Noise Group, 623 F.2d
at 117.

I respectfully dissent.