INVESTIGATING MICHIGAN PRAIRIE WARBLER BREEDING TERRITORY HABITAT
COMPOSITION, WITHIN-TERRITORY RESOURCE USE, AND
FACTORS INFLUENCING OCCUPANCY
By
Katherine J. Kahl

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Fisheries and Wildlife
2012

ABSTRACT
INVESTIGATING MICHIGAN PRAIRIE WARBLER BREEDING TERRITORY HABITAT
COMPOSITION, WITHIN-TERRITORY RESOURCE USE, AND
FACTORS INFLUENCING OCCUPANCY
By
Katherine J. Kahl
Recent Prairie Warbler (Dendroica discolor) observations in Michigan are almost
exclusively from Lake Michigan early-successional dune ecosystems in the Lower Peninsula.
However, data describing breeding locations, population abundance, and habitat composition
influencing breeding territory selection are largely unknown and thus critical information needs
for this state-endangered, Neotropical migratory songbird. Understanding Prairie Warbler
habitat selection during the breeding season will better inform conservation and management
planning. Avian point counts were conducted in 9 study sites along the Lake Michigan coast in
2004 and 2005 to locate Prairie Warblers. Breeding territories were mapped using GPS and GIS.
Plant composition was quantified within breeding territories and in unoccupied areas to
determine whether geographic variation existed in vegetation associated with Prairie Warbler
breeding areas from north to south. Logistic regression was used to develop a predictive model
of Prairie Warbler occurrence. Using 1 m resolution imagery, distances from edge of territories
and unoccupied areas to 7 environmental covariates were used as variables to forecast Prairie
Warbler occupancy in the west Michigan dune ecosystems. Resource utilization functions
(RUF) were developed to explore resource use by Prairie Warblers within highest-use areas of
their breeding territories, thus relating geographic space use to environmental attributes as
portrayed by 30 m land cover data. Comparisons were made across all breeding territories and
for southern and northern sites exclusively to examine the potential for a north-to-south gradient

in resource use. Thirty-eight Prairie Warblers were detected, with high detection probability, in
5 of the 9 sites surveyed. Breeding areas tended to be clustered within sites, with breeding
territory boundaries often adjacent to each other, suggesting potential for conspecific attraction.
Results generally support previously observed relationships between Prairie Warbler occurrence
and a mixture of herbaceous and low-growing woody vegetation. However, my study was the
first to document a latitudinal gradient in Prairie Warbler habitat use related to vegetation
composition; with northern territory space use correlated with denser, woody vegetation and
southern territory space use associated with more open, grass dominated sites. Plant composition
in areas occupied by Prairie Warblers was different from plant composition in unoccupied areas.
However, those differences were not consistent from north to south. The best indicators of
Prairie Warbler occupancy were proximity to other Prairie Warblers and proximity to shrub
cover. I found an 80% probability of occupancy when another Prairie Warbler was within 91 m
of an existing territory. I found a 39% probability of occupancy when shrubs were within 1 m.
There were no significant environmental covariates from RUFs, however, four observations
warrant attention: 1) northern sites have more woody vegetation cover than southern sites, and
southern sites seem to be more open, 2) grass cover appears to be important on southern sites,
potentially because shrubs are less ubiquitous, 3) higher soil moisture in northern territories was
suggested as an important indicator of habitat use, and 4) lack of significant RUF coefficients
suggest that the spatial resolution of land cover data used (30 m) may not have been appropriate
to describe within territory space use for Prairie Warblers. Key among management
recommendations, conservation professionals are encouraged to maintain existing, large areas of
contiguous habitat with well dispersed shrubs to accommodate multiple breeding territories and
thus maintain coastal Prairie Warbler population viability.

ACKNOWLEDGEMENTS

Sincere thanks to the Michigan Department of Natural Resources, Wildlife Division, who
funded this research. Thank you to Becky Humphries, Bill Moritz and the Wildlife Management
Team for the opportunity to see and experience first-hand the challenges of balancing and
prioritizing ecological need with social and political realities. You were great mentors. The
Department of Fisheries and Wildlife, College of Agriculture and Natural Resources, Kalamazoo
Audubon Society, and Kalamazoo Nature Center provided additional funding support for which I
am very grateful. Thank you to my Ph.D. committee: Drs. Pat Brown, Pat Lederle, David Lusch,
Brian Maurer, Gary Roloff, and my primary advisor, Dr. Kelly Millenbah for their continued
encouragement and perseverance on the long road that the completion of this dissertation has
been. Thanks to Bill Taylor for introducing me to some wonderful fishermen and women who
have provided inspiration and invigorated appreciation for quiet rivers, big lakes and big dreams.
Thank you to my friends in the Millenbah lab and in the Fisheries and Wildlife Department for
making things fun. My appreciation to past colleagues at West Michigan Strategic Alliance,
Heart of the Lakes Center for Land Conservation Policy and currently at The Nature
Conservancy for your support along the way. Heartfelt thanks go out to my family, a constant
support network in every way, and especially to my Mom for listening to the play-by-play every
afternoon. Love to my sweet Sam who rolls merrily along. Last but not least thank you to my
husband, Jody, for always believing in me, being the voice of reason, and enduring my schedule.
Words cannot convey how lucky I feel to be your friend and wife.

iv

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................................ vii
LIST OF FIGURES ....................................................................................................................... ix
CHAPTER 1
EXAMINING THE STATUS OF STATE-ENDANGERED PRAIRIE WARBLER
POPULATIONS IN MICHIGAN
NEED FOR RESEARCH ....................................................................................................... 1
DISTRIBUTION AND POPULATION TRENDS ................................................................ 2
PRAIRIE WARBLER ECOLOGY ........................................................................................ 7
GOALS AND OBJECTIVES ............................................................................................... 10
STUDY AREA ..................................................................................................................... 11
LITERATURE CITED ......................................................................................................... 17

CHAPTER 2
UNDERSTANDING VEGETATION COMPOSITION IN MICHIGAN PRAIRIE WARBLER
BREEDING AREAS
INTRODUCTION ................................................................................................................ 19
STUDY AREA ..................................................................................................................... 20
METHODS ........................................................................................................................... 23
Avian Surveys ............................................................................................................... 23
Mapping Occupied Areas ............................................................................................. 24
Unoccupied Sites .......................................................................................................... 25
Vegetation Sampling and Survey Design ..................................................................... 26
Statistical Analysis ........................................................................................................ 27
Detection Probability ................................................................................................ 27
Vegetation Analyses ................................................................................................. 29
RESULTS ............................................................................................................................. 30
Locating Prairie Warbler Breeding Areas..................................................................... 30
Detection Probability .................................................................................................... 30
Vegetation Analysis ...................................................................................................... 32
DISCUSSION ....................................................................................................................... 40
LITERATURE CITED ......................................................................................................... 44

v

CHAPTER 3
PRAIRIE WARBLER OCCUPANCY AND RESOURCE USE IN MICHIGAN DURING THE
BREEDING SEASON
INTRODUCTION ................................................................................................................ 47
STUDY SITE ........................................................................................................................ 49
METHODS ........................................................................................................................... 52
Avian Surveys ............................................................................................................... 52
Territory Mapping......................................................................................................... 54
Unoccupied Sites .......................................................................................................... 55
Occupancy Analysis ..................................................................................................... 55
Within-Territory Resource Utilization.......................................................................... 57
RESULTS ............................................................................................................................. 63
Occupancy Analysis...................................................................................................... 64
Resource Utilization Functions ..................................................................................... 70
DISCUSSION ....................................................................................................................... 74
MANAGEMENT RECOMMENDATIONS ........................................................................ 76
LITERATURE CITED ......................................................................................................... 79

CHAPTER 4
SYNTHESIS OF FINDINGS AND MANAGEMENT RECOMMENDATIONS FOR
MICHIGAN PRAIRIE WARBLER POPULATIONS
INTRODUCTION ................................................................................................................ 84
BACKGROUND AND RESEARCH NEED ....................................................................... 84
SYNTHESIS OF FINDINGS ............................................................................................... 87
Population Distribution, Relative Abundance and Breeding Territory Sizes ............... 87
West Michigan Coastal Habitat Composition............................................................... 90
Assessing Occupancy.................................................................................................... 92
Resource Utilization and Spatial Resolution Considerations ....................................... 93
CONSERVATION AND MANAGEMENT RECOMMENDATIONS .............................. 95
DATA LIMITATIONS ....................................................................................................... 102
CONCLUSIONS................................................................................................................. 103
APPENDIX………………………………………………………………………………. 107
LITERATURE CITED ....................................................................................................... 115

vi

LIST OF TABLES

Table 1. Total area (ha) and potential habitat (ha)* of nine study locations used to survey for
Prairie Warblers in Michigan from May-August in 2004 and 2005. ............................................ 26
Table 2. Detection probability and occupancy estimates for Prairie Warbler point count surveys
in Sleeping Bear Dunes in 2004 (n = 292) and the 8 southern sites (n = 172). ............................ 31
Table 3. Plant species most influential in describing differences across vegetation survey plots.
Plants are shown with correspondence analysis 4-letter code, common name, genus species, and
eigenvector loadings for axes 1 and 2 which represent 15.1% and 11.3% of the proportional
variance in the dataset respectively. Large eigenvector loadings, driving differences, are bolded.
....................................................................................................................................................... 33
Table 4. Two-way analysis of variance (ANOVA) for correspondence axes 1-4 (dependent
variables) to test for differences between areas where Prairie Warblers were present and absent
(presence) and whether differences varied between northern Sleeping Bear Dunes and southern
sites (site). Significant (α <0.05) results are bolded with axes 1 and 2 showing significant
interactions. DF = degrees of freedom. SS=sum of squares. MS = mean square. ..................... 36
Table 5. Total area (ha) and potential habitat (ha)* of 9 study locations used to survey Prairie
Warblers in Michigan from May-August in 2004 and 2005......................................................... 53
Table 6. Common, general land cover classes observed around Prairie Warbler breeding areas
along Michigan’s west coast. ........................................................................................................ 56
Table 7. Twenty-four original C-CAP land cover classes (30 m resolution) were reclassified into
more general classes to represent resource attributes within Prairie Warbler breeding territories
along the west coast of Michigan, 2004 and 2005. ....................................................................... 62
Table 8. Logistic regression results for land cover variables and Prairie Warbler presenceabsence data. The full model contained 6 independent variables. The reduced model was
calculated using significant variables from the full model. Coefficient significance was tested
using Wald’s chi-square test (z scores). Approximation of fit was calculated as (1- residual
2
deviance/null deviance) indicating the percent of variability explained by each model. X notes
whether the 2 models were significantly different (full residual deviance – reduced residual
deviance, 3df)................................................................................................................................ 66
Table 9. Estimates of standardized RUF coefficients (β) for Prairie Warblers across all territories
(n = 38), Sleeping Bear Dunes only (n = 20) and southern sites (n = 18). Relative importance of
resources is indicated by magnitude of β. ..................................................................................... 71
Table 10. Mean percent of each covariate by region. Relative availability of each covariate are
compared across all territories (n = 38), for Sleeping Bear Dunes only (n = 20) and for the

vii

southern sites (n =1 8). Mean TWI values indicate level of wetness for an area on a possible
scale from 0 to 224 (high = wet, low = dry). ................................................................................ 72
Table 11. Mean area of Prairie Warbler 95% and 50% territory use percentiles (see Chapter 2 for
methods). SD = standard deviation. Range of territory areas are shown for each southern survey
site, mean across southern territories, mean across Sleeping Bear Dunes territories, and for each
grouping of territories in the northern, central, and southern areas of Sleeping Bear Dunes (see
Appendix for territory maps). ....................................................................................................... 89

viii

LIST OF FIGURES

Fig. 1. North American Breeding Bird Survey summer Prairie Warbler distribution map (19942003). Numbers in legend refer to numbers of birds, with darkest colors representing greatest
numbers observed. For interpretation of the references to color in this and all other figures, the
reader is referred to the electronic version of this dissertation. ...................................................... 4
Fig. 2. Prairie Warbler occurrence confirmations by county in Michigan between 1986 and
2011. Data may not reflect true current distribution since much of the state has not been
thoroughly surveyed. Data provided by Michigan Natural Features Inventory. ........................... 6
Fig. 3. North American Breeding Bird Survey trend map (1966-2003). Areas of increasing
Prairie Warbler populations are shown in blue. Areas of population decline are shown in red. ... 7
Fig. 4. Prairie Warbler and vegetation survey locations along the Lower Peninsula of Michigan.
Surveys took place from May-August in 2004 and 2005. Distance between Warren Dunes State
Park and Sleeping Bear Dunes National Lakeshore is approximately 386 km (240 miles). ........ 12
Fig. 5. Michigan average July temperature. ................................................................................. 14
Fig. 6. Michigan average annual precipitation............................................................................. 15
Fig. 7. Nine Prairie Warbler survey locations along the Lower Peninsula of Michigan. Surveys
took place from May-August in 2004 and 2005. Distance between Warren Dunes State Park and
Sleeping Bear Dunes National Lakeshore is approximately 386 km (240 miles). Prairie Warblers
were detected in Sleeping Bear Dunes in 2004 and in Kitchel-Lindquist Dunes Preserve,
Saugatuck Dunes State Park, Oval Beach and Warren Dunes State Park in 2005. ...................... 21
Fig. 8. Nested non-aligned block design used for vegetation sampling in Prairie Warbler
territories and unoccupied sites. Vegetation sampling occurred in the 0.5 m circles. There were
52 0.5 m survey plots per 30 m radius plot. .................................................................................. 28
Fig. 9. Difference between Prairie Warbler occupied (solid shape) and unoccupied (hollow
shape) areas in the north (red circles) vs. the south (black triangles). Mean eigenvector values
were plotted for northern Sleeping Bear Dunes and southern sites with standard error bars for the
first two correspondence axes. ...................................................................................................... 37
Fig. 10. Correspondence analysis plot for the 46 plant species. Plant species driving differences
across Prairie Warbler presence and absence sites are represented by 4-letter codes in each
quadrant (see Table 3 for plant species names and eigenvector loadings corresponding to axes).
Sites occupied by Prairie Warbler in Sleeping Bear Dunes are represented by solid (red) circles
and in southern sites by solid (black) triangles. Unoccupied sites in Sleeping Bear Dunes are
represented by hollow (red) circles and in southern sites by hollow (black) triangles.
Correspondence axis 1 (x axis) and 2 (y axis) represent 15.1% and 11.3% of the proportional
variance in the dataset respectively............................................................................................... 38
ix

Fig. 11. Nine Prairie Warbler survey locations along the Lower Peninsula of Michigan. Surveys
took place from May-August in 2004 and 2005. Distance between Warren Dunes State Park and
Sleeping Bear Dunes National Lakeshore is approximately 386 km (240 miles). Prairie Warblers
were detected in Sleeping Bear Dunes in 2004 and in Kitchel-Lindquist Dunes Preserve,
Saugatuck Dunes State Park, Oval Beach and Warren Dunes State Park in 2005. ...................... 50
Fig. 12. (A) Example of multi-dimensional utilization distribution of a Prairie Warbler breeding
territory compared to (B) territory mapping registration points for the same territory. The
highest density of points in (B) is represented by the “highest” peak in (A). ............................... 60
Fig. 13. Example of Prairie Warbler breeding territories calculated using a fixed kernel
estimation with least squares cross validation. Eight breeding territories were mapped in Warren
Dunes State Park, Berrien County, Michigan (southwest Michigan) in 2005. ............................. 65
Fig. 14. Probability of Prairie Warbler presence as a function of distance to (A) shrub and to (B)
closest Prairie Warbler breeding territory; the 2 significant variables in the reduced logistic
regression model. .......................................................................................................................... 68
Fig. 15. Trend suggesting high northern forest use (solid circles) and high southern sand use
(hollow triangles) by Prairie Warblers along Michigan’s west coast. Prairie Warblers tended to
use either forest or sandy areas of their breeding territories. Northern Prairie Warblers tended to
concentrate use in grass when forest use was also high. Southern Prairie Warblers tended to
concentrate use in grass when sand use was also high. ................................................................ 73
Fig. 16. Eight Prairie Warbler breeding territories mapped in 2005, Warren Dunes State Park,
Berrien County, Michigan. Each breeding territory is designated by an individual color. The
95% use percentile for an individual Prairie Warbler with its interior 50% (“core”) use percentile
is shown overlaid on 1 m resolution, 2005 United States Department of Agriculture, National
Agriculture Imagery Program, land cover data........................................................................... 107
Fig. 17. Two Prairie Warbler breeding territories mapped in 2005, Oval Beach, Allegan County,
Michigan. Each breeding territory is designated by an individual color. The 95% use percentile
for an individual Prairie Warbler with its interior 50% (“core”) use percentile is shown overlaid
on 1 m resolution, 2005 United States Department of Agriculture, National Agriculture Imagery
Program, land cover data. ........................................................................................................... 108
Fig. 18. Six Prairie Warbler breeding territories mapped in 2005, Saugatuck Dunes State Park,
Allegan County, Michigan. Each breeding territory is designated by an individual color. The
95% use percentile for an individual Prairie Warbler with its interior 50% (“core”) use percentile
is shown overlaid on 1 m resolution, 2005 United States Department of Agriculture, National
Agriculture Imagery Program, land cover data........................................................................... 109
Fig. 19. Two Prairie Warbler breeding territories mapped in 2005, Kitchel-Lindquist Dunes
Preserve, Ottawa County, Michigan. Each breeding territory is designated by an individual
color. The 95% use percentile for an individual Prairie Warbler with its interior 50% (“core”)

x

use percentile is shown overlaid on 1 m resolution, 2005 United States Department of
Agriculture National Agriculture Imagery Program land cover data. ........................................ 110
Fig. 20. Three Prairie Warbler breeding territories mapped in 2004, southern portion of Sleeping
Bear Dunes National Lakeshore, Benzie County, Michigan. Each breeding territory is
designated by an individual color. The 95% use percentile for an individual Prairie Warbler with
its interior 50% (“core”) use percentile is shown overlaid on 1 m resolution, 2005 United States
Department of Agriculture National Agriculture Imagery Program land cover data. ................ 111
Fig. 21. Six Prairie Warbler breeding territories mapped in 2004, central portion of Sleeping
Bear Dunes National Lakeshore, Leelanau County, Michigan. Each breeding territory is
designated by an individual color. The 95% use percentile for an individual Prairie Warbler with
its interior 50% (“core”) use percentile is shown overlaid on 1 m resolution, 2005 United States
Department of Agriculture National Agriculture Imagery Program land cover data. ................ 112
Fig. 22. Eleven Prairie Warbler breeding territories mapped in 2004, northern portion of
Sleeping Bear Dunes National Lakeshore, Leelanau County, Michigan. Each breeding territory
is designated by an individual color. The 95% use percentile for an individual Prairie Warbler
with its interior 50% (“core”) use percentile is shown overlaid on 1 m resolution, 2005 United
States Department of Agriculture National Agriculture Imagery Program land cover data....... 113

xi

CHAPTER 1

EXAMINING THE STATUS OF STATE-ENDANGERED PRAIRIE WARBLER
POPULATIONS IN MICHIGAN

NEED FOR RESEARCH
Data describing Michigan’s state-endangered Prairie Warbler (Dendroica discolor)
populations are limited. The status of Michigan breeding locations, population abundance, and
habitat composition influencing breeding territory selection are critical management and
conservation needs for this state-endangered, Neotropical migratory songbird (Walkinshaw
1991, Nolan et al. 1999, Cooper 2000, Southwell 2001).
In recent years, the majority of Michigan breeding season observations have been
reported from the western coastal counties along Lake Michigan (Brewer et al. 1991, Michigan
Natural Features Inventory 2011). Gaining an understanding of where west Michigan Prairie
Warbler breeding populations are specifically located, and in what abundance, will allow a
baseline from which to monitor population status. Analyzing similarities and differences in plant
composition between breeding locations will provide increased understanding of Michigan
Prairie Warbler ecology. Describing habitat composition within breeding territories will help
identify and specify Prairie Warbler habitat needs, thus informing conservation practitioners as
they design long term management plans and are asked to respond to proposed land use changes
(e.g., mining, wind tower siting, development).
Understanding the scale at which a species uses the landscape influences the way habitat
composition data is collected (i.e., survey design) and ultimately provides for more relevant
conservation and resource management recommendations for a species (Wiens 1989, BohningGaese 1997) or the habitat on which it depends. Home range estimates describe an animal’s use
1

of space (Aebischer et al. 1993), however, resource use within an animal’s home range is not
often uniform (Marzluff et al. 1997). Exploring not only where Prairie Warblers are located, but
also where specifically within the home range they are spending the majority of their time, is an
important step in understanding key drivers of habitat selection in west Michigan coastal
breeding areas. Linking habitat use information to the resource composition in those highestused areas will increase our overall knowledge of this species and inform subsequent
conservation actions.
Some have speculated that abundant Prairie Warbler habitat exists (Walkinshaw 1991,
Cooper 2000, Southwell 2001), but habitat requirements for Prairie Warbler, and their selection
strategies, may be more specific than previously anticipated. Examination of environmental
factors (e.g., plant composition, proximity to specific land cover types or other Prairie Warblers)
within breeding territories, compared to areas where Prairie Warbler are absent, will allow
conservation practitioners to manage with a greater understanding of Prairie Warbler habitat
selection strategies and thus how to maintain habitat for this state-endangered species in concert
with other conservation objectives.

DISTRIBUTION AND POPULATION TRENDS
Prairie Warbler is one of many species that make an annual spring migration northward,
over thousands of miles of land and water, to their breeding grounds. Prairie Warblers primarily
winter in the West Indies but also in coastal Florida and coastal Central America including parts
of Belize and Costa Rica (Dunn and Garret 1997). The core of the Prairie Warbler breeding
range lies in the central southeastern United States with northern boundaries spanning from
South Dakota, southern Ontario and southern New Hampshire (Walkinshaw 1991, Dunn and

2

Garret 1997) (Figure 1). They arrive on breeding grounds in the southeastern and south-central
United States beginning in March. The Prairie Warblers that make their way to Michigan will
have traveled farther than most, arriving at the northern edge of the breeding range during the
first two weeks of May (Dunn and Garret 1997).
The first survey record for a Prairie Warbler in the United States (April 18, 1810) was
from southwestern Kentucky in what was described by Alexander Wilson as “barrens” and
“prairie country. . .evidently resembling and perhaps representative of the tall-grass prairie”
(Mengel 1965). The bird’s name is derived from this description by Wilson (Nolan 1978). At
the time of its discovery the species was likely a rare and local bird, with most populations
scattered around the margins of its present range (Nolan 1978). Among other early records
throughout the current breeding range, an 1839 record of a Prairie Warbler was noted by Abram
Sager during travels between Detroit, Saginaw Bay and Jackson, Michigan (Wood 1951). It is
thought that if a Michigan breeding populations existed at that time, it was likely disjunct (Nolan
1978).
Prairie Warblers were recorded nesting in the southern Lower Peninsula counties of
Ottawa and Montcalm in the early 1900’s (Barrows 1912). Wood (1951) noted Berrien County
as important to Prairie Warbler, as did Zimmerman and VanTyne (1959) in addition to reporting
nesting records from the west Michigan counties of Ottawa, Muskegon, and Oceana. Wood
(1951), Zimmerman and Van Tyne (1959), and Mayfield (1960) all reported occurrence in
northeastern Lower Peninsula counties of Oscoda, Crawford, Otsego, Cheboygan, and Presque
Isle, primarily associated with jack pine (Pinus banksiana) plains.

3

Fig. 1. North American Breeding Bird Survey summer Prairie Warbler distribution map (19942003). Numbers in legend refer to numbers of birds, with darkest colors representing greatest
numbers observed. For interpretation of the references to color in this and all other figures and
tables, the reader is referred to the electronic version of this dissertation.

4

Though globally secure (NatureServe 2011), Prairie Warbler populations have been
declining nation-wide since 1970 (Robbins et al. 1986, Saurer et al. 2008). Midwest populations
are of moderate to high management concern (Robinson et al. 1999). This trend holds true for
Michigan populations (Michigan Natural Features Inventory 2011), which are thought to have
always been small and localized (Nolan 1978). Between 1986 and 2011, Michigan has had
occurrences confirmed in 18 out of 83 counties (Michigan Natural Features Inventory 2011)
(Figure 2).
The North American Breeding Bird Survey’s (BBS) best estimate of population change
for Prairie Warbler over its range between 1966 and 2003 shows increases in New England,
coastal regions of the Carolinas, southern Louisiana, southern Alabama and southern Georgia,
the southern Florida panhandle and in other smaller patches within its current range (Saurer et al.
2008). The central Michigan Lake Michigan coast shows both decline and increase in Prairie
Warblers (Figure 3).

PRAIRIE WARBLER ECOLOGY
Prairie Warbler breeding territories have been characterized throughout their range as
open, brushy, early successional habitat with scrub-shrub species on poor soils with little or no
tree canopy (Nolan 1978, Walkinshsaw 1991, Nolan et al. 1999, Cooper 2000). Overall, the
Lake Michigan sand dunes have been denoted as critical breeding areas for this species in
Michigan (Walkinshaw 1991). Wood (1951), Zimmerman and Van Tyne (1959), and Mayfield
(1960) all suggested that Michigan’s jack pine plains may be important to the species in the
early- to mid-1900’s.

5

Occurrence

Fig. 2. Prairie Warbler occurrence confirmations by county in Michigan between 1986 and
2011. Data may not reflect true current distribution since much of the state has not been
thoroughly surveyed. Data provided by Michigan Natural Features Inventory.

6

Fig. 3. North American Breeding Bird Survey trend map (1966-2003). Areas of increasing
Prairie Warbler populations are shown in blue. Areas of population decline are shown in red.

7

Michigan populations have been observed breeding in three general vegetation types
(Walkinshaw 1991, Cooper 2000), although this is not viewed as an exhaustive list of potential
habitat:
Early successional stages of Great Lakes sand dunes where deciduous or coniferous
shrubs are intermixed with dune grass and other herbs,
Early-successional, shrubby openings associated with previously-burned jack pine
plains, and
Previously burned conifer stands dominated by deciduous shrubs and small trees.
Prairie Warblers feed primarily on insects and sometimes spiders, snails, or worms; they
are generalists. Nestlings are primarily fed caterpillars (Cooper 2000). The species forages by
gleaning insects from leaves and branches and will occasionally take prey mid-air (DeGraaf et al.
1991, Nolan et al. 1999). A male will forage a few meters off the ground or from higher perches
during the breeding season when he is advertising (Nolan 1978). Mean breeding season feeding
height in Florida for breeding males is 2.19 m; for females 1.40 m (Prather 1994).
Territories are established in Michigan by males upon arrival in early May (Wood 1951,
Nolan 1978, Nolan et al 1999). They advertise for females, who arrive about one week later, and
defend their territory by singing from tall, exposed perches. Fall records suggest that Prairie
Warblers tend to leave Michigan breeding grounds to migrate south by the third week of August
(Wood 1951, Nolan 1978, Nolan et al. 1999).
Prairie Warbler breeding territory size varies between 0.24 – 3.5 ha (roughly half of a
football field to about 8 football fields) (Nolan 1978). The species has been found to return to
the same territory year after year until the land cover has succeeded beyond habitation (Nolan
1978). Males show a high rate of natal site fidelity (Nolan 1978).

8

Nests are open-cup, normally found 1-3 m (roughly 3 to 9 feet) off the ground, and are
composed of woven plant fibers lined with hair or feathers (Cooper 2000). Nests are typically
placed (dependent on geographic region) in a shrub, sapling, thicket, or fern clump. Typically,
3-5 eggs are laid in June, incubated solely by the female, and hatched within 11-15 days (Nolan
1978). Young are tended by both parents and fledge between 8-10 days, remaining dependent on
the parents for an additional 30-35 days (Nolan 1978, Baicich and Harrison 1997).
After habitat loss and fragmentation in winter and breeding ranges, nest parasitism and
predation are the two greatest threats to the Prairie Warbler. Primary predators of nestlings in
Michigan include snakes, chipmunks and corvids (Nolan 1978). Females in Michigan
populations usually only have one brood as the northern breeding season does not often provide
ample time to re-brood. Brown headed cowbirds are known to be nest parasites (Nolan 1978,
Nolan et al. 1999)
Males will sometimes have two territories (and two females). Females (rarely males) will
move and change mates when a nest fails or before attempting a second brood (Nolan et al.
1999). Nolan’s (1978) 20-year study on Indiana Prairie Warbler populations (1951-1972),
indicated that 46% of 137 males were monogamous over one season; only 35% of 176 females
exhibited season-long monogamy with a monogamous male.
Historical (early- to mid-1900’s) Michigan occurrence patterns noted Prairie Warblers
using the sandy outwash plains of the northeast Lower Peninsula; those areas now commonly
associated with the federally endangered Kirtland’s Warbler (Wood 1951, Zimmerman and Van
Tyne 1959, Mayfield 1960). Records show that by the late 1950’s, breeding Prairie Warblers
were not as commonly observed the northeast Lower Peninsula (Mayfield 1960, Nolan 1978).

9

GOALS AND OBJECTIVES
The overarching goal of this research is to advance understanding of Prairie Warbler
needs in the context of the western Michigan coastal landscape. This information will contribute
to our understanding of Prairie Warbler resource use dynamics on the northern edge of the
breeding range as well as provide this information at a spatial resolution that was previously
unknown.
The chapters in this dissertation address the following objectives:
Chapter 2:
Locate Prairie Warbler breeding areas along the west coast of Michigan,
Quantify plant composition associated with Prairie Warbler breeding and unoccupied
sites, and
Determine whether geographic variation exists in the plant composition associated with
Prairie Warbler breeding areas.
Chapter 3:
Determine which land cover types can forecast Prairie Warbler occupancy in west
Michigan dune landscapes,
Develop resource utilization functions for Prairie Warbler breeding territories along
Michigan’s west coast, and
Gain insight into how resource use may differ for Prairie Warblers between northern and
southern coastal breeding territories in Michigan.
Chapter 4 will summarize research findings from Chapters 2 and 3 and translate findings
into management recommendations.

10

Some repetition could occur between chapters, as chapters 2 and 3 are written in
manuscript style, prepared with intent toward submission to peer-reviewed journals.

STUDY AREA
Nine study sites were located in the Lower Peninsula of Michigan along the Lake
Michigan coast (Figure 4). Study sites were chosen because they were publically accessible and
contained large areas of potential habitat for Prairie Warblers (identified by cover type (i.e., jack
pine, sand dunes, scrub-shrub) (Nolan 1978, Nolan et al. 1999, Cooper 2000, Southwell 2001).
Survey sites included six State Parks, Sleeping Bear Dunes National Lakeshore, Oval Beach (a
public beach owned by the city of Saugatuck), and Kitchel-Lindquist Dunes Preserve (a
publically-accessible preserve near Grand Haven). The distance between the southern-most site
(Warren Dunes State Park) and the northern-most site (Sleeping Bear Dunes National Lakeshore)
was approximately 386 km (240 miles).
The mean temperature during survey years (2004 and 2005) at Sleeping Bear Dunes
between May 1-14 (Prairie Warbler estimated arrival on Michigan breeding grounds; Dunn and
Garret 1997) was 11°C and 9°C, respectively. Mean temperatures at Warren Dunes were 14°C
and 12°C, respectively. These year-specific temperatures compare closely to average annual
temperatures at Sleeping Bear Dunes and Warren Dunes during May: 13°C and 14°C
respectively (Figure 5). Average precipitation for the month of May at Sleeping Bear Dunes is
6.6 cm and at Warren Dunes is 8.1 cm (Figure 6). Temperatures increase throughout the Prairie
Warbler breeding season in Michigan with variable rainfall.

11

0

50

100

200 Kilometers

¯

Fig. 4. Prairie Warbler and vegetation survey locations along the Lower Peninsula of Michigan.
Surveys took place from May-August in 2004 and 2005. Distance between Warren Dunes State
Park and Sleeping Bear Dunes National Lakeshore is approximately 386 km (240 miles).

12

Michigan’s Great Lakes shoreline, foredune, and backdune forest systems are a relatively
small but ecologically valuable component of the state’s natural features (Albert 2000).
Distinctive land cover types and hydrologic systems that support floral and faunal species in
these areas provide stop-over resources for annually dependent migrating songbirds, hawks, and
owls from South and Central America, as well as for native shorebirds (Brewer et al. 1991). The
unique combinations of habitat structure and composition are instrumental and serve as breeding
grounds for hundreds of these species.
Foredune vegetation communities consist of specialized plants and some shrubs adapted
to sand burial and fast growth with deep, often rhyzomal, root systems that stabilize blowing
sand. Pioneering grasses such as marram grass (Ammophila breviligulata) and sand-reed grass
(Calamovilfa longifolia) allow sand to accumulate and stabilize. Common Michigan foredune
shrubs include hearty, woody species such as sand cherry (Prunus pumila), red osier dogwood
(Cornus stolonifera), bearberry (Arctostaphylos uva-ursi), several dune willows (Salix spp.), and
creeping juniper (Juniperus horizontalis). These plants must tolerate extreme temperatures, low
moisture levels, and low levels of nutrients in addition to blowing sand.
The foredune zone undergoes continual change due to natural weather events, erosion,
and water level fluctuations. Larger dunes form behind the protective foredune, where sand
accumulation is slower, allowing dune grasses, herbaceous plants and shrubs and forests to
establish. Oaks (Quercus spp.), bigtooth aspen (Populus grandidentata) and sassafras (Sassafras
albidum), in addition to stands of jack pine (Pinus banksiana) and common juniper (J.s
communis), grow closest to the shoreline on wind-ward slopes. In contrast, the landward side of
the dunes is more moist and nutrient-rich, supporting mesic forest species like sugar maple (Acer

13

Fig. 5. Michigan average July temperature.

14

Fig. 6. Michigan average annual precipitation.

15

Saccharum), American beech (Fagus grandifolia), basswood (Tilia americana), and red oak (Q.
rubra).
Anthropogenic effects from development patterns and introduction of non-native, fastspreading plants are stressors on this fragile ecosystem. Incompatible development resulting in
land use change could threaten the variety and availability of habitats, and thus threaten some of
the only areas in which Prairie Warblers have been observed in Michigan in recent years (Brewer
et al. 1991, Michigan Natural Features Inventory 2011).

16

LITERATURE CITED

17

LITERATURE CITED

Aebischer, N. J., P. A. Robertson, and R. E. Kenward. 1993. Compositional analysis of habitat
use from animal radio-tracking data. Ecology 74:1313-1325.
Albert, D. A. 2000. Born of the wind: an introduction to the ecology of Michigan sand dunes.
Michigan Natural Features Inventory, Lansing, Michigan, USA.
Baicich, P. J., and C. O. Harrison. 1997. A guide to nests, eggs, and nestlings of North
American birds. Academic Press, New York, New York, USA.
Barrows, W. B. 1912. Michigan bird life. Special bulletin of the Department of Zoology and
Physiology of the Michigan Agricultural Collage. Michigan Agricultural College, East
Lansing, Michigan, USA.
Bohning-Gaese, K. 1997. Determinants of avian species richness at different spatial scales
Journal of Biogeography 24:49-60.
Brewer, R., G. A. McPeek, and R. J. Adams, Jr. 1991. Atlas of breeding birds in Michigan.
Michigan State University Press, East Lansing, Michigan, USA.
Cooper, J. L. 2000. Special animal abstract for Dendroica discolor (Prairie Warbler). Michigan
Natural Features Inventory, Lansing, Michigan, USA.
DeGraaf, R. M., V. E. Scott, R. H. Hamre, L. Ernst, and S. H. Anderson. 1991. Forest and
rangeland birds of the United States: natural history and habitat use. U.S. Department of
Agriculture, Forest Service Handbook 688, Washington, D.C., USA.
Dunn, J. L., and K. L. Garrett. 1997. A field guide to warblers of North America. Houghton
Mifflin Co., Boston, Massachusetts, USA.
Marzluff, J. M., S. T. Knick, M.S. Vekasy, L. S. Schueck, and T. J. Zarriello. 1997. Spatial use
patterns and habitat selection of golden eagles in southwestern Idaho. Auk 114:673-687.
Mayfield, H. 1960. The Kirtland’s warbler. Cranbrook Institute of Science Bulletin 40,
Bloomfield Hills, Michigan, USA.
Mengel, R. M. 1965. The birds of Kentucky. American Ornithologists Union Monograph, No.
3.
Michigan Natural Features Inventory. 2011. Rare species explorer. http://web4.msue.msu.edu/
mnfi/explorer. Accessed 2004-2011.
NatureServe. 2011. NatureServe explorer: An online encyclopedia of life. Version 7.1.
17

www.natureserve.org/explorer. NatureServe, Arlington, Virginia, USA. Accessed April
2012.
Nolan, V., Jr. 1978. The ecology and behavior of the Prairie Warbler Dendroica discolor.
Ornithological Monographs No. 26.
Nolan, V., Jr., E. D. Ketterson, and C. A. Buerkle. 1999. The birds of North America: Prairie
Warbler (Dendroica discolor). Pages 1-26 in A. Poole and F. Gill, editors. The Birds of
North America, No. 584. Philadelphia, Pennsylvania, USA.
Prather, J. W. 1994. A comparison of the ecology and behavior of two species of congeneric
wood warblers in the mangrove habitats of south Florida. M. A. thesis, University of
Colorado, Denver, Colorado, USA.
Robbins C. S., D. Bystrak, and P. H. Geissler. 1986. The breeding bird survey: its first fifteen
years, 1965-1979. U.S, Fish and Wildlife Service Resource Publication 157.
Washington, D.C., USA.
Robinson, S. K., E. J. Heske, and J. D. Brawn. 1999. Factors affecting the nesting success of
edge and shrubland birds. Report submitted to Illinois Department of Natural Heritage,
Springfield, Illinois, USA.
Saurer, J. R., J. E. Hines, and J. Fallon. 2008. The North American breeding bird
survey, results and analysis 1966 - 2007. Version 5.15.2008. USGS Patuxent Wildlife
Research Center, Laurel, Maryland, USA.
Southwell, D. K. 2001. Conservation assessment for Prairie Warbler (Dendroica discolor).
USDA Forest Service Eastern Region. USDA Forest Service, Hiawatha National Forest,
Michigan, USA.
Walkinshaw, L. H. 1991. Prairie Warbler. Pages 418-419 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.
Wiens, J. A. 1989. The ecology of bird communities. Vol. 1. Foundations and
Patterns. Cambridge University Press, Cambridge, England, UK.
Wood, N. A. 1951. The birds of Michigan. University of Michigan Museum of Zoology
Miscellaneous Publication 75, Ann Arbor, Michigan, USA.
Zimmerman, D. A., and J. Van Tyne. 1959. A distributional checklist of the birds of Michigan.
University of Michigan Museum of Zoology, Ann Arbor, Michigan, USA.

18

CHAPTER 2

UNDERSTANDING VEGETATION COMPOSITION IN MICHIGAN PRAIRIE WARBLER
BREEDING AREAS

INTRODUCTION
The Prairie Warbler (Dendroica discolor) is one of 92 state-endangered species in
Michigan (Michigan Natural Features Inventory 2011). Understanding plant composition in
areas selected by Prairie Warblers in Michigan for breeding territories has been specifically
identified as a critical information need to better understand habitat requirements for natural
resource management and conservation planning (Nolan et al. 1999, Cooper 2000, Southwell
2001). Currently no quantifiable analysis of habitat selection exists for Prairie Warblers in
Michigan (Walkinshaw 1991, Cooper 2000, Southwell 2001). Determining the composition of
resources used by Prairie Warblers can aid our understanding of important elements associated
with habitat use and needs in the broader context of the landscape. Information on withinterritory habitat use will contribute to our understanding of habitat selection on the northern edge
of the Prairie Warbler breeding range.
Prairie Warbler breeding territories have been characterized as open, brushy, early
successional vegetation containing scrub-shrub plant species on poor soils with little or no tree
canopy (Nolan 1978, Walkinshsaw 1991, Nolan et al. 1999, Cooper 2000). Lake Michigan sand
dunes have been denoted as critical breeding areas (Walkinshaw 1991). Michigan Prairie
Warbler populations have been noted to breed in three general vegetation types (Walkinshaw
1991), although this may not be an exhaustive list of potential breeding habitat in Michigan: 1)
early successional stages of Great Lakes sand dunes where deciduous or coniferous shrubs are
intermixed with dune grass and other herbs, 2) early-successional, shrubby openings associated

19

with previously-burned jack pine plains, and 3) previously burned conifer stands dominated by
deciduous shrubs and small trees. Prairie Warbler occurrences have been confirmed in 18 of 83
counties between 1986 and 2011 (Michigan Natural Features Inventory 2011) with more recent
records associated almost exclusively with the west Michigan counties bordering Lake Michigan
(Brewer et al. 1991, Michigan Natural Features Inventory 2011).
The goal of this research was to quantify breeding area information for Prairie Warblers
along the eastern Lake Michigan coast. Objectives were to 1) locate Prairie Warbler breeding
areas, 2) quantify plant composition associated with Prairie Warbler breeding areas and
unoccupied sites, and 3) determine whether geographic variation existed in the plant composition
associated with Prairie Warbler breeding areas.

STUDY AREA
Nine study sites were located in the Lower Peninsula of Michigan along the Lake
Michigan coast (Figure 7). Study sites were chosen because they were publically accessible and
contained potential Prairie Warbler habitat (identified by cover type (i.e., jack pine, sand dunes,
scrub-shrub) (Nolan 1978, Nolan et al. 1999, Cooper 2000, Southwell 2001). Study sites
included 6 State Parks, Sleeping Bear Dunes National Lakeshore, Oval Beach (a public beach
owned by the city of Saugatuck), and Kitchel-Lindquist Dunes Preserve (a publically-accessible
preserve near Grand Haven). The distance between the southern-most site (Warren Dunes State
Park) and the northern-most site (Sleeping Bear Dunes National Lakeshore) was approximately
386 km (240 miles).

20

0

50

100

200 Kilometers

¯

Fig. 7. Nine Prairie Warbler survey locations along the Lower Peninsula of Michigan. Surveys
took place from May-August in 2004 and 2005. Distance between Warren Dunes State Park and
Sleeping Bear Dunes National Lakeshore is approximately 386 km (240 miles). Prairie Warblers
were detected in Sleeping Bear Dunes in 2004 and in Kitchel-Lindquist Dunes Preserve,
Saugatuck Dunes State Park, Oval Beach and Warren Dunes State Park in 2005.

21

Michigan’s Great Lakes shoreline, foredune, and backdune forest systems are a relatively
small but ecologically valuable component of the state’s natural features (Albert 2000).
Foredune vegetation communities consist of specialized plants and some shrubs adapted to sand
burial and fast growth with deep, often rhyzomal, root systems that stabilize blowing sand.
Pioneering grasses such as marram grass (Ammophila breviligulata) and sand-reed grass
(Calamovilfa longifolia) cause sand to accumulate and stabilize. Common Michigan foredune
shrubs include hearty, woody species such as sand cherry (Prunus pumila), red osier dogwood
(Cornus stolonifera), bearberry (Arctostaphylos uva-ursi), several dune willows (Salix spp.), and
creeping juniper (Juniperus horizontalis). These plants must tolerate extreme temperatures, low
moisture levels, and low levels of nutrients in addition to blowing sand (Hazlett 1991, Albert
2000).
The foredune zone undergoes continual change due to natural weather events, erosion,
and water level fluctuations. Larger dunes form behind the protective foredune, where sand
accumulation is slower, allowing dune grasses, herbaceous plants and shrubs and forests to
establish. Oaks (Quercus spp), bigtooth aspen (Populus grandidentata) and sassafras (Sassafras
albidum), in addition to stands of jack pine (P. banksiana) and common juniper (J. communis),
grow closest to the shoreline on wind-ward slopes. In contrast, the land-ward side of the dunes is
more moist and nutrient-rich, supporting mesic forest species like sugar maple (Acer
saccharum), American beech (Fagus grandifolia), basswood (Tilia americana), and red oak (Q.
rubra).
Anthropogenic effects from development patterns and introduction of non-native, fastspreading plants are stressors on this ecosystem. Anthropogenic development could threaten the
variety and availability of habitats, and thus threaten some of the only areas in which Prairie
Warblers have been observed in Michigan in recent years (Michigan Natural Features Inventory
22

2011).
Prairie Warblers arrive on Michigan breeding grounds within the first 2 weeks of May
(Dunn and Garret 1997). Mean temperature in 2004 and 2005 at Sleeping Bear Dunes during
this time (May 1-14) was 11°C (52°F) and 9°C (49°F), respectively; Warren Dunes temperatures
were 14°C (58°F) and 12°C (53°F), respectively. These year-specific temperatures compare
closely to average annual temperatures at Sleeping Bear Dunes and Warren Dunes during May:
13°C (55°F) and14°C (57°F) respectively. Average precipitation for the month of May is 6.6 cm
at Sleeping Bear Dunes and 8.1 cm at Warren Dunes. Temperatures increase throughout the
Prairie Warbler breeding season in Michigan with variable rainfall. Fall avian observation
records suggest that Prairie Warblers tend to leave breeding grounds to migrate south by the third
week of August (Michigan Natural Features Inventory 2011).

METHODS
Avian Surveys
The Michigan Department of Natural Resources’ (DNR) Integrated Forest Monitoring,
Assessment and Prescription (IFMAP) land cover data were used to guide the selection of avian
point count survey locations sampled in 2004 and 2005 (Michigan Department of Natural
Resources 2001). The IFMAP data were derived from classified Landsat Thematic Mapper 5
and 7 imagery, dated 1999-2001. Overall accuracy of IFMAP data was 88% at Anderson Level
2 (major land cover classes; scale of 1:40,000) (Space Imaging 2004). The IFMAP land cover
classes were aggregated from 35 classes into 9 similar, more general classes (Urban, Agriculture,
Herbaceous-Shrub Upland, Upland Deciduous Forest, Upland Coniferous Forest, Water,
Lowland Forest, Nonforested Wetland, Sand/Bare Soil) for the purposes of avian survey point
selection. Survey points were equally distributed within 3 of the 9 land cover classes (Sand/Bare
23

Soil, Herbaceous-Shrub Upland, and Upland Coniferous Forest) to represent all possible land
cover classes that Prairie Warblers have been known to inhabit in Michigan and throughout their
breeding range (Nolan 1978, Nolan et al. 1999, Cooper 2000, Southwell 2001).
Avian point count survey locations were randomly chosen using a GIS random point
generator (Jenness 2003) in suitable land cover types. Study site boundaries were acquired from
state, federal or municipal entities. Two-hundred ninety-two point count survey points were
established in Sleeping Bear Dunes in 2004. The number of survey points in each of the 8 sites
south of Sleeping Bear Dunes (hereafter referred to as the southern sites) was selected in
proportion to the total area of each site, totaling 172 point count survey points in 2005. Number
of sampling sites in Sleeping Bear Dunes (2004) and southern sites (2005) were maximized
based on the number of samples possible during the breeding season.
Locating Prairie Warblers was the primary objective of the point count surveys, therefore
when a Prairie Warbler was detected while walking between point count locations, a GPS point
was taken and added to the point count survey schedule. All avian survey points were at least
200 m apart. Prairie Warblers were identified using 10-minute point counts (Ralph et al. 1995,
Cooper 2000) May 5 – July 28 in 2004 and May 3 – June 22 in 2005. Surveys took place from
30 minutes before sunrise (civil twilight) until 10:30AM in the absence of rain, heavy fog or high
winds (Ralph et al. 1995). Point count locations were surveyed with a rotating time and observer
schedule to reduce biases. All points were observed on at least 4 occasions.
Mapping Occupied Areas
Areas occupied by Prairie Warblers were mapped using GPS to provide descriptive
spatial data estimating the shape and size of the area the birds were using for later vegetation
sampling in those areas. Mapping began when Prairie Warbler presence was recorded during
repeated point count surveys. Mapping visits continued until at least 25 registration (GPS)
24

points per occupied area were collected. Registrations included male singing locations, feeding
sites, or locations where defense of the area boundaries was observed. Though breeding was not
confirmed, these areas were likely breeding territories. Each occupied area was visited for
mapping at least 3 times. Each mapping visit lasted approximately 1 hour. The time of day a
particular bird was mapped alternated so that early morning, mid-morning, and afternoon
behaviors and movements were all included in the area’s delineation.
The Animal Movement extension (Hooge and Eichenlaub 1997) was used in GIS to
determine area size and shape based on registration points. Occupied area estimates were
calculated using a fixed kernel home range estimator with least squares cross validation
(Silverman 1986). Occupied areas were calculated at 95% and 50% probability contours. Fifty
percent probability contours were chosen for use in vegetation sampling (described below)
because these areas represented highest use.
Unoccupied Sites
Unoccupied survey points were chosen randomly from avian point count survey points on
which Prairie Warblers were not detected. Unoccupied points were buffered by the average
radius of the 50% probability contours (polygons) for each site (Table 1) to approximate
occupied area for use in vegetation sampling. Occupied area size (assumed to be a breeding
territory) tended to increase as area of the site increased, therefore the average territory radius for
a site of similar total area was used to buffer unoccupied points on sites with no Prairie Warblers
detected (Hoffmaster, Muskegon, Silver Lake and Ludington) (Table 1).
Vegetation Sampling and Survey Design
A 60 x 60 m grid was overlaid on each occupied area polygon using GIS. Three grid
cells, occurring completely within the occupied area boundary, were randomly selected in each
occupied area to serve as locations for vegetation sampling arrays. Survey arrays were centered
25

Table 1. Total area (ha) and potential habitat (ha)* of 9 study locations used to survey for Prairie
Warblers in Michigan from May-August in 2004 and 2005.

Site
Sleeping Bear Dunes
National Lakeshore
Ludington State Park
& Nordhouse Dunes
Silver Lake State Park
Muskegon State Park
Hoffmaster State
Park
Kitchel-Lindquist
Dunes Preserve
Saugatuck Dunes
State Park
Oval Beach
Warren Dunes State
Park
TOTAL

Territory
(n)
20

Unoccupied
(n)
27

Total
Area
(ha)
28,808

Potential
Habitat
(ha)
5,289

Average
50%
Territory
Radius (m)
56

Avian
survey
points
(n)
292

0

14

1,797

1,371

32

43

0
0
0

4
4
4

545
410
464

528
160
163

32
32
32

17
23
18

2

3

104

56

24

16

6

4

438

245

32

28

2
8

2
2

42
1,456

16
184

19
32

8
19

38

64

34,064

8,012

32

464

*Amount of potential habitat was calculated by summing sand, scrub-shrub herbaceous and
conifer land cover classes derived from Michigan Department of Natural Resources Integrated
Forest Monitoring, Assessment and Prescription land cover dataset. Site boundaries determined
by state, federal or municipal parcel boundary GIS shapefiles.

26

in selected grid cells. Each unoccupied point was used as the center of an unoccupied vegetation
sampling array, hereafter referred to as a plot.
Vegetation composition (presence – absence for each plant species) was measured in 0.5
m radius circles, serving as subsamples for 2 m- and 30 m-radius circles within each occupied
area (Figure 8). The nested sampling design provides for flexibility in future analyses that could
use multiple resolutions of satellite or aerial imagery (Urban 2001). Differences between
occupied and unoccupied areas were quantified 1) within a site and 2) across sites. Plant species
presence data by survey plot was transformed into relative abundances for each species to
standardize sampling effort between plots. Relative abundances were calculated by dividing the
number of times a plant species was present on each plot by the number of samples in the plot.
Statistical Analysis
Detection Probability
Detection probability generally can be defined as the probability of detecting the species
one is surveying. The goal of calculating detection probability was to determine a level of
confidence associated with the number of Prairie Warblers located on study sites based on the
number of locations surveyed and the number of survey repetitions at each point count point.
Using detection probability models is one way to compare field survey techniques and
subsequent findings (i.e., species detections) to a more accurate and precise estimate of species
occupancy in study areas. Detection-corrected estimates of occupancy were calculated using
Program PRESENCE (MacKenzie and Hines 2008). Models developed by MacKenzie et
al. (2002) were used to calculate detection probability and estimate site occupancy for Prairie
Warbler data collected in years 2004 and 2005 for Sleeping Bear Dunes and the southern sites,
respectively. One sampling covariate, standardized Julian date, was used in the models to assess

27

30m
315°

0º
29m

2m
45°

15m
5m
0.5m

270°

90°

225°

135°
180°

Fig. 8. Nested non-aligned block design used for vegetation sampling in Prairie Warbler
territories and unoccupied sites. Vegetation sampling occurred in the 0.5 m circles. There were
52-0.5 m survey plots per 30 m radius plot.

28

whether date played a significant role in detection.
Vegetation Analyses
Exploratory analyses suggested that many plant species were not distributed across all
survey plots with sufficient relative frequency to differentiate between Prairie Warbler presence
and absence. To reduce the dimensionality of the vegetation data, a preliminary correspondence
analysis was conducted in R (R Development Core Team 2008) using all plant species observed.
Based on this analysis a subset of 46 plant species was selected for further analysis.
Using these 46 species, patterns in plant species composition were examined across the
survey plots in a second correspondence analysis. Although there were 38 occupied and 64
unoccupied 30m radius vegetation survey arrays sampled (totaling 102 plots) (Figure 8), only the
76 survey plots on which Prairie Warblers were detected were used in this correspondence
analysis (Sleeping Bear Dunes, Kitchel-Lindquist, Saugatuck, Oval Beach, Warren Dunes).
The goal of the correspondence analysis was to reduce the dimensionality of the dataset
(76 plots x 46 plants data matrix) to describe how the plant composition differed across the 76
plots. Each axis generated by correspondence analysis represents a linear combination of the
variables describing the abundances of each species. The amount of dimensional reduction
represented by an axis is given by the eigenvalues of a matrix representing dissimilarities in
species composition among plots (Legendre and Legendre 1998). The elements of the
eigenvectors of this matrix associated with plant species represent the relative weighting of each
plant species for a given axis.
A two-way analysis of variance (ANOVA), with an interaction term using the first 4
correspondence axes (dependent variables), was performed to determine 1) whether significant
differences existed in areas where Prairie Warblers were present and absent and 2) whether those
differences varied from north to south. A significant interaction between these 2 factors would
29

indicate that the differences between areas where Prairie Warblers were present and absent
varied between northern (Sleeping Bear Dunes) and southern sites.

RESULTS
Locating Prairie Warbler Breeding Areas
The greatest number of Prairie Warbler occupied areas was found in Sleeping Bear
Dunes (n = 20). Eighteen occupied areas were collectively found in the southern sites, with
Warren Dunes having the most (n=8), followed by Saugatuck Dunes State Park (n=6) and
Kitchel-Lindquist Dunes Preserve and Oval Beach (n=2). Ludington State Park/Nordhouse
Dunes, Silver Lake State Park, Muskegon State Park, and Hoffmaster State Park each had zero
Prairie Warblers detected (Table 1). The average number of mapping visits per occupied area
was 4.35. The average number of GPS registrations per territory was 38 (range 10 – 92 (SE =
3.4) across all sites; range 19 – 81 (SE = 3.7) in Sleeping Bear Dunes; range 10 – 81 (SE = 5.8)
in the southern sites).
Detection Probability
Detection probability for Sleeping Bear Dunes was calculated using 292 point count
locations with 5 sampling occasions per point in 2004. Prairie Warblers were detected at 28
point count locations in Sleeping Bear Dunes. The global model using standardized Julian date
as a covariate on detection was overdispersed (c-hat = 1.947) and performed differently than
expected based on bootstrap verification (chi-squared = 0.010). An intercept-only model, was
subsequently calculated using the c-hat value from the global model to increase power in the
detection probability model (MacKenzie et al. 2006). Both models performed equally well,
leading to selection of the least parameterized model as the best model in this situation (Table 2).
Detection probability for Prairie Warblers in Sleeping Bar Dunes was 0.340 (SE = 0.049),
30

Table 2. Detection probability and occupancy estimates for Prairie Warbler point count surveys
in Sleeping Bear Dunes in 2004 (n = 292) and the 8 southern sites (n = 172).

Model
AIC
Sleeping Bear Dunes
Psi(.),p(.)
362.988
Psi(.),p(Date)
362.480
Southern sites
Psi(.),p(.)
148.630
Psi(.),p(Date)
150.570

AIC wt.

Occupancy
Psi
SE

Detection
Probability
p
SE

0.500
0.500

0.123
0.122

0.023
0.023

0.340

0.049

0.725
0.275

0.138
0.137

0.028
0.028

0.554

0.058

31

indicating that the likelihood of detection, given that a Prairie Warbler occupy a site, is 0.34.
Reliable survey methods should be able to detect a species 15% of the time when the species
occupies an area (Bailey et al. 2004, O’Connel et al. 2006). Occupancy corrected for detection
in Sleeping Bear Dunes was 0.123 (SE = 0.023; Table 2).
Detection probability for the southern sites in 2005 was based on 172 point count locations
with 4 sampling occurrences per site. Prairie Warblers were detected at 22 southern point count
locations. Similar to Sleeping Bear Dunes, I calculated an intercept-only and Julian model
(Table 2). The intercept-only model ranked highest but was overdispersed (c-hat=1.694),
suggesting more variation in the observed data than could be accounted for by the model.
Detection probability on the southern sites was 0.554 (SE = 0.058). Occupancy corrected for
detection probability was assessed at 0.138 (SE = 0.028).
Results indicate that Prairie Warbler detectability is relatively high (i.e, > 0.34) using the
point count methodology. Hence, repeated sampling visits (i.e., > 3) should result in reasonably
accurate occupancy status for a given location.
Vegetation Analysis
Correspondence axis 1 represented the largest amount of variability in the dataset. Axes
1 and 2 represent 15.1% and 11.3% of the variability in the dataset, respectively. Only
results for correspondence axes 1 and 2 will be discussed below as they were the only axes
where differences among breeding and non-breeding areas and differences between geographic
locations were found.
Axis 1 is predominantly influenced by large, positive eigenvector loadings associated
with tree species including sugar maple, Austrian pine (P. nigra), red pine (P. resinosa), jack
pine, green ash (Fraxinus pennsylvanica), red oak and black cherry (P. serotina) (Table 3).
Garlic mustard (Alliaria petiolata), considered a nonnative, invasive species, also had a large
32

Table 3. Plant species most influential in describing differences across vegetation survey plots.
Plants are shown with correspondence analysis four-letter code, common name, genus species,
and eigenvector loadings for axes 1 and 2 which represent 15.1% and 11.3% of the proportional
variance in the dataset respectively. Large eigenvector loadings, driving differences, are bolded.

Spp
ANSP
AUPI
BARB
BEAR
BLCA
BLCH
BOBE
BRFE
CHCH
CHWE
COJU
COLU
CRJU
DWGR
FAHE
FLSP
FMGR
GAMU
GRAS
GRSP
HAWE
HOTA
JAPI
JUGR
LIBL
LOTV
MAGR
MEAD
ORGR
PITH
POIV
QUGR
REOA
REPI
RTGR
RUBU
RUSH

Common name
anenome spp
Austrian pine
barberry species
bearberry
bladder campion
black cherry
bouncing bet
bracken fern
choke cherry
chickweed species
common juniper
columbine
creeping juniper
dune wheat grass
false heather
flowering spurge
fowl meadow grass
garlic mustard
green ash
grass species
hawkweed
horse tail
jack pine
june grass
little bluestem
lily of the valley
marram grass
meadowsweet
orchard grass
pitcher's thistle
poison ivy
quack grass
red oak
red pine
red top grass
rubus species
rush species

Genus species
Anemone spp.
Pinus nigra
Berberis species
Arctostaphylos uva-ursi
Silene latifolia
Prunus serotina
Saponaria officinalis
Ptridium aquilinum
Prunus virginiana
Stellaria spp.
Juniperus communis
Aquilegia canadensis
Juniperus horizontalis
Agropyron dasystachyum
Hudsonia tomentosa
Euphorbia corollata
Poa trivialis
Alliaria petiolata
Fraxinus pennsylvanica
grass spp.
Hieracium spp.
Equisetum hyemale
Pinus banksiana
Koeleria macrantha
Schizachyrium scoparium
Convallaria majalis
Ammophila breviligulata
Spiraea alba
Dactylis glomerata
Cirsium pitcheri
Toxicodendron radicans
Agropyron repens
Quercus rubra
Pinus resinosa
Agrostis gigantea
Rubus spp.
Juncus spp.
33

Axis
1
-0.05
0.43
0.01
-0.08
-0.06
0.15
0.00
0.00
0.08
-0.03
-0.04
0.02
-0.04
-0.03
0.00
0.00
0.00
0.27
0.19
-0.01
-0.04
-0.03
0.13
-0.05
-0.08
0.09
-0.04
0.02
-0.03
-0.03
-0.01
0.01
0.15
0.25
-0.01
0.01
0.00

Axis
2
0.09
0.02
0.00
0.40
-0.05
-0.01
-0.12
-0.02
-0.05
-0.13
0.23
0.01
0.50
-0.08
-0.06
-0.08
-0.02
0.00
-0.01
0.13
0.13
-0.17
0.00
-0.14
-0.08
0.00
-0.38
-0.03
-0.09
-0.06
0.14
-0.01
-0.03
0.02
0.25
0.10
-0.10

Table 3. (cont’d)
SFSS
SNOW
SOGE
SPKN
SRGR
SUMA
WHAS
WORM

star.f.solomon's seal
snowberry, common
soapwort gentian
spotted knapweed
sand reed grass
sugar maple
white ash
wormwood

Smilacina stellata
Symphoricarpos albus
Gentiana saponaria
Centaurea maculosa
Calamovilfa longifolia
Acer saccharum
Fraxinum americana
Artemisia campestris

34

0.06
0.01
0.00
-0.08
-0.06
0.71
0.06
-0.08

0.16
0.01
-0.03
0.05
-0.27
0.00
-0.03
-0.15

positive eigenvector loading associated with axis 1. Choke cherry (P. virginiana) and white ash
(F. americana), though not as strongly associated, are 2 additional tree species with positive
loadings on axis 1.
Axis 2 was influenced by a broader diversity of plant types with large loadings (Table 3).
Large, positive eigenvector loadings were associated with 3 shrub and 1 grass species: creeping
juniper, bearberry, common juniper and redtop grass (Agrostis gigantea) (Table 3). Axis 2 had
large negative loadings associated with other grasses including marram grass, sand reed grass
and june grass (Koeleria macrantha). Three herbaceous plant species have large positive
loadings on axis 2: starry false Solomon’s seal (Smilacina stellata), poison ivy (Toxicodendron
radicans) and hawkweed (Hieracium spp.), while several more herbs have large negative
loadings on axis 2: horsetail (Equisetum hyemale), wormwood (Artemisia campestris) and
various chickweed species (Stellaria spp.).
There was a significant interaction between sites where Prairie Warblers were present and
absent and differences between sites from north to south existed for axis 1 (p < 0.0001) (Table
4). The ANOVA results for axis 2 also showed a significant interaction, indicating a significant
difference between occupied and unoccupied sites as well as significant variation between those
differences in occupied and unoccupied sites from north to south (p < 0.0001) (Table 4).
Means were plotted for axis 1 and illustrate the difference between southern sites where
Prairie Warblers were present and absent (Figure 9). Plant species associated with sites where
Prairie Warblers were present in the south were primarily negatively loaded on both axes 1 and 2
and included grass and herbaceous species like marram grass, sand reed grass, wormwood,
horsetail and chickweed species (Table 3, Figure 10). However, plant species associated with
southern sites where Prairie Warblers were absent, while not clearly defined, were primarily
influenced by tree species positively loaded on axis 1. These species include sugar maple,
35

Table 4. Two-way analysis of variance (ANOVA) for correspondence axes 1-4 (dependent
variables) to test for differences between areas where Prairie Warblers were present and absent
(presence) and whether differences varied between northern Sleeping Bear Dunes and southern
sites (site). Significant (α <0.05) results are bolded with axes 1 and 2 showing significant
interactions. DF = degrees of freedom. SS = sum of squares. MS = mean square.
Axis 1
presence
site
presence:site
Residuals
Axis 2
presence
site
presence:site
Residuals
Axis 3
presence
site
presence:site
Residuals
Axis 4
presence
site
presence:site
Residuals

DF
1
1
1
72
DF
1
1
1
72
DF
1
1
1
72
DF
1
1
1
72

SS
3.0620
13.5960
10.5210
53.7250
SS
7.2400
8.2211
7.3923
10.6366
SS
0.7214
1.1754
0.3285
28.3938
SS
1.4060
0.0756
0.0296
30.6442

MS
3.0620
13.5960
10.5210
0.7460
MS
7.2400
8.2211
7.3923
0.1470
MS
0.7214
1.1754
0.3285
0.3940
MS
1.4060
0.0756
0.0296
0.4250

36

F Value
4.1042
18.2211
14.0999

P Value
0.0465
<0.0001
<0.0001

F Value
49.0080
55.6500
50.0390
7.0000
F Value
1.8293
2.9806
0.8330
4.0000
F Value
3.3036
0.1777
0.0695
6.0000

P Value
<0.0001
<0.0001
<0.0001
P Value
0.1805
0.0886
0.3645
P Value
0.0733
0.6746
0.7929

Axis 1
South

1.5
1
0.5
Mean
0

South
SBD
North

SBD
North

-0.5
-1

Presence
Territory

Absence
Unoccupied
Axis 2

1.5
1

SBD
North

0.5
Mean

0

-0.5

South

South
SBD
North

-1

Presence
Territory

Absence
Unoccupied

Fig. 9. Difference between Prairie Warbler occupied (solid shape) and unoccupied (hollow
shape) areas in the north (red circles) vs. the south (black triangles). Mean eigenvector values
were plotted for northern Sleeping Bear Dunes and southern sites with standard error bars for the
first 2 correspondence axes.

37

CRJU
BEAR
COJU
RTGR
SFSS
HAWE
POIV

SUMA
AUPI
REPI
JAPI
GAMU

MAGR
SRGR
JUGR
HOTA
WORM
CHWE

GRAS
REOA
BLCH
CHCH
WHAS

Fig. 10. Correspondence analysis plot for the 46 plant species. Plant species driving differences
across Prairie Warbler presence and absence sites are represented by 4-letter codes in each
quadrant (see Table 3 for plant species names and eigenvector loadings corresponding to axes).
Sites occupied by Prairie Warbler in Sleeping Bear Dunes are represented by solid (red) circles
and in southern sites by solid (black) triangles. Unoccupied sites in Sleeping Bear Dunes are
represented by hollow (red) circles and in southern sites by hollow (black) triangles.
Correspondence axis 1 (x axis) and 2 (y axis) represent 15.1% and 11.3% of the proportional
variance in the dataset, respectively.

38

Austrian pine, red pine, jack pine, choke cherry, white ash, and garlic mustard (Table 3, Figure
10). This suggests that Prairie Warblers in the south may be avoiding areas with tree species and
selecting areas with more grass and herbaceous plant species.
Means were plotted for axis 2 and illustrate the difference between northern areas in
Sleeping Bear Dunes where Prairie Warblers were present and absent (Figure 9). Plant species
associated with areas where Prairie Warblers were present in the north were positively loaded on
axis 2 (and negative on axis 1). These plants included shrubs like creeping juniper, bearberry
and common juniper (Figure 9). Redtop grass, hawkweed and poison ivy were also associated
with areas where Prairie Warblers were present in Sleeping Bear Dunes (Table 3, Figure 10).
Plant species associated with sites where Prairie Warblers were absent in the north were
primarily negatively loaded on both axes 1 and 2 and included many of the same grass and
herbaceous species present marram grass, sand reed grass, wormwood, horsetail, chickweed
species). This suggests that Prairie Warblers in the north may be selecting sites with more shrub
species than Prairie Warblers in the south.
Some of the same plants were common drivers between occupied and unoccupied
southern sites and unoccupied areas of Sleeping Bear Dunes. While it can be difficult to
summarize the complete variability of areas where Prairie Warblers (or any species) are absent, it
is important that this study indicates that unoccupied areas are significantly different from areas
where Prairie Warblers are present, both in the north and in the south. Additionally, plant
composition associated with Prairie Warbler territories in the north is significantly different than
plant composition associated with Prairie Warbler territories in the south.

39

DISCUSSION
The objectives of this study have been met. Prairie Warblers were located along the west
Michigan shoreline with high detection probability. Plant composition varied across the plots
surveyed and aspects of the differences can be related to specific relationships illustrated by
correspondence analysis as outlined above. Plant composition in areas occupied by Prairie
Warblers was found to be significantly different from plant composition in unoccupied areas.
However, those differences were not consistent from north to south; Prairie Warblers breeding in
northern areas had different plant species associated with them than did Prairie Warblers
breeding areas in the southern part of the state. Overall, findings from this study should aid
understanding of the differences in plant composition associated with Prairie Warbler breeding
areas in different geographies along Michigan’s west coast. Findings should better inform
management and land use decision-making in areas this state-endangered species is, or could
potentially be, using.
A main difference between northern and southern breeding areas tended to be the
presence of shrubs versus grass; breeding areas in Sleeping Bear Dunes had more shrubs present
than southern breeding areas did. Shrubs were largely absent on northern unoccupied plots,
which shared some of the same plant species (grasses) as southern breeding areas, reinforcing the
idea that there could be a change in habitat selection strategies from north to south along
Michigan’s west coast.
Why might these differences in vegetation from north to south exist and what does that
mean for Prairie Warbler conservation and management? Michigan’s climate and its glacial
history yield a tension zone running east to west, roughly through the center of Michigan’s
Lower Peninsula, lending to some transition in vegetation community types distributed north to

40

south along the lakeshore. Forest type transitions of primarily hardwood and associated
herbaceous assemblages in the south, to mixed hardwood and coniferous forests with associated
plant assemblages in the north are general descriptions of this transition. It is possible the northto-south gradient in vegetation communities combined with differences in land use patterns (i.e.,
development, fragmentation) could result in different habitat selection strategies for Prairie
Warblers north to south along the lakeshore. With regard to understanding these differences in
the context of Prairie Warbler ecology, it may be important to gain additional insight into the
specific plant assemblages serving the most essential functions in the breeding areas (i.e., nesting
structure) so conservation and management actions consider these fine scale, high use habitat
components. Reducing anthropogenic stressors on the dune ecosystem, such as incompatible
development resulting in land use change and introduction of non-native, fast-spreading plants,
may safeguard the variety and availability of habitats that exist, and thus protect some of the only
areas in which Prairie Warblers have been observed in Michigan in recent years (Michigan
Natural Features Inventory 2011).
This study’s findings suggest that differences in vegetation composition between northern
and southern breeding territories are a finer definition of habitat parameters for this species than
has previously been documented. Learning more about the diversity of habitat Prairie Warblers
may be using throughout Michigan could lend insight into future management opportunities. For
example, surveying for Prairie Warblers in jack pine plains burned 10-20 years previously (i.e.,
Kirtland’s Warbler habitat in the northeast Lower Peninsula). Prairie Warblers were more
prevalent in the northeast Lower Peninsula in the early 1900’s. Special attention to monitoring
for Prairie Warblers in addition to Kirtland’s Warblers during annual Kirtland’s Warbler surveys

41

is recommended to assess whether populations of Prairie Warblers may still be using parts of this
area.
Nolan’s (1978) 20-year study on Indiana Prairie Warbler populations (1951-1972)
suggests that Prairie Warblers and their offspring will continue to return to the same breeding
grounds until the area succeeds. Michigan’s west coast may be providing a constant, or
guaranteed, source of early successional habitat due to the disturbance inherent to the ecosystem.
This could be the reason for their success in this region of the state. However, while the coastal
areas surveyed in this study seem to support the majority of this species’ population in Michigan
(Michigan Natural Features Inventory 2011), surveys in other regions of the state could help
better document the potential range of variability in plant composition associated with other
areas Prairie Warblers find suitable.
This information will become increasingly important given the range of potential
landscape composition changes projected in the Midwest and Great Lakes region under a
changing climate (U.S. Global Change Research Program 2009). Michigan lies along the
northern edge of the Prairie Warbler breeding range and the contribution of this study to Prairie
Warbler ecology is the first of its kind in Michigan and the northern Great Lakes region.
Understanding Prairie Warbler habitat selection strategies at the northern edge of the breeding
range will expand our overall understanding of Prairie Warbler ecology for future natural
resource management and conservation planning. Differences in habitat selection strategies we
observe to the north and to the south of the glacial tension zone in Michigan, may be a unique
opportunity to examine species adaptation strategies at the edge of adjacent plant communities,
potentially informing adaptive behaviors as species ranges continue northward shifts under a
changing climate.

42

LITERATURE CITED

43

LITERATURE CITED

Albert, D. A. 2000. Born of the wind: an introduction to the ecology of Michigan sand
dunes. Michigan Natural Features Inventory, Lansing, Michigan, USA.
Bailey, L. L., T. R Simons, and K. H. Pollock. 2004. Estimating site occupancy and detection
probability parameters for terrestrial salamanders. Ecological Applications 14:692-702.
Brewer, R., G. A. McPeek, and R. J. Adams, Jr. 1991. Atlas of breeding birds in Michigan.
Michigan State University Press, East Lansing, Michigan, USA.
Cooper, J. L. 2000. Special animal abstract for Dendroica discolor (prairie warbler). Michigan
Natural Features Inventory, Lansing, Michigan, USA.
Dunn, J. L., and K. L. Garrett. 1997. A field guide to warblers of North America. Houghton
Mifflin Co., Boston, Massachusetts, USA.
Hazlett, B. T. 1991. The flora of Sleeping Bear Dunes National Lakeshore, Benzie and
Leelanau Counties, Michigan. The Michigan Botanist 30(4):139-207.
Hooge, P. N., and B. Eichenlaub. 1997. Animal movement extension to ArcView 3.2, Version
1.1. Alaska Science Center - Biological Science Office, U.S. Geological Survey.
Anchorage, Alaska, USA.
Jenness, J. 2003. Random point generator (randpts.avx) extension for ArcView 3.x, Version
1.27. Jenness Enterprises. www.jennessent.com/arcview/ random_ points.htm. Accessed
2004 and 2005.
Legendre, P., and L. Legendre. 1998. Numerical Ecology. Second English edition. Elsevier,
Amsterdam, The Netherlands.
MacKenzie D. I., J. D. Nichols, G. B. Lachman, S. Droege, J. A. Royle, and C. A.
Langtimm. 2002. Estimating site occupancy rates when detection probabilities are less
than one. Ecology 83:2248-2255.
MacKenzie, D. I., J. D. Nichols, J. A. Royale, K. H. Pollock, L. L. Bailey, and J. E.
Hines. 2006. Occupancy estimation and modeling: inferring patterns and dynamics of
species occurrence. Academic Press, New York, New York, USA.
MacKenzie, D., and J. Hines. 2008. Program PRESENCE (2). www.mbrpwrc.usgs.gov/software/doc/presence/presence.html. Accessed September 2008.

44

Michigan Department of Natural Resources. 2001. Integrated forest monitoring, assessment,
and prescription (IFMAP) land classification system. www.michigan.gov/dnr/0,1607,7153-30301_39170-144990--,00.html. Accessed 2004-2008.
Michigan Natural Features Inventory. 2011. Rare species explorer. http://web4.msue.msu
.edu/mnfi/explorer. Accessed 2007-2011.
Nolan, V., Jr. 1978. The ecology and behavior of the prairie warbler Dendroica
discolor. Ornithological Monographs, No. 26.
Nolan, V., Jr., E. D. Ketterson, and C. A. Buerkle. 1999. The birds of North America: Prairie
Warbler (Dendroica discolor). Pages 1-26 in A. Poole and F. Gill, editors. The Birds of
North America, No. 584. Philadelphia, Pennsylvania, USA
O’Connel, A. F., Jr., N. W. Talancy, L. L. Bailey, J. R. Saurer, R. Cook, and A. T. Gilbert.
2006. Estimating site occupancy and detection probability parameters for meso- and
large mammals in a coastal ecosystem. Journal of Wildlife Management 70(6):16251633.
R Development Core Team. 2008. R: a language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria. www.R-project.org. Accessed
2007-2008
Ralph, C. J., S. Droege, and J. Sauer. 1995. Managing and monitoring birds using point
counts: standards and applications. U.S. Forest Service General Technical Report. PSWGTR-149:161-169.
Silverman, B.W. 1986. Density estimation for statistics and data analysis. Chapman and
Hall, London, UK. http://nedwww.ipac.caltech.edu/level5/March02/Silverman/
Silver_contents.html. Accessed October 2008.
Southwell, D. K. 2001. Conservation assessment for Prairie Warbler (Dendroica discolor).
USDA Forest Service Eastern Region. USDA Forest Service, Hiawatha National Forest,
Michigan, USA.
Space Imaging. 2004. Review of remote sensing technologies used in the IFMAP project. Final
report. Ann Arbor, Michigan, USA.
Walkinshaw, L. H. 1991. Prairie warbler. Pages 418-419 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.
Urban, D. L. 2001. Tactical monitoring of landscapes. Pages 294-311 in J. Liu and W.
W. Taylor, editors. Integrating landscape ecology into natural resource management.
Cambridge University Press, Cambridge, England, UK.

45

U.S. Global Change Research Program. 2009. Global climate change impacts in the
United States. Regional climate impacts: Midwest. Cambridge University Press, New
York, New York, USA.

46

CHAPTER 3

PRAIRIE WARBLER OCCUPANCY AND RESOURCE USE IN MICHIGAN DURING THE
BREEDING SEASON

INTRODUCTION
A diversity of migratory birds depend on the relatively small but ecologically valuable
Lake Michigan shoreline, foredune, and backdune forest systems (Albert 2000). Migratory birds
are known to use these dune ecosystems as stopover (Mehlman et al. 2005) and breeding habitats
(Brewer et al. 1991). These unique systems (275,000 acres of sand dune along Michigan’s Great
Lakes shoreline; Michigan Sea Grant 2012), sustain specialized plant communities that are
adapted to the extreme temperature and weather patterns inherent to dune landscapes (Albert
2000). State-endangered Prairie Warbler (Dendroica discolor) observations in Michigan are
almost exclusively from the Lake Michigan early-successional dune ecosystems (Michigan
Natural Features Inventory 2011). Describing plant composition in areas selected by Prairie
Warblers is a critical information need (Nolan et al. 1999, Cooper 2000, Southwell 2001).
Understanding Prairie Warbler habitat selection during the breeding season can inform
conservation and management.
A breeding territory describes an animal’s spatial and temporal use of space (Aebischer et
al. 1993). However, resource use within a territory or home range is rarely uniform (Marzluff et
al. 1997). Differential resource use can be quantified using utilization distributions (UD; van
Winkle 1975, Kernohan et al. 2001). A UD is a continuous relative frequency distribution
derived from point locations recorded for an animal over a period of time (van Winkle 1975).
The highest value in the UD corresponds to the area(s) most heavily used by an animal.
Utilization distributions have regularly been used to define core use areas within home ranges
47

(Samuel et al. 1985), assess animal interactions (Seidel 1992, Millspaugh et al. 2000), and
incorporate geographically-referenced time budgets into home range calculations (Samuel and
Garton 1987). Developing UDs for Prairie Warblers and relating the UDs to environmental
covariates in a spatially explicit manner can be achieved using resource utilization functions
(RUF) (Manly et al. 2003, Marzluff et al. 2004). RUFs enable continuous, rather than discrete
(used-not used), resource use assessments across landscapes. The regression coefficients in the
RUF indicate the importance of each resource to variation in the UD.
For Michigan Prairie Warbler populations, information on within-territory habitat use
will contribute to our understanding of habitat selection strategies. Additionally, data on withinterritory habitat use will provide information at a resolution previously not evaluated. Prairie
Warbler breeding habitat is generally characterized as having little tree canopy cover and poor
dry sandy soils that support nesting vegetation structure between 1 and 3 m tall (Cooper 2000,
Dunn and Garrett 1997, Walkinshaw 1991). The first survey record for a Prairie Warbler in the
United States (April 18, 1810) described habitat as “barrens” and “prairie country…evidently
resembling and perhaps representative of the tall-grass prairie” (Mengel 1965). The bird’s name,
derived from this description by Alexander Wilson (Nolan 1978), does not fully describe the
diversity of locations the bird inhabits within its current breeding range.
Overall, the Lake Michigan sand dunes have been designated as critical breeding areas
for this species in Michigan (Walkinshaw 1991). Mayfield (1960) also suggested that
Michigan’s jack pine (Pinus banksiana) plains may be important. Michigan Prairie Warbler
populations have been observed breeding in three general vegetation types (Walkinshaw 1991,
Cooper 2000), although this is not viewed as an exhaustive list of potential habitat: 1) early
successional stages of Great Lakes sand dunes where deciduous or coniferous shrubs are

48

intermixed with dune grass and other herbs, 2) early-successional, shrubby openings associated
with previously-burned jack pine plains, 3) previously burned conifer stands now dominated by
deciduous shrubs and small trees.
The objectives of this study were to 1) determine which land cover types can forecast
Prairie Warbler occupancy in west Michigan dune landscapes, 2) develop resource utilization
functions for Prairie Warbler breeding territories along Michigan’s west coast, and 3) gain
insight into how resource use may differ for Prairie Warblers between northern and southern
coastal breeding territories in Michigan.

STUDY SITE
Nine study sites were located throughout the Lower Peninsula of Michigan along the
Lake Michigan coast (Figure 11). Study sites were chosen because they were publically
accessible and contained potential Prairie Warbler habitat (identified by cover type (i.e., jack
pine, sand dunes, scrub-shrub; Nolan 1978, Nolan et al. 1999, Cooper 2000, Southwell 2001).
Study sites included six State Parks, Sleeping Bear Dunes National Lakeshore, Oval Beach (a
public beach owned by the city of Saugatuck), and Kitchel-Lindquist Dunes Preserve (a
publically-accessible preserve near Grand Haven) (Figure 11).
Foredune vegetation communities consisted of specialized plants adapted to withstand
sand burial and exhibit fast growth with deep, often rhyzomal, root systems that stabilize
blowing sand (Albert 2000). Pioneering grasses such as marram grass (Ammophila
breviligulata) and sand-reed grass (Calamovilfa longifolia) cause sand to accumulate and
stabilize. Common Michigan foredune shrubs include woody species such as sand cherry
(Prunis pumila), red-osier dogwood (Cornus stolonifera), bearberry (Arctostaphylos uva-ursi),

49

0

50

100

200 Kilometers

¯

Fig. 11. Nine Prairie Warbler survey locations along the Lower Peninsula of Michigan. Surveys
took place from May-August in 2004 and 2005. Distance between Warren Dunes State Park and
Sleeping Bear Dunes National Lakeshore is approximately 386 km (240 miles). Prairie Warblers
were detected in Sleeping Bear Dunes in 2004 and in Kitchel-Lindquist Dunes Preserve,
Saugatuck Dunes State Park, Oval Beach and Warren Dunes State Park in 2005.

50

several dune willows (Salix species), and creeping juniper (Juniperus horizontalis). These plants
must tolerate extreme temperatures, low moisture levels, and low levels of nutrients in addition
to blowing sand.
The foredune zone undergoes continual change due to natural weather events, erosion,
and water level fluctuations. Larger dunes form behind the protective foredune where sand
accumulation is slower, allowing dune grasses, herbaceous plants, shrubs and forests to establish.
Oaks (Quercus species), bigtooth aspen (Populus grandidentata) and sassafras (Sassafras
albidum), in addition to patches of jack pine and common juniper (J. communis), grow closest to
the shoreline on wind-ward slopes. In contrast, the land-ward side of the dunes is moister and
more nutrient-rich, supporting mesic forest species like sugar maple (Acer Saccharum),
American beech (Fagus grandifolia), basswood (Tilia americana), and red oak (Q. rubra).
Prairie warblers are most typically found in the early successional, transition zone between open
sand and forest.
Anthropogenic effects from development and introduction of exotic invasive plants are
stressors on the dune ecosystem. Anthropogenic development potentially threatens the variety
and availability of habitats, thus threatening some of the only areas in which Prairie Warblers
have recently been observed in Michigan (Michigan Natural Features Inventory 2011).
Introduction of exotic invasive plants can alter the plant community composition and diversity
(Levine et al., 2003); with potential negative consequences on Prairie Warbler habitat suitability.
Prairie Warblers arrive on Michigan breeding grounds within the first 2 weeks of May
(Dunn and Garret 1997). The mean temperature in 2004 and 2005 at Sleeping Bear Dunes
during this time (May 1-14) was 11°C and 9°C, respectively; Warren Dunes temperatures were
14°C and 12°C, respectively. These year-specific temperatures compare closely to average

51

annual temperatures in Sleeping Bear Dunes and Warren Dunes during May: 13°C and 14°C,
respectively. Average precipitation for May is 6.6 cm in Sleeping Bear Dunes and 8.1 cm in
Warren Dunes. Temperatures increase throughout the Prairie Warbler breeding season in
Michigan with variable rainfall. Fall avian observation records suggest that Prairie Warblers
leave breeding grounds to migrate south by the third week of August (Michigan Natural Features
Inventory 2011).

METHODS
Avian Surveys
I used Michigan Department of Natural Resources (DNR) Integrated Forest Monitoring,
Assessment and Prescription (IFMAP; Michigan Department of Natural Resources 2001) land
cover data to inform the selection of avian point count locations that were sampled in 2004 and
2005. The IFMAP data were derived from classified Landsat Thematic Mapper 5 and 7 imagery,
dated 1999-2001. Overall accuracy was 88% for Anderson Level 2 categories (major land cover
classes; scale of 1:40,000; Space Imaging 2004). I aggregated IFMAP data from 35 original
land cover classes into 9 more general classes (Urban, Agriculture, Herbaceous-Shrub Upland,
Upland Deciduous Forest, Upland Coniferous Forest, Water, Lowland Forest, Nonforested
Wetland, Sand/Bare Soil). Equal numbers of survey points were distributed within 3 of the 9
land cover classes (Sand/Bare Soil, Herbaceous-Shrub Upland, and Upland Coniferous Forest)
known to represent Prairie Warbler habitat throughout their breeding range (Nolan 1978, Nolan
et al. 1999, Cooper 2000, Southwell 2001).
Avian point count locations were randomly chosen using a GIS random point generator
(Jenness 2003) in suitable land cover types (Table 5). Study site boundaries were acquired

52

Table 5. Total area (ha) and potential habitat (ha)* of 9 study locations used to survey Prairie
Warblers in Michigan from May-August in 2004 and 2005.

Site
Sleeping Bear Dunes
National Lakeshore
Ludington State Park
& Nordhouse Dunes
Silver Lake State Park
Muskegon State Park
Hoffmaster State
Park
Kitchel-Lindquist
Dunes Preserve
Saugatuck Dunes
State Park
Oval Beach
Warren Dunes State
Park
TOTAL

Territory
(n)
20

Unoccupied
(n)
27

Total
Area
(ha)
28,808

Potential
Habitat
(ha)
5,289

Average
50%
Territory
Radius (m)
56

Avian
survey
points
(n)
292

0

14

1,797

1,371

32

43

0
0
0

4
4
4

545
410
464

528
160
163

32
32
32

17
23
18

2

3

104

56

24

16

6

4

438

245

32

28

2
8

2
2

42
1,456

16
184

19
32

8
19

38

64

34,064

8,012

32

464

* Amount of potential habitat was calculated for initial point count surveys by summing sand,
scrub-shrub herbaceous and conifer land cover classes derived from Michigan Department of
Natural Resources Integrated Forest Monitoring, Assessment and Prescription land cover dataset.
Site boundaries determined by state, federal or municipal parcel boundary GIS shapefiles.

53

from state, federal or municipal entities. Two-hundred ninety-two point count survey points
were established in Sleeping Bear Dunes in 2004 (5 sampling occurrences per point). The
number of survey points in each of the 8 sites south of Sleeping Bear Dunes was selected in
proportion to the total area of each site, totaling 172 point count survey points across those 8
sites in 2005 (4 sampling occurrences per point).
Locating Prairie Warblers was the primary objective of the point counts, therefore when a
Prairie Warbler was detected while walking between point count locations, a GPS location was
recorded and added to the point count schedule. All avian survey points were at least 200 m
apart. Prairie Warblers were identified using 10-minute point counts (Ralph et al. 1995, Cooper
2000) May 5 – July 28 in 2004 and May 3 – June 22 in 2005. Surveys occurred from 30 minutes
before sunrise (civil twilight) until 10:30AM in the absence of rain, heavy fog or high winds
(Ralph et al. 1995). Point count locations were surveyed with a rotating time and observer
schedule to reduce biases.
Territory Mapping
I conducted territory mapping in all areas occupied by Prairie Warblers in 2004 and 2005.
Mapping began when Prairie Warbler presence was consistently recorded during repeated point
count surveys. Mapping visits continued until ≥25 registration points (collected with GPS) per
territory were collected. Registrations included male singing locations, feeding sites, or
locations where defense of the territory boundaries was observed. Each territory was visited ≥3
times. Each mapping visit lasted approximately 1 hour. The time of day a bird was mapped
alternated so that early morning, mid-morning, and afternoon behaviors and movements were all
included in territory delineation.

54

Unoccupied Sites
Unoccupied survey points were chosen randomly from avian point count survey points on
which Prairie Warblers were not detected. Unoccupied points were buffered by the average
radius of the 50% use percentile, per site, as calculated from territories (Table 5). For sites
without territories I used the average radius from the most proximate occupied site (Table 5). I
used a fixed kernel home range estimator with least squares cross validation to estimate the 50%
use percentile (Silverman 1986). Fifty percent probability contours were chosen to represent
areas of concentrated use.
Occupancy Analysis
I used logistic regression to develop a predictive model of Prairie Warbler occurrence
based on environmental covariates. I used 1 m National Agriculture Imagery Program (NAIP;
United States Department of Agriculture 2005) imagery around each occupied and unoccupied
point to quantify distances to specific cover types. Michigan NAIP imagery was collected
between June and August, 2005. The images were available in county mosaics in natural color
or color-infrared. The data met the 1:12,000 National Map Accuracy Standard (84% at
Anderson level 4) (United States Bureau of the Budget 1947).
I explored the hypothesis that Prairie Warbler breeding territory occurrence was
positively related to proximity of early successional land cover and to other Prairie Warbler
territories. I defined 6 land cover types that could be identified from NAIP and were commonly
observed around avian survey points: 1) developed, 2) sand, 3) grass, 4) shrub, 5) forest, and 6)
water (Table 6). I measured distances from the nearest territory core area (50% use percentile)
and from the edge of buffered unoccupied points to each land cover type using ArcMap 9.2
(ruler tool). I also measured the distance to nearest Prairie Warbler territory.

55

Table 6. Common, general land cover classes observed around Prairie Warbler breeding areas
along Michigan’s west coast.
Land Cover Type
Sand
Grass-herbaceous
Shrub

Forest
Lake Michigan
Developed

Closest Prairie Warbler
occupied area

Definition
Open, “bright”sand with no visible vegetation
Short and tall grasses and herbaceous vegetation indicated by
“grayish” patches of sand that have no visible trees or large shrubs
Patches of shrubs, saplings and occasional large trees that are
disjunct from large, contiguous forest patches. This included large
islands of “shrub” within the dune ecosystem.
Large, contiguous patches of dense canopy forest that boarder the
sand dune ecosystem.
Water line visible in the NAIP image for Lake Michigan
Paved or unpaved roads, buildings, State campgrounds, Stateowned parks with mowed lawns; note “developed” structures (e.g.,
cabin) under the canopy could not be detected
Kernel-estimated 50% territory probability contours (see Chapter 2)

56

I used distances to each land cover type and to Prairie Warbler territory as the predictor
variables in a generalized logistic regression (Legendre and Legendre, 1998). Logistic
regression is a special case of the generalized linear model (McCullaugh and Nelder 1983). It
uses binary response variables (presence and absence) and estimates the probability of an effect
(presence) given quantifiable circumstances (distances to land cover types or to another Prairie
Warbler). I tested for multicollinearity among independent variables using Pearson correlations
(Legendre and Legendre, 1998) in the R statistical and graphing environment (R Development
Core Team 2011). Correlated variables were not included in the same candidate model.
I evaluated a full model (with all 6 independent variables) and a reduced model that
included only significant variables (α < 0.05). Model goodness-of-fit was assessed by
calculating the amount of variation accounted for by each model. Coefficient significance was
tested using Wald’s chi-square test (z scores). A chi-squared test was used to determine whether
the 2 models were significantly different from each other. Significant variables from the reduced
model were plotted against the regressed occupied and unoccupied survey points to illustrate the
probability of Prairie Warbler territory establishment as a function of distance to each variable.
Within-Territory Resource Utilization
I followed methods described by Marzluff et al. (2004) to explore resource use by Prairie
Warblers within their breeding territories. I related geographic space use by Prairie Warblers, as
portrayed by a UD, to environmental attributes using resource utilization functions (RUF; Manly
et al. 2003, Marzluff et al. 2004). Utilization distributions can be related to environmental
attributes using multiple regression (Marzluff 1986, Marzluff et al., 2004). The resulting
regression equation is the RUF, where variation in the height of the UD (dependent variable) is
attributed to variation in measured resources (independent variables) that underlie each UD

57

location. The regression coefficients in the RUF indicate the magnitude and direction of animal
responses to resources within a breeding territory.
I used R (R Development Core Team 2011) to calculate a UD for each Prairie Warbler
breeding territory using a fixed-kernel estimation with least squares cross validation (LSCV)
(Silverman 1986, Seaman and Powell 1996, Kernohan et al. 2001, Gitzen et al. 2006). Kerneldensity estimation is favored over other territory estimation approaches (e.g., minimum convex
polygon, bivariate-normal methods) for data with complex spatial patterns (Marzluff et al. 2004).
The kernel method is sensitive to choice of bandwidth (or smoothing parameter)
(Silverman 1986, Seaman et al. 1999, Kernohan et al. 2001, Gitzen et al. 2006). The bandwidth
controls the width of the individual kernels and therefore determines the amount of smoothing
applied to the data (Kernohan et al. 2001). A large bandwidth value produces a smooth UD,
which overestimates breeding territory size (Kernohan et al. 2001). Small bandwidths produce
potentially fragmented breeding territories (Kernohan et al. 2001, Gitzen et al. 2006). Recent
research indicates that the LCSV bandwidth performs best when distributions of registration
points are in tight clumps (Millspaugh et al. 2006, Gitzen et al. 2006), consistent with Prairie
Warbler registration points. The LSCV bandwidth also tends to perform better than other
methods tested (e.g., solve-the-equation, plug-in) for estimating areas of peak use (e.g.,
measuring UD; Gitzen et al. 2006).
The kernel smoothing (KS) package was used in R, with least-squares-cross-validation
(hlscv.diag R command) bandwidth. Hawth’s Tools (Beyer 2004) was used to calculate 99%
breeding territory boundaries from Prairie Warbler registration points (Kernel Tools – Percent
Volume Contour) and I subsequently used ArcMAP 9.3 to clip the UD with the 99% fixed-kernel
territory boundary. This process resulted in a continuous UD surface that corresponded to the

58

99% territory. Cell size for the grid that represented each territory ranged from 2 - 8 m (varied
depending on the size of the territory). I quantified the heights of the UD (e.g., Figure 12) for
each grid cell in each breeding territory. The UD height was used as the response variable for
the RUF.
I evaluated multiple land cover base map data sources (e.g., Michigan National
Agricultural Imagery Program, Coastal Change Analysis Program (C-CAP), IFMAP, National
Land Cover Database) for this project. Although I used 1 m NAIP county mosaics for the
occupancy analysis, I chose 30 m C-CAP data for the within-territory analysis. The advantage of
C-CAP imagery was that it was classified by 24 land cover types, which could be aggregated
into land cover classes that theoretically influenced Prairie Warbler space use. The classified
data provided an available, large-scale data source for repeating my analytical methods for all
study locations (compared to NAIP unclassified images). C-CAP had a high accuracy rating and
was deemed suitable for portraying the early successional components of the dune landscape
important to Prairie Warblers.
The 30 m resolution C-CAP imagery was collected April 7, 2005 through October 1,
2006 (“leaf on” and “leaf off”), using Landsat Thematic Mapper and Landsat Enhanced
Thematic Mapper (National Oceanic and Atmospheric Administration, Coastal Services Center
2008). The data met the 1:12,000 National Map Accuracy Standard (85% overall accuracy at
Anderson level 4) (United States Bureau of the Budget 1947) with an overall accuracy
assessment of 91.47% (National Oceanic and Atmospheric Administration, Coastal Services
Center 2008). The C-CAP imagery contained 24 original land cover classes. I used 7 general
land cover classes for this study: 1) developed, 2) sand, 3) grass, 4) shrub, 5) forest, 6) water, 7)

59

A.

B.
Fig. 12. (A) Example of multi-dimensional utilization distribution of a Prairie Warbler breeding
territory compared to (B) territory mapping registration points for the same territory. The
highest density of points in (B) is represented by the “highest” peak in (A).

60

wetland (Table 7). I quantified the percent of land cover classes for each cell within Prairie
Warbler territories using a focal mean on a 3x3 neighborhood of cells.
I calculated a topographic wetness index (TWI) for the study area using the Topographic
Wetness Index ArcScript in ArcGIS 9.3 (ESRI 2010). The TWI is a function of the natural
logarithm of the ratio of local upslope contributing area and slope (ESRI 2010, Beven and Kirby
1979). The TWI was used as an index of soil moisture for 90 m cells within the study area and
was calculated using a 30 m digital elevation model. High TWI values indicate wetness while
low TWI values indicate dryness; TWI should thus portray wetness as influenced by topography.
I overlaid Prairie Warbler UDs and GIS layers of the independent variables (i.e., focal
mean percent cover for each land cover class, focal mean of TWI, and nearest distance to other
Prairie Warbler territory) and attributed each location in the UD with the independent variable
value. Prior to RUF analysis (Handcock 2004, Handcock and Stein 1993), regression
assumptions of normality, homoscedasticity and multicollinearity of independent variables were
evaluated. The data were standardized to a mean of 0 and standard deviation of 1. I checked for
multicolinearity among independent variables using a correlation matrix of Pearson correlation
coefficients. I checked for homoscedasticity by evaluating residual plots. Normality was
examined by viewing histograms of the data distribution for each independent variable. For
right-skewed data, I used a Box Cox transformation (MASS package in R) to normalize the data.
The RUF is based on multiple linear regression (Legendre and Legendre, 1998) and
produces maximum likelihood estimates for regression coefficients (Marzluff et al. 2004).
Resource values at each cell within a breeding territory were regressed on the corresponding
value of the UD at that cell. The sign associated with the regression coefficient for land cover

61

Table 7. Twenty-four original C-CAP land cover classes (30 m resolution) were reclassified into
more general classes to represent resource attributes within Prairie Warbler breeding territories
along the west coast of Michigan, 2004 and 2005.
Land Cover
Type

Definition

Developed

The majority of “developed” land cover in selected
survey sites is roads (often unpaved), parking lots
and an occasional house or building. Some managed
grasses or low-lying vegetation is planted in
developed areas for recreation.

Sand

Primarily composed of sand dunes. Generally,
vegetation accounts for less than 10% of total cover.
Areas dominated by grammanoid or herbaceous
vegetation, generally greater than 80% of total
vegetation. Although unlikely to occur, “cultivated
crops” and “pasture/hay” were included in the event
back-dune grassland was misclassified.
Areas dominated by shrubs less than 5 m tall with
shrub canopy typically greater than 20% of total
vegetation. Includes tree shrubs, young trees in an
early successional stage, or trees stunted from
environmental conditions (12). Includes wetlands
dominated by woody vegetation less than 5 m in
height. Total vegetation coverage is greater than
20%. Species present could be true shrubs, young
trees and shrubs, or trees that are small or stunted
due to environmental condition.
Areas dominated by trees generally greater than 5 m
tall and greater than 20% of total vegetation cover.
Includes wetlands dominated by woody vegetation
greater than or equal to 5 m in height where total
vegetation coverage is greater than 20%.

Grass

Shrub

Forest

Water
Wetland

Areas of open water, generally with less than 25%
over of vegetation or soil (e.g., Lake Michigan).
Wetlands dominated by persistent emergent
vascular plants, emergent mosses or lichens. Total
vegetation cover is greater than 80%. Plants
generally remain standing until the next growing
season.
62

C-CAP classes
merged (class
number)
High- (2), Medium(3),and LowIntensity
development(4).
Developed Open
Space (5)
Bare Land (20)
Grassland/Herbaceo
us (8), Cultivated
Crops (6),
Pasture/hay(7)
Scrub/Shrub (12),
Paulstrine
Scrub/Shrub
wetland (14)

Deciduous (9),
Evergreen (10),
Mixed Forest (11),
Palustrine
Forested
Wetland(13)
Open Water (21)
Palustrine Emergent
Wetland (15)

types in this analysis indicates whether use increases (+) or decreases (-) with an increase in the
quantity of the resource. For TWI, the sign designates increase in use with wetter (+) or drier (-)
soil. Significance was estimated by noting whether the 95% confidence interval for the
coefficient (mean ± SE*1.96) overlapped zero.
Mean standardized RUFs were assessed by comparing the regression coefficients in 3
ways: 1) across all 38 breeding territories (pooled), 2) for southern and northern sites exclusively
to examine the potential for a north-to-south gradient in resource use (i.e., compare Sleeping
Bear Dunes (n = 20) to the southern sites (n = 18)), and 3) individually for each of the 5 sites
where Prairie Warblers were detected to assess more localized differences in site use.
Comparison of standardized coefficients quantified the relative importance of each resource to
space use in the breeding territory.
Relative availability of each covariate was compared across all territories (n = 38), for
Sleeping Bear Dunes only (n = 20) and for the southern sites (n = 18). Mean TWI values for an
area are on a possible scale from 0 to 224 (high = wet, low = dry). A t-test with Welch (1938)
correction for unequal sample sizes was used to test for significant differences between mean
percent land cover type for northern and southern territories. The Welch correction involves
using degrees of freedom that are corrected for the sample size discrepancy (Zar 1996).

RESULTS
Thirty-eight Prairie Warblers were detected in breeding territories in 5 of the 9 sites
surveyed (Table 5). The average number of mapping visits per territory was 4.35. The average
number of GPS registrations per territory was 38 (range 10 – 92 (SE = 3.4) across all sites; range

63

19 – 81 (SE = 3.7) in Sleeping Bear Dunes; range 10 – 81 (SE = 5.8) in the southern sites).
Breeding areas tended to be clustered within sites with breeding territory boundaries often
adjacent to each other (e.g., Figure 13).
Occupancy Analysis
Percent cover of grass and shrubs were correlated (Pearson’s correlation, r = 0.817). I
removed the grass land cover variable because Prairie Warbler’s are known to positively
associate with shrub cover. Results from the fully parameterized logistic regression model
indicated that distance to forest edge (forest; z = -2.001, df = 116, P = 0.045), distance to shrub
(shrub; z = -2.394, df = 116, P = 0.017), and distance to closest Prairie Warbler occupied
territory (praw; z = -2.294, df=116, P = 0.022) (Table 8) were negatively associated with Prairie
Warbler occupancy along the western Michigan coastline. The negative association indicates
that the probability of Prairie Warbler occupancy increased at shorter distances from the edge of
another Prairie Warbler territory, to forest, and to shrub. My results confirm the hypothesis that
Prairie Warbler breeding territories are associated with early successional woody land cover in
the dune ecosystem. Additionally, my results indicate that Prairie Warblers establish breeding
territories in close proximity to one another. The full model had an approximation of fit of 0.622
(1- residual deviance/null deviance), suggesting that 62% of the observed variability in the data
was accounted for by this model.
I constructed a less parameterized logistic regression model using the 3 significant
variables from the fully parameterized model (forest, shrub, distance to other prairie warbler
territory). The reduced model indicated that distance to shrub (z = -1.975, df = 119, P = 0.048)
and distance to closest Prairie Warbler territory (z = -2.308, df = 119, P = 0.021) were more

64

Fig. 13. Example of Prairie Warbler breeding territories calculated using a fixed kernel
estimation with least squares cross validation. Eight breeding territories were mapped in Warren
Dunes State Park, Berrien County, Michigan (southwest Michigan) in 2005.

65

Table 8. Logistic regression results for land cover variables and Prairie Warbler presenceabsence data. The full model contained 6 independent variables. The reduced model was
calculated using significant variables from the full model. Coefficient significance was tested
using Wald’s chi-square test (z scores). Approximation of fit was calculated as (1- residual
2
deviance/null deviance) indicating the percent of variability explained by each model. X notes
whether the 2 models were significantly different (full residual deviance – reduced residual
deviance, 3df).

(Intercept)

Full Model
Estimate
SE
1.788
0.749

forest

-0.003

0.001

shrub
sand

-0.117
0.020

0.049
0.017

lake

-0.001

0.001

praw
-0.002
developed
0.001
Null deviance:
Residual deviance:
Approximation of fit:
2
X = 0. 259

0.001
0.001
155.16
58.6
0.622

Reduced Model
SE
z
0.516
4.044

z
P
Estimate
2.387 0.017
2.086
2.001 0.045
-0.002 0.001
2.394 0.017
-0.086 0.043
1.211 0.226
0.781 0.435
2.294 0.022
-0.002 0.001
1.418 0.156
on 122 df
Null deviance:
on 116 df
Resid deviance :
Approx. of fit:

*PRAW is distance to nearest Prairie Warbler occupied area

66

P
0

-1.783

0.075

-1.975

0.048

-2.308

0.021

155.16 on 122 df
62.62 on 119 df
0.596

important than forest cover to Prairie Warbler occupancy (Table 8). The approximation of fit for
the reduced model was 0.596 (1- residual deviance/null deviance), suggesting that nearly 60% of
the observed variation in the data was described using this model. Residual deviance for the
reduced model was 62.62 on 119 degrees of freedom, which was larger than that of the full
model (58.5, 116 df.), indicating that the full model performs better. However, results from
these 2 models are similar; chi-square analysis showed that the fully parameterized model did not
significantly differ from the less parameterized model (P = 0.259, df = 3; Table 8).
I plotted the relationships between Prairie Warbler occupancy probability and distance to
the 2 significant predictors from the reduced model: shrub and distance to nearest Prairie
Warbler territory (Figure 14). I found 0% probability of Prairie Warbler presence when shrubs
were >103 m away from the edge of an existing Prairie Warbler territory. The probability of
Prairie Warbler occupancy was 39% when shrubs were within 1 m (Figure 14). The model also
predicts that Prairie Warblers will occupy sites that are <75 m away from shrubs, but probability
of occupancy at these farther distances is low (1%). The farthest I measured shrubs from the
edge of a Prairie Warbler territory in the field was 30 m.
Distance to the closest Prairie Warbler territory from the edge of another Prairie Warbler
territory was the most significant predictor of Prairie Warbler presence (Table 8). I observed a
0% probability of Prairie Warbler presence when distance to another Prairie Warbler was beyond
1,284 m. However, I found an 80% probability of Prairie Warbler presence when another Prairie
Warbler was within 91 m (Figure 14). These results indicate that Prairie Warbler breeding
territories are clumped in the dune ecosystem, and because shrubs are relatively ubiquitous, I
speculate that there is a conspecific attraction for territory establishment.

67

A.
Fig. 14. Probability of Prairie Warbler presence as a function of distance to (A) shrub and to (B)
closest Prairie Warbler breeding territory; the two significant variables in the reduced logistic
regression model.

68

B.
Fig. 14. (cont’d).

69

Resource Utilization Functions
None of the covariates I evaluated to describe within-territory Prairie Warbler space use
were significant (i.e., Wald Test, P>0.05; Table 9). Similarly, RUFs for the north, south, and
individual sites were not significant, although TWI may be informative (i.e., confidence interval
does not overlap zero) on Sleeping Bear Dunes (Table 9). On the northern Prairie Warbler
territories, resource use was positively correlated with moister sites (i.e., higher TWI values)
(Table 9). These results suggest that habitat use within a Prairie Warbler territory is based on
finer scale vegetation pattern than can be detected via C-CAP (30 m).
Northern and southern Prairie Warbler territories differed in cover type composition.
Mean forest cover for Sleeping Bear Dunes (43%) was higher than that observed for southern
territories (29%, Table 6; Welch’s t = 2.28, df = 33, p = 0.029). Mean grass cover also differed
between Sleeping Bear Dunes (32%) and southern territories (15%, Table 10; Welch’s t = 2.83,
df = 30, p = 0.008). Southern territories had greater mean percent sand than northern territories
(65% vs. 40%; Table 10; Welch’s t = -4.37, df = 35, p = 0.0001). I attribute these differences to
higher TWI (noted in RUF output) that were observed in Sleeping Bear Dunes in that moister
soils will likely support greater vegetation biomass.
Examination of the relationship among territory location and the amount of grass, forest
and sand cover suggests a that northern Prairie Warbler territories contain more forest (Figure
15, solid circles) when grass cover is high, while southern Prairie Warbler territories contain
more sand cover (Figure 15, hollow triangles) when grass cover is high. Thus, my results
indicate that factors associated with space use by Prairie Warblers within a breeding territory
vary by latitude, with more vegetated territories in the north.

70

Table 9. Estimates of standardized RUF coefficients (β) for Prairie Warblers across all territories
(n = 38), Sleeping Bear Dunes only (n = 20) and southern sites (n = 18). Relative importance of
resources is indicated by magnitude of β.
Resource Attribute

All Territories
(n=38)
TWI
Grass
Developed
Shrub
Water
Forest
Wetland
Sand
Sleeping Bear
Dunes (n=20)
TWI
Grass
Developed
Shrub
Water
Forest
Wetland
Sand
Southern sites
(n=18)
TWI
Grass
Developed
Shrub
Water
Forest
Wetland
Sand

Mean
standardized
RUF
coefficients
(β)

95% confidence
interval

+
0.0164
0.0595
0.0015
0.0051
0.0077
0.0018
-0.0084
-0.0123

0.0425
0.1501
0.0298
0.0444
0.0357
0.0339
0.0155
0.0146
+

0.0162
0.0157
-0.0171
0.0256
-0.0122
0.0220
-0.0051
-0.0361

0.0321
0.0319
-0.0013
0.0960
0.0010
0.0468
0.0024
0.0022
+

0.0167
0.1082
0.0222
-0.0178
0.0299
-0.0206
-0.0121
0.0140

0.0696
0.2990
0.0787
0.0093
0.0862
0.0406
0.0385
0.0489

P

(β=0)*

–
-0.0096
-0.0311
-0.0268
-0.0342
-0.0202
-0.0302
-0.0323
-0.0393

0.77
0.99
0.53
0.59
0.64
0.53
0.35
0.29

–
0.0003
-0.0005
-0.0330
-0.0447
-0.0254
-0.0027
-0.0126
-0.0744

0.53
0.53
0.47
0.55
0.48
0.54
0.49
0.44

–
-0.0362
-0.0826
-0.0343
-0.0448
-0.0264
-0.0817
-0.0627
-0.0208

0.66
0.99
0.71
0.33
0.76
0.31
0.39
0.63

*P-values test the null hypothesis that the standardized RUF coefficient (β) is zero, given sample
sizes of n = 38, n = 20 and n = 18.
71

Table 10. Mean percent of each covariate by region. Relative availability of each covariate are
compared across all territories (n = 38), for Sleeping Bear Dunes only (n = 20) and for the
southern sites (n = 18). Mean TWI values indicate level of wetness for an area on a possible
scale from 0 to 224 (high = wet, low = dry).

TWI
All territories
Min
15.91
Max 107.90
Mean
50.81
Sleeping Bear
Min
14.68
Max
91.91
Mean
40.29
Southern sites
Min
17.29
Max 125.68
Mean
62.49
Warren Dunes
Min
11.64
Max 116.26
Mean
46.41
Saugatuck
Min
18.68
Max 136.68
Mean
65.50
Oval Beach
Min
24.01
Max 122.01
Mean
99.07
Kitchel-Lindquist
Min
29.01
Max 134.01
Mean
81.25

SAND

FOREST

GRASS

SHRUB DEVLP*

WETLND

WATER

1.17
99.43
51.92

0.01
96.83
36.12

0.01
80.75
23.59

0.01
44.75
3.35

0.01
18.22
2.13

0.01
12.22
0.81

0.01
13.14
0.67

0.01
98.91
40.17

0.01
96.71
42.88

0.01
87.81
31.54

0.01
47.31
3.38

0.01
27.36
3.08

0.01
9.91
0.88

0.01
17.76
0.42

2.45
100.01
64.97

0.01
96.95
28.60

0.01
72.90
14.76

0.01
41.90
3.31

0.01
8.07
1.08

0.01
14.79
0.73

0.01
8.01
0.95

5.51
100.01
65.54

0.01
94.51
30.72

0.01
79.26
20.53

0.01
30.51
2.38

0.01
0.01
0.01

0.01
5.51
0.03

0.01
12.51
2.06

0.01
100.01
60.53

0.01
98.18
31.98

0.01
64.84
11.18

0.01
53.68
4.75

0.01
0.01
0.01

0.01
11.01
0.76

0.01
3.68
0.06

0.01
100.01
73.35

0.01
100.01
15.79

0.01
61.01
7.61

0.01
61.01
4.38

0.01
39.01
5.29

0.01
16.51
0.90

0.01
11.01
0.16

0.01
100.01
67.65

0.01
100.01
22.80

0.01
83.51
9.52

0.01
33.01
1.66

0.01
33.51
4.36

0.01
61.51
3.28

0.01
0.01
0.01

*Developed land
** Southern sites include Warren Dunes State Park, Saugatuck Dunes State Park, Oval Beach
and Kitchel –Lindquist Dunes Preserve.

72

High
northern
grass use

High
southern
grass use

Fig. 15. Trend suggesting high northern forest use (solid circles) and high southern sand use
(hollow triangles) by Prairie Warblers along Michigan’s west coast. Prairie Warblers tended to
use either forest or sandy areas of their breeding territories. Northern Prairie Warblers tended to
concentrate use in grass when forest use was also high. Southern Prairie Warblers tended to
concentrate use in grass when sand use was also high.

73

DISCUSSION
This research provides new information about Prairie Warbler resource utilization based
on analysis of 38 west Michigan coastal breeding territories. The best indicators of Prairie
Warbler occupancy were proximity to other Prairie Warblers and proximity to shrub cover type.
I found an 80% probability of Prairie Warbler occupancy when another Prairie Warbler territory
was within 91 m. I found a 39% probability of Prairie Warbler presence when shrubs were
within 1 m of an existing Prairie Warbler territory. I found no significant environmental
covariates from standardized RUFs, however, 4 observations warrant attention: 1) northern sites
have more woody vegetation cover than southern sites, and southern sites seem to be more open,
2) grass cover appears to be important on southern sites, potentially because shrubs are less
ubiquitous, 3) higher soil moisture in northern territories was suggested as an important indicator
of habitat use, and 4) lack of significant (standardized) RUF coefficients suggests that the spatial
resolution of land cover data (30 m C-CAP) may not have been appropriate to describe within
territory space use for Prairie Warblers.
When determining which land cover types to use to assess Prairie Warbler resource use, I
drew from Michigan casual observation records and species summaries (Brewer et al. 1991,
Cooper 2000, Michigan Natural Features Inventory 2011), research in other breeding locations
(Nolan 1978, Nolan et al. 1999, Cooper 2000, Southwell 2001) and my survey experience.
Prairie Warblers require a specific combination of land cover features. Prairie Warblers glean
insects from grass, herbaceous plants, shrubs and trees (Nolan 1978, Rappole et al. 1983) and
they defend territories from and nest in shrubs and trees 1-3 m in height (Nolan 1978,
Walkinshaw 1991, Dunn and Garrett 1997, Cooper 2000). Generally, my results support the
previously observed relationships between Prairie Warbler occurrence and a mixture of

74

herbaceous and low-growing woody vegetation. Additionally, my study was the first to
document a latitudinal gradient in Prairie Warbler habitat use that was related to vegetation
composition; with northern territory space use correlated with denser, woody vegetation and
southern correlated with more open, grass dominated sites.
I found that proximity to another Prairie Warbler (within 91 m) was the most important
predictor of occupancy in dune landscapes along the west coast of Michigan. Research suggests
that conspecific attraction influences avian habitat selection and occupancy more than previously
considered or reported (Danchin et al. 2004, Hahn and Silverman 2006, Betts et al. 2008).
Studies on Kittiwakes (Rissa tridactyla) (Danchin et al. 1998) and Collared Flycatchers
(Ficedula albicollis) (Doligez et al. 2002) have both shown that individuals will replicate habitat
selection choices of conspecifics that experience high breeding success, using social cues as their
primary decision-making strategy. Hahn and Silverman (2006) found that conspecific attraction
was an important mechanism for breeding habitat selection for established populations of
Redstarts (Setophaa ruticilla) in Michigan, but the effect was moderated by age, reproductive
experience and timing of arrival. Harrison et al. (2009) warn that traditional habitat models,
which consider only resource distributions and not social factors, may be inadequate tools for the
conservation of Brewers Sparrows (Spizella breweri breweri) at the northern edge of their range.
Lack of significant RUF findings could imply little consistency in resource selection at
the territory level, but more likely suggests that the data chosen to portray habitat conditions
within a territory were likely not suited for this analysis. The 30 m C-CAP data was likely too
coarse for describing Prairie Warbler habitat use within territories. Grass and sand land cover
classes (as portrayed by 2006 C-CAP imagery) likely contained shrubs and saplings that would
go undetected with 30 m resolution data. Finer-resolution land cover data may provide more

75

useful insight into Prairie Warbler space use within a breeding territory. Key factors involved in
habitat selection may not be detected without using the appropriate spatial scales of analysis
(Orians and Wittenberg 1991). Using the grain at which organisms perceive resources is
difficult to know but is certainly related to their sensory abilities (Vos et al. 2001) and thus
should influence the scale at which resources are measured (i.e., the spatial resolution of land
cover data). Marzluff et al. (2004) states that it would be prudent to select land cover data at a
scale fine enough to capture important resource variation, even if one thinks this may not
perceptible to the study organism.

MANAGEMENT RECOMMENDATIONS
The west Michigan dune landscape is providing a source of early successional habitat for
Prairie Warblers that is maintained through wind and wave dynamics. Habitat fragmentation of
dune ecosystems due to development or land conversion limits the amount of available habitat.
Prairie Warblers occurred on 5 of the 9 dune ecosystems I surveyed along the western Michigan
coastline and their 50% territory estimates ranged from an average radius of 19 - 56m, providing
guidance on minimum area requirements for habitat management and restoration. The full suite
of environmental factors useful for describing habitat suitability for Michigan coastal Prairie
Warbler populations are unknown, however my results suggest that shrub cover should be
abundant and well dispersed (<75 m separation) and that Prairie Warblers will likely occupy
landscapes in a clumped pattern. Hence, Prairie Warbler habitat management plans should
emphasize larger areas (i.e., multiple territories) of contiguous habitat with well dispersed
shrubs.

76

New surveys for Prairie Warblers should be located in areas where Prairie Warblers have
most recently been detected (Nolan 1978, Nolan et al., 1999). Prairie Warblers exhibit site
fidelity, so offspring return to the same places year after year if breeding was successful, until
habitat succeeds (Nolan 1978). The return rate to previous year’s territory was 65% based on
Nolan’s 20-year Indiana study (Nolan 1978).

77

LITERATURE CITED

78

LITERATURE CITED

Aebischer, N. J., P. A. Robertson, and R. E. Kenward. 1993. Compositional analysis of habitat
use from animal radio-tracking data. Ecology 74:1313-1325.
Albert, D. A. 2000. Born of the wind: an introduction to the ecology of Michigan sand dunes.
Michigan Natural Features Inventory, Lansing, Michigan, USA.
Betts, M. G., A. S. Hadley, N. Rodenhouse, and J. J. Nocera. 2008. Social information trumps
vegetation structure in breeding-site selection by a migrant songbird. Proceedings of the
Royal Society of Biological Sciences 275:2257-2263.
Beven, K. J. and M. J. Kirkby. 1979. A physical-based variable contributing area model of
basin hydrology. Hydrology Science Bulletin 24:43-69.
Beyer, H. L. 2004. Hawth's analysis tools for ArcGIS. www.spatialecology.com /htools.
Accessed August 2011.
Brewer, R., G. A. McPeek, and R. J. Adams Jr. 1991. Atlas of breeding birds in Michigan.
Michigan State University Press, East Lansing, Michigan, USA.
Cooper, J. L. 2000. Special animal abstract for Dendroica discolor (Prairie Warbler). Michigan
Natural Features Inventory, Lansing, Michigan, USA.
Danchin, E., T. Boulinier, and M. Massot. 1998. Conspecific reproductive success and breeding
habitat selection: implications for the study of coloniality. Ecology 79:2415-2428.
Danchin, E., L. A. Giraldeau, T. J. Valone, and R. H. Wagner. 2004. Public information: from
nosy neighbors to cultural evolution. Science 305:487-490.
Doligez, B., E. Danchin, and J. Clobert. 2002. Public information and breeding habitat selection
in a wild bird population. Science 297:1168-1170.
Dunn, J. L., and K. L. Garrett. 1997. A field guide to warblers of North America. Houghton
Mifflin Co., Boston, Massachusetts, USA.
ESRI. 2010. Topographic wetness index. ArcScript download. www.arcscripts.
esri.com/details.asp?dbid=16750. Accessed 2011.
Gitzen, R. A., J. J. Millspaugh, and B. J. Kernohan. 2006. Bandwidth selection for fixed-kernel
analysis of animal utilization distributions. The Journal of Wildlife Management
70(5):1334-1344.

79

Hahn, B. A. and E. D. Silverman. 2006. Social cues facilitate habitat selection: American
redstarts establish breeding territories in response to song. Biology Letters 2:337-340.
Handcock, M. S. 2004. RUF: An R package for the estimation of the resource
utilization function. http://csde.washington.edu/~handcock/ruf. Accessed 2011.
Handcock, M. S. and M. L. Stein. 1993. A Bayesian analysis of kriging. Technometrics
35(4):403-410.
Harrison, M. L., D. J. Green, and P. G. Krannitz. 2009. Conspecifics influence the settlement
decisions of male Brewer’s sparrows at the northern edge of their range. The Condor
111(4):722-729.
Jenness, J. 2003. Random point generator (randpts.avx) extension for ArcView 3.x, Version
1.27. Jenness Enterprises. www.jennessent.com/arcview/ random_ points.htm. Accessed
2004 - 2005.
Kernohan , B. J., R. A. Gitzen, and J. J. Millspaugh. 2001. Analysis of animal space use
and movements. Pages 126-166 in J. J. Millspaugh and J.M. Marzluff, editors.
Radio tracking and animal populations. Academic Press, San Diego, California,
USA.
Legendre, P., and L. Legendre. 1998. Numerical Ecology. Second English edition. Elsevier,
Amsterdam, The Netherlands.
Levine, J. M., M. Vila, C. M. D’Antonio, J. S. Dukes, K. Grigulis, and S. Lavorel. 2003.
Mechanisms underlying the impacts of exotic plant invasions. Proceedings of the Royal
Society of London 270:775-781.
Manly, B. F. J., L. L. McDonald, and D. L. Thomas. 2002. Resource selection by
animals: statistical design and analysis for field studies. Revised edition.
Kluwer Academic, Boston, Massachusetts. USA.
Marzluff, J. M. 1986. Assumptions and design of regression experiments. Pages 165-170 in J.
Verner, M. L. Morrison, and C. J. Ralph, editors. Wildlife 2000: modeling habitat
relationships of terrestrial vertebrates. University of Wisconsin Press, Madison,
Wisconsin, USA.
Marzluff, J. M., S. T. Knick, M. S. Vekasy, L. S. Schueck, and T. J. Zarriello. 1997. Spatial use
patterns and habitat selection of golden eagles in southwestern Idaho. Auk 114:673-687.
Marzluff, J. M., J. J. Millspaugh, P. Hurvitz, and M. S. Handcock. 2004. Relating resources to a
probabilistic measure of space use: forest fragments and Steller’s jays. Ecology 85:14111427.
Mayfield, H. 1960. The Kirtland’s warbler. Cranbrook Institute of Science Bulletin 40,
80

Bloomfield Hills, Michigan, USA.
McCullaugh, P. and J. A. Nelder. 1983. Generalized linear models. Chapman and Hall,
London, England, UK.
Mehlman, D. W., S. E. Mabey, D. N. Ewert, C. Duncan, B. Abel, D. Cimprich, R. D. Sutter, and
M. Woodrey. 2005. Conserving stopover sites for forest-dwelling migratory landbirds.
Auk 122(4):1281-1290.
Mengel, R. M. 1965. The birds of Kentucky. American Ornithologists Union Monograph No.
3.
Michigan Department of Natural Resources. 2001. Integrated forest monitoring, assessment,
and prescription (IFMAP) land classification system. www.michigan.gov/dnr/0,1607,7153-30301_39170-144990--,00.html. Accessed 2004-2008.
Michigan Natural Features Inventory. 2011. Rare species explorer. http://web4.msue.msu.edu
/mnfi/explorer. Accessed 2004-2011.
Michigan Sea Grant. 2012. Explore Great Lakes coastal habitats. www.miseagrant.umich.edu/
explore/coastal-habitat/habitat-sand-dunes.html. Accessed March 2012.
Millspaugh, J. J., G. C., Brundige, R. A., Gitzen, and K. J., Raedeke. 2000. Elk and hunter
space-use sharing in South Dakota. Journal of Wildlife Management 64:994-1003.
Millspaugh, J. J., R. M. Nielson, L. McDonald, J. M. Marzluff, R. A. Gitzen, C. D. Rittenhouse,
M. W. Hubbard, S. L. Sheriff. 2006. Analysis of resource selection using utilization
distributions. Journal of Wildlife Management 70(2):384-395
National Oceanic and Atmospheric Administration, Coastal Services Center. 2008. NOAA
Coastal Change Analysis Program (C-CAP) 2006 Era Land Cover Data of the C-CAP
zone 51 (upper midwest). www.csc.noaa.gov/crs/lca. Charleston, South Carolina, USA.
Nolan, V., Jr. 1978. The ecology and behavior of the prairie warbler Dendroica discolor.
Ornithological Monographs No. 26.
Nolan, Jr., V., E. D. Ketterson, and C. A. Buerkle. 1999. The birds of North America:
prairie warbler (Dendroica discolor). Pages 1-26 in A. Poole and F. Gill, editors. The
birds of North America, No. 584. Philadelphia, Pennsylvania, USA.
Orians, G. H. and J. F. Wittenberger. 1991. Spatial and temporal scales in habitat selection.
The American Naturalist 137:S29-S49.
R Development Core Team. 2011. R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria. www.R-project.org. Accessed
2011.
81

Ralph, C. J., S. Droege, and J. Sauer. 1995. Managing and monitoring birds using point counts:
standards and applications. U.S. Forest Service General Technical Report. PSW-GTR149:161-169.
Rappole, J., H., E. S. Morton, T. E. Lovejoy III, and J. L. Ruos. 1983. Neartic avian migrants in
the Neotropics. U.S. Department of Interior, Fish and Wildlife Service, Washington,
D.C., USA.
Samuel, M. D., D. J. Pierce, and E. O. Garton. 1985. Identifying areas of concentrated
use within the home range. Journal of Animal Ecology 54:711-719.
Samuel, M., D., and E. O. Garton. 1987. Home range: a weighted normal estimate and
tests of underlying assumptions. Journal of Wildlife Management 49:513-519.
Seaman, D. E., and R. A. Powell. 1996. An evaluation of the accuracy of kernel density
estimators for home range analysis. Ecology 77:2075-2085.
Seidel, K. S. 1992. Statistical properties and applications of a new measure of joint
space use for wildlife. M.S. thesis, University of Washington, Seattle, Washington, USA.
Silverman, B. W. 1986. Density estimation for statistics and data analysis. Chapman and
Hall, London, England, UK. www.ipac.caltech.edu/level5/March02/Silverman/Silver
_contents. Accessed October 2008.
Space Imaging. 2004. Review of remote sensing technologies used in the IFMAP project. Final
report. Ann Arbor, Michigan, USA.
Southwell, D. K. 2001. Conservation assessment for Prairie Warbler (Dendroica discolor).
USDA Forest Service Eastern Region. USDA Forest Service, Hiawatha National Forest,
Michigan, USA.
United States Bureau of the Budget. 1947. United States national map accuracy standards:
U.S. Bureau of the Budget, Washington, D.C.
United States Department of Agriculture. 2005. National Aerial Imagery Program
1METER 2005 MI: 2005 National Agricultural Imagery Program Aerial Photographs,
Michigan (Indiana Geological Survey, SDE Raster Catalog). U.S. Department of
Agriculture, Farm Service Agency, Aerial Photography Field Office, Bloomington,
Indiana, USA. www.fsa.usda.gov/FSA/apfoapp?area=home&subject=prog&topic=nai.
Accessed 2008.
van Winkle, W. 1975. Comparison of several probabilistic home-range models. Journal
of Wildlife Management 39:118-123.
Vos, C. C., J. Verboom, P. F. M. Opdam, and C. J. F. Ter Braak. 2001. Toward ecologically
82

scaled landscape indices. The American Naturalist 157(1):24-41.
Walkinshaw, L. H. 1991. Prairie warbler. Pages 418-419 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.
Welch, B. L. 1938. The significance of the difference between two means when the population
variances are unequal. Biometrika 29:350-361.
Zar, J. H. 1996. Biostatistical analysis. Third edition. Prentice-Hall, Upper Saddle River,
New Jersey, USA.

83

CHAPTER 4

SYNTHESIS OF FINDINGS AND MANAGEMENT RECOMMENDATIONS FOR
MICHIGAN PRAIRIE WARBLER POPULATIONS

INTRODUCTION
This chapter is written for natural resource managers and conservation practitioners. It
summarizes background information for Prairie Warbler, the need for this study, and synthesizes
research findings as context for conservation and management recommendations. Readers
should consult referenced dissertation chapters, figures and tables for further detail on research
methods and findings as indicated.

BACKGROUND AND RESEARCH NEED
The status of Michigan Prairie Warbler (Dendroica discolor) breeding locations,
population abundance, and habitat composition influencing breeding territory selection are
critical management and conservation needs for this state-endangered, neotropical migratory
songbird (Walkinshaw 1991, Nolan et al. 1999, Cooper 2000, Southwell 2001).
Though globally secure (NatureServe 2011), populations of Prairie Warbler have
declined throughout the breeding range (Robbins et al. 1986, Saurer et al. 2011). Midwest
populations are of moderate to high management concern (Robinson et al. 1999). This trend
holds true for Michigan populations at the northern edge of the breeding range (Michigan
Natural Features Inventory 2011), though Michigan populations most likely always have been
small and localized (Nolan 1978).
Michigan’s earliest bird list (Sager 1839) reported Prairie Warbler somewhere between
Detroit, Saginaw Bay, and Jackson. Prairie Warblers were recorded nesting in the southern
84

Lower Peninsula counties of Ottawa and Montcalm in the early 1900’s (Barrows 1912). Wood
(1951) noted Berrien County as important to Prairie Warbler, as did Zimmerman and VanTyne
(1959) in addition to reporting nesting records from the west Michigan counties of Ottawa,
Muskegon, and Oceana. Wood (1951), Zimmerman and Van Tyne (1959), and Mayfield (1960)
all reported occurrence in northeastern Lower Peninsula counties of Oscoda, Crawford, Otsego,
Cheboygan, and Presque Isle, primarily associated with jack pine (Pinus banksiana) plains.
Today, the majority of casual observations come from Lake Michigan dune habitat in the
southwest and northwest Lower Peninsula (Michigan Natural Features Inventory 2011). Prairie
Warbler became more rare in the Michigan jack pine plains after the 1960’s (Nolan 1978,
Walkinshaw 1991), was listed as state threatened in 1992 (Southwell 2001), and as state
endangered in 1999 (Natural Resources and Environmental Protection Act 1994).
Prairie Warblers glean insects from leaves and branches of shrubs and trees and
occasionally take prey mid-air (Nolan 1978, Rappole et al. 1983). Nestlings are primarily fed
caterpillars (Nolan 1978). Mean breeding season feeding height in Florida for breeding males is
2.19 m; for females 1.40 m (Prather 1994). They build open cup nests in shrubs and trees 1-3 m
in height (Nolan 1978, Walkinshaw 1991, Dunn and Garrett 1997, Cooper 2000,) and defend
territories from the highest perches (shrubs or trees) in the area along territory boundaries (Nolan
1978). Males arrive on Michigan breeding grounds from the West Indies during the first two
weeks of May with the female arriving about one week later (Wood 1951, Nolan 1978). Prairie
Warblers typically depart Michigan breeding grounds in late August, with extreme early and late
departure records of August 10 and September 23 respectively (Wood 1951, Nolan 1978).
Throughout their breeding range Prairie Warbler habitat preferences have generally been
characterized as open, early-successional areas with scrub-shrub plant species found in poor soils

85

and little or no tree canopy (Nolan 1978, Walkinshaw 1991, Nolan et al. 1999, Cooper 2000).
Walkinshaw (1991) noted three distinct shrubby habitats in which Michigan Prairie Warblers had
been recorded in recent decades: 1) early successional stages of Great Lakes dunes where
deciduous or coniferous bushes are intermixed with dune grass and other herbs, 2) earlysuccessional, shrubby openings associated with previously-burned jack pine plains, and 3)
previously burned conifer stands dominated by deciduous shrubs and small trees.
Gaining an understanding of where west Michigan Prairie Warbler breeding populations
are located (as the majority of occurrence records are reported from this region of the state;
Michigan Natural Features Inventory 2011), and in what abundance, will allow a baseline from
which to monitor population status. Describing habitat composition within breeding territories
will help identify and specify Prairie Warbler habitat needs, thus informing conservation
practitioners as they design long term management plans and are asked to respond to proposed
land use changes (e.g., mining, wind tower siting, development).
Some have speculated that abundant Prairie Warbler habitat exists (Walkinshaw 1991,
Cooper 2000, Southwell 2001), but habitat requirements for Prairie Warbler, and their selection
strategies, may be more specific than previously thought. Examination of environmental factors
(e.g., plant composition, proximity to specific land cover types or other Prairie Warblers) within
breeding territories, compared to areas where Prairie Warbler are absent, will allow conservation
practitioners to manage with a greater understanding of Prairie Warbler habitat selection
strategies and thus how to maintain habitat for this state-endangered species in concert with other
conservation objectives.

86

I chose to examine nine study sites located along the Lake Michigan coast throughout the
Lower Peninsula of Michigan in 2004 and 2005 (Figure 4) to address the following needs for
Michigan Prairie Warbler populations:
Breeding location, distributions and relative abundance along Michigan’s west coast,
(Nolan 1978, Walkinshaw 1991, Felsing 1994, Cooper 2000, Southwell 2001),
Breeding season plant composition associated with breeding territory selection and
differences that exist in plant composition between territories and in unoccupied areas
(Walkinshaw 1991, Nolan et al. 1999, Southwell 2001),
An assessment of land cover predictors for Prairie Warbler occupancy (Nolan et al. 1999,
Southwell 2001), and
Insights into spatial resolution necessary to detect Prairie Warbler habitat needs.

SYNTHESIS OF FINDINGS
Population Distribution, Relative Abundance and Breeding Territory Sizes
My study sites were chosen because they were publically accessible and contained
potential Prairie Warbler habitat (identified by cover type (i.e., jack pine, sand dunes, scrubshrub; Nolan 1978, Nolan et al. 1999, Cooper 2000, Southwell 2001)). The nine study sites
included 6 State Parks, Sleeping Bear Dunes National Lakeshore, Oval Beach (a public beach
owned by the city of Saugatuck), and Kitchel-Lindquist Dunes Preserve (publically-accessible
preserve near Grand Haven) (Figure 4).
The greatest number of Prairie Warbler occupied areas was found in Sleeping Bear
Dunes (n = 20). Eighteen occupied areas were collectively found in southern sites, with Warren
Dunes having the most (n = 8), followed by Saugatuck Dunes State Park (n = 6) and Kitchel-

87

Lindquist Dunes Preserve and Oval Beach (n = 2). Ludington State Park/Nordhouse Dunes,
Silver Lake State Park, Muskegon State Park, and Hoffmaster State Park each had zero Prairie
Warblers detected (Table 1). Though I did not monitor nests, these occupied areas were assumed
to be breeding territories.
The average number of GPS registrations per breeding territory to estimate territory size
and shape was 38 (range 10 – 92 (SE = 3.4) across all sites; range 19 – 81 (SE = 3.7) in Sleeping
Bear Dunes; range 10 – 81 (SE = 5.8) in the southern sites). Breeding territories tended to be
clustered within sites surveyed, with breeding territory boundaries often adjacent to each other
(e.g., Figure 13; Appendix, Figures 16-22). Prairie Warbler mean 95% territory use percentile,
as calculated from territories (Table 1, Table 11,) ranged from 1.01 ha in Oval Beach (n = 2,
range 0.84 – 1.12; SD 0.245) to 3.170 ha in Sleeping Bear Dunes (n = 20, range 0.16 – 7.93 ha,
SD = 2.281) (Table 11; Appendix, Figures 16-22), providing guidance on minimum area
requirements for Prairie Warbler.
The smallest territory sizes recorded in the literature are: mean 0.47 ha (85 pair/ 40 ha
tract in Maryland; Robbins et al. 1947), 0.24 ha on island in Georgia (Beer et al. 1956). Ranges
of territory size in two adjacent long term study sites in Indiana (Nolan 1978) were: 3.5 and 0.5
ha (n = 111, mean = 1.62, SD = 0.72 ha) on younger successional stage site, and 2.4 and 0.4 ha
(n = 60, mean = 1.47 ha, SD = 0.47 ha) on older successional stage site (more and larger trees).
Territories of young males are often smaller than older males (Nolan et al. 1999). Prairie
Warbler territory sizes in Indiana were noted to be smallest when roughly square and compact,
intermediate when oblong, and largest when very elongated (Nolan 1978, Nolan et al. 1999).
Territories adjoining unsuitable habitat (where defense is thus unnecessary) tend to be larger
(Nolan 1978, Nolan et al. 1999). These size comparisons were roughly consistent with average

88

Table 11. Mean area of Prairie Warbler 95% and 50% territory use percentiles (see Chapter 2 for
methods). SD = standard deviation. Range of territory areas are shown for each southern survey
site, mean across southern territories, mean across Sleeping Bear Dunes territories, and for each
grouping of territories in the northern, central, and southern areas of Sleeping Bear Dunes (see
Appendix for territory maps).
Mean territory area
Southern Sites
(n = 18)
Warren Dunes (n = 8)
Oval Beach (n = 2)
Saugatuck Dunes (n = 6)
Kitchel-Lindquist (n = 2)
Sleeping Bear Dunes
(n = 20)
North (n = 11)
Central (n = 6)
South (n = 3)

95%
(ha)
1.769

SD

range (ha)

SD

range (ha)

0.537 - 3.971

50%
(ha)
0.297

0.910

0.237

0.107 -0.771

2.139
1.012
1.859
1.659
3.170

0.965
0.245
0.936
0.991
2.281

1.368 - 3.971
0.839- 1.185
0.537 - 3.037
0.958 - 2.360
0.158 - 7.931

0.355
0.117
0.365
0.187
0.512

0.251
0.014
0.270
0.111
0.376

0.128 - 0.771
0.107 - 0.128
0.131 - 0.709
0.109 - 0.266
0.032 - 1.286

3.023
3.560
4.925

2.053
1.089
4.175

1.058 - 6.897
1.060 - 3.931
0.158 - 7.931

0.547
0.473
0.535

0.424
0.229
0.535

0.120 - 1.286
0.130 - 0.709
0.032 - 1.098

*southern sites are Warren Dunes and Saugatuck Dunes State Parks, Oval Beach, and KitchelLindquist Dunes Preserve

89

territory sizes recorded for the five sites where Prairie Warblers were found in this coastal, West
Michigan study (Table 1, Table 11). I found Michigan coastal breeding territory size to increase
with increasing size of the survey area available habitat (Appendix, Figures 16-22; Table 1).
West Michigan Coastal Habitat Composition
Generally, my results support the previously observed relationships between Prairie
Warbler occurrence and a mixture of herbaceous and low-growing woody vegetation (Nolan
1978, Walkinshaw 1991, Nolan et al. 1999, Cooper 2000). However, my study was the first to
document a latitudinal gradient in Prairie Warbler habitat use related to vegetation composition;
with northern territory space use correlated with denser, woody vegetation and southern territory
space use associated with more open, grass dominated sites.
Analysis of vegetation survey data (1 m diameter plots; see Chapter 2) showed that plant
composition in areas occupied by Prairie Warblers was different from plant composition in
unoccupied areas (Table 4, Figures 9 and 10). However, those differences were not consistent
from north to south; Prairie Warbler breeding areas in the north had a different group of plants
associated with them than did Prairie Warbler breeding areas in the southern part of the state
(Table 3, Figures 9 and 10).
Prairie Warblers in Sleeping Bear Dunes were likely to be found in areas with woody
cover assemblages including: bearberry (Arctostaphylos uva-ursi), common juniper (Juniperus
communis), creeping juniper (J. horizontalis), and Rubus spp. Redtop grass (Agtostis gigantean),
hawkweed (Hieracium spp.), and poison ivy (Toxicodendron radicans) were also commonly
associated with areas of Prairie Warbler presence (Table 3, Figure 10).
Southern Prairie Warbler breeding territories were dominated by grasses and a mix of
native and exotic herbaceous plants (Table 3); specifically, marram grass (Ammophila

90

breviligulata), sand reed grass (Calamovilfa longifolia), wormwood (Artemisia campestris),
horsetail (Equisetum spp.) and chickweed (Stellaria spp.) (Table 3, Figure 10). Many of these
same plant species were associated with unoccupied areas in Sleeping Bear Dunes.
Plant species associated with southern sites where Prairie Warblers were absent, included
sugar maple (Acer saccharum), Austrian pine (P. nigra), red pine (P. resinosa), jack pine, choke
cherry (Prunus virginiana), white ash (Fraxinum Americana), and garlic mustard (Alliaria
petiolata) (Table 3, Figure 10). This suggests that Prairie Warbler territories in the south were
not associated with tree species and selected areas with more grass and herbaceous plant species.
Southern territory plant composition suggested less dense vegetation overall, more open sand
and more grass than northern Sleeping Bear Dunes territories.
Land cover analysis using 30 m Coastal Change Assessment Program (C-CAP; National
Oceanic and Atmospheric Administration, Coastal Services Center 2008) imagery (Chapter 3)
supported the difference in cover type composition in northern and southern Prairie Warbler
territories. Mean forest cover for Sleeping Bear Dunes territories (43%) was higher than that
observed for southern territories (29%, Table 10; Welch’s t = 2.28, df = 33, p = 0.029). Mean
grass cover also differed between Sleeping Bear Dunes territories (32%) and southern territories
(15%, Table 10; Welch’s t = 2.83, df = 30, p = 0.008). Southern territories had greater mean
percent sand than northern territories (65% vs. 40%; Table 10; Welch’s t = -4.37, df =35, p =
0.0001). I attribute these differences in part, to higher topographic wetness index values that
were observed in Sleeping Bear Dunes (see RUF output Chapter 3) in that moister soils will
likely support greater vegetation biomass.

91

Assessing Occupancy
I found 0% probability of another Prairie Warbler being present when shrubs were >103
m away from the edge of an existing Prairie Warbler territory (Figure 14). The probability of
another Prairie Warbler being present was 39% when shrubs were detected within 1 m of an
existing Prairie Warbler territory (Figure 14). My model also predicted that Prairie Warblers
will occupy sites that are <75 m away from shrubs, but probability of occupancy at these farther
distances was low (1%). The farthest I measured shrubs from the edge of a Prairie Warbler
territory in the field was 30 m.
Distance to the closest Prairie Warbler territory from the edge of another Prairie Warbler
territory was the most significant predictor of Prairie Warbler presence (Table 8). I observed a
0% probability of Prairie Warbler presence when distance to another Prairie Warbler was beyond
1,284 m. However, I found an 80% probability of Prairie Warbler presence when another Prairie
Warbler was within 91 m (Figure 14). These results indicate that Prairie Warbler breeding
territories are clumped in the dune ecosystem, and because shrubs are relatively ubiquitous, I
speculate there is a conspecific attraction (i.e., the tendency for Prairie Warblers to settle near
one another) for territory establishment.
Research suggests conspecific attraction influences avian habitat selection and occupancy
more than previously considered or reported (Danchin et al. 2004, Hahn and Silverman 2006,
Betts et al. 2008). Studies on Kittiwakes (Rissa tridactyla) (Danchin et al. 1998) and Collared
Flycatchers (Ficedula albicollis) (Doligez et al. 2002) have both shown that individuals will
replicate habitat selection choices of conspecifics that experience high breeding success, using
social cues as their primary decision-making strategy. Hahn and Silverman (2006) found
conspecific attraction was an important mechanism for breeding habitat selection for established

92

populations of Redstarts (Setophaa ruticilla) in Michigan, but the effect was moderated by age,
reproductive experience, and timing of arrival. Harrison et al. (2009) warn that traditional
habitat models, which consider only resource distributions and not social factors, may be
inadequate tools for the conservation of Brewers Sparrows (Spizella breweri breweri) at the
northern edge of their range.
Prairie Warblers were commonly found in the more stabilized early-successional,
transition areas between open sand and forest edge or in patches of shrubs within dune blowouts.
As referenced above, plant species associated with Prairie Warbler presence differed from north
to south, but the commonality between territory establishment was availability of woody cover
and most importantly, close proximity to other Prairie Warblers.
The full suite of environmental factors useful for describing habitat suitability for
Michigan coastal Prairie Warbler populations are unknown, however my results suggest that
shrub cover should be abundant and well dispersed (<75 m separation) and that Prairie Warblers
will likely occupy landscapes in a clumped pattern. Hence, Prairie Warbler habitat management
plans should emphasize larger areas of contiguous habitat with well dispersed shrubs to
accommodate multiple territories.
Resource Utilization and Spatial Resolution Considerations
Resource Utilization Functions (RUF) indicate the importance of each land cover
covariate (e.g., land cover type, soil moisture) within the breeding territories (Marzluff et al.
2004). I found no significant environmental covariates from RUFs, however, 4 observations
warrant attention: 1) northern sites have more woody vegetation cover than southern sites, and
southern sites seem to be more open, 2) grass cover appears to be important on southern sites,
potentially because shrubs are less ubiquitous, 3) higher soil moisture in northern territories was

93

suggested as an important indicator of habitat use, and 4) lack of significant RUF coefficients
suggests that the spatial resolution of land cover data (30 m) may not have been appropriate to
describe within territory space use for Prairie Warblers.
Lack of significant RUF findings could imply little consistency in resource selection at
the territory level, but more likely suggest the data chosen to portray habitat conditions within a
territory were likely not suited for this analysis. The 30 m C-CAP data used was likely too
coarse for describing Prairie Warbler habitat use within territories. Grass and sand land cover
classes (as portrayed by 2006 C-CAP imagery) likely contained shrubs and saplings that went
undetected with 30 m resolution data. Finer-resolution land cover data may provide more useful
insight into Prairie Warbler space use within a breeding territory.
Individual Prairie Warblers tended to vary in their use of specific resources. Further
exploration of drivers behind resource use could provide greater insight into landscape patterns
associated with Prairie Warbler breeding territory establishment, given the appropriate spatial
data resolution is utilized (I suggest 1m). The characterization of use for RUFs describing
Michigan Prairie Warbler populations will be an iterative process. Further refinement in how
best to describe specific landscape conditions/covariates and then determine the correct model of
use for those conditions should continue to be applied and reassessed to create better RUFs for
this species. Increased understanding of factors associated with variation in each RUF
coefficient will improve representation of how each coastal Prairie Warbler population uses
resources. Once additional sources of difference or variation are found, additional or refined
RUFs that target those factors could provide greater explanation of resource use in specific areas
that may be applied in similar areas.

94

CONSERVATION AND MANAGEMENT RECOMMENDATIONS
Protect what we have. The threat of coastal habitat loss may be the biggest threat to
coastal Michigan Prairie Warbler populations. The west Michigan dune landscape is providing a
source of early successional habitat for Prairie Warblers that is maintained through wind and
wave dynamics, therefore no specific land management technique to increase coastal west
Michigan habitat appears to be needed. Land use changes that fragment current, large
contiguous areas of habitat and surrounding land cover could pose new threats to coastal Prairie
Warbler populations and potentially alter habitat suitability. As Southwell noted in her 2001
U.S. Forest Service Eastern Region report, “preservation of current habitat is a management
necessity”.
The extent to which land use changes such as urban development (Klepinger 2002), the
siting of wind towers, or new cultivation of cropland (Johnson et al. 2003) could impact habitat
suitability for Prairie Warbler (and other species) is not fully known. Land use development in
west Michigan coastal communities is growing and lakeshore properties in Michigan continue to
receive a disproportionate amount of development pressure relative to inland (United States
Environmental Protection Agency 2001, 2008). The 2002 U.S. Census data confirm that
Michigan’s top 10 coastal cities are losing population while coastal rural areas are rapidly
suburbanizing (United States Census Bureau 2002), indicating that people are moving out of
cities and small towns to develop new coastal areas for cottages, primary residences and business
(United States Census Bureau 2002). This type of development, while recognized as a source of
economic development, degrades the solitude many come to find along Michigan’s western
lakeshore by limiting coastal access and fragmenting viewscapes of continuous natural beach
(Klepinger 2002). These land use patterns, if continued, are predicted to cause greater

95

fragmentation of natural systems (e.g., coastal wetlands, open dune, and forested dune;
Klepinger 2002) for Prairie Warbler and other species.
Wind tower threats to migrating birds (Drewett et al. 2006, Klepinger 2007, American
Bird Conservancy 2010) and bats (Johnston et al. 2003, Baerwald et al. 2008) are well
documented and tower siting is under debate by the environmental community and landowners
concerned with change to aesthetics, quality of life factors, or concern over economic land value
brought about by wind tower siting.
New agricultural development that would fragment current land use has the potential to
increase Brown headed cowbird (Molothrus ater) abundance and thus increase Prairie Warbler
nest parasitism rates (Nolan 1978, Nolan et al. 1999). Throughout most of the breeding range,
Prairie Warbler nests are known to be parasitized by cowbirds (Nolan 1978), though nest
predation by snakes, chipmunks and corvids has been noted to be a bigger threat to Prairie
Warbler than cowbird parasitism; predation “swamping out” effects of parasitism in Indiana
(Nolan et a. 1999). Response of cowbird parasitism by female Prairie Warblers in Indiana was
desertion of 48 (46%) of 105 nests. Parasitism probably has no effect on Prairie Warbler
hatching success (hatching rate in Indiana study in unparasitized nests 95%, in parasitized nests
92%). Mean number of fledglings in 76 nests producing only Prairie Warblers was 3.36; in 9
nests also producing cowbirds, 0.78 warblers (a 73% reduction; Nolan 1978). Among 102
cowbird eggs observed in Prairie Warbler nests in Indiana (Nolan 1978), only 5% produced
fledglings. Mean annual simulated cowbird-caused reduction of Prairie Warbler productivity in
Indiana was 13.3% (Nolan 1978).
Habitat management plans including considerations for Prairie Warbler population
viability should emphasize larger areas of contiguous habitat with well dispersed shrubs to

96

accommodate multiple territories. A comprehensive list of environmental factors describing
habitat suitability for Michigan coastal Prairie Warbler populations is unknown, however my
results suggest that shrub cover should be abundant and well dispersed. I found a 39%
probability of Prairie Warbler occupancy when shrubs were detected within 1 m of an existing
Prairie Warbler territory (Figure 14). My model also predicted that Prairie Warblers will occupy
sites that are <75 m away from shrubs, but probability of occupancy at these farther distances
was low (1%). The farthest I measured shrubs from the edge of a Prairie Warbler territory in the
field was 30 m. Prairie Warblers will likely occupy landscapes in a clumped pattern. My
findings show there was an 80% probability of Prairie Warbler presence when another Prairie
Warbler was within 91 m. Mean 95% Prairie Warbler territory area estimates ranged from 1.01
in Oval Beach to 3.17 in Sleeping Bear Dunes (Table 11), providing conservation planning
guidance for minimum area use requirements. Therefore, accommodating larger, contiguous
areas of early successional habitat has the potential to accommodate larger populations of Prairie
Warblers long term.
Surveys for Prairie Warblers should be located in areas where Prairie Warblers
have been detected in recent years. All surviving male adults likely return to the previous
year’s territory (Nolan 1978) as long as early successional habitat remains. The male return rate
to previous year’s territory was 65% based on Nolan’s 20-year Indiana study (Nolan 1978, Nolan
et al. 1999). After returning (n = 75 cases), 27% of unmated males moved within 2 - 8 days.
Known distance separating old and new territories ranged from 10 - 3,400 m (mean = 710 m, n =
12; Nolan 1978). Nolan noted that previous reproductive success, age, and number of previous
moves, and territory suitability were not correlated with such moves. The female return rate was
smaller; 19 of 105 females returned to the study area in one or more years following banding

97

(Nolan 1978, Nolan et al. 1999). Fourteen females nested on breeding territories that overlapped
their territory from the preceding year (Nolan 1978). Distances between nesting sites for the 2
years ranged from 12 to 325 m (mean = 90 m) (Nolan 1978). Reproductive success and
remaining on the study area after breeding were positively correlated with returns.
Prairie Warblers may disperse from their breeding territory once the nest has fledged,
indicating the importance of surrounding habitat beyond the breeding territory before fall
migration. Nolan (1978) showed that after breeding, 79% of 63 banded males were found before
migration. Some males had stayed near or on the territory, some had moved and then returned to
the territory and some spent the post-breeding time on a previous year’s territory. These areas
were not correlated with reproductive success, date that breeding ended, or date of molt. The
same pattern was found for 43% of 54 females located post-breeding (Nolan 1978).
Future territory mapping, to estimate territory boundaries or utilization
distributions (UD), should collect at least 30 registration (GPS) points if kernel-estimation
techniques similar to the type used in this study will be used. Accurate kernel estimation
assumes that sampling is sufficient to quantify relative differences in resource use (Garton et al.
2001). Seaman et al. (1999) suggests that with “adequate sample sizes” of >30 - 50 registration
points, fixed kernel estimators better represent differential space use than other UD techniques
and perform well at the outer boundary. Failure to secure an adequate sample size could have
bearing on the kernel estimate of the territory UD and thus resource use estimates needed for
future updates in analyzing Michigan Prairie Warbler populations.
Future survey, monitoring, or habitat protection design, using GIS applications to
automate location or identification of potential habitat, should use spatial data with a
resolution finer than 30 m. The 30 m, 2006 C-CAP data used in RUF analysis (Chapter 3) was

98

likely too coarse for describing Prairie Warbler habitat use within territories. Grass and sand
land cover classes (as portrayed by 2006 C-CAP imagery) used in my survey of land cover for
RUF assessment likely contained shrubs and saplings that went undetected with 30 m resolution
data. While on-the-ground vegetation surveys within breeding territories (and in unoccupied
areas for comparison) were thorough, the more sparsely-distributed shrub and sapling component
in southern territories the birds were likely using for nesting structure, did not translate as
dominant drivers of Prairie Warbler presence in those areas. Prairie Warblers nest in shrubs and
trees 1 - 3 m in height (Nolan 1978, Walkinshaw 1991, Dunn and Garrett 1997, Cooper 2000)
and defend territories from the highest perches in the area, along territory boundaries (Nolan
1978). They prefer trees and shrubs with numerous branches, twigs and leaves distributed
throughout (Nolan 1978). Nolan (1978) noted that among 577 nest plants in Indiana, about 27%
were ≥5 m from nearest tree or shrub, 34% were in the middle of a small tree clump, 37% were
at the edge of a large tree clump or woods, 2% were within woods. Among 608 nests in an
Indiana study, 51% were placed in American elm, which constituted 25% of the early
successional fallow fields in the study area (Nolan 1978). Placement of 555 Indiana nests were
as follows: 47% in upright fork; 27% on twigs, against vertical branch; 18% on horizontal or
diagonal branch; 5% in vines or wedged among small branches; 3% other. Though I did not
search for nests, the 2 nests I observed during this study were in common juniper and red-osier
dogwood (Cornus stolonifera), placed in forks of branches. Thus, based on Nolan’s Indiana
study and my own observations, I suggest that being able to discern the shrub component on the
landscape is critical. Spatial resolution finer than 30 m will allow identification of these areas in
future survey design and analysis using GIS and habitat modeling applications.

99

Prairie Warbler management plans should focus on the health and connectivity of
whole systems (i.e., western Lake Michigan dune ecosystem and adjacent potential habitat)
to provide access to a diversity of future habitat options, given overarching threats like
invasive species and climate change and their unknown future impacts, to maintain and
conserve landscapes important to Prairie Warbler. Michigan lies along the northern edge of
the Prairie Warbler breeding range and the contribution of this study to Prairie Warbler ecology
is the first of its kind in Michigan or the northern Great Lakes region. While understanding
Prairie Warbler habitat selection strategies at the northern edge of the breeding range could
expand our overall understanding of Prairie Warbler ecology, fiscal and staff limitations provide
little capacity to address individual species management efforts. Additionally, the broader and
overarching threats of invasive species and climate change to the dune ecosystem and
surrounding systems, will challenge the adaptive tolerance of many niche-specific species and
those already negatively affected by existing stressors (e.g., pests, disease, competition with
invasive species).
The range of potential landscape composition changes projected in the Midwest and
Great Lakes region under a changing climate (United States Global Change Research Program
2009, Lofgren et al. 2011) will be influenced by the projected 1- 3ºC degree temperature increase
by the end of the 21st century as well as increases in heavy precipitation events and storm
intensities with longer periods of drought between storm events (United States Global Change
Research Program 2009). Addition of species moving into the Midwest will also contribute to
competition for resources (Janetos et al. 2008).
While it is important to monitor rare species and design management plans with the needs
of individual species in mind, an ever-evolving shift in focus from individual species

100

management efforts toward protecting connected corridors of natural communities (Parmesan
1996, Union of Concerned Scientists 2010) and potentially their geophysical underpinnings
(Anderson and Ferree 2010) will allow plants and animals the opportunity to move and adjust to
a range future habitat and climate conditions. Protecting the ultimate drivers of biodiversity will
be key to sustaining diverse, and thus healthy, systems over the long term.
Considering Climate Change Vulnerability Index (CCVI; NatureServe 2012) ratings
for Prairie Warbler in the context of 1) species rarity rankings on a state and national level,
2) relative range position in Michigan compared to the full extent of the range, and 3)
potential for breeding range shift into and out of the state should be included in future
Prairie Warbler assessments and long term management planning (e.g., incorporation into
State Wildlife Action Plan update considerations). The CCVI ratings for Prairie Warbler (as
with all species) should be updated as new information becomes available. Examining how
phenological changes are tracking with breeding season temperature or precipitation dynamics in
Michigan may inform scoring.
Communicate with landowners adjacent to areas supporting Prairie Warblers,
landowners with potential habitat, and the public who visits and uses state and federal
parks to notify them of Prairie Warbler occurrence records and encourage conservation
action and partnership when possible. Public awareness of rare (or common) species, their
habitat needs, and research and conservation efforts underway, allow the public the opportunity
to “buy in” to conservation and associate recommended conservation land use practices with
their actions and the health and well-being of places they care about.

101

DATA LIMITATIONS
This study focused on West Michigan, coastal Prairie Warbler breeding habitat and
cannot provide management recommendations for non-coastal occurrences. Casual
observations in other areas of the state have been reported to the Michigan Breeding Bird Atlas
Project (Brewer et al. 1991), Michigan Natural Features Inventory (2011), and on various birding
list serves; specifically, Kirtland’s Warbler (D. kirtlandii) breeding areas in the northeast Lower
Peninsula, various locations in Washtenaw County, and north-central Mecosta County.
Speculation on why Prairie Warblers are not more abundant on “inland” early successional areas
similar to those where Prairie Warbler are found in Indiana, for example, cannot be made beyond
re-stating that Michigan lies at the northern extent of the breeding range (Dunn and Garrett
1997), Prairie Warblers have never been abundant in the state (Nolan 1978), and populations
have likely always been disjunct (Nolan 1978).
Some have speculated that “areas of apparently suitable habitat (i.e., jack pine plains)”
exist but remain unoccupied (Cooper 2000) or that “much suitable habitat appears to be
available” (Southwell 2001) or that “there appears to be no shortage of Prairie Warbler habitat in
Michigan…unoccupied habitat that seems suitable remains” (Walkinshaw 1991). Prairie
Warbler habitat requirements may be more specific than previously thought. The
aforementioned accounts often link potential to increase Prairie Warbler abundance to Kirtland’s
Warbler management efforts and liken their habitat requirements (Cooper 2000, Walkinshaw
1991). These conclusions seem to be primarily based on reported Prairie Warbler use of jack
pine plains prior to 1960, with Prairie Warbler likely using the open, shrubby areas found within
jack pine stands as they were structured at that time.

102

The plant composition and structure of the jack pine plains in 1950’s and 1960’s was
different than it is today. Fire suppression efforts had reduced the jack pine range at that time,
and thus Kirtland’s Warbler habitat availability (Olson 2002, Michigan Department of Natural
Resources 2012). The logging, burning, seeding, and replanting of jack pine necessary for
Kirtland’s Warbler management increased the amount of jack pine forest over time, and changed
the forest and understory composition as well as the landscape structure (Michigan Department
of Natural Resources 2012).
Historically Prairie Warblers in Michigan were more common in the jack pine plains
(Mayfield 1960) than they are today, with Mayfield noting Prairie Warblers nesting on 3 of 10
“Kirtlands Warbler areas” (specific locations or habitat conditions were not stated). Walkinshaw
(1991) and Felsing (1994) note that Prairie Warbler populations in the northeast Lower Peninsula
of Michigan likely peaked in the 1950’s and 1960’s. Timing of Mayfield’s observations
coincide with establishment of Kirtland’s Warbler habitat management by the U.S. Forest
Service and Michigan Department of Natural Resources between 1957 and 1962 (Michigan
Department of Natural Resources 2012). These changes may have contributed to fewer Prairie
Warbler observations in these areas after 1960. Brown headed cowbird prevalence in the area
could have also contributed to fewer Prairie Warblers in the region.

CONCLUSIONS
In conclusion, protecting the early successional dune ecosystems from fragmentation will
provide west Michigan coastal Prairie Warbler populations the opportunity to continue accessing
these areas. Breeding territories have been found in groups along the lakeshore, suggesting the
tendency for Prairie Warblers to settle near one another (conspecific attraction). Because Prairie

103

Warblers tend to return to the same breeding territories annually, continuing to maintain large,
contiguous areas of early successional dune landscapes has the potential to accommodate larger
populations of Prairie Warblers long term. Use of spatial data with resolution finer than 30 m (I
recommend 1 m) may best identify potential habitat when using GIS applications. Threats of
coastal development and subsequent potential for reduced habitat suitability due to area
constraints, nest predation, and nest parasitism, compounded by threats Prairie Warbler may face
on wintering grounds and along migratory routes (which were not examined in this study) may
reduce Prairie Warbler abundance in some of the only areas in Michigan Prairie Warblers are
currently observed.
Additions to our basic understanding of Prairie Warbler conservation could be gained
through long-term examination of wintering habitat status and population dynamics there.
Examination of potential changes in phenology (i.e., spring leaf-out and insect emergence)
experienced along migratory routes and subsequent effects on breeding population viability and
relative abundance would also be valuable. Further refinement of habitat models for Prairie
Warbler should consider social factors in addition to resource distributions to explore the
importance of conspecifics attraction in breeding site determination.

104

APPENDIX

105

0

300

600

1,200 Meters

¯

Fig. 16. Eight Prairie Warbler breeding territories mapped in 2005, Warren Dunes State Park,
Berrien County, Michigan. Each breeding territory is designated by an individual color. The
95% use percentile for an individual Prairie Warbler with its interior 50% (“core”) use percentile
is shown overlaid on 1 m resolution, 2005 United States Department of Agriculture, National
Agriculture Imagery Program, land cover data.

107

0

100

200

400 Meters

¯

Fig. 17. Two Prairie Warbler breeding territories mapped in 2005, Oval Beach, Allegan County,
Michigan. Each breeding territory is designated by an individual color. The 95% use percentile
for an individual Prairie Warbler with its interior 50% (“core”) use percentile is shown overlaid
on 1 m resolution, 2005 United States Department of Agriculture, National Agriculture Imagery
Program, land cover data.

108

0

365

730

1,460 Meters

¯

Fig. 18. Six Prairie Warbler breeding territories mapped in 2005, Saugatuck Dunes State Park,
Allegan County, Michigan. Each breeding territory is designated by an individual color. The
95% use percentile for an individual Prairie Warbler with its interior 50% (“core”) use percentile
is shown overlaid on 1 m resolution, 2005 United States Department of Agriculture, National
Agriculture Imagery Program, land cover data.

109

0

100

200

400 Meters

¯

Fig. 19. Two Prairie Warbler breeding territories mapped in 2005, Kitchel-Lindquist Dunes
Preserve, Ottawa County, Michigan. Each breeding territory is designated by an individual
color. The 95% use percentile for an individual Prairie Warbler with its interior 50% (“core”)
use percentile is shown overlaid on 1 m resolution, 2005 United States Department of
Agriculture National Agriculture Imagery Program land cover data.

110

0

600

1,200

2,400 Meters

¯

Fig. 20. Three Prairie Warbler breeding territories mapped in 2004, southern portion of Sleeping
Bear Dunes National Lakeshore, Benzie County, Michigan. Each breeding territory is
designated by an individual color. The 95% use percentile for an individual Prairie Warbler with
its interior 50% (“core”) use percentile is shown overlaid on 1 m resolution, 2005 United States
Department of Agriculture National Agriculture Imagery Program land cover data.

111

0

600

1,200

2,400 Meters

¯

Fig. 21. Six Prairie Warbler breeding territories mapped in 2004, central portion of Sleeping
Bear Dunes National Lakeshore, Leelanau County, Michigan. Each breeding territory is
designated by an individual color. The 95% use percentile for an individual Prairie Warbler with
its interior 50% (“core”) use percentile is shown overlaid on 1 m resolution, 2005 United States
Department of Agriculture National Agriculture Imagery Program land cover data.

112

0

600

1,200

2,400 Meters

¯

Fig. 22. Eleven Prairie Warbler breeding territories mapped in 2004, northern portion of
Sleeping Bear Dunes National Lakeshore, Leelanau County, Michigan. Each breeding territory
is designated by an individual color. The 95% use percentile for an individual Prairie Warbler
with its interior 50% (“core”) use percentile is shown overlaid on 1 m resolution, 2005 United
States Department of Agriculture National Agriculture Imagery Program land cover data.

113

LITERATURE CITED

114

LITERATURE CITED

American Bird Conservancy. 2010. Wind energy. www.abcbirds.org/abcprograms/policy/
collisions/wind_faq.html. Accessed March 2012.
Anderson M. G., and C. E. Ferree. 2010. Conserving the stage: climate change and the
geophysical underpinnings of species diversity. PLoS ONE 5(7):e11554.doi:10.1371/
journal.pone.0011554.
Baerwald, E. F., G. H. D’Amours, B. J. Klug, and R. M. R. Barclay. 2008. Barotrauma is a
significant cause of bat fatalities at wind turbines. Current Biology 18(16):695-696
Barrows, W. B. 1912. Michigan bird life. Special bulletin of the Department of Zoology and
Physiology of the Michigan Agricultural Collage. Michigan Agricultural College, East
Lansing, Michigan, USA.
Beer, J. R., L. D. Frenzel, and N. Hanson. 1956. Minimum space requirements of some nesting
passerine birds. Wilson Bulletin 68:200-209.
Betts, M. G., A. S. Hadley, N. Rodenhouse, and J. J. Nocera. 2008. Social information trumps
vegetation structure in breeding-site selection by a migrant songbird. Proceedings of the
Royal Society of Biological Sciences 275:2257-2263.
Brewer, R., G. A. McPeek, and R. J. Adams, Jr. 1991. Atlas of breeding birds in Michigan.
Michigan State University Press, East Lansing, Michigan, USA.
Cooper, J. L. 2000. Special animal abstract for Dendroica discolor (Prairie Warbler). Michigan
Natural Features Inventory, Lansing, Michigan, USA.
Danchin, E., T. Boulinier, and M. Massot. 1998. Conspecific reproductive success and breeding
habitat selection: implications for the study of coloniality. Ecology 79:2415-2428.
Danchin, E., L. A. Giraldeau, T. J. Valone, and R. H. Wagner. 2004. Public information: from
nosy neighbors to cultural evolution. Science 305:487-490.
Doligez, B., E. Danchin, and J. Clobert. 2002. Public information and breeding habitat selection
in a wild bird population. Science 297:1168-1170.
Drewett, A. L. and R. H. W Langston. 2006. Assessing the impacts of wind farms on birds.
Ibis 148:29–42.
Dunn, J. L., and K. L. Garrett. 1997. A field guide to warblers of North America. Houghton
Mifflin Co., Boston, Massachusetts, USA.

115

Felsing, J. 1994. Prairie warbler (Dendroica discolor) in G. A. McPeek and R. J. Adams,
editors. The birds of Michigan. Indiana University Press, Bloomington, Indiana, USA.
Garton, E. O., M. J. Wisdom, F. A. Leban, and B. K. Johnson. 2001. Experimental design for
radiotelemetry studies. Pages 16-42 273 in J. J. Millspaugh and J. M. Marzluff, editors.
Radio tracking and animal populations. Academic Press, San Diego, California, USA.
Hahn, B. A. and E. D. Silverman. 2006. Social cues facilitate habitat selection: American
redstarts establish breeding territories in response to song. Biology Letters 2:337-340.
Harrison, M. L., D. J. Green, and P. G. Krannitz. 2009. Conspecifics influence the settlement
decisions of male Brewer’s Sparrows at the northern edge of their range. The Condor,
111(4):722-729.
Janetos, A., L. Hansen, D. Inouye, B. P. Kelly, L. Meyerson, B. Peterson, and R. Shaw, 2008:
Biodiversity. Pages 151-181 in P. Backlund, A. Janetos, D.Schimel, J. Hatfield, K.
Boote, P. Fay, L. Hahn, C. Izaurralde, B. A. Kimball, T. Mader, J. Morgan, D. Ort, W.
Polley, A. Thomson, D. Wolfe, M. G. Ryan, S. R. Archer, R. Birdsey, C. Dahm, L.
Heath, J. Hicke, D. Hollinger, T. Huxman, G. Okin, R. Oren, J. Randerson, W.
Schlesinger, D. Lettenmaier, D. Major, L. Poff, S. Running, L. Hansen, D. Inouye, B. P.
Kelly, L. Meyerson, B. Peterson, and R. Shaw, editors. The effects of climate change on
agriculture, land resources, water resources, and biodiversity in the United States.
Synthesis and Assessment Product 4.3. U.S. Department of Agriculture, Washington,
D.C., USA.
Johnson, G. D., W. P. Erickson, M. D. Strickland, M. F. Shepherd, D. A. Shepherd, and S. A.
Sarappo. 2003. Mortality of bats at a large-scale wind power development at Buffalo
Ridge, Minnesota. American Midland Naturalist 150:332-342.
Klepinger, M. E. 2002. Status of planning and zoning in Michigan’s Great Lakes shoreline
communities. Sustainable Costal Community Development Initiative, Michigan Sea
Grant College Program. www.miseagrant.umich.edu/downloads/coastal/planning_
zoning_ 02_Klepinger.pdf. Accessed March, 2012.
Klepinger, M. E. 2007. Michigan land use guidelines for siting wind energy systems.
Michigan State University Extension Bulletin WO-1053. http://miwind.msue.msu.
edu/uploads /files/msue_bulletin.pdf. Accessed March 2012
Lofgren, B. M., T. S. Hunter, and J. Wilbarger. 2011. Effects of using air temperature as a proxy
for potential evapotranspiration in climate change scenarios of Great Lakes basin
hydrology. Journal of Great Lakes Research 37(4):744-752.
Marzluff, J. M., J. J. Millspaugh, P. Hurvitz, and M. S. Handcock. 2004. Relating resources to a
probabilistic measure of space use: forest fragments and Steller’s jays. Ecology 85:14111427.

116

Mayfield, H. 1960. The Kirtland’s warbler. Cranbrook Institute of Science Bulletin 40,
Bloomfield Hills, Michigan, USA.
Michigan Department of Natural Resources. 2012. Featured species and wildlife habitat
management: Kirtland’s warbler (Dendroica kirtlandii). www.michigan.gov/dnr/0,4570,
7-153-10370_12145_12202-32591--,00.html. Accessed March 2012.
Michigan Natural Features Inventory. 2011. Rare species explorer. http:// web4.msue.msu.edu/
mnfi/explorer. Accessed 2004-2011.
National Oceanic and Atmospheric Administration, Coastal Services Center. 2008. NOAA
Coastal Change Analysis Program (C-CAP) 2006 Era Land Cover Data of the C-CAP
zone 51 (upper midwest). www.csc.noaa.gov/crs/lca. Charleston, South Carolina, USA.
Natural Resources and Environmental Protection Act. 1994. Act 451 by section 36503, MCL
324.36503, and listed under R 299.1026 as amended of the Michigan Administrative
Code.
NatureServe. 2011. NatureServe explorer: An online encyclopedia of life. Version 7.1.
www.natureserve.org/explorer. NatureServe, Arlington, Virginia, USA. Accessed April
2012.
NatureServe. 2012. NatureServe Climate Change Vulnerability Index. Version 2.1.
www.natureserve.org/prodServices/climatechange/ccvi.jsp. Natureserve, Arlington,
Virgiana, USA. Accessed March 2012.
Nolan, V., Jr. 1978. The ecology and behavior of the prairie warbler Dendroica discolor.
Ornithological Monographs, No. 26.
Nolan, Jr., V., E. D. Ketterson, and C. A. Buerkle. 1999. The birds of North America:
prairie warbler (Dendroica discolor). Pages 1-26 in A. Poole and F. Gill, editors. The
birds of North America, No. 584. Philadelphia, Pennsylvania, USA.
Olson, J. A. 2002. Special animal abstract for Dendroica kirtlandii (Kirtland’s warbler).
Michigan Natural Features Inventory, Lansing, Michigan, USA.
Parmesan, C. 1996. Climate change and species’ range. Nature 382:765-766.
Prather, J. W. 1994. A comparison of the ecology and behavior of two species of congeneric
wood warblers in the mangrove habitats of south Florida. M. A. thesis, University of
Colorado, Denver, Colorado, USA.
Rappole, J., H., E. S. Morton, T. E. Lovejoy III, and J. L. Ruos. 1983. Neartic avian migrants in
the Neotropics. U.S. Department of Interior, Fish and Wildlife Service, Washington,
D.C., USA.

117

Robbins, C. S., D. Bystrak, and P. H. Geissler. 1986. The breeding bird survey: its first fifteen
years, 1965-1979. U.S. Fish and Wildlife Service Resource Publication No. 157.
Washington, D.C., USA.
Robbins, C. S., R. E. Steward, and M. Karplus. 1947. Breeding bird census of dry deciduous
scrub. Audubon Field Notes 1:200-201.
Robinson, S. K., E. J. Heske, and J. D. Brawn. 1999. Factors affecting the nesting success of
edge and shrubland birds. Report submitted to Illinois Department of Natural Heritage,
Springfield, Illinois, USA.
Sager, A. 1839. Report of Dr. Abraham Sager, Zoologist of Geological Survey. House
Documents, State of Michigan, USA 1839:410-421.
Sauer, J. R., J. E. Hines, J. E. Fallon, K. L. Pardieck, D. J. Ziolkowski, Jr., and W. A. Link.
2011. The North American breeding bird survey, results and analysis 1966 - 2009.
Version 3.23.2011. www.mbr-pwrc.usgs.gov/bbs/bbs.html. USGS Patuxent Wildlife
Research Center, Laurel, Maryland, USA. Accessed March, 2012.
Seaman, D. E., J. J. Millspaugh, B. J. Kernohan, G. C. Brundige, K. J. Raedeke, and R. A.
Gitzen. 1999. Effects of sample size on kernel home range estimates. Journal of
Wildlife Management 63:739-747.
Southwell, D. K. 2001. Conservation assessment for Prairie Warbler (Dendroica discolor).
USDA Forest Service Eastern Region. USDA Forest Service, Hiawatha National Forest,
Michigan, USA.
Union of Concerned Scientists. 2010. Are butterflies secret harbingers of global warming?
www.ucsusa.org/global_warming/science_and_impacts/science/climate-scientist-camilleparmesan.html. Accessed, March 2012
United States Census Bureau. 2002. Michigan. Geograpahy Division. www.census.gov/geo/
www/guidestloc/st26_mi.html. Accessed March 2012.
United States Environmental Protection Agency. 2001. National coastal conditions report.
Office of Research and Development/Office of Water. http://water.epa.gov/type/oceb/
assessmonitor/nccr/index.cfm. Washington, D.C., USA.
United States Environmental Protection Agency. 2008. National coastal conditions report III.
Office of Research and Development/Office of Water. http://water.epa.gov/type/oceb/
assessmonitor/nccr/index.cfm. Washington, D.C., USA.
United States Global Change Research Program. 2009. Global climate change impacts in the
United States. Regional climate impacts: Midwest. Cambridge University Press, New
York, New York, USA.

118

Walkinshaw, L. H. 1991. Prairie warbler. Pages 418-419 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.
Wood, N. A. 1951. The birds of Michigan. University of Michigan Museum of Zoology
Miscellaneous Publication 75, Ann Arbor, Michigan, USA.
Zimmerman, D. A., and J. Van Tyne. 1959. A distributional checklist of the birds of Michigan.
University of Michigan Museum of Zoology, Ann Arbor, Michigan, USA.

119