Bald and Golden Eagles - Status, trends, and estimation of sustainable take rates in the United States (Status Report)

U.S. Fish & Wildlife Service

Bald and Golden Eagles

Population demographics and estimation of
sustainable take in the United States, 2016
update

ii

Bald and Golden Eagles

Population demographics and estimation of

sustainable take in the United States, 2016

update

April 26, 2016

Authors
Brian A. Millsap, Emily R. Bjerre, Mark C. Otto, Guthrie S. Zimmerman, and Nathan L. Zimpfer
U.S. Fish and Wildlife Service

Division of Migratory Bird Management

Disclaimer:
The information in this report is intended to aid in the development of regulations and inform eagle
management decisions by the U.S. Fish and Wildlife Service. The recommendations and findings in the
report do not constitute U.S. Fish and Wildlife Service policy, but they will be considered by the U.S. Fish
and Wildlife Service as it sets eagle management policies. Any use of trade, product, or firm names is for
descriptive purposes only and does not imply endorsement by the U.S. Government.
Suggested Citation:
U.S. Fish and Wildlife Service. 2016. Bald and Golden Eagles: Population demographics and estimation of
sustainable take in the United States, 2016 update. Division of Migratory Bird Management, Washington
D.C., USA.

iii

Acknowledgments

We gratefully acknowledge contributions of data from tagged golden eagles by Bryan Bedrosian, Pete Bloom,
James Cain, Ross Crandall, Robert Domenech, Daniel Driscoll, Jamey Driscoll, Mark Fuller, Rick Gerhardt,
Al Harmata, Todd Katzner, Robert Knight, Craig Koppie, Brian Latta, Mike Lockhart, Mark Martell, Carol
McIntyre, Libby Mojica, Robert Murphy, Gary Roemer, Steve Slater, Jeff Smith, Dale Stahlecker, Brian
Washburn, and Jim Watson. Andrew Dennhardt provided additional information on eastern U.S. golden eagle
population estimates. Bruce Peterjohn and Danny Bystrak with the U.S. Geologic al Survey Bird Banding
Lab helpfully provided banding data. The analyses and report greatly benefited from reviews and assistance
provided by members of the Service’s Eagle Technical Assessment Team, and we are especially grateful for
a thorough review by Todd Katzner. Analyses of the bald eagle post-delisting survey data were improved
by input from John Sauer and William Link, and population models and harvest analyses were improved by
suggestions and input from Michael Runge, John Sauer, and Leslie New. Gabriela Chavarria (U.S. Fish and
Wildlife Service, Science Applications) was instrumental in funding a significant portion of the on-going
PTT-tagging study for golden eagles. We appreciated comments on an earlier draft by Leon Kolankiewicz,
and Eliza Savage. Any errors or omissions are the responsibility of the senior author.

iv

Executive Summary

In June 2014 the U.S. Fish and Wildlife Service (Service) announced its intent to consider several revisions to
regulations at 50 CFR, part 22 that pertain to permits to take bald (Haliaeetus leucocephalus) and golden
(Aquila chrysaetos) eagles. The Service is preparing a Programmatic Environmental Impact Statement
(PEIS) to evaluate the potential effects of the revised regulations on eagle population status. The PEIS
will analyze alternatives that include both conservative and liberal take rates for both species, consistent
with the overall management objective of maintaining stable or increasing populations relative to estimated
population levels in 2009. The liberal alternatives will use take rates estimated from the median values
for relevant parameters (e.g., population size, growth rates), and the conservative alternatives will use the
20𝑡ℎ quantile values of parameter estimates. The alternatives will also consider different configurations
of eagle management units (EMUs): (1) the current EMUs, which are Bird Conservation Regions (BCRs)
for golden eagles, and approximately Service regional boundaries for bald eagles based on nest densities;
and (2) the four administrative migratory bird flyways (i.e., Atlantic, Mississippi, Central, and Pacific). To
inform the evaluation of the PEIS alternatives, a subgroup of the Service’s Eagle Technical Assessment Team
compiled recent information on population size and trend of both species of eagle, generated estimates of
recent survival and fecundity rates, and used these data in models to predict future population trends and
the ability of each species to withstand additional mortality in the form of permitted take. This document
summarizes the findings from those analyses.
The team (hereafter we) estimated population size for the bald eagle in the coterminous United States
(U.S.) using a population model in conjunction with estimates of the number of occupied nesting territories
(representing the number of breeding pairs) in 2009 from a comprehensive dual-frame aerial survey. That
population size estimate combined with a previous estimate of population size for Alaska was 143,000
(20𝑡ℎ quantile = 126,000) bald eagles for the entire U.S. in 2009. This represents an increase in population
size since 2007 in the coterminous U.S. (the year the final rule for delisting under the Endangered Species Act
was published). We attribute the difference to improved survey and estimation efforts, as well as increases
in bald eagle numbers. Consistent with the population model, independent Breeding Bird Survey (BBS)
indices indicated bald eagles are continuing to increase over much of the U.S. We used a potential biological
removal model to estimate sustainable take rates and limits with the goal of maintaining at least the 2009
population level, and concluded that under the liberal alternative bald eagles over most of the country can
support an annual take rate of 8% (20𝑡ℎ quantile = 6% under the conservative alternative). The exceptions
are the Southwestern U.S., where population growth potential is lower, and Alaska, where limited survey
information led managers to select a lower management objective factor; there, the sustainable take rates
are 4.5% (20𝑡ℎ quantile = 3.8%) and less than 1%, respectively. Nationally, the annual bald eagle take limit
with these rates would be approximately 6,300 eagles under the liberal alternative and 4,200 eagles under the
conservative alternative.
We estimated population size for the golden eagle by first estimating a population size for the western
coterminous U.S. using a composite model that integrated multi-year information from a late summer aerial
transect survey over the interior western U.S. with information from the BBS. Population size for Alaska
could not be estimated directly. Instead, we used results from mid-winter aerial transect surveys in 2014
and 2015 over the same area as the interior western U.S. summer transect survey to estimate the increase
in population size between late summer and winter. The increase was used as a coarse estimate of the
size of the overwintering migrant population. We allocated 24% of the winter increase to Alaska as a
conservative population estimate, assuming migrants originated proportionately across western Canada and
v

Alaska. A population size estimate for eastern North America was available from the literature. Combining
the western coterminous U.S., Alaska, and eastern U.S. estimates, total population size for the golden
eagle in the U.S. (including Alaska) was approximately 39,000 (20𝑡ℎ quantile = 34,000) in 2009 and 40,000
(20𝑡ℎ quantile = 34,000) in 2014. The population trend estimate from the composite model was stable, but
an estimate from a population model similar to that used for the bald eagle suggested the population in the
coterminous western U.S. might be declining towards a lower equilibrium size. Thus, taking into account the
uncertainty, the available data for golden eagles are somewhat equivocal, with count data suggesting a stable
population but with demographic data forecasting a slight decline.
We used banding data obtained from the United States Geological Survey Bird Banding Lab from
1968–2014 to estimate contemporary age-specific survival rates. We also used a data set of unbiased cause-of­
mortality information for a sample of 386 satellite-tagged golden eagles from 1997–2013 to estimate the effect
of current levels of anthropogenic mortality on those survival rates. Anthropogenic factors were responsible
for about 56% of satellite-tagged golden eagle mortality, but rates of anthropogenic mortality varied among
age-classes, ranging from 34% for first-year eagles to 63% for adults. We estimated the maximum rate
of population growth for the golden eagle in the U.S. in the absence of existing anthropogenic mortality
was 10.9% (20𝑡ℎ quantile = 9.7%). Sustainable take under these conditions is close to 2,000 individuals
(20𝑡ℎ quantile = 1,600). However, available information suggests ongoing levels of human-caused mortality
likely exceed this value, perhaps considerably. Thus, the data from satellite tags lends further support to the
suggestion from the demographic models that current survival rates may be leading to a decline in population
size.
The Service also has a need to apply take values to nest disturbance and loss. We updated metrics for
converting take via nest disturbance and nesting territory loss to debits from the EMU take limits for bald and
golden eagles. The current policy is that for each instance of authorized take through disturbance in each year
the nest is disturbed, the Service reduces EMU take limits by the median number of young that would have
been expected to fledge from the disturbed territory. The updated median productivity values are 1.12 for the
bald eagle (0.73 in the Southwest region only), and 0.54 for the golden eagle. By carrying forward the above
debits from the EMU take limits for a period of years equal to the species or population-specific generation
time (10 years for the bald eagle—12 years in the Southwest, and 11 years for the golden eagle), we also
calculated a take value for nesting territory loss (i.e., the territory becomes permanently vacant).
In addition to setting EMU take limits, the Service has established local-area population (LAP) thresholds
for permitted take when authorized take in a local area might have long-term negative consequences at that
scale. The primary objective of LAP take limits is to minimize chances of extirpation of local breeding or
wintering populations of eagles. The LAP take thresholds are cumulative, such that all ongoing Servicepermitted take and any new take under consideration for a permit is taken into account. This take is in addition
to any existing ongoing unpermitted take that is occurring in the LAP. As such, the LAP take analysis is a
form of cumulative effects analysis for each eagle take permit. Unlike EMUs, the LAP area is unique to each
prospective permit and is defined as the area of the permitted activity bounded by the 90𝑡ℎ quantile of the
natal dispersal distance for golden eagles (109 mi), and the median natal dispersal distance of females for
bald eagles (86 mi). The Service has identified LAP take-rates of ≥1% as being of concern, and rates of 5%
being the maximum of what should be considered. We analyzed the effects of the 5% take threshold on LAPs
for each species of eagle and showed that for bald eagles the additional take could result in a reduction of
the equilibrium population size in the LAP area of 38%. For golden eagles, which currently appear to be at
quasi-equilibrium, the 5% threshold could result in a decline of 80% to a new lower equilibrium. In both
cases, extirpation of the local area population appeared unlikely under this policy.
When authorized take exceeds EMU take limits, Service policy is that take must be effectively offset
by compensatory mitigation such that there is no net increase in mortality. Currently, the only offsetting
mitigation measure the Service has enough information to confidently apply in this manner is retrofitting of
power lines to reduce eagle electrocutions (although the Service does consider other offsetting mitigation
vi

options on an experimental basis). Offsetting mitigation is mostly an issue affecting take authorization
for golden eagles, as EMU take limits are set at zero requiring all authorized take to be offset. Based on
the cause-specific mortality rates analyzed, we estimated that 500 (20𝑡ℎ quantile = 280) golden eagles are
electrocuted in the U.S. annually. This estimate provides an indication of the number of golden eagle deaths
the Service can expect to offset though electrocution abatement until proven methods to reduce other forms
of mortality are available.
The Service currently implements the eagle take permit program within the context of an adaptive
management framework that requires regular updates of population size estimates. We offer a possible regime
for conducting these surveys that balances information needs with costs and logistics, and that would allow
updating of population size estimates every six years. Bald eagle surveys would be conducted in years three
and six, and paired summer-winter golden eagle surveys in the first/second, and fourth/fifth years of each
six-year period. Data collected should be used to re-assess population status, revise and update population
size estimates by EMU, and to update—and if necessary modify—EMU or LAP take restrictions.

vii

Table of Contents

Acknowledgments

iv

Executive Summary

v

Introduction
Eagle Management Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

1

Bald Eagle
Demographic Rates and Characteristics . . .
Survival . . . . . . . . . . . . . . . . .
Productivity . . . . . . . . . . . . . . .
Matrix Population Model . . . . . . . .
Population Size . . . . . . . . . . . . . . . .
Number of Occupied Nesting Territories
Total Population Size . . . . . . . . . .
Population Trajectory . . . . . . . . . . . . .
Management Unit Comparison . . . . . . . .

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

3

3

3

3

3

6

6

8

8

9

Golden Eagle
Demographic Rates and Characteristics
Survival . . . . . . . . . . . . . .
Causes of Mortality . . . . . . . .
Productivity . . . . . . . . . . . .
Matrix Population Model . . . . .
Population Size . . . . . . . . . . . . .
Western U.S. . . . . . . . . . . . .
Alaska . . . . . . . . . . . . . . .
Eastern U.S. . . . . . . . . . . . .
Total U.S. . . . . . . . . . . . . .
Population Trajectory . . . . . . . . . .
Management Unit Comparison . . . . .

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

12

12

12

13

13

15

15

15

17

18

18

18

18

.
.
.
.
.
.
.
.

22

22

22

23

24

25

26

27

29

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

Effects of Take
Take Limits at the Scale of Eagle Management Units .
Approach for Estimating Sustainable Take Limits
Sustainable Take for Bald Eagles . . . . . . . . .
Sustainable Take for Golden Eagles . . . . . . . .
Metrics for Take as a Result of Nest Disturbance . . . .
Metrics for Take as a Result of Territory Loss . . . . .
Take Limits at the Scale of the Local Eagle Population
The Role of Compensatory Mitigation . . . . . . . . .
viii

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

Population Monitoring

31

Literature Cited

32

Appendices

36

A1 Review of Eagle Productivity Data
36

Literature Review and Data Compilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

Bibliography: Bald Eagle Productivity Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

Bibliography: Golden Eagle Productivity Data . . . . . . . . . . . . . . . . . . . . . . . . . . 38

A2 Bald and Golden Eagle

Productivity Meta-Analysis
Abstract . . . . . . . . .
Introduction . . . . . . .
Methods . . . . . . . . .
Data . . . . . . . .
Model . . . . . . .
Results . . . . . . . . . .
Discussion . . . . . . . .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

40

40

40

41

41

41

42

43

A3 The 2009 National Bald Eagle Post-Delisting Survey and Estimation Results
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Survey Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Dual-frame Survey Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Stratification for the 2009 Survey and Analysis . . . . . . . . . . . . . . . . . . . .
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sample Unit Selection and Sample Size . . . . . . . . . . . . . . . . . . . . . .
Surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Estimating Occupied Nests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
List Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Area Estimation: Correcting for Detection Probability With Multiple Observers
Area Estimation: New Occupied Nest Estimation . . . . . . . . . . . . . . . . .
Area Estimation: Area-Only Estimation . . . . . . . . . . . . . . . . . . . . . .
Dual-Frame Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
List Coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
List-Frame Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Area-Frame Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Dual-Frame Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The Low-Density Strata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Area-Only Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
List Coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

58

58

58

58

59

59

59

60

61

61

62

64

64

65

65

65

65

67

68

68

68

71

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

A4 Updated Golden Eagle Population Size Estimate in the Western U.S.
73

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

ix

A5 USFWS Policy Update to the Use of Eagle Natal Dispersal Distances in Permitting Decisions
Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Decision Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Technical Issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Inconsistent Terms Issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Alternative 1a (ETAT-recommended alternative) . . . . . . . . . . . . . . . . . . . . . .
Alternative 1b . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Alternative 2a . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Alternative 2b . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Decision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

x

97

97

98

98

100

100

100

100

100

101

101

List of Tables

1
2
3
4
5
6
7
8
9
10
11
12
13
14
A2-1
A2-2
A2-3
A2-4
A2-5
A2-6
A2-7
A3-1
A3-2
A3-3
A3-4
A3-5
A3-6
A3-7
A3-8
A3-9
A4-1
A4-2
A4-3

Candidate models for bald eagle survival rates and band recovery probabilities, 1995–2014.
Annual survival rate estimates for bald eagles, 1995–2014. . . . . . . . . . . . . . . . . .
Estimated total bald eagle population size in 2009, sustainable harvest rates, and harvest

limits by potential EMU. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Summary of bald eagle banding and recovery locations summarized by 2009 EMUs. . . . .
Summary of bald eagle banding and recovery locations summarized by Flyway. . . . . . .
Candidate models for golden eagle survival rates and band recovery probabilities, 1968–2014.
Annual survival rate estimates for golden eagles, 1968–2014. . . . . . . . . . . . . . . . .
Estimated total annual golden eagle mortality attributable to different causes. . . . . . . . .
Golden eagle annual survival rate estimates with and without anthropogenic mortality. . . .
Estimated total golden eagle population size in 2014, sustainable harvest rates, and harvest

limits by potential EMU. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Golden eagle banding locations and subsequent mortality recoveries by 2009 EMU. . . . .
Golden eagle banding locations and subsequent mortality recoveries summarized by Flyway.
Take associated with disturbance to nesting bald eagles and the loss of occupied nesting

territories. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Take associated with disturbance to nesting golden eagles and the loss of occupied nesting

territories. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
AIC values for models to explain variation in bald and golden eagle productivity. . . . . . .
Regional prediction estimates of bald and golden eagle productivity. . . . . . . . . . . . .
Productivity model random effect standard errors with 95% credible intervals. . . . . . . .
Bald eagle productivity model median random effects apart from region. . . . . . . . . . .
Golden eagle productivity model median random effects apart from region. . . . . . . . . .
Bald eagle studies, simple productivity ratios, and final random effect model median

estimates for each area and year combination. . . . . . . . . . . . . . . . . . . . . . . . .
Golden eagle studies, simple productivity ratios, and final random effect model median

estimates for each area and year combination. . . . . . . . . . . . . . . . . . . . . . . . .
Geographic sampling areas for the 2009 post-delisting survey. . . . . . . . . . . . . . . . .
Survey capture history summaries and estimated detection probabilities. . . . . . . . . . .
List-frame estimates of total occupied bald eagle nests by strata. . . . . . . . . . . . . . . .
Estimates of new, occupied nests by area-frame strata. . . . . . . . . . . . . . . . . . . . .
List-frame, area-frame, and dual-frame estimates of occupied nests by stratum. . . . . . . .
Number of recorded occupied nests for areas of low nest density. . . . . . . . . . . . . . .
Estimates of occupied nests by strata based on area search plot data only. . . . . . . . . . .
Total list-only, area-only, and dual-frame estimates of occupied nests for the high-density

strata. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Proportion of the total estimated nests included on nest lists by strata. . . . . . . . . . . . .
Proportion of BCRs used to post-stratify BCR-specific golden eagle population size esti­
mates to a Flyway scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Population size estimates for golden eagles in the Central and Pacific Flyways, and the total

western U.S., 1967–2014. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Population size estimates of golden eagles in each BCR in the western U.S., 1967–2014. . .
xi

4

5

8

11

11

12

13

14

15

19

20

21

27

28

44

45

45

48

49

52

56

61

63

66

67

68

69

70

70

71

74

79

83

A5-1 Examples of bald and golden eagle natal dispersal distance criteria and implications for take

benchmarks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

xii

List of Figures

1
2
3
4
5
6
7
8
9
10
11
12
13
A2-1
A2-2
A2-3
A2-4
A2-5
A2-6
A3-1
A4-1
A4-2
A4-3
A4-4
A4-5
A4-6

Eagle management unit configurations under consideration by the Service. . . . . . . . . .
Diagram of the Lefkovitch (stage-structured) population model for the bald eagle, and

corresponding population projection matrix. . . . . . . . . . . . . . . . . . . . . . . . . .
Example of the linear relationship used to incorporate density-dependent effects on fecundity

into eagle population projection models. . . . . . . . . . . . . . . . . . . . . . . . . . . .
Estimated number of occupied bald eagle nesting territories by 2009 bald eagle management

units and the apparent change from 2007 to 2009. . . . . . . . . . . . . . . . . . . . . . .
Projected bald eagle population in the U.S. and the Southwest from 2009–2109. . . . . . .
Diagram of the stage-structured population model for the golden eagle, and corresponding

population projection matrix. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Comparison of time series for golden eagles in the western U.S. . . . . . . . . . . . . . . .
Population projection matrix trajectory for the golden eagle allowing for density-dependent

effects on fecundity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Harvest yield curve for the bald eagle in the U.S., excluding the Southwest region, under

liberal and conservative alternatives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Histogram of the difference between estimated current human-caused mortality of golden

eagles and the estimated sustainable take rate at maximum sustainable yield. . . . . . . . .
Simulated effect of added take on golden eagle populations. . . . . . . . . . . . . . . . . .
Effect on bald eagle local area populations (LAP) of a 5% additive harvest rate. . . . . . .
Effect on golden eagle local area populations (LAP) of a 5% additive harvest rate. . . . . .
Bald eagle productivity for the Southwest U.S. and the U.S. excluding the Southwest. . . .
Bald eagle productivity model random effects due to year given study. . . . . . . . . . . .
Bald eagle productivity model random effects due to study apart from regional and area

differences. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Golden eagle productivity for the U.S. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Golden eagle productivity model random effects due to year given study. . . . . . . . . . .
Golden eagle productivity model random effects due to the study random effects apart from

regional differences. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Survey strata for the 2009 bald eagle post-delisting survey. . . . . . . . . . . . . . . . . .
North American Bird Conservation Initiative bird conservation regions (BCRs). . . . . . .
Updated golden eagle BCR-specific trends compared to those estimated by Millsap et al.

(2013). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Comparison of golden eagle trends in the Pacific and Central Flyways and the total trend for

the entire western U.S. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Comparison of Flyway-specific time series of golden eagle population size estimates from

1967–2014. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Comparison of time series for golden eagles in the western U.S. based on data from

1967–2010 (Millsap et al. 2013) and updated data (1967–2014). . . . . . . . . . . . . . . .
Comparison between Millsap et al. (2013) and updated scaling factors used to adjust BBS

indices to a population estimate for golden eagles. . . . . . . . . . . . . . . . . . . . . . .

xiii

2

5

7

7

10

16

16

17

24

25

26

29

30

44

46

47

49

50

51

60

75

76

77

77

78

78

Introduction

In June 2014 the U.S. Fish and Wildlife Service (Service) announced its intent to consider several revisions to
regulations at 50 CFR, part 22 that pertain to permits to take bald (Haliaeetus leucocephalus) and golden
(Aquila chrysaetos) eagles. As part of this process, the Service is preparing a Programmatic Environmental
Impact Statement (PEIS) to assess potential effects of the revised regulations on bald and golden eagle
populations. The PEIS will analyze alternatives that include both conservative and liberal take rates for both
eagle species, consistent with the overall management objective of maintaining the potential for stable or
increasing populations relative to 2009 estimates. The Service’s Eagle Management Team (EMT) decided that
for the liberal alternatives take rates would be estimated using the median values for relevant parameters (e.g.,
population size, growth rates), and that conservative alternatives would use the 20𝑡ℎ quantiles of parameter
estimates. The PEIS alternatives will also consider two different configurations of eagle management
units (EMUs): 1) the current EMUs, which are Bird Conservation Regions (BCRs) for golden eagles, and
management units based on nest densities for bald eagles (U.S. Fish and Wildlife Service 2009b); and 2) the
four administrative migratory bird flyways (i.e., Atlantic, Mississippi, Central, and Pacific; U.S. Fish and
Wildlife Service 2014, 2015).
The PEIS must consider the most current relevant information for both eagle species. A subteam of
the Service’s Eagle Technical Assessment Team (ETAT) began work in February 2015 assembling relevant
scientific data and conducting analyses to support the PEIS. Much of this work has focused on gathering data
to estimate sustainable take rates and take limits for both eagle species. The subteam (hereafter team, or we)
compiled recent information on population size and trend of both species of eagle, generated estimates of
recent survival and fecundity rates, and used these data in models to predict future population trends and the
ability of each species to withstand additional mortality in the form of permitted take.
Herein, we describe approaches used in conducting technical analyses to inform the PEIS, then summarize
the results with consideration of the Service’s proposed management objectives. The trend and population
status information are presented together under relevant subheadings for each species of eagle. Because the
data available for bald and golden eagles differs, the subheadings and approaches used for the two species
differ in some cases. The analysis of resilience to additional permitted take (harvest) by both species is
covered in the final section of this document. Throughout, we present means or medians and 20𝑡ℎ quantiles
for parameter estimates that are used directly in the calculation of values under liberal and conservative PEIS
alternatives, otherwise we present 95% confidence limits (or credible intervals when Bayesian methods are
used) for estimates.

Eagle Management Objectives
In 2009, the Service established management objectives for bald and golden eagles as part of the Final
Environmental Assessment on the Nonpurposeful Eagle Take Regulations (U.S. Fish and Wildlife Service
2009a,b). The management objectives at that time were to maintain stable or increasing populations of both
species of eagle within a set of described EMUs, with 2009 as the baseline, for 100 years into the future. In
2009, EMUs for bald eagles were based on nest densities and approximated Service regions, and EMUs for
golden eagles were BCRs in the western U.S. (Figure 1). The use of different management units for the two
eagle species reflected biological differences and differences in data available.
The Service is proposing to retain the 2009 management objectives, but is considering revising the EMUs
to follow the four administrative migratory bird flyways with some modifications (available data presented
1

later in this report suggest that breaking the Pacific Flyway into three separate EMUs may be warranted for
the bald eagle, and combining the Mississippi and Atlantic Flyways may be warranted for the golden eagle).
In addition, the Service is proposing to address uncertainty in estimates of population size and resiliency to
increased permitted take by considering distribution medians for relevant parameters for liberal alternatives,
and 20𝑡ℎ quantiles for conservative alternatives. We use 𝑁2009 to refer to Service population objective(s) in
formulas and in the text.

Figure 1. The 2009 eagle management units (EMUs) for bald eagles (top left) and golden eagles (top right, U.S. Fish
and Wildlife Service 2009b), and the administrative flyways (bottom, U.S. Fish and Wildlife Service 2014, 2015),
which are under consideration as alternative EMU configurations by the Service. The blue lines indicate 100∘ W
longitude, and the red line indicates 40∘ N latitude.

2

Bald Eagle

Demographic Rates and Characteristics
Survival
We estimated bald eagle survival rates using banding data provided by the U.S. Geological Survey Bird
Banding Laboratory (BBL). Given evidence of continued population growth (see Population Trajectory),
we limited the survival analysis to 1995–2014. This data set included 14,805 banding records and 296 dead
recoveries. We estimated annual survival rates using a dead-recovery model with the Seber parameterization
in Program MARK (Cooch and White 2014). We evaluated a set of 10 candidate models that included age,
year, geographic covariates on survival, and geographic variation in recovery probability (the probability that
dead, marked individuals are reported). We then used the best-supported model (based on model-selection
using an overdispersed and small-sample adjustment [QAIC𝑐 ] of the Akaike information criterion [Burnham
and Anderson 2002]) from this set to evaluate a second set of six models that included linear and quadratic
time trends (Table 1). The best-supported models from this analysis included two age-classes, first-year
(hatching year [HY]) and older (after-hatching-year [AHY]), and two geographic areas—the Southwest region
west of the 100𝑡ℎ meridian and south of 40∘ N latitude (hereafter Southwest), and the rest of the U.S. We used
the most parsimonious model within two QAIC𝑐 units of the top model to estimate annual survival rates. For
the bald eagle, the best supported model included a geographic effect on recovery probability, but a model
with constant recovery probability that required estimates of fewer parameters had an only slightly greater
QAIC value, so we used the latter model. We used the Markov chain Monte Carlo (MCMC) method in
Program MARK so our final estimates were in the form of samples from the Bayesian posterior distributions
of each parameter. Our estimates of annual survival were 66% for HY and 93% for AHY bald eagles in the
Southwest, and 86% for HY and 91% for AHY bald eagles over the rest of the U.S. (Table 2).
Productivity
We conducted a thorough literature review and obtained estimates of bald eagle productivity from 17 study
areas in the U.S. from 1995–2014 (Appendix A1). Productivity ranged from 0.52–2.29 young fledged per
occupied nesting territory (Table A2-6). We then used a random-effects meta-analysis model and estimated
predictive distributions for bald eagle productivity (Appendix A2). Productivity differed by region with
lower productivity in the Southwest (median = 0.73, 95% credible interval = 0.40–1.36) than the rest of the
U.S. (median = 1.12, 95% credible interval = 0.73–1.72).
Matrix Population Model
We used the demographic rates and variances described above to parameterize a post-breeding Lefkovitch
(stage-structured) matrix to model potential population growth and the stable-age distribution for the bald
eagle (Caswell 2001). For population projections, our model included age categories 0–1, 1–2, 2–3, and
> 3 years (Figure 2). The model allowed bald eagles > 3 years old to breed, but limited the number of total
breeding opportunities to the estimated number of occupied nesting territories in 2009. This limitation reflects
that availability of suitable nesting sites is a proximate limiting factor in many raptor populations (Hunt 1998,
Millsap and Allen 2006), and that younger individuals are less competitive than older individuals for breeding
slots (Turrin 2014, Lien et al. 2015). By using the estimated number of occupied bald eagle nesting territories
in 2009, we conservatively capped projected population growth to levels consistent with known numbers of
3

Table 1. Candidate models evaluated to explain variation in bald eagle survival rates and band recovery probabilities, based on dead band recoveries from
1995–2014 analyzed using a Seber parameterization in Program MARK.

4

AIC𝑐 Weights

Model
Likelihood

Number of
Parameters

3,555.41
0.00
3,556.64
1.23
3,557.16
1.75
3,557.58
2.17
3,557.77
2.36
3,558.73
3.32
3,559.53
4.12
3,559.83
4.42
3,560.26
4.86
5,979.11 2, 423.70

0.32
0.17
0.13
0.11
0.10
0.06
0.04
0.04
0.03
0.00

1.00
0.54
0.42
0.34
0.31
0.19
0.13
0.11
0.09
0.00

6
5
3
4
6
5
5
5
5
1,248

0.00
19.82
20.29
21.74
23.12
494.45

1.00
0.00
0.00
0.00
0.00
0.00

1.00
0.00
0.00
0.00
0.00
0.00

5
6
7
5
7
312

Modelsb

QAIC𝑎𝑐

S(age2_hy&ahy+zone2_SW) r(zone2_E&W)
S(age2_hy&ahy+zone2_SW) r(.)
S(age2_hy&ahy) r(.)
S(age3_hy&SA&ahy+zone2_SW) r(.)
S(age2_hy&ahy+zone2_SW) r(zone2_SW)
S(age4) r(.)
S(age3_hy&SA&ahy+zone2_SW) r(zone2_SW)
S(age2_hy&ahy+zone2_S&N) r(.)
S(age2_hy&ahy+zone2_E&W) r(.)
S(age+time+zone+age*time*zone) r(age+time+zone+age*time*zone)
Models with Time Trends
S(zone2_SW) r(zone2_E&W)
S(zone2_SW+quadratic) r(.)
S(zone2_SW+quadratic) r(zone2_E&W)
S(zone2_SW+linear) r(.)
S(age_quadraticAHY + zone2_SW r(zone2_E&W)
S(age+time+zone+age*time*zone) r(age+time+zone+age*time*zone)

3,556.64
3,576.46
3,576.93
3,578.38
3,579.76
4,051.09

a

Δ QAIC𝑐

QAIC𝑐 is an adjusted AIC𝑐 value to account for overdispersion in the data, as measured by the value 𝑐^. 𝑐^ = 1.00 in the absence of overdispersion. 𝑐^ = 1.23 for the global
model in this case.
b
Abbreviations are as follows: S= survival; r= recovery probability; age= age covariate, with the number of age-classes in the model specified by the numeral and the
age-classes identified as hy= hatching-year, ahy= after-hatching-year, SA= subadult (ages 2–4); zone= geographic covariate, with the numeral indicating the number of
geographic zones in the model and the geographic zones identified as E=east of the 100𝑡ℎ meridian, W= west of the 100𝑡ℎ meridian, S=south of 40∘ N latitude, N= north
of 40∘ N latitude, SW= south of 40∘ N latitude and west of the 100𝑡ℎ meridian. Symbols are as follows: . = constant, +=additive effect, *= interactive effect. Linear refers
to a model with a linear time effect over the full time-series, and quadratic refers to a model with a quadratic time effect over the full time-series.

Table 2. Annual survival rate estimates for bald eagles, 1995–2014, based on a dead-bird band
recovery model with the Seber parameterization in Program MARK. Estimates are from the
best-supported model in Table 1.

Annual Survivala
HY, U.S. excluding Southwest
AHY, U.S. excluding Southwest
HY, Southwest
AHY, Southwest
Recovery Probability
a

Estimate

Lower 95%
Credible Interval

Upper 95%
Credible Interval

0.86
0.91
0.66
0.93
0.03

0.80
0.86
0.31
0.73
0.03

0.90
0.94
0.87
0.99
0.04

Abbreviations are: HY = hatching-year; AHY = after-hatching-year; Southwest = U.S. west of the 100𝑡ℎ
meridian and south of 40∘ N latitude.

⎛

⎞ ⎛
⎞ ⎛
⎞
𝑁0,𝑡+1
0
0
0
𝑏(218, 23) × 𝑛(0.59, 0.13)
𝑁0,𝑡
⎜ 𝑁1,𝑡+1 ⎟ ⎜𝑏(158, 26)
⎟ ⎜ 𝑁1,𝑡 ⎟
0
0
0

⎜
⎟ =
⎜
⎟ ×
⎜
⎟
⎝
𝑁2,𝑡+1 ⎠
⎝

⎠
⎝
𝑁2,𝑡 ⎠

0
𝑏(218, 23)
0
0

𝑁≥3,𝑡+1
0
0
𝑏(218, 23)
𝑏(218, 23)
𝑁≥3,𝑡

Figure 2. Diagram of the Lefkovitch (stage-structured) population model for the bald eagle, and corresponding
population projection matrix. On the diagram, stages correspond to year age-classes 0–1, 1–2, and 2–3. Stage > 3
includes all subsequent age classes. S denotes survival rates, where 𝑆1 is the rate for first-year survival and 𝑆2 is the
rate for survival in all subsequent years (see Table 1 for justification for the 2-age class model and Table 2 for rates). F
denotes per-individual fecundity, with reproduction contingent on breeding slots being available. The matrix shown is
parameterized with values for bald eagles outside the Southwest U.S.; the model for the Southwest used demographic
values specific to that region. Survival rates (rows 2–4) were sampled from beta distributions and fecundity (row 1) is
shown as sampled from a normal distribution, but in the actual models we used the specific random-effects predictive
distribution. To estimate the stable-age distribution, we used a similar model but with a non-reproductive 3–4 year-old
stage, and a reproductive > 4 year-old stage (see text for details).

5

nesting territories (including the number of “floating" adults not associated with nesting territories) rather
than assuming continued increases in breeding opportunities.
We used the stable-age distribution in combination with the direct estimates of numbers of occupied
nesting territories in 2009 to estimate total bald eagle population size (see below). In using the stableage distribution we made the implicit assumption that bald eagle populations are at equilibrium (Caswell
2001). We acknowledge this is unlikely given evidence of continued population growth, but believe this
is a conservative assumption that likely leads us to underestimate total population size because in growing
populations a larger percentage of the population is in the younger age classes (Lande et al. 2003). For this
model we were not able to estimate the number of non-breeding individuals in the adult stage using the
minimum number of known nesting territories. Rather, because bald eagles < 4 years old breed infrequently
(Bowman et al. 1995, Turrin 2014), we revised our projection model to include a 3–4 year non-reproductive
age class, and we added a fifth age class that included all adults ≥ 4 years of age. We assumed all individuals
in the adults ≥ 4 years stage were associated with a nesting territory.
We ran 10,000 iterations of 100-year simulations for each population projection. Although we projected
forward 100 years, we note that future predictions are only valid and relevant to the degree that environmental
and biological conditions remain as they were over the time period when vital rates were measured. This
critical assumption is less likely to be met the further into the future the projections go and should be kept in
mind when evaluating this information. For each simulation, we sampled survival rates from beta distributions
with shape parameters derived from the pertinent survival rates, and we randomly sampled the appropriate
random-effects predictive distributions for fecundity values.
We incorporated a density-related response in fecundity by allowing mean productivity to increase
as the proportion of adult floaters decreased. We implemented this as a linear increase in per-individual
fecundity, ranging from the median of the estimated predictive distribution at current (2015) population
levels assuming all territories are occupied, to the maximum annual observed rate (see ‘year:study area
productivity’ combinations cited in Appendix A2 [Table A2-6]) when territory occupancy rates approached
zero (Figure 3). There is strong evidence for density-related dampening of fecundity rates in increasing
raptor populations (Kauffman et al. 2004, Bretagnolle et al. 2008, Fasce et al. 2011), as well as evidence
for increasing per-individual fecundity rates in some decreasing populations (Whitfield et al. 2004b, 2007,
Baldwin et al. 2012). This type of response in fecundity may be the result of interference competition
(Kauffman et al. 2004, Bretagnolle et al. 2008) or increasing nesting habitat heterogeneity as lower-quality
nesting territories are occupied at higher densities (Ferrer and Donazar 1996, Carrete et al. 2006).
In addition to using this basic model to estimate the population growth potential and stable age distribution
of bald eagles, we used it to estimate demographic carrying capacity. Our estimates of carrying capacity
assume nest site availability rather than food or other resources will be the proximate factor limiting growth,
consistent with Moffat’s equilibrium theory (Hunt 1998).

Population Size
Number of Occupied Nesting Territories
We obtained estimates of the number of occupied bald eagle nesting territories in the coterminous
U.S. from a dual-frame survey coordinated by the Service in 2009 (Tables A3-5, A3-8). We used the stratified
survey estimates to calculate the number of occupied nesting territories for each EMU under consideration
by redistributing the strata estimates to the EMUs according to the proportion of the total strata area within
each EMU (Figure 4). The estimated number occupied nesting territories across EMUs in the coterminous
U.S. was slightly lower than the dual-frame survey estimate of occupied nesting territories due to rounding
and imperfect alignment of the survey strata and EMUs. We combined the EMU estimates for the coterminous
U.S. with an existing estimate for Alaska from 2009 (U.S. Fish and Wildlife Service 2009b) and calculated
6

Figure 3. Example of the type of linear relationship used to incorporate density-dependent effects on fecundity into
eagle population projection models. This example is for the golden eagle, where per-individual fecundity ranges from
0.275 young per year with 100% of nesting territories occupied (the mean of the predictive distribution derived from a
literature review of contemporary fecundity rates), to 0.62 as the occupancy rate approaches zero (the maximum
fecundity rate observed in any one year in the studies included in the literature review). Thus, fecundity in the model
increases linearly according to the described linear model as populations fall and nesting territories go unoccupied.

EMU

2007

2009

2009 CI

Alaska
15,000 15,000 (12,471–17,529)
Great Lakes
3,452 5,879 (4,769–6,989)
Lower Mississippi
447 1,207
(753–1,661)
Mid-Atlantic
952 1,766 (1,373–2,159)
New England
603
645
(577–713)
Northern Rocky Mtns.
564
339
(0–751)
Pacific
1,039 2,587 (2,073–3,101)
Rocky Mtns. & Plains
200
338
(281–3950
Southeast
1,210 2,611 (2,180–3,042)
Southwest
51
176
(119–233)
Total

23,518 30,548 (24,524–36,572)

Figure 4. A map (left) showing the apparent change in estimated occupied bald eagle nesting territories by bald eagle
management unit (U.S. Fish and Wildlife Service 2009b) between the time of delisting in 2007 (data used were from
2007 or earlier) and 2009 (except Alaska, estimated from the post-delisting survey flown in 2009). The table (right)
shows the number of estimated occupied nesting territories for both time periods and the 95% credible intervals for the
2009 estimates (the delisting numbers did not include explicit quantification of uncertainty). The different methods
used to estimate nesting population size between the two intervals likely contribute to the differences shown here.
Estimates for Alaska are based on limited local survey information (U.S. Fish and Wildlife Service 2009b).

7

nearly 30,600 (95% confidence interval = 24,500–36,600) occupied bald eagle nesting territories in the U.S. in
2009.
Total Population Size
The stable-age distribution from the five-stage matrix population model estimated that 40% of bald eagles
in the Southwest and 43% elsewhere were ≥ 4 years old. Under the assumption that all of these individuals
occupied nesting territories, we estimated the total population size for each region in the coterminous
U.S. using the formula:
𝑁𝑂𝑐𝑐.𝑇 𝑒𝑟𝑟 * 2
𝑁𝑇 𝑜𝑡𝑎𝑙 =
,
𝑝(𝐴𝑔𝑒 ≥ 4)
where 𝑁𝑂𝑐𝑐.𝑇 𝑒𝑟𝑟 is the estimated number of occupied nesting territories (approximately 200 for the Southwest
and 30,400 elsewhere in the U.S.), and 𝑝(𝐴𝑔𝑒 ≥ 4) is the proportion of the population ≥ 4 years old. We
estimated a median bald eagle population size of approximately 143,000 nationally (20𝑡ℎ quantile= 126,000);
estimates for each prospective bald eagle EMU are provided in Table 3.

Population Trajectory
The BBS index trend estimate for the bald eagle over the entire BBS coverage area for the period
1966–2012 is 5.3% (95% confidence interval = 4.1–6.6%), though trends for the area that includes Alaska
Table 3. Estimated total bald eagle population size in 2009 at the median (N) and 20𝑡ℎ quantile (𝑁20 ) by potential eagle
management unit (EMU). Estimated sustainable harvest rates (h) and harvest limits (H) are also presented with the median
and 20𝑡ℎ quantile for each EMU. Harvest rates and limits are constrained by a management objective factor (𝐹0 ) such that
take is consistent with the objective of maintaining the potential for an equilibrium population size ≥ 𝑁 .

N

𝑁20

h

ℎ20

H

𝐻20

Source

Great Lakes
Lower Mississippi
Mid-Atlantic
New England
Northern Rocky Mountains
Pacific
Rocky Mountains and Plains
Southeast
Southwest
Alaska-FWa
Atlantic Flyway
Central Flyway
Mississippi Flyway
Pacific Flyway, South
Pacific Flyway, North

70,544
27,440
5,640
8,244
3,017
1,569
12,102
1,583
12,190
648
70,544
22,279
3,209
31,706
447
14,792

62,935
24,065
4,622
7,201
2,729
720
10,504
1,411
10,788
533
62,935
20,387
1,163
27,334
391
13,296

0.007
0.080
0.080
0.080
0.080
0.080
0.080
0.080
0.080
0.045
0.007
0.080
0.080
0.080
0.045
0.080

0.008
0.060
0.060
0.060
0.060
0.060
0.060
0.060
0.060
0.038
0.008
0.060
0.060
0.060
0.038
0.060

494
2,195
451
660
241
126
968
127
975
29
494
1,782
257
2,537
20
1,183

494
1,444
277
432
164
43
630
85
647
20
494
1,223
70
1,640
15
798

USFWS (2009b)
Post-Delisting Survey
Post-Delisting Survey
Post-Delisting Survey
Post-Delisting Survey
Post-Delisting Survey
Post-Delisting Survey
Post-Delisting Survey
Post-Delisting Survey
Post-Delisting Survey
USFWS (2009b)
Post-Delisting Survey
Post-Delisting Survey
Post-Delisting Survey
Post-Delisting Survey
Post-Delisting Survey

Total US
Total US (excluding AK)

142,977
72,434

125,508
62,572

6,273
5,772

4,240
3,742

Management Unit
Alaskaa

a

Population size estimates for Alaska are approximations based on limited survey information. Because of this added uncertainty, the
Service proposes to use a lower management objective factor for Alaska that results in a take limit comparable with that estimated in
2009 (U.S. Fish and Wildlife Service 2009b). The median value of h is used in every case.

8

have been closer to stable at 0.08% (95% confidence interval = -8.41–5.44%; Sauer et al. 2014). We observed
increases in the number of occupied nesting territories and inferred population size between pre-2007 (the
time of delisting under the Endangered Species Act; U.S. Fish and Wildlife Service 2009b) and 2009 in all
current bald eagle EMUs except the Northern Rockies (Figure 4). Data used to support delisting were nest
counts provided by the states, whereas the 2009 estimates are based on the dual-frame survey which corrected
for detection and dealt with issues of sampling effort. The percent list coverage (the proportion of the total
estimated nests, regardless of occupancy status, represented on state nest lists) for the dual-frame survey
ranged from 48–100%, meaning some state nests lists were missing as many as 52% of the total estimated
nest structures (Appendix A3). Thus, it is likely that some part of the difference in population size between
the two time periods is a result of differences in survey and analysis methodologies. In particular, the decline
indicated for the Northern Rockies EMU is not reflected in the BBS data, which shows a population change
of 8.7% (95% confidence interval = 5.1–13.1%) from 2003–2013 (Sauer et al. 2014).
We estimated future bald eagle populations using the previously described population projection matrix
model and the conservative assumption that the number of suitable bald eagle nesting territories will not
increase above the 2009 estimate. Given limitations of the Alaska data and evidence from BBS indices that
bald eagle populations are growing more slowly there, we did not model projections for Alaska and assumed
that Alaska’s bald eagle population will remain stable (though demographic rates suggested continued
growth is possible). With these constraints, the model forecasts that the number of bald eagles in the
U.S. outside the Southwest will continue to increase until populations reach equilibrium at about 228,000
(20𝑡ℎ quantile = 197,000; Figure 5). The model predicts that bald eagles in the Southwest will continue to
increase until reaching equilibrium at about 1,800 (20𝑡ℎ quantile = 1,400). Again, these projections assume
underlying demographic rates and other environmental factors remain unchanged through time and assume
food and other factors will not become limiting. Additionally, these projections do not take into account
forecasted changes in climate nor how such changes may affect bald eagle population vital rates, population
size, food availability, or other factors.

Management Unit Comparison
To assess whether the EMU configurations under consideration (2009 EMUs and Flyway EMUs) differed
in terms of capturing bald eagle movements across seasons and life stages, we used 1,021 band recovery
records from the BBL data set from 1931–2014. We compared the frequency with which banded bald eagles
were recovered within the same EMU as they were originally banded. Eagles were not banded systematically
or randomly with respect to EMUs, however, if eagles frequently move distances or with directionality that
is incongruent with the shape and size of a particular management unit configuration, we would expect a
difference even in basic summary metrics. We found that 84% (range = 43–100%) of bald eagles were banded
and later recovered in the same 2009 EMU (Table 4) compared to 94% (range = 67–96%) recovered in the
same Flyway EMU (Table 5). In part, the difference may be a reflection of the larger geographic size of the
Flyway EMUs. This is supported by the increase in the percentage of recoveries (98%) when adjacent 2009
EMUs are also considered. However, the consistently higher percentage of recoveries in the same Flyway
EMU may also suggest a general association between bald eagle movements and Flyway EMUs.

9

Figure 5. Projected bald eagle population in the United States excluding the Southwest (top) and in the Southwest
(bottom) from 2009–2109 using a stage-structured population projection matrix and demographic rates derived from
data over the period from 1995–2015. The blue shading indicates the 95% confidence intervals around the estimates
(dark blue lines). The gray, dashed lines are the 95% confidence intervals for population size in 2009, which is the
minimum bald eagle population size objective. Model projections assume demographic rates remain as estimated, and
that the number of suitable nesting territories does not increase above 2009 levels, which accounts for the plateau in
population size.

10

Table 4. Bald eagle banding locations by 2009 eagle management unit (EMU; Figure 1)
and subsequent location of band recoveries (total recovered). Band recovery is classified
as the percentage of the total recovered bands that were recovered in the same EMU, the
same EMU or an adjacent EMU, or a non-adjacent EMU (other) relative to the banding
location.

Recovered (%):
EMU Banded

Total
Recovered

Same
EMU

Same EMU or
Adjacent EMU

Other
EMU

16
510
51
133
140
46
20
36
39
30

100
90
43
85
86
70
100
58
74
73

100
99
100
100
99
96
100
100
85
87

0
1
0
0
1
4
0
0
15
13

1,021

84

98

2

Alaska
Great Lakes
Lower Mississippi
Mid-Atlantic
New England
Northern Rocky Mountains
Pacific
Rocky Mountains and Plains
Southeast
Southwest
Total

Table 5. Bale eagle banding locations by adminstrative Flyway (Figure
1) and subsequent location of band recoveries (total recovered). Band
recovery is classified as the percentage of the total recovered bands that
were recovered in the same Flyway, the same Flyway or an adjacent
Flyway, or a non-adjacent Flyway (other) relative to the banding
location.

Recovered (%)
EMU Banded
Atlantic
Central
Mississippi
Pacific
Total

Total
Recovered

Same
EMU

Same or
Adjacent EMU

Other
EMU

319
46
540
116

94
67
96
93

100
100
100
98

0
0
0
2

1,021

94

99.7

0.3

11

Golden Eagle

Demographic Rates and Characteristics
Survival
We estimated golden eagle survival rates using banding data from 1968–2014 provided by the BBL. A
longer time series was necessary for reliable estimates than for the bald eagle, however, evidence suggests
that golden eagle populations across the western U.S. have been largely stable over that longer period of time
(Millsap et al. 2013), thus we assume survival rates have also been relatively stable. The data set included
10,627 banding records and 565 dead recoveries. As with bald eagles, we estimated annual survival rates
using a dead-recovery model with the Seber parameterization in Program MARK (Cooch and White 2014).
We evaluated a set of 10 candidate models that included age, time, linear and quadratic time trends in survival,
and—for all but the global model—a constant recovery probability (Table 6). The best-supported model
included four age-classes: <1 year (HY), second-year (SY), third-year (TY), and after-third-year (ATY). We
used this model and the approach described previously for bald eagles to estimate annual survival of 70%,
77%, 84%, 87% for the respective golden eagle age classes (Table 7).

Table 6. Candidate models evaluated to explain variation in golden eagle survival rates and band recovery
probabilities, based on dead band recoveries from 1968–2014 analyzed using a Seber parameterization in
Program MARK.

AIC𝑐
Weights

Model
Likelihood

Number of
Parameters

0.00
2.93
14.29
18.33
22.50
26.85
42.53
42.98
59.42

0.81
0.19
0.00
0.00
0.00
0.00
0.00
0.00
0.00

1.00
0.23
0.00
0.00
0.00
0.00
0.00
0.00
0.00

5
4
3
7
6
6
7
6
2

8,432.31 1, 146.70

0.00

0.00

728

Modelsb

QAIC𝑎𝑐

S(age4) r(.)
S(age3_hy&sa&aty) r(.)
S(age2 + hy&ahy) r(.)
S(age4 + quadratic) r(.)
S(age4 + linear_age2_hy) r(.)
S(age4 + linear) r(.)
S(age4 + quadratic_age2 _aty) r(.)
S(age4 + linear_age2_ahy) r(.)
S(.) r(.)
S(age4 + time + age4*time)
r(age4 + time + age4*time)

7,285.61
7,288.53
7,299.89
7,303.94
7,308.11
7,312.46
7,328.14
7,328.59
7,345.03

Δ QAIC𝑐

a

QAIC𝑐 is an adjusted AIC𝑐 value to account for overdispersion in the data, as measured by the value 𝑐^. 𝑐^ = 1.00 in
the absence of overdispersion. 𝑐^ = 1.02 for the global model in this case.
b
Abbreviations are as follows: S = survival; r = recovery probability; age = age covariate, with the number of
age-classes in the model specified by the numeral and the age-classes identified as hy = hatching-year, ahy=
after-hatching-year, SA = subadult (ages two through four). Symbols are as follows: . = constant, + = additive
effect, * = interactive effect. Linear refers to a model with a linear time effect over the full time-series, and
quadratic refers to a model with a quadratic time effect over the full time-series.

12

Table 7. Annual survival rate estimates for golden eagles, 1968–2014, based on a
dead-bird band recovery model with the Seber parameterization in Program
MARK. Estimates are from the best-supported model in Table 6.

Annual Survivala
HY
SY
TY
ATY
Recovery Probability
a

Estimate

Lower 95%
Credible Interval

Upper 95%
Credible Interval

0.70
0.77
0.84
0.87
0.06

0.66
0.73
0.79
0.84
0.06

0.74
0.81
0.88
0.89
0.07

Abbreviations are: HY = hatching-year; SY = second year; TY = third-year,
ATY = after-third-year.

Causes of Mortality
For golden eagles, the Service had access to a database of information (see Acknowledgments) on 386
satellite-tagged golden eagles over the period 1997–2013. As of 2013, 139 of those birds had died and were
recovered; cause-of-death was known for 97 eagles. Radio- and satellite-tagged raptors are an important
source of unbiased information on causes of death compared to bands, for which recovery probability varies
by the type of death (e.g., raptors struck by vehicles are more likely to be encountered than raptors that
die of starvation [Kenward et al. 1993]). The Service was particularly interested in the relative extent of
human-caused (anthropogenic) mortality, in that this mortality is generally considered additive to natural
mortality rates (but see Chevallier et al. 2015). Using the 139 eagles that had died and were recovered, we
computed estimates of the overall proportion of annual mortality that was attributable to different factors
using a binomial generalized linear model (GLM) in a Bayesian framework; this approach allowed us to
include and account for the 42 eagles that died from unknown causes. Then, we used posterior distributions
of the rates of cause-specific mortality, age-specific survival rates, and population size (partitioned by age
class according to the stable-age distribution) to derive an estimate of the number of golden eagle mortalities
annually by cause of death for each golden eagle age class for the total population (Table 8).
The relative importance of anthropogenic mortality increased with age, with 63% of adult golden eagle
mortality caused by humans compared to 34% for HY golden eagles (Table 9). Moreover, annual golden
eagle survival rates would be approximately 10% higher without human-caused mortality, assuming the
reduction in anthropogenic mortality is additive, which may not be entirely the case, particularly for juvenile
golden eagles (Chevallier et al. 2015).
Productivity
As with bald eagles, we reviewed the literature and obtained estimates of golden eagle productivity
from 1995–2014. We included data from 12 study areas in the U.S. (Appendix A1) and used the same
meta-analysis framework as for bald eagles to characterize the distributions of golden eagle productivity
(Appendix A2). The best model for predicting productivity was a random-effects model with overdispersion
(Table A2-1) that estimated median productivity for the continental U.S. (Table A2-2) as 0.54 (95% credible
intervals = 0.40–0.75) young fledged per occupied nesting territory. Model selection did not support use of
region-specific productivity values, though this may in part be to the limited productivity data available.

13

Table 8. Causes of death for satellite-tagged golden eagles in North America, 1997–2013, extrapolated to estimate total annual golden eagle mortality attributable
to different causes. Extrapolations are based on estimated cause-specific proportions from a Bayesian binomial generalized linear model, and also account for
uncertainty in annual mortality rates and in age-class population sizes. The extrapolation assumes that this sample of satellite-tagged deaths are representative of
deaths overall, and was computed using the stable age distribution and age-specific survival rates described in the text. Age classes are age <1 year (HY), 1–3
years (Subadult), and >3 years (ATY).

Observed Deaths
Factor

14

Shot
Electrocution
Poisoning
Collision
Trap
Lead Toxicosis
Starvation/disease
Injury
Fighting
Predation
Drowning
Total

Type
Anthropogenic
Anthropogenic
Anthropogenic
Anthropogenic
Anthropogenic
Anthropogenic
Natural
Natural
Natural
Natural
Natural

Total Deaths Projected per Year (95% Credible Interval)

HY Subadult ATY
6
8
1
4
1
1
35
3

2
4
1
2
1
1

2
1

1

62

12

5
1
4
3
1
1
2
2
4

23

HY
141
193
21
93
21
21
876
68
3
44
21

(55–292)
(88–364)
(2–100)
(27–218)
(1–98)
(2–97)
(605–1,205)
(17–180)
(0–47)
(7–142)
(2–97)

Subadult
7
180
393
7
78
7
177
79
78
8
80

(0–153)
(27–526)
(126–806)
(0–154)
(5–362)
(0–160)
(28–538)
(5–370)
(5–363)
(0–152)
(5–369)

ATY
777
131
611
444
132
131
282
280
609
17
18

Total

(280–1,600)
926 (336–2,046)
(9–604)
504 (124–1,494)
(188–1,360) 1,025 (316–2,266)
(106–1,137)
545 (133–1,509)
(9–611)
231 (15–1,071)
(9–609)
160 (10–867)
(48–883)
1,334 (681–2,626)
(44–888)
427 (66–1,437)
(188–1,377)
690 (193–1,787)
(0–277)
69 (8–571)
(0–281)
119 (6–747)

1,503 (806–2,840) 1,093 (200–3,954) 3,432 (882–9,628) 6,029 (1,888–16,422)

Table 9. Golden eagle annual survival rate estimates with and without anthropogenic
mortality. The proportion of mortality caused by humans was estimated from a sample of
satellite-tagged golden eagles that died (see Table 8), and results presented here account
for uncertainty in the proportions of cause-specific mortality, survival rates, and population
size. Base survival rates were estimated from dead band recoveries using a Seber
parameterization in Program MARK.

Age Class
First-year
Cause-of Death
Anthropogenic
Natural
Survival Rate
Only natural mortality
All mortality

Subadult

Adult

0.34 (0.23–0.46) 0.57 (0.32–0.81) 0.63 (0.44–0.80)
0.66 (0.54–0.77) 0.43 (0.19–0.68) 0.37 (0.20–0.56)
0.80 (0.76–0.85) 0.92 (0.86–0.96) 0.93 (0.89–0.96)
0.70 (0.66–0.74) 0.80 (0.77–0.83) 0.87 (0.84–0.89)

Matrix Population Model
We used the above demographic rates and variances to parameterize a post-breeding Lefkovitch matrix
model for the golden eagle (Figure 6). We followed the same approach as for the bald eagle population
projection model. As discussed above (see Bald Eagle: Matrix Population Model), we assumed no growth in
the number of suitable nesting territories above 2009 levels. As for the bald eagle, we incorporated densitydependent effects on fecundity (Figure 3) such that per-individual fecundity ranged from approximately 0.28
young per year (the mean of the predictive distribution for productivity from the meta-analysis) with 100%
of nesting territories occupied and an estimated floater to breeder ratio of 1.13:1 (the estimated condition
in 2015), to 0.62 (the maximum fecundity rate observed in any one year for the studies included) as the
territory occupancy rate approaches zero. Thus, fecundity in the model increases linearly as populations fall
and nesting territories go unoccupied.
We used this model to estimate the stable age distribution of golden eagles and to project future population
trajectory, assuming environmental conditions remain as they were over the time these data were collected,
and that nest site availability and survival rates and not food will be the proximate factors limiting growth
(Figure 7). For golden eagles we also have direct estimates of current and historical population size (e.g.,
Millsap et al. 2013), and we were able to project future population trajectory based on empirical trend
estimates (Figure 8; Appendix A4).

Population Size
Western U.S.
Since 2006, the Service has funded an annual late-summer aerial transect survey to estimate golden
eagle population size over four BCRs in the interior western U.S. that account for about 80% of the western
U.S. golden eagle population (Nielson et al. 2014). Recently, Millsap et al. (2013) combined these data with
BBS indices in a hierarchical model to produce a composite estimate of golden eagle population size and
trend for the entire coterminous U.S. west of the 100𝑡ℎ meridian over the years 1967–2010. We updated the
Millsap et al. (2013) composite model estimates of golden eagle population size and trend through 2014 for
this analysis (Appendix A4). The updated summer population size estimates do not differ substantially from
those reported by Millsap et al. (2013), and indicate a late summer population averaging 31,000 (20𝑡ℎ quantile
= 29,000) individuals over the most recent decade (Figure 7). The updated composite model estimated the
15

⎞
⎛

⎞

0
0
0
𝑏(243, 47) × 𝑛(0.27, 0.03)

𝑁0,𝑡+1
𝑁0,𝑡
⎜𝑏(353, 159)
⎟
0
0
0

⎜ 𝑁1,𝑡+1 ⎟ ⎜
⎟ ⎜ 𝑁1,𝑡 ⎟
⎜
⎟ =
⎜
⎟ ×
⎜
⎟
0
𝑏(287,
85)
0
0
⎟ ⎝
𝑁2,𝑡 ⎠

⎝
𝑁2,𝑡+1 ⎠
⎜
⎝

⎠

0
0
𝑏(243, 47)
0
𝑁≥3,𝑡+1
𝑁≥3,𝑡
0
0
0
𝑏(816, 127)
⎛

⎞

⎛

Figure 6. Diagram of the stage-structured population model for the golden eagle, and corresponding population
projection matrix. See Figure 2 and text for a description of the model framework and vital rates. Fecundity, F (row 1)
is shown as sampled from a normal distribution, but in the actual models we used the specific random-effects predictive
distribution.

Figure 7. Comparison of time series for golden eagles in the western U.S. based on data from 1967–2010 (dashed line
with blue shading, Millsap et al. 2013) and updated data for the period 1967–2014 (red line with red shading, Appendix
A4). The lines are mean population estimates and colored shading represents the 95% credible intervals (CIs), with the
1967–2014 time series CIs shaded red, and the 1967–2010 time series CIs shaded blue. Note the large amount of
overlap between the CIs.

16

Figure 8. Population projection matrix trajectory for the golden eagle after model was adapted to allow for
density-dependent effects on fecundity, as described in the text. The blue area represents the upper and lower 95%
confidence limits, and the solid line is the median. The dashed lines represent the 95% credible intervals for the 2009
population estimate, which is the population objective.

total coterminous western U.S. population as 30,000 (20𝑡ℎ quantile = 27,000) for 2009 (Figure A4-2).
We combined the 2009 and 2014 western U.S. composite estimates with contemporary estimates of the
eastern U.S. and Alaska golden eagle populations to calculate the population goal (𝑁2009 ) and harvest limits,
respectively (see below).
Alaska
In 2014 and 2015, the Service funded aerial transect surveys over the same four-BCR area of the interior
west in January to estimate mid-winter population size (Nielson and McManus 2014, Nielson et al. 2015).
Golden eagles from natal areas above 60∘ N latitude are usually migratory (McIntyre et al. 2008), as are many
individuals from the subarctic regions of Canada and Alaska (Kochert et al. 2002). Thus, the mid-winter
population in the coterminous U.S. includes resident birds that remain in the coterminous U.S. year-round
and migrants that occur at more northern latitudes in the summer, but migrate into the coterminous U.S. for
the winter. Increases in counts from late summer to mid-winter likely provide a lower bound on the size of
the northern migratory population of western golden eagles. The survey estimated increases in the number of
golden eagles between late summer (August/September) and mid-winter (January) of 4,000 (95% credible
interval = 3,800–4,100) in 2013–2014, and 17,000 (95% credible interval = 14,900–20,200) in 2014–2015.
This mid-winter survey has not been conducted frequently enough to evaluate the meaning and significance
of the annual variability in the change in numbers of eagles between late-summer and winter. However, these
are the first data that allow approximation of the size of the high-latitude migratory golden eagle population
17

in western North America. Assuming the presumed northern migrant golden eagles are originating from
natal areas in Canada (west of the 100𝑡ℎ meridian) and Alaska in proportion to the relative area of those
regions (76% Canada, 24% Alaska), then in 2013–2014 and 2014–2015 around 1,000–4,000 mid-winter
migrants originated from Alaska. We used the larger estimate as our population size for Alaska for the liberal
PEIS alternatives, and the midpoint (2,500) as the population estimate for the conservative PEIS alternatives.
This assumes that all golden eagles in Alaska in the late summer are wintering in the coterminous U.S. In
comparison, in 2009, the Service coarsely estimated the size of the Alaskan golden eagle population at 2,400
individuals (U.S. Fish and Wildlife Service 2009b).
Eastern U.S.
Golden eagles occur frequently in eastern North America, primarily as winter migrants from breeding and
natal areas in eastern Canada (Morneau et al. 2015). Recently, the size of this population has been estimated
at 5,000 (20𝑡ℎ quantile = 4,000; Dennhardt et al. 2015), which corresponds with what is known about the
number of occupied nesting territories in the breeding range of this population in eastern Canada (Morneau
et al. 2015).
Total U.S.
We pooled western U.S., Alaska, and eastern U.S. population estimates to develop a total estimate of
golden eagle population size for the U.S. in 2014 for the purpose of computing contemporary harvest limits
(Table 10), presented in the Effects of Take section below. We used the 2009 estimate for the coterminous
western U.S. and contemporary estimates for Alaska and the eastern U.S. as our population goal for the
golden eagle (39,000, 20𝑡ℎ quantile = 34,000).

Population Trajectory
The updated summer golden eagle population trend for the coterminous western U.S. from the composite
model did not differ substantially from the trend reported by Millsap et al. (2013), with an annual rate-of­
change of 1.0 (95% credible interval = 0.99–1.01) over the most recent decade (Figure 7). The annual rate­
of-change from the demographic (population projection) model, however, averaged 0.998 (95% confidence
interval 0.997–0.999), and suggested that golden eagles in the coterminous western U.S. might be gradually
declining toward a new, lower equilibrium of about 26,000 individuals (Figure 8). Confidence limits for the
demographic model projection broadly overlapped the credible interval for the composite model projection,
so the results are generally consistent despite their differing ramifications. As noted previously, the validity
of future predictions under both models are dependent on continuation of the biological and ecological
conditions under which the vital rates were estimated.

Management Unit Comparison
To compare the different EMU configurations under consideration, we used 683 golden eagle band
recovery records from the BBL data set from 1926–2014. As with the similar data set for bald eagles, we
compared the frequency with which banded golden eagles were recovered within the same EMU as they
were originally banded (640 of the 683 banded were recovered within the U.S.). We found that 73% (range =
0–86%) of golden eagles were banded and recovered within the same 2009 EMU (Table 11), whereas 84%
(range = 50–87%) were in the same Flyway EMU (Table 12). Again, as with bald eagles, golden eagles were
not banded systematically or randomly with respect to EMUs.
18

Table 10. Estimated total golden eagle population size in 2014 at the median (N) and 20𝑡ℎ quantile (𝑁20 ) by potential
eagle management unit (EMU). Estimated sustainable harvest rates (h) and harvest limits (H) are also presented for each
quantile for each EMU. Harvest rates and limits are constrained by a management objective factor (𝐹0 , see text) such that
take is consistent with the objective of maintaining the potential for an equilibrium population size greater than or equal to
𝑁2009 , 29,659, the population objective for the coterminous western U.S.

N

𝑁20

h

ℎ20

H

𝐻20

Source

Alaska
Eastern
BCR 5
BCR 9
BCR 10
BCR 11
BCR 15
BCR 16
BCR 17
BCR 18
BCR 32
BCR 33
BCR 34
BCR 35
Atlantic/Mississippi Flyways
Central Flyway
Pacific Flyway

4,091
5,122
189
6,596
5,675
836
72
4,258
9,837
1,459
718
418
411
786
5,122
15,327
15,927

2,544
4,002
114
5,682
4,851
519
38
3,585
8,091
1,091
549
247
229
528
4,002
13,210
14,437

≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0

≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0
≈0

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

Nielson et al. 2014, 2015
Dennhardt et al. 2015
Attachment 4
Attachment 4
Attachment 4
Attachment 4
Attachment 4
Attachment 4
Attachment 4
Attachment 4
Attachment 4
Attachment 4
Attachment 4
Attachment 4
Dennhardt et al. 2015
Attachment 4
Attachment 4

Total (US west)
Total (Contiguous US and Alaska)

31,254
40,467

30,191
34,193

0
0

0

0

Management Unit

19

Table 11. Golden eagle banding locations and subsequent location of
mortality recoveries summarized by 2009 eagle management units (EMU),
which are Bird Conservation Regions (BCRs) for the golden eagle (Figure 1).
BCRs are referred to here by their BCR number. Band recovery is classified
as the percentage of the total recovered bands that were recovered in the
same BCR, the same BCR or an adjacent BCR, or a non-adjacent BCR
(other) relative to the banding location. BCRs that did not have any banded
eagles subsequently recovered were omitted.

Recovered (%)
BCR Banded

Total
Recovered

Same BCR

Same or
Adjacent BCR

Other
BCR

3
4
5
6
8
9
10
11
12
13
14
16
17
18
19
20
22
23
24
28
29
30
32
33
34
36

5
10
3
3
3
245
53
82
2
4
1
52
23
52
13
1
3
4
4
21
3
1
40
5
4
3

0
30
0
0
0
86
64
78
50
0
0
65
65
77
54
0
67
25
25
76
0
0
85
60
50
0

0
30
100
67
67
98
94
91
50
75
0
100
100
96
92
100
100
100
75
90
67
100
93
100
100
0

100
70
0
33
33
2
6
9
50
25
100
0
0
4
8
0
0
0
25
10
33
0
8
0
0
100

Total

640

73

93

7

20

Table 12. Golden eagle banding locations and subsequent mortality
recoveries summarized by adminstrative Flyway (Figure 1). Band
recovery is classified as the percentage of the total recovered bands that
were recovered in the same Flyway, the same Flyway or an adjacent
Flyway, or a non-adjacent Flyway (other) relative to the banding
location.

Recovered (%)
Total
Recovered

Same
EMU

Same EMU or
Adjacent EMU

Other
EMU

Atlantic
Central
Mississippi
Pacific

23
135
14
408

74
82
50
87

91
90
86
90

9
10
14
10

Total

580

84

90

10

EMU Banded

21

Effects of Take

The Service currently manages eagle take at two geographic scales, regional EMUs and the ‘local-area
population’ (LAP). As noted previously, for the PEIS the Service is considering two alternative EMU
configurations for the regional management scale—the EMUs established in 2009 and the four administrative
flyways, which may better represent geographic use across seasons. Unlike EMUs, the LAP is unique to
each prospective permit and is defined as the area of the permitted activity bounded by the 90𝑡ℎ quantile of
the natal dispersal distance for golden eagles, 109 mi, and the median female natal dispersal distance for
bald eagles, 86 mi (Millsap et al. 2014). These values were adopted by the EMT based on recommendations
by ETAT (Appendix A5) to update the values discussed in the Eagle Conservation Plan Guidance (ECPG)
following updated analyses published by Millsap et al. (2014).
Eagle take at the EMU-scale is governed by a take rate that is compatible with maintaining an equilibrium
population size equal to or greater than the population objective. Take limits at the LAP-scale, on the other
hand, apply only to take permitted or authorized by the Service and, while they allow for local population
declines under some situations, they are intended to prevent local extirpation of eagles—both breeding and
non-breeding. The Service has acknowledged that some, perhaps even most, eagles taken at a permitted
project will originate from natal areas outside the LAP. However, given fidelity to migration corridors and
wintering areas by both bald and golden eagles (McIntyre et al. 2008, Mojica et al. 2008), limiting take at the
LAP-scale has conservation benefits—which likely accrue to more than just eagles breeding within the LAP.
With take limits at both the EMU- and LAP-scales, across an EMU we would expect a landscape with
some areas in proximity to permitted projects with comparatively high levels of authorized anthropogenic
mortality, but offset by other areas where authorized anthropogenic take is low, averaging to a maximum
across the EMU equal to or less than the EMU take limit. In cases where take exceeds the EMU take limit, all
excessive take must be offset by mitigation that will commensurately reduce ongoing mortality from other
sources, such that there is no authorized increase in net mortality (U.S. Fish and Wildlife Service 2009a).

Take Limits at the Scale of Eagle Management Units
Approach for Estimating Sustainable Take Limits
We used a potential biological removal (PBR) model to estimate sustainable lethal take rates for both
species of eagle (Williams et al. 2002, Dillingham and Fletcher 2008, Runge et al. 2009). The PBR model
produces an estimate of the sustainable harvest rate (h, hereafter take rate) using the formula:
ℎ=

𝑟𝑚𝑎𝑥
𝐹0 ,
2

where 𝑟𝑚𝑎𝑥 = the maximum rate of population growth, sampled from the uncertainty distribution; and 𝐹0 = a
management objective factor, ranging from 0 (no harvest) to 2 (harvest rate = 𝑟𝑚𝑎𝑥 ), where a value of 𝐹0 = 1
is the maximum sustainable yield (MSY) and a value of 0.5 is 1/2 MSY. MSY occurs at a population size of
1/2 carrying capacity.
The maximum rate of population growth (𝑟𝑚𝑎𝑥 ) is rarely observed in nature, and no published estimates
exist for bald or golden eagles or any closely related eagle species. Under these circumstances 𝑟𝑚𝑎𝑥 is
typically estimated from available demographic data using one of several approaches; we tested two such
22

approaches. First, following Runge et al. (2009), we estimated 𝑟𝑚𝑎𝑥 using Slade’s equation (Slade et al.
1998):
1 = 𝑝𝜆−1 + 𝑙𝛼 𝑏𝜆−𝛼 − 𝜆𝛼 𝑏𝑝(𝜛−𝛼+1) 𝜆−𝜛+1 ,
where 𝑝 = the annual adult survival rate from the Seber dead-recovery models, sampled as a uniform
distribution between the lower and upper 95% credible limits on the estimates; 𝑙𝛼 = the product of annual
survival rates for ages 0–4 from the Seber dead-recovery models, sampled as uniform distributions between
the lower and upper 95% credible limits; 𝑏 = per individual fecundity, sampled as a uniform distribution
from the mean to the upper 95% credible limit of the random-effect predictive distribution for fecundity;
𝜛 = estimated maximum lifespan, which we obtained by expanding the annual survival rates, and based
on that expansion, sampled from a uniform distribution between 25–30 years; 𝛼 = age-at-first breeding,
estimated from Birds of North America accounts for each species (Buehler 2000, Kochert et al. 2002) as 4–6
years sampled as a continuous uniform distribution; and 𝜆 (the intrinsic growth rate) = 𝑟𝑚𝑎𝑥 + 1. The formula
is solved for 𝜆 by optimizing a solution on that parameter. Using Monte Carlo methods, we simulated each
parameter 10,000 times and solved for 𝜆, which provided 𝑟𝑚𝑎𝑥 . For the second approach, we parameterized
our demographic population models with uniform samples between the mean and upper 95% credible intervals
of the fecundity and survival distributions and solved for 𝜆. Both approaches produced similar results, thus
we used the demographic model estimates because they required fewer assumptions.
We used the sustainable take rate distribution, h, to calculate sustainable take limits (H) at time t using
the formula:
𝐻𝑡 = ℎ𝑁𝑡 ,
where 𝑁𝑡 = population size at time t, sampled from the uncertainty distribution for 𝑁𝑡 as described previously.
Thus, the estimate of 𝐻𝑡 for both eagle species is represented by a distribution that accounts for uncertainty
in both the sustainable take rate and population size at time t. The subscript is intended as a reminder that
both H and N are not constants, and must be updated regularly with monitoring information (see Population
Monitoring). Unless otherwise noted, we further constrained our estimates of h by setting 𝐹0 to a value
consistent with the objective of maintaining an equilibrium population size ≥ 𝑁2009 (Figure 9). For the liberalalternative, we used the medians of the parameter distributions to estimate H. For the conservative-alternative,
we accounted for uncertainty by estimating ℎ20 using the 20𝑡ℎ quantiles of 𝑟𝑚𝑎𝑥 and carrying capacity,
then multiplying by the 20𝑡ℎ quantile of the appropriate population size estimate to obtain 𝐻20 . Figure 9
provides an example of how h, H, ℎ20 , and 𝐻20 , were determined relative to the population objective using
standard harvest yield curve for the bald eagle (Williams et al. 2002, Runge et al. 2009). Estimates of carrying
capacity are a key component of the equilibrium harvest curve. We used our demographic model estimates of
equilibrium population size as our values of carrying capacity in harvest rate analyses, though actual carrying
capacity might occur at lower population levels if resources become limiting before demographic rates.
Ideally, 𝑟𝑚𝑎𝑥 is calculated in the absence of anthropogenic mortality. We had no data to estimate current
anthropogenic take for bald eagles. However, for golden eagles we were able to estimate the proportion
of mortality in each age class attributable to anthropogenic versus natural causes (Table 9). We used this
information to estimate “natural” 𝑟𝑚𝑎𝑥 (𝑟𝑚𝑎𝑥 in the absence of anthropogenic mortality) for the golden eagle,
as well as to quantify the amount of the potential sustainable take that is already occurring following an
approach similar to that used by Whitfield et al. (2004a) for Scottish golden eagles.
Sustainable Take for Bald Eagles
Outside the Southwest region, we estimated that 𝑟𝑚𝑎𝑥 for bald eagles with all current mortality was 20.6%
quantile = 18.4%), yielding h = 10.3% (ℎ20 = 9.2%). Our demographic-model estimate of carrying
capacity was 227,800 (20𝑡ℎ quantile = 197,500). To remain consistent with management objectives, we then
adjusted h to a level compatible with maintaining an equilibrium population ≥ 𝑁2009 by setting set 𝐹0 to 0.78

(20𝑡ℎ

23

Figure 9. Harvest yield curve for the bald eagle in the U.S., excluding the Southwest region, under liberal (coral) and
conservative (blue) alternatives. The sustainable take limit is the value of the y-axis at the intersection of the yield curve
and the population size objective (after Runge et al. 2009). In the case of the liberal alternative, the sustainable harvest
rate (h) at maximum sustainable yield (MSY) is 10%. However, we adjusted h to 8% unsing a management objective
factor of 0.78, which corresponds with an equilibrium population size of 143,000, which is our management objective.
We followed the same approach for the conservative alternative, but using the 20𝑡ℎ quantiles of the relevant estimates
rather than the medians. The take limits shown here do not correspond with the final take limits in Table 3 because the
take limit for Alaska was further constrained to meet specific regional management objectives.

(𝐹020 = 0.63). Using this approach, h = 8% (ℎ20 = 6%) for the bald eagle outside the Southwest. In the
Southwest, we estimated that 𝑟𝑚𝑎𝑥 = 17.9% (20𝑡ℎ quantile = 15%). Our demographic model suggested 𝑁2009
in the Southwest was less than 1/2 demographic carrying capacity. As it was of interest to managers to allow
for further bald eagle population growth in the Southwest, we set h to 1/2 the harvest rate at MSY (4.5%), and
ℎ20 to the 20𝑡ℎ quantile of 1/2 the MSY harvest rate (3.75%), rather than the higher take rates associated with
the 2009 population estimate. In Alaska, because of uncertainties in the population size estimate, managers
opted to maintain H and 𝐻20 at approximately 500, as was recommended in 2009 (U.S. Fish and Wildlife
Service 2009a); we adjusted 𝐹0 accordingly. Collectively, across EMUs the estimated bald eagle take limits
in the United States are 6,273 and 4,240 under the liberal and conservative alternatives, respectively (Table 3).
Sustainable Take for Golden Eagles
We estimated that natural 𝑟𝑚𝑎𝑥 (calculated in the absence of anthropogenic mortality) for the golden
eagle was 10.9% (20𝑡ℎ quantile = 9.7%), yielding h = 5.4% (ℎ20 = 4.9%). Our demographic-model estimate
of carrying capacity was 73,000 (20𝑡ℎ quantile = 64,000). H at our population objective would be 2,000
(ℎ20 = 1, 600) and MSY under these conditions would be 2,200. However, we estimated that currently
about 3,400 (95% credible interval = 935–9,253) golden eagles die annually from anthropogenic causes in
the U.S. (Table 8). Despite the considerable uncertainty in estimates of both MSY and current levels of
24

Figure 10. Histogram of the difference between the posterior distribution of the estimated current amount of
human-caused mortality of golden eagles in the United States (see Table 8) and the posterior distribution of the
estimated sustainable take rate at maximum sustainable yield. The blue-shaded area indicates the portion of the
probability distribution that lies within the 95% credible interval, the dashed vertical blue line is the median, and the
solid vertical blue line indicates zero. Note that most of the distribution lies above zero, which means estimated
sustainable take is less than the estimated levels of current human-caused mortality.

mortality, these data suggest golden eagles in the U.S. are currently experiencing more take than can be
sustained at the population objective or at MSY (Figure 10). This result is somewhat at odds with our estimate
of the stable long-term population trend from the composite model. It is possible that golden eagles are
compensating for the high unnatural mortality rate with increases in survival or fecundity to a greater degree
than we have allowed for in our demographic model. Regardless, adding additional unmitigated mortality
will either exacerbate the potential for declines, or steepen the rate of any decline that is presently occurring.
To illustrate this further, we computed the effect of added take for the golden eagle in 1% increments up to a
10% harvest rate using the population model described previously (Figure 11). All added take resulted in
population declines to new, lower equilibrium sizes. The upper 95% confidence intervals under all of the
scenarios tested were below 𝑁2009 and therefore not consistent with the Service’s population objective. Given
this, we use zero as the take value for both h and ℎ20 for the golden eagle (Table 10).

Metrics for Take as a Result of Nest Disturbance
For disturbance to have a population-level effect, it has to result in a loss of potential productivity. In
2009, the Service used the EMU-specific productivity (mean number of young fledged per occupied nesting
territory) for each species per year as the expected loss for each instance of authorized nest disturbance
(U.S. Fish and Wildlife Service 2009b). Here we follow the same approach with updated take values from the
appropriate random-effects predictive distributions from the productivity meta-analysis (Figures A2-1 and
A2-4). We used the median values of the distributions for the liberal alternatives, and the 80𝑡ℎ quantiles for
25

Figure 11. Simulated effect of added take on golden eagle populations. The top line is the projected golden eagle
population trend with no additional take. Each line below the top line represents the population projection with
increased take in 1% increments, up to a take rate of 10% in the bottom line. Shaded areas represent the 95%
confidence intervals on the projections and transition from blue to purple as the take rate increases.

the conservative alternatives to maintain a protective 20% probability of underestimating the productivity
potentially lost as a result of disturbance. Following this approach, for each instance of nest disturbance
predicted to result in loss of productivity, take thresholds for bald eagles outside the Southwest are debited by
1.12 or 1.33 eagles, under the liberal and conservative alternatives respectively, per year that the disturbance
occurs. For bald eagles in the Southwest, take thresholds are reduced by 0.73 or 0.95, and for golden eagles
by 0.53 or 0.59, respectively (Tables 13 and 14).

Metrics for Take as a Result of Territory Loss
Loss of an occupied nesting territory results in the recurring loss of annual production from that territory.
However, this loss of future production is difficult to estimate and account for in debiting take thresholds.
In 2009, the Service quantified future production lost from loss of an occupied territory by comparing
equilibrium population size with N and 𝑁 − 1 nesting territories, then debiting EMU take limits by the
difference (U.S. Fish and Wildlife Service 2009b). This approach assesses the effects of loss indirectly and
relates it to a future equilibrium population size rather than the population objective. Here, for each instance
of occupied territory loss, we subtract the mean annual per nesting-territory productivity from the EMU take
limit annually for the generation time of the eagle species (Tables 13 and 14). We define generation time as
the average age of breeders in the population (Caswell 2001, Bienvenu and Legendre 2015). Using this as the
temporal scale over which we account for productivity lost is biologically relevant and sufficiently long to
assure that potential longer-term effects can be accounted for by future adjustments to the EMU take limits
based on reassessments of eagle populations (see Population Monitoring below).
26

Table 13. Take limits associated with take as a result of disturbance to nesting bald eagles, and the loss of occupied bald
eagle nesting territories.
Debits to Take Limits

Eagle Management Unit
Alaskaa
Great Lakes
Lower Mississippi
Mid-Atlantic
New England
Northern Rocky Mountains
Pacific
Rocky Mountains and Plains
Southeast
Southwest
Alaska-FWa
Atlantic Flyway
Central Flyway
Mississippi Flyway
Pacific Flyway, South
Pacific Flyway, North
a

Per Instance
Per Instance
Generation Cumulative
Cumulative
Disturbance Take, Disturbance Take,
Time
Per Territory, Per Territory,
80𝑡ℎ Quantile
(years)
50𝑡ℎ Quantilea 80𝑡ℎ Quantilea
50𝑡ℎ Quantile
1.12
1.12
1.12
1.12
1.12
1.12
1.12
1.12
1.12
0.73
1.12
1.12
1.12
1.12
0.73
1.12

1.33
1.33
1.33
1.33
1.33
1.33
1.33
1.33
1.33
0.95
1.33
1.33
1.33
1.33
0.95
1.33

10
10
10
10
10
10
10
10
10
12
10
10
10
10
12
10

11.20
11.20
11.20
11.20
11.20
11.20
11.20
11.20
11.20
8.76
11.20
11.20
11.20
11.20
8.76
11.20

13.30
13.30
13.30
13.30
13.30
13.30
13.30
13.30
13.30
11.40
13.30
13.30
13.30
13.30
11.40
13.30

The per-instance take debit is applied annually for the generation time. These are the cumulative debits at the end of the specified
generation time.

We recognize that for golden eagles in particular, nesting territories are often occupied by successive
generations of individuals. Additionally, for both species, some nesting territories hold more value than others
(Millsap et al. 2015, Watts 2015). Moreover, it is often difficult to predict in advance whether an activity will
result in loss of a nesting territory, or simply the loss of a nest structure and cause a shift in use to an existing
or new alternative nest—which may have little or no consequence to the eagle population (Watts 2015). For
these reasons, each instance where loss of a nesting territory is a possible outcome requires additional review
on the part of Service biologists. Permitting the loss of high-value nesting territories with a long history of
occupancy and production could have greater population-level consequences.
We used the mean of the fertility rate schedule from the matrix demographic models (effectively the mean
age of breeders in the population) as the generation time. Generation time is 12 years for bald eagles in the
Southwest and 10 years for bald eagles in the rest of the U.S. Golden eagle generation time is 11 years. The
corresponding debits to take limits by EMU are given in Tables 13 and 14.

Take Limits at the Scale of the Local Eagle Population
The objective of the LAP take limit is to regulate take such that local populations are protected from
extirpation due to Service-authorized activities. Although the primary aim is to prevent extirpation of local
nesting populations, there is increasing evidence of strong philopatry to non-breeding areas in both species
of eagle (McIntyre et al. 2008, Mojica et al. 2008), and the LAP take limits also provide protection from
overharvest of wintering and migrating eagles. As noted above, LAP take limits pertain only to take permitted
or authorized by the Service, and are cumulative, taking into consideration all Service-authorized activities
affecting the LAP.
27

Table 14. Take limits associated with take as a result of disturbance to nesting golden eagles, and the loss of
occupied golden eagle nesting territories.
Debits to Take Limits
Per Instance
Per Instance
Generation Cumulative
Cumulative
Disturbance Take, Disturbance Take,
Time
Per Territory, Per Territory,
80𝑡ℎ Quantile
years
50𝑡ℎ Quantilea 80𝑡ℎ Quantilea
Eagle Management Unit
50𝑡ℎ Quantile
Alaska
Eastern
BCR 5
BCR 9
BCR 10
BCR 11
BCR 15
BCR 16
BCR 17
BCR 18
BCR 32
BCR 33
BCR 34
BCR 35
Atlantic/Mississippi
Central Flyway
Pacific Flyway
a

0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53

0.59
0.59
0.59
0.59
0.59
0.59
0.59
0.59
0.59
0.59
0.59
0.59
0.59
0.59
0.59
0.59
0.59

11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11

5.83
5.83
5.83
5.83
5.83
5.83
5.83
5.83
5.83
5.83
5.83
5.83
5.83
5.83
5.83
5.83
5.83

6.49
6.49
6.49
6.49
6.49
6.49
6.49
6.49
6.49
6.49
6.49
6.49
6.49
6.49
6.49
6.49
6.49

The per-instance take debit is applied annually for the generation time. These are the cumulative debits at the end of the
specified generation time.

The Service identified LAP take-rates of ≥ 1% as being of concern, and rates of 5% being at the maximum
of what should be considered (U.S. Fish and Wildlife Service 2013). Note that the take authorized (within
the LAP take limits) is in addition to the average background rate of anthropogenic mortality—for golden
eagles, this is about 10% (Table 9). Thus, total anthropogenic mortality for a golden eagle LAP experiencing
the maximum permitted take rate of 5% is likely about 15%. We do not have similar mortality information
for bald eagles, but the same principle (that take authorized would be in addition to an existing background
rate of anthropogenic mortality) applies to that species as well. As part of the LAP analysis for both species,
Service biologists also consider available information on unpermitted take occurring within the LAP area and
carefully evaluate evidence of excessive unpermitted take during the permitting process.
The population size of the LAP is estimated by expanding the density estimates for EMUs to the LAP area.
We acknowledge this approach is simplistic for at least two reasons: (1) given the eagle density estimates
come from nesting or late-summer population surveys, they do not account for seasonal influxes of eagles
that occur through migration and dispersal; (2) this approach assumes eagle density is uniform across the
EMU, which we know is inaccurate. In most cases the first simplification leads to an underestimate of true
density, particularly in core wintering areas during the non-breeding months. As such, this serves as an
added buffer against over-harvest of local nesting eagles. The second assumption of uniform density leads to
greater relative protection of areas with higher than average eagle density within an EMU, and less relative
protection in areas of lower density. Over time, with better information on resource selection and factors
accounting for variation in density (e.g., Tack and Fedy 2015), as well as improved knowledge of seasonal
changes in eagle density and population-specific movement patterns, we can improve the LAP analysis to
more realistically account for the true LAP impact by projects under consideration. For now, however, LAP
take thresholds allow the Service to authorize limited take of eagles while favoring eagle conservation in the
28

Figure 12. Effect on bald eagle local area populations (LAP) of a 5% additive harvest rate. The top figure is for a small
project (6,870 km2 local area) in a low-density EMU (0.008 bald eagles per km2 ), whereas the bottom figure is for a
large project (37,551 km2 local area) in a high-density EMU (0.05 bald eagles per km2 ). The black line and
blue-shaded area shows the trajectory and 95% confidence interval at the 5% LAP take limit, the red line and
pink-shaded area is the trajectory and 95% confidence interval in the absence of added take.

face of uncertainty.
To understand the potential consequence to the LAP of authorizing take up to the levels of the LAP take
thresholds, we conducted a series of simulations using our demographic models to add a 5% take-rate to
background take levels for a hypothetical LAP of both species of eagle (Figure 12, Figure 13). We looked at
hypothetical large and small project footprints in high- and low-density EMUs. For the golden eagle, adding
5% take results in a decline in the LAP and eventually lowers the equilibrium as much as 80%. However, the
LAP did not go to extirpation for the scenarios considered. For the bald eagle, an additive 5% take does not
cause declines in projected LAPs, but reduces the size of the eventual equilibrium LAP by 38% from the
equilibrium in the absence of added take.

The Role of Compensatory Mitigation
Authorized take above EMU take limits has to be offset by compensatory mitigation that will produce
a commensurate decrease in a pre-existing mortality factor, or increase in carrying capacity (U.S. Fish and
Wildlife Service 2009a, 2013). The effect of this mitigation must be that no net increase in mortality occurs
within the EMU where the take is authorized. In the case of golden eagles, our analyses suggest even current
29

Figure 13. Effect on golden eagle local area populations (LAP) of a 5% additive harvest rate (above a background rate
that is, on average, about 10%). The top figure is for a small project (37,299 km2 local area) in a low-density EMU
(0.008 golden eagles per km2), whereas the bottom figure is for a large project (126,950 km2 local area) in a
high-density EMU (0.027 golden eagles per km2 ). Dotted gray lines show the starting and equilibrium population
levels, whereas black lines and blue shading denote the median population size and 95% confidence limits.

levels of take may not be sustainable. Offsetting mitigation for golden eagles at a rate of > 1:1 may be
necessary to be compatible with the Service’s population objective.
The factor that most limits how much golden eagle take the Service can permit is the amount of ongoing
unpermitted take or natural mortality that can reasonably be expected to be offset. Quantifying the real effects
of conservation actions in reducing mortality has proven difficult to date. Electric distribution power line
retrofitting to reduce electrocutions (Avian Power Line Interaction Committee (APLIC) 2006, U.S. Fish and
Wildlife Service 2013) remains the best understood existing mortality source with thorough representation in
the scientific literature, and our analyses suggest that if perfectly effected, about 500 (20𝑡ℎ quantile = 280)
golden eagle deaths could be offset through this approach annually (Table 8). Work to develop other
approaches for implementing and quantifying the performance of other compensatory mitigation mechanisms
is an area of active research, and promising advances are being made (e.g., Cochrane et al. 2015).
The Service currently requires that offsetting mitigation be undertaken in the same EMU where the take
is authorized (U.S. Fish and Wildlife Service 2013). Our analysis suggests this spatial scale is still reasonable,
especially under the Flyway EMU alternatives which take into account the full annual cycle of both eagle
species. However, because a substantial proportion of the mortality of golden eagles originating in Alaska
occurs on migration or during winter in the interior western coterminous U.S. and north-central Mexico
(McIntyre 2012), effective mitigation for take of Alaskan golden eagles could occur in these areas as well.
30

Population Monitoring

As noted previously, the take limits are time-sensitive and require regularly updated estimates of population
size. More generally, the Service has also implemented the eagle take permit process under a formal adaptive
management framework, such that monitoring eagle populations and updating population estimates and take
limits are critical parts of the adaptive management feedback loop (U.S. Fish and Wildlife Service 2013). For
these reasons, the Service is interested in formalizing its eagle population monitoring commitments as part of
the PEIS process.
The existing golden eagle assessment approach (using a hierarchical model to combine density estimates
from the summer aerial-transect survey with BBS indices [Appendix A4]) provides reasonable information
on golden eagle population size and trend at the coarse scale considered under the national permit program.
The winter golden eagle survey also provides useful information on the number and distribution of golden
eagles in the core of the species’ range in winter. This information about wintering eagles is essential for
more accurate accounting of the effects of take at different locations on different natal populations of golden
eagles; pairing summer and winter surveys maximizes the opportunity to quantify wintering golden eagle
populations that move into the coterminous U.S. from northern latitudes (see Golden Eagle: Population Size).
As part of future bald eagle nesting territory survey efforts, the Service will investigate the potential for
combining the dual-frame survey estimates of occupied nesting territories with BBS indices to better link
the dual-frame results to changes in total population size (expanding beyond the current focus on breeding
numbers); additionally, this capitalizes on the rigorous and standardized data set from the BBS. Because our
conversion of the dual-frame survey results to total population size estimates depends on accurate EMU-scale
estimates of productivity, the Service will investigate adjustments to the dual-frame survey design that will
provide information on nest success and brood sizes in a sample of occupied nesting territories.
There are several other areas of active inquiry that should improve the Service’s ability to effectively
manage eagles in the future. In particular, resource utilization functions have the potential to vastly improve
the accuracy of LAP analyses. Additionally, surveys and studies to locate, map, and prioritize nesting
territories are important in that they can serve to identify and direct projects away from important high-density
areas or high-performing nesting territories. Service biologists will continue to look for ways to implement
these surveys as efficiently and effectively as possible, including periodic reassessments of statistical power
and reliability, and integrating other sources of information (e.g., Christmas Bird Counts) with ongoing
surveys to improve power, representativeness, and to expand the scale of inference.

31

Literature Cited

Avian Power Line Interaction Committee (APLIC). 2006. Suggested Practices for Avian Protection on
Power Lines: The State of the Art in 2006. Edison Electric Institute, APLIC, and the California Energy
Commission., Washington, DC, and Sacramento, California, USA. URL http://www.dodpif.org/
downloads/APLIC_2006_SuggestedPractices.pdf.
Baldwin, J. D., J. W. Bosley, L. Oberhofer, O. L. Bass, and B. K. Mealey. 2012. Long-term changes,
1958–2010, in the reproduction of bald eagles of Florida Bay, southern coastal Everglades. Journal of
Raptor Research 46:336–348.
Bienvenu, F., and S. Legendre. 2015. A new approach to the generation time in matrix population models.
The American Naturalist 185:834–843.
Bowman, T. D., P. F. Schempf, and J. A. Bernatowicz. 1995. Bald eagle survival and population dynamics in
Alaska after the “Exxon Valdez" oil spill. Journal of Wildlife Management 59:317–324.
Bretagnolle, V., F. Mougeot, and J. C. Thibault. 2008. Density dependence in a recovering osprey population:
demographic and behavioural processes. Journal of Animal Ecology 77:998–1007.
Buehler, D. A. 2000. Bald Eagle. The Birds of North America Online (A. Poole and F.Gill, Eds.), Ithaca:
Cornell Laboratory of Ornithology; Retrieved from The Birds of North America Online database. URL
http://bna.birds.cornell.edu/bna/species/506/articles/breeding.
Burnham, K. P., and D. R. Anderson. 2002. Model Selection and Multimodel Inference: A Practical
Information-Theoretic Approach. Springer, New York.
Carrete, M., J. A. Donázar, and A. Margalida. 2006. Density-dependent productivity depression in Pyrenean
bearded vultures: implications for conservation. Ecological Applications 16:1674–1682.
Caswell, H. 2001. Matrix Population Models: Construction, Analysis, and Interpretation. Sinauer Associates,
Sunderland, Massachusetts, USA.
Chevallier, C., A. Hernández-Matías, J. Real, N. Vincent-Martin, A. Ravayrol, and A. Besnard. 2015.
Retrofitting of power lines effectively reduces mortality by electrocution in large birds: an example with
the endangered Bonelli’s eagle. Journal of Applied Ecology 52:1465–1473. URL http://dx.doi.
org/10.1111/1365-2664.12476.
Cochrane, J. F., E. V. Lonsdorf, T. D. Allison, and C. A. Sanders-Reed. 2015. Modeling with uncertain science:
estimating mitigation credits from abating lead poisoning in golden eagles. Ecological Applications
25:1518–1513.
Cooch, E., and G. C. White. 2014. Program MARK. Colorado State University, Fort Collins, Colorado USA,
9𝑡ℎ edition. URL http://www.cnr.colostate.edu/~gwhite/mark/mark.htm.
Dennhardt, A. J., A. E. Duerr, D. Brandes, and T. E. Katzner. 2015. Integrating citizen-science data with
movement models to estimate the size of a migratory golden eagle population. Biological Conservation
184:68–78.

32

Dillingham, P., and D. Fletcher. 2008. Estimating the ability of birds to sustain additional human-caused
mortalities using a simple decision rule and allometric relationships. Biological Conservation 141:1783–
1792.
Fasce, P., L. Fasce, A. Villers, F. Bergese, and V. Bretagnolle. 2011. Long-term breeding demography and
density dependence in an increasing population of golden eagles Aquila chrysaetos. Ibis 153:581–591.
Ferrer, M., and J. A. Donazar. 1996. Density-dependent fecundity by habitat heterogeneity in an increasing
population of Spanish imperial eagles. Ecology 77:69–74.
Hunt, W. G. 1998. Raptor floaters at Moffat’s equilibrium. Oikos 82:191–197.
Kauffman, M. J., J. F. Pollock, and B. Walton. 2004. Spatial structure, dispersal, and management of a
recovering raptor population. The American Naturalist 164:582–597.
Kenward, R. E., V. Marcström, , and M. Karlbom, 1993. Causes of death in radio-tagged northern goshawks.
Chapter 13, pages 57–61 in P. Redig, J. E. Cooper, J. D. Remple, and D. B. Hunter, editors. Raptor
Biomedicine. University of Minnesota Press, Minneapolis, Minnesota.
Kochert, M. N., K. Steenhof, C. L. McIntyre, and E. H. Craig. 2002. Golden Eagle. The Birds of North
America Online (A. Poole, Ed.), Ithaca: Cornell Laboratory of Ornithology; Retrieved from The Birds
of North America Online database. URL http://bna.birds.cornell.edu/bna/species/684/
articles/introduction.
Lande, R., S. Engen, and B. E. Saether. 2003. Stochastic Population Dynamics in Ecology and Conservation.
Oxford University Press, New York, New York, USA.
Lien, L. A., B. A. Millsap, K. Madden, and G. W. Roemer. 2015. Male brood provisioning rates provide
evidence for inter-age competition for mates in female Cooper’s Hawks Accipiter cooperii. Ibis 157:860–
870.
McIntyre, C. 2012. Quantifying sources of mortality and winter ranges of golden eagles from interior Alaska
using banding and satellite tracking. Journal of Raptor Research 46:129–134.
McIntyre, C. L., D. C. Douglas, and M. W. Collopy. 2008. Movements of golden eagles (Aquila chrysaetos)
from interior Alaska during their first year of independence. Auk 125:214–224.
Millsap, B. A., and G. T. Allen. 2006. Effects of falconry harvest on wild raptor populations in the
United States: theoretical considerations and management recommendations. Wildlife Society Bulletin
34:1392–1400.
Millsap, B. A., T. G. Grubb, R. K. Murphy, T. Swem, and J. W. Watson. 2015. Conservation significance of
alternative nests of golden eagles. Global Ecology and Conservation 3:234–241.
Millsap, B. A., A. R. Harmata, D. W. Stahlecker, and D. G. Mikesic. 2014. Natal dispersal distance of bald
and golden eagles originating in the coterminous United States as inferred from band encounters. Journal
of Raptor Research 48:13–23.
Millsap, B. A., G. S. Zimmerman, J. R. Sauer, R. M. Nielson, M. C. Otto, E. Bjerre, and R. Murphy.
2013. Golden eagle population trends in the western United States: 1968–2010. Journal of Wildlife
Management 77:1436–1448.
Mojica, E. K., J. M. Meyers, B. A. Millsap, and K. L. Haley. 2008. Migration of Florida sub-adult bald
eagles. Wilson Journal of Ornithology 120:304–310.
33

Morneau, F., J. A. Tremblay, C. Todd, T. E. Chubbs, C. Maisonneuve, J. Lemaître, and T. Katzner. 2015.
Known breeding distribution and abundance of golden eagles in eastern North America. Northeastern
Naturalist 22:236–247.
Nielson, R. M., G. DiDonato, L. McManus, and L. L. McDonald.
2015.
A survey of
golden eagles (Aquila chrysaetos) in the western U.S.: mid-winter 2015. Western Ecosys­
tems and Technology, Inc., Cheyenne, Wyoming, USA. URL http://www.west-inc.com/
wildlife-life-monitoring-and-research/.
Nielson, R. M., and L. McManus. 2014. A survey of golden eagles (Aquila chrysaetos) in the western
U.S.: mid-winter 2014. Western Ecosystems and Technology, Inc., Cheyenne, Wyoming, USA. URL
http://www.west-inc.com/wildlife-life-monitoring-and-research/.
Nielson, R. M., L. Mcmanus, T. Rintz, L. L. McDonald, R. K. Murphy, W. H. How, and R. E. Good. 2014.
Monitoring abundance of golden eagles in the western United States: golden eagle abundance in the
western U.S. Journal of Wildlife Management 78:721–730.
Runge, M. C., J. R. Sauer, M. L. Avery, B. F. Blackwell, and M. D. Koneff. 2009. Assessing allowable take
of migratory birds. Journal of Wildlife Management 73:556–565.
Sauer, J. R., J. E. Hines, J. E. Fallon, K. L. Pardieck, D. J. Ziolkowski, and W. A. Link. 2014. The North
American breeding bird survey, results and analysis 1966–2012, version 02.19.2014. USGS Patuxent
Wildlife Research Center, Laurel, MD. URL http://www.mbr-pwrc.usgs.gov/bbs/bbs.html.
Slade, N. A., R. Gomulkiewicz, and H. M. Alexander. 1998. Alternatives to Robinson and Redford’s method
of assessing overharvest from incomplete demographic data. Conservation Biology 12:148–155.
Tack, J. D., and B. C. Fedy. 2015. Landscapes for energy and wildlife: conservation prioritization for golden
eagles across large spatial scales. PLoS One 10:e0134781.
Turrin, C., 2014. Rise of a floater class: Behavioral adjustments by breeding bald eagles in a population
approaching saturation. Master’s thesis, College of William and Mary.
U.S. Fish and Wildlife Service. 2009a. 50 CFR Parts 13 and 22. Eagle permits; take necessary to protect
interests in particular localities. Federal Register 74(175):46836–46879. URL http://www.fws.gov/
policy/library/2009/E9-21589.pdf.
U.S. Fish and Wildlife Service. 2009b. Final environmental assessment. Proposal to permit take provided un­
der the Bald and Golden Eagle Protection Act. Division of Migratory Bird Management, Washington, DC,
USA. URL http://www.fws.gov/alaska/eaglepermit/pdf/environmental_assessment.
pdf.
U.S. Fish and Wildlife Service. 2013. Eagle conservation plan guidance. Module 1–land-based wind energy.
Version 2. Division of Migratory Bird Management, Washington, DC, USA. URL http://www.fws.
gov/migratorybirds/pdf/management/eagleconservationplanguidance.pdf.
U.S. Fish and Wildlife Service. 2014. 50 CFR Part 20. Migratory bird hunting; final frameworks for
late-season migratory bird hunting regulations. Federal Register 79(184):56864–56890.
U.S. Fish and Wildlife Service. 2015. USFWS administrative waterfowl flyway boundaries. Vector digital
data. Falls Church, Virginia, USA. URL http://ecos.fws.gov/ServCatFiles/reference/
holding/42216/WaterfowlFlyways.zip.
34

Watts, B. D. 2015. Estimating the residual value of alternate bald eagle nests: implications for nest protection
standards. Journal of Wildlife Management 79:776–784.
Whitfield, D. P., A. H. Fielding, M. J. Gregory, A. G. Gordon, D. R. A. Mcleod, and P. F. Haworth. 2007.
Complex effects of habitat loss on golden eagles Aquila chrysaetos. Ibis 149:26–36.
Whitfield, D. P., A. H. Fielding, D. R. A. Mcleod, and P. F. Haworth. 2004a. The effects of persecution on age
of breeding and territory occupation in golden eagles in Scotland. Biological Conservation 118:249–259.
Whitfield, D. P., A. H. Fielding, D. R. A. Mcleod, and P. F. Haworth. 2004b. Modelling the effects
of persecution on the population dynamics of golden eagles in Scotland. Biological Conservation
119:319–333.
Williams, B. K., J. D. Nichols, and M. J. Conroy. 2002. Analysis and Management of Animal Populations:
modeling, estimation, and decision making. Academic Press, San Diego, California, USA.

35

Appendix A1. Review of Eagle Productivity Data

Mark Brennan and Brian Millsap
U.S. Fish and Wildlife Service

Division of Migratory Bird Management

Literature Review and Data Compilation
In March 2015, the Service conducted a comprehensive review of available peer-reviewed publications
that included data on eagle productivity. The data were evaluated, filtered, and, if appropriate, used to update
and improve estimates of eagle productivity–including variability and the range of observed productivity
across the United States. The Service used the improved productivity estimates to update demographic
population models and evaluate the effect of mortality rates under various management scenarios.
We conducted a thorough literature search through a variety of available databases and search tools
(Google Scholar, Web of Science, BioOne, Biosis, and ProQuest) using various combinations of the following
key words: eagle, bald eagle, golden eagle, productivity, nest productivity, nest success, nest monitoring,
population analysis, and population status. We also searched prominent authors of eagle biology and ecology.
We filtered the literature search results to focus on publications containing productivity data from 1995 to
2015. Given potential long-term fluctuations in productivity (e.g., due to changes in conservation practices,
land use, or environmental factors), focusing on data from the last two decades helped ensure a contemporary
estimate of eagle productivity. We further limited results to bald and golden eagle productivity observed in
the conterminous United States and Alaska. We also initially selected references based on titles or abstracts,
and later excluded publications that did not contain data sufficient to determine nest or territory occupancy
and young fledged. We discovered 98 productivity references for bald eagles and 70 for golden eagles that
met our criteria.
There was considerable variability in the reporting of nest or territory activity and inconsistency in the
use of the terms “active” and “occupied” by different authors in describing breeding status. We specifically
looked for publications that included the number of young fledged from occupied nesting territories based on
criteria for occupancy initially described by Postupalsky (1974) and later by Steenhof and Newton (2007).
Limiting our results to papers that specifically allowed determination of territorial occupancy for determining
productivity further reduced the number of studies in our summary.
We defined productivity as the total young fledged per occupied nesting territory (Steenhof and Newton
2007). However, we found inconsistencies between the papers we reviewed in the way productivity was
calculated; in some cases it was possible to base productivity estimates on the number of occupied nesting
territories sampled, whereas in other cases productivity was clearly weighted towards successful nesting
territories. When necessary we back-calculated productivity using the total occupied nesting territories and
total number of young counted. We also excluded studies that had manipulative components (e.g., egg
removal, experimental disturbance) since the manipulations could affect productivity.
Though we preferred to include studies that reported annual estimates of productivity (or the data that
could be used to compute them), some studies only reported aggregated data or estimates for multiple years
within a specific area or a subset of known nesting territories. We excluded any redundant data identified in
the course of the review.
We found a greater number of publications reporting productivity for bald eagles than golden eagles, as
well as uneven coverage of the known range of both species. We subsequently used State or other agency
36

monitoring reports for breeding eagles where the data available would provide more representative spatial
coverage of the known breeding range. Most peer-reviewed studies focused on local sub-populations or
specific nesting territories monitored over time.
The relatively small number of papers—18 for bald eagles and 12 for golden eagles—included in our
final compilation of productivity data from the literature and subsequent meta-analysis (see Appendix A2)
reflects our stringent data quality standards. Many of these final papers contained data spanning multiple
years and included a large number of nesting territories. We believe these data are representative of available
data on eagle productivity across the known bald and gold eagle breeding ranges in the U.S.

Literature Cited
Postupalsky, S. 1974. Raptor reproductive success: some problems with methods, criteria, and terminology.
Pages 21–31 in F. N. Hamerstrom Jr., B. E. Harrell, and R. R. Olendorff, editors. Raptor Research Report
No. 2. Management of raptors. Vermillion, South Dakota, USA.
Steenhof, K., and I. Newton. 2007. Assessing raptor reproductive success and productivity. Pages 184–192
in K. L. Bildstein and D. M. Bird, editors. Raptor research and management techniques. Hancock House,
Surrey, B.C.

Bibliography: Bald Eagle Productivity Data
Allison, L. J., J. T. Driscoll, and K. V. Jacobson. 2008. Demographic analysis of the bald eagle in Arizona.
Nongame and Endangered Wildlife Program Technical Report 221. Arizona Game and Fish Department,
Phoenix, Arizona, USA.
Badzinski, D., and I. Richards. 2002. Southern Ontario bald eagle monitoring project, 2002 report. Report by
Bird Studies Canada to the Ontario Ministry of Natural Resources and the Canadian Wildlife Service,
Ontario, Canada.
Bowerman, W. W., D. A. Best, T. G. Grubb, G. M. Zimmerman, and J. P. Giesy. 1998. Trends of contaminants
and effects in bald eagles of the Great Lakes basin. Environmental Monitoring and Assessment 53:197–
212.
Buck, J. A., R. G. Anthony, C A. Schuler, F. B. Isaacs, and D. E. Tillitt. 2005. Changes in productivity and
contaminants in bald eagles nesting along the lower Columbia River, USA. Environmental Toxicology
and Chemistry 24:1779–1792.
Jenkins, M. A., and S. K. Sherrod. 2005. Growth and recovery of the bald eagle population in Oklahoma.
Wildlife Society Bulletin 33:810–813.
Millsap, B., T. Breen, E. McConnell, T. Steffer, L. Phillips, N. Douglass, and S. Taylor. 2004. Comparative
fecundity and survival of bald eagles fledged from suburban and rural natal areas in Florida. Journal of
Wildlife Management 68:1018–1031.
New Jersey Department of Environmental Protection. 1997. New Jersey bald eagle management project,
1997. New Jersey Department of Environmental Protection, Department of Fish, Game and Wildlife,
Trenton New Jersey, USA.
New Jersey Department of Environmental Protection. 2000. New Jersey bald eagle management project,
2000. New Jersey Department of Environmental Protection, Department of Fish and Wildlife, Trenton,
New Jersey, USA.

37

New Jersey Department of Environmental Protection. 2005. New Jersey bald eagle management project,
2005. New Jersey Department of Environmental Protection, Department of Fish and Wildlife, Trenton,
New Jersey, USA.
New Jersey Department of Environmental Protection. 2007. New Jersey bald eagle project, 2007. New Jersey
Department of Environmental Protection, Department of Fish and Wildlife, Trenton, New Jersey, USA.
New Jersey Department of Environmental Protection. 2013. New Jersey bald eagle project, 2013. New Jersey
Department of Environmental Protection, Department of Fish and Wildlife, Trenton, New Jersey, USA.
Nye, P. 2010. New York State bald eagle report. New York State Department of Environmental Conservation,
Albany, New York, USA.
Route, B., and R. Key. 2009. Contaminants in bald eagle nestlings at Apostle Islands National Lakeshore,
Mississippi National River and Recreation Area, and the St. Croix National Scenic Riverway: Data
summary 2006–2008. Natural Resource Data Series NPS/GLKN/NRDS–2009/001. National Park
Service, Fort Collins, Colorado, USA.
Stinson, D. W., J. W. Watson, and K. R. McAllister. 2007. Washington State status report for the bald eagle.
Washington Department of Fish and Wildlife, Olympia, WA, USA.
Todd, C. S. 2004. Bald eagle assessment. Maine Department of Inland Fisheries and Wildlife, Bangor, Maine,
USA.
Watkins, M. A., and D. W. Mulhern. 1999. Ten years of successful bald eagle nesting in Kansas. Kansas
Ornithological Society Bulletin 50:29–33.
Watts, B. D., G. D. Therres, and M. A. Byrd. 2008. Recovery of the Chesapeake Bay bald eagle nesting
population. Journal of Wildlife Management 72:152–158.
Zwiefelhofer, D. 2007. Comparison of bald eagle (Haliaeetus leucocephalus) nesting and productivity at
Kodiak National Wildlife Refuge, Alaska, 1963–2002. Journal of Raptor Research 41:1–9.

Bibliography: Golden Eagle Productivity Data
Craighead Berengia South. 2014. Golden eagle breeding ecology in South Central Montana 2014 summary.
Unpublished report, Craighead Berengia South, Kelly, Wyoming, USA.
Hawks Aloft, Incorporated. 2002. Nesting and productivity of golden eagles in Northwestern and WestCentral, New Mexico. Report submitted to the Bureau of Land Management, Socorro and Farmington
Field Offices, New Mexico, USA.
Hawks Aloft, Incorporated. 2006. Monitoring nesting golden eagles for the Farmington Field Office, and
nesting raptors for the Taos Field Office, BLM Resource Areas, New Mexico, USA.
Hawkwatch International, Incorporated. 2009a. Nesting ecology of raptors in northwest Utah:1998–2007. A
report prepared for Utah Department of Natural Resources, Division of Wildlife Resources and USDI
Bureau of Land Management, Salt Lake Field Office, Salt Lake City, Utah, USA.
Hawkwatch International, Incorporated. 2009b. Northwest Nevada raptor nest survey 2008. A report prepared
for USDI Bureau of Land Management, Elko District Office, Elko, Nevada, USA.
Isaacs, F. B. 2011. Golden eagles (Aquila chrysaetos) nesting in Oregon, 2011. 1𝑠𝑡 annual report, revised, 16
April 2012. Oregon Eagle Foundation, Inc., Klamath Falls, Oregon, USA.
McIntyre, C. L. 2002. Patterns in nesting area occupancy and reproductive success of golden eagles (Aquila
chrysaetos) in Denali National Park and Preserve, Alaska, 1988–99. Journal of Raptor Research 36:50–54.

38

McIntyre, C. L., and L. G. Adams. 1999. Reproductive characteristics of migratory golden eagles in Denali
National Park, Alaska. Condor 101:115–123.
McIntyre, C. L., and J. H. Schmidt. 2012. Ecological and environmental correlates of territory occupancy
and breeding performance of migratory golden eagles (Aquila chrysaetos) in interior Alaska. Ibis
154:124–135.
Morneau, F., B. Gagnon, S. Poliquin, P. Lamothe, N. D’Astous, and J. A. Tremblay. 2012. Breeding status
and population trends of golden eagles in northeastern Québec, Canada. Avian Conservation and Ecology
7(2):4.
Preston, C. R. 2014. Golden eagle nesting ecology in the Bighorn Basin. 2014 progress report. Draper
Natural History Museum, Cody, Wyoming, USA.
Ritche, R. J., A. M. Wildman, and D. A. Yokel. 2003. Aerial surveys of cliff-nesting raptors in the National
Petroleum Reserve–Alaska, 1999, with comparisons to 1977. Technical Note 413. Bureau of Land
Management, Denver, Colorado, USA.
U.S. Fish and Wildlife Service. 2013. Final environmental assessment. Issuance of a Bald and Golden Eagle
Protection Act permit to take golden eagles by the Hopi Tribe for Native American religious purposes.
Washington, D.C., USA.

39

Appendix A2. Bald and Golden Eagle

Productivity Meta-Analysis

Mark Otto
U.S. Fish and Wildlife Service

Division of Migratory Bird Management

Abstract
As part of a larger effort to update and improve productivity parameters used in eagle population
modeling efforts, Brennan and Millsap (Appendix A1) compiled a dataset of contemporary productivity
information for bald and golden eagles, Haliaeetus leucocephalus and Aguila chrysaetos respectively, across
the U.S. from 1995–2014. With these data, I used a random effects meta-analysis model and estimated the
predictive distributions for bald eagle and golden eagle productivity. Bald eagle productivity differed by
region with lower productivity in the Southwest (mean = 0.77, SE = 0.249) than in the rest of the continental
U.S. (mean = 1.15, SE = 0.252), whereas golden eagle productivity did not differ by region (mean = 0.55,
SE = 0.087). Apart from the fixed stratum differences for bald eagles, the best-supported models included
standard errors for the random effects for study, area (bald eagles only), year given study, and overdispersion;
the extent to which the random effect credible intervals overlapped zero varied by species.

Introduction
Meta-analyses combine results of different studies of the same subject in order provide stronger and more
robust inferences (Borenstein et al. 2010). Meta-analyses are often used in the medical field to combine
clinical trial data. Just as medical trials may have different methodologies, target populations, and sampling
designs and selections, so do wildlife demographic studies (Johnson 2002). There are two main types of
meta-analyses: fixed and random effect. Fixed effect models are used when the studies are thought to be
functionally equivalent, whereas the random effect model assumes that they have common characteristics but
are not the same (Borenstein et al. 2010). Thus a fixed effect model assumes that a single value is common
to all studies, in contrast to a random effects model which assumes that the values belong to a common
distribution (Higgins et al. 2009).
Summarizing a range of studies over different areas and time spans, accounting for the study, area,
and annual components of variation also complicates the analysis with decisions on how to separate and
characterize the different forms of variation. Rather than estimate the common value (in our case productivity),
Higgins et al. (2009) recommend using predictive distributions. They also recommend for a small number
of studies, using the t-distribution with k minus 4 degrees of freedom (k is the number of studies), instead
of the normal distribution. Due to the combined complexities of deciding on the proper prediction variance
and other model choices, I used the classic normal distribution for this analysis. The approach is similar
to what is presented in New et al. (2015) for the prior parameters for the eagle example where the authors
created a mixture distribution from the small number of projects available and estimated parameters for a
common distribution from the mixture. My methods here similarly yield a common predictive distribution for
productivity from the projects available.

40

Methods
Data
Brennan and Millsap (Appendix A1) searched the published literature for bald and golden eagle
productivity data and compiled datasets for each species from studies within the U.S. from 1995–2014.
I categorized the target populations for the included studies in terms of area and time span and accounted
for separate values for multi-area and multi-year data. I used sample size (the number of nesting territories
or nests), number of fledglings, productivity values, and standard errors reported in the studies. When not
reported, I back-calculated sample size from number of fledglings and productivity. In one case where only
the productivity value was reported, the sample size became the inverse of the productivity value—resulting
in one fledgling in the study and the smallest weight possible given to that study.
There were 18 studies included in the bald eagle analysis: one multi-area study, nine multi-year studies,
and two multi-area and multi-year studies. In cases where studies included multi-area or multi-year data,
I used random effects for area or year nested within study. The data did not support interactions between
area and year in the 2 multi-area and multi-year studies. There were 12 studies included in the golden eagle
analysis: nine multi-year studies but no multi-area and multi-year studies. This limited the golden eagle
analysis to only considering study-to-study and year-to-year variation.
Model
The productivity random effects model is a Poisson log-normal hierarchical model (although a gamma
distribution could replace the log normal). The data are the number of successful fledglings in each study
(with values separated by areas and years in multi-strata studies). The log sample sizes, 𝑆𝑖𝑗𝑘𝑙 number of
nesting territories, are treated as offsets but are shown here on the original scale,
𝐹𝑖𝑗𝑘𝑙 ∼ 𝑃 𝑜𝑖𝑠𝑠𝑜𝑛 (𝑅𝑖𝑗𝑘𝑙 𝑆𝑖𝑗𝑘𝑙 ) .
𝐹𝑖𝑗𝑘𝑙 is the number of fledglings in the 𝑘 𝑡ℎ area and 𝑙𝑡ℎ year of the 𝑗 𝑡ℎ study in the 𝑖𝑡ℎ region. Not all
subscripts are necessary if it is not multi-area and multi-year study. 𝑅𝑖𝑗𝑘𝑙 is the estimated random effect
productivity estimate, and 𝑆𝑖𝑗𝑘𝑙 is the sample size in number of occupied nesting territories. Since the model
conditions on occupied nesting territories, we only make the basic assumption that the likelihood occupied
nesting territories were observed was not linked to the productivity rate. If the chances of detecting an
occupied nesting territory early, even if it later fails, are good then the potential for such detection bias should
be low. Log productivity is affected by the region, the study within that region, and if applicable a year within
a given study.
(︀
)︀
2
.
log(𝑅𝑖𝑗𝑘𝑙 ) = N 𝑟𝑖 + 𝜓𝑗|𝑖 + 𝛼𝑘|𝑖𝑗 + 𝜏𝑙|𝑖𝑗 , 𝜎Overdispersion
Study, 𝜓𝑗|𝑖 ; area, 𝛼𝑘|𝑖𝑗 ; and year, 𝜏𝑙|𝑖𝑗 , are nested random effects, with study nested within region, and area
2
and year nested within study; there were no multi-region studies. The overdispersion variance is 𝜎Overdispersion
.
The random effects use an Ottomert transformation that converts 𝑛1 random variables into n centered
variables with the same standard deviation and the same correlations among all the effects. The transformation
corrects for the under-estimation of the standard deviation caused by generating and centering n random
variables. The area in multi-area random effects and year in multi-year random effects are nested within
study, so their effects are centered within each study.
)︀
(︀
2
𝜓𝑗|𝑖 = 𝑂𝑡𝑡𝑜𝑚𝑒𝑟𝑡 𝑆𝑡𝑢𝑑𝑦𝑗 |𝑅𝑒𝑔𝑖𝑜𝑛𝑖 )𝑁 (0, 𝜎𝑆𝑡𝑢𝑑𝑦
)︀
(︀
2
𝛼𝑘|𝑖𝑗 = 𝑂𝑡𝑡𝑜𝑚𝑒𝑟𝑡 𝐴𝑟𝑒𝑎𝑘 |𝑆𝑡𝑢𝑑𝑦𝑗 )𝑁 (0, 𝜎𝐴𝑟𝑒𝑎
(︀
)︀
𝜏𝑙|𝑖𝑗 = 𝑂𝑡𝑡𝑜𝑚𝑒𝑟𝑡 𝑌 𝑒𝑎𝑟𝑙 |𝑆𝑡𝑢𝑑𝑦𝑗 )𝑁 (0, 𝜎𝑌2 𝑒𝑎𝑟
41

Because studies are nested within region, the area and year effects are also nested within region. I assumed the
study, area, and year variation were the same across regions. For bald eagles, there was only one multi-year
study representing the Southwest and only one study representing the East for golden eagles.
I ran glm and glmer models in R (R Core Team 2014) to discriminate among models using AIC and
then estimated the best-supported models using Stan (Stan Development Team 2015), which is equivalent
to a Bayesian estimation with non-informative priors. I included overdispersion by adding a random effect
where the effect was different for every observation and tested models for an overall mean only and an overall
mean with overdispersion. I also calculated simple estimates of productivity by aggregating the fledged and
occupied territory counts for each area and year of each study by region then taking the ratio.

Results
Overdispersion gave a vast reduction in AIC (ΔAIC) for both the bald eagle and golden eagle models
(118.49 and 35.37, respectively). For bald eagles, the best-supported model was a random effects model with
overdispersion that included a fixed effect for region (separating the Southwest from the rest of the U.S.;
Table A2-1). The Southwest had lower overall productivity (Table A2-2) than the rest of the U.S., but there
was wide overlap between the predictive distributions (Figure A2-1). Both prediction distributions are right
skewed and leptokurtic, therefore the best way to use the productivity information as part of a demographic
model is to sample the posterior simulations.
All of the random effects (study, area, year, and overdispersion) were important to the model (25.68);
the estimates of the standard errors for the random effects are in Table A2-3. The random effects from the
final model were more spread out than the simple estimates for both regional distributions (Figure A2-1).
Normally we would expect random effects estimates to shrink or be less spread out, but the differences are
small and likely due to separating the study, area, and year random effects. The study, area, and year effects
are all significant and the credible intervals do not overlap zero. The total random effect variance is the sum
of the variances of all the random effects (Table A2-2).
Figure A2-2 shows the percent change due to the study random effects apart from regional differences
and the percent change due to the year given study. The productivity estimates by study from the random
effects model include the region effects and the study, area, and year random effects (Table A2-4); they vary
from 0.48 to 0.57. The model random effect estimates by study, area, and year are presented in Table A2-5.
For golden eagles, the best-supported model was the random effects model with overdispersion (Table
A2-1). The study and year random effects were important to the model (4.02), but there were no multi-area
studies so I did not include area random effects. I explored models with a regional effect (e.g., Eastern U.S.,
Western U.S., Alaska) but there was no support for including any regional differences (3.56 for 4 degrees of
freedom; Table A2-6) so the final model estimated an overall productivity for the entire U.S. including Alaska
(Table A2-2). The overall prediction estimate along with the 95% prediction intervals is shown in Figure
A2-4. The estimates from the final model are a bit lower than simple estimates taken by aggregating the
fledged and occupied territory counts then taking the ratio (Figure A2-4). Explaining this will require further
exploration. The distribution is right skewed, skewness = 2.09, and is highly leptokurtic, kurtosis = 22.59,
therefore sampling the posterior simulations is the best way to use the productivity estimates in other models,
since they do not fit a common distribution.
The estimates of the standard errors for the random effects are in Table A2-2. The study and year
random effects had low variation (medians 0.1 and 0.29, respectively), and all random effect credible intervals
overlap zero (Figure A2-5). The non-significance of the study and year effects and the significance of the
overdispersion reinforce the AIC differences observed in the model comparisons. The productivity estimates
from the random effects model by study include the region effects and the study random effects; they vary
from 0.48 to 0.57 (Table A2-5). The random effect model estimates which include the study, area, and

42

year random effects for each study and year combination are included in Table A2-6 along with the simple
productivity ratios.

Discussion
I conducted this modeling effort with the specific goal of rapidly producing a usable predictive distribution
for productivity that could be used in subsequent population modeling efforts. Though the approach was
logical, there were a number of decisions that could be explored further. I only included the study random
effect variances in the predictions. Though this is consistent with most meta-analysis models, it is unusual
to have the additional complexities of multi-area and multi-year studies. An alternative approach may be
to include both random effects and overdispersion in the prediction variation along with the additional
consideration of using a t-distribution instead of a normal distribution. However all of this would make the
already large prediction intervals larger, possibly to the point of no longer being useful. The current approach
used to estimate the predictive distribution is consistent with other meta-analysis models and sampling the
posterior simulations will provide reasonable productivity estimates given the data available.

Literature Cited
Borenstein, M., J. P. T. H. L. V. Hedges, and H. R. Rothstein. 2010. A basic introduction to fixed-effect and
random-effects models for meta-analysis. Research Synthesis Methods 1:97–111.
Higgins, J. P. T., S. G. Thompson, and D. J. Spiegelhalter. 2009. A re-evaluation of random-effects
meta-analysis. Journal of the Royal Statistical Society 172:137–159.
Johnson, D. H. 2002. The importance of replication in wildlife research. Journal of Wildlife Management
66:919–932.
New, L., E. Bjerre, B. Millsap, M. C. Otto, and M. C. Runge. 2015. A collision risk model to predict
avian fatalities at wind facilities: an example using golden eagles, Aquila chrysaetos. PLoS One
10(7):e0130978. URL http://dx.doi.org/10.1371/journal.pone.0130978.
R Core Team. 2014. R: A Language and Environment for Statistical Computing. R Foundation for Statistical
Computing, Vienna, Austria. URL http://www.R-project.org/.
Stan Development Team. 2015. Stan Modeling Language Users Guide and Reference Manual, Version 2.8.0.
URL http://mc-stan.org/.

43

Figure A2-1. Bald eagle productivity for the Southwest U.S.(left) and the U.S. excluding the Southwest (right). The blue
curve is the empirical density distribution of the estimates—which are shown via the rug just above the x-axis. The
vertical blue line is the median with the area within the 95% credible intervals shaded blue. The red and green curves
represent the log normal and normal distributions (respectively) defined by the estimated means and standard deviations.

Table A2-1. AIC values for bald and golden eagles from glm and glmer models which included overdispersion, study,
area, year, and region effects with a mean. Region included Alaska, the Southwest (SW), the conterminous
U.S. excluding the Southwest (Lower 48) and the entire continental U.S. (Overall) for bald eagles. Region included
Eastern U.S. (East), Western U.S. (West), Alaska (AK) and Overall for golden eagles.

Species

Overdispersion

Bald Eagle
x
x
x
x
x
Golden Eagle
x
x
x
x
x

Fixed-Effects

Random-Effects Difference-DoF

AIC

Overall
Overall
Alaska+Lower48+SW
Alaska+Lower48+SW
Overall
Lower48+SW
Alaska+Lower48+SW

None
None
None
All
All
All
All

1
2
4
6
4
6
7

1,449.50
924.08
907.00
999.81
883.57
881.75
881.32

Overall
Overall
Alaska+East+West
Alaska+East+West
Overall
East+West
Alaska+East+West

None
None
None
All
All
All
All

1
2
4
5
4
5
6

688.50
506.56
510.01
541.47
502.53
504.20
506.09

44

Table A2-2. Regional prediction means, standard errors (SE), medians,
and lower and upper limits (LCL, UCL) of the 95% credible intervals
from the random effects models for bald and golden eagle productivity.
The bald eagle model included a fixed effect for region and estimated
productivity for the U.S. excluding the Southwest (U.S.–SW) and the
Southwest (SW). The golden eagle model is an overall random effects
model.

Species
Bald eagle

Golden eagle

Region

Mean

SE

Median (LCL–UCL)

U.S.–SW 1.15 0.252
SW
0.77 0.249
Overall

0.55 0.087

1.12
0.73

(0.73–1.72)
(0.40–1.36)

0.54

(0.40–0.75)

Table A2-3. Productivity model random effect standard
errors and lower and upper limits (LCL, UCL) of the 95%
credible intervals for a) bald eagles and b) golden eagles.
The total standard error is the square root of the sum of all
the random effect variances.
(a) Bald Eagle

Random
Effect SE

Mean

SE

Study
Area
Year
Overdispersion
Total

0.21
0.13
0.14
0.02
0.26

0.047
0.056
0.020
0.016
0.041

Median (LCL–UCL)
0.20
0.12
0.14
0.02
0.25

(0.14–0.32)
(0.05–0.26)
(0.11–0.18)
(0.01–0.06)
(0.19–0.35)

(b) Golden Eagle

Random
Effect SE

Mean

SE

Study
Year
Overdispersion
Total

0.11
0.27
0.31
0.47

0.079
0.132
0.12
0.059

45

Median (LCL–UCL)
0.10
0.29
0.32
0.46

(0.00–0.30)
(0.02–0.49)
(0.07–0.51)
(0.36–0.60)

Percent Year Effect Change
Allison et al 2008

Buck et al 2005

Jenkins and Sherrod 2005

Millsap et al 2004

Nye 2010

Route and Key 2009

Stinson et al 2007

Todd 2004

Watkins and Mulhern 1999

Watts et al 2008

Zwiefelholder 2007

1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
-40

-20

0

20

40

-40

-20

0

20

40

Percent Change
Figure A2-2. Bald eagle productivity model random effects (percent change) and 95% credible intervals due to year
given study.

46

Figure A2-3. Bald eagle productivity model random effects (percent change) and 95% credible intervals due to study
after accounting for regional (top) and area (bottom) differences in the model (see Table A2-6 for the full list of studies
with the associated region, area, and year). The effects have wide credible intervals, some of which do not overlap zero.

47

Table A2-4. Bald eagle productivity model median random effects (and lower and upper limits, LCL and UCL, of the
95% credible intervals) apart from region. The effect medians are presented in descending order. Fledged and nest
counts are aggregated over all areas and years for each study. Ratio is the simple ratio of the total fledged to the total
occupied nesting territories across all areas and years.

Study

Ratio

Median

(LCL–UCL)

Fledged

Allison et al. 2008
Zwiefelholder 2007
Buck et al. 2005
Jenkins and Sherrod 2005
Todd 2004
Clark et al. 2007
Stinson et al. 2007
McDowell et al. 2000
McDowell and Itchmoney 1997
Bowerman et al. 1998
McHugh and Chanda 2005
Badzinski and Richards 2002
Watts et al. 2008
Millsap et al. 2004
Nye 2010
Clark et al. 2013
Watkins and Mulhern 1999
Route and Key 2009

0.74
0.84
0.93
0.88
0.92
0.97
1.04
1.16
1.21
1.21
1.21
1.24
1.26
1.32
1.31
1.38
1.71
1.55

0.73
0.84
0.85
0.88
0.91
1.00
1.03
1.12
1.16
1.16
1.17
1.20
1.20
1.28
1.30
1.35
1.42
1.48

(0.64–0.83)
(0.78–0.92)
(0.76–0.94)
(0.78–0.99)
(0.87–0.96)
(0.81–1.24)
(1.00–1.05)
(0.84–1.49)
(0.84–1.57)
(1.08–1.25)
(0.95–1.45)
(0.93–1.53)
(1.16–1.25)
(1.10–1.49)
(1.22–1.37)
(1.15–1.55)
(1.09–1.86)
(1.30–1.67)

234
836
766
241
1, 916
62
8, 074
29
17
1, 817
64
41
4, 001
158
1, 540
177
41
254

48

Occupied
Nesting
Territories
317
998
828
274
2, 091
64
7, 784
25
14
1, 497
53
33
3, 181
120
1, 178
128
24
164

Figure A2-4. Golden eagle productivity for the U.S. The blue curve is the empirical density distribution of the
estimates—which are shown via the rug just above the x-axis. The vertical blue line is the median with the area within
the 95% credible intervals shaded blue. The red and green curves represent the log normal and normal distributions
(respectively) defined by the estimated mean and standard deviation.

Table A2-5. Golden eagle productivity model median random effects (and lower and upper limits, LCL
and UCL, of the 95% credible intervals) apart from region. The effect medians are presented in
descending order. Fledged and nest counts are aggregated over all areas and years for each study. Ratio
is the simple ratio of the total fledged to the total occupied nesting territories across all areas and years.

Productivity random effects

Ratio

Median

(LCL–UCL)

Fledged

Hopi Navajo 2013
Hawkwatch International 2009a
Morneau et al. 2012
Hawks Aloft 2002
McIntyre and Schmidt 2012
Preston 2014
Hawks Aloft 2006
McIntyre and Adams 1999
Isaacs 2011
Berengia 2014
Ritchie et al. 2003
Hawkwatch International 2009b

0.51
0.50
0.49
0.50
0.61
0.56
0.64
0.61
0.60
0.60
1.18
0.92

0.49
0.51
0.53
0.53
0.53
0.54
0.54
0.54
0.54
0.55
0.56
0.58

(0.40–0.58)
(0.41–0.60)
(0.40–0.67)
(0.41–0.68)
(0.46–0.62)
(0.44–0.65)
(0.43–0.75)
(0.44–0.69)
(0.43–0.73)
(0.45–0.69)
(0.45–0.87)
(0.47–0.84)

362
257
24
38
692
149
27
112
169
117
13
85

49

Occupied
Nesting
Territories
715
510
49
76
1, 140
264
42
184
280
196
11
92

Percent Year Change Effect | Study Effects
Berengia 2014

Hawks Aloft 2002

Hopi Navajo 2013

1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014

Hawkwatch International 2009a Hawkwatch International 2009b

McIntyre and Adams 1999

1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014

McIntyre and Schmidt 2012

Morneau et al. 2012

Preston 2014

1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
-50

0

50

100

-50

0

50

100

-50

0

50

100

Percent Change
Figure A2-5. Golden eagle productivity model random effects (percent change) and 95% credible intervals due to year
given study. The random effect credible intervals all overlap zero.

50

Figure A2-6. Golden eagle productivity model random effects (percent change) and 95% credible intervals due to study
after accounting for regional differences in the model (see Table A2-7 for the full list of studies with the associated
region and year). The credible intervals all overlap zero.

51

Table A2-6. Bald eagle studies included in the analysis, their simple productivity ratios, and the final random effect model median estimates for each
area and year combination of the studies.

52

Study

Region

Watts et al. 2008
Allison et al. 2008
Allison et al. 2008
Allison et al. 2008
Allison et al. 2008
Buck et al. 2005
Buck et al. 2005
Allison et al. 2008
Todd 2004
Allison et al. 2008
Buck et al. 2005
Allison et al. 2008
Jenkins and Sherrod 2005
Allison et al. 2008
Jenkins and Sherrod 2005
Allison et al. 2008
Zwiefelholder 2007
Jenkins and Sherrod 2005
Jenkins and Sherrod 2005
Zwiefelholder 2007
Todd 2004
Jenkins and Sherrod 2005
Todd 2004
Jenkins and Sherrod 2005
Todd 2004
Jenkins and Sherrod 2005
Stinson et al. 2007
Stinson et al. 2007
Todd 2004

Other
SW
SW
SW
SW
Other
Other
SW
Other
SW
Other
SW
Other
SW
Other
SW
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other

Area

Lower Columbia River
Lower Columbia River

Lower Columbia River

Year

Fledged

Sample
Size

Productivity

Model
Median

1997
2003
2000
1998
1997
1997
1996
1996
1996
2001
1995
1995
1996
1999
1999
2002
1997
1998
1997
2002
2000
2000
2003
1995
1995
2001
1995
1995
1998

227
25
23
21
23
32
39
23
141
28
22
23
13
31
20
37
368
21
19
468
205
29
273
17
176
31
509
509
189

416
42
38
34
32
54
48
30
203
36
35
28
25
36
32
41
460
28
26
538
234
33
309
19
192
32
558
558
202

1.40
0.60
0.61
0.62
0.72
0.59
0.81
0.77
0.69
0.78
0.63
0.82
0.52
0.86
0.63
0.90
0.80
0.75
0.73
0.87
0.88
0.88
0.88
0.89
0.92
0.97
0.91
0.91
0.94

0.624
0.677
0.680
0.692
0.722
0.737
0.741
0.742
0.747
0.748
0.759
0.764
0.778
0.780
0.796
0.800
0.812
0.836
0.837
0.875
0.884
0.891
0.891
0.895
0.916
0.917
0.917
0.917
0.933

Table A2-6. Bald eagle studies included in the analysis, their simple productivity ratios, and the final random effect model median estimates for each
area and year combination of the studies. (continued)

53

Study

Region

Jenkins and Sherrod 2005
Stinson et al. 2007
Stinson et al. 2007
Todd 2004
Buck et al. 2005
Todd 2004
Buck et al. 2005
Todd 2004
Stinson et al. 2007
Stinson et al. 2007
Buck et al. 2005
Todd 2004
Clark et al. 2007
Bowerman et al. 1998
Jenkins and Sherrod 2005
Stinson et al. 2007
Stinson et al. 2007
Stinson et al. 2007
Stinson et al. 2007
Stinson et al. 2007
Stinson et al. 2007
McDowell et al. 2000
McDowell and Itchmoney 1997
Bowerman et al. 1998
McHugh and Chanda 2005
Bowerman et al. 1998
Badzinski and Richards 2002
Bowerman et al. 1998
Nye 2010

Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other

Area

Oregon
Oregon

Oregon

Year

Fledged

Sample
Size

Productivity

Model
Median

2002
1996
1996
1999
1997
2002
1996
2001
1997
1997
1995
1997

38
564
564
207
244
280
215
266
565
565
214
179
62
81
53
713
713
925
925
761
761
29
17
207
64
694
41
38
87

38
599
599
216
248
290
230
269
574
574
213
176
64
90
41
648
648
840
840
673
673
25
14
176
53
583
33
30
75

1.00
0.94
0.94
0.96
0.98
0.97
0.93
0.99
0.98
0.98
1.00
1.02
0.97
0.90
1.29
1.10
1.10
1.10
1.10
1.13
1.13
1.16
1.21
1.19
1.21
1.19
1.20
1.27
1.16

0.945
0.946
0.946
0.950
0.960
0.962
0.964
0.979
0.986
0.986
0.989
0.993
1.005
1.006
1.083
1.097
1.097
1.098
1.098
1.127
1.127
1.129
1.157
1.168
1.174
1.186
1.193
1.195
1.208

Michigan Great Lakes
2003
1998
1998
2005
2005
2001
2001

Michigan Interior
Wisconsin
Ohio
2003

Table A2-6. Bald eagle studies included in the analysis, their simple productivity ratios, and the final random effect model median estimates for each
area and year combination of the studies. (continued)

54

Study

Region

Millsap et al. 2004
Millsap et al. 2004
Watts et al. 2008
Millsap et al. 2004
Millsap et al. 2004
Nye 2010
Nye 2010
Millsap et al. 2004
Millsap et al. 2004
Route and Key 2009
Bowerman et al. 1998
Nye 2010
Millsap et al. 2004
Millsap et al. 2004
Nye 2010
Watts et al. 2008
Nye 2010
Nye 2010
Nye 2010
Watkins and Mulhern 1999
Nye 2010
Route and Key 2009
Clark et al. 2013
Nye 2010
Watts et al. 2008
Watts et al. 2008
Route and Key 2009
Route and Key 2009
Route and Key 2009

Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other

Area

Apostle Island NRA
Minnesota

Year

Fledged

Sample
Size

Productivity

Model
Median

1998
1998
1998
1999
1999
2005
2007
2000
2000
2007

12
16
563
13
15
112
153
15
15
10
797
244
13
19
223
490
111
190
83
5
94
18
177
71
464
650
28
6
8

12
12
462
12
12
92
124
12
12
9
618
192
12
12
173
377
84
145
62
5
70
14
128
51
340
472
19
4
8

1.00
1.33
1.20
1.08
1.25
1.22
1.24
1.25
1.25
1.10
1.29
1.27
1.08
1.58
1.29
1.30
1.32
1.31
1.34
1.00
1.34
1.30
1.38
1.35
1.40
1.40
1.50
1.50
1.00

1.213
1.213
1.216
1.222
1.222
1.235
1.245
1.246
1.246
1.249
1.276
1.278
1.280
1.280
1.286
1.291
1.310
1.311
1.317
1.327
1.329
1.330
1.343
1.347
1.347
1.362
1.372
1.378
1.386

Lake Superior shore

2010
2001
2001
2009
1996
2004
2008
2001
1995
2002
2007

St. Croix NRA upper
St. Croix NRA lower
Apostle Island NRA

2000
1995
1999
2007
2007
2008

Table A2-6. Bald eagle studies included in the analysis, their simple productivity ratios, and the final random effect model median estimates for each
area and year combination of the studies. (continued)

55

Study

Region

Watts et al. 2008
Watkins and Mulhern 1999
Millsap et al. 2004
Millsap et al. 2004
Route and Key 2009
Watkins and Mulhern 1999
Watts et al. 2008
Route and Key 2009
Nye 2010
Route and Key 2009
Route and Key 2009
Watkins and Mulhern 1999
Route and Key 2009
Route and Key 2009
Route and Key 2009
Route and Key 2009

Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other
Other

Area

Apostle Island NRA

Mississippi River pools
Mississippi River NRA
St. Croix NRA lower
St. Croix NRA upper
St. Croix NRA lower
Mississippi River NRA
Mississippi River NRA

Year

Fledged

Sample
Size

Productivity

Model
Median

2001
1998
1997
1997
2006
1996
2000
2008
2006
2007
2008
1997
2006
2006
2008
2006

849
11
17
23
16
9
758
36
172
22
16
16
31
9
30
24

601
7
12
12
11
5
513
24
110
15
10
7
20
5
14
11

1.40
1.57
1.42
1.91
1.50
1.80
1.50
1.50
1.56
1.50
1.60
2.29
1.60
1.80
2.10
2.20

1.398
1.405
1.426
1.426
1.440
1.449
1.454
1.494
1.494
1.532
1.534
1.556
1.582
1.589
1.704
1.770

Table A2-7. Golden eagle studies included in the analysis, their simple productivity ratios, and the final
random effect model median estimates for each year of the studies.

Study
McIntyre and Schmidt 2012
Hopi Navajo 2013
Hawkwatch International 2009a
Hopi Navajo 2013
Hopi Navajo 2013
McIntyre and Schmidt 2012
McIntyre and Schmidt 2012
Hopi Navajo 2013
Preston 2014
Hawkwatch International 2009a
Hawkwatch International 2009a
Hopi Navajo 2013
Hawks Aloft 2002
Morneau et al. 2012
Preston 2014
Morneau et al. 2012
McIntyre and Adams 1999
Preston 2014
Morneau et al. 2012
Hawkwatch International 2009b
Hopi Navajo 2013
McIntyre and Schmidt 2012
McIntyre and Adams 1999
McIntyre and Schmidt 2012
McIntyre and Schmidt 2012
McIntyre and Schmidt 2012
McIntyre and Schmidt 2012
Hawkwatch International 2009a
Hopi Navajo 2013
Berengia 2014
Hawkwatch International 2009a
Preston 2014
Hopi Navajo 2013
Hopi Navajo 2013
Berengia 2014
Hawks Aloft 2002
Morneau et al. 2012
Hawks Aloft 2006
Isaacs 2011
Hawkwatch International 2009b
Morneau et al. 2012
Ritchie et al. 2003

Region

Year

Fledged

Sample
Size

Productivity

Model
Median

AK
West
West
West
West
AK
AK
West
West
West
West
West
West
East
West
East
AK
West
East
West
West
AK
AK
AK
AK
AK
AK
West
West
West
West
West
West
West
West
West
East
West
West
West
East
AK

2002
2003
2001
2010
2002
2003
2004
1997
2012
2003
2002
2012
2002
2007
2011
2002
1995
2013
1997
2008
2001
1995
1996
2001
1996
1998
2005
2004
1996
2012
2007
2014
1999
2004
2011
2000
1998

4
9
13
6
20
19
20
7
15
26
23
23
12
4
17
2
25
17
2
8
35
24
30
31
28
33
38
42
14
21
34
29
39
43
23
12
3
27
169
8
6
13

73
60
60
29
71
71
73
25
48
78
68
60
33
14
44
8
59
42
7
16
75
56
62
68
61
66
76
84
26
41
67
54
70
76
42
22
5
42
280
11
9
11

0.05
0.15
0.22
0.21
0.28
0.27
0.27
0.28
0.31
0.33
0.34
0.38
0.36
0.29
0.39
0.25
0.42
0.41
0.29
0.50
0.47
0.43
0.48
0.46
0.46
0.50
0.50
0.50
0.54
0.51
0.51
0.54
0.56
0.57
0.55
0.55
0.60
0.64
0.60
0.73
0.67
1.18

0.296
0.358
0.397
0.415
0.417
0.420
0.424
0.445
0.447
0.448
0.456
0.468
0.474
0.474
0.480
0.481
0.484
0.485
0.492
0.494
0.496
0.497
0.504
0.505
0.506
0.518
0.520
0.520
0.521
0.521
0.522
0.530
0.531
0.535
0.535
0.539
0.540
0.542
0.544
0.546
0.557
0.558

2007
2004

56

Table A2-7. Golden eagle studies included in the analysis, their simple productivity ratios, and the final
random effect model median estimates for each year of the studies. (continued)

Study
Berengia 2014
Berengia 2014
Berengia 2014
McIntyre and Schmidt 2012
Hawks Aloft 2002
Hopi Navajo 2013
Hopi Navajo 2013
McIntyre and Schmidt 2012
McIntyre and Schmidt 2012
Preston 2014
Hopi Navajo 2013
Hawkwatch International 2009a
Hawkwatch International 2009a
Morneau et al. 2012
McIntyre and Schmidt 2012
Hawkwatch International 2009b
McIntyre and Schmidt 2012
McIntyre and Schmidt 2012
McIntyre and Schmidt 2012
McIntyre and Adams 1999
McIntyre and Schmidt 2012
Preston 2014
Hawkwatch International 2009b

Region

Year

Fledged

Sample
Size

Productivity

Model
Median

West
West
West
AK
West
West
West
AK
AK
West
West
West
West
East
AK
West
AK
AK
AK
AK
AK
West
West

2013
2014
2010
2010
2001
2005
2000
2008
2000
2010
1998
2005
2006
2000
1997
2005
2007
2009
2006
1997
1999
2009
2004

26
26
21
49
14
59
55
52
51
34
52
67
52
7
58
32
73
67
76
57
69
37
37

41
41
31
75
21
84
76
75
70
43
63
87
66
6
69
35
81
74
80
63
72
33
30

0.63
0.63
0.68
0.65
0.66
0.70
0.72
0.69
0.73
0.79
0.83
0.77
0.79
1.17
0.84
0.91
0.90
0.91
0.95
0.90
0.96
1.11
1.23

0.561
0.564
0.572
0.575
0.579
0.583
0.584
0.585
0.596
0.615
0.621
0.621
0.624
0.633
0.636
0.640
0.658
0.660
0.675
0.675
0.677
0.718
0.732

57

Appendix A3. The 2009 National Bald Eagle

Post-Delisting Survey and Estimation Results

Mark Otto1 , John Sauer2 , Emily Bjerre1 and Brian Millsap1
2 U.S.

1 U.S.

Fish and Wildlife Service
Division of Migratory Bird Management

Geological Survey
Patuxent Wildlife Research Center

Abstract
In 2009, with assistance from many States, Tribes, and other collaborators, the U.S. Fish and Wildlife
Service conducted a national-scale survey to estimate the total number of occupied bald eagle nesting
territories in the coterminous U.S. The survey followed a dual-frame sampling design developed by Haines
and Pollock (1998) that we modified to account for detectability of nest structures and tested with pilot studies
in several states from 2004–2006. We estimate 16,048 (SE 727) occupied bald eagle nesting territories in
the coterminous U.S. in 2009. The estimate includes 13,025 occupied nesting territories estimated with the
dual-frame design (overall coefficient of variation of 6) for areas with a high abundance of nest structures,
and 3,023 occupied nesting territories, which we did not estimate with the dual-frame design and consider a
minimum count, in areas with a low abundance of nest structures. The dual-frame approach illustrates the
difficulties of adequately estimating the true number of nests from existing nest lists and the challenges of
effectively sampling across a large geographic extent. The substantial increase in the number of occupied
nesting territories from previous estimates reflects improvements in sampling design and estimation, as well
as increases in the overall number of breeding pairs of bald eagles.

Survey Overview
In the ‘Post-Delisting Monitoring Plan for the Bald Eagle’ (Post-Delisting Plan), the U.S. Fish and
Wildlife Service (Service) describes the post-delisting survey (PDS), a national-scale survey designed to
monitor the status of the bald eagle by collecting occupancy data on nest structures (nests) over a 20-year
period, beginning in early 2009 with a baseline survey (U.S. Fish and Wildlife Service 2009). In this case,
occupied nests are equivalent to occupied nesting territories since there is only one occupied nest structure
per nesting territory (see Steenhof and Newton 2007). The baseline PDS design provides an efficient and
unbiased estimate of the number of occupied nests. It was designed, in part, to meet specific objectives for
detecting population change with future surveys which affected survey sampling effort. The survey design
takes advantage of the efficiency of surveying known nest locations with using randomly selected area plot
searches to account for nests not included on the lists, while also accounting for detectability of nests using a
modification of Haines and Pollock (1998) dual-frame survey approach.

Dual-frame Survey Design
Many States, Tribes, or other partner organizations maintain lists of current, and often historical, bald
eagle nest locations within states and the status of those nests in certain years (e.g., whether or not they
were used by a nesting pair of bald eagles). The effort put toward checking the status of these known nests
varies depending on the State and the resources available and ranges from intensive State-wide censuses
to opportunistic checks or reports by members of the public. The recent recovery success of bald eagles
58

combined with budget limitations has caused many states to discontinue long-term annual nest occupancy or
production surveys. The cumulative information from these nest lists, though impressive, is incomplete (not
all nests were found and checked), inconsistent (information recorded differs among states), may be biased
(in particular in areas where nests are found opportunistically rather than with a dedicated search effort), and
frequently out-of-date in terms of occupancy information. Because of these limitations, it is difficult to tease
out unbiased estimates of nest occupancy and there is only a limited ability to extrapolate beyond the lists
themselves to an overall breeding population of bald eagles.
The dual frame survey design provides a means of using the existing data on the number of occupied nests
available from State lists and estimating the number of additional occupied nests not accounted for by the
lists. This requires sampling survey plots for occupied nests, removing known nests (those on the lists) from
the samples, and estimating the number of occupied nests that are not on the lists. Those numbers are then
added to the list data to get the total number of occupied nests. Complications arise in this estimation because
1) not all states have updated their lists, so additional sampling of the list nests was necessary in some areas to
estimate occupancy status for the lists; 2) not all nests are seen during sampling, necessitating the estimation
of detectability of nests (occupied and unoccupied) during the area sampling using a multiple-observer
procedure; 3) the occupancy status of observed nests is not certain; and 4) bald eagle abundance varies greatly
within the coterminous U.S. and plot surveys could only be implemented in areas where bald eagle nesting
populations could be surveyed with reasonable effort. The methods described in this paper accommodate all
of these issues.
In 2004–2006, USFWS and the U.S. Geological Survey (USGS) conducted several pilot studies to test
the dual-frame sampling approach by combining methods traditionally used by states to monitor occupied
nests, standardizing the survey protocols, and adding area-based sample plot searches for nests not included
on the nest lists (U.S. Fish and Wildlife Service 2009). Area sample plots also included a multiple-observer
detectability component, which allowed the analysis to include an estimate of the proportion of nests missed
during the area-based plot search surveys. These pilot surveys confirmed the utility of the dual-frame design
for eagle nest surveys and provided the basis for the baseline PDS design, including the expected nest list
coverages (percent of the total nests represented on State nest lists) used to design appropriate sampling levels
given the survey objectives (U.S. Fish and Wildlife Service 2009).

Stratification for the 2009 Survey and Analysis
Because bald eagles are still rare in some portions of the United States, the PDS was implemented within
strata defined by physiographic regions (FigureA3-1). Strata sampled as part of the dual frame survey design
were only in regions where eagles are relatively abundant (Table A3-1). The final strata were based on
a cluster analysis of bald eagle nest densities within Bird Conservation Regions (BCRs) and States with
additional editing in consultation with survey biologists to incorporate knowledge of local natural history and
habitat information (see U.S. Fish and Wildlife Service [2009] and U.S. North American Bird Conservation
Initiative Committee [2000]). Area strata may include all or part of multiple State list frames since nest
lists are nearly always maintained at individual State levels (Table A3-1). State list-frames are thus divided
and combined as appropriate to create a strata-specific list-frame. The strata list-frame estimate is then
combined with the area-frame survey estimate (corrected for detectability of nests based on multiple observer
procedures) to compute the dual-frame estimates by strata.

Methods
Sample Unit Selection and Sample Size
We selected sample plots for the survey using the Generalized Random Tessellation Stratification (GRTS)
methodology (Kincaid and Olsen 2013) as described in the Post-Delisting Plan. The dual-frame sample is
59

Figure A3-1. Survey strata for the 2009 bald eagle post-delisting survey. We defined the strata for area plot samples in
terms of Bird Conservation Regions (U.S. North American Bird Conservation Initiative Committee 2000) and densities
of known nests. We further subdivided the area-frame strata to accommodate differences in nest lists maintained by the
component states or to meet information needs of cooperators. Geographic areas are thus identified as AA.BB, where
AA defines the list sample frame and BB defines the area sample frame. We did not sample light grey areas with area
plot surveys and estimated occupied nests only from list frame data.

based on an area sample that is augmented with information from known nest locations (the list). Because
list information is often not current, additional sampling may be required to estimate the total number of
occupied nests in the list. Here, we used GRTS to select plots for the area sampling, but also selected plots to
estimate the number of the occupied nests in the list. Sample size determination for both the area sampling
and list sampling are described in detail in the Post-Delisting Plan; list sampling included any known (list)
nests included in selected area sample plots and the additional list sample plots. In list sample plots we only
assessed the status of known nests, whereas in area plots we searched the entire plot area for any previously
unknown (new) nests and also assessed the occupancy status of any known nests (U.S. Fish and Wildlife
Service 2009). Known nests were removed from the area sample and made part of the list sample prior to
data analyses. We drew a 20% over-sample for each sampling frame by strata; these were replacements for
selected plots that could not be flown for a variety of access, safety, or logistical issues.
Surveys
We conducted survey flights in the early part of the breeding season when the majority of bald eagle
breeding pairs are closely tied to nests and detectability should highest (before leaf-out). Survey crews
recorded GPS locations and nest status observations following protocols described in the Post-Delisting Plan.
Post-survey, we used flight tracks and nest information to reconcile nest observations with known nests and
60

Table A3-1. Geographic sampling areas with the associated list frame (LL), area frame (AA), dual-frame
(LL.AA) and a short locality description based on the associated Bird Conservation Region (BCR).

List
Frame

Area
Frame

Dual

Frame

Northwest (NW)

WA
WA
CR
OR

SC
WC
CR
OR

WA.SC
WA.WC
CR.CR
OR.OR

South Cascades
Olympic Sound & Northeast Cascades
Columbia River
Pacific Rainforest

Northern Rockies (NR)

NR

NR

NR.NR

Northern Rockies

Great Lakes (GL)

WI
MI
OT

WI
OT

Sampling Area

BT
PT

WI.BT

MI.BT
OT.BT
WI.PT
OT.PT

BCR Description

Boreal Hardwood Transition

Pine Hardwood Transition

Louisiana (LA)

LA
LA

MV
LC

LA.MV
LA.LC

Mississippi Alluvial Valley
Western Gulf Coastal Plain

Maine (ME)

ME
ME
ME

MH
ML
UC

ME.MH
ME.ML
ME.UC

Maine Upper Middle Coast & Highlands
Maine including Aroostook
Maine Down East

Chesapeake Bay (CB)

CH

CH

CH.CH

Chesapeake (Mid-Atlantic Coast)

Coastal Plain (CP)

SC
GA

LL

SC.LL
GA.LL

Southeastern Coastal Plain

Florida (FL)

FL
FL

FC
FN

FL.FC
FL.FN

Central Florida
Northern Florida

determine the sampling frame assignment and status for each nest.

Estimating Occupied Nests
List Estimation
Due to the variation in list quality, we used several different methods to determine the number of occupied
nests in the lists. Some states conducted a census of their lists in 2009 and could provide the number of
known, occupied nests directly; in other states list sampling was conducted by the Service in conjunction
with the area plot surveys to estimate the proportion of nests in the lists that were occupied. Finally, some
states only record occupancy by territory rather than by nest, requiring additional assumptions (e.g., that other
known nests within the territory were unoccupied) or adjustments in order to estimate the total number of
occupied nests.
The proportion of occupied nests is:
∑︀𝑛𝐿𝑖
𝑝𝑂𝑐𝑐
𝑖

𝑖

=

𝑛𝐿
𝑖

61

𝐿
𝑦𝑖𝑗

,

𝑡ℎ
where 𝑛𝐿
𝑖 is the number of sampled list nests for the 𝑖 stratum. To simplify the estimate, we are ignoring
𝐿 is a variable to indicate if
that the list sample unit is a plot with list nests, rather than the nests themselves. 𝑦𝑖𝑗
𝑡ℎ
the 𝑗 nest in the stratum is occupied. The total number of occupied nests is:

𝑌𝑖𝐿 = 𝑁𝑖𝐿 𝑝𝑖𝑂𝑐𝑐 ,
where the proportion of occupied nests is multiplied by 𝑁𝑖𝐿 , the total number of list nests in the 𝑖𝑡ℎ stratum.
The variance is:
(︀
)︀ (𝑆𝑖𝐿 )2
𝑉 𝑎𝑟(𝑌𝑖𝐿 ) = 𝑁𝑖𝐿 𝑁𝑖𝐿 − 𝑛𝑖𝐿
,
𝑛𝐿
𝑖
where sample standard deviation of the proportion is:
√︃
(︀
)︀
𝑛𝐿
𝐿
𝑆𝑖 = (︀ 𝐿 𝑖 )︀ 𝑝𝑖𝑂𝑐𝑐 1 − 𝑝𝑖𝑂𝑐𝑐 .
𝑁𝑖 − 1
Finally, the stratum estimates and variances are added to get the survey total occupied list nests and its
variance. The square root of the variance is the standard error of the total occupied list nests.
Area Estimation: Correcting for Detection Probability With Multiple Observers
We used a capture-recapture based approach for estimating detectability of nests by observers when
surveying plots. For each nest naively observed (that is the observers did not have knowledge of the nest
location at the time they recorded the observation), we used information about which observers detected
the nest to create a capture history (in this context “capture” is an observation state). The capture histories
provided sufficient information to estimate detection as described in the Post-Delisting Plan. Pilots were
included in the capture histories even though they were not observing at all times by including a “not looking”
code in the capture histories in addition to the ‘detected’ and ‘did not detect’ codes. Recorded ‘not looking’
observations were considered non-detections for that observer in the analysis. We did not record capture
histories for nests where the observers were aware of the nest location prior to their observations.
We developed a Bayesian capture-recapture model with non-informative priors based on Link and Barker
(2010) in JAGS (Plummer 2003) to estimate detection probabilities among two or three independent observers.
We summarized capture histories by combinations of observers in the pilot, front and rear seats (Table A3-2).
Captures have three possible states: seen (1), not seen (0), and not looking (x). Capture histories for each nest
observation are presented as 3-tuples (pilot observation, front seat observation, rear seat observation) and
can have seven possible values—for example: a nest seen only by the front observer (010), only by the rear
observer (001), or by all of the observers (111). The model accounted for observer, seat (pilot, front, or rear),
and platform (fixed-wing airplane or helicopter) as random-effects off an overall mean, 𝜇, on the logit scale:
logit(𝑝𝑑𝑒𝑡
𝑖𝑗𝑘 ) = 𝜇 + platform𝑖 + seat𝑗 + observer𝑘
Each platform-observer-combination was fit with a multinomial for seven combinations of detections. The
multinomial fits the number of capture histories of a given observer combination in a given platform and a
given seat position. The probabilities are conditioned on the nest being observed. We estimated the probability
of at least one of the three observers seeing a nest, irrespective of nest status, for each platform and observer
combination (i.e., group-specific detection rates), along with 95% credible estimates. We then adjusted nest
counts for detectability using the estimated detection probabilities for the observer-seat-platform combination
that surveyed each plot.

62

Table A3-2. Survey capture history summaries and estimated detection probabilities by observer combination, seat,
and platform. Seat positions include, in order, pilot, front seat observer (Front), and rear seat observer (Rear).
Survey platform is either fixed-wing plane (FW) or helicopter (Hel). ‘0’ in the capture history indicates the nest
was ‘not seen’ by the observer and a ‘1’ indicates ‘seen’. Detection probability is the probability at least one of the
observers (given observers, platform and seats) would see a nest, and includes the estimated standard error.

Platform

Pilot

Front

Rear

001

010

011

100

101

110

111

Detection
Probability

SE

FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
FW
Hel
Hel
Hel
Hel

JKB
JKB
JKB
VRB
VRB
BBD
MDK
MDK
TSL
TSL
TSL
FHR
FHR
FHR
FHR
FHR
FHR
WER
WER
WER
WER
DJS
JWS
JWS
PPT
PPT
PPT
PPT
PPT
RT
MBS
MBS
MBS
MBS

DED
CST
CST
DJ
DSP
SAN
CAK
HHO
CAK
CAK
JCO
SPE
GEM
GEM
JGM
MCO
MS
CH
CAK
CAK
CAK

CST
DED
KS
DSP

7
9
0
3
0
8
0
3
0
10
0
1
2
2
0
0
1
0
1
0
4
2
0
4
0
4
0
0
3
3
6
2
1
0

1
28
2
1
0
4
6
1
1
21
1
5
0
7
0
3
1
2
15
0
11
0
3
4
5
0
6
0
4
4
13
2
13
14

2
20
0
12
0
4
3
5
1
6
1
2
0
2
0
0
1
2
4
3
9
0
0
3
2
2
0
1
13
8
24
2
4
0

0
0
0
8
1
5
4
6
0
2
0
0
1
1
0
8
1
0
2
0
0
0
3
1
1
0
1
1
1
2
3
3
3
5

0
0
0
0
0
13
1
1
0
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
7
0
0
0
0

0
6
0
2
0
1
1
4
0
4
0
2
1
1
0
5
0
0
0
0
0
0
2
1
0
0
2
0
0
2
0
7
4
25

0
9
0
6
0
15
1
1
2
3
0
6
2
2
0
0
1
0
0
0
0
0
0
1
1
1
0
0
0
31
5
27
3
0

90
91
89
90
80
97
85
78
86
82
85
94
89
85
76
56
76
85
80
87
78
64
66
82
84
93
77
83
93
98
96
97
94
90

3.0
2.9
4.3
3.5
4.7
0.8
4.0
5.6
3.9
3.5
10.1
3.2
5.5
5.4
7.3
7.1
9.2
10.4
3.8
6.7
3.9
18.2
6.9
6.0
5.6
3.3
6.8
6.1
3.4
0.7
1.1
0.8
1.5
2.0

ERB
JWG
DSB
DSB
DSB
NK
RM
SAN
DSP
DSP
DSP
DSP

JHW
JK
RDR
JGM
RDR
JB
JGM
TJH
CWJ

MCO
GR
ERB
KH
RDR
SMP
ERB
NK
RM
DSB
DSB
JHW
ERB
VRB
MCO

63

Area Estimation: New Occupied Nest Estimation
We removed any known nests (list nests) that were located in area survey plots from the area-frame data
set prior to the analysis in order to ensure independence of the two data frames. The area-frame estimates
thus represent the number of new occupied nests. By ‘new’, we mean new to the list, not necessarily newly
constructed (e.g., one could observe a well-established nest that is clearly ‘old’ but previously unknown,
especially in areas not traditionally included in State surveys). New nests found in 2009 were added to the
list frame for future surveys.
Thompson (1992, pg. 168 equation 9) provides formulas for estimating population size under simple
random sampling, accounting for detection uncertainty which can be applied independently to each stratum.
The estimator assumes detection is constant over the stratum, however, detection can vary by observer, seat,
and platform. To estimate population (nest) totals, we obtained the mean or density of occupied new nests
and the variance for each stratum given a detection probability,
𝑦¯𝑖𝑁 |𝑝𝐷𝑒𝑡

⎛ 𝐴
⎞−1 𝐴
𝑁𝑖
𝑁𝑖
𝑁
∑︁
∑︁
𝑤𝑖𝑗 𝑦𝑖𝑗
⎝
⎠
=
𝑊𝑖𝑗
𝑝𝐷𝑒𝑡
𝑖𝑗
𝑗
𝑗

)︀
𝑉 𝑎𝑟 𝑦𝑖𝑁 |𝑝𝐷𝑒𝑡 =
(︀

𝑛𝐴
𝑖
𝐴
(𝑁𝑖 −

⎛
1)

⎞−1

𝐴

𝑁𝑖
∑︁

⎝

𝑤𝑖𝑗 ⎠

𝑗

𝐴

𝑁𝑖
∑︁
(︀

(︀ 𝑁 𝐷𝑒𝑡 )︀)︀2
𝑁
𝐷𝑒𝑡
𝑦𝑖𝑗
− 𝑝𝑖𝑗
𝑦¯𝑖 |𝑝
,

𝑗

where the weights are the ratio of the total number of plots in the stratum over the number of sample plots,
𝑤𝑖𝑗 = 𝑁𝑖𝐴 /𝑛𝐴
𝑖 for all j. Then, we used Monte Carlo integration (Givens and Hoeting 2005, pg. 144) over the
distributions of the detections to get the unconditional population densities and variances using formulas 1.6
and 1.7 in Gelman et al. (1995, pg. 20). The distributions of the detections were simulated from the Bayesian
model by platform-observer-seat combination.
The totals and variances were obtained by expanding the densities,
(︀
)︀
(︀ (︀
)︀)︀
𝑦¯𝑖𝑁 = 𝐸𝑝𝐷𝑒𝑡 𝑦𝑖𝑁 = 𝐸𝑝𝐷𝑒𝑡 𝑦¯𝑖𝑁 |𝑝𝐷𝑒𝑡 = 𝐸𝑝𝐷𝑒𝑡 𝐸 𝑦𝑖𝑁 |𝑝𝐷𝑒𝑡
(︀
(︀
)︀)︀
(︀
(︀
)︀)︀
𝑉 𝑎𝑟(𝑦𝑖𝑁 ) = 𝑉 𝑎𝑟𝑝𝐷𝑒𝑡 𝐸𝑝𝐷𝑒𝑡 𝑦𝑖𝑁 |𝑝𝐷𝑒𝑡 + 𝐸𝑝𝐷𝑒𝑡 𝑉 𝑎𝑟 𝑦𝑖𝑁 |𝑝𝐷𝑒𝑡
which were applied to get the unconditional totals and variances. We report the standard errors by taking the
square root of the variances.
𝑌𝐼𝑁 = 𝑁𝑖𝐴 𝑦¯𝑖𝑁
𝑉 𝑎𝑟(𝑌𝑖𝑁 ) = 𝑁𝑖𝐴 (𝑁𝑖𝐴 − 𝑛𝑖𝐴 )

𝑉 𝑎𝑟(𝑦𝑖𝑁 )
,
𝑛𝐴
𝑖

Area Estimation: Area-Only Estimation
We also generated an estimate of the area-plot counts (as though there were no list data) for comparison.
We used all nests observed in the area plots (including list nests) and the observer-detection correction
to calculate the totals and standard errors as described above and compared estimates to the list-only and
dual-frame estimates. Applying the observer-detection correction to the known list nests may make estimates
slightly high since detection rates may be higher than for unknown nests. It is also important to include an
equivalent number of sampled area plots relative to the effort put into the list sample in order to make a
reasonable comparison between the estimates.
64

Dual-Frame Estimation
Since the estimates of occupied, known nests from the list frame and the estimates of occupied, new nests
from the area-frame are independent, we sum the totals and their variances to get the dual-frame estimates
(see U.S. Fish and Wildlife Service [2009] for additional details). We added the list-frame and area-frame
stratum estimates to get the national total following Haines and Pollock (1998):
𝑌^𝐷 𝐹 =

𝐼𝐿
∑︁

𝑌^𝐿 𝑖 +

𝑖

𝐼𝐴
∑︁

𝑌^𝑁 𝑖

𝑖

We similarly added the variances to get the variance of the total,
𝑉 𝑎𝑟(𝑌^𝐷𝐹 ) =

𝐼𝐿
∑︁

𝑉 𝑎𝑟(𝑌^𝐿 𝑖) +

𝐼𝐴
∑︁

𝑖

𝑉 𝑎𝑟(𝑌^𝑁 𝑖)

𝑖

List Coverage
The list coverage is the percent of the total estimated nests (irrespective of occupancy status) that were
represented in the list frame; it highlights the relative number of nests missing from the list. The number of
new nests is estimated using
𝑁𝐿
(𝐿𝑖𝑠𝑡 𝐶𝑜𝑣𝑒𝑟𝑎𝑔𝑒 %)𝑖 = 𝐿 𝑖 𝑁 .
^
𝑁 +𝑁
𝑖

𝑖

The number of nests in the list is fixed, but the number of new nests is estimated. We simulated the new nest
distribution with a log normal with the log mean of the new nest estimate and its relative variance. The means
and standard errors are taken from simulations of the above ratios.

Results and Discussion
The 2009 PDS was designed to 1) estimate the total number of occupied nesting territories for bald eagles
in the coterminous U.S. using occupied nests as a measure of occupied nesting territories, 2) ensure estimates
are unbiased and account for uncertainty, and 3) collect data such that the estimates allow for detection of
population decline with future surveys as specified in the Post-Delisting Plan. We achieved these objectives
using a dual-frame analysis to estimate the number of occupied nests and the uncertainty of the estimates.
This approach allowed us to combine information from State nest lists (the list-frame), which was efficient to
sample but was not standardized across States and can be biased, with information about nests that are not
represented on the lists (the area-frame), which was less efficient to collect but used standardized protocols
and accounted for detectability of nests. This approach also allowed for a direct comparison of estimates
based solely on sampling the list-frame or the area-frame to the dual-frame estimates. Stratification using
known nest densities and physiographic boundaries helped ensure representative and efficient sampling and
estimation which are important given the PDS goals and highly variable bald eagle abundance within the U.S.
List-Frame Estimates
For list-frame estimates, when appropriate, we used State survey data for list nests rather than list-frame
sample data collected specifically for the PDS. Often States surveyed a larger proportion of their total list
and, in cases where we used State survey data, the observed proportion of list nests that were occupied was
comparable to what we observed in the PDS sample (Table A3-3). We included observations of 7,461 nests
65

Table A3-3. List-frame estimates and nest observation totals of occupied bald eagle nests by strata. Table A3-1 describes the strata.
‘Survey Data’ indicates whether the data are from a State survey or the post-delisting survey.

66

Sampling
Area

Area
Strata

List
Strata

Survey
Data

Occupied
Seen

Total
Seen

Occupied (%)

SE

Total
Not Seen

Total
Occupied

SE

CV (%)

NW

SC
WC
CR
OR

WA
WA
CR
OR

PDS
PDS
State
PDS

7
32
117
57

11
87
275
130

62
37
43
44

13
5
3
4

513
2, 277
150
1, 181

103
873
181
575

20.5
117.0
4.5
51.0

20
13
2
9

NR

NR

NR

PDS

14

18

76

10

645

506

61.3

12

BT

WI
MI
OT

State
State
PDS

683
528
22

2, 100
2, 104
54

33
25
41

1
1
7

583
230
1, 069

873
586
459

6.0
2.2
70.2

1
0
15

PT

WI
OT

State
PDS

383
5

732
11

54
46

2
14

184
610

492
285

3.4
84.3

1
30

LA

MV
LC

LA
LA

PDS
PDS

25
1

36
5

69
25

8
46

448
25

334
7

33.6
4.1

10
56

ME

MH
ML
UC

ME
ME
ME

State
State
State

44
48
408

127
91
1, 145

35
53
36

4
5
1

23
2
60

52
49
429

1.0
0.1
0.8

2
0
0

CB

CH

CH

PDS

36

79

46

6

1, 011

497

56.0

11

CP

LL

SC
GA

State
State

252
42

294
62

86
67

2
6

4
0

255
42

0.1
0.0

0
0

FL

FC
FN

FL
FL

PDS
PDS

57
10

85
15

67
66

5
12

1, 279
599

912
403

64.5
69.0

7
17

GL

Table A3-4. Estimates of new, occupied nests by area-frame strata. Plot densities are not corrected for observer
detection, but new occupied nest totals are corrected for detection by the observer-seat-platform combination used
to sample the plot.

Area
Stratum

New
Occupied

Sampled
Plots

Plot
Density

SE

Total
Plots

Total New
Occupied

SE

CV (%)

NW

SC
WC
CR
OR

1
10
1
3

5
18
6
22

0.20
0.56
0.17
0.14

0.20
0.15
0.17
0.07

203
452
115
1,244

42
262
20
127

40.5
65.7
19.1
81.6

96
25
93
64

NR

NR

2

14

0.14

0.10

2,603

401

250.7

62

GL

BT
PT

13
3

27
9

0.48
0.33

0.17
0.17

2,549
1,775

1, 614
660

461.3
291.8

29
44

LA

MV
LC

9
0

15
5

0.60
0.00

0.32
0.00

610
849

429
0

198.3
0.0

46
0

ME

MH
ML
UC

1
0
1

13
7
15

0.08
0.00
0.07

0.08
0.00
0.07

268
418
411

23
0
31

20.6
0.0
27.4

90
0
90

CB

CH

33

33

1.00

0.30

570

702

172.4

25

CP

LL

4

17

0.24

0.14

991

280

139.6

50

FL

FC
FN

8
5

19
16

0.42
0.31

0.12
0.15

613
801

264
257

71.4
120.6

27
47

Region

(2,781 occupied nests) out of 17,994 known nests on State nest lists for the high-density strata. Occupancy
rates ranged from 25–86% among strata and averaged approximately 50%. Within strata, we extrapolated the
occupancy rates and standard errors for the sampled nests to the nests not included in the sample to get the
total occupied nests for the list-frame—7,913 (SE 727) occupied nests.
Area-Frame Estimates
Table A3-4 shows the total new nests observed in area plots by stratum and the resulting plot densities
which ranged from 0–1 new nests per plot (average 0.3). We applied detection probabilities that accounted
for observer combination, seat, and platform to the plot densities to get the area-frame estimates. New nest
estimates ranged from 0 new nests in parts of Maine and Louisiana where nest lists were actively updated
through 2009 to 1,614 new nests in the Boreal Transition area, which is the northern part of the Great Lakes
region.
Detection rates varied by observer-seat-platform combination with the probability of at least one observer
detecting a nest ranging from as low as 56% (SE 7.1%) to as high as 98% (SE 0.7%). Platform had a
noticeable impact on detectability, with helicopter crew detection rates of 90% (SE 0.8%) to 97% (SE 1.5%).
This was consistent with the general impression of survey crews that searching for nest in helicopters was
more accommodating in terms of general maneuverability and visibility in spite the helicopters often being
used for plots with habitat that would be considered to have lower detectability (conifer-dominant tree stands).

67

Table A3-5. List-frame (List), area-frame (Area), and dual-frame (DF) estimates of occupied nests
and standard errors by stratum for the 2009 National post-delisting survey. The list-frame
estimates are known nests, the area-frame estimates are only new nests (known nests excluded),
and the dual-frame estimates combine both new and known nests from the two independent
sampling frames.

Region

Area Stratum

List

SE

Area

SE

DF

SE

Northwest

SC
WC
CR
OR

103
873
181
575

20.5
116.5
4.5
51.0

42.0
262.0
20.0
127.0

40.5
65.7
19.1
81.6

145
1,135
201
702

45.4
133.7
19.6
96.2

Northern Rockies

NR

506

61.3

401.0

250.7

907

258.1

Great Lakes

BT
PT

1,918
776

70.5
84.4

1,614.0
660.0

461.3
291.8

3,532
1,436

466.7
303.8

Louisiana

MV
LC

334
7

33.6
4.1

429.0
0.0

198.3
0.0

762
7

201.1
4.1

Maine

MH
ML
UC

52
49
429

1.0
0.1
0.8

23.0
0.0
31.0

20.6
0.0
27.4

75
49
460

20.6
0.1
27.4

Chesapeake

CH

497

56.0

702.0

172.4

1,200

181.3

Coastal Plain

LL

297

0.1

280.0

139.6

577

139.6

Florida

FC
FN

912
403

64.5
69.0

264.0
257.0

71.4
120.6

1,177
660

96.2
139.0

Dual-Frame Estimates
The dual-frame estimates represent the total occupied nests for each stratum (Table A3-5), and ranged
from only 7 occupied nests in the Western Gulf Coastal Plain of Louisiana to 3,532 occupied nests in the
Boreal Transition area in the northern portion of the Great Lakes region (SE 4.1 and 466.8, respectively). The
overall estimate for all of the high-density strata combined was 13,025 occupied nests (SE 727). We combine
this with the minimum number of occupied nests for the low-density strata to get the total national estimate
of occupied nests for the coterminous U.S.
The Low-Density Strata
The occupied nest estimates for the low-density strata are based entirely on the list data for the States (or
portions of States) included geographically in the strata (Figure A3-1). We totaled the number of occupied
nests on each of the State nest lists that best represented the number of occupied nests in 2009 (Table A3-6).
There were 3,023 additional occupied nests in the low-density portions of the coterminous U.S. which when
added to the 13,025 occupied nests estimated for the high-density areas, yields a national total of 16,048
occupied nests or breeding pairs of bald eagles in 2009.
Area-Only Estimates
To evaluate the efficacy of the dual-frame survey and better understand the contribution of each sample
frame to the analysis, we also analyzed the area survey plot data as though we only conducted the area
68

Table A3-6. The recorded number of occupied nests (considered minimums) for low-density
areas (areas in the coterminous U.S. not included in the high-density strata in Figure A3-1).
Reporting years other than 2009 are shown in parentheses. ‘Low’ after the State name indicates
the low-density portion of a state partially included in a high-density area.

State List
Alabama (2006)
Arizona
Arkansas (2008)
California
Colorado
Connecticut (2010)
Florida South
Idaho Low
Illinois Low
Indiana
Iowa Low
Kansas
Kentucky
Massachusetts (2010)
Michigan Low
Minnesota Low
Mississippi
Missouri (2011)
Montana Low
Nebraska

Occupied Nests
77
50
110
64
51
18
120
84
59
194
199
33
56
30
77
61
31
165
97
48

State List
Nevada
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon Low
Pennsylvania
Rhode Island
South Dakota (2012)
Tennessee
Texas (2005)
Utah
Vermont
Washington Low
West Virginia (2010)
Wisconsin Low
Wyoming Low (2005)
Total

69

Occupied Nests
3
11
69
5
173
113
67
151
72
20
173
1
128
130
160
11
3
18
36
5
50
3,023

Table A3-7. ‘Area-only’ estimates of occupied nests, which use
only information collected during area plot surveys (assumes no
prior knowledge of nest locations from nest lists), by area-frame
stratum.

Region

Area
Stratum

Occupied

SE

CV (%)

Northwest

SC
WC
CR
OR

169
969
84
1, 221

162
224
38
619

96
23
45
51

Northern Rockies

NR

2, 462

980

40

Great Lakes

BT
PT

3, 110
869

744
302

24
35

Louisiana

MV
LC

1, 002
0

507
0

51
0

Maine

MH
ML
UC

115
0
515

65
0
150

56
0
29

Chesapeake

CH

1, 477

293

20

Coastal Plain

LL

349

147

42

Florida

FC
FN

1, 288
462

333
146

26
32

Table A3-8. The overall list-only, area-only, and
dual-frame estimates of occupied nests and standard
errors for the high-density strata.

Type
List-only
Area-only
Dual Frame

Occupied Nests
7, 913
14, 091
13, 025

70

SE
214
1,610
727

CV (%)
3
11
6

Table A3-9. The number list nests (List), new nests (New), and percent list
coverage (proportion of the total estimated nests that were represented on the
nest list) by area-frame stratum. The list nest and new nest columns include all
nests, regardless of occupancy status. We present the means and standard error
for each stratum.

List

New

List
Coverage (%)

SE

Northwest

SC
WC
CR
OR

164
2,364
425
1,311

42
367
42
127

76
86
90
90

15.5
3.0
5.3
6.0

Northern Rockies

NR

663

823

45

9.4

Great Lakes

BT
PT

6,140
1,537

2,621
1,352

70
53

6.2
10.4

Louisiana

MV
LC

484
30

524
0

48
100

10.8
0.0

Maine

MH
ML
UC

150
93
1,205

23
0
31

83
100
96

11.5
0.0
3.3

Chesapeake

CH

1,090

1,001

52

4.9

Coastal Plain

LL

360

280

56

11.6

Florida

FC
FN

1,364
614

264
257

83
70

3.7
9.5

17,994

7,752

68

3.1

Area

Overall

Area
Stratum

plot survey; we refer to this as the ‘area-only’ analysis (Table A3-7). The area-only estimates are generally
consistent with the dual-frame estimates but with much greater uncertainty around the estimates (Table A3-8).
The greatest disparity in the area-only and dual-frame estimates occurs in strata where we were able to include
data from a large sample of the nest list (e.g., the Columbia River strata [CR] in the Northwest). There is also
disparity where no occupied nests were detected in the sample plots, as in parts of Maine (ML), though this is
an artifact of the estimation of a zero variance where there should be some chance of a non-zero estimate
even when the estimate itself is zero. In all strata, however, the inclusion of the list frame in the dual-frame
analysis generally reduced the uncertainty in the estimates by 5%.
List Coverage
By estimating the total number of nests, regardless of status, using the same dual-frame approach for the
high-density strata we can also evaluate how well State nest lists represent the true number of nests on the
landscape (Table A3-9). The percent of the total estimated nests that were included on the nest lists, the list
coverage, ranged from 48–100%, with 68% coverage overall (SE 10.8%, 0, and 3.1%, respectively). Known
nest locations are generally efficient to sample but nest lists can be problematic since the lists are often not
constructed using unbiased or geographically representative sampling and therefore the lists tend to be biased
in unpredictable ways. A good example of this is Louisiana, where we found no new nests in the Western

71

Gulf Coastal Plain (100% list coverage) but estimated 52% more nests in the Mississippi Alluvial Valley
strata. The State nest list provided a better representation of the numbers of nests in certain areas of the State
than others, however, without the area plot sampling data to reveal the bias, there would have been no way to
account for the difference.
Based on data from many other sources such as the USGS Breeding Bird Survey (BBS, Sauer et al. 2014)
and many State monitoring reports, bald eagle populations continue to thrive in many areas. This survey
was intended to provide a greatly-improved estimate of occupied nests to inform on-going management
assumptions and efforts, but also to serve as an important baseline for comparison with similar estimates from
future monitoring efforts. We estimated a total of 16,048 bald eagle breeding pairs in the coterminous U.S. in
2009. These numbers are higher than previous estimates, but are not out of scale with the levels of population
increase seen with the BBS.

Literature Cited
Gelman, A., J. B. Carlin, H. S. Stern, and D. B. Rubin. 1995. Bayesian Data Analysis,. 1𝑠𝑡 edition. London,
Chapman & Hall.
Givens, G. H., and J. A. Hoeting. 2005. Computational Statistics. John Wiley & Sons, Inc.
Haines, D. E., and K. H. Pollock. 1998. Estimating the number of active and successful bald eagle nests: an
application of the dual-frame method. Environmental and Ecological Statistics 5:245–256.
Kincaid, T. M., and A. R. Olsen. 2013. spsurvey: spatial survey design and analysis. R package version 2.6.
URL http://www.epa.gov/nheerl/arm/.
Link, W. A., and R. J. Barker. 2010. Bayesian Inference with Ecological Applications. Academic Press,
London, United Kingdom.
Plummer, M., 2003. JAGS: A program for analysis of Bayesian graphical models using Gibbs sampling.
Page 10 in K. Hornik, F. Leisch, and A. Zeileis, editors. Proceedings of the 3𝑟𝑑 International Workshop
on Distributed Statistical Computing (DSC 2003). Vienna, Austria.
Sauer, J. R., J. E. Hines, J. E. Fallon, K. L. Pardieck, D. J. Ziolkowski, and W. A. Link. 2014. The North
American breeding bird survey, results and analysis 1966–2012, version 02.19.2014. USGS Patuxent
Wildlife Research Center, Laurel, MD. URL http://www.mbr-pwrc.usgs.gov/bbs/bbs.html.
Steenhof, K., and I. Newton. 2007. Assessing raptor reproductive success and productivity. Pages 184–192
in K. L. Bildstein and D. M. Bird, editors. Raptor research and management techniques. Hancock House,
Surrey, B.C.
Thompson, S. K. 1992. Sampling. John Wiley & Sons, Inc., Hoboken, New Jersey, USA.
U.S. Fish and Wildlife Service. 2009. Post-delisting monitoring plan for the bald eagle (Haliaeetus
leucocephalus) in the contiguous 48 states. U.S. Fish and Wildlife Service, Divisions of Endangered
Species and Migratory Birds and State Programs, Midwest Regional Office, Twin Cities, Minnesota,
USA.
U.S. North American Bird Conservation Initiative Committee. 2000. North American bird conservation
initiative bird conservation regions map. Division of Migratory Bird Management, U.S. Fish and Wildlife
Service, Arlington, Virginia, USA. URL http://www.nabci-us.org.
72

Appendix A4. Updated Golden Eagle Population
Size Estimate in the Western U.S.

Guthrie Zimmerman1 , Brian Millsap1 , Mark Otto1 , and John Sauer2
1 U.S.

2 U.S.

Fish and Wildlife Service
Division of Migratory Bird Management

Geological Survey
Patuxent Wildlife Research Center

Summary
Millsap et al. (2013) derived population size estimates for golden eagles in 12 Bird Conservation Regions
(BCRs) throughout the western US. In their analysis they combined indices from the breeding bird survey
(BBS, Pardieck et al. 2015) and population size estimates from the Western EcoSystems Technology summer
golden eagle survey (WGES, Nielson et al. 2012) where the two surveys overlapped, BCRs 9, 10, 16, and
17 (Figure A4-1), to 1) combine information from the two surveys into composite trend and population
size estimates, and 2) estimate a scaling factor that could adjust BBS indices in non-overlapping BCRs
for estimating a population size in those regions. Millsap et al. (2013) included data from 1967–2010 and
provided population time series for those years, and trend estimates for 1968–2010.
We used the methods described by Millsap et al. (2013) to update golden eagle population sizes throughout
the western U.S. with four additional years (2011–2014) of BBS and WGES data. The WGES has conducted
distance sampling-based surveys since 2006 in the 4 overlap BCRs, except in 2011 when they did not conduct
surveys in BCR 17. Therefore, we did not consider BCR 17 an overlap area in 2011 and adjusted the BBS
index using the BCR 17 scaling factor for that year.
We compared our updated western U.S. time series, and short-term (1990–2014) and long-term (1968–
2014) BCR-specific trends to those estimated by Millsap et al. (2013). In addition, we explored the feasibility
of Flyway-scales as potential units for management. We estimated Central and Pacific Flyway population
estimates by post-stratifying BCR-specific populations into the Flyways based on the proportion of area of
each BCR within each Flyway (Table A4-1). We provide population estimates for each BCR (Table A4-3),
Flyway (Table A4-2), and the entire western U.S. from 1967–2014.
The updated trend estimates for all BCRs were closer to stable than those reported by Millsap et al.
(2013), with the exception of BCRs 32 and 33 which were slightly more negative. The BCR-specific trends
were not statistically different based on the overlap of the 95% credible intervals (Figure A4-2). We detected
a slightly increasing trend in the Central Flyway and slightly decreasing trend in the Pacific Flyway, but these
were not statistically different than a stable population and were not different than the overall estimate for the
entire western U.S. (Figure A4-3). The estimated time series indicated a slightly higher population size in
the Pacific Flyway compared to the Central Flyway, but this difference was not significant (Figure A4-4).
The western U.S. population estimates were similar to those reported in Millsap et al. (2013), except that the
slight increasing population trend they discussed was not evident in the updated time series (Figure A4-5).
We also note the additional years of data do not change the scaling factors for the individual BCRs or the
overall average (Figure A4-6).

Literature Cited
Millsap, B. A., G. S. Zimmerman, J. R. Sauer, R. M. Nielson, M. C. Otto, E. Bjerre, and R. Murphy.
2013. Golden eagle population trends in the western United States: 1968–2010. Journal of Wildlife
Management 77:1436–1448.
73

Nielson, R. M., L. McManus, T. Rintz, and L. L. McDonald. 2012. A survey of golden eagles (Aquila
chrysaetos) in the western US, 2006–2012. Western EcoSystems Technology, Inc., Cheyenne, Wyoming,
USA.
Pardieck, K. L., D. J. Ziolkowski Jr., and M.-A. R. Hudson. 2015. North American breeding bird survey
dataset 1966–2014, version 2014.0. U.S. Geological Survey, Patuxent Wildlife Research Center. URL
http://www.pwrc.usgs.gov/BBS/RawData/.

Table A4-1. Proportion of BCRs in the Pacific
and Central Flyways used to post-stratify
BCR-specific golden eagle population size to a
Flyway scale, 1968–2014.

Proportion of BCR
BCR
5
9
10
11a
15
16
17
18
32
33
34
35
a

Pacific Flyway

Central Flyway

1
1
0.78
0.08
1
0.69
0.01
0
1
1
0.91
0.04

0
0
0.22
0.92
0
0.31
0.99
1
0
0
0.09
0.96

Approximately 24% of BCR occurred in the
Mississippi Flyway; for the purposes of this
analysis, we included that with the Central
Flyway

74

BCR
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

Name

BCR

Aleutian/Bering Sea Islands
Western Alaska
Arctic Plains and Mountains
Northwestern Interior Forest
Northern Pacific Rainforest
Boreal Taiga Plains
Taiga Shield and Hudson Plains
Boreal Softwood Shield
Great Basin
Northern Rockies
Prairie Potholes
Boreal Hardwood Transition
Lower Great Lakes/St. Lawrence Plain
Atlantic Northern Forest
Sierra Nevada
Southern Rockies/Colorado Plateau
Badlands and Prairies
Shortgrass Prairie
Central Mixed-grass Prairie

20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37

Name
Edwards Plateau
Oaks and Prairies
Eastern Tallgrass Prairie
Prairie Hardwood Transition
Central Hardwoods
West Gulf Coastal Plain/Ouachitas
Mississippi Alluvial Valley
Southeastern Coastal Plain
Appalachian Mountains
Piedmont
New England/Mid-Atlantic Coast
Peninsular Florida
Coastal California
Sonoran and Mohave Deserts
Sierra Madre Occidental
Chihuahuan Desert
Tamaulipan Brushlands
Gulf Coastal Prairie

Figure A4-1. North American Bird Conservation Initiative (NABCI) bird conservation regions (BCRs). The horizontal
lines indicate the 4 BCRs where the breeding bird survey (BBS) and Western EcoSystems Technology summer golden
eagle survey (WGES) overlap.

75

Figure A4-2. Updated golden eagle BCR-specific trends compared to those estimated by Millsap et al. (2013). The
BCRs where both breeding bird survey (BBS) and Western EcoSystems Technology summer golden eagle survey
(WGES) data were collected are at the top in bold.

76

Figure A4-3. Comparison of golden eagle trends in the Pacific and Central Flyways and the total trend for the entire
western U.S.

Figure A4-4. Comparison of Flyway-specific time series of golden eagle population size estimates from 1967–2014.

77

Figure A4-5. Comparison of time series for golden eagles in the western U.S. based on data from 1967–2010 (Millsap
et al. 2013) and updated data (1967–2014).

Figure A4-6. Comparison between Millsap et al. (2013) and updated scaling factors used to adjust BBS indices to a
population estimate for golden eagles.

78

Table A4-2. Population size estimates for golden eagles in the
Central and Pacific Flyways, and the total western U.S.,
1967–2014.

Quantiles
Region

Year

Mean

SD

2.50%

50%

97.50%

CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF
CF

1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007

14,272
14,299
14,210
14,188
14,321
14,292
14,293
14,326
14,552
14,585
14,282
14,528
14,621
14,868
14,554
14,344
14,749
14,359
14,723
14,522
14,375
14,697
14,707
14,643
15,048
14,934
15,059
14,723
14,863
14,947
15,093
14,976
15,532
15,044
15,683
15,036
15,254
15,285
15,323
16,529
15,769

3,457
3,392
3,329
3,295
3,249
3,210
3,184
3,141
3,160
3,112
2,989
2,982
3,016
3,050
2,883
2,791
2,861
2,742
2,761
2,686
2,621
2,622
2,613
2,578
2,606
2,577
2,573
2,448
2,460
2,419
2,411
2,432
2,506
2,394
2,517
2,365
2,365
2,412
2,371
2,352
2,401

8,752
8,949
8,925
9,004
9,146
9,106
9,174
9,319
9,540
9,639
9,398
9,711
9,765
10,050
9,866
9,750
10,169
9,852
10,221
9,992
9,967
10,343
10,364
10,383
10,814
10,750
10,893
10,581
10,767
10,865
11,040
10,957
11,385
11,025
11,636
11,020
11,315
11,304
11,383
12,578
11,789

13,869
13,874
13,786
13,795
13,924
13,908
13,944
13,961
14,182
14,218
13,960
14,152
14,244
14,477
14,232
14,037
14,402
14,037
14,415
14,251
14,103
14,416
14,443
14,352
14,743
14,650
14,747
14,480
14,600
14,716
14,829
14,699
15,245
14,811
15,366
14,810
15,001
15,030
15,058
16,319
15,531

22,322
22,211
21,878
21,960
21,785
21,701
21,686
21,616
21,858
21,824
21,168
21,340
21,599
21,917
21,180
20,802
21,276
20,558
21,125
20,629
20,259
20,663
20,595
20,520
21,036
20,846
20,938
20,202
20,469
20,390
20,650
20,475
21,182
20,404
21,454
20,297
20,690
20,720
20,580
21,735
21,250

79

Table A4-2. Population size estimates for golden eagles in the
Central and Pacific Flyways, and the total western U.S.,
1967–2014. (continued)

Quantiles
Region

Year

Mean

SD

2.50%

50%

97.50%

CF
CF
CF
CF
CF
CF
CF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF

2008
2009
2010
2011
2012
2013
2014
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000

15,107
14,582
15,780
16,218
14,629
17,229
15,554
17,541
17,097
17,317
17,191
17,315
17,117
17,227
16,826
17,011
17,128
16,505
16,758
16,506
16,540
16,853
17,045
16,686
15,948
16,039
16,162
16,248
16,220
16,365
16,352
16,683
16,221
17,135
16,692
16,872
16,586
15,617
16,097
16,198
15,916

2,194
2,137
2,324
2,558
2,197
2,904
2,475
2,760
2,621
2,620
2,546
2,533
2,457
2,465
2,336
2,356
2,368
2,229
2,228
2,141
2,129
2,195
2,223
2,106
1,977
1,973
1,946
1,921
1,891
1,927
1,878
1,908
1,825
2,002
1,876
1,983
1,806
1,701
1,764
1,729
1,689

11,465
11,017
11,878
12,080
10,966
12,503
11,523
12,855
12,600
12,846
12,821
12,977
12,831
12,989
12,728
12,953
13,004
12,602
12,881
12,769
12,814
13,017
13,228
13,020
12,478
12,607
12,765
12,905
12,885
13,033
13,053
13,318
12,993
13,630
13,419
13,438
13,412
12,629
13,004
13,120
12,889

14,901
14,363
15,566
15,914
14,391
16,884
15,276
17,285
16,856
17,113
16,982
17,129
16,925
17,016
16,646
16,846
16,952
16,333
16,598
16,371
16,388
16,698
16,863
16,518
15,821
15,905
16,034
16,108
16,070
16,220
16,229
16,529
16,078
16,993
16,546
16,733
16,475
15,488
15,975
16,090
15,829

20,035
19,385
21,034
22,195
19,611
23,871
21,156
23,604
22,840
23,032
22,702
22,839
22,411
22,594
21,844
22,039
22,242
21,259
21,544
20,995
21,067
21,553
21,884
21,378
20,212
20,246
20,426
20,380
20,255
20,471
20,362
20,756
20,078
21,532
20,695
21,137
20,398
19,331
19,905
19,857
19,461

80

Table A4-2. Population size estimates for golden eagles in the
Central and Pacific Flyways, and the total western U.S.,
1967–2014. (continued)

Quantiles
Region

Year

Mean

SD

2.50%

50%

97.50%

PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
PF
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total

2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993

16,375
16,109
16,606
15,728
16,611
15,574
15,825
15,499
15,595
15,886
16,428
16,117
17,023
15,924
31,813
31,396
31,528
31,380
31,637
31,409
31,520
31,152
31,563
31,713
30,787
31,286
31,126
31,408
31,407
31,389
31,435
30,307
30,761
30,684
30,623
30,917
31,072
30,995
31,731
31,155
32,194

1,754
1,724
1,780
1,689
1,842
1,574
1,612
1,573
1,553
1,605
1,660
1,652
1,878
1,759
5,173
5,030
4,962
4,854
4,801
4,748
4,725
4,570
4,612
4,564
4,389
4,373
4,314
4,337
4,255
4,201
4,140
3,984
3,988
3,908
3,817
3,814
3,844
3,776
3,817
3,707
3,814

13,306
13,092
13,502
12,736
13,424
12,775
13,032
12,729
12,904
13,069
13,465
13,184
13,728
12,846
22,947
22,890
23,036
23,049
23,416
23,210
23,376
23,260
23,671
23,796
23,159
23,805
23,774
23,990
24,007
24,163
24,236
23,432
23,867
23,853
23,947
24,279
24,414
24,429
25,176
24,829
25,641

16,242
15,985
16,468
15,599
16,430
15,475
15,720
15,404
15,480
15,765
16,322
16,011
16,879
15,806
31,398
30,976
31,069
30,990
31,179
31,049
31,119
30,772
31,178
31,322
30,435
30,931
30,762
31,025
31,069
31,061
31,057
29,947
30,429
30,354
30,358
30,631
30,790
30,723
31,411
30,832
31,892

20,164
19,725
20,413
19,364
20,732
18,995
19,340
18,876
18,952
19,366
20,003
19,651
21,080
19,737
43,304
42,687
42,287
42,164
42,322
41,831
41,972
41,118
41,662
41,796
40,383
40,746
40,474
40,830
40,617
40,696
40,406
39,148
39,611
39,142
38,776
39,051
39,437
39,077
40,094
39,131
40,519

81

Table A4-2. Population size estimates for golden eagles in the
Central and Pacific Flyways, and the total western U.S.,
1967–2014. (continued)

Quantiles
Region

Year

Mean

SD

2.50%

50%

97.50%

Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total
Total

1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014

31,415
31,735
31,533
30,710
31,073
31,731
30,960
32,058
31,145
31,861
31,013
31,934
32,104
31,594
30,606
30,177
31,666
32,646
30,746
34,252
31,477

3,646
3,732
3,575
3,488
3,549
3,579
3,488
3,597
3,465
3,509
3,481
3,528
3,368
3,396
3,182
3,153
3,343
3,556
3,276
3,998
3,562

25,051
25,149
25,192
24,716
24,950
25,457
24,881
25,932
25,075
25,789
24,973
25,791
26,228
25,769
25,076
24,720
25,832
26,591
25,102
27,382
25,293

31,172
31,447
31,315
30,459
30,814
31,474
30,732
31,757
30,875
31,619
30,705
31,659
31,852
31,307
30,395
29,917
31,418
32,334
30,492
33,910
31,182

39,250
39,750
39,130
38,358
38,699
39,427
38,533
40,106
38,620
39,582
38,702
39,649
39,479
39,147
37,537
37,060
39,067
40,668
37,845
42,956
39,334

82

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014.

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5

1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008

190
217
183
182
182
171
180
170
189
191
179
170
180
201
181
171
172
185
182
165
166
166
171
190
199
198
169
200
171
177
169
202
176
196
196
174
197
179
188
193
207
171

120
155
97
97
95
80
84
76
88
90
79
71
75
90
74
66
66
72
72
62
63
62
64
77
98
84
63
92
63
66
63
91
67
84
84
67
85
72
80
84
103
76

59
69
59
60
61
57
67
59
75
77
72
66
75
91
80
69
72
85
83
64
61
67
69
90
95
95
69
95
69
78
67
91
76
89
90
71
88
73
79
81
86
56

165
182
164
162
163
157
165
156
172
172
165
159
166
181
168
162
163
171
168
156
157
158
161
174
179
180
161
180
162
166
161
183
166
179
179
164
181
167
174
177
185
160

467
593
424
414
411
370
385
355
408
420
372
341
361
424
361
329
326
362
359
312
310
308
322
384
424
412
314
429
316
336
314
431
336
406
402
335
407
355
382
402
445
349

83

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

5
5
5
5
5
5
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9

2009
2010
2011
2012
2013
2014
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002

173
183
184
174
184
195
5,962
5,647
5,988
6,039
6,240
6,072
6,112
6,071
6,051
6,264
5,719
5,938
5,854
5,952
6,193
6,505
6,233
5,626
5,679
5,841
5,865
5,922
6,263
6,134
6,332
6,170
7,068
6,559
7,049
6,220
5,652
6,420
6,304
6,040
6,355
6,387

77
83
89
83
89
97
1,547
1,415
1,464
1,462
1,512
1,397
1,394
1,358
1,376
1,415
1,280
1,291
1,248
1,259
1,320
1,421
1,292
1,164
1,156
1,162
1,146
1,129
1,202
1,138
1,183
1,113
1,383
1,198
1,379
1,096
1,026
1,123
1,090
1,028
1,086
1,081

56
65
64
53
64
68
3,559
3,363
3,672
3,729
3,891
3,842
3,889
3,909
3,850
4,025
3,595
3,816
3,807
3,871
4,071
4,292
4,182
3,637
3,707
3,871
3,921
4,024
4,298
4,215
4,395
4,291
4,865
4,577
4,852
4,368
3,882
4,572
4,446
4,256
4,533
4,523

160
167
168
160
167
175
5,743
5,475
5,806
5,851
6,044
5,908
5,956
5,909
5,873
6,080
5,599
5,786
5,736
5,816
6,046
6,314
6,081
5,525
5,581
5,740
5,759
5,826
6,129
6,023
6,202
6,067
6,890
6,421
6,863
6,121
5,571
6,309
6,207
5,954
6,247
6,282

363
386
395
374
399
438
9,630
8,938
9,418
9,418
9,834
9,308
9,360
9,198
9,252
9,622
8,631
8,882
8,708
8,800
9,210
9,855
9,252
8,179
8,238
8,482
8,463
8,419
9,016
8,671
8,995
8,653
10,308
9,266
10,233
8,673
7,887
8,981
8,748
8,294
8,762
8,780

84

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

9
9
9
9
9
9
9
9
9
9
9
9
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10

2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996

6,703
6,098
7,009
5,599
6,211
5,880
6,112
6,038
6,757
6,426
7,083
6,578
4,955
4,988
4,974
4,986
5,046
5,045
5,097
5,102
5,125
5,145
5,310
5,329
5,339
5,323
5,384
5,373
5,385
5,415
5,410
5,576
5,627
5,643
5,639
5,610
5,876
5,841
5,841
5,708
5,773
5,983

1,153
1,039
1,272
864
984
910
928
918
1,009
980
1,196
1,098
1,346
1,327
1,295
1,292
1,279
1,252
1,251
1,218
1,199
1,196
1,198
1,187
1,164
1,138
1,135
1,106
1,110
1,090
1,075
1,089
1,058
1,056
1,030
1,016
1,058
1,029
1,013
987
972
1,001

4,783
4,323
4,997
4,112
4,559
4,271
4,518
4,462
5,015
4,730
5,064
4,739
2,868
2,888
2,946
2,920
2,985
3,020
3,074
3,107
3,171
3,162
3,335
3,364
3,416
3,444
3,518
3,498
3,513
3,552
3,581
3,737
3,830
3,863
3,889
3,897
4,077
4,100
4,105
4,004
4,096
4,267

6,590
6,011
6,834
5,528
6,112
5,821
6,026
5,955
6,680
6,344
6,961
6,467
4,783
4,819
4,825
4,835
4,914
4,906
4,964
4,973
5,001
5,017
5,194
5,206
5,227
5,214
5,274
5,272
5,271
5,314
5,317
5,467
5,537
5,553
5,537
5,526
5,778
5,758
5,754
5,624
5,696
5,892

9,278
8,393
9,931
7,487
8,394
7,871
8,149
8,062
8,989
8,603
9,746
9,011
8,052
8,034
7,921
7,936
7,974
7,875
7,929
7,818
7,802
7,834
7,980
8,041
7,988
7,829
7,936
7,817
7,847
7,836
7,818
7,994
8,001
7,942
7,925
7,849
8,254
8,092
8,039
7,856
7,855
8,147

85

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

10
10
10
10
10
10
10
10
10
10
10
10
10
10
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11

1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994

5,906
5,751
5,911
6,097
6,152
5,938
6,310
6,154
6,080
6,435
6,265
6,642
6,466
6,870
198
202
205
206
234
219
275
234
243
272
263
273
260
322
299
294
298
330
343
342
357
331
404
362
412
396
503
536

975
948
951
969
981
942
987
968
945
983
963
1,022
974
1,032
124
126
124
123
136
121
166
129
132
148
139
154
126
185
158
145
141
168
160
156
164
151
186
156
185
167
244
262

4,197
4,074
4,270
4,422
4,465
4,267
4,632
4,452
4,414
4,713
4,593
4,872
4,791
5,113
45
46
47
49
67
55
87
62
66
83
77
72
68
110
91
82
85
99
124
118
127
86
163
123
148
130
206
220

5,819
5,683
5,835
6,018
6,064
5,868
6,205
6,071
6,017
6,355
6,203
6,545
6,383
6,789
170
175
178
182
205
195
236
209
219
242
235
244
239
283
269
270
275
300
315
315
328
311
369
340
379
373
449
477

8,045
7,771
7,968
8,170
8,266
7,936
8,453
8,241
8,090
8,574
8,341
8,855
8,569
9,097
515
530
519
505
574
519
680
543
564
641
606
636
556
787
685
642
637
738
734
723
743
671
850
724
861
786
1,141
1,207

86

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
12
13
14
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15

1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2011
2012
2013
2014
2012
2013
2014
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984

568
531
507
497
709
608
564
748
638
773
655
691
631
661
747
760
807
6,575
684
902
840
6,216
6,472
5,683
193
187
182
177
171
164
169
154
160
147
143
140
147
135
129
132
123
121

285
256
214
216
386
279
253
398
310
387
296
308
292
315
386
374
385
1,008
342
441
399
971
1,134
1,003
110
114
102
114
84
76
78
68
70
63
60
57
66
54
49
53
46
44

232
213
202
183
282
249
202
300
235
316
239
268
197
192
243
251
297
4,830
179
325
289
4,544
4,558
3,977
59
58
59
58
59
59
65
58
66
58
56
55
65
59
54
59
52
54

505
479
469
462
607
551
523
653
581
685
602
632
587
613
677
690
735
6,491
630
816
768
6,138
6,362
5,599
169
165
161
158
155
149
154
141
147
136
133
130
135
127
121
123
117
114

1,340
1,172
1,027
1,010
1,762
1,320
1,174
1,782
1,388
1,730
1,382
1,439
1,321
1,391
1,674
1,649
1,742
8,775
1,491
2,013
1,812
8,363
9,001
7,924
457
436
421
396
374
346
364
321
326
301
288
277
300
258
246
256
230
225

87

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
15
16
16
16
16
16
16
16
16
16
16
16
16

1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978

133
116
122
110
121
109
103
101
99
98
95
98
93
96
94
86
90
83
81
82
78
83
78
77
76
73
78
70
72
70
6,311
6,218
6,067
6,007
5,834
5,944
5,896
5,658
5,729
5,710
5,480
5,677

60
43
48
39
50
39
38
36
36
37
35
39
36
39
38
35
38
35
35
37
35
41
37
38
37
37
43
38
39
39
1,955
1,880
1,804
1,744
1,680
1,696
1,648
1,547
1,535
1,523
1,427
1,450

62
51
58
48
58
52
45
44
43
43
40
44
41
43
41
34
38
31
30
30
27
31
27
25
25
23
25
20
21
20
3,349
3,362
3,299
3,317
3,239
3,366
3,373
3,212
3,341
3,366
3,221
3,424

121
110
113
105
111
103
98
96
94
92
90
92
88
89
87
81
83
77
76
76
72
75
71
70
68
65
69
62
63
62
6,046
5,946
5,819
5,755
5,605
5,715
5,667
5,468
5,530
5,501
5,299
5,473

270
216
238
201
235
201
190
187
183
182
176
188
179
189
184
171
183
167
166
172
165
182
169
171
170
167
185
167
170
170
10,818
10,671
10,210
10,055
9,710
9,850
9,697
9,172
9,248
9,172
8,672
9,021

88

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
16
17
17
17
17
17
17

1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
1967
1968
1969
1970
1971
1972

5,419
5,372
5,391
5,294
5,236
5,081
5,106
5,087
4,932
5,006
4,688
4,870
4,883
4,597
4,598
4,744
4,374
4,837
4,559
4,282
4,377
4,214
4,238
4,120
4,042
3,926
4,031
4,069
3,787
3,471
3,556
3,612
3,580
4,129
4,200
4,223
9,060
9,075
9,045
9,143
9,131
9,126

1,355
1,334
1,312
1,259
1,225
1,168
1,149
1,139
1,057
1,068
995
1,006
993
902
882
919
832
949
844
780
774
742
739
724
697
681
671
716
627
566
566
580
551
690
794
762
3,080
3,002
2,966
2,944
2,880
2,838

3,269
3,240
3,315
3,281
3,263
3,196
3,263
3,269
3,211
3,275
3,031
3,228
3,298
3,073
3,142
3,252
2,968
3,365
3,180
2,946
3,087
2,954
2,982
2,890
2,826
2,722
2,887
2,874
2,745
2,516
2,598
2,624
2,641
2,955
2,934
2,969
4,387
4,497
4,509
4,712
4,704
4,706

5,246
5,208
5,231
5,130
5,102
4,955
4,966
4,945
4,810
4,870
4,591
4,744
4,762
4,504
4,503
4,636
4,287
4,707
4,460
4,211
4,293
4,147
4,163
4,055
3,981
3,876
3,966
3,992
3,725
3,417
3,506
3,559
3,535
4,073
4,098
4,146
8,610
8,621
8,634
8,727
8,739
8,740

8,501
8,443
8,464
8,209
8,070
7,671
7,746
7,690
7,307
7,467
6,919
7,216
7,117
6,622
6,581
6,857
6,216
7,117
6,450
5,992
6,078
5,854
5,864
5,740
5,586
5,398
5,534
5,673
5,173
4,737
4,817
4,876
4,800
5,632
6,087
5,924
16,393
16,106
15,987
16,045
15,775
15,741

89

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17
17

1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014

9,129
9,235
9,459
9,472
9,200
9,415
9,541
9,803
9,390
9,161
9,693
9,312
9,546
9,280
9,212
9,570
9,526
9,577
9,848
9,863
9,924
9,511
9,665
9,537
9,716
9,606
10,075
9,655
10,311
9,655
9,822
9,915
10,012
10,968
10,406
9,721
9,066
10,133
10,624
9,113
11,437
10,012

2,816
2,789
2,810
2,757
2,623
2,638
2,670
2,707
2,516
2,415
2,509
2,388
2,394
2,305
2,252
2,249
2,218
2,211
2,232
2,218
2,201
2,063
2,061
2,021
2,007
1,994
2,099
1,968
2,134
1,941
1,946
1,982
1,976
1,945
2,009
1,775
1,677
1,875
2,138
1,755
2,491
2,030

4,784
4,910
5,110
5,309
5,052
5,296
5,366
5,680
5,399
5,297
5,804
5,416
5,780
5,506
5,496
5,944
5,922
6,037
6,318
6,353
6,450
6,105
6,302
6,211
6,377
6,303
6,711
6,336
7,004
6,362
6,572
6,623
6,755
7,728
7,125
6,807
6,283
7,001
7,241
6,266
7,469
6,755

8,749
8,838
9,088
9,125
8,865
9,055
9,170
9,425
9,079
8,874
9,357
9,025
9,233
9,013
8,968
9,312
9,271
9,302
9,555
9,586
9,638
9,275
9,448
9,330
9,508
9,385
9,816
9,440
10,034
9,466
9,605
9,689
9,780
10,765
10,202
9,518
8,877
9,943
10,329
8,900
11,103
9,764

15,708
15,790
16,056
16,016
15,295
15,545
15,696
16,033
15,230
14,820
15,535
14,790
15,166
14,551
14,258
14,721
14,632
14,627
15,088
14,955
15,041
14,128
14,389
14,102
14,307
14,044
15,005
14,118
15,362
14,053
14,323
14,447
14,494
15,384
15,118
13,728
12,879
14,403
15,633
13,183
17,214
14,675

90

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18
18

1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008

1,170
1,151
1,166
1,148
1,212
1,215
1,206
1,164
1,239
1,198
1,229
1,216
1,248
1,214
1,257
1,377
1,197
1,235
1,326
1,338
1,234
1,268
1,403
1,296
1,254
1,313
1,248
1,333
1,374
1,393
1,386
1,350
1,483
1,437
1,499
1,417
1,442
1,358
1,458
1,456
1,492
1,377

470
442
434
421
439
435
420
402
416
390
395
379
383
355
361
440
343
342
376
362
327
330
373
332
322
324
315
318
335
341
337
330
375
364
392
358
367
359
376
382
397
381

501
506
527
524
573
581
588
562
620
610
626
638
661
648
701
764
626
675
759
791
672
730
845
752
701
783
684
801
848
858
855
793
917
865
906
834
852
731
858
834
852
719

1,093
1,080
1,096
1,085
1,141
1,146
1,138
1,104
1,170
1,139
1,175
1,165
1,192
1,166
1,207
1,298
1,157
1,196
1,270
1,284
1,205
1,230
1,341
1,257
1,223
1,275
1,222
1,298
1,331
1,348
1,342
1,315
1,426
1,387
1,438
1,374
1,398
1,328
1,410
1,415
1,444
1,344

2,303
2,208
2,194
2,137
2,274
2,252
2,207
2,103
2,217
2,107
2,166
2,091
2,184
2,037
2,102
2,427
1,965
2,002
2,213
2,204
1,966
2,032
2,291
2,050
1,952
2,046
1,932
2,059
2,150
2,185
2,155
2,089
2,368
2,296
2,434
2,244
2,283
2,150
2,346
2,339
2,411
2,210

91

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

18
18
18
18
18
18
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32

2009
2010
2011
2012
2013
2014
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002

1,598
1,561
1,645
1,444
1,501
1,502
934
913
927
918
874
894
905
848
911
863
878
872
852
891
848
862
807
840
837
819
831
796
815
817
792
757
817
799
768
809
772
750
772
767
779
766

466
436
500
423
441
450
269
257
252
244
228
227
225
210
222
205
207
204
197
210
187
190
179
184
185
174
179
167
173
180
167
164
178
170
164
180
169
165
171
173
180
178

908
876
917
730
789
772
513
500
525
529
491
526
539
489
559
523
539
540
526
556
534
552
497
539
531
524
530
505
531
526
510
464
530
521
490
523
494
470
491
480
490
476

1,529
1,500
1,564
1,398
1,449
1,449
901
884
899
890
853
869
881
830
882
840
854
849
833
865
829
839
791
821
816
801
811
781
798
797
776
748
795
780
753
786
756
737
754
750
759
745

2,701
2,580
2,850
2,401
2,502
2,517
1,551
1,499
1,506
1,473
1,388
1,417
1,416
1,312
1,423
1,340
1,346
1,327
1,287
1,380
1,263
1,309
1,199
1,263
1,262
1,208
1,236
1,168
1,209
1,225
1,166
1,108
1,233
1,192
1,136
1,228
1,162
1,116
1,158
1,167
1,196
1,166

92

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

32
32
32
32
32
32
32
32
32
32
32
32
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33

2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996

784
737
748
742
734
700
715
711
728
715
699
719
947
911
925
841
888
810
823
785
783
805
777
723
745
692
698
707
772
663
696
622
671
656
628
635
592
562
584
615
604
598

191
176
182
183
185
178
185
189
196
193
194
205
538
492
492
418
440
380
392
358
343
353
328
304
308
275
278
274
352
245
271
226
251
240
221
227
202
196
202
226
223
216

482
445
451
443
433
401
412
401
419
402
386
401
287
280
308
277
315
284
302
285
304
321
320
271
307
272
293
309
356
283
325
263
324
313
299
310
282
243
284
307
297
292

760
720
727
722
712
684
695
692
702
692
675
690
826
809
822
761
798
742
749
724
724
739
720
676
694
652
654
662
703
628
646
593
626
616
594
596
563
534
554
575
563
560

1,229
1,140
1,168
1,161
1,156
1,099
1,141
1,147
1,189
1,154
1,156
1,204
2,285
2,133
2,159
1,893
1,985
1,771
1,769
1,671
1,623
1,672
1,574
1,460
1,508
1,357
1,365
1,367
1,614
1,243
1,362
1,157
1,282
1,245
1,150
1,186
1,071
1,020
1,053
1,173
1,148
1,129

93

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
33
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34

1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990

533
555
568
507
527
512
488
496
474
509
494
485
451
458
449
449
439
435
1,050
987
994
945
938
909
931
843
902
831
814
762
743
719
806
744
670
695
674
644
708
614
632
633

185
200
219
189
202
198
192
198
192
223
214
217
204
214
212
221
219
219
789
689
711
633
608
585
600
487
537
464
431
399
380
352
450
379
305
317
303
277
352
250
261
275

247
260
266
208
233
219
199
208
181
210
202
187
160
165
158
153
141
138
216
212
230
209
232
214
254
227
264
251
262
233
225
236
288
264
222
267
260
245
289
243
261
269

506
522
530
482
495
480
460
462
446
466
453
443
417
418
410
405
396
391
848
819
814
794
790
775
793
740
779
734
723
684
669
656
708
670
619
639
621
598
635
574
587
579

977
1,041
1,095
950
1,006
992
938
974
930
1,039
1,027
1,021
937
979
962
979
989
985
3,030
2,793
2,790
2,599
2,493
2,405
2,428
2,080
2,286
2,030
1,911
1,750
1,704
1,581
1,903
1,651
1,424
1,459
1,410
1,332
1,568
1,208
1,301
1,306

94

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
34
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35

1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984

600
599
547
580
541
560
519
500
513
524
579
576
503
476
447
462
455
463
427
414
476
445
418
418
843
900
872
787
887
840
795
887
770
815
795
771
800
782
831
768
848
805

245
237
221
228
202
227
195
200
205
210
282
304
210
209
196
206
209
221
205
204
250
239
226
230
433
480
409
355
418
380
345
386
313
328
320
301
310
294
310
284
323
297

252
263
216
262
230
246
217
187
215
229
246
239
209
184
151
173
168
169
141
132
164
144
124
123
284
322
321
274
347
321
290
368
284
330
325
303
337
325
374
321
388
363

559
559
518
542
514
523
490
475
482
488
517
509
466
440
418
428
417
421
391
378
425
392
372
369
763
803
794
726
807
768
739
814
725
763
744
727
757
742
784
732
797
762

1,201
1,170
1,034
1,135
1,026
1,093
970
957
1,006
1,034
1,285
1,329
1,011
984
909
941
960
1,000
916
896
1,082
1,050
981
983
1,853
2,037
1,875
1,620
1,901
1,739
1,612
1,827
1,505
1,614
1,551
1,471
1,494
1,469
1,584
1,429
1,616
1,493

95

Table A4-3. Population size estimates of golden eagles in each
BCR in the western U.S., 1967–2014. (continued)

Quantiles
BCR

Year

Mean

SD

2.50%

50%

97.50%

35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35
35

1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014

830
855
898
834
783
763
839
759
796
733
753
789
899
1,064
750
827
768
767
849
819
753
897
834
957
790
855
744
882
846
802

294
322
341
303
265
269
295
250
264
258
249
260
324
484
251
283
262
263
308
292
265
333
306
403
300
334
300
348
338
323

386
408
444
403
353
326
410
335
373
275
329
371
461
521
318
397
332
335
412
375
314
435
386
459
336
386
271
396
366
326

782
800
831
788
752
733
791
734
763
710
729
756
839
940
723
784
738
736
797
775
726
836
783
872
745
796
703
819
787
748

1,535
1,634
1,751
1,542
1,396
1,397
1,547
1,324
1,421
1,300
1,312
1,375
1,675
2,352
1,329
1,478
1,363
1,370
1,593
1,498
1,362
1,727
1,559
1,963
1,494
1,655
1,442
1,728
1,644
1,586

96

Appendix A5. USFWS Policy Update to the
Use of Eagle Natal Dispersal Distances in
Permitting Decisions
A recommendation from the U.S. Fish and Wildlife Service’s1 Eagle Technical Assessment Team for
consideration by the U.S. Fish and Wildlife Service’s Eagle Management Team
1

U.S. Fish and Wildlife Service, Division of Migratory Bird Management

Background
The Service’s approach to managing eagle populations, as outlined in its 2009 Eagle Permit Rule Final
Environmental Assessment (hereafter, 2009 FEA), is to do so primarily at broad, regional eagle management
unit (EMU) scales. EMUs for golden eagles are Bird Conservation Regions. For bald eagles, EMUs are
based on nest densities and correspond with Service Region boundaries, with slight modifications. However,
the Service also is concerned with possible declines in eagle breeding populations at smaller geographic
scales. Within a stable regional population there could be a local area where substantial mortality or decreased
productivity develops. Initially, there may be enough floaters in this local area population (LAP) to fill
breeding area vacancies and compensate for a loss of recruits, but as losses accelerate some vacancies may
not be not filled. Thus, the potential “rescue effect” attributed to eagles recruiting into the LAP by returning
to breed at sites near their natal areas probably is diminished, i.e., probability of recruiting likely decreases
with increasing distance from the center of the LAP. In this way, the population is functionally closed to
ingress of individuals. This potential depression in resiliency of LAPs is of concern, so the Service uses a
biological basis for defining LAPs. Specifically, the Service uses metrics based on natal dispersal distance
(NDD), defined as the linear gap between a bird’s location of origin and its first breeding or potential breeding
location (Greenwood and Harvey 1982, Annual Review of Ecology, Evolution, and Systematics 13:1–21).
An NDD metric is used as a buffer distance from a given site or project of concern due to potential for take
of eagles; the landscape encompassed by the buffer thus represents the LAP area associated with the site or
project. Determination of LAP size and benchmark levels is described in Appendix F of the Service’s 2012
Eagle Conservation Plan Guidance, Module 1–Land-based Wind Energy.
Results reported in the recent paper, Natal Dispersal Distance of Bald and Golden Eagles Originating in
the Coterminous United States as Inferred from Band Encounters, by Millsap et al. (2014, Journal of Raptor
Research 48:13–23) should prompt the Service to update the NDD criteria currently used for determining
geographic boundaries and estimating size of LAPs of golden eagles and bald eagles. Using a refined dataset
and improved analytical methods, the authors re-analyzed NDDs of golden eagles and bald eagles based on
band recoveries. The key statement in Millsap et al. (2014) on implications of their analysis of natal dispersal
of each eagle species follows:
“Depending on the management policy and circumstances, choice of a natal dispersal value
in the range of the 50𝑡ℎ to 90𝑡ℎ quantile of the distribution as an effect-distance for breeding
populations of both species of eagle seems reasonable. For Golden Eagles, this range is 46–175
km [29–109 mi], and for Bald Eagles, 69–346 km [43–215 mi]. This range includes the natal
dispersal value of 69 km [42 mi] currently in use for Bald Eagles, but the 90𝑡ℎ quantile for
Golden Eagles is slightly less than the 225 km [140 mi] currently in use for this species (U.S. Fish
and Wildlife Service 2013).”

97

Of additional importance in the updated analysis by Millsap et al. (2014) is documentation of a difference
in NDD between sexes of bald eagles, with females having significantly longer NDDs than males. This is
typical across Class Aves and is likely the case for golden eagles as well, though there were not enough
recoveries of bands from known-sex golden eagles for the authors to conduct an analysis of sex-biased NDD
for the species.
Another issue should be addressed in this update. During this review, an inconsistency was discovered
in the 2009 FEA with respect to the values used for NDDs of bald eagles and golden eagles. Of note in the
2009 FEA’s “Definitions and Interpretations Used. . . ” is this: “Natal dispersal distance–extent of movement
between the place of birth and place of first breeding”. The following, excerpted directly from the 2009 FEA,
demonstrate the inconsistent application of NDD:
(p. 24) “We used natal populations (eagles within the median [emphasis added] natal dispersal
range of each other) in our evaluation in order to look at distribution across the landscape”
. . . which is consistent with this for bald eagles:
(p. 24) “We used natal populations (eagles within the median natal dispersal range of each other)
in our evaluation. . . ”
. . . but not consistent with this for golden eagles:
(p. 26) “. . . 90% of mature golden eagles reencountered during the breeding season were within
140 miles [emphasis added] of their natal site. We will consider the natal dispersal distance of
golden eagles when evaluating effects to local area populations.”
then, finally. . .
(p. 37) “For overall permit management, we will consider local area population effects within the
species specific natal dispersal distances (43 miles for bald eagles, 140 miles for golden eagles).”
Per the above passages from pages 26 and 37, the intent of the 2009 FEA was to use the 90𝑡ℎ quantile
value from the (normal) distribution of NDD records available for golden eagles. Reference to the point
value used for NDD as a “median” from page 24 was incorrect for this species. The intent of the 2009
FEA was to also use the 90𝑡ℎ quantile for bald eagles (not reflected in the above excerpts) and thus address
NDD consistently between species. Unfortunately, the median output from a related analysis was shifted
inadvertently to the text for bald eagles in the Draft Environmental Assessment stage and the discrepancy was
overlooked during completion of the 2009 FEA.

Decision Point
Technical Issue
Technical issues to be considered are the updated distributions of NDDs for each eagle species, and the
finding that natal dispersal differs by sex for at least the bald eagle.1 The latter is not an insignificant issue;
for bald eagles, the current practice of pooling data from both sexes to estimate the median NDD ignores
72% of the distance over which female recruits would originate, per the update by Millsap et al. (2014) Use
of either the median of the distribution of female NDDs (Table A5-1, option 4) or the 90𝑡ℎ quantile of the
1

Important note: NDD criteria considered herein and their implications for estimating LAP size in decisions regarding “benchmark”
levels of take of golden eagles and bald eagles are summarized, using a hypothetical example, in Table A5-1.

98

Table A5-1. Examples of bald and golden eagle natal dispersal distance criteria and the implications for take
benchmarks.

Species
Golden eagle

Bald eagle

NDD metrica

NDD
LAP
LAP size
5% benchmark
value (mi) area (mi2 )b (n eagles, example)c
(n eagles)

current 2009 FEA
pooled 90𝑡ℎ
quantile
(option 1) revised–pooled
median
( 2) revised–pooled
90𝑡ℎ quantile

140

61,575

1,000

50

29

2, 642

43

2

109

37, 325

606

30

current 2009 FEA
pooled median
(option 1) revised–pooled
median
( 2) revised–pooled
90𝑡ℎ quantile
( 3) revised–male,
mediand
( 4) revised–female,
median
( 5) revised–male,
90𝑡ℎ quantilee
( 6) revised–female,
90𝑡ℎ quantilee

43

5,809

500

25

42

5, 542

477

24

215

145, 220

12, 497

625

37

4, 301

370

19

86

23, 235

2, 000

100

200

125, 664

10, 814

541

419

551, 541

47, 465

2, 373

a

Currently used values are derived from a normal distribution. Revised values are from lognormal distributions in Millsap et al.
(2014).
b
Area calculation is based on a circle with the respective NDD value as its radius, extending from a central point instead of from a
polygon as normally would be the case.
c
LAP size is based on simple extrapolation of density estimates to Eagle Management Units, as described in the 2009 FEA: (1) for
golden eagles, density estimates for Bird Conservation Regions were from Partners In Flight; (2) for bald eagles, density
estimates for Service Regions were derived from minimum number of occupied breeding areas at the time of delisting. Because
this example is hypothetical, no EMUs are identified.
d
The sample of bald eagle band recoveries used by Millsap et al. (2014) was large enough to assess NDD for each sex. This was
not the case for golden eagles, however.
e
Sex-specific, 90𝑡ℎ quantile NDD values for bald eagles were not directly conveyed in Millsap et al. (2014), but were calculated
by Brian A. Millsap for this document based on the dataset and approach used in the publication.

pooled distribution of NDDs (option 2) would capture more of this area, resulting in a LAP size about 4 and
25 times greater, respectively, than currently used (column 5). The median of the distribution of male NDDs
(option 3) would capture slightly less than the area currently used.
Under Service policy for benchmark guidelines authorizing take of up to 1% and 5% of the size of a given
LAP, an increase in LAP size allows for correspondingly greater levels of take under permit. An increase in
authorized take of bald eagles seems appropriate, given the species’ current robust population status across
its range in the U.S. When the median distribution of the female NDD is used, a 4-fold increase in take
authorized at the 5% benchmark levels seems reasonable (example in Table A5-1, column 6: increase from
current 25 to 100 eagles under option 4). However, a 25-fold increase resulting from use of a 90𝑡ℎ quantile of
99

the pooled distribution of NDDs (column 6: increase from current 29 to 625 under option 2) seems excessive,
at least until uncertainty surrounding NDD (Millsap et al. 2014), including regional variation, is reduced.
The 90𝑡ℎ quantile NDD value for male bald eagles is nearly as excessive, while that for females is far more
excessive (options 5 and 6).
Again, for golden eagles, current data are too few to discern whether NDD differs by sex, though a
difference likely occurs. Use of the pooled 90𝑡ℎ quantile is consistent with the Service’s current approach for
the golden eagle. Moreover, the smaller LAP area, population size, and 5% benchmark level that result (40%
decrease in example in Table 1, option 2, columns 4–6) seem appropriate, given the species’ more tentative
population status. However, this status does not warrant use of the median (option 1), which could lead to
almost no take at the 5% benchmark level.
Inconsistent Terms Issue
In current LAP analyses, the Service uses the median and the 90th quantile of the distribution of NDDs
for bald eagles and golden eagles, respectively, per the 2009 FEA. In both cases, data from sexes are pooled.
As noted above, this stems from an undetected error in the 2009 FEA; the intent at the time was to use the
90𝑡ℎ quantile of the distribution of NDD data, pooled from both sexes, for each species.

Recommendations
There are four reasonable alternatives for addressing these two issues:
Alternative 1a (ETAT-recommended alternative)
Acknowledge and reconcile errors in use of terms in the 2009 FEA; the original intent was to use the
90𝑡ℎ quantile of the distribution of NDDs for both species. Be consistent with the 2009 FEA’s approach to
NDD criteria for golden eagles by using the updated 90𝑡ℎ quantile NDD value. For bald eagles, be reasonably
consistent by using the updated median NDD value but base this on data from females to better account for
their significantly greater NDDs. This approach is recommended because it would incorporate new evidence
of sex-biased NDD of bald eagles reported by Millsap et al. (2014), acknowledging that use of a pooled
value is no longer consistent with best available science for this species. Use of the females’ greater NDD
is recommended because it is a more liberal approach, appropriate given the species’ current robust status;
use of the males’ NDD would, in contrast, result in reduced levels of authorized take. Under this alternative,
the Service would complete this policy revision at the earliest possible date to incorporate the best and most
recent information in decision-making.
Alternative 1b
Same as Alternative 1a, except inconsistent terminology in the 2009 FEA would be reconciled during the
NEPA process associated with the ongoing Rule revision rather than as an earlier policy decision.
Alternative 2a
Acknowledge and reconcile errors in use of terms in the 2009 FEA, just as in Alternative 1a; again, the
original intent was to use the 90𝑡ℎ quantile of the distribution of NDDs for both species. Adopt as the updated
NDD criterion the 90𝑡ℎ quantile values for the pooled NDD distributions for both of the respective species,
reported by Millsap et al. (2014). However, the bald eagle’s new NDD under this alternative would result in an
excessively large local area size, exceeding the area encompassed by regional EMUs in some cases. Moreover,
100

this approach would not incorporate new evidence of sex-biased NDD of bald eagles reported by Millsap
et al. (2014) and thus would fail to incorporate the latest, best available information. Regardless, under this
alternative, the Service would complete this policy revision at the earliest possible date to incorporate the best
and most recent information in decision-making.
Alternative 2b
Same as Alternative 2a, except inconsistent terminology in the 2009 FEA would be reconciled during the
NEPA process associated with the ongoing Rule revision rather than as an earlier policy decision.

Decision
The EMT supported the ETAT-recommended alternative (1a) and the process to update the ECPG
appendix is underway. This approach to updating the NDD will acknowledge and reconcile errors the errors
in the 2009 FEA, it will be consistent with the 2009 FEA’s approach to NDD criteria for golden eagles by
using the updated 90𝑡ℎ quantile NDD value, it will be reasonably consistent for bald eagles by using the
updated median NDD value for females (thereby incorporating the new evidence of sex-biased NDD of bald
eagles), it is a more liberal approach for bald eagles which is appropriate given the species’ current robust
status, and a slightly more conservative approach with golden eagles as the revised NDD would result in
reduced levels of authorized take. Under this alternative, the Service will complete this policy revision at the
earliest possible date by revising the ECPG (U.S. Fish and Wildlife Service 2013, Appendix F) to reflect the
incorporation of the best and most recent information in decision-making.

Literature Cited
Greenwood, P. J., and P. H. Harvey. 1982. The natal dispersal of breeding birds. Annual Review of Ecology,
Evolution, and Systematics 1:1–21.
Millsap, B. A., A. R. Harmata, D. W. Stahlecker, and D. G. Mikesic. 2014. Natal dispersal distance of bald
and golden eagles originating in the coterminous United States as inferred from band encounters. Journal
of Raptor Research 48:13–23.
U.S. Fish and Wildlife Service. 2013. Eagle conservation plan guidance. Module 1–land-based wind energy.
Version 2. Division of Migratory Bird Management, Washington, DC, USA. URL http://www.fws.
gov/migratorybirds/pdf/management/eagleconservationplanguidance.pdf.

101

U.S. Fish & Wildlife Service Headquarters
5275 Leesburg Pike, MS: MB
Falls Church, VA 22041-3803
U.S. Fish & Wildlife Service
http://www.fws.gov

Migratory Bird Program
http://www.fws.gov/birds

For state transfer relay service
TTY/Voice: 711