.7:

L? 5.,

if“.
k. ‘ ”.3”, n

. L”
. wk”.
.14.:
Reciti v
. n? .

 

.c:

 

 

u!

. F r .,s.%..w.. g”...
.3. .. . 3 .
. .54321
. 2m;
u 4 ‘ ,
..

y a 017%
v9... 3&1?“
mm. fix a»?
fimmawuflflx ,

am ‘

c.
{v
.

.
a
i :53“.
t fay: ‘
, .
. . .

._ .11. .1
.7 :1:

. «33
z..- in 4....
t. 53%... 62...

W

 

m.

3%;
9};

V 1%:
:2”;

 

alit Ill-hue.»
, .. A... - .. .
r ‘ ‘

 

“WSW
Z

200?
5394—23”?

[Date

This is to certify that the

thesis entitled

Land Use/Land Cover Patterns of Bald Eagle
(Haliaeetus leucocephalus) Nesting Habitat
in the Northern Lower Peninsula of Michigan

 

presented by

Carrie Rose Steen

has been accepted towards fulfillment
of the requirements for

 

 

 

 

Master's degnmin Fisheries & Wildlife
'\
/ Major pro essor
May 6, 2003

 

0-7639

MS U is an Affirmative Action/Equal Opportunity Institution

 

 

Michigan State

LIBRARY

University

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

rd

 

i, {)1

b .. .-_~,._ 34.
iificxszmt

’3’]
Lo

 

 

 

 

 

 

 

 

 

 

 

 

6/01 C'JCIRC/DateDue.p65-p.15

 

LAND USE / LAND COVER PATTERNS OF BALD EAGLE (Haliaeetus
Ieucocephalus) NESTING HABITAT IN THE NORTHERN LOWER PENINSULA OF
MICHIGAN
By

Carrie Rose Steen

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE
Department of Fisheries and Wildlife

2003

ABSTRACT
LAND USE / LAND COVER PATTERNS OF BALD EAGLE (Haliaeetus
Ieucocephalus) NESTING HABITAT IN THE NORTHERN LOWER PENINSULA OF
MICHIGAN
By

Carrie Rose Steen

A GIS was used to determine if a pattern in bald eagle nest areas and reproduction
exists in relation to land use / land cover patterns in Michigan. Habitat analyses were
conducted at three level: nests sites in the Black River Watershed, breeding areas in the
northern lower peninsula and nest sites in the northern lower peninsula (NLP) of
Michigan. I examined the change in land cover over time, proportion of land cover types
in nest and breeding areas and the relationship between nesting and production associated
with surrounding land cover. I used 1993 GAP land cover data to develop a habitat
model for nests in the NLP. Over all scales, bald eagles preferred land cover that
included lowland forested areas, and avoided land cover that included urban and non-
forested areas. Logistic regression models demonstrated distance to water as the most
important variable in nest location. Results suggest that proportions of land cover types
among nests is highly variable, but general nest site preferences do not seem to be
changing. As the eagle population continues to expand, the availability of optimal
breeding areas will decrease, and breeding pairs settling in sub-optimal areas may
experience lower reproductive rates. Results of the analyses and methods that were
developed should provide information useful in identifying future potential habitat for

nesting bald eagles in Michigan.

But you never were made, as I,
On the wings of the wind to fly!
The eagle said.

Will Carleton

To all those who kept me grounded

iii

ACKNOWLEDGMENTS

This project resulted from an ongoing program conducted since 1961 to monitor
the Michigan bald eagle population. It is impossible to name everyone who has been
involved over the years, but the program has largely been a cooperative effort of
biologists at the Michigan Department of Natural Resources Wildlife Division, United
States Fish and Wildlife Service, Michigan State University and Clemson University.
This particular study was primarily funded by the Michigan Natural Heritage Program
through contributions to the Nongame Wildlife Fund. Supplemental funding was
provided by the George J. Wallace and Martha C. Wallace Endowed Scholarship Fund.

I wish to thank my committee members, Dr. Brian Maurer, Dr. Catherine Lindell
and Dr. Bill Bowerman. I appreciate the guidance and assistance they provided for
incorporating landscape scale methods and for keeping me on track with statistics. Dr.
Bowerman, through his years of research on Michigan bald eagles, has added valuable
insight to the dynamics of the Michigan population.

Ray Rustem of the MDNR Wildlife Division is responsible for initiating this
study and has always been very supportive and helpful in directing me to data sources
and providing any other assistance needed. And he never asked what took so long.

Dave Best of the USFWS has been invaluable in providing necessary data,
helping to correct the dataset, providing insight of the data and the monitoring program,
as well as the birds themselves. He was always cooperative and quick to answer my
numerous questions, no matter how much I pestered him. I appreciate his support and

willingness to work with me in providing any information and assistance requested.

iv

Sergej Postupalsky has been involved with the monitoring program from the
beginning. If any one person has been responsible for collecting a large portion of nest
location and production data, and compiling productivity reports and population trends, it
would have to be him. He has been very helpfirl and supportive, always willing to fill in
the blanks and provide knowledge that comes only through experience.

Thanks to the MDNR Wildlife Division, especially Marshall Strong and Kevin
Gardner for their help with GIS and land cover data; and Steve Beyer and Arnie Karr for
letting me tag along on aerial surveys. The staff of the Department of Fisheries and
Wildlife also deserves my thanks for keeping things on track, particularly Jane Thompson
and Sarah Cline. And I’m completely indebted to Dr. Mike Jones for always coming to
the rescue and making the computer do what I wanted it to.

Team Millenbah has my undying gratitude for their friendship, technical
assistance and willingness to laugh at anything. And to my Rose Lake friends, I couldn’t
have done this without the encouragement and necessary diversion that you provided. A
special thanks to Al Stewart for convincing me that going back to school would be a good
idea, and then starting the wheels turning.

A few friends in particular deserve my sincerest appreciation. I have to thank
Steph Hogle and Earl Flegler for listening to anything, laughing at stupid jokes,
reminding me to graduate and keeping me sane in general. Heather Seymour — wish I
could say more, but she is the definition of a best friend and I can’t possibly thank her
enough.

Thanks to all my family and friends who have supported and believed in me, and

leave no doubt that they will always do so.

Saving the best for last, to my advisor, Dr. Kelly Millenbah, for knowing when to
let me work and when to make me work. She has been irreplaceable in career guidance
and moral support, an excellent teacher and mentor. This would not have been possible if
she hadn’t taken me on and stuck with me as long it took. Thanks Kelly.

And finally, my family. To my brother, Ricky, for being — well — my little
brother. His music, through the stereo and the phone, accompanied many hours at the
computer, and made the distance a little shorter. My parents, who gave me the
confidence and faith to always pursue my goals, whatever the setbacks. For the constant
support in every way possible, no matter how big the sacrifice, they are the ultimate
reason for this accomplishment. When things got overwhelming, I could think of what
Mom said, “you can always come home”; and what dad said, “keep your nose to the
grindstone, and all you get’s a flat nose.” Both in their own ways kept me going. Love
you.

And to Joshua. For the humility, perspective and companionship that can only be

imparted by a furry, four-legged fiiend.

vi

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. ix

LIST OF FIGURES ........................................................................................................... xi

CHAPTER 1

SPECIES DESCRIPTION AND HISTORICAL TRENDS ................................................ 1
Natural History and Distribution ...................................................................................... 1
History and Status in the United States ............................................................................. 2
History and Status in Michigan ......................................................................................... 4
Key to Abbreviations and Definitions .............................................................................. 8

LITERATURE CITED ...................................................................................................... 10

CHAPTER 2

LAND USE / LAND COVER PATTERNS OF BALD EAGLE NEST LOCATION,
ACTIVITY AND SUCCESS IN THE BLACK RIVER WATERSHED, MICHIGAN ...12

Introduction ..................................................................................................................... 12
Study Area ...................................................................................................................... 13
Methods ........................................................................................................................... 15
Data Preparation ........................................................................................................... 15
GIS and Statistical Analysis ......................................................................................... 17
Results ............................................................................................................................. 20
Discussion ....................................................................................................................... 27
Limitations and Management Implications .................................................................... 32
LITERATURE CITED ...................................................................................................... 35
CHAPTER 3
LAND USE / LAND COVER ASSOCIATED WITH BALD EAGLE BREEDING
AREAS IN THE NORTHERN LOWER PENINSULA OF MICHIGAN ........................ 38
Introduction ..................................................................................................................... 38
Study Area ...................................................................................................................... 40
Methods ........................................................................................................................... 40
Data Preparation and Breeding Area Delineation ........................................................ 4O
GIS and Statistical Analysis ......................................................................................... 45
Results ............................................................................................................................. 48
Discussion ....................................................................................................................... 53
Limitations and Management Implications .................................................................... 58
LITERATURE CITED .......................................................... 61
CHAPTER 4
LAND USE / LAND COVER PATTERNS AND HABITAT MODEL FOR BALD
EAGLE NESTS IN THE NORTHERN LOWER PENINSULA OF MICHIGAN .......... 64
Introduction ..................................................................................................................... 64

vii

Study Area ...................................................................................................................... 65

Methods ........................................................................................................................... 66
Data Preparation ........................................................................................................... 66
GIS and Statistical Analysis ......................................................................................... 67

Results ............................................................................................................................. 70

Discussion ....................................................................................................................... 78

Limitations and Management Implications .................................................................... 85

LITERATURE CITED ...................................................................................................... 88
CHAPTER 5

MANAGEMENT IMPLICATIONS OF USING LAND USE / LAND COVER DATA
FOR DETERMINING NESTING HABITAT OF BALD EAGLES [N MICHIGAN ..... 90

Summary ......................................................................................................................... 90

Limitations ...................................................................................................................... 91

Management Implications ............................................................................................... 93

LITERATURE CITED ...................................................................................................... 96
APPENDIX I

METADATA FOR MICHIGAN BALD EAGLE NEST LOCATIONS -— DIGITAL
POINT FILE ...................................................................................................................... 97
APPENDIX 11

MIRIS land cover classes occurring in Black River Watershed bald eagle nests and
grouped classes used in resource selection analysis ........................................................ 102

viii

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Table 7

Table 8

Table 9

Table 10

Table 1 1

Table 12

LIST OF TABLES

Number of cover classes, and minimum / maximum percent total
area of land cover classes occurring in each time step in the Black
River Watershed, Michigan and bald eagle nests within the watershed....22

Number of nests in Black River Watershed with land cover class
occurring within 500 meters of nest site in 1963, 1978 and 1993 ............. 23

Land cover class as percent of total Black River Watershed area and
percent of combined bald eagle nest site buffer areas. .............................. 24

Means of percent land cover by cover class for Black River Watershed
bald eagle nests within 3 time steps ........................................................... 25

Mean percent land cover by cover class for Black River Watershed
bald eagle nests within activity and production groups ............................. 26

Habitat selection based on land cover composition within 500 meters
of bald eagle nest locations in the Black River Watershed, Michigan
represented by confidence intervals of proportion used ............................ 28

Number of bald eagle breeding areas (BAS) in the northern lower
peninsula of Michigan with land cover class occurring within BA ........... 46

Percent land cover composition of study area and combined area of
bald eagle breeding areas in the northern lower peninsula of Michigan ...49

Mean percent land cover by cover class within 2 time steps for bald
eagle breeding areas in the northern lower peninsula of Michigan ........... 51

Mean percent land cover by cover class within activity and production

groups for bald eagle breeding areas in the northern lower peninsula of
Michigan .................................................................................................... 52

Habitat selection of nesting bald eagles in the northern lower peninsula

of Michigan, determined by confidence intervals of proportion of land
cover composition within 5-kilometer breeding areas representing used
resource ...................................................................................................... 54

Number and percent of bald eagle nests with land cover classes

occuning within 500 meters of nests in the northern lower peninsula of
Michigan .................................................................................................... 71

ix

Table 13

Table 14

Table 15

Table 16

Table 17

Table 18

Table 19

Table 20

Mean percent land cover by cover class within 500 meters of bald eagle
nests in the northern lower peninsula of Michigan for each production
group .......................................................................................................... 72

Habitat selection of nesting bald eagles in the northern lower

peninsula of Michigan, determined by confidence intervals of proportion
of land cover composition within 500 meters of nests representing used
resource ...................................................................................................... 73

Land cover class groupings used for bald eagle nesting habitat model
of the northern lower peninsula of Michigan ............................................. 75

Number and percent of bald eagle nests with model land cover classes
occurring within 500 meters of nests in the northern lower peninsula of
Michigan .................................................................................................... 76

Mean and median percent land cover of grouped land cover classes for
bald eagle nests existing in the northern lower peninsula of Michigan,
1992-1994 .................................................................................................. 77

Logistic regression model classification rates for bald eagle nests
(used) and random points (unused) in the northern lower peninsula
of Michigan ................................................................................................ 79

Percent classification and omission error for bald eagle nesting
habitat model in relation to average rates over 20 iterations ..................... 80

Summary of logistic regression model variables for bald eagle nesting
habitat potential in the northern lower peninsula of Michigan .................. 81

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

LIST OF FIGURES

Location of Black River Watershed bald eagle nest site study
area in the northern lower peninsula of Michigan ..................................... l4

Predominant land cover classes and percent of total area of the
Black River Watershed .............................................................................. 16

Land use / cover used to conduct bald eagle nesting habitat analyses
in the Black River Watershed, Michigan ................................................... 19

Study area for bald eagle breeding areas in the northern lower peninsula
of Michigan ................................................................................................ 41

Bald eagle breeding buffer showing alternate nests, nest activity
weights (# above nests), activity center and buffer boundary ................... 44

xi

CHAPTER 1

SPECIES DESCRIPTION AND HISTORICAL TRENDS

Natural History and Distribution

As one of the most widely distributed and well recognized raptors in North
America, the bald eagle (Haliaeetus leucocephalus) has been the subject of much study
and observation. It is estimated that prior to European settlement, there were a quarter
million to a half million eagles throughout the continent (Gerrard and Bortolotti 1988).
Although the species has undergone dramatic population fluctuations since, current
geographic distribution is similar to historic distribution. Because the main food source
for bald eagles is fish, they are primarily associated with areas of extensive aquatic
habitats such as coastlines, major rivers and lakes. Breeding occurs across most of
Alaska, Canada and the contiguous United States (US), while wintering birds tend to
congregate mostly in the lower 48 states where open water can be found.

As the second largest raptor on the continent, the size of the bald eagle ranges
widely, weighing 3.0 - 6.3 kg (~ 6.5 — 14 lbs.) and having a wingspan of 168 — 244 cm (~
6 — 7 fl.) (Buehler 2000). These large birds build equally impressive nests, generally
ranging 1.5 — 1.8 m in diameter and 0.7 — 1.2 m tall. The nest structure is constructed
primarily of twigs placed just below the crown of alive, overstory tree (Buehler 2000).
Due to their size and preference for fish, bald eagles typically nest in mature, forested
areas within 3 km of shallow (S 2 m) water (Buehler 1995). High foliage height diversity

and relatively open canopies characterize forest tracts where eagles nest (Buehler 2000).

Nests occur in a variety of tree species, both coniferous and deciduous, varying in height
from 20 to 60 m (x = 82 cm dbh, 28 m tall), providing adequate cover while allowing
good visibility and flight access (Buehler 2000). A pair may maintain and reuse alternate
nests within its breeding area, with as many as seven nests at a time associated with one
breeding pair (Postupalsky 1983).

In the upper Midwest, nest building typically begins in February, with l — 3 eggs
laid in March or April and incubated for 35 days. Successful nests may produce 1 — 3
young, which fledge at 8 — 14 weeks of age. After dispersal, immature eagles tend to
range through the summer and wintering areas typical of their particular population,
although there does not seem to be much fidelity to any specific site. This period
generally lasts until maturity is reached at 5 — 6 years. At maturity, the eagle attains the
signature full white head and tail plumage and is ready to begin breeding (Weise 1998,

Buehler 2000).

History and Status in the United States

While it is impossible to know how many eagles existed in pre-settlement US, it
is believed that there were 25,000 to 75,000 breeding eagles in the lower 48 states when
the bird was adopted as the national symbol in 17 82. At this time, the eagle population
began to experience habitat degradation due to expanding settlement, and shooting
mortality justified by the exaggerated perception that they killed livestock and posed a
threat to salmon fisheries (US. Fish and Wildlife Service (USFWS) 1995). In fact,
Alaska posted a bounty on bald eagles in 1917, further encouraging their destruction.

This combination of factors led to a noticeable decline in the population by the early

1900’s. As people became more aware of the magnitude of the loss, the species began to
receive increasing attention for its conservation.

The Bald Eagle Protection Act of 1940 and the 1953 repeal of the Alaska bounty
law made direct harm, harassment, possession and trading of eagles illegal, but did not
address habitat concerns or a future threat that was to have as devastating an effect. After
a brief period of stabilization, the population again began a downward trend. Throughout
the 1950's, the decline of bald eagle populations across the lower 48 states was
accelerated by decreased reproductive success associated with the use of organochloride
compounds such as dichlorodiphenyltrichloroethane (DDT) and polychlorinated biphenyl
(PCB). When the National Audubon Society conducted a breeding pair census in 1963,
only 417 pairs were found in the lower 48 states (US. Federal Register 59:35585).

In 1967, the southern bald eagle (defined as those located south of the 40th
parallel) was officially listed as endangered in the conterrninous US. under the
Endangered Species Protection Act (U .S. Federal Register 32:4001). The ratification of
the newer Endangered Species Act (BSA) in 1973 mandated the national protection of a
species and its habitat if its population levels were reduced to the point of possible
extinction. The southern bald eagle carried its endangered status to listing under the new
act. A proposed modification was finalized in 1978 that classified all eagles in 43 states
as endangered, and threatened in the remaining five (Michigan, Minnesota, Oregon,
Washington and Wisconsin) (U .8. Federal Register 43:6230-6233). Under authority of
the ESA, recovery regions were designated (Michigan and 23 other states were included
in the Northern States region), and working groups were assigned to develop a recovery

plan for each region.

With the implementation of recovery plans, habitat protection and ban of DDT
use in 1972, a 1994 census found 4,452 eagle pairs in the lower 48 states (US. Federal
Register 60: 35999—36010). Based on this ten-fold increase in only 30 years, the species
was reclassified as threatened in all 48 contiguous states in 1995. By 1999, population
estimates exceeded 5,700 breeding pairs, with a consistent 8% average yearly increase for
the previous 10 years. This prompted a proposal for delisting the species across the lower
48 states (U .8. Federal Register 64:36453-36464). The Bald and Golden Eagle
Protection Act, the Migratory Bird Treaty Act and the Lacey Act would still protect the
bald eagle from direct harm and illegal trade, but these do not offer complete habitat
protection as does the ESA. As continued development increases land conversion,
concerns arose that delisting may ultimately lead to a loss of suitable breeding areas. The
1999 proposal, while still active, has not yet been adopted and the bald eagle currently

remains federally listed as threatened in the lower 48 states.

History and Status in Michigan

In pre-settlement Michigan, the bald eagle historically nested throughout the state
where suitable habitat occurred. Through the early and mid-1900's, the population
gradually declined due to loss of habitat, prey resources and direct control by humans
(Barrows 1912, Brewer and McPeek 1991). By the early-1940's, it was estimated that
approximately 50 breeding pairs existed in the state, with the species restricted to the
northern half of the state by 1959 (www.michigan.gov/dnr). Although the federal Bald
Eagle Protection Act helped stabilize Michigan’s eagle population, the use of DDT

during this time impeded the population’s recovery.

The effect of DDT on raptors caused thinning of eggshells, and with thinner
shells, eggs were likely to break before hatching. Due to this compound’s accumulation
in the Great Lakes, environmental persistence and bioaccumulation in Great Lakes fish,
Michigan eagles were particularly affected. Even after-the ban of DDT, reproduction in
Michigan remained severely depressed, especially in pairs nesting near the Great Lakes
and anadromous rivers (Bowerman 1993). To address this, an annual monitoring
program was initiated in 1961, confirming low reproductive rates through the 1960’s.

In 1974, Michigan legislators passed the Endangered Species Act (PA 203), and
in 1976 the bald eagle was listed as threatened within the state. During the mid- and late-
1970's, the number of bald eagles increased slightly and remained steady around 86
breeding pairs. By 1983, it was determined that a region-wide goal of 1,200 breeding
pairs by 2000 was adequate to consider the bald eagle population fully recovered in the
Northern States (Grier et a1. 1983). Michigan was to contribute 140 breeding pairs to the
total recovery goal.

Surveys conducted in Michigan in 1985 revealed 121 breeding pairs producing
0.97 young per occupied nest. The bald eagle population had shown a marked increase
and continued to exceed the production levels necessary for population maintenance (0.7
young per occupied nest) (Postupalsky 1985). In the late-1980's, the Michigan
Department of Natural Resources (MDNR) increased its own recovery goal to 300
nesting pairs based on historical bald eagle nesting records and the number of suitable
territories still in existence (T. Wiese, MDNR, personal communication).

The goal of 300 nesting pairs was initially reached in Michigan in the mid-19905,

and the current status is 388 nesting pairs producing 352 young (0.93 young per occupied

. nest) in 2001 (Postupalsky 2001). Since the mid-1980's, nesting distribution has also
expanded southward, and active nests have now been recorded in 61 counties (D. Best,
USFWS, personal communication). The bald eagle currently remains listed as threatened
in the state under the revised 1994 protected species list of the Michigan Natural
Resources and Environmental Protection Act (Part 365 of PA 451). While the proposal
for federal delisting and achievement of recovery goals is a Sign of successful recovery in
Michigan, it is expected to pose new problems in management. If the species and its
habitat are not firlly protected by law, then management will likely focus more on
maintaining suitable bald eagle habitat within changing, and often conflicting, land use
patterns.

Since the re-establishment of the bald eagle in Michigan has occurred in a time
and location of continued or increasing human development, it is important to determine
what factors have affected nest site selection. Monitoring efforts that began in 1961 have
resulted in data collection on bald eagles for variables such as nest location and nest tree
characteristics. In addition, nestlings have been banded has been done as a means of
following population movements, determining productivity, and formulating population
projections to determine long-term survival (T. Wiese, personal communication). Data
collection and handing have continued and currently provide over 40 years of information
on the decline and recovery of Michigan’s eagles. This data set provides an excellent
resource in determining patterns of bald eagle re-establishment, including how land use
has affected recovery.

This study was an opportunity to analyze the dynamics of a population whose

rapid expansion (after removal of the primary stressor, organochloride contaminants) has

occurred in a landscape much different from what it had previously known. As both
human and eagle populations continue to expand, variables such as level and time of
disturbance (Wright 1986, McGarigal et al. 1991), spatial factors and habitat quality will
continue to affect bald eagle distribution and productivity. The specific goal of this thesis
is to examine the effects of land use / cover on nest site selection and reproductive

success of bald eagles in Michigan.

Key to Abbreviations and Definitions

NLP — northern lower peninsula

SLP — southern lower peninsula

UP — upper peninsula

BRW — Black River watershed

MDNR - Michigan Department of Natural Resources

USFWS — United States Fish and Wildlife Service

MIRIS — Michigan Resource Inventory System — digital land cover maps derived
from 1:24,000 scale color-infiared and black-and-white aerial photographs

GAP — GAP Analysis Project — digital land cover derived from classification of
30 meter Landsat Thematic Mapper imagery

BA — bald eagle breeding area

Nest, nest site — geographical location of an individual nest

Nest area — area around an individual nest bounded by a SOC-meter circle centered
on the nest

Breeding area —— formally, “the local area associated with one territorial pair of
eagles and containing one or more nest structures” (Postupalsky, 1983); for land
cover analysis, this also contains the extended area that may provide potential
foraging habitat

Alternate nests — all nests within a single breeding area, only one of which is used

for nesting in a given breeding season

Occupied, active — evidence of breeding activity was observed; this includes a
pair of adult eagles, one eagle observed incubating, new nest built or an old nest
with signs of repair in the current season; may refer to either a single nest or
breeding area

Inactive — no evidence of breeding activity observed; includes situations where no
eagles are seen, a pair is observed with one individual being immature with no
breeding behavior exhibited or one or more immature individuals are observed;
may refer to either a single nest or breeding area

Successful, productive — young were observed fledged or adequate evidence was
observed to have assumed fledging; may refer to either a single nest or breeding
area

Failed — pair observed at some stage of active breeding, with no evidence of
young fledged; may include nest destruction, failure to lay or hatch eggs, or
young not surviving to fledging; may refer to either a single nest or breeding area
Activity center — weighted geographic center of a breeding area determined by the
number of years each nest was in existence and the number of years each nest was

active within a given time period

LITERATURE CITED

Barrows, W. B. 1912. Michigan bird life. Michigan Agricultural College, East
Lansing, Michigan, USA.

Bowennan, W. W. 1993. Regulation of bald eagle (Haliaeetus leucocephalus)
productivity in the Great Lakes basin: an ecological and toxicological approach.
Dissertation, Michigan State University, East Lansing, Michigan, USA.

Brewer, R., and G. A. McPeek. 1991. Bald eagle - species account. Page 160 in R.
Brewer, G. A. McPeek, and R. J. Adams, editors. The atlas of breeding birds of
Michigan. Michigan State University Press, East Lansing, Michigan, USA.

Buehler, D. A. 1995. A geographic information system to identify potential bald eagle
breeding habitat for southeastern United States rivers and reservoirs. Proceedings
Annual Conference of Southeast Association of Fish and Wildlife Agencies
49:292-302.

. 2000. Bald Eagle (Haliaeetus leucocephalus). No. 506 in A. Poole and
F. Gill, editors. The Birds of North America. The Birds of North America,
Philadelphia, Pennsylvania, USA.

Gerrard, J. M., and G. R. Bortolotti. 1988. The bald eagle: haunts and habits of a
wilderness monarch. Smithsonian Institute Press, Washington, D. C ., USA.

Grier, J. W., J. B. Elder, F. J. Gramlich, N. F. Green, J. V. Kussman, J. E. Mathisen, and
J. P. Mattsson. 1983. Northern states bald eagle recovery plan. U. S. Fish and
Wildlife Service, Denver, Colorado, USA.

McGarigal, K., R. G. Anthony, and F. B. Isaacs. 1991. Interactions of humans and bald
eagles on the Columbia River estuary. Wildlife Monographs 115:1-47.

Postupalsky, S. 1983. Techniques and terminology for surveys of nesting bald eagles.
In Grier et al. Northern states bald eagle recovery plan. U. S. Fish and
Wildlife Service, Denver, Colorado, USA.

. 1985. Eagle and osprey research in Michigan. Michigan Department of
Natural Resources Wildlife Division Report 3102. East Lansing, Michigan, USA.

. 2001. Bald eagle numbers and reproduction — Michigan - 2001. Michigan
Department of Natural Resources Wildlife Division. East Lansing, Michigan,
USA.

US. Fish and Wildlife Service. 1995. Biologue Series -— Endangered species success
story: bald eagle. US. Fish and Wildlife Service, U. S. Department Interior.

10

Weise, T. 1998. Michigan bald eagle information - web site report. Michigan
Department of Natural Resources Wildlife Division, East Lansing, Michigan,
USA.

Wright, M. F. 1986. Simulating the search image of nesting bald eagles in western
Montana: a management model. University of Montana, Missoula, Montana,
USA.

11

CHAPTER 2
LAND USE / LAND COVER PATTERNS OF BALD EAGLE NEST LOCATION,

ACTIVITY AND SUCCESS IN THE BLACK RIVER WATERSHED, MICHIGAN

INTRODUCTION

Many studies have been conducted on the bald eagle in the United States. These
studies cover a variety of subjects including contaminant and human disturbance effects
on production, habitat use by non-breeding birds (sub-adults and wintering birds),
foraging habits and population dynamics. Nest habitat characteristics have not received
as much attention as these other factors even though nest sites are integral for
reproduction and their protection and management are mandated by the ESA. In over
300 references used by Buehler (2000), approximately 20 focused specifically on nest
site selection and nesting habitat. Of these, nesting habitat has been described primarily
in the Pacific Northwest (Anthony and Isaacs 1989, Garrett et al. 1993), Northeastern
U.S. (Andrew and Moshier 1982, Livingston et al. 1990) and Southeastern U.S. (Wood et
al. 1989, Hardesty and Collopy 1990, Buehler 1995). However, bald eagle research in
the Great Lakes region, particularly Michigan, has concentrated on contaminant effects
(Bowerman 1993; Bowerman et al. 1994, 1998, 2000), response to human disturbance
(Mathisen 1968, Grubb et a1. 1992, Grubb and Bowerman 1997) and microhabitat
variables such as nest tree characteristics (Grewe 1980, Mathisen 1983, Bowerman
1993)

In pre-settlement Michigan, the bald eagle historically nested throughout the state

where suitable habitat occurred. Through the early and mid-1900's, the population

12

gradually declined due to loss of habitat, prey resources, direct control by humans and
low productivity due to environmental contaminants (Barrows 1912, Brewer and McPeek
1991). To address this, an annual statewide monitoring program was initiated in 1961,
providing over 40 years of information on nest location and production. These data,
coupled with the increasing availability of digital spatial data, provides an excellent
resource for determining both spatial and temporal patterns of land use / land cover
associated with bald eagle nest location and productivity. This study presents an analysis
of nesting habitat of a population whose rapid expansion (after removal of the primary
decimating factor, pesticides) has occurred in a landscape much different from what had
previously been available. Specific objectives were to determine patterns of land cover
preferences associated with bald eagle nest locations, temporal changes in nest site land
cover and differences in land cover between active and inactive nests, and between

productive and failed nests.

STUDY AREA

The data on nest location and production have been collected across the state of
Michigan. Because standard, annual data collection did not begin until after the species
had become restricted to the north, the majority of eagle breeding locations are
distributed throughout the northern Lower Peninsula (N LP) and the Upper Peninsula
(UP). The present analysis focused on nest sites in the Black River watershed (BRW).

The BRW encompasses approximately 155,800 hectares in northern Michigan
and includes parts of Cheboygan, Montmorency, Otsego and Presque Isle counties

(Figure 1). Approximately half the watershed area is under state ownership in the

13

 

 

 

 

 

 

 

 

 

Fig. 1. Location of Black River Watershed bald eagle nest site study area in the
northern lower peninsula of Michigan.

Mackinac State Forest. Land cover is predominantly aspen / white birch and associates,
northern hardwoods and lowland conifers, which together comprise 47% of the watershed
area (Figure 2). Although this area did not contain many eagle nests through the 1960’s
and 1970’s, the area was the subject of a land use / land cover time series analysis
(Rutledge 2001). The BRW data provide detailed land cover data in a GIS format, using
the same classification system (Michigan Resource Inventory System (MIRIS)) for three
time periods (1963, 1978 and 1992). This allowed examination of patterns in bald eagle
nesting areas using more detailed land cover classes than that of other available Michigan
spatial data during the entire time that nest locations and production have been

monitored.

METHODS
Data Preparation

To use a geographic information system (GIS) for comparison of nest locations
and associated land cover, a digital point file was created of all recorded bald eagle nest
locations in Michigan. Nest locations were digitized in ArcView 3.2 (ESRI, Redlands,
California) using a combination of GPS coordinates, public land survey description and
additional reference notes. Nest points were checked against hydrology shapefiles to
ensure that nests did not fall offshore or in the middle of a lake with respect to existing
base maps. A USFWS biologist (D. Best, personal communication) was consulted to
correct discrepancies (Appendix I).

The ArcView extension XTools (DeLaune 2001) was used to select nests falling

within the study area by using the clip command to clip the BRW area from the statewide

15

 

C] Other —— 53%

— 26%

I: Aspen, white birch and associates

- Northern Hardwood — 10%

- Lowland Conifer — 11%

Fig. 2. Predominant land cover classes and percent of total area of the Black River

Watershed, Michigan.

16

nest location coverage. Production, start year and end year records for these nests were
used to determine whether each nest was active, inactive, productive or failal for each
year it existed. Sample selection was further refined by using only those nests in the
BRW that existed at some time within a three-year period centered on each land cover
time step. This provided three 3-year time steps for nests (1962-1964, 1977-1979 and
1991-1993), maximizing sample size without temporally extrapolating land cover data
too far.

These nests were firrther classified according to activity (A -— active, I -- inactive)
and production (P — productive, F — failed). Because activity and production are variable
across years for individual nests, only those nests that existed in the same year as the land
cover data time steps (1963, 1978 and 1992) were used for activity and production
analysis.

GIS and Statistical Analysis

Land cover was summarized by MIRIS level 4 cover class for the entire BRW
area (Appendix II), and percent of total watershed area was calculated for each class for
each year (1963, 1978 and 1992). Nest sites and associated land use / cover were
evaluated by buffering each nest with a circular buffer of 500 m radius. This distance
was chosen because most disturbance responses of nesting eagles to humans or
conspecifics are induced within 500 m (Fraser et. al. 1985, Livingston et. al. 1990, Grubb
et al. 1992). This distance is also comparable to a tertiary disturbance buffer zone
described in the management guidelines of the Northern States Bald Eagle Recovery Plan
(Grier et al. 1983).

XTools was used to intersect buffer zones for each nest time step with the

17

 

associated land cover time step to determine land cover within 500 m surrounding each
nest (Figure 3). The resulting nest site land cover was summarized by MIRIS level 4
cover class and percent area of each class was calculated for each nest area. Some land
cover classes were combined because of the number of classes and range of values
occurring in the BWR and nest sites. For example, for resource selection analysis, all
wetland classes (i.e., emergent, scrub / shrub, forested) were grouped into one general
wetland class. Percent cover class of combined nest area within each time step was also
calculated as a means of initially determining which classes could be grouped or left out
of further statistical tests.

Nest site land cover was also classified by time step and each of the 26 land cover
classes occurring in nest areas were tested using a Kruskal-Wallis (or = 0.10) to detect
temporal changes in nesting land cover. Nests existing in multiple time steps were
treated as separate occurrences since land cover around those nests may have changed
over time, and this analysis is based on use of nests rather than just where they were
initially placed.

To determine if nest site land cover had any significant affect on activity or
production, nests from all time steps (1963, 1978 and 1992) were grouped. A Mann-
Whitney U (or = 0.10) was used to test land cover classes for differences within activity
and production categories. Twenty-two land cover classes occurred in at least one nest
area and were tested for active / inactive nest comparisons. Two classes (northern
hardwood — no predominant and upland conifers — white spruce predominant) only
occurred in nests that could not be classified by production category, leaving 20 land

cover classes for productive / failed nest comparisons.

18

 

0 nest site

land cover

 

Fig. 3. Example of 500 m land use / cover buffer used to conduct bald eagle nesting
habitat analyses in the Black River Watershed, Michigan.

19

Preference or avoidance in nest site land cover was analyzed using Resource
Selection Analysis Software (Leban 1999). Nest area land cover compositions from all
time periods were used. Similar cover classes were grouped providing 21 classes for
analysis (Appendix 11). Land cover composition of the entire BRW area was designated
as available resource. The data matrix was prepared by pairing each nest area
composition (used resource) with its corresponding time period’s BRW composition
(available resource). This was entered into the sofiware and analyzed using the Neu
method (Neu et al. 1974; a = 0.10), which combines the use of chi-square goodness-of-fit
analysis and construction of Bonferroni simultaneous confidence intervals. The result is
preference, avoidance, or neither for each cover class based on where the predicted use
value falls in reference to the confidence interval. For instance, if the lower confidence
limit of used resource is greater than the predicted value, that class is considered

preferred.

RESULTS

Twenty-one nests existed in the BRW during the given time periods. All nests
were used for use / availability and temporal analysis, 8 in 1962-1964, 3 in 1977-1979
and 10 in 1991-1993. Only 15 nests existed in the same years as the land cover data,
leaving 9 active and 6 inactive nests for activity comparisons. Of the 9 active nests, one
had an unknown outcome, leaving 3 productive and 5 failed nests for production
comparisons. Over all time steps, 91 MIRIS Level 4 cover classes occurred within the
BRW, and 32 occurred in nest areas. No time steps include all classes, indicating the

addition and removal of some classes over time, and nesting areas in all time steps

20

include < 34% of possible cover classes. The total number of cover classes occurring in
nest areas was highest during the 1993 time step and lowest during the 1978 time step.

Percent total area of land cover classes was highly variable in all time periods,
ranging from near zero to > 60% (Table 1). Of all cover classes occurring within the total
nest area of any time step, most were only present in <50% of the nest sites (Table 2).
Aspen / white birch, lowland conifers and northern hardwoods occur in the greatest
percentage across the watershed, while aspen / white birch, lake and lowland conifer are
greatest among the nest areas (Table 3). Only 4 other cover classes were present at nest
areas during all time periods (scrub/shrub wetland, upland jack pine, lowland hardwood -
aspen predominant, herbaceous). Twenty-two cover classes occurred at >2% of total nest
area in at least one time period (Table 3). Only 7 of these classes occurred in nests areas
in all time periods. All classes fluctuated greatly over time in nest areas even as they
remained consistent in the watershed (Table 3).

Five classes (extractive — wells; northern hardwoods; aspen / white birch and
associates; red pine; jack pine) showed significant changes over time (P < 0.10). The
percent of all five classes changed within nest areas from 1963 to 1992 with northern
hardwoods, aspen / white birch and associates, and red pine decreasing over time and
extractive and jack pine increasing. Aspen / white birch also showed a significant
decrease from 1978 to 1992 (P = 0.02) and red pine decreased both from 1963 to 1978
and 1978 to 1992 (P = 0.02; Table 4). No significant differences (P > 0.10) occurred in
any land cover classes between active and inactive nests (Table 5). Only one cover class
(extractive — wells) showed a significant difference (P = 0.08) between productive and

failed nests, occurring in greater proportion in productive nests than in failed (Table 5).

21

Table 1. Number of cover classes, and minimum / maximum percent total area of land
cover classes occurring in each time step in the Black River Watershed, Michigan and
bald eagle nest areas within the watershed.

 

 

# MIRIS Level 4 Minimum Maximum
Time Step Land Covers Classes % Area % Area
Black River 1963 83 <0.01 26.50
Watershed 1978 80 <0.01 26.53
1992 86 <0.0l 23.87
500 meter 1962-1964 19 0.19 57.78
nest site 1977-1979 8 0.76 57.28
buffer 1991-1993 20 0.17 61.74

 

22

Table 2. Number of nests in Black River Watershed, Michigan with land cover class
occurring within 500 meters of nest site in 1963, 1978 and 1992.

 

No. of nests

 

 

Level 4 1963 1978 1992
MIRIS code Description (11 = 8) (n = 3) (n = 10)
1 130 Residential 1 2
1730 Extractive - wells 4
3100 Non-forested - herbaceous 3 l 2
3200 Shrub 3 5

4110 Northern hardwood 3

411 1 Sugar maple predominant 1

4130 Aspen, white birch and assoc. 8 3 6
4132 Bi gtooth aspen predominant 2

4143 Lowland hardwood - sofi maple predominant 3 2
4146 Aspen predominant 3 2 5'
4149 Other lowland hardwood 1

4210 Coniferous - pine l

4212 Red pine predominant 4

4213 Jack pine predominant l 1 5

4221 Upland conifer - white spruce predominant l

4230 Lowland conifers 5 2 4
4231 Cedar predominant l

4232 Black spruce predominant 1

4900 Other forested 1

5 100 Stream 6 2 7
5200 Lake 4 2 4
5300 Reservoir 1

6110 Forested wetland — wooded 2 4
6120 Scrub, shrub 5 2 6
6210 Non-forested wetland - aquatic bed 3 2
6220 Emergent 2

 

23

 

09w cod cod who Omd de Bowuofio - van—33 cuamoeoHaoZ ommc

 

8... a: 9.2 NS OS was BE... seam cm;
8.4 cod £6 ”no woo 3o 883 - 2303 BEE 2 3
SN 86 28 mg mg 3 .o scream 8mm
32 $22 ”3: mam 42 m2 £3 8%
So 86 Se 2: :1 :: aaaneea 38o :3.
23 w? _ 5% n; cos c; seas 3233 Om?
2.2 new Ba Ba 2e 2 e aaeeoeaa Ea as: 2?
Se 85 a2 and as was €25.83 2% Ba .2?
a2 8.0 86 as 3% SA Ea - aeeaoo 2 9.
Ba 2:: Se 3% as. m Em aafioeaa use... 9: 4
8s 2:. an. : 2.. 42 MS 522805 is: com a: v
86 8.0 23 :2. m _ .N m L .N $8322 9333 8;
3m 23 Se 02 22 SN EEEOBE some aosam mm _ v
is $2 3.2 3.: $8 on: 8...: Ba :25 as? .552 2:.
8.0 . 8o 86 OS :8. SN afiaeoeea x8 Be a a 4
8o 85 86 a .m 3.». mom “ESE 29E swam _ H 3
85 Se was 8s Gs Gs 802:2 E0552 2 2.
mi. 85 as was was EN Beam 8%
8a :3 EN s3 28 36 Ease»; - 8:28-82 2: m
cod 2:. 85 one :e as 8285 8:.
as 86 as as 2: was .3533. OS _
$2.32 $25 32.8.: 82 was 82 83:88 88 mg:
8&3 88 H8: 888 can 35883 52% xoflm

 

.5505 2a meow 08: 2.8 H32 3 E Av\..o.NA wagooo 8 88: E “samba momma—o 3:0 .933 Stan 86 «we: Emma
23 35988 no E083 can 88 gamma—22 622033 53M xoflm :38 no 5083 me $20 830 284 .m 2an

24

 

 

 

 

 

mac 3 od ed I “80.08”,” 3 3
$6 ad o.o _.~ I won 093—; 00608802 3
:3 ed v; v.2 III 95? £30m 3 3
9.6 06 ed ad I 80803 003.08» £82.03
mmd _.~ o.o od III .II :oE0m0M 3
nod w.m_ v.2 Nam— III I: 003‘— 3
86 5o wd : I .II 8025 0033
wmd @— od ad I I" .050 3
38 9o 3 2v 83% 3.85 .. .. ..
wfio od 0.0 Na 0800 3 3 3
32¢ v.0 v; g Wm 8882008 oz 00.880 van—33 3 3
33 ed o.o ad 003% 0033 $0088 983: 0050 3 3
mod 02 in od 0:5 x03 3 3 3
86 ed od w.~ 0:5 80¢ 3 3 3
wmd 04 ed ad 8058.808 02 0am 2800.380 3
$6 _.o ad ad 0050 3 3 3
£5 E fl: 3 5&3 3 .. 3
93.0 ON 06 NS 0302 com 80330: van—Boa 3 3
mmd o.m od od 0094 fioogm m0§8mm< 98 3 3
mod w.c msm WE 80582008 02 :85 833 d0am< 3 3
33 ed od “.ov 0802 Bmsm 3 3 3
36 ed o.o m8 8382008 02 89592 E088 Z 0030685 006008
de Nd o.o Wm II I 255 3
who 2.. m.m o.~ Ii ll 0800.300: “003.08%8 Z
mod md od od ll m=0>> 0>coabxm 3
Ed ~.w o.o ad II III 3802,03“ 5303
A 32.53 aha—-32 $2-82 v _0>0A m .054 N _0>0A a _0>0A
doum 0:5. $20 952

 

.305 08: m 55:? 060:

0&00 23 60500003 003M 805 8.“ 05: 008 098:0 98 000.0 0080 33 0080 9:: E0000.“ .8 0502 .v 050%

25

888:5 8508— 65 88: 038m 5 Samoa 8: mam—U m

BEE t ”— .uESSvSQ I m ”3:on E50255 +
2685 I _ .953 I < 695% b33333 _._

 

 

 

 

 

Ed o.w ed the ms TV I EomBEm 3 3
3.6 3 ed Ed ..o o; I can 0:33. 33.8882 3
$6 3 he as 3 Z I can... 5.5m 3 3
Sue md v.2 :5 No 0.3 II 3803 37.23 €828?
5.0 3. ed om o o o m N I III hotumom 3
. . . III II: o a 3
8.— odm 92 «no nNN oo— x A
. . 36 ed o4 11' I Emubm .583
2: m o _ _ . .
mm o 95 N. m 3288qu 02 5380 6:233 3 3
8.3 3 9o . . .
I. I mm o _ _ o o 835m 0:55 30,380 ESE: 350 3 3
9o 5; E of. S3 q: a: 33.: 3 3 3
Ed Nd ed the wN _.o 3E com 3 3 3
5.0 NM o.o end od w; “cacgocoa 02 0:5 aneuubcou 3
and o.o no o3 o.o S 560 3 3 3
as «a 3 3o 2: E 8%< 3 3 3
o m. o o.o cdm Sue m.~ mgr.— oEnE com @839“: van—Bog 3 3
86 EN ed 2. o m N m _ 593‘ £8qu 8888m< 28 3 3
. . ovd w.w_ WE 83.5883 02 scum 85>? doqm< 3 3
05.0 v ON m a
. I. a I Duo o.m I EEEBBE oz vooBEE 505.52 323685 6238,,“
E o v o _ m who 3 .2 1! I £5 3
2: m . m N3 cod a. _ v. m -rlu .|.| Scooafiom 3328-5 Z
mod od m6 :6 N6 N6 III w=o>> o>uombxm 3
end od 52 36 ed 9m II III .anaoEmom 583
m m + m m _ ..< v .964 m 625 N :53 ~ ~26:
cocoa—c8; bréo< 3.20 mg:

 

£5on coco—60E can @358 £535 mafia “no: 2me 23 53:32 £05683 83M xoflm go.“ .560 9:: «coupon :82 .m 033.

26

Resource selection analysis indicated preferred land cover classes for nest areas
including urban, lowland hardwood (soft maple and aspen), lowland conifer (cedar),
water and wetland. A preference for upland jack pine was indicated but not significant.
Aspen / white birch (bigtooth aspen) and upland conifer (white spruce) where neither

preferred nor avoided, and all other classes were avoided (P < 0.10, Table 6).

DISCUSSION 5—

In general, land cover classes associated with bald eagle nest sites in the BRW are i..-
consistent with general habitat preferences of forested areas near (< 2 km) large bodies of
water (Buehler 2000). Bowerman (1993) described Michigan bald eagle nesting habitat
but only for nests associated with the Great Lakes and the focus was on microhabitat
variables (e. g., availability of nest and perch trees, type and proximity of disturbance,
amount of forested area). Another study currently in review (K. Eisenreich, Clemson
University, personal communication) used GIS methods and land use / land cover data
for the entire state. However, the land cover data was from the US. Environmental
Protection Agency’s 1983 coverage in the Anderson classification system. The coverage
was developed from larger scale 1:250,000 quadrangles and only broad cover
classifications such as agriculture, forested and wetland were used. Using such broad
classes does not identify more specific habitat associations characteristic of the Great
Lakes region. This is the first time bald eagle nest habitat has been described for
Michigan using land cover data at a finer resolution (124,000 scale).

Land cover classes that were identified either as preferred (Table 6) or

consistently present through time (Table 4) tend to be forest classes that are associated

27

FE

Baccawu 8: :5 ©8865 sauce—om .._

 

 

 

58¢ £35 88 38¢ 3: 8 528.8 833338 .8883 .53 8383 5:383
58¢ poem 25¢ ~5¢ 88 86:82 .8323. 38.3: 3233
58¢ £35 35¢ 38¢ m~¢¢ 380 3 3
.m2 233. 38¢ 38¢ 58 8355235 oz 858 83:33 3
~8¢ 5¢¢ ¢8¢ 8&8 253 33:38 835: sec 3
..mz £35 85¢ 38¢ 58 a: 332 3 3
88¢ 29E 55 3 :5 38¢ 85 Ex 3 3
oomod Eo>< mmod m 56 mood “888088 02 vim 808528 8529a
88 ~25 88 3350 3 3
58¢ 585 58¢ 38¢ ¢8¢ 5&3. 3 3
58¢ 5.35 m :5 38¢ 58¢ 353: com 3 3
88¢ 28,3 58 88 88 6358823. oz 82232 8333 3
28 -¢¢ 5¢¢ 382 588m 322833 83 3
58¢ 202 88 8 8 ¢ 22¢ 8358825 oz :85 223 .532 3
28¢ EOE 38¢ 88 88 35 Ba 838330 3
58¢ 222 95¢ 58 88 2%: 3&8 3 3
58¢ 23.3. 35¢ 38¢ 35¢ 33355825 oz 58323; 835.02 8333585 8328
225 22.3 _:¢ 38 35¢ 53352:... 388355 8388-82
58¢ 222 38¢ 85¢ 88 288.83.
58¢ .558 25¢ 88 5¢¢ 233 I 023838 .3828: :35
m 8828 333:3}. 8 8 3 83 m :33 N a. g :33
8508M 83.885 883 833 $20 mEE

 

63258 83.58:— 8 889:8 88 85.585 mo mgr—BE ooeowmaoo an 858852 EmEQME 82—8083
83d x85 05 E 80:82 58 280 23 mo 8208 00m 555» 835888 8>8 82 8 88a 8828 233mm 6 2an

28

with lakes and streams in Michigan (lowland hardwoods, lowland conifers and wetlands).
Some surprising results were noted in classes representing factors of particular
importance (i.e., water (Buehler 1995, Swenson et al. 1986), urban (residential and
industrial; Mathisen 1968, Grubb et al. 1992, Grubb and Bowerman 1997)). Residential
areas, while occurring at low levels in all categories, increased in nest areas more so than
in the watershed from 1978 to 1992, leading to the urban class being designated as
preferred (Table 6). This would seem to indicate increasing tolerance for disturbance and
would be counter-intuitive. More likely, rather than a true change in disturbance
tolerance, this is probably a statistical artifact due to increased development and land
conversion during the same time that the number of nesting areas increased.

Though changes in water cover classes over time were not significant, an
unexpected trend was noticed. The percentage of lakes, and reservoirs to a lesser extent,
showed an overall decline in nest areas over time. While water was a preferred cover
class (Table 6) and consistently comprised a greater percentage of nesting areas than
BRW area through time, the percentage of water in BRW area did not change (Table 3).
This could indicate a reduction of nests located on inland lakes, possibly due to
increasing development and recreation on lakeshores. However, this could simply be due
to the increase in the bald eagle nesting population. While eagle nesting habitat over the
entire state was not saturated by 1992, the number of nesting pairs had nearly tripled
since 1978. By the last year of this analysis, nest areas near water in the BRW may have
all been occupied, forcing newer pairs to establish nests farther away from lake and river
shorelines.

Other than the slight discrepancies in what would be expected with urban and

29

water classes, most of the preferred land cover classes (water, wetland, lowland
hardwood and lowland conifer; Table 6) are consistent with what is generally considered
suitable bald eagle nesting habitat. However, it cannot be assumed that eagles are
selecting nest areas based on these habitat variables alone. It is more likely that their
tendency to nest near water plays a greater role and the preferred land cover classes
indicated are simply characteristic of lowland areas surrounding water bodies in
Michigan.

Occurrence of j ack pine cover steadily increased in nest areas over time, but in the
BRW it remained unchanged fiom 1963 to 1978 and declined slightly from 1978 to 1992.
Even with this increase, 1992 was the only time period in which it occurred at a greater
percentage in the nest areas than in the watershed (Table 4). While a preference for
upland jack pine was indicated, this class occurred at a low percentage (~ 6 %) of nest
area composition, and probably does not have a significant relationship to nesting biology
(Table 6). This is more likely caused by both the increased number of breeding pairs and
increased lake shore development that would lead to nests being established farther from
water bodies, increasing upland cover classes surrounding nests. As such, it could be an
indicator (or consequence) of the pattern noted in urban and water classes. Cover classes
that were avoided (Table 6) are also consistent with general nesting area preferences in
that avoided classes tend to be upland classes (greater distance to water) or agriculture
and non-forested (lack of suitable nest trees).

All wetland cover classes also fluctuated over time in nest areas, while remaining
unchanged in the watershed (Table 3). This is probably an effect of the small sample size

for each time period and the relatively low occurrence of these classes in nest area

30

composition which would exaggerate the fluctuations. When all wetland classes were
grouped and all nests used for resource selection analysis, wetlands were generally
preferred. Again, this is most likely an effect of their proximity to large bodies of water
rather than an active selection for the wetlands specifically.

Classes showing a significant temporal change (extractive — wells; northern
hardwoods; aspen, white birch and associates; red pine; jack pine) generally occurred at
fairly low levels during all time periods and may not indicate a true change in nest area
land cover composition. Cover classes comprising the greatest proportion of combined
nest areas (aspen / white birch and associates, and soft maple and aspen predominated
lowland hardwoods) tended to vary in the same way but to a greater extent than in the
BRW (Table 3). This generally involved an increase fi'om 1963 to 1978, followed by a
decrease from 1978 to 1992, and is most likely a reflection of regional land management
rather than changes in nest use. However, nest tree selection could be more of a factor in
this pattern as surrounding nest area. Forest management practices may have led to a
greater abundance of large, mature aspen and maple in the late 1970’s, while harvest and
senescence could have reduced the availability of these trees by the 1990’s. Rutledge
(2001) found that aspen / white birch had the greatest total area of land conversion, and
lowland hardwoods had the third greatest area converted within the BRW. The amount
of these cover classes in the watershed would not have changed much in total area, as
regenerating stands would be included in land cover descriptions. However, the trees
would not be large enough yet to support an eagle nest, and could account for the greater
decrease in nesting area aspen / maple cover classes in 1992.

The lack of differences between active and inactive nests has two possible causes.

31

 

First, nest establishment and use is probably based more on a smaller scale of habitat
variables and proximity to open water early in the season (Buehler 2000, Swenson et al.
1986). If surrounding land cover has an affect on activity (i.e., prone to disturbance, lack
of adequate cover) it may not be realized until later in the season after a nesting attempt
has been made. These cases are still considered active though they may have failed.
Secondly, because breeding pairs of bald eagles maintain and may reuse alternate nests,
activity and inactivity are not mutually exclusive within breeding areas. This would have
affected statistical analysis since the two categories are not independent samples.

There were also no differences in land cover between productive and failed nests
other than in the extractive — wells cover class. This class was higher in productive nests
and showed a significant increase over time, but probably has no effect since it only
occurred at extremely low levels (< 1.0%). Some differences were expected as the
presence of some classes and resulting edge effects could influence foraging quality,
cover, and likelihood of predation or disturbance affects. The small sample size certainly
affected the results of this analysis as well, so the results of production comparisons were

considered inconclusive.

LIMITATIONS AND MANAGEMENT IMPLICATIONS

Results of this analysis do indicate some obvious patterns, but these should be
considered cautiously due to limitations of the analysis. While the land cover time series
allowed examination of temporal changes, the geographic extent is limiting. The
combination of a small area (relative to bald eagle habitat use) and the reduced eagle

population led to very small sample sizes. This in turn reduces the power of statistical

32

 

analysis. This is somewhat alleviated by the quality of the land cover data. The land
cover database was developed specifically to study changes over time in this region of the
state (Rutledge 2001). Therefore, the land cover associated with eagle nests in the BRW
can be considered reliable even if the lack of nest sites does lend itself to hypothesis
testing.

Resource selection analysis should also be considered carefully. A problem
typically encountered with compositional data is that proportional composition is not
independent. An increase in one cover class will lead to a decrease in another cover
class. For this reason, a preference in one class will inevitably lead to an apparent
avoidance of another class (Aebischer et al. 1993), so there were some concerns in using
the Neu method (N eu et al. 1974). This constraint was ultimately considered acceptable
though following a comparison of selection analysis methods conducted by McClean et
al. (1998). The authors found that the Neu method was the only method that identified
selection patterns consistent with known requirements and did so across several spatial
scales.

In spite of these limitations, there are trends that may be useful. Although land
cover among nests is highly variable, the overall habitat preferences do not seem to have
truly changed over time. Land cover classes that seem to be preferred consistently
comprise the greatest proportion of nest areas but are typically those that have also been
the most intensively managed or subject to conversion (e. g., jack pine, aspen, wetlands).
Because of this, the temporal changes in the amount of those land cover classes should be
interpreted as being due to land management practices rather than changing habitat

preference of bald eagle breeding pairs. This implies that breeding pairs are not

33

necessarily becoming more tolerant of less suitable areas. Some biologists have noted
that nesting eagles may be more tolerant of human activities than once believed (D. Best,
personal communication).

Michigan’s breeding eagle population has increased very rapidly, clouding the
line between intolerance or simply taking what is available. Timing and type of
disturbance may play a greater role in nest success than just the presence of human
activity. A nest is less likely to be abandoned later in the season, but pairs may be more
likely to accept some disturbance if it is already there when they begin nesting (S.
Postupalsky, personal communication). This does not necessarily mean they will use the
same nest again in following years if a more suitable area is available. Managers should
keep in mind that as the eagle population continues to expand, the availability of optimal
breeding areas will decrease, and as breeding pairs begin to settle in sub-optimal areas,
they may experience lower reproductive rates. If the population experiences another
serious decline caused by increased mortality (e.g., disease factors, environmental
contaminants) optimal habitat may play an important role in maintaining reproductive

SUCCESS.

34

LITERATURE CITED

Aebischer, N. J ., P. A. Robertson, and R. E. Kenward. 1993. Compositional analysis of
habitat use from animal radio-tracking data. Ecology 74(5):1313-1325.

Andrew, J. M., and J. A. Mosher. 1982. Bald eagle nest site selection and nesting habitat
in Maryland. Journal of Wildlife Management 46(2): 383- 390.

Anthony, R. G., and F. B. Isaacs. 1989. Characteristics of bald eagle nest sites in
Oregon. Journal of Wildlife Management 63(1):]48-159.

Barrows, W. B. 1912. Michigan bird life. Michigan Agricultural College, East
Lansing, Michigan, USA.

Bowerman, W. W. 1993. Regulation of bald eagle (Haliaeetus leucocephalus)
productivity in the Great Lakes basin: an ecological and toxicological approach.
Dissertation, Michigan State University, East Lansing, Michigan, USA.

, D. A. Best, J. P. Giesy, T. J. Kubiak, and J. G. Sikarskie. 1994. The influence
of environmental contaminants of bald eagle (Haliaeetus leucocephalus)
populations in the Laurentian Great Lakes, North America. Pages 703-707 in
B.U. Meyburg and R. D. Chancellor, editors. Raptor conservation today. Pica
Press, Moose Jaw, Saskatchewan, Canada.

, 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
Management and Assessment 53:197-212.

, D. A. Best, T. G. Grubb, J. G. Sikarskie, and J. P. Giesy. 2000.
Assessment of environmental endocrine disruptors in bald eagles of the Great
Lakes. Environmental Management and Assessment 53:197-212.

Brewer, R., and G. A. McPeek. 1991. Bald eagle - species account. Page 160 in R.
Brewer, G. A. McPeek, and R. J. Adams, editors. The atlas of breeding birds of
Michigan. Michigan State University Press, East Lansing, Michigan, USA.

Buehler, D. A. 1995. A geographic information system to identify potential bald eagle
breeding habitat for southeastern United States rivers and reservoirs. Proceedings
Annual Conference of Southeast Association of Fish and Wildlife Agencies
49:292-302.

. 2000. Bald Eagle (Haliaeetus Ieucocephalus). No. 506 in A. Poole and

F. Gill, editors. The Birds of North America. The Birds of North America,
Philadelphia, Pennsylvania, USA.

35

 

DeLaune, M. 2001. XTools ArcView extension. Version 6/1/2001. State Forest
Management Program, Oregon Department of Forestry. Salem, Oregon, USA.

Fraser, J. D., L. D. Frenzel, and J. E. Mathisen. 1985. The impact of human activities on
Breeding bald eagles in North-central Minnesota. Journal of Wildlife
Management 49(3):585-592.

Garrett, M. G., J. W. Watson, and R. G. Anthony. 1993. Bald eagle home range and
habitat use in the Columbia River estuary. Journal of Wildlife Management
57(1):]9-27.

Grewe, A. J. 1980. Habitat requirements of nesting populations. Pages 7-10 in T. N.
Ingram, editor and coordinator. Proceedings of the bald eagle conference on

” ‘6

“raptor rehabilitation , eagle nesting biology”.

Grier, J. W., J. B. Elder, F. J. Gramlich, N. F. Green, J. V. Kussman, J. E. Mathisen, and
J. P. Mattsson. 1983. Northern states bald eagle recovery plan. U. S. Fish and
Wildlife Service, Denver, Colorado, USA.

Grubb, T. G., W. W. Bowerman, J. P. Giesy, and G. A. Dawson. 1992. Response of
Breeding bald eagles, Haliaeetus leucocephalus, to human activities in
Northcentral Michigan. The Canadian Field-Naturalist 106:443-453.

, and W. W. Bowerman. 1997. Variations in breeding bald eagle responses to
jets, light planes and helicopters. Journal of Raptor Research 31(3):213-222

Hardesty, J. L. and M. W. Collopy. 1990. Nesting ecology, habitat use, and dispersal of
the southern bald eagle on John F. Kennedy Space Center (Merritt Island National
Wildlife Refuge) and Vicinity, Florida: Phase One. Annual Report to the
National Fish and Wildlife Foundation.

Leban, F. 1999. Resource selection for Windows 95/NT. Version 1.0, Beta 8.1.
University of Idaho, Moscow, Idaho, USA.

Livingston, S. A., C. S. Todd, W. B. Krohn, and R. B. Owen, Jr. 1990. Habitat model
for nesting bald eagles in Maine. Journal of Wildlife Management 54(4):644-653.

Mathisen, J. E. 1968. Effects of human disturbance of nesting of bald eagles. Journal of
Wildlife Management 32(1): 1-6.

. 1983. Nest site selection by bald eagles on the Chippewa National Forest.

Pages 95-100 in D. M. Bird, editor. Biology and management of bald eagles and
osprey. Harpbell Press, St. Anne de Bellevue, Quebec, Canada.

36

McClean, S. A., M. A. Rumble, R. M. King, and W.L. Baker. 1998. Evaluation of
resource selection methods with different definitions of availability. Journal of
Wildlife Management 62(2):793-801.

Neu, C. W., C. R. Byers, J. M. Peek. 1974. A technique for analysis of utilization —
availability data. Journal of Wildlife Management 38(3):54l-545.

Rutledge, D. T. 2001. Changes in land cover and wildlife habitats in two watersheds in
the lower peninsula of Michigan. Dissertation, Michigan State University, East
Lansing, Michigan, USA

Swenson, J. E., K. L. Alt and R. L. Eng. 1986. Ecology of bald eagles in the greater
Yellowstone ecosystem. Wildlife Monographs 95.

Wood, P. B., T. C. Edwards, Jr., and M. W. Collopy. 1989. Characteristics of bald eagle
nesting habitat in Florida. Journal of Wildlife Management 53(2):441-449.

37

 

CHAPTER 3
LAND USE / LAND COVER ASSOCLATED WITH BALD EAGLE BREEDING

AREAS IN THE NORTHERN LOWER PENINSULA OF MICHIGAN

INTRODUCTION

Many studies have been conducted on the bald eagle in the United States. These
studies cover a variety of subjects including contaminant and human disturbance effects
on production, habitat use by non-breeding birds (sub-adults and wintering birds),
foraging habits and population dynamics. Nest habitat characteristics seem to have not
received as much attention as these other factors, even though nest sites are integral for
reproduction and their protection and management are mandated by the ESA. In over
300 references used by Buehler (2000), approximately 20 focused specifically on nest
site selection and nesting habitat. Of these, nesting habitat has been described primarily
in the Pacific Northwest (Anthony and Isaacs 1989, Garrett et al. 1993), Northeastern
U.S. (Andrew and Mosher 1982, Livingston et al. 1990) and Southeastern U.S. (Wood et
al. 1989, Hardesty and Collopy 1990, Buehler 1995).

Nesting habitat studies also concentrate on variables associated with individual
nests where breeding activity occurs. This is certainly legitimate and even necessary
when most data used in these analyses are collected over a period of 2 — 4 years. Over
time though, bald eagle breeding pairs build up clusters of nests, all of which may be
maintained and reused. These clusters are referred to as breeding areas (BAs) that are by
definition “the local area associated with one territorial pair of eagles and containing one

or more nest structures” (Postupalsky 1983). This “local area” may contain habitat

38

variables beyond an individual nest site that may still factor into nest habitat suitability.
Furthermore, some state’s monitoring programs track production by nest, but also
maintain information on which nests are associated to form BAs (Michigan, New York,
Ohio; D. Best, personal communication).

Because eagle pairs will alternate nests over time, some types of data may also be

better analyzed on the level of BA (contaminants, productivity; D. Best, personal

“5'."

communication), especially when analysis is done over several years or several different

time periods. Truly linking these variables with habitat components of the same spatial

..3
‘D

extent is typically not done, however, for two reasons. First, a fairly long-terrn dataset
for these variables is required to take into account the area associated with alternate nests
that are built over time. Second, there are no real geographic boundaries for BAs since
only an active nest is actively defended in terms of what is commonly thought of as a
“territory”.

Some bald eagle research in the Great Lakes region, particularly Michigan, have
been conducted in relation to BAs rather than nests, but still have not been linked to
broader scale habitat variables such as land cover data. Furthermore, these analyses have
concentrated on contaminant effects (Bowerman 1993; Bowerman et al. 1994, 1998,
2000), response to human disturbance (Mathisen 1968, Grubb et al. 1992, Grubb and
Bowerman 1997) and microhabitat variables such as nest tree characteristics (Grewe
1980, Mathisen 1983, Bowerman 1993).

In 1961, an annual statewide monitoring program was initiated in Michigan to
address concerns of a dramatically declining eagle population. This program currently

provides over 40 years of information on eagle nest location and production. This data

39

set, coupled with the increasing availability of digital spatial data, provides an excellent
resource in determining both spatial and temporal patterns of land use / land cover
associated with bald eagle nesting habitat and productivity, and makes analyses possible
for a large geographic area. This study is an opportunity to analyze nesting habitat during
two time periods on a BA level. Specific objectives were to develop a method of
delineating bald eagle BAs and use this to determine patterns in land cover associated
with habitat preferences in BAs, temporal changes in BA land cover and differences in
land cover between active and inactive, and productive and failed BAs on a regional

scale.

STUDY AREA

Because standard, annual data collection on bald eagles did not begin until after
the species had become restricted to northern Michigan, the majority of breeding
locations are distributed throughout the NLP and UP. To conduct analysis on this broad a
spatial scale and include a temporal component, regional land cover data was needed for
more than one time period. The availability of existing GIS coverages of this nature
focused the study on the NLP (Figure 4). The most common land cover types in this
region are cropland, northern hardwood forests and aspen / white birch forests, which

together comprise approximately 45% of the region’s land cover.

METHODS
Data Preparation and Breeding Area Delineation

Regional land use / land cover data for two time periods was obtained from the

MDNR Wildlife Division. The 1978 dataset is a 124,000 scale statewide coverage

40

 

 

 

 

 

 

 

 

 

 

Fig. 4. Study area for bald eagle breeding areas in the northern lower peninsula of
Michigan.

41

developed from color-infrared and black-and-white aerial photography in the MIRIS
classification scheme. The 1993 coverage is for the NLP in the Gap Analysis Project
(GAP) classification scheme. This dataset was obtained as a grid deve10ped from 30 m
LandSat satellite imagery. The grid was converted to a polygon coverage using Arc /
Info, and a crosswalk was developed between the land classification systems to convert
MIRIS land cover codes to equivalent GAP land cover codes.

To analyze land cover associated with BAs rather than individual nests, BA
boundaries had to be established. However, there are several difficulties in determining
this area. Due to the dynamic nature of nest building and use within a BA from year to
year, there are no true spatial boundaries for an entire BA. Only one nest is used per year
but the number of alternate nests existing and which one is used in any given year can be
highly variable. All alternate nests associated with a breeding pair of eagles are
unavailable for use by another pair, but it is the active nest that will be the most actively
defended. Inactive nests within active BAs may still be used for perching.

Determining a spatial reference associated with BAs on which to construct a
buffer also posed problems. If only nests existing in one year are used, then variability in
nest use is not taken into account, and some established BAs might not have any nest
structures present in that year to provide a spatial reference. To overcome this, nests
existing at any time within a 5-year period centered on the land cover year (1976 — 1980
and 1991 — 1995) were used to establish spatial references for BAs. ArcView 3.2 was
used to clip the geographic extent of the NLP from a statewide digital point file of all nest
locations (Appendix I). Production records were used to limit the selected nests to those

existing at some time within the two S-year periods. Because the number of nests varies

42

 

among BAs, simply using existing nests to establish a preliminary boundary (i.e.,
minimum convex polygon method) for buffering would cause inconsistencies in final BA
area. To maintain consistency in the area of land cover analyzed, each BA was buffered
around a single point. To avoid centering the buffer on a single nest when multiple nests
may be used over a 5-year period, a weighted center for each BA was used.

The number of years each nest existed and the number of years each was active
within the corresponding time period were calculated. An activity weight was calculated
based on the number of years a nest existed out of 5 years possible and the number of
years a nest was active for the time it existed:

Weight = (# years existing / 5) + (# years active / # years existing)

Using this calculation alone led to instances of nests that were present but inactive all 5
years receiving a greater weight than nests that were active but only present for 1 or 2
years. To ensure that completely inactive nests never received a greater weight than that
of active nests, a correction factor was added to nests that were inactive for the entire
time they were present. Since no references could be found regarding time allocation
between alternate nests, a correction factor of 0.5 was chosen to ensure that completely
inactive nests were weighted less than any active nests. The initial weights of these
inactive nests were essentially reduced by half.

Weight factors were calculated for the selected nests and added to the attribute
table. The ArcView extension Weighted Mean of Points v 1.1 (J enness 2000) was used
to determine a weighted center (activity center) of each BA. Because there is no
established size for BAs, the activity centers were buffered with a circular buffer of 5 km

radius (Figure 5). This distance would likely include all alternate nests as well as

43

 

 

0 Activity center

® Alternate nests

Fig. 5. Bald eagle breeding buffer showing alternate nests, nest activity weights
(# above nests), activity center and buffer boundary.

potential foraging area where land cover may also be an important factor. It has also
been noted that some eagles in Michigan have been seen foraging up to 8 km from the
nearest nest (W. Bowerman, Clemson University, personal communication), and
Swenson et a1. (1986) found that a characteristic of BAs in the Yellowstone ecosystem
was the presence of open water within 5 km. The final BA coverage consisted of 5 km
BA boundaries centered on activity centers.
GIS and Statistical Analysis

Land cover was summarized by 32 GAP land cover classes for the entire study
area, and percent total area was calculated for each class for each year (1978 and 1993).
BA land cover was determined by using Arc / Info (ESRI, Redlands, California) to
intersect each 5 km BA buffer coverage with its respective year’s land use / cover
coverage. The resulting BA land cover was summarized by GAP land cover class and
percent area of each class was calculated for each BA. Several classes were grouped for
analysis using BAs only (temporal, activity and production) to allow better conversion of
MIRIS classes to GAP classes. This involved grouping “lowland evergreen needle-
leaved shrub” with “shrubland”, and “lowland broad-leaved deciduous shrub” and
“lowland broad-leaved evergreen shrub” with “other lowland shrub”, resulting in 29 GAP
classes (Table 7). Two GAP classes (northern hardwood/conifer and mixed lowland
conifer/hardwood) had no equivalent MIRIS classes, so the 1978 time period had only 27
GAP classes. For analyses, a value of zero was added to these two classes for all 1978
BAs. Because some BA buffer boundaries were cut off by the Great Lakes shoreline, the
difference between the area of these BAs and the area of a full 5 km radius buffer was

calculated and added to the water cover class.

45

Table 7. Grouped GAP land cover classes and number of bald eagle breeding areas
(BAs) in the northern lower peninsula of Michigan with land cover class occurring

 

 

 

 

within BA.
No. of BAs

1978 1992
GAP class description (11 = 42) (n = 107)
High intensity urban 41 105
Low intensity urban 21 29
Extractive 30 76
Agricultural crops 33 107
Orchard/vineyard 1 1 10
Herbaceous openland 42 107
Shrubland 42 104
Other broad-leaved deciduous forest 3 48
Northern hardwood 33 107
Northern hardwood/ conifer * — 80
Aspen/birch 39 107
Oak 36 99
Oak/j ack pine l 12
Other coniferous forest 27 82
White pine 20 42
Red pine 31 106
Upland jack pine 20 98
Cedar Spruce Fir 20 23
Emergent wetland/wet meadow 42 106
Other lowland shrub 42 104
Other forested wetland 29 61
Mixed lowland hardwood 42 60
Lowland jack pine 26 20
Black spruce 2 55
Northern white cedar 1 50
Mixed lowland conifer/hardwood * — 103
Barren land 6 38
Water 42 107
Urban grassland 29 80
* No MIRIS equivalent

46

BA land cover was grouped by time step and each of the 29 resulting land cover
classes were tested using a Mann-Whitney U (or = 0.10) to detect significant temporal
changes in BA land cover. BAs existing in multiple time steps were treated as separate
occurrences since land cover may have changed over time, and this analysis is based on
use of alternate nests rather than their establishment. To determine if BA land cover had
any significant affect on activity or production, activity and production categories were
assigned to BAs. Activity and production categories were based on whether each BA
was active (A) or inactive (I) in 1978 and / or 1993, and active BAs were classified as ‘
productive (P) or failed (F). Because activity and production are variable across years for
individual nests, BAs established after the year of land cover data were not used. For
example, a BA established in 1979 would only be included for its activity in the 1993
time period. A Mann-Whitney U (or = 0.10) was used to test each of the 29 land cover
classes for differences within activity and production categories.

Preference or avoidance in BA land cover was analyzed using Resource Selection
Analysis Software (Leban 1999). Because difficulties converting MIRIS classes to GAP
classes led to questionable results in temporal and activity analyses, only 1993 BA land
cover composition was used. All 32 GAP classes were used without grouping. Land
cover composition of the entire NLP study area was designated as available resource.
The data matrix was prepared by pairing each BA composition (used resource) with NLP
composition (available resource). This was entered into the software and analyzed using
the Neu method (N eu et al. 1974; a = 0.10), which combines the use of chi-square
goodness-of-fit analysis and construction of Bonferroni simultaneous confidence

intervals to represent actual proportion of resource used. The result is preference,

47

avoidance, or neither for each cover class based on where the predicted use value falls in
reference to the confidence interval. For instance, if the lower confidence limit of used
resource is greater than the predicted value, that class is considered preferred. Great
Lakes shoreline BAs were used without adding Great Lakes composition, since
establishing a boundary beyond the Great Lakes shoreline to delineate as associated

available resource would be somewhat arbitrary.

RESULTS

One-hundred forty-nine BAs existed in the NLP during the given time periods.
All BAs were used for temporal analysis, 42 in 1976-1980 and 107 in 1991-1995. Only
141 BAs existed in the same years as the land cover data, leaving 110 active and 31
inactive BAs for activity comparisons. For production comparisons, 64 BAs were
productive and 46 failed. After conversion of MIRIS land cover classes to GAP classes,
30 GAP classes occurred in 1978 and all 32 classes occurred within the NLP and BAs in
1993. Of all cover classes occurring within the total area of BAs in either time step, most
were present in approximately two-thirds of BAs. Water, herbaceous openland,
Shrubland, emergent/wet meadow, forested wetland and mixed lowland hardwood
occurred in all BAs in 1978. By 1993, the classes occurring in all BAs were water,
aspen/birch, northern hardwood, herbaceous openland and agricultural crops (Table 7).

Aspen / birch, northern hardwoods and agriculture occur in the greatest
percentage across the NLP, while aspen / birch and oak are greatest among the BAs.
Seventeen cover classes occurred at >2% of total BA area in at least one time period

(Table 8). All classes fluctuated over time in both the NLP and BAs. For most classes,

48

Table 8. Percent land cover composition of study area and combined area of bald
eagle breeding areas (BA) in the northern lower peninsula (NLP) of Michigan.

 

 

 

 

NLP BA
GAP code description 1978 1993 1978 1993
High intensity urban 2.83 3.45 2.64 2.62
Agricultural crops 14.01 18.12 2.93 8.18
Herbaceous openland 7.60 6.46 3.89 5.82
Shrubland 4.44 1.67 3 .27 1.72
Northern hardwood 14.11 17.44 5.91 11.34
Aspen/Birch 14.71 11.87 17.48 14.05
Oak 8.94 3.86 15.70 5.87
Other coniferous forest 2.71 1.25 3.91 1.61
White pine 2.98 0.05 6.10 0.05
Red pine 5.69 4.26 6.48 4.70
Upland jack pine 1.54 6.15 1.98 9.40
Emergent wetland 0.79 1.44 1.57 2.36
Lowland shrub 2.88 3.12 4.23 3.08
Mixed lowland hardwood 7.40 1.84 6.23 1.70
Lowlandjack pine 1.91 0.12 4.10 0.10
Mixed lowland conifer/hardwood 0.00 7.95 0.00 12.03
Water 3.40 5.54 6.49 6.93
classes <2.0% of total area 4.06 5.41 4.14 5.88

 

49

temporal changes in BAs mimicked changes in the NLP. While most classes increased or
decreased similarly over time in both the NLP and BAs, changes in BAs tended to be
more extreme.

Two classes showed temporal changes in BAs that were the reverse of their
change in the NLP. Herbaceous openland, while decreasing slightly in the NLP,
increased in BAs, and lowland shrub decreased in BAs while increasing in the NLP
(Table 8). Amount of water in combined BAs remained the same from 1978 to 1993
even though it increased in the NLP. Amount of agricultural cropland in BAs remained
less than half the percent it occurred in the NLP, but the amount of increase in BAs was
greater than in the NLP (Table 8). Upland jack pine also increased to a greater degree in
BAs, and went fi'om approximately equal proportions in the NLP and BAs in 1978 to
50% greater in BAs than the NLP in 1992. Oak decreased over time in both the BAs and
NLP, but dropped to a greater extent in BAs. High intensity urban increased only slightly
in the NLP and remained unchanged in BAs (Table 8).

Twenty-one classes showed significant (P < 0.10) changes over time in BAs. Of
these, 11 classes increased in percent area from 1978 to 1993, and the remaining 10
decreased (Table 9). Most lowland and wetland classes were among those that decreased
in area. Those that increased included upland, mixed conifer/hardwood, herbaceous
openland, urban and agriculture classes. Eight classes were different between activity
groups (P < 0.10), with mixed hardwood/conifer classes among those with greater area in
active BAs. Orchard, lowland hardwood and lowland jack pine comprised a greater area
of inactive BAs (Table 10). Differences between production categories occurred in red

pine, black spruce and northern white cedar, which made up a greater percent area in

50

Table 9. Mean percent land cover by cover class within 2 time steps for bald eagle
breeding areas in the northern lower peninsula of Michigan.

 

 

 

Time Step

GAP class description 1976-1980 1991-1995 P

High intensity urban 2.64 2.65 0.91
Low intensity urban 0.26 0.39 0.04
Extractive 0.10 0.45 0.30
Agricultural crops 2.93 8.26 <0.01
Orchard / vineyard 0.35 0.09 0.01
Herbaceous openland 3.89 5.88 <0.01
Shrubland 3.27 1.73 0.01
Other broad-leaved deciduous forest 1.58 0.25 <0.01
Northern hardwood 5.91 11.45 <0.01
Northern hardwood / conifer 0.00 3.23 <0.01
Aspen / birch 17.48 14.19 0.14
Oak 15.70 5.92 <0.01
Oak / Jack pine 0.01 0.29 0.08
Other coniferous forest 3.91 1.63 0.15
White pine 6.10 0.05 <0.01
Red pine 6.48 4.74 0.12
Upland jack pine 1.98 9.49 <0.01
Cedar Spruce Fir 1.04 0.13 <0.01
Emergent wetland / wet meadow 1.57 2.38 0.13
Other lowland shrub 4.23 3.11 <0.01
Other forested wetland 0.55 0.18 <0.01
Mixed lowland hardwood 6.23 1.72 <0.01
Lowland jack pine 4.10 0.10 <0.01
Black spruce 0.02 0.41 <0.01
Northern white cedar 0.01 0.16 <0.01
Mixed lowland conifer/hardwood 0.00 12.14 <0.01
Barren land 0.02 0.16 <0.01
Water 9.42 8.64 0.69
Urban grassland 0.20 0.19 0.46

 

51

Table 10. Mean percent land cover by cover class within activity and production
groups for bald eagle breeding areas in the northern lower peninsula of Michigan.

 

 

Activity Production

GAP class description A!“ I P P + F P

High intensity urban 2.61 2.79 0.19 2.77 2.38 0.19
Low intensity urban 0.41 0.18 0.89 0.35 0.49 0.16
Extractive 0.44 0.10 0.17 0.42 0.47 0.42
Agricultural crops 6.77 5.86 0.97 6.70 6.87 0.47
Orchard / vineyard 0.08 0.47 0.05 0.07 0.10 0.13
Herbaceous openland 5.46 4.79 0.47 5.73 5.08 0.86
Shrubland 2.07 2.38 0.42 1.76 2.49 <0.10
Other broad-leaved deciduous forest 0.54 1.04 0.45 0.20 1.02 0.62
Northern hardwood 9.79 10.19 0.71 9.63 10.00 0.25
Northern hardwood / conifer 2.20 1.66 0.02 2.02 2.45 0.71
Aspen / birch 15.00 16.09 0.81 15.00 15.00 0.97
Oak 7.95 11.36 0.48 7.06 9.19 0.80
Oak / jack pine 0.14 0.39 0.31 0.20 0.06 0.32
Other coniferous forest 2.28 2.33 0.69 2.38 2.13 0.63
White pine 1.24 3.81 0.46 0.66 2.06 0.31
Red pine 5.56 3.81 0.26 6.75 3.90 0.06
Upland jack pine 8.15 3.67 <0.01 9.75 5.91 0.19
Cedar Spruce Fir 0.41 0.30 0.62 0.13 0.79 0.03
Emergent wetland / wet meadow 2.31 1.78 0.80 2.12 2.57 0.46
Other lowland shrub 3.48 3.60 0.14 3.15 3.96 0.33
Other forested wetland 0.29 0.26 0.39 0.26 0.35 0.52
Mixed lowland hardwood 2.70 4.66 0.03 2.13 3.49 0.26
Lowland jack pine 1.13 1.89 0.01 1.32 0.86 0.43
Black spruce 0.35 0.12 0.05 0.47 0.19 0.05
Northern white cedar 0.15 0.02 0.05 0.22 0.05 0.03
Mixed lowland conifer / hardwood 9.31 5.88 <0.01 9.46 9.09 0.32
Barren land 0.15 0.05 0.35 0.13 0.17 0.12
Water 8.84 10.34 0.70 9.00 8.62 0.34
Urban grassland 0.20 0.20 0.22 0.15 0.29 0.91

 

* Activity groups: A — active, I — inactive

+ Production groups: P - productive, F - Failed

52

productive BAs. Shrubland and cedar/spruce/fir occurred in greater percent area in failed
BAs (Table 10).

Resource selection analysis resulted in 7 preferred classes including mixed
northern hardwood/conifer, aspen/birch, oak, upland jack pine, emergent/wet meadow,
mixed lowland conifer/hardwood and water (Table 11). Agricultural cropland, orchard
and northern hardwood were avoided (P < 0.10). Preference and avoidance were
indicated but not significant in 5 other classes, including urban, extractive, black spruce

and other conifers (Table 11).

DISCUSSION

One of the clearest patterns from this analysis is that BA land cover composition
is representative of land cover composition of the NLP and changes over time are similar
for both areas. This is possibly due to sampling effect since the larger buffer distance (5
km radius) used for BAs translates into a circular area of 78 kmz. Over all BAs, this is a
large area and accounts for a greater proportion of the study area than using individual
nest areas (generally 500 meters). While this needs to be taken into consideration, some
results are notable.

While temporal changes of land cover in BAs tended to occur in the same
direction as in the NLP, there were two exceptions. Herbaceous openland decreased
slightly in the NLP, probably due to succession. There was no preference for this class,
but it increased by nearly 50% in BAs from 1978 to 1993, and was present in nearly the
same proportion as in the NLP in 1993 (Table 8). This disagrees with the general

preference for forested areas but is probably a result of the increasing number of BAs

53

Table 11. Habitat selection of nesting bald eagles in the northern lower peninsula
of Michigan, determined by confidence intervals of proportion of land cover

composition within 5-kilometer breeding areas representing used resource.

 

 

Lower Upper Proportion Resource
GAP class description CL CL Available Selection P
High intensity urban 0.022 0.032 0.035 Avoid N.S.*
Low intensity urban 0.002 0.006 0.005
Extractive 0.003 0.007 0.002 Prefer N.S.
Agricultural crops 0.075 0.093 0.181 Avoid 0.0001
Orchard / vineyard 0.000 0.002 0.007 Avoid 0.0500
Herbaceous openland 0.053 0.067 0.065
Shrubland 0.013 0.021 0.016
Other broad-leaved deciduous forest 0.001 0.004 0.004
Northern hardwood 0.107 0.126 0.174 Avoid 0.0001
Northern hardwood / conifer 0.027 0.038 0.020 Prefer 0.0001
Aspen / birch 0.133 0.155 0.119 Prefer 0.0500
Oak 0.053 0.068 0.039 Prefer 0.0001
Oak / jack pine 0.001 . 0.005 0.002
Other coniferous forest 0.013 0.021 0.013 Prefer N.S.
White pine 0.000 0.001 0.001
Red pine 0.042 0.055 0.043
Upland jack pine 0.087 0.106 0.062 Prefer 0.0001
Cedar Spruce Fir 0.000 0.003 0.001
Emergent wetland / wet meadow 0.019 0.029 0.014 Prefer 0.0001
Other lowland shrub 0.003 0.008 0.005
Lowland broad-leaved deciduous shrub 0.020 0.029 0.025
Lowland broad-leaved evergreen shrub 0.000 0.003 0.001
Other forested wetland 0.001 0.003 0.002
Mixed lowland hardwood 0.013 0.022 0.018
Lowland jack pine 0.000 0.002 0.001
Black spruce 0.002 0.006 0.002 Prefer N.S.
Northern white cedar 0.000 0.003 0.001
Mixed lowland conifer/hardwood 0.113 0.134 0.080 Prefer 0.0001
Barren land 0.000 0.003 0.004 Avoid N.S.
Water 0.063 0.079 0.055 Prefer 0.0500
Urban grassland 0.001 0.003 0.005 Avoid N.S.
Lowland needle-leaved evergreen shrub 0.000 0.001 0.000

 

* N.S. - Selection indicated but not significant

54

 

 

covering more of the total NLP area. The fluctuations of this class are probably due to
sampling effect rather than changing preferences. However, it could also be indicative of
pairs settling farther from optimal habitat as the breeding population approached
saturation. Lowland shrub also showed a reverse temporal trend in BAs, but only
changed minimally in both BAs and the NLP (Table 7). There were also difficulties in
assigning this GAP class to the 1978 MIRIS classes and most likely explains the
inconsistency.

Significant temporal changes in land cover composition of BAs seems to show a
slight trend away fi'om general nesting habitat preferences of forested areas near water
(Buehler 2000) with little disturbance (Grubb et al. 1992, Grubb and Bowerman 1997).
This typically involved a decrease in lowland classes accompanied by an increase in
upland and urban classes (Table 9), but is probably not indicative of changing preferences
for several reasons. Intolerance of disturbance is usually exhibited within 500 meters of
active nest sites (Fraser et. al. 1985, Grubb et al. 1992, Livingston et. al. 1990) but the
BAs encompass a much larger area including land cover within 5 km. Increase in urban /
industrial and cropland within BAs is probably due more to the overall increase in these
classes in the NLP. This will not necessarily have an affect if active nests are still more
than 500 meters from these areas.

Some classes (e.g., barren land, white cedar, oak / jack pine) that increased in
BAs from 1978 to 1993 occurred at fairly low levels in both time periods and probably
have no real affect even though statistically significant (Table 9). The mixed hardwood /
conifer classes were GAP classes that had no equivalent MIRIS classes so it is impossible

to determine the true extent of the change in these areas. Of those classes that decreased

55

in BAs, most were only present at a low percent area in both time periods and probably
do not represent any true correlation. These tended to be lowland classes and may be
another indication of the increased number of breeding pairs in 1993, forcing a greater
proportion of pairs to nest farther from bodies of water.

Both oak and white pine showed a more extreme decrease in area within BAs than
in the NLP, but may again be a result of misclassification in converting MIRIS classes to
GAP classes. The MIRIS classification scheme distinguished between upland and
lowland areas predominated by these two species while the GAP classification scheme
did not. This led to some lowland MIRIS classes being converted to lowland GAP
classes without a predominant species. However, this could be indicative of succession
and land management causing relatively pure stands in 1978 to become more mixed by
1993. Either way, it is difficult to determine if there is a real relationship to bald eagle
BAs in this trend.

Of the 8 classes that occurred in greater percentages in active BAs than in inactive
BAs, only mixed lowland conifer / hardwood and upland jack pine averaged more than
2.0%. However, mixed lowland conifer / hardwood was one of the classes without any
equivalent MIRIS class. The time periods were grouped for this analysis, and there were
more active BAs from the 1993 time which would have included this class. This result is
again inconclusive as there is no way to determine to what extent this type of mixed class
occurred in the 1978 BAs. Of the 3 classes occurring in greater percentages in inactive
BAs, all averaged less than 5.0% in both activity categories (Table 10). Mixed lowland
hardwood was present to the greatest extent in inactive BAs but this could again be due to

problems in cover class scheme conversion rather than a true difference. The 1978 cover

56

classes were specific to predominant species so that classes such as lowland oak would
have been converted to the oak GAP class rather than being grouped with other lowland
hardwoods.

The same problems occurred in analyzing differences between production
categories, but are at least a bit more consistent with what would be expected. Of the 3
classes greater in productive than failed BAs, only red pine was present in any
appreciable amount (Table 10). However, it is unclear if this is due to the discrepancies
in classification schemes when there were more productive BAs from the 1993 time
period. Even if there is a real difference in this class, it is again more likely due to the
increased number of BAs being established farther from water and including more upland
area. Only two classes (Shrubland and cedar/spruce/fir) were present in greater amounts
in failed BAs. However, both averaged less than 3.0% of BA composition and probably
had no affect on productive outcome.

Because of the inconsistencies incurred by using two classification schemes, it is
difficult to determine which results are truly indicative of temporal changes or differences
in activity and production categories. Resource selection analysis results were more
consistent though. Since only 1993 BAs and land cover were used, the results can also be
considered more dependable. Preferred land cover classes are primarily those that would
be associated with lowland areas in close proximity to water and relatively undisturbed
areas (Table 11). Even for the larger scale BAs, this is still consistent with preferences in
nest site selection (Buehler 1995, Swenson et al. 1986, Mathisen 1968, Grubb et al.

1992)

The preference for mixed classes agrees with the quality of high foliage height

57

diversity and relatively open canopies that tend to characterize nesting areas (Buehler
2000). Even the preference for oak and upland jack pine are not inconsistent, as these
classes can indicate the increased upland area covered by a greater number of BAs.

These classes may also be included in areas that are forested and relatively undisturbed.
These reasons may have led to the classification of these two as preferred even without
the closer proximity to water. Most avoided classes (both significant and non-sigrrificant)
are also consistent with nesting habitat characteristics, as most are non-forested or
associated with greater human activity (agriculture, urban, barren land, urban grassland;
Table 11). Avoidance of northern hardwoods was not expected, but may be due to the
relative nature of compositional data (Aebischer et a1. 1993). Preference in other forested
classes will necessarily lead to a supposed avoidance of others. Overall, resource
selection results are consistent with both general nesting area preferences and the results

of the smaller scale analysis conducted on nest sites in the Black River Watershed.

LIMITATIONS AND MANAGEMENT IMPLICATIONS

Results of this analysis do indicate some obvious patterns, but these should be
considered cautiously due to limitations of the analysis. While the use of two time
periods for land cover allowed examination of temporal changes, the attempt to equalize
two land classification systems caused many difficulties in interpreting results. Because
of this, it is impossible to determine if true differences occurred, or if they were simply
different because of the way cover classes were converted. This means that analyses
involving both time periods should be considered inconclusive even when consistent with

what would be expected of bald eagle breeding habitat.

58

Resource selection analysis should also be considered carefully. A problem
typically encountered with compositional data is that proportional composition is not
independent. An increase in one cover class will lead to a decrease in another cover
class. For this reason, a preference in one class will inevitably lead to an apparent
avoidance of another class (Aebischer et al. 1993), so there were some concerns in using
the Neu method (N eu et al. 1974). This constraint was ultimately considered acceptable
though following a comparison of selection analysis methods conducted by McClean et
al. (1998). The authors found that the Neu method was the only method that identified
selection patterns consistent with known requirements and did so across several spatial
scales. The relatively large sample size (n=107) also adds to the power of this analysis,
in that a statistical preference or avoidance is likely to be indicative of true selection over
this number of BAs.

In spite of these limitations, there are trends that may be useful. Land cover
among BAs is less variable than among nests (BRW), but the overall habitat preferences
typically agree. For the most part, analysis of BA composition seemed to result in a
generalized version of the results indicated in individual nest composition (BRW). This
is probably a combination of the more general GAP classification scheme and the larger
area encompassed by BAs. The greater amount of area will cause BA composition to
approach the composition available across the entire NLP, so results will not be as
focused as with individual nest areas. This does not imply that this analysis cannot be
useful though. While specific nesting habitat should be managed for its affect on nest
placement, the larger BA habitat should not be overlooked. This broader area may play a

role in providing additional foraging opportunities, amount of cover, proximity of

59

potential disturbance, predator abundance and other factors that may affect successful
reproduction.

The use of BAs rather than individual nest sites adds the component of
considering all nests associated with a bald eagle breeding pair. This allows
consideration of bald eagle breeding habitat variables beyond an actively defended nest.
The method used for delineating BAs is by no means the best possible, but it can be
viewed as a starting point. Relatively simple modifications can be made that may more
accurately reflect the dynamics that occur within BAs. For example, the buffer radius
could be changed and based more specifically on home range data or the average distance
between breeding pairs. More linear buffers could be could be constructed for Great
Lakes and riverine BAs. This would allow for more accurate compositions of these areas
where eagle movements also tend to be more linear (Harmata and Montopoli 2001).

Weighting nests based on existence and activity is a fairly simple idea. However,
it shifts placement of BA boundaries based on the most consistently active area while still
accounting for all nests within the BA. It also adds the possibility of including BAs that
have been inactive for several years. This may be a better measure of habitat affects on
activity since activity / inactivity is independent at the level of BA but not for individual
nests within BAs. While the results of these analyses may be largely inconclusive, this
study offers a method of adding a spatial component to BAs rather than just using active

nests that tend to change over time.

60

LITERATURE CITED

Aebischer, N. J ., P. A. Robertson, and R. E. Kenward. 1993. Compositional analysis of
habitat use from animal radio-tracking data. Ecology 74(5): 1313-1325.

Andrew, J. M., and J. A. Mosher. 1982. Bald eagle nest site selection and nesting habitat
in Maryland. Journal of Wildlife Management 46(2):383-390.

Anthony, R. G., and F. B. Isaacs. 1989. Characteristics of bald eagle nest sites in
Oregon. Journal of Wildlife Management 63(1):]48-159.

Bowerman, W. W. 1993. Regulation of bald eagle (Haliaeetus leucocephalus)
productivity in the Great Lakes basin: an ecological and toxicological approach.
Dissertation, Michigan State University, East Lansing, Michigan, USA.

, D. A. Best, J. P. Giesy, T. J. Kubiak, and J. G. Sikarskie. 1994. The influence
of environmental contaminants of bald eagle (Haliaeetus leucocephalus)
populations in the Laurentian Great Lakes, North America. Pages 703-707 in
B.U. Meyburg and R. D. Chancellor, editors. Raptor conservation today. Pica
Press, Moose Jaw, Saskatchewan, Canada.

, 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
Management and Assessment 53:197-212.

, D. A. Best, T. G. Grubb, J. G. Sikarskie, and J. P. Giesy. 2000.
Assessment of environmental endocrine disruptors in bald eagles of the Great
Lakes. Environmental Management and Assessment 53:197-212.

Buehler, D. A. 1995. A geographic information system to identify potential bald eagle
breeding habitat for southeastern United States rivers and reservoirs. Proceedings
Annual Conference of Southeast Association of Fish and Wildlife Agencies
49:292-302.

. 2000. Bald Eagle (Haliaeetus leucocephalus). No. 506 in A. Poole and
F. Gill, editors. The Birds of North America. The Birds of North America,
Philadelphia, Pennsylvania, USA.

Fraser, J. D., L. D. Frenzel, and J. E. Mathisen. 1985. The impact of human activities on
Breeding bald eagles in North-central Minnesota. Journal of Wildlife
Management 49(3):585-592.

Garrett, M. G., J. W. Watson, and R. G. Anthony. 1993. Bald eagle home range and

habitat use in the Columbia River estuary. Journal of Wildlife Management
57(1):19-27.

61

Grewe, A. J. 1980. Habitat requirements of nesting populations. Pages 7-10 in T. N.
Ingram, editor and coordinator. Proceedings of the bald eagle conference on
“raptor rehabilitation”, “eagle nesting biology”.

Grubb, T. G., W. W. Bowerman, J. P. Giesy, and G. A. Dawson. 1992. Response of
Breeding bald eagles, Haliaeetus leucocephalus, to human activities in
Northcentral Michigan. The Canadian Field-Naturalist 106:443-453.

, and W. W. Bowerman. 1997. Variations in breeding bald eagle responses to
jets, light planes and helicopters. Journal of Raptor Research 31(3):213-222

Hardesty, J. L. and M. W. Collopy. 1990. Nesting ecology, habitat use, and dispersal of
the southern bald eagle on John F. Kennedy Space Center (Merritt Island National
Wildlife Refirge) and Vicinity, Florida: Phase One. Annual Report to the
National Fish and Wildlife Foundation.

Harmata, A. R, and G. J. Montopoli. 2001. Analysis of bald eagle spatial use of linear
habitat. Journal of Raptor Research 35(3): 297- 213

J enness, J. 2000. Weighted Mean of Points, Version 1.1. United States Forest Service,
Rocky Mountain Research Station. Flagstaff, Arizona, USA.

Leban, F. 1999. Resource selection for Windows 95/NT. Version 1.0, Beta 8.1.
University of Idaho, Moscow, Idaho, USA.

Livingston, S. A., C. S. Todd, W. B. Krohn, and R. B. Owen, Jr. 1990. Habitat model
for nesting bald eagles in Maine. Journal of Wildlife Management 54(4):644-653.

Mathisen, J. E. 1968. Effects of human disturbance of nesting of bald eagles. Journal of
Wildlife Management 32(1): 1 -6.

. 1983. Nest site selection by bald eagles on the Chippewa National Forest.
Pages 95-100 in D. M. Bird, editor. Biology and management of bald eagles and
osprey. Harpbell Press, St. Anne de Bellevue, Quebec, Canada.

McClean, S. A., M. A. Rumble, R. M. King, and W.L. Baker. 1998. Evaluation of
resource selection methods with different definitions of availability. Journal of
Wildlife Management 62(2):793-801.

Neu, C. W., C. R. Byers, J. M. Peek. 1974. A technique for analysis of utilization —
availability data. Journal of Wildlife Management 38(3):541-545.

Postupalsky, S. 1983. Techniques and terminology for surveys of nesting bald eagles.
In Grier et al. Northern states bald eagle recovery plan. U. S. Fish and
Wildlife Service, Denver, Colorado, USA.

62

Swenson, J. E., K. L. Alt and R. L. Eng. 1986. Ecology of bald eagles in the greater
Yellowstone ecosystem. Wildlife Monographs 95.

Wood, P. B., T. C. Edwards, Jr., and M. W. Collopy. 1989. Characteristics of bald eagle
nesting habitat in Florida. J oumal of Wildlife Management 53(2):441-449.

63

CHAPTER 4
LAND USE / LAND COVER PATTERNS AND HABITAT MODEL FOR BALD

EAGLE NESTS IN THE NORTHERN LOWER PENINSULA OF MICHIGAN

INTRODUCTION

In pre-settlement Michigan, the bald eagle historically nested throughout the state
where suitable habitat occurred. Through the early and mid-1900's, the population
gradually declined due to loss of habitat and prey resources, direct control by humans and
low productivity due to environmental contaminants (Barrows 1912, Brewer and McPeek
1991). To address this, an annual statewide monitoring program was initiated in 1961,
providing over 40 years of information on nest location and production.

The Michigan bald eagle breeding population has rapidly expanded (after removal
of the primary stressor, organochloride contaminants) but this has occurred in a landscape
much different from what had previously been available. As the landscape continues to
change, identification of unused but suitable nest habitat becomes more important.
Habitat models can be a useful tool in determining potential of existing habitat.
Bowerman (1993) developed a nesting habitat model for the Great Lakes region using
Pattern Recognition to model the search image used in aerial habitat surveys. The model
successfully classified 82% of known breeding areas as being located in good habitat, but
the model was developed to identify potential nesting habitat only along the Great Lakes
shoreline. Since this model only included area within 1.6 km of the Great Lakes

shoreline, it may not apply as accurately to inland areas. Another difficulty in applying

64

this method is that it is based on aerial surveys and the cost of flight time can be
prohibitive.

One method that may be more efficient is to model bald eagle nesting habitat
using digital land use / cover data obtained from remote sensing techniques. The current
Michigan bald eagle data set, coupled with the increasing availability of digital spatial
data, provides an excellent resource in determining spatial patterns of land use / cover
associated with bald eagle nest location and productivity. Specific objectives of this
analysis were to apply a GIS to determine landscape scale patterns of land cover
associated with bald eagle nests, and to determine differences in land cover between
productive and failed nests. A nesting habitat model was developed to determine if land
cover data derived from satellite imagery could sufficiently identify potential bald eagle

breeding habitat in Michigan on a regional scale.

STUDY AREA

Because standard, annual data collection on bald eagles did not begin until after
the species had become restricted to northern Michigan, the majority of breeding
locations are distributed throughout the NLP and UP. To conduct analysis on this broad a
spatial scale, regional land cover data was needed. The availability of existing GIS
coverages of this geographic extent focused the study on the NLP (Figure 4, Chapter 3).
The most common land cover types in this region are cropland, northern hardwood
forests and aspen / white birch forests, which together comprise approximately 45% of

the region’s land cover.

65

METHODS
Data Preparation

A GIS coverage of regional land use / land cover data for 1993 in the GAP land
classification scheme was obtained from the MDNR Wildlife Division. The ArcView
extension XTools (DeLaune 2001) was used to select nests falling within the study area
by using the clip command to clip the NLP area from a statewide nest location coverage
(Appendix I). Sample selection was further refined by using only those nests in the NLP

that existed at some time within a 3-year period centered on the land cover data year (11 =

 

211), providing a sample of NLP nests existing from 1992 to 1994. Production records
for these nests were used to determine whether each nest was productive (P) or failed (F)
for each year it existed. Because activity and production are variable across years for
individual nests, only those nests that existed in the same year as the land cover data
(1993) were used for production analysis.

To use logistic regression for modeling nest areas compared to unused sites,
random points were needed to represent unused sites. The ArcView extension Random
Points Generator (J enness 2001) was used to generate a set of 211 random points (equal
to the number of nests). Since bald eagles are known to be closely associated with major
water bodies, distance to water was also calculated for inclusion in the logistic regression
model.

A hydrology coverage was developed using a hydrology polygon coverage
created from 1983 US. Geological Survey (USGS) 1:100,000 digital line graphs (DLG).
Great Lakes shoreline was added by creating an ArcView polygon shapefile of a box

larger than the geographic extent of the entire state. This was converted to an Arc / Info

66

coverage. Using the 1978 statewide MIRIS land cover coverage to delineate the state
boundary, the Arc / Info erase command was executed to erase the extent of the state
coverage from the box. The resulting coverage provided a polygon of the Great Lakes
shoreline. This was merged with the 1983 USGS hydrology coverage to provide a
complete hydrology coverage. An AML script was used in Arc / Info to calculate the
distance from each nest and random point to the nearest body of water using the point
locations and hydrology coverage.

Since some classes are highly variable across nests and there wereover 30 land
classes, the decision was made to reduce the number of variables to 20 cover classes.
This would reduce the affect of species-predominant classes which are biased toward
pine in the GAP scheme and provide more consistent variables across nests. Similar land
cover classes were grouped to reduce the number of land cover variables to 19. For
instance, all forested and shrub wetland classes were combined into a single

forested/shrub wetland class. Distance to water was used as the last variable.

GIS and Statistical Analysis

Land cover was summarized by 32 GAP land cover classes for the entire study
area, and percent of total area was calculated for each class. Nest sites and associated
land use / cover were evaluated by buffering each nest with a circular buffer of SOO-meter
radius. This distance was chosen because most disturbance responses of nesting eagles to
humans or conspecifics are induced within 500 meters (Fraser et. al. 1985, Grubb et al.
1992, Livingston et. al. 1990). This distance is also comparable to a tertiary disturbance

buffer zone described in the management guidelines of the Northern States Bald Eagle

67

Recovery Plan (Grier et al. 1983). The ArcView extension XTools (DeLaune 2001) was
used to intersect nest site buffer zones with GAP land cover data to determine land cover
within 500 meters surrounding each nest (Figure 3, Chapter 3). The resulting nest site
land cover was summarized by GAP land cover class and percent area of each class was
calculated for each nest area. Percent land cover associated with random points was
determined using the same method as that used for nests.

To determine if nest site land cover had any significant affect on production, a
Mann-Whitney U (or = 0.10) was used to test all land cover classes for differences
between production categories. Preference or avoidance in nest site land cover was
analyzed using Resource Selection Analysis Software (Leban 1999). All 32 GAP cover
classes and 111 randomly selected nests (those used to test the logistic regression model)
were analyzed. Land cover composition of the entire NLP was designated as available
resource. The data matrix was prepared by pairing each nest area composition (used
resource) with NLP composition (available resource). The matrix was entered into the
software and analyzed using the Neu et al. (1974) method (a = 0.10), which combines the
use of chi-square goodness-of-fit analysis and construction of Bonferroni simultaneous
confidence intervals. The result is preference, avoidance, or neither based on where the
predicted use value falls in reference to the confidence interval. For instance, if the lower
confidence limit of used resource is greater than the predicted value, that class is
considered preferred.

Logistic regression was evaluated using a cross-validation procedure to determine
the predictive value of this approach. Twenty models were generated and tested using

random selections of both nests and random points. To accomplish this, approximately

68

 

half (100 of 21 l) of existing nests were randomly selected using the ArcView extension
Animal Movement (Hooge 2001). These were designated as “model” nests that would be
used for estimating land cover parameters through logistic regression. The remaining
nests were designated “test” nests that would be used to validate the model. The random
selection process was repeated with the random point set to provide a model and test set
of unused resources. All model points (nests) were designated as used (1) and the model
random points were designated as unused (0). The remaining 111 nests (1) and 111
random points (0) were used as test points to validate the corresponding model. The
entire process was iterated 20 times to develop 20 model data sets of used and used
points, each with a corresponding test data set.

Matrices were developed for each of the model data sets using distance to water
and land cover composition of each point represented by percent area of the 19 grouped
land cover classes. Each dataset was entered into NCSS 2000 (NCSS Statistical
Software, Kaysville, Utah) and run through the logistic regression analysis. The data
were analyzed using backward variable selection with a 0.20 selection cutoff and 0.50
classification midpoint and a maximum of 20 iterations. The average percent correct
classification by group (0 or 1), percent correct total classification, percent omission error
(number of nests incorrectly classified as unused out of the total number of nests) and
percent total omission error (number of nests incorrectly classified as unused out of the
total number of point classified as unused) was calculated across all 20 of the resulting
models.

Land cover composition and distance to water of each test dataset was run

through the model developed by its corresponding model points. The average was

69

calculated for these test sets for percent correct classification by group (0 or 1), percent
correct total classification, percent omission error and percent total omission error. The
averages, percent classification rates and number of times each variable occurred in the
model were examined across all iterations for both model and test data to provide a

measure of the predictive accuracy of this method.

RESULTS

Two-hundred eleven nests existed in the NLP between 1992 and 1994.
Occurrence of land cover classes was highly variable, ranging from those occurring in
only 1% of nests to those occurring in 80% of nests (Table 12). No classes occurred
within 500 m of all nests. F our classes (Aspen/Birch, emergent wetland, water and
mixed lowland conifer/hardwood) were present in at least two-thirds of the nest areas
(Table 12). Eighty-seven nests were active in 1993, 55 productive and 32 failed.
Differences between productive and failed nests occurred in 5 land cover classes (P <
0.10, Table 13). Only 2 of these classes, northern hardwood and lowland broad-leaved
deciduous shrub, averaged >5% of nest areas in either production group. The percent of
water within 500 m of nests, while not significantly related to production, was more than
5 0% greater at productive nests than failed nests.

Resource selection analysis resulted in 8 preferred cover classes (P < 0.10),
primarily those associated with lowland areas (e.g., emergent wetland, northern white
cedar, mixed lowland conifer/hardwood, Table 14). Seven classes, mostly urban,
openland and agriculture, were avoided (Table 14). Preference was indicated but not

significantly in red pine, forested wetland and lowland evergreen shrub. Avoidance was

70

Table 12. Number and percent of bald eagle nests with land cover classes occurring
within 500 meters of nests in the northern lower peninsula of Michigan.

 

 

GAP class description # of nests % of nests
High intensity urban 53 25
Low intensity urban 3 1
Extractive 7 3
Agricultural crops 105 50
Orchard / vineyard 2 l
Herbaceous openland 128 61
Shrubland 94 45
Other broad-leaved deciduous forest 33 16
Northern hardwood 128 61
Northern hardwood / conifer 80 38
Aspen / birch 168 80
Oak 1 00 47
Oak / jack pine 9 4
Other coniferous forest 64 30
White pine 6 3
Red pine 122 58
Upland jack pine 136 64
Cedar Spruce Fir 15 7
Emergent wetland / wet meadow 139 66
Other lowland shrub 34 16
Lowland broad-leaved deciduous shrub 79 37
Lowland broad-leaved evergreen shrub 23 11
Other forested wetland 24 1 1
Mixed lowland hardwood 65 31
Lowland jack pine 5 2
Black spruce 52 25
Northern white cedar 19 9
Mixed lowland conifer / hardwood 168 80
Barren land 6 3
Water 141 67
Urban grassland 6 3
Lowland needle-leaved evergreen shrub 10 5

 

71

Table 13. Mean of percent land cover by cover class within 500 meters of bald eagle
nests in the northern lower peninsula of Michigan for each production group.

 

 

GAP class description P * F P

High intensity urban 0.98 0.51 0.17
Low intensity urban 0.03 0.05 0.30
Extractive 0.32 0.85 0.28
Agricultural crops 2.12 4.51 0.09
Orchard / vineyard 0.00 0.02 0.20
Herbaceous openland 4.90 5.21 0.85
Shrubland 0.55 2.27 0.01
Other broad-leaved deciduous forest 0.12 0.40 0.16
Northern hardwood 7.40 10.32 0.03
Northern hardwood / conifer 3.41 2.63 0.41
Aspen / birch 13.00 11.94 0.73
Oak 4.33 8.31 0.67
Oak / jack pine 0.10 0.07 0.30
Other coniferous forest 1.23 0.40 0.23
White pine 0.01 0.01 0.91
Red pine 5.49 2.60 0.18
Upland jack pine 7.85 5.16 0.85
Cedar Spruce Fir 0.09 0.30 0.44
Emergent wetland / wet meadow 7.06 6.88 0.40
Other lowland shrub 0.94 1.70 0.83
Lowland broad-leaved deciduous shrub 3.28 5.82 0.07
Lowland broad-leaved evergreen shrub 0.55 0.05 0.16
Other forested wetland 0.52 0.30 0.96
Mixed lowland hardwood 2.37 1.64 0.75
Lowland jack pine 0.24 0.07 0.62
Black spruce 1.08 0.15 0.27
Northern white cedar 0.58 0.02 0.14
Mixed lowland conifer / hardwood 17.84 18.17 0.96
Barren land 0.00 0.63 0.02
Water 13.32 8.68 0.13
Urban grassland 0.23 0.05 0.30
Lowland needle-leaved evergreen shrub 0.03 0.27 0.90

 

* Production groups: P — productive, F - Failed

72

Table 14. Habitat selection of nesting bald eagles in the northern lower peninsula of
Michigan, determined by confidence intervals of proportion of land cover composition
within 500 meters of nests representing used resource.

 

 

Lower Upper Proportion Resource
GAP class description CL CL Available Selection P
High intensity urban 0.004 0.009 0.035 Avoid 0.0001
Low intensity urban 0.000 0.001 0,005 Avoid 0.0500
Extractive 0.002 0.006 0.002
Agricultural crops 0.020 0.029 0.181 Avoid 0.0001
Orchard/vineyard 0.000 0.000 0.007 Avoid 0.0001
Herbaceous openland 0.034 0.046 0.065 Avoid 0.0001
Shrubland 0.005 0.010 0.016 Avoid 0.0500
Other broad-leaved deciduous forest 0.000 0.002 0.004 Avoid N.S.*
Northern hardwood 0.073 0.089 0.174 Avoid 0.0001
Northern hardwood/conifer 0.026 0.036 0.020 Prefer 0.0001
Aspen/Birch 0.102 0.121 0.119
Oak 0.050 0.064 0.039 Prefer 0.0001
Oak/j ack pine 0.000 0.001 0.002 Avoid N.S.
Other coniferous forest 0.007 0.013 0.013
White pine 0.000 0.000 0.001 Avoid N.S.
Red pine 0.046 0.060 0.043 Prefer N.S.
Upland jack pine 0.060 0.075 0.062
Cedar Spruce Fir 0.000 0.003 0.001
Emergent wetland/wet meadow 0.051 0.065 0.014 Prefer 0.0001
Other lowland shrub 0.003 0.007 0.005
Lowland broad-leaved deciduous shrub 0.031 0.042 0.025 Prefer 0.0500
Lowland broad-leaved evergreen shrub 0.001 0.004 0.001 Prefer N.S.
Other forested wetland 0.002 0.006 0.002 Prefer N.S.
Mixed lowland hardwood 0.013 0.021 0.018
Lowland jack pine 0.000 0.002 0.001
Black spruce 0.007 0.013 0.002 Prefer 0.0001
Northern white cedar 0.006 0.011 0.001 Prefer 0.0001
Mixed lowland conifer/hardwood 0.213 0.239 0.080 Prefer 0.0001
Barren land 0.001 0.003 0.004 Avoid N.S.
Water 0.121 0.141 0.055 Prefer 0.0001
Urban grassland 0.001 0.004 0.005 Avoid N.S.
Lowland needle-leaved evergreen shrub 0.000 0.001 0.000

 

* N.S - selection indicated but not significant

73

[m

 

indicated for white pine, oak/jack pine, urban grassland and barren land. All other
classes were neither preferred nor avoided (Table 14).

Occurrence of the 19 grouped land cover classes (Table 15) used in developing
the logistic regression model was highly variable among nests. The presence of these
classes ranged fi'om lowland jack pine present in only 2% of nests to aspen/birch and
mixed lowland conifer/hardwood occurring in 80% of nests (Table 16). No classes
occurred within 500 m of all nests. Five classes (aspen/birch, emergent wetland, water,
mixed lowland conifer/hardwood and barren/open/shrubland) were present in at least
two-thirds of the nest areas (Table 16).

Distance to water was the only variable present in all models and always had a
negative coefficient, indicating that as the distance decreases the suitability of the
location increase (Table 17). All other variables had positive coefficients for most of the
models in which they occurred, so that the increased percent of these classes should
indicate greater suitability. However, model iterations 4 - 6 exhibited negative
coefficients for all variables included. The number of variables occurring in each of the
20 model iterations ranged from 6 (occurring in 4 model iterations) to 20 (occurring in 6
iterations) (Table 17). Four of the land cover variables (oak/jack pine/upland jack pine,
mixed lowland conifer/hardwood, forested/shrub wetland and emergent wetland)
occurred in >75% and lowland jack pine occurred in 50% of the 20 iterations (Table 17).
Two cover classes, aspen/birch and oak, occurred in 14 of the 20 iterations, but tended to
be included in the more complex models consisting of 10 or more variables (Table 17).
While distance to water occurred in all 20 iterations, water as a cover class occurred in

only 13 models.

74

 

 

 

5600 0:62 E058:

003% :53

E 003:...“ 5600

50:0,: 505::00 550 .05: 0::3

003::00 \600265: E0500:

65335 65330 500035: .65: :05:

6520526555 .3000 5.53055

65555 5:5 02:05:00 .58: 35:02:: 220: .5:5 55:00:: :mE

0E: :03 65—220:

600265: 6532 600:8

600265: E055: .5000: 5002006 6020076500: 5:8

0:3 60:

553

050

:0::\:0:5

266008 0025053022 E03080

:25 50555 6050—0600: 65:26— .65503 60528 00:00 58% :005520
6050—6505 65—220: .855 506506 6050—6005 65—26— 535 65:26— 00:20
60026565.:600 65:26: 600:8

03: 035.5 653: 0:3 0:05.550

5600 05:2, 805.52

008% :55

:m 003% 5600

305.2:00 55905: 02:3
5::00600265: E055 Z
83335580838:
65:05:50B_:0tw<
026050032583

0:3 :02 6533
600265: 65:26: 60:22

.2006 605076505 550600265: E05002

05: 60M
55%
050

600582
26600:: “02265302, E03083”:

65:03 :E:m\6050:0m
60036555.:600 65:20— 6052
0:3 :03 653205: :05 .500

 

820 5608 E 60620:: 8.05:0 @3430

360:: 55850: 0550: :_ 605 00505

 

.5m_:0_2 :0 235:0: 526: E055: 06 :0 5608 5:5: 3:00: 0&3 65: 5.: 600.: 559005 820 5200 65: .3 03.3.

75

Table 16. Number and percent of bald eagle nests with model land cover classes
occurring within 500 meters of nests in the northern lower peninsula of Michigan.

 

 

GAP class description # of nests % of nests
Urban / Extractive 61 29
Agricultural / Orchard 105 50
Barren / Open / Shrubland 145 69
Northern hardwood / Other broad-leaved deciduous 129 61
Northern hardwood / conifer 80 38
Aspen / birch 168 80
Oak 100 47
Oak / jack pine / Upland jack pine 136 64
White pine / Other coniferous 67 32
Red pine 122 58
Cedar Spruce Fir 15 7
Emergent wetland / wet meadow 139 66
Forested / Shrub wetland 124 59
Mixed lowland hardwood 65 31
Lowland jack pine 5 2
Black spruce 52 25
Northern white cedar 19 9
Mixed lowland conifer / hardwood 168 80
Water 141 67

 

76

 

 

 

 

0 6 cm a on o: w cu cm 0 6 N. cm 6: S N. on 6 5605 E 83252 .5 5:502
6 + + + + + + 5600 02:3 E0552
N. + + + + + - + 05:5 2055
w + + + + + + + + :E 025$ 5600
w + + + + + - + 88058 :25 \ as: 223
a + + + + + + - + 5:500 \ 600365: E0552
3 + + + + + + + - + 6:25:51 :05 \ 555
a + + + + + - + 232° \ 2332:?
a + + + + + - + 02:05:50 \ 5:5
o: + + + + + + + + + + 0:3 205.: 65304
0: + + + + + + + + - + 600365: 6:232 600:2
0: + + + + + + - + .6506 550 \ 600365: E0552
_ _ + + + + + + + - + 0:3 60.x
M: + + + + + + + + + - + 553
E + + + + + + + + + + + - + + 050
E + + + + + + + + + + - + + :02: \ :33.
6: + + + + + + + + + + + + - + + + 30650E :03 \ 6:503 ::0w5::m
C + + + + + + + + + + + + + - + + + 6:253 5:5 \ 60505.:
w: + + + + + + + + + + + + + + - + + + 600365: \ 5:300 6532 6052
0: + + + + + + + + + + + + + + - + + + 0:3 :05” 6:53: \ 0:3 205.: 050
cm - - - - - - - - - - - - - - - - - - E: 553 0: 00555
560:: 5 60555 cm a: w: E 6: m 2 E m: N: 2 o: a m N. m m m _ $20 5200 6:5..—
mm20 5:5 .5 .02 5:55:

 

.5555: :0 25:50: 532 E055:

05 E 55:05: 52:5: 550: 295 65: :0: 52» 5205000 53 53552 _060:: 550050: 2552 «:0 555m .2 2:52.

77

Distance to water averaged 741 In over all existing nests (Table 17). Of the 4
classes present in > 75% of models, only mixed lowland conifer / hardwood had a
relatively high mean percent area (Table 18). The other 3 classes had only 5 — 7 % mean
percent areas (Table 18). Water as a class and aspen / birch also were also among the
highest in mean percent area, but occurred in only 60 — 65 % of models (Table 17).
Lowland jack pine had the lowest mean percent area over all nests but occurred in half
the models (Table 17). Mixed lowland hardwood and northern hardwood / other
deciduous also occurred in half the models (Table 17).

P-values for distance to water were < 0.01 in all models. Land cover variables
tended to fluctuate around 0.10 across all models in which those variables occurred
(Table 19). On average, the models correctly classified 79% of all locations (nests and
random points) and 83% of nests used to construct the models. This left 17% of model
nest locations incorrectly classified as unused locations, and of all locations classified as
unused, 19% were actually nests (Table 20). When the models were applied to test
locations (nests and random points not used in model development), the model correctly
classified 75% of all locations and 79% of nests (Table 20). For the test locations, 21%
of nests were incorrectly classified as unused and 22% of all locations classified as
unused were actually nests (Table 20). Random points classificationrates were slightly
lower than nests with 75% of the model locations and 72% of the test locations correctly

classified (Table 20).

DISSCUSSION

Bald eagle nest area land cover was highly variable in the NLP when considering

78

Table 18. Mean and median percent land cover of grouped land cover classes and
distance to water for bald eagle nests existing in the northern lower peninsula of

Michigan, 1992-1994.

 

 

 

GAP class description mean median
Urban / Extractive 1.28 0.00
Agricultural / Orchard 3.06 0.00
Barren / Open / Shrubland 5.23 1.50
Northern hardwood / Other broad-leaved deciduous 9.14 0.95
Northern hardwood / conifer 2.92 0.00
Aspen / Birch 1 1.22 5.00
Oak 4.86 0.00
Oak / jack pine / Upland jack pine 6.13 1.56
White pine / Other coniferous 0.94 0.00
Red pine 4.54 0.58
Cedar Spruce Fir 0.10 0.00
Emergent wetland / wet meadow 6.80 1.61
Forested/ Shrub wetland 5.40 0.46
Mixed lowland hardwood 2.04 0.00
Lowland jack pine 0.09 0.00
Black spruce 0.80 0.00
Northern white cedar 0.90 0.00
Mixed lowland conifer / hardwood 20.53 13.88
Water 14.01 5.15
Distance to water 741.22 327.10

 

79

:3 v as“; 9:865 I ._

 

 

 

Ed N_d m _d m _d mod Ed :88 223 80532

Ed N_ d m E d m _ d mod 3 d w E d 88% xoflm

Ed NEd mod 2d m _d E d mod Ed h”— oopam 830

Ed Ed 2 d m _d mod mod :d Ed 38888 BEOBEQ 8.53

Ed Ed 8d m _ d m _ d mod Ed NE d 3d hobcoohcooBEE 505.52

E d N E d o E d m E d m _ d mod cod _ _ d w E d van—nEmEumoEobmm

Ed N E d m E d m _ d mod cod 2 d ddd w _ d EEEOBBH—zotw’x

E d N E d m E d m E d wdd :d E d E d w E d u>cumbxo>8€3

E d odd m _ d m E d wod wdd odd w E d o E d can x3“ 2233

Ed N E d mod n E d 2 d wdd :d 3 d w E d 80329. 98—32 @322

Ed ONd Ed m _d m E d wdd ddd 2d 8d w _ d .2qu 35.568323 .552

Ed N E d odd m E d m E d mod nod :d 3 d w _ d ofiq vex

Ed mod N _ d mod Ed m E d m E d mod odd 3d :d n E d m _ d 835

add Ed 3d Ed Ed 2d Ed 3d Ndd mod 2d Ed mod Ed xuo
3d Ed 5d Ed vod .n _ d m E d mod mod ddd :d E d w _ d coaguama‘
E d -- N E d mod mod n E d m _ d mod Nod mod Ndd N _ d d E d w _ d mod 38on 83%2283 .wBEm

5d Ed 5d Ed -- m _ d m _ d Ed Ndd -- wdd nod Ed Ndd w _ d mdd van—Ho? £57. \ “6835..”
-- Ed -- N—d .. -- 3d m _ d -- -- -- wdd ddd :d -- E d vod @8332 toquo .32 352
m _d Ed vod NEd ddd 8d 2d Ed 8d .. vod mod Ndd Ed N—d mod Ed 3d can 32:. gingham x8330
-- -- -- -- -- -- -- -- -- -- -- -- -- -- -- : -- -- _. -- e5 $33 2 3.865
E M: E 2 E E E N— I E o m N. d w v m N E 830 B>oU v55

nouns:
.cwwEomE do 22.0.:qu

$32 E058: on“ E 36:83 “833 wcumo: 2mg 23 8d 82m>¢ 032.2? 388 5:32on Biwe— do $883 .E 038.

80

 

 

NR. mu 33:3

NN EN mm E. E E. mm 533
:ommmmEo :88 :39 38w as 8:0 :39 9.5% .3
539 .x. 865:: .x. 89:8 c\o 82.80 ex. :owmmmfio X. 82.80 3 8280 .x.

 

 

HOW Mumm— “worm.

How «:5 8:03.:ng 3on

 

.53522 do 535:0: .632 505.5:

05 :5 522538 2052: 3:53 wfiuo: 2mg 23 :5 8:0 :ofiflEo 5:: :oumowfimfio 5:38: :32 dN 25d.

81

all 32 GAP cover classes (Table 12), but also when only 19 grouped classes were
considered (Table 16). However, in both cases the same 4 classes were consistently
present in the majority of nests. Given that nesting bald eagles are known to be closely
associated with water (Buehler 1995, Swenson et a1. 1986), it was expected that the water
cover class would tend to be present within 500 m of nests. The other 3 classes are also
consistent as emergent wetland, mixed lowland conifer/hardwood and aspen/birch are
typical of areas surrounding bodies of water in the NLP.

Of the 5 cover classes present in different proportions between productive and
failed nests, all occurred at higher levels in failed nests. Agricultural cropland, Shrubland
and barren land occurred at < 5% in both categories so it is difficult to determine if the
presence of these classes had any real affect on production. However, they may represent
those nests that were placed in less than optimal areas. Agriculture near these nests may
make them more prone to disturbance and increase the potential for nest abandonment
(Fraser et. al. 1985, Grubb et al. 1992). Shrubland and barren land also occurred in low
percent area but again may be indicative of nests that are in less forested areas making
them more susceptible to disturbance or harsh weather.

The percent of northern hardwoods was also higher in failed nests but probably
has little affect on production. It seems that a mixed class of this nature would provide
the high foliage height diversity characteristic of bald eagle nest sites (Buehler 2000).
This is a very broad class that does not distinguish between upland and lowland areas
though. It is more likely that this class represents more upland area farther from water,
which would also be less suitable for nesting. While the difference in amount of water

within 500 m was not significant, productive nest areas consisted of 50% more water than

82

failed nests. This is also consistent with the tendency of eagles to nest < 2 km fi'om a
body of water (Buehler 2000), but also indicates that the actual distance can play a role.

Resource selection analysis also provided results consistent with general nest site
preferences (Buehler 1995, Swenson et al. 1986, Mathisen 1968, Grubb et al. 1992) but
characterizes habitat use more specifically for Michigan. Preferred land cover classes are
primarily those that would be associated with lowland areas in close proximity to water
and relatively undisturbed areas (Table 14). The preference for mixed classes lends the
quality of high foliage height diversity and relatively open canopies that tend to
characterize nesting areas (Buehler 2000). Even the preference for oak is not
inconsistent, as this class may also be included in areas that are forested and relatively
undisturbed. This may have led to the classification of oak as preferred even without the
closer proximity to water.

All 7 classes that were avoided represented areas prone to disturbance (urban) or
were non-forested (agriculture, Shrubland) (Table 14). Where avoidance was indicated
but not significant tended to include classes that are more typically upland areas. This
could be a result of nests occun'ing in areas that are usable but of marginal quality. The
population had still not reached saturation at this time (1993) so it cannot be said that
these areas were truly avoided. This trend may be due to a greater number of nests being
placed farther away from water as the population expanded, but does not necessarily
mean those nests were less successful. Avoidance of northern hardwoods was not
expected, but may be due to the relative nature of compositional data (Aebischer et a1.
1993). Preference in other forested classes will necessarily lead to a supposed avoidance

of others. Overall, resource selection results are consistent with both general nesting area

83

preferences and the results of the analyses conducted on nest sites in the Black River
Watershed and BAs in the NLP.

Variables consistently present in the habitat models were again consistent with
habitat preferences. These included mixed and 10wland classes that would be typical of
areas near water (Table 17). Distance to water was the only variable present in all 20
iterations of the model and with P-values < 0.01, indicating that this is probably the
strongest factor in nest site selection for bald eagles (Table 19). Average omission errors
in the model (known nests classified as unused locations), were fairly low for a model
encompassing this large a spatial extent. This can be considered acceptable when the
model was developed using nests over a short period of time. Some of those nests may
be in lesser quality habitat, but pairs using those nests may have moved to other locations
in the following years. This could be another result of population expansion as well,
since increasing numbers of pairs will be forced to settle in sub-optimal areas as the
optimal areas become saturated.

Commission errors (unused locations classified as nests) tended to be 10% higher
than omission errors. This was not considered a problem though because the population
in 1992 — 1994 had not even reached the recovery goal of 300 pairs yet. Some of those
random points misclassified could very well have been suitable nesting areas and there
just were not any pairs using them at that time. However, it would seem that if these
areas were suitable they would have been used before those areas where known nests
were misclassified as unused. It is not clear if this is due to inaccuracy of the model or
some other confounding factor. Experience can play a role in nest site selection as well

as distance to other breeding pairs. Random points classified as nests may have been in

84

suitable areas concerning land cover but the density of previously existing nest areas may
have precluded new pairs from utilizing these areas. In general, the model provided
relatively high correct classification rates for nests in both the model and test data sets. If
goodness-of-fit statistics are applied to determine a best model, this process could then be
an adequate method for developing a model for identifying remaining nesting habitat

potential in Michigan.

LIMITATIONS AND MANAGEMENT IMPLICATIONS

Results of this study indicate certain characteristics of bald eagle nesting habitat
in the NLP. However, resource selection and production analyses should be considered
carefully. A problem typically encountered with compositional data is that proportional
composition is not independent. An increase in one cover class will lead to a decrease in
another cover class. For this reason, a preference in one class will inevitably lead to an
apparent avoidance of another class (Aebischer et a1. 1993), so there were some concerns
in using the Neu method (N eu et a1. 1974). This constraint was ultimately considered
acceptable though following a comparison of selection analysis methods conducted by
McClean et al. (1998). The authors found that the Neu method was the only method that
identified selection patterns consistent with known requirements and did so across several
spatial scales.

Another problem in interpreting these results is that there is not necessarily a
causal effect. As indicated in the model, distance to water appears to be the key factor in
where bald eagles will nest. Preferred land cover classes most likely are not being

selected but are simply characteristic of lowland areas near water in the NLP. Still,

85

 

 

suitable nest trees in relatively undisturbed areas are needed for nesting. This may place
more emphasis on those areas that are avoided or that tend to occur in greater proportion
in failed nests. It may, therefore, be just as important to identify suitable areas based on a
lack of these classes.

Michigan’s breeding eagle population has increased very rapidly, clouding the
line between intoleranCe or simply taking what is available. Timing and type of
disturbance may play a greater role in nest success than just the presence of human
activity. A nest is less likely to be abandoned later in the season, but pairs may be more
likely to accept some disturbance if it is already there when they begin nesting (S.
Postupalsky, personal communication). This does not necessarily mean they will use the
same nest again in following years if a more suitable area is available. Managers should
keep in mind that as the eagle population continues to expand, the availability of optimal
breeding areas will decrease, and as breeding pairs begin to settle in sub-optimal areas,
they may experience lower reproductive rates. If the population experiences another
serious decline caused by increased mortality (e. g., disease factors, environmental
contaminants) optimal habitat may play an important role in maintaining reproductive
success.

Whether the population experiences another dramatic decrease or continues
expanding until usable habitat is saturated, a method for identifying suitable nesting areas
will be a useful tool in management. The cross-validation procedure used for this study
can provide direction in nest habitat identification, but these specific models should not
be applied at this time. These models were based on 1993 data and cannot be accurately

attributed to present conditions. Instead, they should be seen as a test of a method that

86

can be performed efficiently while still providing sufficiently accurate results. Using
only land cover classification and distance to water correctly identified approximately
80% of existing nests. This is comparable to Bowerman’s (1993) pattern recognition
model that used aerial habitat survey.

Since it would be difficult to apply a model fi'om this study to land cover in a
different classification system, it is recommended that the process be repeated first based
on the newer land cover data that has only recently been released for Michigan. Until
updated data can be obtained, the information gained from these models can still be used
as a supplement to resource selection results to point out those land cover classes that
appear to have the most affect on nest site suitability. For now, the results provide
confirmation that distance to water is the most important factor, and that nests average ~
750 m from the nearest water body. Areas matching this description can then be further
investigated for presence or absence of land cover classes that are preferred or avoided.
Results from production analysis can also provide greater insight in identifying those
areas that are not only usable but are the most suitable.

Overall, a logistic regression model based on land use / cover data can provide an
accurate means of identifying bald eagle nesting habitat over a large geographic extent,
while focusing on land cover particular to the region, but may not be applicable to current
conditions. If current land cover data can be obtained, it will more accurately represent
the land cover that exists at present. Coupled with a nesting population that is now
nearing habitat saturation, a more recent modeling process should provide a better
measure of what is currently happening and provide better results in identifying

remaining suitable habitat.

87

LITERATURE CITED

Aebischer, N. J ., P. A. Robertson, and R. E. Kenward. 1993. Compositional analysis of
habitat use from animal radio-tracking data. Ecology 74(5):1313-1325.

Barrows, W. B. 1912. Michigan bird life. Michigan Agricultural College, East
Lansing, Michigan, USA.

Bowerman, W. W. 1993. Regulation of bald eagle (Haliaeetus leucocephalus)
productivity in the Great Lakes basin: an ecological and toxicological approach.
Dissertation, Michigan State University, East Lansing, Michigan, USA.

Brewer, R., and G. A. McPeek. 1991. Bald eagle - species account. Page 160 in R.
Brewer, G. A. McPeek, and R. J. Adams, editors. The atlas of breeding birds of
Michigan. Michigan State University Press, East Lansing, Michigan, USA.

Buehler, D. A. 1995. A geographic information system to identify potential bald eagle
breeding habitat for southeastern United States rivers and reservoirs. Proceedings
Annual Conference of Southeast Association of Fish and Wildlife Agencies
49:292-3 02.

. 2000. Bald Eagle (Haliaeetus leucocephalus). No. 506 in A. Poole and
F. Gill, editors. The Birds of North America. The Birds of North America,
Philadelphia, Pennsylvania, USA.

DeLaune, M. 2001. XTools ArcView extension. Version 6/1/2001. State Forest
Management Program, Oregon Department of Forestry. Salem, Oregon, USA.

Fraser, J. D., L. D. Frenzel, and J. E. Mathisen. 1985. The impact of human activities on
Breeding bald eagles in North-central Minnesota. Journal of Wildlife
Management 49(3):585-592.

Grier, J. W., J. B. Elder, F. J. Gramlich, N. F. Green, J. V. Kussman, J. E. Mathisen, and
J. P. Mattsson. 1983. Northern states bald eagle recovery plan. U. S. Fish and
Wildlife Service, Denver, Colorado, USA.

Grubb, T. G., W. W. Bowerman, J. P. Giesy, and G. A. Dawson. 1992. Response of
Breeding bald eagles, Haliaeetus leucocephalus, to human activities in
Northcentral Michigan. The Canadian Field-Naturalist 106:443-453.

Hooge, P. N. 2001. Animal movement analysis ArcView extension, Version 2.0.
United States Geological Service — Biologic Resource Discipline, Alaska
Biological Science Center, Glacier Bay Field Station, Gustavus, Alaska, USA.

88

 

J enness, J. 2001. Random Points Generator, Version 1.1. United States Forest Service,
Rocky Mountain Research Station. Flagstaff, Arizona, USA.

Leban, F. 1999. Resource selection for Windows 95/NT. Version 1.0, Beta 8.1.
University of Idaho, Moscow, Idaho, USA.

Livingston, S. A., C. S. Todd, W. B. Krohn, and R. B. Owen, Jr. 1990. Habitat model
for nesting bald eagles in Maine. Journal of Wildlife Management 54(4):644-653.

Mathisen, J. E. 1968. Effects of human disturbance of nesting of bald eagles. Journal of
Wildlife Management 32(1):1-6.

McClean, S. A., M. A. Rumble, R. M. King, and W.L. Baker. 1998. Evaluation of
resource selection methods with different definitions of availability. Journal of

Wildlife Management 62(2):793-801.

Neu, C. W., C. R. Byers, J. M. Peek. 1974. A technique for analysis of utilization —
availability data. Journal of Wildlife Management 38(3):54l-545.

Swenson, J. E., K. L. Alt and R. L. Eng. 1986. Ecology of bald eagles in the greater
Yellowstone ecosystem. Wildlife Monographs 95.

89

CHAPTER 5
MANAGEMENT IMPLICATIONS OF USING LAND USE / LAND COVER DATA

FOR DETERMINING NESTING HABITAT OF BALD EAGLES IN MICHIGAN

SUMMARY

In general, land cover among bald eagle nests and BAs is highly variable.
However, land cover associated with nest sites in the BRW and NLP, and BAs in the
NLP, are consistent with general habitat preferences of forested areas near (< 2 km) large
bodies of water (Buehler 2000). These preferences also do not appear to be changing
over time. Those classes that tend to occur most consistently during all time periods
analyzed (Table 3, Table 8) also tend to characterize areas that are relatively undisturbed,
near water, and predominated by super-canopy tree species (wetlands, aspen / birch,
lowland conifer, lowland hardwood) with high foliage height diversity. These classes are
also among those consistently preferred at all spatial scales, and classes that were avoided
(urban / industrial, agriculture, open / shrub land) were typical of areas that are non-
forested or prone to disturbance (Table 6, Table 11, Table 14). Although the habitat
model was somewhat variable depending on the selection of nests used, the most
consistent variables (Table 20) also agreed with general nest site preferences.

Land cover classes showing temporal trends counter to these preferences (urban,
agriculture, uplands — jack pine / oak, and aspen / birch) are also those that typically have
been the most intensively managed or subject to conversion. This could indicate a
reduction of nests located on inland lakes, possibly due to increasing development and

recreation on lakeshores. However, this could simultaneously be due to the increase in

90

the bald eagle nesting population. While eagle nesting habitat over the entire state was
not saturated by 1992, the number of nesting pairs had nearly tripled since 1978. By the
last period of this analysis, nest areas near water in the BRW and the entire NLP may
have all been occupied, forcing newer pairs to establish nests farther away fiom lake and
river shorelines.

Because of this, the temporal changes for those land cover classes should be

interpreted as being due to land management practices rather than changing habitat

 

preference of bald eagle breeding pairs. This implies that breeding pairs are not
necessarily becoming more tolerant of less suitable areas, and those using less preferred
habitat may experience a reduction in successful reproduction.

Differences in land cover between productive and failed nests (or BAs) tend to
agree with nest site preference. Failed nests had higher proportions of open land and
productive nests had higher proportions of forested area. However, in both groups of
nests, significantly different land cover classes occurred in very low proportion (Table 5,
Table 10, Table 13). Because so many other variables can affect reproduction as well, it
is difficult to say if those classes that were different were truly affecting production.
While classes higher in failed nests do not occur in any appreciable amount, these
differences may still be evidence of lower success from factors associated with more

open areas (i.e., lack of cover, increased disturbance or predation).

LIMITATIONS
Results of this analysis indicate some obvious patterns, but these should be

considered cautiously due to limitations of the analysis. While the BRW land cover time

91

series allowed examination of temporal changes, the geographic extent is limiting. The
combination of a small area (relative to bald eagle habitat use) and the reduced eagle
population led to very small sample sizes. This in turn reduces the power of statistical
analysis. This is somewhat alleviated by the quality of the land cover data. The land
cover database was developed specifically to study changes over time in this region of the
state (Rutledge 2001). Therefore, the land cover associated with eagle nests in the BRW
can be considered reliable even if it the lack of nest sites does lend itself to hypothesis
testing.

The affect of small sample size was overcome in considering the entire NLP when
analyzing BAs, and the use of two time periods allowed examination of temporal changes
as well. However, the attempt to equalize two land classification systems in this case
caused many difficulties in interpreting results. Because of this, it is impossible to really
determine if true differences occurred, or if they were simply different because of the
way cover classes were converted. But most of the temporal changes in NLP BAs agreed
with those in the BRW, adding some confidence to the results at both scales.

There was also a limitation in resource selection analysis, since proportional
compositional data was used. A problem typically encountered with compositional data is
that pr0portional composition is not independent. An increase in one cover class will
lead to a decrease in another cover class. For this reason, a preference in one class will
inevitably lead to an apparent avoidance of another class (Aebischer et al. 1993), so there
were some concerns in using the Neu method (N eu et al. 1974). This constraint was
ultimately considered acceptable though following a comparison of selection analysis

methods conducted by McClean et al. (1998). The authors found that the Neu method

92

was the only method that identified selection patterns consistent with known
requirements and did so across several spatial scales. Although this limitation should be
considered when interpreting and applying habitat selection results, consistency across
spatial scales and between nest sites and BAs reinforce those classes that appear to be
most important.

Another problem in interpreting these results is that there is not necessarily a
causal effect. As indicated in the model, distance to water appears to be the key factor in
where bald eagles will nest. Preferred land cover classes most likely are not being
selected but are simply characteristic of relatively undisturbed, lowland areas near water
in the NLP. This may place more emphasis on those areas that are avoided or tend to
occur in greater proportion in failed nests. These classes appear to be more consistent in
both spatial and temporal patterns than those classes that are preferred. It may then be
just as important to identify suitable areas based on a lack of these avoided classes near
suitable water bodies, rather than trying to identify all combinations of classes that are

preferred.

MANAGEMENT IMPLICATIONS

In spite of these limitations, there are some usefirl trends in land cover patterns
associated with nesting areas specifically in the NLP. More importantly, methods that
were explored could also prove to be useful in the future. Whether the population
experiences another dramatic decrease or continues expanding until usable habitat is
saturated, a method for identifying suitable nesting areas will be a useful tool in

management as land conversion continues to occur. The use of BAs rather than

93

 

individual nest sites adds the component of considering all nests associated with a bald
eagle breeding pair, and removes the dependency between activity / inactivity among
associated nests. This allows consideration of bald eagle breeding habitat variables
beyond an actively defended nest. The method used for delineating BAs is by no means
the best possible, but it can be viewed as a starting point. The results of BA analyses are
difficult to interpret because of different land classification schemes. However, this
method offers addition of a spatial component to BAs rather than just using active nests
to identify habitat use by breeding bald eagles.

Although based on individual nest sites, the models developed for this study can
also provide direction in nest habitat identification. However, it is not recommended that
any particular model be applied at this time. This process was based on 1993 data and
cannot be accurately attributed to present conditions. Instead, the information gained
from these models should be seen as a test of a method that can be performed efficiently
while still providing sufficiently accurate results. Using only land cover classification
and distance to water correctly identified approximately 80% of existing nests. This is
comparable to other nesting habitat models based on variables typically require site visits
and field data collection (Bowerman 1993, Livingston et al. 1990).

Overall, a logistic regression model based on land use / cover data can provide an
accurate means of identifying bald eagle nesting habitat over a large geographic extent,
while focusing on land cover particular to the region. However, to use this method in
Michigan, it is recommended that a model be constructed based on the newer land cover
data that is due for release. Using BA land cover instead of nests to develop the model

would also more adequately take into account inactive BAs and maintain an independent

94

 

 

sample. A more recent model coupled with a nesting population that is nearing habitat
saturation, should provide a better measure of what is currently happening and provide

better results in identifying remaining suitable habitat.

 

 

95

LITERATURE CITED

Aebischer, N. J ., P. A. Robertson, and R. E. Kenward. 1993. Compositional analysis of
habitat use from animal radio-tracking data. Ecology 74(5):1313-1325.

Buehler, D. A. 2000. Bald Eagle (Haliaeetus leucocephalus). No. 506 in A. Poole and
F. Gill, editors. The Birds of North America. The Birds of North America,
Philadelphia, Pennsylvania, USA.

Bowerman, W. W. 1993. Regulation of bald eagle (Haliaeetus Ieucocephalus)
productivity in the Great Lakes basin: an ecological and toxicological approach.
Dissertation, Michigan State University, East Lansing, Michigan, USA

Livingston, S. A., C. S. Todd, W. B. Krohn, and R. B. Owen, Jr. 1990. Habitat model
for nesting bald eagles in Maine. Journal of Wildlife Management 54(4):644-653.

McClean, S. A., M. A. Rumble, R. M. King, and W.L. Baker. 1998. Evaluation of
resource selection methods with different definitions of availability. Journal of
Wildlife Management 62(2):793-801.

Neu, C. W., C. R. Byers, J. M. Peek. 1974. A technique for analysis of utilization —
availability data. Journal of Wildlife Management 38(3):541-545.

Rutledge, D. T. 2001. Changes in land cover and wildlife habitats in two watersheds in

the lower peninsula of Michigan. Dissertation, Michigan State University, East
Lansing, Michigan, USA

96

 

APPENDIX I

Metadata for Michigan Bald Eagle Nest Mcflrns - Digital Point File
Title: mi_benests
Format: ArcView 3.2 shapefile and all associated files ( *.shp, *.dbf, *.shx)

Originator: Michigan State University and Michigan Department of Natural Resources
Carrie R. Steen, Graduate Research Assistant, Dept. Fisheries and Wildlife, 13
Natural Resources, Michigan State University, East Lansing, MI 48824-
1222; steencar@msu.edu (517) 353-5135
Kelly F. Millenbah, Assistant Professor, Dept. Fisheries and Wildlife, 13 Natural
Resources, Michigan State University, East Lansing, MI 48824-1222;
millenba@msu.edu (517) 353-4802

Funding: This project was funded by the Michigan Natural Heritage Program through
contributions to the Nongame Wildlife Fund.

Created: 2002

Description: Digital point file of all reported bald eagle nests in Michigan with
locational information (excluding Isle Royale). Attributes include coordinate location
when available, Public Land Survey description (PLS) and other information related to
specific nest sites.

Spatial Reference: Michigan GeoRef from Oblique Mercator projection
Scale factor at center = 0.9996
Azimuthal angle = 337.25556
False easting = 2546731.496
False northing = - 43 54009.816
Horizontal datum name = North American Datum 1983 (NAD83).

Map Units: meters

Contact: Technical questions regarding Questions regarding nest
digital nest locations: location source data:
Carrie Steen Dave Best
13 Natural Resources US. Fish and Wildlife Service
Michigan State University 2651 Coolidge Rd.
E. Lansing, MI 48824 E. Lansing, MI 49783

(517) 353-5135
steencar@msu.edu

97

 

 

Methods:

Nests with coordinate locations

Nests with coordinate locations were digitized by adding an event theme in
ArcView, referenced to the decimal degree latitude / longitude coordinates, and
converting the theme to a shapefile. It was assumed that these coordinates had
been taken in the WGS 1984 datum since that is the default setting for most GPS
units. Some coordinate locations were extrapolated from paper maps, but since
the source map data were not available, there were relatively few locations
derived by this method (~5) and WGS 1984 matches closely to NAD 1983, those
locations were treated the same as GPS locations. The datum for this shapefile
was converted using the ESRI Projection Utility, to convert from geographic
lat/long WGS 1984 to geographic lat/long NAD 1983. The shapefile was not
projected at this time — only a datum conversion was performed using a
geographic transformation of WGS 1984 (1) to NAD 1983. The NAD 1983
shapefile was added to a view in ArcView and projected to Michigan GeoRef
using the MI DNR Projection Extension (Michigan Department of Natural
Resources).

Nests previously digitized

Some nests had been previously digitized in CMAP using approximately the same
methods as mentioned above (J. Stuht 2000). These files (three regions — UP,
NLP and SLP) were converted to ArcView shapefiles by J. Wilson (USFWS, East
Lansing, Michigan), from whom they were obtained. The shapefiles, all in NAD
1927, were converted individually to Michigan GeoRef using the MI DNR
Projection Extension. The resulting regional GeoRef files were merged to form a
statewide file. The attribute table for this file was linked to the coordinate
location file, and a series of selections and queries were used to determine which
nests in the Stuht file were already in the coordinate location file. These nests
were removed fiom the Stuht file to allow the most recent coordinates to take
precedence (aerial GPS is updated by ground GPS when possible). The
remaining nests from the Stuht file were merged with the coordinate location file.

Nests not previously digitized

The original location source spreadsheet (D. Best, USFWS, East Lansing) was
reorganized as a flat file in *.dbf format, and imported to ArcView. It was linked
to the attribute table from the nests already digitized to determine the remaining
nests that had only a PLS description and had not been part of the Stuht file.
These were digitized manually in ArcView, directly in Michigan GeoRef, using
PLS township / range, section and subsection grids. For each nest, a point was
placed at the center of the smallest PLS unit provided. In some cases, if other
information was provided in the original location spreadsheet, Michigan Resource
Inventory System (MIRIS) base maps (roads, hydrology and land use / land
cover) were added to the view to gain more accurate placement. The base maps
were created in 1978, so were used only when there was reference to features that
are relatively unchanging (i.e., N of river, 100m fi'om shoreline). Once
completed, the manually digitized nest file was merged with the previous
shapefile.

98

 

 

Final shapefile
XY coordinate fields (in Michigan GeoRef coordinates) were added to the

shapefile .dbf using the ArcView extension Coordinate Tool Box Ext. 1.0
(www.csri.com). This was done by setting the view properties map and distance
units to meters, then using the extension tool “Add XY coordinates in decimal
degrees” to add coordinates in decimal format. All fields except Nest_Site,
X_Coord and Y_Coord were removed to allow replacement by the corrected
source spreadsheet (see Data Quality). The final corrected spreadsheet was
converted to .dbf format, imported to ArcView and joined by Nest_Site to the
digital shapefile attribute table. The joined table was exported as a new .dbf and
used to add a new event theme in ArcView using the Michigan GeoRef XY
coordinates. This was converted to a shapefile to produce the final digital
location file with complete and corrected attributes. The GeoRef was reproj ected
to Geographic NAD 1983 to add fields for latitude and longitude based on the
digitized position of nests. These fields were also added to the GeoRef shapefile.

Data Quality:
All coordinate locations were checked against PLS locations to eliminate original
data entry errors. If any GPS locations did not agree with the recorded PLS
position, PLS descriptions took precedence. Usually this was only a matter of the
GPS being in an adjacent subsection, and the point was nudged across the line. In
cases where the GPS was believed to be more accurate based on additional
information, the PLS description was modified in the attribute table.

Nests points taken from a Stuht file were also checked against PLS grids to detect
any locational shifts that may have occurred during re-proj ection, or conversion
from CMAP to ArcView. Nests that were not centered, but still in the smallest
legal unit provided were not re-centered. This was based on the assumption that
there was a paper map showing nest location used in the initial digitizing. A
small number of nests (~ 5) occurred in a different subsection than what was
reported in the source data. These were assumed to have been near a subsection
line originally and shifted during re-proj ection. These were nudged back across
the line, but not centered.

The next check was performed by querying nests by county, to confirm that all
locations fell within the county boundaries for which they were reported.
Discrepancies were noted, discussed with D. Best and corrected.

Nests were also checked against 1:24,000 great lakes shoreline base maps
(MDNR) and 1:100,000 base hydrology (polygon, USGS). Nests that fell
offshore or in inland lakes were moved if it was possible to keep them within the
same PLS location. In some cases, nests were not moved because locations
occurred in wetland areas that were digitized as part of a lake in the hydrology
base maps.

99

Note: Nests were placed as accurately as possible according to the recorded
location. Some may or may not fall within a lake depending on the accuracy and
resolution of base maps used (for example, not all base maps include smaller
islands or delineation between lake and swamp). This should be taken into
consideration when conducting analysis.

Nests associated with Michigan breeding areas, but actually located in Wisconsin
have PLS reported in Wisconsin township / range. For example, a nest in a
Menominee County breeding area but across the state line has a reported range of
“East” while other nests in Menominee County have reported ranges of “West”.
This is not an error, and should be read according to a Wisconsin PLS grid rather
than Michigan.

All discrepancies that were found either in coordinate locations or PLS were
reported to D. Best. Once the correct information was determined, the nest was
re-positioned in the digital file, and the information in the source data spreadsheet
was corrected. The corrected spreadsheet was used as the final source .dbf
mentioned above to create the final shapefile.

 

Final digital shanefile:
1659 nests
593 breeding areas

Not digitized:
l7 nests did not have enough information to be placed
Isle Royale nests are not included
Nests in Wisconsin or Canada breeding areas with no Michigan breeding
area designation were not included

Data Accuracy:
PLS locations were considered most accurate when there were discrepancies

between GPS and PLS. If GPS coordinates did not agree with other information,
PLS and referential information took precedence.

 

PIA Nests without GPS readings were digitized according to PLS and any other
locational information provided. When no additional reference was available,
such as distance from a river, the nest was placed at the center of the smallest unit
in the PLS description. The greatest possible error would then occur for a nest
that is actually at the comer of the smallest unit recorded, as follows:

Section - 0.7 miles (1126.5 meters)

‘A section — 0.35 miles (563.2 meters)

V4 ‘A section — 0.18 miles (289.7 meters)

100

Ground GPS: Nests with ground coordinates reported had GPS readings taken at
the nest tree and are accurate within a few meters. Those taken before 2000 when
selective availability was turned off were not always corrected and may be
approximately 200-300 meters off.

Aerial GPS: Aerial GPS readings were taken while circling a nest at
approximately 100 meters, however aerial readings taken before 2000 were not
always corrected. Due to the readings being taken in flight and selective
availability, readings before 2000 may be up to 0.5 miles (805 meters) off.
However, when these nests were checked against PLS descriptions, most fell at
least within the correct 1A section reported. Manual replacement of aerial GPS
positioned nests was usually due to the coordinates placing them in a body of
water, rather than on shore.

101

APPENDIX Il

MIRIS land cover classes occurring in Black River Watershed bald
eagle nests and grouped classes used in resource selection analysis

 

 

 

 

 

 

 

858:5 588:2-82 5:28 3

505 05:82 58885-82 5:283

5:283 820.235 5888 m 5:283

05:28.3 5:283 5058 3 5888 "2 5:28 3

89,885 8:25

0202 88>?

8:03 808m 8:03

8500 8500 8.580 5:233 888200 5888 m

8280 5:232 5022 5022 8.580 5:233 888200 5888 n—
00:80 823 - :8200 5:253 008% 82:5 88200 5:28 850 088.280 5888 "2
0:8 202. 0:8 202. 02: 088.280 5888 “2

0:8 505 0:8 505 0:5 8885200 5888 ..—

02: 5022 5022 0:5 08:82:00 58885

850 - 583582 .302 850 583582 5:233 5082585 5888 m
:88 - 583582 .302 8%]: 583582 5:233 5082585 5888 "2
28:: ~80. - 583582 .302 2:8 tom 583582 5:230..— 50>a2585 5888.5
583582 5:232 58:2 5082 583582 5:233 5082585 5888 m
8:8 58:85 :88 58:85 8:20:08 58 20:2 823 .:0%< 5082585 5888 .5

20:2 823 \ 8:8 5022 5022 82088 58 20:2 823 .853: 50302585 5888 m
28 505 28 502 80202200 5082585 5888 m

2:8 895 - 583582 .2 28:: 886 583582 :8582 5082585 5888 "2
583582 :858: 5022 5022 583582 :8582 5082585 5888 ..—
58m 588:2-82

58885-82 8800885 58885-82

838:8 58:80 83858

0503 02:00:55 82:3

52:3 x285 5.82 285 280282 85:3

88:0 :2820m 8:885 v .82 252 m .082 2522 N _0>2 2522 2 88. £522

102

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

lllllalllllljljllhllljlflljllllllllfl‘ll