as...» .
mamas.
kg.

.7: I

.
Z
«a

El ;

..
A

.3... . 6.
3......

s» .
x. 55.5.”

a n ,
9 “1-2.;
34?

.5 at. ... . .
5.5.5.}.
_ x93. 3..

4%). Air! ..
Phat/{r hi nut...
\ ck..." 4? :33 .w
a\_ ”Erin-Ru , .. . . 11......
. lxlfi up. in. u
r 3......

. “:9
:11“.Lfl.’u2_ mfi
i...) .21":-
aalfin?x1&:.u¢
nil-7....
V t
, . Lk
. . .3 Vim
E . n
2 v... v... N a E
2 . . , 11.: z x J \.
K3... :5 :Jtunfluu irisevflw. . 1.3.:
.mfix: . n .V 1.!

:.
I. .

. .15..
.1 it“... i

. 3
.531 rd
in: 13:42}.

a." 34%.. Rafi... 1.
Jr. st. 224} .2.
*A’!

I; .1 It
2.3.178 3 i:
#5... .C: it .

15.714“... 3

1.
35 £5. £1;
twat-5.9.5113!

15m. if? x 1. 43:39.5.“

+ it; 1".3- 1 . In t. ..

.4 3| A . 1
3w. ‘ 5..
a $.31: vi... urtvn
viz-11.3.1546 nu .If
\ . .5. I. . .3).
.xonhtfinfllgfl..l.o akin .
tall: ..
vars-ID
£325... . I
.v. A-

It...
1. .1!

In: 4.;
:1 2.
1.

 

W._...E p fifiyfigwfi‘ infigt; a . E. .

. , . .1}. .11! £31115”: u .300.-
.rm‘fiflg . . . . invagnix‘ét

IUIIIIHIHHIIIUIWill”lllll||||illlllllllllHlillllil‘l

3 31293020483

:2. CLO

 

LIBRARY
Michigan State
University

 

 

 

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

 

DATE DUE

DATE DUE

DATE DUE

 

3575 W3 yoga

 

 

 

 

 

 

 

 

 

 

 

 

 

11/00 chIRCJDanOmpes-p. 14

 

THE EFFEl

THE EFFECTS OF LANDSCAPE CONTEXT AND COASTLINE COMPLEXITY ON
BIRDS IN GREAT LAKES COASTAL WET MEADOWS

By

Samuel Keith Riffell

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Zoology

2000

EFFECTS C
THE Bl]

Landscape
processes (eg ar
context. Landsca
landscape, and di
species richness.
located in 40 Grea
1993)- Twenty wi
each rear. A 018
patches and the as
each set of depend
[number of variable
regressions, Regrc
efiecrs of patch-let

coastline complex:

SPCCles IiC:

coastline compone

 

associated With co

wetland COUICXIS’ ll

ABSTRACT

THE EFFECTS OF LANDSCAPE CONTEXT AND COASTLINE COMPLEXITY ON
BIRDS IN GREAT LAKES COASTAL WET MEADOWS

By

Samuel Keith Riffell

Landscape studies on birds have focused on effects of patch- and regional-level
processes (e. g. area or fragmentation), but little is known about the effects of landscape
context. Landscape context refers to the composition of the adjacent and surrounding
landscape, and different contexts can have different effects on within-patch processes like
species richness. To test for effects of landscape context, I censused birds along transects
located in 40 Great Lakes coastal wet meadows during two breeding seasons (1997 and
1998). Twenty within-patch habitat characteristics (e. g. grass density) were measured
each year. A GIS database was used to calculate 24 characteristics of the adjacent
patches and the associated coastline. I used separate principal component analyses on
each set of dependent variables (i.e. habitat, landscape and coastline) to reduce the
number of variables and remove inter-correlations before using linear and logistic
regressions. Regressions involved a two-stage approach to adjust bird variables for
effects of patch-level variables before testing for effects of landscape context and
coastline complexity.

Species richness, total abundance and most species were related to landscape and
coastline components in consistent fashions. Higher richness and abundance were
associated with complex contexts (high number of adjacent patches and patch types),

wetland contexts, human-developed contexts, and less complex coastlines. Similarly,

 

 

species turnove
Other analyses
located in com

coastal lobes.

Great Lakes co
patches located
coastlines may
of species and l
contexts more :
kllOWlCdge abo

(turnover), mar

acCurately pred

species turnover was lower in large patches located in complex or wetland contexts.
Other analyses indicated that species richness and species composition of wet meadows
located in coastal coves was not substantially different than wet meadows located out on
coastal lobes.

These results indicate that conservation strategies for wetland birds in
Great Lakes coastal regions should consider effects of landscape context. Specifically,
patches located in complex and/or wetland contexts, but associated with simpler
coastlines may be better conservation choices because they support the greatest numbers
of species and birds. Similarly, wet meadows located in complex and/or wetland
contexts more stable species composition (i.e. lower turnover rates). Without explicit
knowledge about how landscape context affects species richness, abundance, and stability
(turnover), managers will be unable effectively select conservation areas and unable to

accurately predict how the species composition of these areas will change over time.

 

 

 

My disser
my major profess
Angie. Generous
doctoral program
Liu. Brian Keas ;

general encourage

 

 

 

database. This re
individuals: Dr, I
Rutledge. The Dr
computer resourc
me to work on pr
Chapter of the N:
the Michigan Pol

schmmhip; the

ACKNOWLEDGEMENTS

My dissertation would not have been possible without the counsel and support of
my major professor, Dr. Tom Burton, nor without the patience and support of my wife,
Angie. Generous guidance and helpful suggestions were provided throughout my
doctoral program by committee members Dr. Don Beaver, Dr. Jim Bence, and Dr. Jack
Liu. Brian Keas and Joe Gathman provided suggestions for sampling and analysis and
general encouragement. Brian Keas helped collect field data and construct the GIS
database. This research also benefited from the advice and assistance of several
individuals: Dr. Dave Ewert, Dr. Kevin Gutzwiller, Dr. Mic Hamas, and Daniel
Rutledge. The Department of Geography at Michigan State University provided
computer resources. Residents of Hessel, Cedarville, and DeTour graciously permitted
me to work on private lands. This research was supported by grants from the Michigan
Chapter of the Nature Conservancy (to Tom Burton); American Ornithologist's Union;
the Michigan Polar-Equator Club; the George and Martha Wallace Ornithology

Scholarship; the Graduate School and the Department of Zoology at Michigan State.

"I can do all things through Christ who strengthens me." Phillipians 4: 13

iv

 

LIST OF TAB
LIST OF FlGl
INTRODUC T

CHAPTER 1

LANDSCAPE

EFFECTS O.\
Introd-
Metho
Result
Discus
Implic

CHAPTER 2
LANDSCAP;
EFFECTS 0:
Introc
Metht
Resul
Discu
Impli

CHAPTER 3
PENIN SLTLJ
ARE COVE
Intro
Meti

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................ vii

LIST OF FIGURES ............................................................................................ ix

INTRODUCTION ............................................................................................... 1

CHAPTER 1

LANDSCAPE CONTEXT AND COASTLINE COMPLEXITY:

EFFECTS ON SPECIES RICHNESS AND ABUNDANCE ............................... 6
Introduction ............................................................................................. 7
Methods ................................................................................................. 10
Results ................................................................................................... 20
Discussion ............................................................................................. 4]
Implications for Conservation and Management .................................... 54

CHAPTER 2

LANDSCAPE CONTEXT AND COASTLINE COMPLEXITY:

EFFECTS ON SPECIES TURNOVER RATES ................................................ 57
Introduction ........................................................................................... 58
Methods ................................................................................................. 61
Results ................................................................................................... 67
Discussion ............................................................................................ 7]
Implications for Conservation and Management .................................... 84

CHAPTER 3

PENINSULA EFFECTS IN A COASTAL LANDSCAPE:

ARE COVES MORE SPECIES RICH THAN LOBES? .................................... 86
Introduction ........................................................................................... 87
Methods ................................................................................................. 88
Results ................................................................................................... 97
Discussion ............................................................................................ 102
Implications for Conservation and Management ................................... 105

SUMMARY AND RECOMMENDATIONS ................................................... 107

APPENDICES ................................................................................................. 110

BIBLIOGRAPHY ............................................................................................ 123

 

TABLE 1. L
northern shor
counties, M 1c

TABLE 2. S
in 40 wet me.“
in Chippewa

TABLE 3. l~
components :
TABLE 4. S
coastline cor
along the nor
Mackinac cc

TABLE 5 1

TABLE 6_ 4
0f hndscape
regression n
Ciated “rith 1
and MaCkin

ireCIIOnS O
bird \rar‘labl

TABLE 7.
compohem.

with that CC

LIST OF TABLES

TABLE 1. Landcover classification system derived for the
northern shoreline of Lake Huron in Chippewa and Mackinac
counties, Michigan ................................................................................................ 16

TABLE 2. Summary statistics for habitat variables measured
in 40 wet meadows along the northern shoreline of Lake Huron
in Chippewa and Mackinac counties, Michigan, 1997-1998. ................................. 21

TABLE 3. Habitat components derived from principal
components analysis on original habitat variables. ................................................. 23

TABLE 4. Summary statistics for landscape context and

coastline complexity variables measured for 40 wet meadows

along the northern shoreline of Lake Huron in Chippewa and

Mackinac counties, Michigan, 1997-1998. ............................................................ 25

TABLE 5. Landscape components and coastline components
derived from principal components analysis on original variables .......................... 27

TABLE 6. Comparison of within-patch habitat effects and effects

of landscape context and coastline complexity in two-stage

regression models of wet meadow birds in 40 wet meadows asso-

ciated with the northern shoreline of Lake Huron in Chippewa

and Mackinac counties, Michigan. (+) or (-) indicates the

directions of the relationship between that component and the

bird variable. ......................................................................................................... 31

TABLE 7. Results of regression modeling summarized by principal

component. Models for each of the listed bird variables contained

that component as a significant predictor (P < 0.10). (+) indicates

landscape or coastline characteristics that are positively associated

with that component; (-) indicates characteristics that are negatively

associated. ............................................................................................................. 38

TABLE 8. Summary statistics for mean habitat variables and habitat

change variables for 40 wet meadows associated with the northern

shoreline of Lake Huron in Chippewa and Mackinac counties,

Michigan, 1997-1998. ........................................................................................... 64

TABLE 9. Mean habitat components and habitat change

components derived from original variables which were retained
for analysis. ........................................................................................................... 68

vi

 

TABLE 10. End:
40 wet meadows
Huron in Chippe‘

TABLE II. Res

rates in wet mea' I
Lake Huron in C l
.1998. For all re;

TABLE 12 Con
between habitat c
and mean habitat
the northern shor
counties, Michigz

 

TABLE 12. C08
cove and lobe we
OfLake Huron, C
all regressions, c

TABLE 13. Res
Presence/absenm
coves or lobes 01
and Mackinac c(

TABLE 10. Extinctions, immigrations, and turnover rate for
40 wet meadows associated with the northern shoreline of Lake
Huron in Chippewa and Mackinac counties, Michigan, 1997-1998. ....................... 70

TABLE 11. Results of logistic regression on bird species turnover

rates in wet meadows associated with the northern shoreline of

Lake Huron in Chippewa and Mackinac counties, Michigan, 1997

-1998. For all regressions, n = 40 .......................................................................... 72

TABLE 12. Correlation coefficients and P-values for correlations

between habitat change components and area, perimeter-area ratio,

and mean habitat components for 40 wet meadows associated with

the northern shoreline of Lake Huron in Chippewa and Mackinac

counties, Michigan, 1997-1998 .............................................................................. 74

TABLE 12. Coefficients of adjusted species-area regression for

cove and lobe wet meadows associated with the northern shoreline

of Lake Huron, Chippewa and Mackinac counties, Michigan. For

all regressions, cove n = 10 and lobe n = 11 ........................................................... 99

TABLE 13. Results of regression on species abundance and

presence/absence variables in 21 wet meadows located in either

coves or lobes of the northern shoreline of Lake Huron, Chippewa

and Mackinac counties, Michigan .......................................................................... 100

vii

 

FIGLRE 1. Diag
Great Lakes coas
shoreline of Lake
Michigan, 1997-
relation to center
sampling techniq

FIGURE 2. Relat
(COASTZ) and tl
scape (l-km frorr

FIGURE 3. Relat
coastline (COAS'
(Him from perin

FIGLRE 4. Theo
island biogeogra;
to near islands; us
far islands ..........

FIGURE 5_ Dias:
meadow Patch d1
palCh not Clearly
meadow Patch in
meeting either c,

III COVES 0r DUI OI
retained fer analt

LIST OF FIGURES

FIGURE 1. Diagram of habitat sampling procedures used in 40

Great Lakes coastal wet meadows associated with the northern

shoreline of Lake Huron in Chippewa and Mackinac counties,

Michigan, 1997-1998. A) Locations of 20 habitat sampling points

relation to center of SO-m transect segments. B) Details of 3

sampling techniques usedvat each of these 20 points. ............................................ 13

FIGURE 2. Relationship between the the total coastline length
(COASTZ) and the mean patch size (ha) in the surrounding land-
scape (l-km from perimeter of the wet meadow). .................................................. 52

FIGURE 3. Relationship between the the fractal dimension of the
coastline (COASTI) and the percent of the surrounding landscape
(Man from perimeter of the wet meadow) comprised of bulrush marsh ................. 53

FIGURE 4. Theoretical species-area relationships as predicted by

island biogeography. Wet meadows located in coves are analogous

to near islands; wet meadows located on lobes are analogous to

far islands. ............................................................................................................. 89

FIGURE 5. Diagram of northern Lake Huron shoreline. A = wet

meadow patch distinctly located within a cove. B = wet meadow

patch not clearly located in either a lobe or out on a cove. C = wet

meadow patch distinctly located out on a lobe. Only wet meadow

meeting either criterion A or C were included in this analysis. ............................... 92

FIGURE 6. Species-area relationships in 30 wet meadows located

in coves or out on lobes. Sites to the left of dashed line are ones
retained for analysis ............................................................................................... 93

viii

 

B
effeC‘S O:
1993; Fre
Gum'flk
1995) 1“
Characterif
Dinsmore
and edge 3f
imponaflce '
et al. I 998 )
also centered
Faaborg et al.

Based t
have been deve
and Finch 1995)
(Schwartz 1 999).
effects ofadjacen
Only recently hat"
remis of species '
al 1999; Liu and
Humphrey 1999)

al 2983; Liu and

INTRODUCTION

Research in avian ecology over the last two decades has extensively explored the
effects of landscape structure on species richness (e. g. Blake and Karr 1987; Pearson
1993; Freemark et al. 1995), patch selection (e. g. Gutzwiller and Anderson 1987a;
Gutzwiller and Anderson 1992; Saab 1999), and breeding success (e. g. Robinson et al.
1995). Initially, research focused on how birds were affected by patch-level landscape
characteristics such as patch area (Galli et al. 1976; Blake and Karr 1987; Brown and
Dinsmore 1991), isolation (e. g. van Dorp and Opdam 1987; Brown and Dinsmore 1986),
and edge effects (see Andren 1995 for review). More recent research addressed the
importance of corridors (e. g. Dunning et al. 1995) and landscape connectivity (e. g. Haig
et al. 1998 )to the successful conservation of bird species. Substantial research effort has
also centered on important regional-level processes such as forest fragmentation (e. g.
Faaborg et al. 1995; Drolet et al. 1999; Rosenberg et al. 1999; Villard et al. ~1999).

Based on information from these and similar studies, specific recommendations
have been developed for the conservation and management of avian species (e. g. Martin
and Finch 1995) and various strategies for the design and management of preserves
(Schwartz 1999). However, such information is incomplete without considering potential
effects of adjacent and surrounding landscape elements, or landscape context, on patches.
Only recently have ecologists recognized that landscape context can affect patches in
terms of species diversity (Webb and Hopkins 1984; Bloutin and Jobin 1998; Cantero et
al. 1999; Liu and Ashton 1999), patch-use by species (Pearson 1993; Forys and
Humphrey 1999) and the sustainability of economic and ecological processes (Risch et

al. 1983; Liu and Ashton 1999) occurring within patches.

 

Corr
to birds has
context has
For example
near edges) 1
ambiguous (
edges (lands.

particular edj
explain varia-
(Gum‘iller a
occurrence of
etTects of lanc
HObson 1997}
Preserve avian
T0 amt
wetland-d 0 mi r
the Upper Penj
Patches deing
Cha‘raCtel’istics a
and associated c

patch a’65:! and

With 1

V

andSCape C

area and habitat.

Compared to other aspects of avian ecology, the importance of landscape context
to birds has received little attention by researchers; but yet, consideration of landscape
context has already helped clarify our understanding of some critical landscape processes.
For example, predicting the occurrence of edge effects on birds (increased nest predation
near edges) has been difficult because results of studies and experiments have been
ambiguous (Paton 1994). However, accounting for characteristics of patches adjacent to
edges (landscape context) has helped predict if and to what extent edge effects occur at a
particular edge (see Andren 1995; Bayne and Hobson 1997). Landscape context can also
explain variation in the bird species composition of habitats (Pearson 1993) and patches
(Gutzwiller and Anderson 1987a; Estandes and Temple 1999; Saab 1999) as well as the
occurrence of individual species. Unfortunately, studies which focus specifically on the
effects of landscape context on birds are uncommon (but see Pearson 1993; Bayne and
Hobson 1997), and this limits our understanding of avian ecology and hinders efforts to
preserve avian diversity.

To ameliorate this lack of information, I conducted a landscape-scale study in a
wetland-dominated, coastal region located along the northern shoreline of Lake Huron in
the Upper Peninsula of Michigan. Birds were censused in 40 coastal wet meadow
patches during 1998 and 1999. I also measured a suite of within-patch habitat
characteristics along with characteristics of the adjacent patches, surrounding landscape,
and associated coastline. For all analyses, bird variables were first adjusted for effects of
patch area and within-patch habitat characteristics. Thus, any significant relationships
with landscape context or coastline complexity are effects above and beyond those of

area and habitat. Three sets of analysis were carried out, each designed to test for effects

 

of landsca]
These anal
C h;
apparent ar
richness an
of a patch r
Pearson 19
(1995) desc
and the nur
urban come
of species a
Effort; but
nonhem La
Mer
One), hOWe
also idemif:
a Patch O\'e
comma C1
the within‘F
species mm

preserved 01

of landscape context and coastline complexity on a different aspect of avian biology.
These analyses correspond to three specific chapters which are described below.

Chapter One addresses the primary objective of my research: to test for any
apparent and strong associations between landscape context and measures of bird species
richness and abundance. Recent research has demonstrated that the species composition
of a patch may vary according to landscape context (e. g. Gutzwiller and Anderson 1987a;
Pearson 1993; Estandes and Temple 1999; Saab 1999). For example, Friesen et al.
(1995) described a negative relationship between bird species richness of forest patches
and the number of adjacent houses, and some species avoided forest patches located in
urban contexts altogether. Information about which contexts support the greatest number
of species and/or the highest density of particular species would benefit conservation
efforts, but this information is lacking for wetland-dominated landscapes such as the
northern Lake Huron coastline.

Merely identifying the contexts that maintain high species richness (i.e. Chapter
One), however, is not enough to ensure success of conservation efforts. Managers must
also identify which landscape contexts facilitate a more stable species composition within
a patch over time (i.e. lower species turnover rates). To help identify more stable
contexts, Chapter Two tests for effects of landscape context and coastline complexity on
the within-patch species turnover rates between 1997 and 1998. Information about
species turnover rates will also help managers predict how the species composition of
preserved or managed patches can be expected to change over time. Without this
knowledge, manager’s predictions about the future status of conservation areas are not

based on accurate information and may lead to poor management decisions.

 

Per
distance fr
l—‘orrnan 19
and penins
richness p2
peninsula e
meadows It
coastal lobt
coves that s

and also prc
Wetland pat
The
Oflandscape
northern Lal
COaStline are
Great Lakes
develomem
migratory. br
I992; Prince i

(Brewer et al .

Peninsula effects are defined as decreasing species richness with increasing
distance from the mainland along a peninsula (Forman and Godron 1986; Milne and
Forman 1986; Forman 1995). The northern Lake Huron coastline is highly interdigitated,
and peninsular lobes and coves abound. In Chapter 3, I test predictions about species
richness patterns along an interdigitated coastline that are based on the presence of
peninsula effects. Specifically, peninsula effects predict that the species richness of wet
meadows located in coves (near the mainland) should be higher than those located out on
coastal lobes (far down the peninsula). Identifying positions along coastal lobes and
coves that support higher species richness will further understanding of landscape context
and also provide more detailed information relevant to the selection and management of
wetland patches.

The information provided by this research will not only aid general understanding
of landscape context effects, but will also directly benefit the regional management of the
northern Lake Huron coastline. Although the wetlands associated with this section of
coastline are some of the most undeveloped freshwater coastal wetlands remaining in the
Great Lakes region, they are also currently threatened by residential and commercial
development. These wetlands are a conservation priority because they represent critical
migratory, breeding, and foraging habitat for a wide variety of bird species (Prince et al.
1992; Prince and Flegel 1995), many of which are declining regionally and/or locally
(Brewer et al. 1991). Unfortunately, knowledge about the influence of landscape
structure on birds in wetland-dominated landscapes is alarmingly scarce (see Freemark et
al. 1995 and Weller 1999 for reviews), and, except for riparian forests (e. g. Gutzwiller

and Anderson 1987a & b, Saab 1999), the effects of landscape context on birds in other

 

  
  
 
  

wetland landsca;
knowledge of thr
success of mana;

regions are serio

wetland landscapes are practically unknown (but see Naugle et al. 1999). Without
knowledge of the effects of landscape context on birds, the effectiveness and ultimate
success of management schemes and conservation efforts in the Great Lakes and other

regions are seriously compromised.

CHAPTER ONE

LANDSCAPE CONTEXT AND COASTLINE COMPLEXITY:

EFFECTS ON SPECIES RICHNESS AND ABUNDANCE

 

 

Over the
landscape comp
Pearson 1993 t F
1987a; Gutzwill
etal1995)‘ Re
characteristics 5‘
Dinsmore 199])
and edge effects
importance of Cl
etal. 199811013
also centered or
Faaborg et al. 1

Based c
developed spec
Species ( e. g. M
management 0
is incomplete x
context, on pa
affect patches

1998; camert

Forys and Hu

(RiSch et a] ‘

INTRODUCTION

Over the last decade, research in avian ecology has emphasized the importance of
landscape composition and structure to species richness (e. g. Blake and Karr 1987;
Pearson 1993; Freemark et al. 1995), patch selection (e. g. Gutzwiller and Anderson
1987a; Gutzwiller and Anderson 1992; Saab 1999), and breeding success (e. g. Robinson
et al. 1995). Research initially focused on how birds were affected by landscape
characteristics such as patch area (Galli et al. 1976; Blake and Karr 1987; Brown and
Dinsmore 1991), isolation (e. g. van Dorp and Opdam 1987; Brown and Dinsmore 1986),
and edge effects (see Andren 1995 for review) but has developed to address the
importance of corridors (e. g. Dunning et al. 1995) and landscape connectivity (e. g. Haig
et al. 1998) to the successful conservation of bird species. Substantial research effort has
also centered on important regional-level processes such as forest fragmentation (e. g.
Faaborg et al. 1995; Drolet et al. 1999; Rosenberg et al. 1999; Villard et al. 1999).

Based on information from these and similar studies, wildlife managers have
developed specific recommendations for the conservation and management of avian
species (e. g. Martin and Finch 1995) and various strategies for the design and
management of preserves (Schwartz 1999). But no matter how detailed, such information
is incomplete without considering effects of adjacent landscape elements, or landscape
context, on patches. Ecologists have only recently recognized that landscape context can
affect patches in terms of species diversity (Webb and Hopkins 1984; Bloutin and Jobin
1998; Cantero et al. 1999; Liu and Ashton 1999), patch-use by species (Pearson 1993;
Forys and Humphrey 1999) and the sustainability of economic and ecological processes

(Risch et al. 1983; Liu and Ashton 1999) occurring within patches.

For b1]
but already co
important lanc
effects on bird
studies and ex
characteristics
to what extent
Hobson 1997)
composition 0
EStandes and '
Friesen et al. (
forest patches
PaIChes locate.
Specificany on
1993; Ba)“ e a
hinders Efforts

T0 leai

SPECIES-occup.
I

For birds, the importance of landscape context is just beginning to be understood,
but already consideration of landscape context has furthered understanding of some
important landscape processes. For example, understanding the occurrence of edge
effects on birds (increased nest predation near edges) has been difficult because results of
studies and experiments have been ambiguous (Paton 1994). However, accounting for
characteristics of patches adjacent to edges (landscape context) has helped predict if and
to what extent edge effects occur at a particular edge (see Andren 1995; Bayne and
Hobson 1997). Landscape context can also explain variation in the bird species
composition of habitats (Pearson 1993) and patches (Gutzwiller and Anderson 1987a;
Estandes and Temple 1999; Saab 1999) as well as the occurrence of individual species.
Friesen et al. (1995) described a negative relationship between bird species richness of
forest patches and the number of adjacent houses, and some species avoided forest
patches located in urban contexts altogether. Unfortunately, studies which focus
specifically on the effects of landscape context on birds are uncommon (but see Pearson
1993; Bayne and Hobson 1997), and this limits our understanding of avian ecology and
hinders efforts to preserve avian diversity.

To learn more about the effects of landscape context on species richness and
species-occupancy in wetland-dominated landscapes, I censused birds in 40 coastal wet
meadow patches and measured characteristics of the adjacent patches, surrounding
landscape, and associated coastline. Variables describing bird species richness and
abundance were first adjusted for wetland area, perimeter, and within-patch habitat
characteristics so that effects of landscape context would not be confounded with these

factors.

M)" pri
landscape com
were also guid
patch composi
patches locatec
patch types) w
patch type (For
contexts would
types (Freisen e
developed cont
developed cont:
hFPOIhesized sh
should have hi g

1987),

compan

a1 1995 and We

Gum-trier and a

like most “'etlanc

dei'leiopment. Th

My primary objective was to test for any apparent and strong associations between
landscape context and measures of bird species richness and abundance, but my analyses
were also guided by three a priori hypotheses about how context might influence within-
patch composition of wet meadow bird communities. First, I expected that wet meadow
patches located in more complex contexts (adjacent to a greater number of patches or
patch types) would have greater species richness than those adjacent primarily to one
patch type (Forman 1995). Second, 1 expected that wet meadows in human-developed
contexts would have lower species richness than those adjacent to more natural patch
types (Freisen et al. 1995); or, that some species would avoid wet meadows in human-
developed contexts while other, more tolerant species would be more abundant in human-
developed contexts (Blair 1996; Bolger et al. 1997; Bock et al. 1999). Third, I
hypothesized that wet meadows associated with more complex and convoluted coastlines
should have higher species richness than those associated with simpler coastlines (Nilsson
1987)

Compared to terrestrial landscapes, knowledge about the influence of landscape
structure on birds in wetland-dominated landscapes is particularly scarce (see Freemark et
al. 1995 and Weller 1999 for reviews). And, with the exception of riparian forests (e. g.
Gutzwiller and Anderson 1987a & b, Saab 1999), the effects of landscape context on birds
in other wetland landscapes are practically unknown (but see Naugle et al. 1999). The
northern shoreline of Lake Huron where I conducted this research contains some of the
most undeveloped freshwater coastal wetlands remaining in the Great Lakes region; but,
like most wetland landscapes, they are threatened by residential and commercial

development. These wetlands are a conservation priority because they serve as critical

migratory, bret
1992; Prince at
(Brewer et al

birds. however

conservation e1

1 worke
(Mine and Albe
shoreline of La]
counties). Wei
seasonally or 5}:
and vegetation '
szadensis) ant
intersPersed Wit
Shrubs (Salix Sp
floating (P013216

between hummc

migratory, breeding, and foraging habitat for a wide variety of bird species (Prince et al.
1992; Prince and Flegel 1995), many of which are declining regionally and/or locally
(Brewer et al. 1991). But, without knowledge of the effects of landscape context on
birds, however, the effectiveness and ultimate success of management schemes and
conservation efforts in the Great Lakes and other regions are seriously compromised.
METHODS
Study Area

I worked in 40 wet meadows that were part of northern Great Lakes marshes
(Minc and Albert 1998) and located within 500 m of the highly convoluted, northern
shoreline of Lake Huron in the Upper Peninsula of Michigan (Mackinac and Chippewa
counties). Wet meadows selected for sampling met several additional criteria: they were
seasonally or shallowly flooded; water levels were influenced by overall lake water levels;
and vegetation was predominantly a mixture of hummock-forming grasses (Calamagrostis
canadensis) and sedges (Carex stricta and C. aquatilis). The grass/sedge vegetation was
interspersed with varying amounts of bulrush (Scirpus spp.), cattail (T ypha spp.), and
shrubs (Salix spp., Alnus spp., and Myrica spp.). Submersed (Potomogeton spp.) and
floating (Polygonum spp., Lemna spp.) vegetation were often present in standing water
between hummocks.

Bird Censuses

Permanent census transects (Verner 1985) oriented perpendicular to the coastline
were established within each wet meadow. I censused each transect for breeding birds on
multiple occasions (4 times during 1997; 5 times during 1998) between 15 May and 4

July. All censuses were conducted by a solitary individual (SKR) between 0530 and 1030

10

 

EST. Bereteth
randomized T]
that eaCh Wet rr.
Censuses were
temperatures <
Because
tranSCCt quickl}
grasslands (Yer
per hour) Afie
minute count fr
remainder of th:
in wide transeCI
wet meadow. F
distinct in time :
Censuse
birds (Virginia I
Meltin 1993). I
point count. Af
conducted while
along the tran

SCC

Species’ calls am

broadcast.

EST. Before the first census each May, the order in which wet meadows were visited was
randomized. This order was then systematically rotated for each successive census visit so
that each wet meadow was equally censused during different times of the morning.
Censuses were not conducted in unsuitable weather conditions (Robbins, 1981):
temperatures < 0°C, strong winds, steady rainfall, or fog limiting visibility to < 300 m.

Because the hummocks and standing water often made it difficult to walk the
transect quickly or quietly, I modified traditional protocols for line-transect sampling in
grasslands (Vemer 1985). Transects were walked at a slower than typical pace (0.5 km
per hour). After progressing 50 m along the transect, I stopped and conducted a 10-
minute count from that point. After 10 minutes had elapsed, I resumed walking the
remainder of the transect. All birds seen or heard were recorded as either within the 100-
m wide transect (fixed-width), or outside the transect but still within the perimeter of the
wet meadow. Records were based on initial detections of individual birds known to be
distinct in time and space.

Censuses were augmented by broadcasting recorded vocalizations of secretive
birds (Virginia Rail, Sora, Yellow Rail, Least Bittem) to elicit responses (sensu Gibbs and
Melvin 1993). Broadcasts were made during the last 2 minutes of the initial IO-minute
point count. After completing the entire census transect, additional broadcasts were
conducted while returning to the start of the transect to ensure that all individuals present
along the transect were detected. Each 100-m segment received a broadcast of all three
species' calls, and I randomized the order of the species' calls at each consecutive

broadcast.

ll

  
  
 

 

1 Because
richness and ab
habitat character
reached mature
Within each seg.
center of that ses
intervals along e
each bird trams
milhng tech
trees and Sn 8
Pom

3"“ mask 1
SQCfiQnS (‘

the 8 56Ct

estlmate

Habitat Characteristics

Because variation in within-patch habitat characteristics can influence bird species
richness and abundance of wetlands (Craig and Beal 1992; Weller 1999), I measured 20
habitat characteristics along each bird census transect in late July after vegetation had
reached mature height. Each bird census transect was divided into SO-m segments.
Within each segment, four habitat-sampling radii were established radiating from the
center of that segment (Figure 1A). Five sampling points were located at lO-meter
intervals along each of the habitat-sampling radii (total of 20 points per 50-m segment of
each bird transect). At each of these sampling points, I used three different types of
sampling techniques: point-intercept sampling, frequency of cover types, and frequency of
trees and snags (Figure 18).

Point-intercept sampling. At each of the sampling points (Figure 18), I passed a
2-m metal rod vertically through the vegetation. The rod was divided into eight 0.25-m
sections (or strata), and I counted the number of hits, or touches, by vegetation in each of
the 8 sections (sensu Rotenberry and Wiens 1980). Number of hits was used as an
estimate of vegetation density. Grass density was calculated as the total number of hits by
grasses or sedges in all eight 0.25-m strata; woody vegetation density was calculated as
the total number of hits by woody vegetation in all strata; and total vegetation density was
the total number of hits by all vegetation types in all strata. I estimated shrub foliage
diversity by summing the number of 0.25-m sections of metal rod with at least one live
woody hit (8 possible) and the number of sections containing at least one dead woody hit
(up to 8 for a total possible of 16). I also used the metal rod to measure water depth,

grass height, and hummock height at each of the sampling points.

12

 

A

Habitat samplii
radius

 

A

Habitat sampling
radius

/ 25m

 

Bird census transect

 

 

 

 

 

50 meters

 
 
    

x

1 m x l in sampling frame Habitat sampling radius

   
  

K

2.5 m—radius sample plot

   

Point-intercept
sampling point

Figure 1. Diagram of habitat sampling procedures used in 40 Great Lakes coastal

wet meadows associated with the northern shoreline of Lake Huron in Chippewa and
Mackinac counties, Michigan, 1997-1998. A) Locations of 20 habitat sampling points
relation to center of 50-m transect segments. B) Details of 3 sampling techniques used
at each of these 20 points.

13

(Be
standardizt
Sampling P‘
have OCCU”

the valley be
sampling, 1 5
measuremem
and variance 1
Freq”:
cover types, 1 I
points (Figure
vegetation (gra
vegetation, subr
moss. Frequenc
Sampling points
Density g
a 2.5-in radius ci
Snags within the
OCCurrence of de
deciduous and
cc

that Stem tV’pe p
. 1'

lirrl
~ 8 or n
0 Cha
ng

Because hummock formation resulted in a highly variable water depth, I
standardized measurements by sampling between hummocks in the valley nearest to the
sampling point. This was not a random approach, but it eliminated variation that would
have occurred if the rod were positioned at the peak of a hummock at some points and in
the valley between hummocks at others. At a remote site prior to conducting habitat
sampling, 1 sampled seven 50-m sections of wet meadow habitat and determined that 20
measurements per 50m segment were sufficient to acquire stable estimates of the mean
and variance for all variables using the metal rod.

Frequency of cover types. To estimate percent cover of various vegetation and
cover types, I centered a l-m2 sampling frame on each of the point-intercept sampling
points (Figure 1b). Within each frame, I recorded the presence or absence of graminoid
vegetation (grass or sedge), cattail (T ypha spp.), bulrush (Scirpus spp.), floating
vegetation, submersed vegetation, willow (Salix spp.), alder (A [nus spp.), open water, and
moss. Frequency of occurrence of each cover type was calculated as the proportion of the
sampling points (20 per 50—m segment) at which that cover type was present.

Density of trees and snags. To estimate the density of trees and snags, I centered
a 2.5-m radius circular plot on each point-intercept sampling point. All trees, shrubs, and
snags within the plot were identified and counted. I calculated the frequency of
occurrence of deciduous and coniferous stems (> 2.0 m in height) and the frequency of
deciduous and coniferous snags (> 2.0 m in height) as the proportion of circular plots with
that stem type present. Trees and shrubs were sampled in 1997 only, because there was

little or no change in these habitat components between years.

14

Landsca
geographic infor
entire coastal lar
interpreted from
georeferenced, r
as a template for
detailed in Tablfi
Visual ground-tr

The nun
the number of p
interfaces “as c
where the a djac
of each Wet me;
Calculated in Ar
surrounding Ian
surroundjng Ian

Calculation of Landscape Variables

Landscape context variables. Landscape context variables were derived using a
geographic information system (ArcView; ESRI, Inc. 1998). Landuse/landcover of the
entire coastal landscape (including wet meadow sites and adjacent habitats) was
interpreted fi'om color aerial photographs taken in 1992. Individual photographs were
georeferenced, rectified and then mosaiked into a composite color image which was used
as a template for digitizing landcover classes. The landcover classification scheme is
detailed in Table 1. All landcover classifications were verified and updated by extensive
visual ground-truthing during the spring of 1999.

The number of adjacent patches, the number of different adjacent patch types, and
the number of patch interfaces were counted for each wet meadow patch. Number of
interfaces was calculated as the number of discrete sections of a wet meadow perimeter
where the adjacent patch type changed. Patch area, perimeter length, and the proportion
of each wet meadow’s perimeter adjacent to each particular landcover type were
calculated in ArcView. Four other landscape context variables were calculated for the
surrounding landscape (within l-km of the edge of each wet meadow): percent of the
surrounding landscape comprised of Lake Huron, percent of the surrounding landscape
comprised of wet meadow habitat, density of roads (m/ha), and density of streams (m/ha).

Coastline complexity variables. Three measures of coastline complexity were
derived from a vector coverage of the northern Lake Huron shoreline included in the U. S.
Army Corps of Engineers Great Lakes coastal geomorphology database (Great Lakes
Environmental Research Center). Three coastline vector segments representing three

spatial scales -- 500-m, 1-km, and 2-km from the wet meadow perimeter -- were created

15

Table 1. Landcover classification system derived for the northern shoreline of Lake Huron
in Chippewa and Mackinac counties, Michigan.

Urban (Residential, Commercial, and Industrial)
Agricultural (Cropland, Pasture, etc.)
Non-Forested Openings
Forested
Lakes, Reservoirs, or other open water
Forested Wetlands
Inland Emergent Wetlands
Coastal Wetlands:
Bulrush Marsh (Scirpus sp. wetland)
Cattail Marsh (T ypha sp. wetland)
Wet Meadow
Fen Wetland
Sand Flats
Alvar Wetland
Cobble Beaches
Sand Beaches or Sand Dunes

 

l6

 

 

 

cal

frc

p0

al

for each wet meadow using buffer operations in Arc/Info. Measures of complexity were
calculated for each coastline segment separately. Total coastline was calculated directly
from attribute tables. Sinuosity (total coastline divided by straight-line distance from end-
point to end-point) and fi'actal dimension of each coastline segment were calculated using
a Fortran program, Fract_Line (Lam and de Cola 1993).

Statistical Analysis

Habitat, landscape and coastline variables. Because many habitat, landscape, and
coastline variables were inter-correlated, I used principal component analysis (PCA) to
remove correlations and to reduce the number of variables for analysis (Proc Princomp,
SAS Institute 1990). Principal component analysis was conducted three separate times:
once each for the habitat, landscape and coastline variables resulting in three sets of
principal components. Hereafter, these sets of components will be referred to as habitat,
landscape and coastline components. Within each set, components with an eigenvalue >
1.0 were retained. Principal components were not rotated; and eigenvector coefficients,
rather than factor loadings, were used to interpret components because factor loadings do
not conserve the multivariate contributions of the original variables (Rencher 1992).

Bird variables. I analyzed data only for those birds detected within the IOO-m
transects. Species richness was the total number of different species detected during all
census visits. Abundance (an estimate of density) was calculated as the mean number of
birds per hectare of transect per visit. I calculated abundance for all birds (total bird
abundance), all nesting birds, all non-nesting birds, and for individual species that were
detected in > 2/3 of the wet meadows. Because wet meadows possess characteristics of

both grasslands and wetlands, a species was designated in the nesting guild if it was an

17

obligate or facultative nester in either grasslands (V ickery et al. 1999) or wetlands (based
on life-history traits listed in Terres 1980). The non-nesting guild included species that
nest exclusively in other habitats but still rely on wet meadows for other requirements such
as foraging (e. g. Great Blue Heron). Nesting guilds are listed in Appendix A.

Linear and logistic regression models. Because I wanted to test for effects of
landscape context and coastline complexity above and beyond variation in bird variables
that could be explained by area and habitat effects, 1 adopted a two-stage regression
approach (Morrison et al. 1998). In the first stage, a bird variable was regressed against
patch area, perimeter-area ratio and 5 habitat principal components. Then, the residuals
from this model were regressed against the 8 landscape and coastline components. For
species richness variables, abundance variables, and abundance of species that occurred on
> 2/3 of the wet meadows, I used stepwise linear regression for variable selection (Draper
and Smith 1981). For species that occurred on less than 2/3 of the wet meadows, I used
logistic regression (Trexler and Travis 1993) on presence/absence data. Instead of using
residuals in the second stage of logistic regression, the habitat components selected in the
first stage were forced into the model first before using stepwise selection on the
landscape and coastline components (sensu Estandes and Temple 1999).

For each stage of linear and logistic regressions, I limited the number of
explanatory variables to no greater than 8 (at least 5 observations per variable). For linear
regression, variable selection was based on several criteria. First, models containing all
possible subsets of variables selected by stepwise selection (using an F-to-stay = 0.15)
were compared using prediction residuals (Draper and Smith 1981). This served as a

method of validation, and the subset model with the lowest (best) sums of squares of

18

prediction residuals was kept (PRESS: Draper and Smith 1981; Myers 1989). Other
criteria for variable selection required that assumptions of regression (normality of
residuals, constant variance) were satisfied and that all explanatory variables had
significant t-statistics (or < 0.10). For logistic regression, models containing all possible
subsets of variables selected by stepwise selection were evaluated using Aikai's
Information Criteria (AIC), and subset models with the lowest (best) AIC were kept (SAS
Institute 1990). Additional criteria for variable selection required that the model pass a
Hosmer and Lemeshow goodness-of-fit test (Hosmer and Lemeshow 1989), that all
explanatory variables had significant Wald xz-statistics (or < 0.10), and that included
components did not have unusually large coefficients or standard errors. Large
coefficients and SEs indicate that the model is overfit or other problems such as complete
separation or colinearity (Hosmer and Lemeshow 1989). These variable selection criteria
were used for both stages of the regression procedures.

Transformation of variables. Many of the original habitat and landscape context
variables were either count data or percentage data which are traditionally log or arcsine
transformed. The reason for such transformations in the context of principal component
analysis is to linearize the variables so that PCA can construct better linear combinations
(maximized variance) of those variables. However, inspection of biplots of both
transformed and untransformed variables revealed that transformations did not further
linearize relationships between variables, and PCA using transformed variables did not
result in qualitatively different principal components nor did those components explain
more of the total variance in the original variables. Thus, I did not transform original

variables prior to PCA. Similarly, counts of species richness and species’ abundance were

19

not automatically log transformed. Log transformations on dependent variables were used
only as necessary to linearize relationships between dependent and regressor variables,
normalize residuals of regression models, or stabilize error variance (Rahbek 1997).
RESULTS
Bird Species Richness and Abundance
Species richness within wet meadows ranged from 1 to 19in 1997 and from 1 to
16 during 1998. Over a two-year period, I made 2179 individual detections of 55
different species (Appendix A) during a total of 360 census visits to 40 wet meadow sites.
Twenty-five species that potentially nest in wet meadow habitats (Appendix A) comprised
just over 83% of all the detections. The most common species were the Yellow Warbler,
Common Yellowthroat, Red-winged Blackbird, Swamp Sparrow, and Song Sparrow.
Scientific names for all species are provided in Appendix A.
Habitat Characteristics
Although wet meadow sites were selected for inclusion in this study based on
criteria of similar habitat features, the 20 habitat characteristics did vary both among sites
and between years (see Table 2 for summary statistics). Five principal components based
on habitat characteristics were retained for analysis, and they accounted for 74% and 77%
of the total variation in the original habitat variables for 1997 and 1998, respectively
(Appendices B and C). Preliminary analyses indicated that linear and logistic regression of
bird variables using these principal components produced models with r2 values that were
not significantly different from models constructed using the original habitat variables (S.
Riffell, unpublished data). Thus, principal components derived from original habitat

components adequately accounted for habitat-related variance in bird variables.

20

._..>.0.E-0E_. .05 .05 0mcms0 .0... 2.. mwmcm E... .00.. .3 0:03.550... 03000.. mom. 5 003800 .02 ..

 

 

-- -- -- 8.28... .m... 8.. as .8... 88.8..

-- -- -- 8...-8... 8... 8.. e. .8... 820.80

-- -- -- 8.8-8... 8... a. as 82. 808.8...

-- -- *r 8.88... a . . . Sm 0.. 82. .85....50
$05 .05. 80.. no 5:030...

8.2-8... 8... 8... 8.8.8.. 0.... 8.0 0.. s.3).

8. 5-8... 3.... mm... 8.. .8... .w... 8.». as 5...... :80

8.8.8.. 8.. .6... 8.2.8.. 8.. 2.8 as .82

8.8-8... mm. 8.8 8.8-8... 8.. 8... e. 32......

8.2.8.. 8.... m3 887...... .2. 8.8 9.. 8.3.002. 8888

8.8.8.. 2.. .3 8.8.8.. 8.. 8.... as 8.8.802. 8.8.8

8. 3-8... 88 N. .8. 8.8.8.. 8... 2.... as €830

82-8... .2 02 8.88... gm 8.. as 8......

8.2.78.8 8... $8 8.8........... 8.. E .8 as 28.8.0
3%. .0>00 mo 5:030.”—
888... ...... mm... 8.2-8... 8... 8... 6.8.... p.23... 08...... 8:...
8.2.2.. 8... 8.0 8.2.8.. .8... 8.... .582... 8.25.. 8.8.82. .28
v... .8... 8... mm... 8. .8... 8... 8... .582... .88.. 8.8.002. 883
8.2.8.... 8... z... 8.2-8... .8... new 8.88.... 5.2.0.. new
382.0. .8 8... 8.8. No: 78.8 R... 8.8. .5 2.0.2. new
5.8-8... 8.. .08 8.8.8.. 8.. 8.8 .5 50...... v.8552:
32-8... 8... 8.. 8.8.8.. 8.. o. .8 .8 58.. .033

0w=nx mm 5.02 08.3. Mm .802 3...: 03....3/ .333.—
wmo. boo.

 

woo—-30— .:meE—z £09500 005x022 .05. 030.320
5 :03: 00—0120 0.50.9.0 E055: 0.... wee—0 0320008 .03 ov E 00530:. 0030...... .350; .8 02.2.0... Eafifism .N 030,—.

21

Habitat components were biologically interpreted with respect to the contributions
of the original variables (Table 3). Habitat components selected in two-stage regressions
of bird variables were consistent with documented bird-habitat relationships in grasslands
and wetlands (Weller 1999). However, because the focus of this study was bird-landscape
relationships, bird-habitat relationships will not be discussed herein. Readers interested in
further information about bird-habitat relationships in northern Lake Huron coastal wet
meadows are referred to Table 3 and Appendices B and C where information about these
components is given.

Landscape Components

The landscape surroundings of the 40 wet meadows varied from simple contexts
(patch perimeter bordered by only 11 patch) to very complex contexts (patch perimeter
bordered by as many as 11 different patches). Some wet meadows were bordered entirely
by forest, others were bordered primarily by various wetland types, and still others by
open water. Although the proportion of the wet meadow perimeter adjacent to urban
patches ranged from 0 - 62%, less than 10% of most wet meadow perimeters were
adjacent to urban patches. Additionally, most urban patches consisted of single-family
residences, summer cabins, or boat dock and marinas which represented only a slight to
moderate human-development of the landscape. The extensive agricultural, industrial, or
commercial development that is characteristic of more populated regions of the Great
Lakes were uncommon or lacking in this landscape. Summary statistics for landscape
context variables are listed in Table 4. Principal component analysis identified 5 landscape
context components (LANDl-LANDS) which comprised 78% of the total variance in the

16 original variables (Appendix D). Each of these components was comprised of a suite

22

Table 3. Habitat components derived from principal components analysis on original
habitat variables.

HABI

HABZ:

HAB3.’

HAB4:

HAB5

HABI :

HABZ:

1997 Habitat Components
Grass Density and Height

Increasing hummock height, grass height & density, total vegetation
density

Trees vs. Deep Water and Bulrush

Increasing fi'equency of deciduous trees
Decreasing water depth and frequency of bulrush

Woody Vegetation

Increasing woody density, shrub foliage diversity, frequency of willow &
alder

Forested (deciduous) vs. Graminoid Coverage

Increasing frequency of deciduous trees and snags
Decreasing graminoid coverage

Open Water and Floating Vegetation vs. Willow

Increasing frequency of moss, open water pockets, and floating vegetation
Decreasing frequency of willow

I998 Habitat Components
Grass Density and Height

Increasing hummock height, grass height & density, total veg. density
Decreasing frequency of bulrush

F orested/shrubby vs Deep Water

Increasing woody, frequency of willow, trees, and snags
Decreasing water depth

 

23

Table 3. Continued.

HAB3: Shrubby, Open vs Graminoid C overager
Increasing shrub foliage diversity, frequency of open water and deciduous

snags
Decreasing graminoid coverage

HAB4: Willowy vs Snags and Cattail

Increasing frequency of willow
Decreasing deciduous snags and cattail

HAB3: Open Water and Floating Vegetation vs. Willow
Increasing frequency of floating vegetation, open water, moss, and

coniferous trees
Decreasing frequency of willow

 

24

 

 

Ima—

Si

Table 4. Summary statistics for landscape context and coastline complexity variables
measured for 40 wet meadows along the northern shoreline of Lake Huron in Chippewa
and Mackinac counties, Michigan, 1997-1998.

 

 

Variable Units Mean SE Range
Landscape Context Variables
# of adjacent patches #/patch 4.08 0.37 1.00-1 1.00
# of adjacent patch types #/patch 3 .25 0.20 1.00-6.00
# of interfaces with other patches #/patch 4.93 0.56 1 .00-17 .00
Percent of patch perimeter adjacent
to:
Urban % 8.55 2.39 000-6240
Non-forested opening % 1.12 0.47 0.00-12.10
Forest % 51.17 4.18 0.00-100.00
Open water % 6.20 2.56 000-7650
Forested wetland °/o 9.13 2.42 0.00-6800
Bulrush marsh % 16.19 2.81 000-5830
Cattail marsh % 6.12 1.87 0.00-46. 10
Total wetland % 32.31 3.76 000-9694
Surrounding landscape (within 1-
km of patch perimeter):
Amount of wet meadow % 3 .70 0.48 0.00-11.00
Amount ofLake Huron % 31.35 2.71 470-7100
Stream density m/ha 4.90 0.30 030-1037
Road density m/ha 12.04 1.10 0.10-3293

Coastline Complexity Variables
Total coastline

500-m m/ha 6.15 0.40 030-1218

l-km m/ha 13.35 0.85 4.41-33.82

2-km m/ha 33.87 2.33 11.85-79.33
Sinuosity

500-m --* 2.84 0.25 1.08-8.35

l-km -- 2.83 0.23 1.06-7.57

2-km -- 2. 82 0.20 1.06-6.99
Fractal dimension

SOO-m --** 1.17 0.02 1.01-1.42

l-km -- 1.16 0.01 1.01-1.39

2-km -- 1.15 0.01 1.01-1.28

 

* Sinuosity is the ratio of total length (total coastline) divided by the distance between the
two endpoints. Sinuosity of 1.00 indicates a straight line.
** Fractal dimension is unitless.

25

of intercorrelated, and often conceptually related, original variables which permitted me to
interpret and name (sensu Saab 1999) each component (see Table 5).

The first component (LAND 1) represented a contrast of complex contexts vs.
simple, forested contexts. High values of LAND] were associated with wet meadows that
were adjacent to a large number of patches and patch types and that had proportionally
more perimeter bordered by wetlands, and low values of LAND] were associated with
patches primarily surrounded by forest habitat. The second component (LAND2) was a
contrast between Lake Huron and stream contexts vs. forested wetland contexts. High
values of LAND2 were associated with wet meadows where a relatively high proportion
of the surrounding landscape was comprised of Lake Huron and also contained relatively
high stream densities. Low values of LAND2 indicated that a relatively high proportion of
wet meadow perimeters was bordered by forested wetland.

The third principal component (LAND3) represented a gradient of increasingly
human-developed contexts. High values of LAND3 were associated with relatively higher
proportions of urban and cattail marsh perimeters as well as greater density of roads
within the surrounding landscape. Anthropogenic effects often facilitate the development
and spread of cattails in wetlands (Newman et al. 1998), so it is not surprising that cattail
perimeters were correlated with other measures of human impact. The fourth principal
component (LAND4) was a contrast of wetland contexts vs. open water contexts. High
values of LAND4 were associated with wet meadow perimeters bordered by relatively
higher proportions of bulrush marsh and all wetland types as a group, while low values
indicated wet meadow perimeters were bordered by relatively higher proportions of open

water and were less connected to other wetlands. The last component (LANDS),

26

Table 5. Landscape components and coastline components derived from principal
components analysis on original variables.

[AND]

LAND2:

LAND3:

LAND4:

IANDS

COAST]:

Landscape Context Components
Complex Contexts vs. Simple. Forested Contexts

Increasing number of patches, patch types, and total wetland perimeter
Decreasing forested perimeter

Lake and Stream Contexts vs. Forested Wetland Contexts

Increasing proportion of Lake Huron and density of streams in the
surrounding landscape

Decreasing forested wetland perimeter.

H uman-developed Contexts

Increasing Urban and Cattail Perimeter and Road Density

Wetland Contexts vs. Open Water Contexts

Increasing bulrush marsh and total wetland perimeter
Decreasing open water perimeter

Forested Contexts vs. Aquatic Contexts
Increasing forested perimeter
Decreasing open water and wetland perimeters; increasing proportion
of Lake Huron in the surrounding landscape
Coastline Complexity Components

Highly Fractal and Complex Coastlines

Increasing fractal dimension, Sinuosity and total coastline length at all
three spatial scales

 

27

Table 5. Continued.

C OA S T2: Increasing Total Coastline
Increasing length of total coastline (all scales)
C OAS T 3 : Scale Contrast / Sinuosity

Increasing Sinuosity and total coastline (l-km and 2-km scales)
Decreasing fractal dimension, Sinuosity and total coastline (500-113

 

28

 

 

me

represented a contrast of terrestrial, forested contexts vs. aquatic contexts. High values
of LANDS were associated with wet meadows that were primarily adjacent to forested
habitats while low values of LANDS were associated with wet meadows adjacent to
bulrush marshes and other wetlands and/or surrounded by a greater proportion of open
water.
Coastline Components

Three measures of coastline complexity were calculated at three spatial scales for
coastline segments associated with each wet meadow. These segments varied from
simple, nearly straight coastlines to highly convoluted and complex coastlines (Table 4).
Complexity measures were highly inter-correlated and principal components analysis
identified three components which contained 77% of the variance in the original variables
(Appendix E and Table 5). The first coastline component (COASTI) represented highly
fractal and complex coastlines at all scales. Large values of COASTI indicated wet
meadows that were associated with highly fractal coastline segments at all three scales.
These segments also tended to be highly sinuous with high amounts of total coastline.
High values of the second component (COASTZ) indicated wet meadows associated with
coastline segments with high amounts of total coastline length at all three spatial scales.
The third component (COAST3) represented a scale contrast. High values of COAST3
indicated wet meadows with high values of total coastline and Sinuosity for associated
coastline segments at scales of l-km and 2-km, but low values of all three coastline

metrics for segments at the scale of 500-km.

29

an:
and
loca
(LL
1C 0,:

WHEN

abundaj

Bird-Landscape Relationships

For 22 of the 26 species and bird variables tested, landscape context or coastline
components significantly explained variation above and beyond that explained by models
containing area, PAR and habitat components during at least 1-yr of the study. The
improvement in variance explained (r2) ranged from 5% to 40% (Table 6). Because of the
large number of models involving community- or species-level bird variables, results will
be summarized in two ways. First, I will summarize the landscape and coastline
components which were selected in regression models of community level variables.
Models involving total species richness and total abundance will be described first,
followed by models involving nesting and non-nesting guild variables. Second, I will
summarize results of species-specific models by each landscape and coastline component
in turn.

Total species richness and total abundance. Area, perimeter-area ratio (PAR) and
habitat componentsalone explained 63% and 64% of the variance in species richness in
1997 and 1998, respectively (Table 6) in the first stage of linear regressions. Landscape
and coastline components significantly increased the variance explained to 75% in 1997
and 78% in 1998. In both years, higher species richness was observed in wet meadows
located in more complex contexts (LANDl) and more human-developed contexts
(LAND3). High species richness was also associated with less fiactal coastlines
(COASTI: Table 6) in 1997, and with less sinuous coastlines (COAST3) and aquatic
contexts (LANDS) in 1998.

Habitat components explained 38% and 10% of the variance in total bird

abundance in 1997 and 1998, respectively. Addition of landscape and coastline

30

.. E: 5.3153
.ccruILfr\ph —v—-flw fix

if};
A... :4. 3.2.0.5.: NU LC «4.
3:39 01:92.35: LC v.

» ..:.tu.:.\.o.;\< .mottzero
33330.: ~03 D? Em Evie >33? .20.: uvgkc 33%....2
3:0 22:1..C £02....T:.:~:>.L2 .~Cv..~L.~l-CC...» .0 0~£fih

.0634... m.
.0920 3:... m3

 

3 £3.00

 

R: .3 325 3 825 3 825 2 .o 3 mm: .3 52 355:5: mo: 52
m2 3 2300 .3 325 .3 525 moo 3 mm: .35: 355:5: mo: 8%
00:59.27: 35 3:82
3 .5300 3 mm:
:3 .3 RES 3 82:... .3 825 $5 3 mm: .3 82 355:5: we: 52
A3 vm:
3 mm: 3 a:
5.0 3 525 2.: .3 mi .3 52 358:5: mo: 82
30:58 860% wccmoz
3 2.300
one .3 323 .3 872.. .3 .025 2.: 3 mm: 355:5: mo: 52
$3 XE
one 3 E72: :3 .3 mm: .3 E: 355:5: mo: 82
oocmvcnndx thm 730,—.
3 $300 3 8:
who .3 325 3 825 .3 572.. 5.0 .3 mm: .3 52 355:5: we: 52
TL :93 A3 mmm
m3 3 £300 3 825 .3 .025 So .3 M2: 3 52 355:5: we: 32
305.3: 860% .80.:
um 3:0:0QE00 09:03::— :::0c_:w_m um 3.0.: E :0: .30 m. 033:? Em

 

35:09:00 09335: 36205 20:02

20:08 35385::

 

033:3 EB 05 :5 30:09:00 SE 5030: 3:20:20: 05 .30 80:00:: 05 8:50:05 3 :0 A3 .::m££2 £055.00

0:50—02). :5 «30:5:0 E :0SI 3:..— .«0 05.80:: 805.8: 05 5:5 “08:68:: 250308 :03 o: :_ £5: 30:86 :03 .«0 20:08
532%: 033-025 5 3:83:80 0:25:00 :5 3883 0:58:52 .30 30:30 :5 30.450 55:: 5.88-555, .30 :0mtaqfioo 0 03:...

31

     

F‘ cu fllw‘

2:7:CZ
1|: ,..l\ (at IV: \.\
1.53:. >.:C

3c ~..E.m>. 35$ :

 

5:022:23. » C .3315

 

 

$5 3 37:3 .3 5:23 5.: 3 E: .3 5:. 55.
-- -- N00 3 .0: .3 5:. 82
30:QO m:0m
$3 vm:
m3 3 825 N20 .3 E: .3 5a 5.5.
m: 3 £300 .3 325 :5 3 E: .3 52 82
30:55 9::3m
one 3 05.5.00 .3 325 3.0 3 mm: 52
$0 3 .5300 n _ .o 3 mm: 82
HMO.£§O__0> COEEOU
35:23.05.
.0. mo: 005.0592
-- -- a. .0 3 mm: .3 mm: 52
-- -- 2.0 3 cm: .3 S: 82
00:20:92 35 w:cm0:-:0z
two 3 525 :5 3 mm: .3 mm: 55.
-- -- 3.0 3 E: .3 .52 82
mmoccomm
8.00am w:.um0:-:0Z
NM 85:09:00 09:83:. “:00u.:w.m my. .3008 :. mom ..0 .4 0.3.8:, .85

 

 

85:09:00 0958.050. w:..0:.0:. 20.003.

20.008 3:030:00...

555550 5 25:.

32

30.00:: 3:312:50:

 

:.:.::.:..v C U~C=.~.

 

 

 

 

8.0 3 200.00 m _ .0 382 000.
3 mm:
3.0 3 $300 00.0 .3 E: .3 82 002
30m
30 3 .023 R0 3 82 000.
TL vm:
-- -- 2.0 .3 E: .3 82 002
:0: 250.5
2.0 3 Sm<00 R0 3 mm: .3 82 000.
~00 3 .0300 00.0 3 E: 000.
03:02
3 mm:
-- -- R0 .3 :5 .3 8% 000—
3.0 3 2.300 2.0 3 E: .3 8.3 82
. 5035 533:2
R0 3 825 2 .0 3 mm: 002
0.00 3 E0400 .3 525 -- -- 82
5:01 02m 386
oocomn<\oo:0mo.£
my. 35:09:00 09300::— EmoE:m_m NM 600:: :m mom .30 m 030:9, 95
35:09:00 0900005. 36205 0.0002 0.009: 35-08301

 

0800000 .0 2,3.

33

;.-..q~;saa. :—:;..I~—n..—n‘:~.&

3.022.223. »

 

C Uttflh

 

 

-- -- 0 _ .0 3 mm: .3 mm: 000.

00.0 3 33000 2.0 3 00: .3 8:0 000.
mEBx03 0.0000

00.0 3 00720 00.0 3 82 32
50>? 0w00m

00.0 3 83000 .3 002.3 :0 3 mm: .3 mm: 000.

-- -- 2.0 3 mm: 000.
5:3003m 502

00.0 3 05.000 .3 325 ~00 3 00: .3 82 002
EEwEv. 503$

-- -- 00.0 3 mm: .3 E: 002

-- -- 0 _ .0 3 00: 000.
50.009503 03300

-- -- 0_ .0 3 002 32
503: 505002

-- -- -- -- moi

-- -- 2. -- 000—
0&:m :08800
E 35:09:00 0900005. 0:00£:w_m mm .0008 E mom .30 u 030:? gm

 

35:09:00 0900305. m:=5_0:_ 20:02

30008 3:030:50:

 

00000000 .0 200...

34

w;

       
   

. t...:~:..\#v >p0~\hvsf
. . . 1:...2 xx ., ,. .
3030: 5:33...

0300.015..th
:22 0 /

 
  

\v.d~:~.~u:nv.v mb U\A-~.~.

 

 

N. .0 3 .0300 -- -- 30.
300.0% 3.9.0.0
-- -- 000 3 03. .3 82 000.
300.8% :0::0>0m
3 0m... .3 mm...
00.0 3 $3.00 00.0 .3 .3. .3 002 000.
.. nu _ —.o A+V no.2. hoo—
;0:...0.00 50.58....
3 0m:
.00 3 0023 .3 57.5 00.0 .3 0m... .3 N3. 000.
00.0 3 .020... 2.0 3 .0: .3 82 .00.
00.8.8.0 0000.303.
3 03.
00.0 3 0:30.00 .3 0025 .3 .023 00.0 .3 00.. .3 mm... 000.
-- -- 00.0 3 0m... 3 mm... 000.
.0553 30:0.»
Ly. 35:09:00 0.50005. .:00....:w.m NM. .0008 :. mon. 00 .0 0.0.0.0? 00m.

 

35:00:50 0900005. w:.0:.0:. 0.0003.

0.000.: 3:03.350...

 

00000000 .0 0.00.0

35

components significantly increased the variance explained to 50% in both years (Table 6).
Total abundance was positively associated with complex contexts (LANDI) in both years.
In 1998, total abundance was also associated with lake and stream contexts (LAND2),
aquatic contexts (LANDS) and less total coastline length (COASTZ).

Nesting and non-nesting guilds. Area, PAR, and habitat components explained
75% and 69% of the variation in nesting species richness in 1997 and 1998, respectively
(Table 6). Inclusion of landscape and coastline components significantly increased the
explained variance to 84% in 1997 and 81% in 1998. In 1997, higher nesting species
richness was associated with complex contexts (LANDI). In 1998, higher nesting species
richness was associated with lake and stream contexts (LAND2), human-developed
contexts (LAND3), aquatic contexts (LANDS) and less fractal coastlines (COASTI).
Components selected in models for nesting bird abundance generally mirrored those
selected in richness models (Table 6).

Area and habitat components explained approximately 30% of the variation in non-
nesting species richness and abundance. Landscape and coastline components did not
significantly increase the variance explained for any non-nesting guild variable with the
exception of non-nesting species richness which was associated with complex contexts
(LANDI) in 1998 (Table 6).

Community and species-specific responses. Landscape and! or coastline
components significantly explained additional variation above and beyond that explained
by area, perimeter, and habitat components for most of the species during at least one year
(Table 6) with the exception of Common Snipe, Northern Flicker, and Downy

Woodpecker. Landscape and coastline components selected in models were highly

36

inconsistent between years (e. g. Common Yellowthroat related to COASTl in 1997 and
LANDS and COAST2 in 1998; see Table 6), but were highly consistent with regard to the
direction of relationships observed for particular principal components. Thus, both
community-level and species-level results of models will be summarized by component
(see Table 7) in the following text.

LAND] .' Complex contexts vs. simple, forested perimeters. When selected in
regression models, this component was always positively associated with bird variables.
No negative associations with this component were observed. In addition to species
richness and total bird abundance (both years) and nesting bird species richness and
abundance (1997), the presence/absence of 4 bird species (Great Blue Heron, Virginia
Rail, Yellow Warbler, and Red-winged Blackbird) were positively associated to LAND]
during one or both years (Tables 6 & 7). Compared to wet meadows primarily
surrounded by forest, these 4 species were more likely to be detected in wet meadows in
complex contexts (adjacent to a greater number and diversity of patches).

LAND2: Lake Huron and stream contexts vs. forested wetland contexts. This
component was related to four bird variables (Table 7). In 1997, Swamp Sparrow
abundance was negatively related to this component. In contrast, nesting bird species
richness and abundance of nesting and total birds in 1998 were positively associated with
this component.

LAND3: Human-developed contexts. Five bird variables were associated with this
variable (Table 7). Total species richness was positively associated with human-developed
contexts during both years, and thus with wet meadows which had relatively more

adjacent urban landuse, adjacent cattail marsh, and more roads in the surrounding

37

Table 7. Results of regression modeling summarized by principal component. Models for
each of the listed bird variables contained that component as a significant predictor (P <
0.10). (+) indicates landscape or coastline characteristics that are positively associated
with that component; (-) indicates characteristics that are negatively associated.

Positive Relationships Negative Relationships

 

LAND! .' Complex Contexts (+) vs. Simple, Forested Contexts (-)

Total Species Richness 1997 & 1998 None
Total Abundance 1997 & 1998

Nesting Bird Species Richness 1997

Nesting Bird Abundance 1997

Non-nesting Bird Species Richness 1998

Great Blue Heron 1997

Virginia Rail 1998

Yellow Warbler 1998

Red-winged Blackbird 1997 & 1998

LANDZ: Lake and Stream Contexts (+) vs. Forested Wetland Contexts (-)

Total Abundance 1998 Swamp Sparrow 1997
Nesting Bird Species Richness 1998
Nesting Bird Abundance 1998

LAND3: H uman-developed Contexts (+)
Total Species Richness 1997 & 1998 Great Blue Heron 1998
Nesting Bird Species Richness 1998
Nesting Bird Abundance 1998

LAND4: Wetland Contexts (+) vs. Open Water Contexts (-)

Alder Flycatcher 1998 Song Sparrow 1998

Sedge Wren 1998
Red-winged Blackbird 1998

 

38

Table 7. Continued.

Positive Relationships Negative Relationships

 

LANDS: Forested Contexts (+) vs. Aquatic Contexts (-)

NONE Total Species Richness 1998
Total Abundance 1998
Nesting Bird Species Richness 1998
Nesting Bird Abundance 1997 & 1998
Eastern Kingbird 1997
Yellow Warbler 1998
Common Yellowthroat 1998
Swamp Sparrow 1998
Song Sparrow 1998

C OAS T1 .' Highly Fractal and Complex Coastlines (+)

Mallard 1997 Total Species Richness 1997

Cedar Waxwing 1998 Nesting Species Richness 1998
Nesting Bird Abundance 1998
Great Blue Heron 1997
Common Yellowthroat 1997
Swamp Sparrow 1997
Chipping Sparrow 1997

C 0A S T 2 .° Increasing Total Coastline (+)

Mallard 1998 Total Bird Abundance 1998
Nesting Bird Abundance 1997
American Bittern 1997
American Goldfinch 1998
Common Yellowthroat 1998

C CA S T3 : Scale Contrast / Sinuosity (+)

Sora 1997 Total Species Richness 1998
Eastern Kingbird 1997 Sora 1998
Cedar Waxwing 1997 Alder Flycatcher 1998

Yellow Warbler 1998

 

39

landscape. In 1998, nesting bird species richness and abundance were positively
associated with human-developed contexts. In contrast, presence/absence of Great Blue
Heron (1998) was negatively associated with human-developed contexts.

LAND4: Wetland contexts vs. open water contexts. Relationships between bird
variables and LAND4 were predominantly positive relationships (Table 7). During 1998,
Alder Flycatcher, Sedge Wren, and Red-winged Blackbird were all more likely to be
detected in wet meadows adjacent to bulrush marsh or other wetland types. In contrast,
higher Song Sparrow abundance (1998) was observed in wet meadows bordered by a
relatively higher proportion of open water.

LANDS: Terrestrial, forested contexts vs. aquatic contexts). This component was
negatively associated with a total of 10 bird variables, and no positive relationships were
observed (Table 7). In addition to total species richness and abundance and nesting bird
species richness and abundance, this component was negatively related to five species:
Common Yellowthroat (I 998), Swamp Sparrow (1998), Song Sparrow (1998), Eastern
Kingbird (1997) and Yellow Warbler (1998). These species were more abundant or more
likely to be detected in wet meadows bordered by higher proportions of aquatic habitats
relative to terrestrial habitats.

C OAS T1 : Highly fiactal and complex coastlines. Significant relationships
between COAST] and bird variables were predominantly negative (Table 7). In addition
to total species richness (1997) and nesting bird species richness and abundance (1998),
Common Yellowthroat (1997), Swamp Sparrow (1997), Great Blue Heron (1997) and
Chipping Sparrow (1997) were less abundant or less likely to be detected in wet meadows

associated with highly fractal or complex coastlines. In contrast, the presence of Mallard

4O

(1997) was positively associated with highly fractal coastline segments (high values of
COASTI).

C OAS T 2: Total coastline. Significant relationships between COAST2 and bird
variables were also predominantly negative (Table 7). Total bird abundance (1998),
nesting bird abundance (1998), Common Yellowthroat abundance (1998) and the
probability of detecting American Bittern (1997) and American Goldfinch (1998) were
negatively associated with the total amount of coastline within the surrounding landscape.
Conversely, the probability of detecting Mallard (1998) positively associated with the total
amount of coastline surrounding wet meadows (COAST2).

C OAS T 3 : High total coastline and sinuosity at I-km/Z-km scale vs. complex
coastlines at 5 00-m scale. Relationships between COAST3 and bird variables were mixed
(Table 7). This component was positively related to the probability of detecting Sora
(1997), Eastern Kingbird (1997), and Cedar Waxwing (1997), but was negatively
associated with the probability of detecting Sora (I998), Alder Flycatcher ( I 998) and
Yellow Warbler (1998) and total species richness (1998).

DISCUSSION
Effects of Area and Perimeter

Much previous work has focused on the importance of patch area as a predictor
for species richness in both wetlands (Brown and Dinsmore 1991) and grasslands (Herkert
I994; Vickery et al. 1994; Helzer and Jelinski 1999). Most recently, Helzer and Jelinski
(1999) demonstrated that the perimeter-to-area ratio (PAR) may be a more apprOpriate
predictor of bird species richness and occurrence in wet grasslands than area, but fiirther

research is needed to confirm their findings. Because I included both area and PAR in the

4]

first stage of regressions, my results can help identify the relative importance of these two
related patch features to the conservation of wetland birds in Great Lakes coastal
wetlands. Although not the primary focus of this paper, the relevance of this information
to conservation planning warrants further discussion.

In contrast to what Helzer and Jelinski (1999) observed, area occurred more
frequently in linear and logistic regression models than did PAR suggesting that area is
better predictor (Table 6) for Great Lakes coastal wet meadows. Total species richness,
nesting birds, non-nesting birds, and 12 species were positively associated with area:
American Bittern, Mallard, Virginia Rail, Sora, Northern Flicker, Eastern Kingbird, Sedge
Wren, Cedar Waxwing, Red-winged Blackbird, American Goldfinch, Savannah Sparrow,
and Swamp Sparrow. In contrast, perimeter-to-area ratio (PAR) occurred in models for
only 4 variables. Species richness in 1997 was negatively associated with PAR indicating
that patch shape (amount of edge relative to area) explained a significant amount of
variation in species’richness above and beyond habitat and area. Swamp Sparrow were
negatively associated with PAR during one year, but positively associated with area in the
other. PAR was a positive predictor for the abundance of Song Sparrow, a known edge
denizen. These results underscore the potential importance of perimeter-to-area ratios for
patch selection in some species of wetland and grassland birds. But, they also indicate that
teasing apart the separate effects of area and PAR may be difficult and that area may still
be the better predictor for many grassland/wetland bird communities. Based on my

results, it would be premature to dismiss the importance of area in favor of PAR.

42

Importance of Landscape Features

Landscape and coastline components significantly increased the variation explained
in the large majority of models. Because I used a two-stage regression approach, efi‘ects
of landscape context and coastline complexity described herein represent true landscape-
efl‘ects, and not merely effects due to correlations with area, perimeter or other habitat
characteristics. Thus, I can confidently assert that landscape context, in the sense of
characteristics of the adjacent landscape and surrounding coastline, can explain variation in
bird richness and species' distributions in Great Lakes coastal wet meadows that is not
explained by area and habitat characteristics.

Overall, landscape variables selected in linear and regression models for a
particular bird variable were not consistent from year-to-year. This is a common problem
with habitat and landscape studies (e. g. Brown and Dinsmore 1991; Villard et al. 1999).
One cause is that birds may select habitats based on a suite of characteristics. at a variety of
spatial scales (Pearson 1993; Freemark et al. 1995; Weller 1999) either in concert or in a
temporal and/or spatial hierarchy (sensu Saab 1995). This problem is exacerbated in
Great Lakes coastal habitats because patch-level characteristics (e.g. water depth or
duration of flooding) can and often do fluctuate widely from year-to-year (Bedford 1992).
In such dynamic systems, birds may use different landscape cues in one year compared to
the next based on changes in underlying habitat conditions at the patch scale. Another
factor contributing to these inconsistencies are characteristics of the statistical methods
themselves. Stepwise regression techniques only recover the combinations of independent
variables that best explain the variance in bird variables. As with most other multivariate

techniques (e. g. ordination, discriminant analysis, etc), there is no assurance that these

43

variables have direct causal relationships to bird variables. In situations where the
organism of interest is likely to use multiple characteristics as selection criteria, one
variable (e. g. grass density) may be selected by regression techniques in one year; and
another variable (e. g. adjacent forest) may be selected the next year. But in reality, the
bird may be using a combination of the two factors measured by researchers during both
years. In light of low annual consistency in regression models, species-specific
associations with landscape and coastline components should be used to generate
hypotheses only and, although they may possibly represent causal relationships, it would
be unwarranted to construe them as such.

However, more robust conclusions about the general effects of landscape context
on birds are possible. The relationships between landscape and coastline components and
bird variables were consistent such that sign (direction) and magnitude of the regression
coefficients were remarkably similar across difi‘erent bird variables (see Table 6). For
example, in every instance where complex contexts (LANDI) was selected as an
explanatory variable, it was positively associated with that bird variable, and no negative
associations were observed (Table 7). Additionally, bird/component relationships were
usually readily explained by existing knowledge and/or theories of landscape ecology.
These properties of the responses I observed lend credence to my conclusions and indicate
that these relationships are not spurious results of the modeling process. Rather they
likely represent at least generalizable, predictive relationships, and potentially causal
mechanisms for some species. Following, 1 highlight four general patterns of bird

response to landscape context which were consistent across bird variables.

44

Complex contexts. My first hypothesis was that the species richness of a patch
should increase with the number of adjacent patches and patch types (Forman 1995).
Increased species richness of wet meadow birds was positively associated with a principal
component (LANDI) representing a gradient of increasing number of adjacent patches
and patch types, or diverse landscape contexts (Table 6), and this association occurred in
both years. One could posit that larger patches tend to have more adjacent patches and
patch types simply because they are larger or have more perimeter, and that the effects of
complex contexts are in reality effects of area and/or perimeter. However, this is highly
unlikely because the two-stage regressions removed effects of both area and perimeter in
the first stage.

Forman (1995) first predicted this relationship, and suggested that different
adjacent patch types each have unique species pools from which they contribute colonizers
to the species richness pool of the patch in question. Although my results are consistent
with this mechanism, I also observed significant relationships between the number of
adjacent patches and the abundance and occurrence of several species (Table 7). Thus this
colonization mechanism, while a likely possibility, is not by itself sufficient to account for
the patterns observed in Great Lakes coastal wetlands. A second alternative mechanism is
that a greater number of adjacent patches provide a more diverse, more stable or more
abundant suite of food resources for birds located immediately adjacent to breeding
habitat. Many wetland bird species depend on adjacent habitats for foraging (e. g. Red-
winged Blackbirds [Orians 1980]; see also Szaro and Jackle 1984), and a diversity of
resources (i.e. diversity of adjacent patch types) would also stabilize available food

resources in highly variable systems (Skagen and Knopf 1994) like Great Lakes coastal

45

wetlands. This would increase the intrinsic habitat quality of wet meadows located in
complex contexts, allow smaller territories (Anderson and Titman 1992) and, hence,
greater abundance of birds within the same patch (Gill 1995).

Human-developed Contexts. In contrast to my original hypothesis that species
richness would decrease in increasingly developed contexts, total species richness during
both years and nesting birds in 1998 were higher in human-development contexts (Table
7). A possible reason is that even the most urbanized wet meadows in the region are only
developed to an intermediate degree. Thus, they may not be degraded enough to exclude
wet meadow breeders (such as American Bittern), but the developed context may attract
additional species (e. g. American Robin sensu Blair 1996) such that species richness is
highest at moderate levels of impact. Low to moderate levels of human impacts can also
repel potential predators (sensu Bowers and Breland 1996), such as the Great Blue Heron
(see next paragraph) which could, in turn, make those wet meadows more attractive to
breeding birds. More research is necessary to determine if such patterns consistently
occur in moderately-developed landscape contexts.

The probability of detecting Great Blue Heron was lower in wet meadows located
in more developed contexts, but this was the only species or bird variable that was
negatively related to developed contexts. Great Blue Herons nest in forest habitats that
are disjunct from the wet meadows I studied, but did use wet meadows regularly for
foraging (S. K. Riffell, personal observations). Hence, human-developed contexts may
exclude herons or other wading birds from valuable foraging territories, but this too,

warrants further investigation.

46

Because negative effects of human-developed contexts were largely absent in these
wet meadows, urban impacts do not appear to be a primary factor in determining habitat
use in this region. However, I warn that potentially negative effects of urbanization
should not be ignored. Negative responses by birds to development-related impacts have
been well-documented in other regions and habitats (F reisen et al. 1995; Blair 1996;
Bolger et al. 1997), and the lack of a relationship I observed could reflect the generally
low levels of development in the northern Lake Huron region. Also, I did not measure
breeding success, so it is not known whether human-developed contexts impact breeding
success in these wet meadows. Impacts on breeding success would not necessarily be
apparent in richness or presence/absence data because the abundance of relatively
unimpacted wet meadow habitat (source habitat) in the region could be supplying many
surplus breeders. These surplus breeders would be relegated to patches in developed
contexts and maintain high abundance in these impacted sites (sink habitats) despite the
lower breeding success therein (Pulliarn 1988).

Wetland contexts. Although I did not theorize a priori that wetland contexts
(connected, adjacent wetlands) would be an important predictor of bird distributions in
Great Lakes coastal wetlands, the combined responses to landscape components 2, 4, and
5 suggest that wetland connectivity may indeed be very important for birds. Three
species, Alder Flycatcher, Sedge Wren, and Red-winged Blackbird, were associated with
wetland contexts compared to open water contexts (LAND4; Table 7). Similarly, l4 bird
variables (5 species) were associated with aquatic contexts compared to terrestrial,

forested contexts (LANDS; Table 7).

47

Wetland and/or aquatic contexts may confer advantages to birds breeding in wet
meadows in two major ways. First, adjacent wetlands may improve the foraging resources
available to birds. Red-winged Blackbirds are known to use adjacent habitat, rather than
the breeding patch, for foraging (Orians 1980) and likely other species do as well (Szaro
and Jackle 1984). Adjacent wetlands and aquatic habitats may be a more productive
source of plants, insects and other invertebrates that are important food resources for
breeding birds compared to terrestrial forests (Vemer and Willson 1966; Tilton and
Schwegler 1978; Mitsch and Gosselink 1993) or open water habitats (Tilton and
Schwegler 1978; Gathman et al. 1998). Emergence of invertebrates is ofien synchronized
such that each species emerges en masse over a short time-period (few days to a few
weeks: Orians 1980), and a particular species would only be available to breeding birds
for a short time. Wetland contexts with a large variety of adjacent wetland types would
contain more invertebrate species and, hence, a more constant supply of emerging
invertebrates. Also, invertebrate abundance within wet meadow patches may decrease
over the course of the summer because of decreasing oxygen concentrations and
decreasing water levels (Kaldec and Smith 1992), so an adjacent wetland which remains
inundated (like bulrush marsh) would provide needed resources in mid- to late-summer
that open water would not. Wet meadows bordered by open water may also be less
protected from lake winds, storm-related fluctuations in water level, and temperature
fluctuations (Herdendorf 1992). For these reasons, wet meadows connected to other
types of wetlands (in a wetland context) may provide higher quality habitat for birds.

Similarly, three bird variables were positively related to contexts with a relatively higher

48

proportion of Lake Huron and higher stream densities (LAND2). These types of aquatic
habitats likely serve many of the same functions just discussed.

Second, wetland and aquatic contexts may provide protection from predators.
Predator densities and predation rates are typically higher in upland habitats compared to
wetland habitats (Picman 1988). Thus, adjacent forest and upland habitats may serve as a
source of terrestrial predators (e.g. squirrels or raccoons) that could access the wet
meadow via the shared boundary. Also, the physical structure of adjacent trees may
provide perches for visual-searching predators (e. g. corvids; Bergin et al. 1997) and/or
brood parasites (e. g. Brown-headed Cowbirds; Hauber and Russo 2000) permitting them
to scan deep into the interior of the wet meadow and enhancing their ability to detect
nests. In contrast, inundated wetlands or open water do not provide either form of access
for these predators and can restrict access merely by the presence of standing water (Jobin
and Picman 1997). Any or all of these mechanisms could decrease predation rates within
a wet meadow and enhance its attractiveness to breeding birds.

Of the species related to wetland contexts (LAND4 or LANDS), not all were true
wetland obligates (e. g. Sedge Wren), but many were facultative species (e.g. Song
Sparrow, Eastern Kingbird) which can and often do breed very successfiilly in entirely
terrestrial habitats. This underscores the importance of wetland contexts and wetland
connectivity (Taylor et al. 1993; Haig et al. 1998) to breeding birds in general including
facultative and even non-wetland birds. Whether or not these birds have higher breeding
success in wetlands than in their more commonly occupied terrestrial habitats is not
currently known, and future studies that tested this question would yield valuable

information.

49

Significant relationships to wetland contexts were observed almost exclusively
during 1998 (Table 7). Overall lake levels were approximately 30 cm lower in 1998 than
in 1997 (NCAA, unpublished data). Consequently, many of the wet meadows that had
been inundated during the breeding season in 1997 were merely saturated or entirely dry
during much of 1998 (Table 2). Because lower water levels may lead to less invertebrate
production fiom within the wet meadow, one would expect birds to seek out wet
meadows located in a wetland context where additional invertebrate resources were
available from the adjacent landscape, and relationships observed in 1997 and 1998 are
consistent with this expectation. In landscapes with high inter-annual variability in habitat
quality like Great Lakes coastal wetlands, birds may change habitat selection criteria
according to underlying habitat conditions. Preserving complexes of connected wetland
types is critical to providing the resources necessary for bird communities over the entire
range of annual conditions that are likely to occur in such systems.

Effects of Coastline Complexity. My third hypothesis was that coastline
complexity (i.e. coastlines that were more fractal or more sinuous) would be positively
associated with species richness and other bird variables. Although a few bird variables
were positively related to coastline components, relationships with coastline components
were predominantly negative (Table 7), exactly opposite of what I hypothesized. These
results are surprising because landscape theory predicts that highly complex landscapes
should be more stable (F onnan 1995) and hence, have higher biodiversity in patches.
Also, complex coastlines are correlated with greater wetland resources (Nilsson 197 8;

Kent and Wong 1982) are important for many wetland birds (Weller 1999).

50

One possible explanation for why birds were associated with less complex
coastlines is that, for the northern Lake Huron coastline, fractal dimension (i.e. COASTI)
was negatively related to the amount of coastal wetlands in the surrounding landscape
(Figure 2), and thus, to wetland contexts. This contrasts the positive relationship between
coastline fractal dimension and wetland area predicted by others (Kent and Wong 1982).
Wetland contexts are important for many species (see discussion earlier), so the negative
responses to COAST] may reflect this relationship.

Information about migratory forest songbirds in terrestrial landscapes may provide
a second explanation for why many wet meadow birds were negatively related to complex
coastlines. As a group, migratory songbirds are highly-sensitive to habitat fragmentation
(see Faaborg et al. 1995), and Flather and Sauer (1996) observed that migratory songbirds
were less abundant in forested landscapes with high complexity or high fractal dimension.
Complex landscapes also had, on average, smaller patch sizes. In this study, COAST2
(total coastline length) was similarly inversely correlated to mean patch size (Figure 3) in
the surrounding landscape, and the majority of the species with negative relationships to
coastline components were migratory songbirds (Table 7). Thus, patches in these
landscapes may be avoided by birds because they are in a landscape context which they
perceive as very fragmented (sensu Flather and Sauer 1996).

These two mechanisms could account for the negative relationships between birds
and coastline components I observed, but some species were positively related to coastline
complexity. Interpretation is further complicated because responses to COAST3 (a
contrast of scale) were mixed indicating that responses may be scale-specific, and the

critical scale may vary from species to species. More research is needed about how

51

0.16

 

0.12 —
.C
9
(U
2 0.08 —
.C
(I)
E
S
:11
°\.. 0.04 ~

0.00 7

 

 

COAST1: Fractal Dimension

Figure 2. Relationship between the the fractal dimension of the coastline
(COASTI) and the percent of the surrounding landscape (l-km from
perimeter of the wet meadow) comprised of bulrush marsh.

52

 

 

_x
01
l

 

Mean Patch Size (ha)
8

01
1

 

 

 

T l l l

COAST2: Total Coastline

Figure 3. Relationship between the the total coastline length (COAST2)
and the mean patch size (ha) in the surrounding landscape (l-km from
perimeter of the wet meadow).

53

coastline complexity and fractal dimension of landscapes affect bird distributions.
Understanding these landscape characteristics will be difficult because responses to
coastline characteristics may be year- and/or scale-specific and complicated by the effects
of different life-history traits (Hanson and Urban 1992) on avian response.
IMPLICATIONS FOR CONSERVATION AND RESEARCH

These results have implications for conserving birds in Great Lakes coastal wet
meadows, but they apply to other regions and habitats as well. Because Great Lakes
coastal wet meadows are emergent wetlands dominated by grasses and sedges, they
possess habitat features and bird species that are characteristic of both wetlands and
grasslands. Thus, my recommendations apply more broadly to the general management
and conservation of avian communities in both wetland and grassland habitats.

Habitat, landscape, and coastline components selected by regression models were
generally not consistent between years, and this illustrates the importance of long-term,
multi-year studies (W iens 1.981). This problem is further compounded in dynamic systems
like Great Lakes coastal wetlands because birds may alter habitat selection criteria as
overall habitat conditions change from year to year. Thus, only long-term, multi-year
research projects can identify the entire complement of habitat types and landscape
structures needed to conserve wetland bird communities over time in dynamic landscapes.
Unfortunately, such long-term studies do not exist, and management of wetland birds must
be guided by results from short-term studies such as mine until longer-running studies
have been completed. Fortunately, the direction of responses were remarkable consistent
across species despite inter-annual inconsistencies within species, for most landscape

components. This strengthens the conclusions that can be drawn about the effects of

54

landscape structure on wet meadow birds in general, and, in the absence of long-term
studies, some preliminary management guidelines can be drawn.

First, area and perimeter were important predictors of both species richness and
the occurrence of many bird species (Table 5). The importance of patch size has been
well-documented in wetlands (Brown and Dinsmore 1986; Weller 1999) as well as in
other habitats (e. g. grasslands [Herkert 1994; Vickery et al. 1994] and forests [e. g. Blake
and Karr 1987]). Similarly, patches with low perimeter-to-area ratio contain less edge-
and more interior-habitats (Helzer and Jelinski 1999). Management plans should continue
to focus on preserving large habitat patches with relatively little perimeter, although the
importance of small wetlands should not be ignored (Gibbs 1993; Semlitsch and Bodie
1998; Schwartz 1999).

Plans for the conservation of wetland birds should go further, however, and
consider the landscape context of patches. Wet meadows adjacent to a large number of
patches and/or patch types should be given priority over wet meadows in simple landscape
contexts because patches in complex contexts consistently had higher species richness and
were more likely to contain several species. My results also demonstrated that wet
meadows connected to other types of wetlands (i.e. in a wetland context) were
consistently associated with the presence or absence of not only obligate
wetland/grassland species, but also certain facultative wetland species. This relationship
was markedly more important in 1998 when water levels were low. Wetland conservation
schemes should strive to protect entire wetland complexes and preserve connections

among wetland types rather than focusing on portions of wetlands or individual wetland

types.

55

Considering landscape context could greatly enhance the effectiveness of
conservation efforts. For example, even small wet meadow patches could be valuable if
located in complex, wetland contexts which would then provide a stable set of resources
in low-water years. Similar relationships may operate in many other landscapes, and these
recommendations could likely be implemented in conservation efforts with a high
probability of success. However, further research about the effects of landscape context is
needed to fully understand these processes and insure the success of conservation efforts.
Without detailed information about the effects of landscape context on within-patch
diversity and process, conservation finds and efforts will not be fiilly effective, predictions
of landscape change will not be accurate, and critical ecosystem fiinctions could be

compromised.

56

CHAPTER TWO

LANDSCAPE CONTEXT AND COASTLINE COMPLEXITY:

EFFECTS ON SPECIES TURNOVER RATES

57

INTRODUCTION

Species turnover is the change in species composition of a patch or community
due to extinction, immigration, or both (Diamond and May 1977). Each individual
extinction (loss of a species) or immigration (addition of a new species) represents one
turnover event, and the sum of these events over a unit of time as a percentage of the total
species pool is the turnover rate. MacArthur and Wilson (1967) focused attention on the
importance of turnover to maintenance of species richness within island communities.
According to their theories of island biogeography, turnover rates on islands decrease
with increasing island area and increasing isolation (distance to the mainland or nearest
island), and these turnover processes interact to produce a predictable, equilibrium
number of species for each island.

Since that time, avian biologists have used island biogeography as a tool for
explaining patterns of avian species richness and as a framework for understanding the
effects of fragmentation in terrestrial habitat patches (see Faaborg et al. 1995 for review).
But surprisingly, little attention has been paid to patch-level turnover rates in avian
communities despite the importance of turnover to the success of conservation efforts
(see Rice et al. 1983; Gutzwiller and Anderson 1987b). Habitats with lower turnover
rates contain relatively more stable communities where individual species are less likely
to go extinct, and hence are more effective choices for habitat preservation. Also,
managers must be able to predict how rapidly communities can be expected to change,
and must also be able to predict the nature of these changes (Russell et al. 1995).
Without an understanding of what drives turnover within avian communities, such

predictions are not possible.

58

Studying avian turnover in most non-oceanic habitat islands is problematic
because they are not true islands, and the relatively straight-forward theories of island
biogeography often do not sufficiently explain observed patterns (e.g. Gutzwiller and
Anderson 1987b). Although area and isolation may influence turnover in avian
communities, both internal features of the habitat (Gutzwiller and Anderson 1987b) and
changes in habitat features over time (e. g. Weller 1994; Patten and Rotenberry 1998) can
influence patch-level turnover of species. Human-induced factors can also alter these
processes, not only by affecting habitat features, but also by promoting invasion of new
predators (Savidge 1987), influencing survival in migratory and wintering habitats (e. g.
Conway et al. 1995), or otherwise disrupting extinction and colonization processes.

The landscape context of a patch refers to the characteristics of the adjacent
habitats and surrounding landscape, and it is unknown how context affects turnover rates
within patches. Some processes critical to avian biology such as edge-related nest
predation can be influenced by the characteristics of the adjacent patches (Andren 1995;
Bayne and Hobson 1997). Also, recent studies have described effects of landscape
context on species richness and species-occupancy of patches for riparian forest patches
in the western United States (Gutzwiller and Anderson 1987a; Saab 1999), forest
fragments in a Canadian agricultural landscape (Freisen et al 1995), and Chilean forest
remnants (Estandes and Temple 1999). These patterns in species richness could be
caused by effects of landscape context altering patch-level turnover rates (colonization
and/or extinction), but such research has not been conducted.

It is not difficult to theorize how landscape context could influence turnover in

avian communities. If the surrounding habitat is inhospitable, it could effectively isolate

59

patches which are not geographically distant (W iens 1996). Certain contexts may
increase predation or mortality (Andren 1995; Bayne and Hobson 1997), provide fewer
foraging resources, or otherwise render a patch less-suitable as habitat. Breeding
individuals may be less likely to return to patches in unfavorable contexts the following
year. This would raise turnover rates in unfavorable contexts. Alternatively, breeding
individuals may be more likely to return to patches in favorable contexts because their
breeding success would be enhanced (Blancher and Robertson 1985; Bensch and
Hasselquist 199]). This would lower turnover rates in favorable contexts.

I tested whether or not landscape context influenced avian turnover rates in 40
Great Lakes coastal wet meadows over a 2-yr time-interval. Turnover rates were first
adjusted for area and within-patch habitat characteristics, and then I tested for additional
effects of landscape context. These analyses were designed to answer three major
questions: (1) Does landscape context affect avian turnover rates in wet meadows? (2)
Does the complexity of the associated coastline affect avian turnover rates? and (3) Does
island biogeography predict avian turnover rates?

These wetlands are associated with the northern shoreline of Lake Huron and are
some of the most undisturbed remaining wetlands in the Great Lakes regions. They are
critical habitat for migratory and breeding birds (Prince and Flegel 1995), and some of
these species are declining locally and/or regionally (Brewer et al. 1991). These wet
meadows are currently threatened by commercial and residential development, so
understanding the processes that drive turnover rates in these wet meadows is critical.
Unfortunately, information about turnover in wetland bird communities is lacking.

Without this understanding, it will be impossible to predict how the bird communities of

60

the region can be expected to change. As a result, the success of management and
conservation efforts will be hindered.
METHODS
Study Area
Data from 40 wet meadows (sedge- and grass-dominated wetlands) associated with the
northern shoreline of Lake Huron in Chippewa and Mackinac counties were used for
these analyses. Sites were selected based on specific vegetational and geographical
criteria. These criteria and a detailed description of the vegetation are listed in Chapter
One, Methods, Study Area.
Bird Censuses
The presence or absence of each bird species was determined each year using line-
transect censuses in each wet meadow. All birds seen or heard within the perimeter of
the wet meadow patch were recorded. Each transect was censused 4 times in 1997 and 5
times in 1998. Only the first 4 visits of 1998 (corresponding temporally to the 4 visits in
1997) were included in these analyses. A detailed description of censusing protocol is
listed in Chapter One: Methods, Bird Censuses.
Habitat Characteristics
Because within-patch habitat characteristics (e. g. Gutzwiller and Anderson 1987b) could
influence patch-level turnover in bird communities, I measured 20 habitat characteristics
along each census transect in both 1997 and 1998. Details of the habitat sampling

methods, as well as habitat variable definitions, are listed in Chapter One: Methods,

Habitat Characteristics.

6]

Calculation of Landscape Variables
Landscape context variables included the number of adjacent patches, the number of
adjacent patch types, and the proportion of the wet meadow perimeter bordered by each
patch type. The classification system used for delineating patches is listed in Table 1.
Landscape context variables also included the density of roads and streams in the
surrounding landscape. Coastline complexity measures were calculated for three
coastline segments associated with each wet meadow: SOO-m, I-km, and 2-km from the
perimeter of each wet meadow. Total coastline, Sinuosity, and fractal dimension were
calculated for each coastline segment using ArcView software and a Fortran program,
Fract_Line (Lam and De Cola 1993). Detailed description of how these variables were
measured is listed in Chapter One: Methods, Calculation of Landscape Variables.

Statistical Methods

Habitat Variables. Great Lakes coastal wetlands are influenced by overall lake water
levels which can fluctuate greatly from one year to the next (Bedford 1992). Thus, there
is the potential that mean water depth and, consequently, other within-patch habitat
characteristics changed from 1997 to 1998. I tested to see if such changes occurred using
two-tailed, paired-t tests on each of the 20 habitat characteristics (Ott 1988). The null
hypothesis was that the mean of a particular habitat characteristic was not different in
1998 compared to 1997. The alternative hypothesis was that it did change. When the
assumptions of the parametric t-test were not clearly satisfied, 1 used the Wilcoxon

signed-rank test instead (Ott 1988). 1 used or = 0.10 for all tests involving habitat

characteristics.

62

Because species turnover could have been affected by changes in habitat
characteristics (Patten and Rotenberry 1998), unchanging (or average) habitat
characteristics (Gutzwiller and Anderson 1987b), or some combination of both, I
conducted a principal components analysis (PCA) separately on two different sets of
variables. First, I identified the set of habitat characteristics that changed significantly
and calculated the difference in the mean of that characteristic in 1998 and the mean of
that characteristic in 1997 (see Table 8). Then I conducted a PCA on this set of variables.
Components derived from this PCA will be referred to hereafter as habitat change
components. Second, I took the mean value for all 20 habitat characteristics over the two
year period and conducted a second PCA on this set. Components derived from this PCA
will be hereafter referred to as mean habitat components. Within each set, components
with an eigenvalue > 1.0 were retained.

Landscape and Coastline Variables. Because landscape context and coastline
complexity variables were intercorrelated, I used principal components analysis to
remove the correlations and reduce the number of variables. I conducted two separate
PCA's, one each on the landscape context variables and the coastline complexity
variables. These analyses are described in detail in Chapter One, Methods, Statistical
Analysis. The same, identical landscape and coastline components are used in the present

analysis.

Calculation of Turnover Rates. I calculated turnover rate following the formula

of Diamond and May (1977):

turnover rate97 to 93 = (# extinctions” + # immigration593) / total # of specie597+9g

63

._0>..0.:.-08.0 .0... :00 0w:0..0 00: 0.0 0m0:0 0:0 000.. .00 0:003:00 00:000.. 000. :. 00.0800 002......
.... :0x00..>> :0 0 0.80020 05.0 0. 0.0.00 .00..- ...

 

 

-- -- -- -- 00.0.-000 00.. 0008 0080.80

-- -- -- - 00.0-00.0 00.. 0008 00800.80

-- -- -- - 00.0-00.0 .0. . 800 00000.000

-- - - 1... 0000-000 00.0 0.02. 00800.80
0w0:0 0:0 000... 00 00:00:00”.

.000 000 u. 0000.000.- 00. 00. . 0-000 .00 000.).

0000 000 n. 00.. .000.- 000 0.0.000 :0 0.03 :80

000.0 00.0 n. 000.000.- 000 0000-000 00.0 002

0000 0... n. 000.00.. .- 00.0 00. . 0-000 .00 30.03

.000 v .0. - n. 00.00.0000- 00.00- 00.0-00.0 :00. 02800? 00205000

.000 v 0. n .0 00.0-00.00- 00.0- 0. 0.80 00.0 0000.002, 00.00.”.

0.00 .00- u. 00.0.00. .0- 00.0- 0000-000 .00. 002.00

0000 00 u .0 000.000.- 00.0- 03.0-00.0 00.0 .0000

0000 .0 n .0 00.0.00...- .v.. 0000.00.00 00.00 0.00.880
00...: .000 -.0 00:00.08...
0000 00. - u. 000.000. 0.0- 000000 00... 0.0020 000:0. 0200
0.00 000 u .0 00.0-00.0- 000 000.000 0. .0 0.20.0 00.03002, .300
0000 000 n. 000-000- 00.0 000-000 000 0.050 0000.002, 0083
.000 000 u .0 00.0-00.0- 000 000.000 00.0 0.050 0020
0.0.0 00.0 n. 00.00-00,. .- 00... 0. 00.00.00 00.00. 0.0.2. 0020
0000 .00- u. 00.00-00.00- 00.0. 0000-... 0 00.00 0.0.2. V00:80..
.000 v .00. u. 0.0-00.0- 0. .0.- 0000-000 00.... 5000 .203

a 0.0.80 .000 00.5. .0... 0080 0|. 0.0005 8.00..
0w:0..0 :003.

 

000.000. 800.00.: 00.8000 0080.003. 0:0 03009.00 :. :83. 00.0.. -.0 08.0.2.0

80.00: 0... 5.3 00.0.8000 0300008 .03 00 00.. 0030...? 0w:0..0 3.0.0.. 0:0 0030...... 0.50.. :008 8.. 00.0080 3088.6 .0 0.00...

64

where # extinctions = the number of species detected in a patch in 1997 that were not
detected in that patch in 1998; # of immigrations = the number of species detected in
1998 that were not detected in 1997; and total # of species = the total number of different
species detected in a patch over both years (or # species detected in 1997 + #
immigrations).

A species was considered present in a patch if it was detected within the perimeter
of the wet meadow patch during at least one of the four census visits during a year. I
used unlimited-distance records rather than fixed-width records (birds detected within the
lOO-m wide transect) because this would have overestimated true patch turnover. For
example, a species recorded within the transect in 1997 could move its territory the next
year such that it was located just outside the lOO-m wide transect. Although that species
persisted in the patch, it would have been erroneously recorded as an extinction if only
fixed-width data had been used. Using data collected within the patch perimeter but over
an unlimited transect width avoids this problem.

Because the factors that influence the turnover of species which nest in the wet
meadows might differ from the factors that influence the turnover of accidental species or
species which use the wet meadows primarily for foraging (e. g. Great Blue Heron), I
calculated 3 different turnover rates. Total species turnover rate was calculated based on
all species detected in the wet meadow regardless of nesting requirements. Nesting
species turnover was calculated using only that subset of species that can potentially
breed in wet meadows (Appendix A). Because habitat features of coastal wet meadows
are characteristic of both wetlands and grasslands, nesting species include obligate and

facultative grassland species (Vickery et al. 1999) in addition to obligate and facultative

65

wetland birds (based on information in Terres 1980). Non-nesting species turnover was
calculated using only occurrences of those species not included in the nesting guild
(Appendix A).

Regression modeling. Because I wanted to test for effects of landscape context
and coastline complexity on avian turnover rate above and beyond that which could be
explained by area and habitat effects, I adopted a two-stage regression approach
(Morrison et al. 1998) similar to that used in Chapter One. In the first stage, I regressed
turnover rate against area, perimeter-area ratio (PAR), habitat change components and
mean habitat components. The residuals from this model were then regressed against
landscape and coastline components. Landscape and coastline components significantly
influenced turnover rates if a component explained a significant portion of the variation
in these residuals (t-statistic with P < 0.10).

At each stage, I used a stepwise selection procedure. Rather than let .the stepwise
algorithm select the variables unsupervised, I set the F-to-stay value at P < 0.15 to select
a suite of variables. I then evaluated models involving all possible subsets of this suite of
variables to find the best model (Draper and Smith 1981). The best model (1) met the
assumptions of regression (constant variance, normal residuals, etc.); (2) had the lowest
sum of prediction residuals (a form of validation [Myers 1987]); and (3) all explanatory
variables had t-ratios significant at P < 0.10. Transformations were conducted if and only

when necessary to satisfy the assumptions of regression (Rahbek 1997).

66

RESULTS
Habitat Characteristics

Although wet meadows included in this study were selected because they shared
similar vegetation structure and composition, there was still considerable variation among
sites (see Tables 2 and 8). Five mean habitat components were retained from principal
component analysis, and these five components accounted for 75% of the total variation
in the original mean values (Appendix F).

Several habitat characteristics did change significantly from 1997 to 1998 (Table
8). Mean water levels were significantly lower in 1998 relative to 1997 (P < 0.001), and
consequently so were the percent cover of submerged (P < 0.001) and floating (P <
0.001) vegetation (Table 8). Conversely, mean grass cover (P < 0.024), mean grass
height (P = 0.010), mean grass density (P = 0.051) and total vegetation density (P =
0.040) increased in 1998 (Table 8). Two habitat change components were retained from
principal component analysis on these 7 variables, and they accounted for 56% of the
total variation in the original habitat change variables (Appendix G).

Each mean habitat component and habitat change component was influenced
primarily by a set of correlated and often biologically- or structurally-related variables, so
I interpreted each component (Table 9).

Landscape Context Components

The context of wet meadows within the surrounding landscape varied from very.
simple contexts (patches bordered entirely by a single habitat type) to complex contexts
(patches bordered by a diversity of habitat types: see Table 4). Principal component

analysis identified 5 components which accounted for 7 8% of the total variation

67

Table 9. Mean habitat components and habitat change components derived from original
variables which were retained for analysis.

 

 

Principal
Component Interpretation
Mean Habitat Components
PC 1 Increasing hummock height, grass height, grass density, and vegetation
density
Decreasing bulrush frequency
PC2 Increasing frequency of trees, snags, and alder and woody vegetation
density
Decreasing water depth and bulrush fi'equency
PC3 Increasing woody vegetation density, shrub foliage diversity, and
frequency of willow
PC4 Increasing frequency of open water, deciduous trees and snags
Decreasing frequency of graminoid vegetation
PCS Increasing frequency of coniferous trees, moss, open water and
floating vegetation
Decreasing frequency of willow
Habitat Change Components
CHANGE] +: Increased grass height, grass density , little or no decrease in water

depth

-: Increased frequency of submersed vegetation, decreased water depth

CHANGE2 +2 Increased grass height and frequency of floating vegetation; little or no

change in water depth

-: Increased frequency of graminoid vegetation, decreased (or little
change) in grass density, decreased water depth

 

68

(Appendix D). Similarly, coastline complexity varied from very simple coastline
segments to very sinuous, highly fractal segments (Table 4). PCA identified 3 coastline
components which accounted for 77% of the total variation in coastline characteristics
(Appendix E). Interpretations and names for landscape and coastline components are
listed in Table 5.
Changes in Species Richness

I estimated species richness using over 3000 individual detections at 40 sites
during 320 census visits. Total species richness declined significantly from a mean of
9.08 species/patch (range: 1-22) in 1997 to 8.25 species/patch (range: 1 to 18) in 1998
(paired-t = -2.05, df = 39, P = 0.048). Nesting species richness declined significantly
fiom a mean of 7.30 species/patch (range: I to 18) in 1997 to 6.75 species/patch (range: 1
to 16) in 1998 (paired-t = -1 .84, df = 39, P = 0.074). Species richness of non-nesters also
declined from a mean of 1.78 species/patch (range: 1 to 7) to 1.50 species/patch (range:
1-5) species/patch, but this decline was not significant (paired-t = -l . 10, df = 39, P =
0.279).

Turnover Rate

Total species turnover rate. All 40 sites experienced at least one turnover event
(extinction or immigration) between 1997 and 1998 (Table 10), and mean turnover in the
wet meadows averaged 0.56 turnover events/species (range: 0.19 - 0.86). The first stage
of the regression procedure indicated that turnover rate was negatively associated with
log(area) and mean habitat component 3 (MHAB3: greater density of woody vegetation,
greater shrub foliage diversity, and coverage of willow). Total turnover was positively

associated with mean habitat component 4 (MHAB4: more open water, deciduous trees

69

Table 10. Extinctions, immigrations, and turnover rate for 40 wet meadows associated
with the northern shoreline of Lake Huron in Chippewa and Mackinac counties,
Michigan, 1997-1998.

 

Wet Meadow Extinctions Immigrations Total Species Turnover Rate
1 1 2 16 0.188
2 S 3 18 0.444
3 S 3 13 0.615
4 2 2 5 0.800
5 7 2 24 0.375
6 1 4 8 0.625
7 O 2 3 0.667
8 4 3 10 0.700
9 2 2 10 0.400
10 2 2 13 0.308
11 4 2 7 0.857
12 S 3 14 0.571
13 3 0 6 0.500
14 3 0 7 0.429
15 3 2 13 0.385
16 6 2 17 0.471
17 2 2 6 0.667
18 S 6 22 0.500
19 l 1 3 0.667

20 3 2 9 0.556
21 4 0 7 0.571
22 6 2 20 0.400
23 6 3 11 0.818
24 1 4 14 0.357
25 2 2 6 0.667
26 7 S 19 0.632
27 2 3 9 0.556
28 9 2 21 0.524
29 S 4 15 0.600
30 l 3 13 0.308
31 6 4 12 0.833
32 S O 6 0.833
33 4 2 9 0.667
34 3 6 11 0.818
35 8 2 12 0.833
36 2 2 13 0.308
37 2 4 14 0.429
38 1 l 3 0.667
39 0 2 7 0.286
40 S 2 11 0.636

 

7O

and snags, less coverage of graminoid vegetation). Area and habitat components
explained 36% of the variability in turnover rate (Table 11).

Landscape components described an additional 14% of the variation in total
species turnover rate (Table l 1). Turnover rate was negatively associated with LAND]
and positively associated with LANDS. Total species turnover rate was lower in complex
contexts (patches adjacent to a large number of patches and patch types), and turnover
rate was lower in wetland contexts (patches with perimeters adjacent to proportionately
more open water and wetland habitats) compared to terrestrial forest contexts. Coastline
components were not related to total species turnover rate.

Nesting species turnover rate. Nesting species turnover was not related to area.
Similar to total species turnover rate however, nesting species turnover was negatively
associated with MHAB3 and positively associated with MHAB4 (Table 1 l). Nesting
species turnover was also positively related to LANDS. Thus, turnover rates of nesting
species were lower in wetland contexts (bordered by a relatively higher proportion of
open water and/or wetland habitats) compared to terrestrial, forested contexts.

Non-nesting species turnover rate. Turnover rate of non-nesting species was
negatively related to area (r2 =- 0.10; Table 11). Mean habitat components, habitat
change components, landscape components and coastline components did not explain
significant amounts of remaining variation in non-nesting species turnover rates.

DISCUSSION
A verage Habitat Conditions vs. Changes in Habitat Conditions
Great Lakes coastal wetlands are dynamic habitats in that water levels fluctuate

from year-to-year (Bedford 1992), and vegetation characteristics can change in response

71

Table 11. Results of logistic regression on bird species turnover rates in wet meadows
associated with the northern shoreline of Lake Huron in Chippewa and Mackinac
counties, Michigan, 1997-1998. For all regressions, n = 40.

 

Habitat Landscape
Turnover Rate Variables r2 Components r2
All Species Area (-) 0.36 LAND] (-) 0.50
HPC3 (-) LANDS (+)
HPC4 (+)
Nesting Species HPC3 (-) 0.21 LANDS (+) 0.27
HPC4 (+)
Non-Nesting Species Area (-) 0.10 -- --

 

72

to these fluctuations. Lake Huron water levels were approximately 20 cm lower in 1998
than in 1997 (NCAA). Consequently, many habitat characteristics including water level,
grass density, and coverage of submerged and floating vegetation were significantly
different in 1998 (see Table 8). Because these specific features provide many services
critical to breeding birds (e. g. protection from predators [Jobin and Picman 1997],
foraging substrates [Brewer et a1. 1991; Weller 1999], etc), one would intuitively expect
these changes to be a major driver of turnover in these wet meadows (Weller 1994).
However, regression models selected only mean habitat components, and not habitat
change components, as predictors of avian turnover (Table 11).

Although this suggests that average habitat conditions influence turnover rates
(Gutzwiller and Anderson 1987b), 1 caution that this does not imply that changing habitat
conditions are unimportant to turnover rates, breeding success or other aspects of avian
ecology. First, this study was conducted over a short time-interval (2 yrs), and the impact
of these habitat changes on, turnover rates and bird distributions may not be apparent over
such a short interval. Second, the magnitude of most habitat changes were correlated
with the mean of that habitat characteristic over the study period (the greater the water
depth on average, the greater the potential for large changes), or were correlated with area
(Table 12). Because of these correlations, mean habitat components may explain, not
only variation in turnover due to average habitat conditions, but also some of the
variation due to changing habitat conditions. Thus, mean habitat components may have
been selected as the best predictors by regression techniques even though habitat change
components may have also influenced turnover rates. I caution that the potential

importance of yearly fluctuations in water levels and other habitat characteristics on

73

Table 11. Correlation coefficients and P-values for correlations between
habitat change components and area, perimeter-area ratio, and mean
habitat components for 40 wet meadows associated with the northern
shoreline of Lake Huron in Chippewa and Mackinac counties, Michigan,
1997-1998.

 

 

Habitat Habitat
Area and Mean Habitat Components Change PCl Change PC2
log(Area) —0.021 -0.3 1 1
0.899 0.051
Perimeter-area ratio (PAR) -0. 151 0.285
0.354 0.074
Mean habitat PC 1 0.127 -0.179
0.437 0.268
Mean habitat PC2 0.363 0.560
0.022 0.000
Mean habitat PC3 0.078 -0.316
0.634 0.047
Mean habitat PC4 0.149 0.208
0.360 0.197
Mean habitat PCS -0.019 0.210
0.909 0.193

 

74

turnover rates and avian community dynamics within Great Lakes coastal wetlands
should still be considered in both fitture research and conservation efforts.
Eflects of Habitat Characteristics on Turnover Rates

Turnover rates were lower in wet meadows characterized by relatively denser
woody vegetation, a more diverse shrub layer (< 2-m in height) and more willow shrubs
(MHAB3). Wet meadows with these characteristics offer more vertical substrates for
foraging and more woody vegetation required by shrub-nesting species like the Yellow
Warbler, and woody wet meadows may be more likely to consistently attract these
species year after year compared to sites with little shrubby vegetation. Alternatively, the
structure of shrub vegetation and shrub-dominated wetlands is less affected by changing
water levels than more herbaceous-dominated wetlands, or at least changes in shrub-
dominated wetlands have a longer lag time (Burton 1985). Thus, species turnover rates
could be lower in shrubby wet meadows because the overall habitat structure and
vegetative composition does not change as rapidly.

Turnover rates werehigher in wet meadows characterized by increased frequency
of pockets of open water and interspersed deciduous trees and snags and a decreased
frequency of graminoid coverage (MHAB4). Two factors could account for this
relationship. First, wet meadows possessing more homogeneous grass coverage may
simply attract wetland/ grassland breeding birds (6. g. Swamp Sparrow, American Bittern)
more consistently than those where graminoid vegetation is heterogeneous or interspersed
with deciduous trees and open water. Second, interspersed trees and snags may provide

perches for predators (like corvids; Bergin et al. 1997) or brood parasites (Hauber and

Russo 2000) within the meadow and, thus, elevate nest predation rates by allowing

75

visual-searching predators to penetrate deep into the interior of the wet meadow patch.
Similarly, open pockets of open water may disrupt the protective cover of the graminoid
vegetation making bird nests easier to locate. Breeding wetland birds that suffer
predation events are not likely to return to the same patch in the following year (Blancher
and Robertson 1985; Bensch and Hasselquist 1991), and either of these mechanisms
could result in increased turnover rates.

Effects of Landscape Context

While recent studies have demonstrated that characteristics of the adjacent
landscape can affect species richness (Pearson 1993; Estandes and Temple 1999; Saab
1999; Chapter One), patch selection by species (Friesen et al. 1994), and edge-related
nest predation (Andren 1995; Bayne and Hopson 1997), mine is the first study to relate
adjacent habitats to within-patch turnover rates. Turnover rates were lower in patches
adjacent to a proportionately greater number and diversity of patches (complex contexts:
LANDI). Turnover rates were also lower for sites adjacent to proportionately more
water and wetland habitats (a wetland context) compared to patches surrounded by forest
habitats (a terrestrial context; LANDS).

The effect of a diverse landscape context was first predicted by Forman (1995)
who theorized that species richness of a patch should increase with the number of
adjacent patches because each adjacent patch contributes different colonizers to the patch
in question. The logical extension of this theory is that turnover rates should be lower in
patches surrounded by a variety of patch types because there is an increased probability
of replacing species which abandon the patch with colonizers from adjacent patches.

Previous analyses (Chapter One) indicated that bird species richness was indeed higher

76

for wet meadows in diverse contexts. The decreased turnover rates for meadows in
complex contexts supports this mechanism and suggests that adjacent patches may indeed
serve as sources for colonizers. Because of the effects of adjacent patches on
colonization rates, wet meadows in complex contexts may maintain higher species
richness and have lower annual turnover rates.

However, previous results suggested that other explanations may also be
appropriate (see Chapter One) because patches located in complex contexts also had
greater abundance (Table 7). Colonization mechanisms predict increased species
richness, but not increased abundance. A possible explanation is that complex contexts
may provide a more diverse, and hence, more stable, suite of resources for foraging. This
would increase the intrinsic habitat quality of a patch and allow breeding bird densities to
be greater (Anderson and Titman 1992; Gill 1995), and could also account, at least in
part, for the decreased turnover rates observed in complex contexts. More stable food
resources would attract potential breeders more consistently, and breeding pairs would be
less likely to abandon these patches in low water years when production of invertebrates
and plants within the meadow itself is low. More research will be required to test if one
or both potential mechanisms are responsible for the patterns I observed.

The second relationship was that turnover rates were lower in patches located in
an aquatic/wetland context compared to those in a terrestrial context (Table 11). Wetland
contexts similarly contained greater species richness, total bird abundance and abundance
of several species (Chapter One, Table 7). Two mechanisms may interact to account for
this relationship. First, many species that breed in wet meadows often forage in nearby

habitats (e. g. Red-winged Blackbird [Orians 1980]), and wetland habitats provide an

77

abundant supply of aquatic invertebrates (Tilman and Schwegler 1978; Gathman et al.
1999). Emergence of aquatic invertebrates is highly seasonal such that different species
emerge over a short time-interval (few days to a few weeks: Orians 1980). Diverse
wetland contexts contain more invertebrate species, and would thus provide a consistent
supply of emerging invertebrates over the duration of the bird-breeding season. Wetland
contexts would be especially important during low water years (like 1998) when the wet
meadows are not flooded, and hence, invertebrate densities and emergence are
substantially lower within the meadow itself. In such years, adjacent wetlands would
insure that food resources remained available, and breeding birds would be less likely to
abandon wet meadows during dry years. This would lower turnover rates in patches
located in wetland contexts.

Second, adjacent wetlands may provide protection from predators simply because
water prevents access by many terrestrial predators (Jobin and Picman 1997). Predation
rates, and presumably either predator densities or predator efficiency, are higher in
upland habitats than in wetland habitats (Picman 1988). Adjacent upland habitats may
provide access to the wet meadow by upland predators whereas adjacent wetland may
buffer the wet meadow from these predators. Also, visual-searching predators and brood
parasites can use trees at forest edges to search the wet meadow for nests (Bergin et al.
1997; Hauber and Russo 2000), and adjacent wetlands do not provide this type of
structure. If the potential for nest predation is higher in forested, terrestrial contexts, then
breeding individuals would be less likely to return to a meadow where they had bred
unsuccessfully the year before (Blancher and Robertson 198; Bensch and Hasselquist

1991), hence, increasing turnover rates in terrestrial contexts.

78

One might argue that these observed effects are really an indirect effect of area
and/or perimeter. Large wet meadows, or oddly shaped meadows with a relatively longer
perimeter, would be adjacent to a greater number of patches and patch types, and would
be more likely to be adjacent to wetland habitats, simply because they were larger.
However, the two-stage regression design removed effects of area and perimeter (in the
form of perimeter/area ratio) before I tested for the effects of landscape context, so this is
unlikely. Thus, these results represent a true "context effect" above and beyond effects of
area and perimeter.

Eflects of Coastline Complexity

Ecological theory suggests that complex, or highly fractal, landscapes are more
stable because the diversity of the landscape dampens environmental and biological
oscillations (F orman 1995). Additionally, some researchers have noted that some
wetland species and habitats tend to be more abundant along complex coastlines (Nilsson
1978; Kent and Wong 1984). Thus, I hypothesized that turnover rates should be lower in
wet meadows associated with complex or highly fractal coastline segments. However,
my results did not support this because coastline components were not related to turnover
rates.

Coastline components were significant predictors of species richness and
presence/absence in these same Great Lakes wet meadows (see Chapter One, Table 7), so
it is still possible that birds do use coastline features as a habitat selection criteria, at least
to some extent. If birds do use coastline features as habitat selection criteria, then these

features may serve only to increase or decrease the maximum species richness of a

79

particular patch. In Great Lakes coastal wet meadows, coastline features were not related
to turnover rates.
Turnover in Nesting vs. Non-nesting Species

Regression analysis selected different variables as predictors for turnover rates in
nesting and non-nesting birds which suggests that different factors influence colonization
and extinction events in these two groups of species. Turnover of nesting birds was
related to habitat characteristics and landscape context. The importance of patch-level
habitat characteristics to nest site selection (Knopf and Sedgewick 1992) and nesting
success (e. g. Picman 1988) of wetland/grassland birds has been well-documented, so it is
not surprising that habitat characteristics are related to turnover rates of nesting species.
Turnover of nesting birds was also lower in wetland contexts. Wetland contexts could
decrease turnover rate through effects on nesting success described earlier: either the
higher quality foraging resources or the protection from nest predators that would be
afforded by wetland contexts would increase nesting success and thus lower turnover.

In contrast, turnover of non-breeding birds was lower in larger patches, but was
not related to any other habitat or landscape component. Foraging birds (such as Great
Blue Herons) traveling from other habitats may be more likely to forage in large wet
meadows simply because of a sampling effect. They would be more likely to encounter a
large wet meadow than a small wet meadow, or perhaps these species would be simply
more likely to be detected by an observer in large patches. Similarly, the probability of
detecting that same species again the following year would be greater in larger wet

meadows, and observed turnover rates would be lower in large meadows. It is also

80

possible that larger patches are more likely to contain the food resources required by
foraging birds. Effects of area would account for this as well.

The habitat and landscape components selected in the model for total species
richness represented a combination of the variables selected in the individual models for
nesting and non-nesting species turnover (Table 11). Thus, turnover of nesting species
and turnover of non-nesting species may be independent processes governed by a
different suite of factors: habitat and landscape context for the former, and area for the
latter. This is not surprising because nesting birds require different characteristics (e.g.
nesting substrates, long-term food availability, etc.) than do non-nesting or foraging birds
(e. g. proximity to their nesting habitat, immediate food supply), and these characteristics
may be associated with different habitat features and landscape contexts. Biologists and
managers should realize that conservation criteria aimed at decreasing turnover rates for
one group of birds would not be effective at decreasing turnover rates in the other.

One landscape component in the overall turnover model, diverse contexts
(LANDl), was not included in either the nesting or non-nesting turnover models.
Complex contexts could have influenced turnover rates of either nesting species, non-
nesting species, or both and still not have been selected as a predictor if it was secondary
in importance to the other factors (habitat characteristics and wetland context for nesters,
area for non-nesters). Only when turnover of nesters and non-nesters are combined into a
single turnover rate, however, the combined effects of complex contexts may have been

great enough for the stepwise algorithm to select that component.

81

Relevance of Island Biogeograph y

Island biogeography predicts that turnover rate should decrease with increasing
island (or patch) area and with increasing isolation (farther from the mainland or nearby
patches of the same type), and that these processes interact to maintain a predictable
equilibrium number of species (MacArthur and Wilson 1967). Researchers have made
repeated attempts to use this theory to predict the species richness and turnover patterns
of birds in various types of non-oceanic "island" patches (Faaborg et al. 1995), but these
attempts have met with only varying success for two major reasons. First, colonization
and extinction rates can be influenced by within-patch habitat features (Gutzwiller and
Anderson 1987b), by changes in habitat features (Weller 1994), or by human-induced
changes (Savidge 1987; Patten and Rotenberry 1998) such that mainland avian
populations are rarely at equilibrium. In the Great Lakes coastal wet meadows I studied,
this assumption was clearly not satisfied. On average, species richness decreased
between 1997 and 1998 which would not be expected if the region were truly at
equilibrium.

Second, the habitats surrounding these wet meadows are not inhospitable for birds
as is the case for the oceanic islands systems used to develop island biogeography
(MacArthur and Wilson 1967). In fact, adjacent habitats are used by many wetland
species throughout the breeding season (Weller 1999). If these aspects of wetland habitat
"islands" are considered, it is quite reasonable to expect landscape context to influence
turnover rates and for island biogeography to poorly predict these turnover rates. This is
precisely what was observed for bird communities in Great Lakes coastal wet meadows.

Total species turnover rate was inversely related to area. However, further analyses

82

indicated that this was due primarily to turnover of non-nesting species whereas turnover
of nesting species was related, not to area, but to habitat characteristics and landscape
context.

Isolation was not related to turnover rates either. As an a posteriori analysis, I
calculated two measures of isolation for each wet meadow: distance to nearest wet
meadow and proportion of the surrounding landscape (within l-km of patch perimeter)
comprised of wet meadow. I then went back, and to each stage of each regression model,
added both measures of isolation to the stepwise algorithm. Neither isolation measure
was significant for any of the three turnover rates. This is not surprising because the
northern Lake Huron coastline is relatively undisturbed and contains a great deal of wet
meadow habitat. In fact, no wet meadow was farther than 1.2 km from another wet
meadow patch and most inter-patch distances were much smaller. These distances are
easily traversed by vagile organisms like birds, so these wet meadows may not be
fiinctionally isolated at all.

Turnover of bird communities in Great Lakes coastal wet meadows was
influenced by many of the same variables that were significantly related to species
richness (compare Table 5, Chapter 1, and Table 10, this chapter). Thus, the theory of
island biogeography applies to these wetlands in that the processes of colonization and
extinction (through turnover rates) are linked to and help determine species richness.
However, the more specific tenets that area and isolation predict turnover rates (and
species richness) did not apply to birds in Great Lakes coastal wet meadows. Rather,
turnover rates in coastal wetlands are more likely driven by habitat conditions and

landscape context.

83

IMPLICATIONS FOR CONSERVATION AND MANAGEMENT

Although recommendations based on my results are most appropriate for
managing avian diversity in Great Lakes coastal wetlands, they may apply to other
habitats as well. Wet meadows are similar to both wetlands and grasslands in terms of
both habitat characteristics and breeding bird species, so species turnover may be
influenced by similar factors in other grassland and wetland habitats in other landscapes.

My results indicate that managers should not assume that island biogeography is
an appropriate model for predicting turnover rates or for formulating conservation plans.
Oftentimes, the primary target of management plans in both grasslands and wetlands are
nesting species. In Great Lakes wetlands, turnover of nesting species was influenced by
habitat characteristics and landscape context, whereas turnover of non-nesting species
was related only to area. Thus, programs aimed at preserving species which breed in
wetland or grasslands should not rely solely on predictions of island biogeography, but
consider critical features of the habitat and a patch's context in the landscape in addition
to wetland area and isolation.

In Great Lakes coastal wet meadows, annual turnover of birds would be
minimized in large wet meadows located in a wetland context where they are adjacent to
other aquatic- and wetland-habitat types. Wet meadows should also contain a relatively
greater density and diversity of woody vegetation in the shrub layer, be comprised of
homogeneous, relatively unbroken stands of grasses and sedges, and lack deciduous trees
and snags in the interior of the wet meadow. Wet meadows meeting these criteria would
attract both breeding and non-breeding species consistently from year-to-year, and

community composition would be more stable.

84

Future studies should be conducted over S-yrs, lO-yrs, or longer if possible. My
interpretations are based on only one time-step (2 yrs) which is often not long enough to
identify long-term patterns because relationships can fluctuate from year-to-year (Wiens
1981). However, because such long-term studies are currently lacking for the Great
Lakes region, recommendations based on short-term studies such as mine should be
heeded until longer-term studies have been completed. Scientists and managers alike
should seek to identify and preserve habitat patches with lower turnover rates because
this represents more stable communities less likely to lose species (biodiveristy) to
extinction (Russell et al. 1995). Without understanding turnover rates and the factors that
influence this process, effective long-term management of wetland birds is difficult
because without this knowledge, it is impossible to predict how communities will change

over time.

85

CHAPTER THREE

PENINSULA EFFECTS IN A COASTAL LANDSCAPE:

ARE COVES MORE SPECIES RICH THAN LOBES?

86

INTRODUCTION

The phenomenon of decreasing species richness with increasing distance along
peninsula lobes is called a peninsula effect (Forman and Godron 1986). Peninsula effects
have been identified for a variety of taxa on large peninsulas like Baja California (e. g.
Taylor and Regal 1978) and Florida (e. g. Keister 1971), but this pattern has not been
consistently observed on all peninsulas or for all taxa (Taylor 1987). Review of the
evidence (Forman 1995) suggests that these patterns may be caused by mechanisms of
colonization and extinction related to dispersal rates and island biogeography (MacArthur
and Wilson 1967; Taylor and Regal 1978), a gradient in climate or environmental factors
from the mainland to the peninsula tip (Taylor and Regal 1978), or a combination of both
(Milne and Forman 1986).

Attention to peninsula effects has been focused on the continental scale, but little is
known about the presence or absence of peninsula effects at smaller, more regional scales
like a Great Lakes coastline. This is surprising in light of commonness of highly
interdigitated landscapes in both coastal and terrestrial regions and the utility of such
knowledge for conservation planning. Milne and Forman (1986) found that plant species
richness decreased along 9 peninsulas 375 m to 16,000 m in length in coastal Maine, but
other studies at this scale are lacking. If such patterns are ubiquitous at both the
continental and regional scales, such information would be extremely valuable in setting
conservation priorities.

The theory of island biogeography can be extended to peninsular landscapes if
patches located on lobes are considered analogous to far islands and patches located

Within coves can be considered analogous to near islands. These analogies suggest two

87

predictions (see Figure 4): (1) species should accumulate faster with area (steeper slopes)
in lobe patches compared to cove patches; and (2) cove patches should be generally more
species rich (higher intercepts). 1 tested these predictions along the northern coast of
Lake Huron in Michigan. I compared bird species richness of coastal wet meadows
located within coves (sensu near islands) to species richness of wet meadows located on
peninsula lobes (sensu far islands). Before testing for differences in species richness, 1
controlled for variation related to habitat and landscape characteristics.
METHODS
Study Area

I worked in wet meadows located along the highly convoluted, northern shoreline
of Lake Huron in the Upper Peninsula of Michigan (Mackinac and Chippewa counties).
All wet meadows were seasonally or shallowly flooded grasslands, were within 500-km of
the coastline, were influenced by lake water levels; and were dominated by a mixture of
tussock-forming grasses (Calamagrostis canadensis) and sedges (Carex stricta and C arex
aquatilis). Wet meadows were interspersed with varying amounts of bulrush (Scirpus
spp.), cattail (T ypha spp.), and shrubs (Salix spp., Alnus spp, and Myrica spp.).
Submersed (Potomageton spp.) and floating (Polygonum spp., Lemna spp.) vegetation
were often present in standing water between hummocks.

Bird Censuses

Within each of the wet meadows, I established a 100-m wide, permanent transect
(Vemer 1985) which I censused censused multiple times between 15 May and 4 July of
each year (4 times during 1997; 5 times during 1998). Census techniques are described in

detail in Chapter One, Methods, Bird Censuses.

88

 

Near Islands

# of Species

Far Islands

 

 

 

Island Area

Figure 4. Theoretical species-area relationships as predicted by island biogeography.
Wet meadows located in coves are analogous to near islands; wet meadows located on

lobes are analogous to far islands.

89

Habitat Characteristics

Because variation in habitat characteristics can influence bird species richness (e. g.
Craig and Beal 1992), I measured 20 habitat characteristics along each bird census
transect in late July after vegetation had reached mature height. Within each segment, 1
estimated grass height and density, density and diversity of shrub vegetation, water depth
and hummock height using point-intercept sampling techniques (sensu Rotenberry and
Wiens 1980). Percent cover of vegetation types (e. g. bulrush, cattail, open water, etc.)
was estimated as the frequency of occurrence in l-mz-sampling plots. Frequency of trees
and snags were estimated from 2.5-m radius sampling plots. Detailed description of
habitat sampling techniques are in Chapter One, Methods, Habitat Sampling. Habitat
variables are listed in Appendices H and 1.

Landscape Characteristics

Surrounding landuse of the northern Lake Huron shoreline was interpreted into an
ArcView GIS from 1:24,000-scale color aerial photographs taken in 1992. Accuracy was
verified by extensive ground-truthing. Wet meadow areas, perimeters and the proportion
of the perimeter adjacent to specific patch types were calculated in ArcView. Buffer
operations were used to calculate density of wet meadow, Lake Huron, roads and streams
within l-km of the wet meadow border. For each wet meadow, 3 coastline segments
(500-m, l-km, and 2-km from each wet meadow perimeter) were derived from a vector
coverage of the Great Lakes shoreline (Great Lakes Environmental Research Center). For
each segment, I calculated total coastline length, Sinuosity, and fractal dimension (Lam and
De Cola 1993). Further details about the derivation of landscape and coastline variables

are in Chapter One, Methods.

90

Construction of Bird Variables

I analyzed data only for those species which were detected within the 100-m wide
transects. Three different measures of species richness were calculated: total species
richness (the total number of different species detected within the wet meadow during a
particular year); nesting species richness (the total number of nesting species detected
within the meadow during a year); and non-nesters species richness (total number of non-
nesting species detected within a year). Because wet meadows have characteristics of
both wetlands and grasslands, a species was designated as a nesting species if it was an
obligate or facultative grassland (Vickery et al. 1999) or wetland (based on Terres 1980)
nester. I also calculated total bird abundance and abundance of species occurring on > 2/3
of the wet meadows. Abundance was calculated as the average number of individuals
detected divided by the total transect area.

Statistical Analysis

Site Selection. Of the 40 wet meadows censused for breeding birds, I identified 30
which were clearly located either in a cove or out on a lobe (see Figure 5). Wet meadows
not clearly located in either a cove or lobe were excluded from these analyses. Initial
examination of the species-area curve for these wet meadow patches revealed that the
range of patch areas was much higher for cove sites than for lobe sites (Figure 6).
Because it is inappropriate to compare regression lines or other statistics which are
estimated over disparate ranges, I restricted statistical analysis to 21 sites that were less
than 5 ha (Figure 6). For cove sites, n = 10; for lobe sites, n = 11. The center of lobe
patches was located at a mean of 817 m (standard error = 221.7 and ranging from 157 m

to 2268 m) from the mainland.

91

_ 1‘..-

 

 

Figure 5. Diagram of northern Lake Huron shoreline. A = wet meadow patch distinctly
located within a cove. B = wet meadow patch not clearly located in either a lobe or out on
a cove. C = wet meadow patch distinctly located out on a lobe. Only wet meadow meeting
either criterion A or C were included in this analysis.

92

Total Species Richness - 1997

25

 

 

 

 

 

 

 

 

l
l
l
i o
204 j .
I o
|
i o
15— I .
I ' 0
I
“II .
I I
l' I
- l
5— . l
O |
' i 0 coves
'- i I LOBES
0a :
l
|
I
I
l l l l I I I
0 5 10 15 2° 25 30

Area (ha)

Figure 6. Species-area relationships in 30 wet meadows located in
coves or out on lobes. Sites to the left of dashed line are ones retained
for analysis.

93

35

Habitat, Landscape, and Coastline Characteristics. Habitat characteristics (e.g.
C: raig and Beal 1992), landscape characteristics (e. g. Brown and Dinsmore 1986; Naugle
at al. 1999, Chapter One), and coastline characteristics (see Chapter One) can affect
p attems of bird species richness and could thus be confounded with or mask true
p eninsula effects. To ameliorate this, I accounted for effects of these variables before
testing for effects of coves and lobes. To reduce the number of variables, I conducted 4
separate principal component analyses: once each for 1997 habitat characteristics, 1998
habitat characteristics, landscape variables, and coastline variables. For each of the 4 sets
of variables, I retained components with eigenvalues > 1 for flirther analysis.

- Species-area Regressions. To test hypotheses about differences in species-area
relationships of coves and lobes, I used a two-stage regression approach (Morrison et al.
1 998) to account for habitat and landscape effects. Because of small sample size and
lirrnited degrees of freedom, I restricted the number of habitat/landscape components to
one. Using Proc RSquare-(SAS Institute, Inc, 1989), I identified the principal that was the
best, single predictor of each richness variable from a set of variables including area,
Perimeter-area ratio, habitat components, landscape components, and coastline
Components. The best predictor had the highest r2, a linear relationship with the richness
Variable, and satisfied all assumptions of regression. Each richness variable was then
regressed against this best predictor, and the residuals from this regression were saved and
Used as the dependent variable in a second regression.

1n the second-stage of regression, I used these adjusted richness variables
(residuals) to test for differences in species richness patterns between cove patches and

lobe patches. I conducted separate adjusted species richness vs. area regressions for cove

94

and lobe patches and then tested for differences in the slope and intercept parameters. 1
tested two null hypotheses for each year and richness variable combination: (1) the
intercept of the species-area regression for cove patches would be equal to or less than the
intercept of the species-area regression for lobe patches; and (2) the slope of the species-
area regression for cove patches would be equal to or greater than the slope of the species
v s - area regression for lobe patches. The alternate hypotheses were that species-area
regressions involving cove patches would have higher intercepts but smaller slopes than
regressions involving lobe patches. I tested for differences in slopes and intercepts
between regressions involving cove and lobe patches using one-tailed t-tests described by
Zar (1984: eqs. 18.1 and 18.25). For all statistical tests, I used or = 0.10 to improve

statistical power (W estrnoreland and Best 1985; Riffell et al. 1996).

Species-area relationships are typically expressed on a log-log scale by
transforming both species number and area. However, 1 did not conduct transformations
unless it was necessary to meet the assumptions of regression. The species-area curve
involving all 31 cove and lobe patches demonstrated the classical curvilinear relationship,
but my analyses were restricted to only the linear part of the relationship (Figure 3). Thus,
I suspected that the assumptions of regression might be satisfied without transformations
(sensu Rahbek 1997). I evaluated each model with respect to the assumptions of linear
regression (Ott 1988): linear relationship between X and Y; constant variance of residuals;
and normality of residuals. For all richness variables and all years, untransformed data met
these assumptions, and transformations did not improve on these criteria relative to the

untransformed model. In many instances, transformations introduced blatant violations of

regression assumptions.

95

.s

To assess the ability of the t-tests to detect true differences in slopes and
intercepts, I calculated minimum detectable differences (63 [Zar 1984]). For these
analyses, 63 represent the smallest absolute difference in a slope or intercept that would
have been detected at or = 0.10 at a power of 80%. I calculated the median and range of
Ss for tests involving slopes and intercepts.

Species' abundance and presence/absence. For total bird abundance and
abundance of species that occurred on at least 2/3 of the sites, I used traditional linear
models to test for differences in abundance between cove and lobe sites. I restricted the
number of covariates to one covariate for each abundance variable (see above). I
identified the single, best predictor from the set of habitat, landscape, and coastline
components. The final linear model included one covariate (the best predictor) and
cove/lobe designation as the main effect. I used or = 0.10 and two-tailed F-statistics for
hypothesis tests about coves and lobes.

For species occurring on < 2/3 of the sites, I used presence/absence data. Before
testing for differences between coves and lobes, I identified the single, best predictor from
among the habitat, landscape and coastline components using a forward selection
procedure in a generalized linear model (PROC GENMOD: SAS Institute Inc, 1997).
This best predictor was then included as a covariate with cove/lobe position as the main
effect. To test for significant effects of cove/love position, I used Wald x2 statistics. All
test statistics and P-values were based on Type III sums of squares. Tests based on Type
III sums of squares account for effects of other variables in the model. Hence, they test
for effects of cove/lobe position above and beyond effects of the habitat or landscape

covariate.

96

 

RESULTS
Habitat Characteristics

Although wet meadows were selected based on similar habitat features, the habitat
variables I measured did vary among wet meadows. Habitat characteristics are
summarized elsewhere (Chapter One and Two, Tables 2 & 8) and are beyond the scope of
this chapter. Principal component analysis identified 6 components in 1997 (Appendix H)
and 5 components in 1998 (Appendix I) to be retained for subsequent analyses. These
components accounted for 81% and 79% of the variation in the original variables.

Habitat characteristics generally did not differ significantly between cove and lobe
patches during either 1997 or 1998. Frequency of moss and submersed vegetation were
greater in lobe patches than in cove patches (P < 0.10) in 1998, but not it 1997. None of
the other 18 habitat characteristics I measured were significantly different between lobes
and coves in either year. These two significant tests are less than the 4 significant tests (2
years x 20 habitat variables x 0.10 = 4) expected due to chance.

Landscape and Coastline Characteristics

Landscape and coastline characteristics varied among wet meadows. However,
discussion of these characteristics is beyond the scope of this paper. Readers are referred
to Chapters One and Two for detailed descriptions. For these analyses, PCA identified 5
landscape components (Appendix J) and 3 coastline components (Appendix K). These
components accounted for 87% and 84% of the variation in the original variables.

Species Richness
I detected a total of 39 species over two years in the 21 wet meadows. Species

residuals vs. area regressions were generally significant (P < 0.10) for total species

97

 

richness and nesting species richness in coves and lobes separately (Table 12), indicating a
relationship between richness and area independent of habitat effects for these richness
measures. Richness of non-nesting species was not significantly related to area (Table 12).

Intercepts from cove-patch regressions were not different from lobe-patch
regressions (Table 12). For non-nesting species richness in 1998, the slope was greater
for lobe sites than for cove sites as predicted (P = 0.086). However, intercepts did not
differ in 1997 for any other variable. This single significant test is only 1 out of a total of
12 tests on slopes and intercepts. Additionally, regression parameters did not differ
between coves and lobes in a consistent direction. Thus, there is little evidence of
peninsula effects on species-area relationships.

Minimum Detectable Eflect Sizes

For tests involving slopes from species residuals - area regressions, the mean
minimum detectable effect size (SS) in slopes was 1.643 (range: 0.816 - 2.362). The
mean 6 for intercepts was 1.804 (range: 0896-2594). Increases in slopes and intercepts
this large or larger would have been detected with a statistical power of 80% at an or =
0.10. These 6s are quite large. Tests would have been reliably significant (80% power)
only when lobe slopes were 72% - 182% steeper than the cove slopes or when lobe
intercepts were 34% - 129% smaller than cove intercepts. Thus, I would not have
detected small or subtle peninsula effects had they occurred.

Individual Species

I conducted a total of 20 tests on abundance and presence/absence data for a total

of 11 different species (Table 13). Some species occurred frequently enough for statistical

analysis during only one year. Swamp Sparrow abundance was higher in cove patches

98

Table 12. Coefficients of adjusted species-area regression for cove and lobe wet meadows
associated with the northern shoreline of Lake Huron, Chippewa and Mackinac counties,
Michigan. For all regressions, cove n = 10 and lobe n = 11.

Parameter Coves Lobes Difference t P-value

 

Total Species Richness - 1997 (HABPC 3)

Slope 1.59 1.36* 0.23 0.241 0.594
Intercept -3 .72 -2.80 -0.92 -0.391 0.649
Total Species Richness -- 1998 (LANDPCS)
Slope 1.29* 200* -0.71 -0.700 0.247
Intercept -2.59 -4.41 1.82 0.721 0.241
Nesting Species Richness - 1997 (LANDPCI)
Slope 0.86 142* -0.56 -0.751 0.232
Intercept -1.58 -3.24 1.66 0.884 0.195
Nesting Species Richness - 1998 (HABPC 3)
Slope 202* 1.06”“ 0.96 1.540 0.929
Intercept -4.7 1 -2.24 -2.47 -1 .592 0.935
Non-nesting Species Richness -- 1997 (LANDPC 5) .
Slope 0.44 0.27 0.17 0.348 0.563
Intercept -0.69 -0.71 0.02 0.017 0.493
Non-nesting Species Richness - 1998 (LANDPCS)
Slope -0.04 0.71 -0.75 -1.428 0.086
Intercept -0.09 -1.22 1 .13 0.849 0.204

 

* Indicates that the slope for that particular species-area regression was significant (P <
0.10).

99

 

Table 13. Results of regression on species abundance and presence absence variables in

21 wet meadows located in either coves or lobes of the northern shoreline of Lake Huron,

Chippewa and Mackinac counties, Michigan.

 

Position
Variable Covariate Parameter F or 3f” P—value
A bundance Variables

Total Abundance

1997 HAB3 (+) -0.56 0.59 0.453

1998 LANDS (-) 0.08 0.00 0.948
Common Yellowthroat

1997 HAB4 (-) 0.28 1.82 0.195

1998 HAB2 (+) -0. 18 0.71 0.411
Swamp Sparrow

1997** HABZ (+) 0.04 0.04 0.836

1998 HAB] (+) 0.66 5.43 0.032
Song Sparrow

1997 Area (-) -0.44 -3.93 0.008

1998 Area (-) -0.50 0.26 0.619

Presence/absence Variables

Great Blue Heron

1997 LAND] (+) 4.89 4.88 0.027
American Bittern

1997 LANDS (-) 1.86 1.81 0.178

1998 HAB2 (+) 1.17 0.79 0.373
Mallard

1997 LAND] (+) -0.15 0.02 0.883
Virginia Rail

1997 COAST2 (+) 2.24 2.29 0.130
Alder Flycatcher

1997 HAB3 (+) 0.28 0.08 0.780

1998 -- -0.59 0.45 0.504

 

lOO

 

 

Table 13. Continued.

 

 

Position
Variable Covariate Parameter F or [7' P-value

Yellow Warbler

1997 LANDS (-) 0.96 0.78 0.371

1998 LANDS (-) 0.30 0.09 0.770
Red-winged Blackbird

1997 LAND] (+) 0.74 0.38 0.537

1998 HAB2 (-) 2.02 3.07 0.080
American Goldfinch

1998 HAB3 (+) 7.82 8.73 0.003

 

* Abundance variables use a standard least squares F-statistic. Presence/absence variables

use a Wald f—statistic

** Data for 1997 Swamp Sparrow abundance is based on presence/absence data.

101

 

than in lobe patches during 1998. Similarly, the probability of detecting Great Blue Heron
(1997), Red-winged Blackbird (1998) and American Goldfinch (1998) was also higher in
cove patches. In contrast, the abundance of Song Sparrow (1997) was greater in lobe
patches. I observed a total of 5 significant tests which is approximately 2.5 times the
number (0.10 x 20 tests = 2) of significant tests expected by chance alone.
DISCUSSION
Habitat Characteristics

Habitat characteristics such as grass density (e. g. Herkert 1994), vertical structure
(e. g. Rotenberry and Wiens 1980) and heterogeneity (Rotenbeny and Wiens 1980; Craig
and Beal 1992) can influence both the likelihood that a species will be present and the
ability of species to be detected by sight and sound (Waide and Naims 1988). Thus,
features of the habitat can influence species richness and estimates of species richness.
Additionally, habitat characteristics may vary with any environmental gradient present
between cove and lobe patches, and this variation could mask or be confounded with true
cove vs. lobe differences. In this study, only 2 out of the 20 measured habitat
characteristics were significantly different between cove and lobe patches, and these were
both during one year. Two significant tests could easily be accounted for by chance.
Thus, there were few, if any major habitat differences between coves and lobes. Although
this does not preclude the presence of environmental differences between coves and lobes
in an unmeasured variable (e. g. air temperature or wind velocity), a gradient was not
reflected in 20 major habitat characteristics. This evidence, and my technique of removing

habitat- and landscape-related variation in species richness using principal components,

102

 

reduces the likelihood that habitat characteristics were confounded with the cove/lobe
designation.
Species Richness vs. Area Regressions

I detected only one statistically significant difference in intercepts (non-nesting
species richness in 1998), and no significant differences in slopes (Table 12). While this
one significant test was consistent with the prediction that species-area slopes would be
steeper in lobe patches, regression parameters did not differ between cove and lobe
regressions in a consistent manner. Additionally, this one significant result can also be
accounted for by chance. Because of the low power of my tests; however, I cannot rule
Out the possibility that small, subtle peninsula effects might still be present in Great Lakes
coastal wet meadows. However, because of the lack of consistent patterns in the
regression parameters, it is highly doubtful that large differences in the species-area
relationships of cove and lobe patches actually existed, and also unlikely that any large
peninsula effects were present at this scale in this system.

There are several factors which may explain why peninsula effects were absent, or
only very subtle, in this study. One reason may be that there was little difference in habitat
characteristics between cove and lobe sites. Gradients in habitat or environmental
conditions along the peninsula may sometimes produce peninsular gradients in species
richness (e. g. Taylor and Regal 1978; Milne and Forman 1987). Because habitat
differences were slight or absent in my study area, it is unlikely that habitat-driven
peninsula effects could have occurred, although differences in environmental variables 1
did not measure (e. g. wind speeds) might have existed. Theoretically, peninsula effects

should still be present because of the effect of distance from the mainland on colonization

103

 

rates (MacArthur and Wilson 1967; Figure 4), but the northern Lake Huron coastline
consists of relatively small peninsulas 100 m to 3 km in length. Peninsula effects have
been observed at this scale for plants (Milne and Forman 1987), but birds are highly vagile
organisms capable of traversing (and hence colonizing) these distances easily. For this
reason peninsula effects may not exist for birds at this spatial scale or be so subtle that
they are difficult to detect. Additionally, in this region migrating birds arrive traveling
north in the springtime, and thus encounter the tips of these peninsulas first after crossing
the lake. This order of arrival may increase the likelihood of a species colonizing a lobe
patch and offset existing peninsula effects.
Response of Individual Species

Great Blue Heron, Red-winged Blackbird, American Goldfinch, and Swamp
Sparrow were more abundant or more likely to be detected in cove patches than in lobe
patches, but none were consistently related to cove/lobe position during both years. These
species may prefer cove sites for several reasons. American Goldfinch may be more likely i
to be detected in cove patches simply because they are closer to the mainland, terrestrial
habitats they typically inhabit, or they may simply avoid habitats close to open water such
as those on coves. Similarly, Red-winged Blackbirds regularly use upland habitat (among
others) for foraging (Orians 1980), and cove patches were likely near to more of these
types of habitats. Great Blue Heron use wet meadows primarily for foraging, and cove
sites may be closer to their forested nesting habitats. Or, cove patches may be closer to
habitats Heron use for retreat when threatened. Swamp Sparrow and Red—winged
Blackbird were the only wetland/ grassland breeding species that were more abundant in

cove patches. Although the habitat features were not generally different between cove

104

 

 

and lobe wet meadows, cove patches may buffer many environmental factors I did not
measure. For instance, wind speeds and air temperatures may vary less in cove patches
than in lobe patches, but without further research this is only speculation. One species,
Song Sparrow, was more abundant in lobe patches compared to those in coves. Song
Sparrows are known for their affinity for edges (Terres 1980), and may perceive peninsula
tips as edge habitats. Another possibility is that, as a generalist species, Song Sparrows
may prefer lobes to avoid interspecific competition because other species (like the
conspecific Swamp Sparrow) may be less abundant in lobes.

Although I observed over twice as many significant tests for abundance and
presence/absence as predicted by chance, still only 4 of the 11 species were significantly
associated with either cove patches or lobe patches. Additionally, total abundance did not
differ between cove patches and lobe patches. Although there may be a few species which
discriminate between coves and lobes at this scale, this difference between coves and lobes
does not appear to be important for birds in Great Lakes coastal wet meadows. More
studies should be conducted at a variety of spatial scales and with greater replication in
order to identify more completely the suite of species which exhibit peninsula effects.

IMPLICATIONS FOR CONSERVATION AND MANAGEMENT

Because coves were not significantly more species rich than lobes and few species
were more abundant in coves, cove patches would not be considered superior
conservation choices in the northern Lake Huron coastal region. However, 1 caution that
dismissing the presence or importance of cove vs. lobe effects in this region or other
regions would be unwise because they have been demonstrated in other systems for other

taxa (e. g. Taylor and Regal 1978; Milne and Forman 1987). In this study, I addressed

lOS

 

only avian richness patterns, and did not investigate other taxa. Avian reproductive
parameters like pairing success, clutch size or reproductive success could vary between
cove and lobe patches, but these differences would not necessarily have been reflected in

differences in species richness. Additionally, statistical power (ability to detect effects)
was quite low, and only large peninsula effects would have been detected. Research about
the effects of cove and lobe locations on all aspects of avian ecology and other taxa should
continue. Until such research is completed, the extent of peninsula effects will remain
unknown.

An important caveat is that all of the wet meadows larger than 5 hectares in the
region were located in coves, and the importance of large patches to the reproductive
success of birds (e. g. Hoover et al. 1995; Burhans and Thompson 1999), persistence of
populations (F orrnan 1995), and conservation reserve design (Schwartz 1999) have been
well established. Based on this information, conservation plans for the northern Lake :
Huron coastline should target large wetlands located in coastal coves. There may also be 1
subtle differences in habitat characteristics or environmental conditions between lobes and
coves (e. g. more floating vegetation in cove patches), and such differences could represent
a difference in habitat quality for wildlife. Without further research, however, the

importance of these differences will not be fully understood.

106

 

‘5'

SUMMARY AND RECOMMENDATIONS

Based on my results, conservation planning for birds in the northern Lake Huron
coastline should focus not simply on large wetlands but go fiirther and consider the
landscape context of particular patches. Two specific types of landscape contexts were
clearly beneficial to birds. Patches located in complex contexts (adjacent to a large
number of patches and patch types) and in wetland contexts contained significantly
higher species richness than patches of similar size in simple and terrestrial contexts.
Complex contexts and wetland contexts were also associated with lower rates of within-
patch species turnover were thus not only more species-rich but also more stable and
predictable in terms of species composition. For these reasons, conservation plans should
seek to preserve whole wetland complexes rather than individual wetlands so that
preserves contain the landscape contexts that maximize both species diversity and
population persistence.

Effects of other types of landscape context were less clear. For example, species
richness was higher in more human-developed contexts, but effects of human-developed
contexts were generally lacking for individual species and totally absent for species
turnover rates. However, the potentially negative impacts of human development in the
northern Lake Huron shoreline should not be dismissed because negative effects of
human activities on birds have been well-documented (Freisen et al. 1995; Blair 1996;
Bolger et al. 1997). Also, the degree of human development the northern Lake Huron
shoreline is slight compared to the heavy commercial and industrial impacts that are
common in other parts of the Great Lakes region. Thus, some species with an affinity for

human-developed areas may have been attracted (e. g. American Robin sensu Blair 1996)

107

 

without excluding sensitive species. For these reasons, human-related impacts on birds in
the region should remain a serious management concern and a primary focus of research.

Another less than general trend involved coastline complexity. Richness and

most bird species were associated with simpler coastlines, but a few bird species were
associated with more complex coastlines. In the northern Lake Huron region, highly
complex and fractal coastlines were negatively correlated to mean patch sizes and amount
of coastal wetland in the surrounding landscape. Hence, wet meadows associated with
complex coastlines were located in landscape contexts which contained less wetland and
appeared more fragmented. For these reasons, many birds may have avoided these
patches. Coastline complexity was not an important predictor of turnover. Conservation
priorities which emphasize simpler coastlines would benefit a large number of species,
but would not benefit other species like Mallards which were associated with complex
coastlines. More research is needed before the effects of coastline complexity on wetland
birds can be fully understood, and management decisions based on coastline features
should not be made until such research is complete.

Peninsula effects on birds appeared to be absent in the northern Lake Huron
coastline in that species richness of wet meadow patches located in coves was not
significantly different from patches located out on lobes. Although some species were
associated with either coves or lobes, responses were not consistent. One important
caveat, however, was that all the large (>5 ha) wet meadows were located in coastal
coves. For birds, peninsula effects do not provide predictions useful for conservation and

management, but future research about peninsula effects on other taxa or other aspects of

avian biology would be worthwhile.

108

Habitat, landscape, and coastline components selected in regression models were
generally not consistent between 1997 and 1998, and this illustrates the importance of
long-term, multi-year studies (W iens 1981). This problem is further compounded in
dynamic systems like Great Lakes coastal wetlands because birds may alter habitat
selection criteria as overall habitat conditions change from year to year (e. g. wetland
contexts were more important in a low water year). Thus, longer-running, multi-year

research projects are needed to identify the entire complement of habitat types and
landscape contexts needed to conserve wetland bird communities over time in dynamic
landscapes. Unfortunately, such long-term studies do not exist, and management of
wetland birds must be guided by results from short-term studies such as mine until

longer-running studies have been completed.

109

APPENDICES

110

Appendix A. List of bird species detected in 40 wet meadows associated with the
northern shoreline of Lake Huron in Chippewa and Mackinac counties, Michigan, 1997-

 

 

1998.

Wet Meadow
Common Name Scientific Name Nester“
Pied-billed Grebe Podilymbus podiceps
Green Heron Butorides striatus
Great Blue Heron Ardea herodias
Black-crowned Night Heron Nycticorax nycticorax
American Bittern Botaurus lentiginosus Nester
Mute Swan Cygnus olor
Canada Goose Branta canadensis
Mallard Anas platyrhynchos Nester
Black Duck Anas rubripes Nester
Blue-winged Teal Anas discors Nester
Wood Duck Aix sponsa
Common Merganser Mergus merganser
Cooper's Hawk Accrpiter cooperii
Bald Eagle Haliaeetus leucocephalus
Northern Harrier Circus cyaneus Nester
Sandhill Crane Grus canadensis Nester
Virginia Rail Rallus limicola Nester
Sora Porzana carolina Nester
Killdeer C haradrius vociferus Nester
Common Snipe Gallinago gallinago Nester
Mourning Dove Zenaida macroura Nester
Ruby-throated Hummingbird Archilochus colubris
Belted Kingfisher Cerlye alcyon
Northern Flicker Colaptes auratus
Red-naped Sapsucker Sphyrapicus varius
Hairy Woodpecker Picoides villosus
Downy Woodpecker Picoides pubescens
Eastern Kingbird T yrannus tyrannus Nester
Great-crested Flycatcher Myiarchus crinitus
Yellow-bellied Flycatcher Empidonaxflaviventris
Alder Flycatcher Empidonax alnorum Nester
American Crow Corvus brachyrhynchos
Black-capped Chickadee Poecile atricapillus
Marsh Wren C istothorus palustris Nester
Sedge Wren C istothorus platensis Nester
Gray Catbird Dumatella carolinensis
American Robin Turdus migratorius
Cedar Waxwing Bombycilla cedrorum

Black-and-white Warbler

Mniotilta varia

Ill

Appendix A. Continued.

 

 

Wet Meadow
Common Name Scientific Name Nester“
Nashville Warbler Vermivora ruficapilla
Yellow Warbler Dendrioca petechia Nester
Chestnut-sided Warbler Dendrioca pensylvanica
Common Yellowthroat Geothlypis trichas Nester
Bobolink Dolichonyx orzyivorus Nester
Red-winged Blackbird Agelaius phoeniceus Nester
Northern Oriole Icterus galbula
Common Grackle Quiscalus quiscula Nester
American Goldfinch Carduelis tristas
Savannah Sparrow Passerculus sandwichensis Nester
LeConte's Sparrow Ammospiza leconteii Nester
Chipping Sparrow Spizella passerina
White-throated Sparrow Zonotrichia albicollis
Lincoln's Sparrow Melospiza lincolnii Nester
Swamp Sparrow Melospiza georgiana Nester
Song Sparrow Melospiza melodia Nester

 

* Designation as nesting species indicates species is either obligate or facultative wetland
or grassland nesting bird. Designations are based on Vickery et al. (1999), life-histroy
traits described in Terres (1980), and personal observations (S. Riffell).

112

Appendix B. Eigenvectors of the first 5 principal components derived from the 1997
habitat variables for 40 wet meadows associated with the northern shoreline of Lake

Huron in Chippewa and Mackinac counties, Michigan.

 

Habitat Variable PC] PC2 PC3 PC4 PCS
Eigenvalue 4.70 3 .94 2.60 1.72 1.22
Percent variance explained 23.51 19.71 12.99 8.59 6.08
Water depth 0.23 -O.33 0.18 -0. 11 -0.03
Hummock height 0.30 -0. 13 0.16 -0.08 0.08
Grass height 0.42 0.08 0.10 -0.06 -0.04
Grass density 0.34 0.28 0.01 -0.03 0.12
Woody vegetation density -0.21 0.13 0.46 -0.07 -0. 12
Total vegetation density 0.32 0.30 0.04 -0.01 0.09
Shrub foliage diversity -0.11 0.04 0.51 -0.02 0.17
Frequency of cover types
Graminoid -0.04 0.20 -0.19 -0.53 0.13
Cattail 0.14 0.10 -0.20 0.20 -0.26
Bulrush -0.27 -0.34 -0.06 0.02 -0.21
Floating vegetation 0.24 -0.14 0.18 0.19 0.39
Submersed vegetation 0.26 0.28 0.05 -0.18 -0.00
Willow 0.04 -0.18 0.43 -0.04 -0.35
Alder -0. 16 0.28 0.31 -0.02 -0.20
Open water 0.0] -0.17 0.20 0.19 0.52
Moss -0.24 0.18 -0.05 -0.00 0.34

Frequency of trees and

snags
Coniferous trees -0.27 0.19 0.02 -0.19 0.29
Deciduous trees -0.06 0.30 0.07 0.46 -0.19
Coniferous snags -0. 16 0.25 0.15 -0. 19 -0.0]
Deciduous snags -0.05 0.23 -0.02 0.51 0.14

 

113

Appendix C. Eigenvectors of the first 5 principal components derived from the 1998
habitat variables for 40 wet meadows associated with the northern shoreline of Lake

Huron in Chippewa and Mackinac counties, Michigan.

 

 

Habitat Variable PC] PC2 PC3 PC4 PCS
Eigenvalue 5.63 3.63 2.37 1.90 1.19
Percent variance explained 28.13 18.14 11.87 9.51 5.95
Water depth 0.05 —0.37 0.25 -0.05 -0. 13
Hummock height 0.38 -0.00 0.01 0.07 0.02
Grass height 0.39 -0.02 0.0] 0.14 0.01
Grass density 0.39 0.06 0.04 0.14 0.15
Woody vegetation density -0. 12 0.32 0.26 0.28 -0. 17
Total vegetation density 0.39 0.1 1 0.08 0.13 0.12
Shrub foliage diversity -0.15 0.17 0.38 0.33 -0.10
Frequency of cover types
Graminoid 0.00 0.22 -0.40 0.25 0.19
Cattail 0.15 -0.03 -0.05 -0.35 -0. 14
Bulrush -0.35 -0.19 0.07 0.02 -0.04
Floating vegetation 0.17 -0.20 0.33 0.19 0.39
Submersed vegetation -0.24 -0. 16 0.18 0.05 0.17
Willow -0.03 -0.1 1 0.09 0.46 -0.49
Alder -0.04 0.42 0.25 -0.03 -0.01
Open water 0.05 -0.29 0.44 -0.06 0.23
Moss -0.27 0.04 0.02 -0.08 0.32

Frequency of trees and

snags
Coniferous trees -0.14 0.25 0.07 -0.07 0.41
Deciduous trees 0.09 0.32 0.20 -0.26 -0.21
Coniferous snags -0.03 0.34 0.11 0.03 0.13
Deciduous snags 0.11 0.13 0.30 -0.48 -0.19

 

114

Appendix D. Eigenvectors of the first five principal components derived from the

landscape context variables for 40 wet meadows associated with the northern shoreline of

Lake Huron in Chippewa and Mackinac counties, Michigan.

 

 

Habitat Variable LAND] LAND2 LAND3 LAND4 LANDS
Eigenvalue 4.14 2.67 1.94 1.59 1.28
Percent variance explained 27.58 17.78 12.97 10.61 8.55
# of adjacent patches 0.40 0.16 -0.23 -0.24 0.15
# of adjacent patch types 0.42 0.13 -0.17 -0.21 0.04
# ofinterfaces 0.38 0.11 -0.24 0.27 0.18
% of patch perimeter
adjacent to:
Urban 0.18 0.03 0.58 -0.03 -0.05
Non-forested opening 0.23 0.16 0.07 -0. 15 0.19
Forest -0.34 0.10 -0.23 0.03 0.51
Open water -0.15 0.05 0.08 -0.56 -0.48
Forested wetland 0.16 -0.43 -0. 16 -0.01 -0.06
Bulrush marsh 0.09 0.26 -0.30 0.50 -0.24
Cattail marsh 0.28 -0.20 0.34 0.08 -0.09
Total wetland 0.33 -0.21 -0.18 0.40 -0.28

% ofwet meadow in -0.03 0.30 0.24 0.10 0.18

surrounding landscape

(within l-km of patch

perimeter) ~

% ofLake Huron in 0.04 0.46 0.00 0.06 -0.28

surrounding landscape

(within 1-km of patch

perimeter)

Road density 0.25 -0.16 0.31 0.19 0.39

Stream density 0.08 0.50 0.18 0.16 -0.01

 

115

Appendix E. Eigenvectors of the first three principal components derived from the
coastline complexity variables for 40 wet meadows associated with the northern shoreline
of Lake Huron in Chippewa and Mackinac counties, Michigan.

 

 

Habitat Variable COAST] COAST2 COAST3
Eigenvalue 3 .27 2.60 1.05
Percent variance explained 36.32 28.93 1 1.70
Total coastline

500-m segments 0.20 0.51 -0.21

l-km segments 0.16 0.56 0.20

2-km segments 0.16 0.42 0.59
Sinuosity

500-m segments 0.45 0.01 -0.34

l-km segments 0.29 -0.27 0.46

2-km segments 0.20 -0.40 0.38
Fractal dimension

500-m segments 0.43 -0. 10 -0.30

l-km segments 0.45 -0.02 -0.08

2-km segments 0.44 -0. 12 0.07

 

116

Appendix F. Eigenvectors of the first five principal components derived fi'om mean
habitat variables for 40 wet meadows associated with the northern shoreline of Lake
Huron in Chippewa and Mackinac counties, Michigan, 1997-1998.

 

 

Habitat Variable PC] PC2 PC3 PC4 PC 5
Eigenvalue 5.49 3.85 2.50 1.89 1.18
Percent variance explained 27.44 19.24 12.51 9.46 5.90
Water depth 0.13 -0.37 0.19 -0.06 -0.02
Hummock height 0.37 -0.05 0.12 -0.05 0.01
Grass height 0.40 0.03 0.06 -0.07 -0.05
Grass density 0.38 0.17 -0.00 -0.03 0.12
Woody vegetation density -0. 16 0.23 0.41 -0.13 -0.15
Total vegetation density 0.37 0.21 0.03 -0.01 0.09
Shrub foliage diversity -0. 12 0.10 0.52 -0.08 0.03
Frequency of cover types
Graminoid -0.00 0.19 -0.26 -0.48 0.14
Cattail 0.13 0.04 -0.22 0.23 -0.22
Bulrush -0.31 -0.26 0.03 0.01 -0.02
Floating vegetation 0.21 -0. 19 0.26 0.13 0.38
Submersed vegetation 0.28 0.19 0.03 -0. 18 0.13
Willow 0.01 -0. 16 0.40 -0.26 -0.37
Alder -0.09 0.39 0.25 0.01 -0.07
Open water 0.03 -0.21 0.29 0.32 0.43
Moss -0.26 0.12 -0.7 0.06 0.33

Frequency of trees and

snags*
Coniferous trees -0.20 0.24 0.01 -0.06 0.43
Deciduous trees 0.02 0.33 0.06 0.37 -0.30
Coniferous snags -0.08 0.32 0.12 -0. 12 0.11
Deciduous snags 0.03 0.21 0.03 0.55 -0.05

 

* Frequency of trees and snags are mean values for 1997 because these variables did not

change over one year.

117

Appendix G. Eigenvectors of the first two principal components derived from habitat
variables that changed significantly between 1997 and 1998 for 40 wet meadows
associated with the northern shoreline of Lake Huron in Chippewa and Mackinac

counties, Michigan, 1997-1 998.

 

 

Habitat Variable CHAN GE] CHAN GE2

Eigenvalue 2. 17 1 .78
Percent variance explained 31.04 25.41
A Water depth 0.24 0.56
A Grass height 0.36 0.42
A Grass density 0.58 -0.37
A Total vegetation density 0.58 -0.33
A Frequency of cover types

Graminoid -0. 16 -0.39

Floating vegetation 0.14 0.34

Submersed vegetation -0.30 -0.03

 

118

I
‘sz

Appendix H. Eigenvectors of the first 6 principal components derived from the 1997
habitat variables for 21 wet meadows located in lobes or coves of the northern shoreline
of Lake Huron in Chippewa and Mackinac counties, Michigan.

 

Habitat Variable HAB] HAB2 HAB3 HAB4 HAB5 HAB6
Eigenvalue 5.36 4.36 2.73 1.59 1.15 1.05
Percent variance explained 26.79 21.78 13.63 7.97 5.73 5.25
Water depth -0.34 -0.21 0.10 -0.10 -O.19 -0. 12
Hummock height -O.36 -0.08 0.11 -0.10 -0.18 -0. 12
Grass height -0.37 0.17 0.09 -0.03 0.04 -0.00
Grass density -0.23 0.34 -0.00 -0.02 -0.09 0.35
Woody vegetation density 0.29 0.08 0.37 -0. 15 -0.13 0.09
Total vegetation density -0.20 0.35 0.05 -0.01 -0.1 1 0.36
Shrub foliage diversity 0.21 0.02 0.45 -0. 12 -0.05 0.14
Frequency of cover types
Graminoid 0.13 0.25 -0.26 -0.36 0.02 -0.14
Cattail -0.06 0.11 -0.13 0.53 0.24 -0.36
Bulrush 0.17 -0.38 0.04 0.15 -0.21 0.09
Floating vegetation -0. 18 -0.17 0.11 0.21 0.33 0.25
Submersed vegetation -0.19 0.34 -0.02 -0.16 0.16 -0.07
Willow 0.00 -0. 12 0.52 0.04 0.01 -0.22
Alder 0.24 0.26 0.25 0.04 -0.11 -0.25
Open water 0.03 -0.11 0.18 -0. 11 0.65 0.34
Moss 0.25 0.02 -0.26 0.16 0.18 0.10

Frequency of trees and

snags
Coniferous trees 0.33 0.02 -0.21 -0.08 -0. 19 0.32
Deciduous trees 0.07 0.22 0.24 0.49 -0.05 0.15
Coniferous snags 0.20 0.24 0.08 -0.22 0.35 -0.32
Deciduous snags 0.08 0.34 0.00 0.30 -0.18 -0.01

 

119

 

Appendix I. Eigenvectors of the first 5 principal components derived from the 1998
habitat variables for 21 wet meadows located in lobes or coves of the northern shoreline
of Lake Huron in Chippewa and Mackinac counties, Michigan.

 

 

j’

Habitat Variable HAB] HAB2 HAB3 HAB4 HAB5
Eigenvalue 6.13 4.66 2.11 1.61 1.22
Percent variance explained 30.64 23.31 10.57 8.06 6.09
Water depth -0.06 -0.33 0.10 0.24 0.33
Hummock height 0.32 -0.20 0.06 -0.11 -0.09
Grass height 0.33 -0.21 0.07 -0.07 0.07
Grass density 0.38 -0.11 0.03 0.00 -0.08
Woody vegetation density 0.00 0.37 0.28 -0.19 0.04
Total vegetation density 0.38 -0.04 0.08 0.00 -0.06
Shrub foliage diversity -0.09 0.31 0.36 -0.17 0.22
Frequency of cover types

Graminoid 0.16
Cattail 0.21 -0.36 -0.03 0.40 -0.06
Bulrush -0.36 -0.04 0.17 0.06 0.10
Floating vegetation 0.00 -0.32 0.29 0.08 -0.37
Submersed vegetation -0.27 -0.10 0.21 0.02 -0.23
Willow -0.04 -0.09 0.49 -0.29 0.32
Alder 0.14 0.35 0.25 0.20 -0.00
Open water 021 -0.20 0.08 0.27 -0.00
Moss -0.32 0.08 -0. 12 0.23 0.0]
Frequency of trees and
snags
Coniferous trees -0. 14 0.26 -0.22 .01 —0.33
Deciduous trees 0.15 0.19 0.31 0.38 -0.08
Coniferous snags 0.11 0.27 0.03 0.01 -0.08
Deciduous snags 0.18 0.19 0.09 0.47 -0.19

 

120

Appendix J. Eigenvectors of the first 5 principal components derived from the landscape
context variables for 21 wet meadows located in coves or lobes of the northern shoreline
of Lake Huron in Chippewa and Mackinac counties, Michigan.

 

 

 

Habitat Variable LAND] LAND2 LAND3 LAND4 LANDS
Eigenvalue 4.87 3.00 2.02 1.56 1.10
Percent variance explained 32.44 20.01 13.45 10.40 7.31
# of adjacent patches 0.40 0.20 -0.03 -0. 16 -0.08
# of adjacent patch types 0.42 0.07 -0.07 -0.14 -0.06
# ofinterfaces 0.40 0.12 -0.13 -0. 16 -0.03
% of patch perimeter
adjacent to:

Urban 0.22 -0.09 -0.32 0.46 -0.38
Non-forested opening 0.19 0.25 -0. 14 0.02 0.03
Forest -0.33 0.26 0.11 0.18 0.25
Open water -0.15 0.00 -0.42 -0.47 -0.09
Forested wetland 0.21 -0.30 0.12 -0.3] 0.49
Bulrush marsh 0.06 0.00 0.61 -0.05 -0.40
Cattail marsh 0.25 -0.30 -0.22 0.15 0.27
Total wetland 0.26 -0.28 0.44 -0. 14 0.02
% ofwet meadow in 0.12 0.35 0.08 0.22 0.54
surrounding landscape
(within l-km of patch
perimeter)
% ofLake I-Iuron in 0.17 0.44 0.02 -0.11 -0.07
surrounding landscape
(within l—km of patch
perimeter)
Road density 0.23 -0.23 0.05 0.50 0.04
fiam density 0.12 0.41 0.16 0.14 0.01
121

a!
5
II
I- J
+

 

Appendix K. Eigenvectors of the first 3 principal components derived from the coastline
complexity variables for 21 wet meadows located in coves or lobes of the northern
shoreline of Lake Huron in Chigipewa and Mackinac counties, Michigan.

 

 

Habitat Variable COAST] COAST2 COAST3
Eigenvalue 4.20 2.40 l .21
Percent variance explained 46.67 26.70 13 .47
Total coastline

500-m segments 0.13 0.54 -0.33

l-km segments 0.09 0.62 0.12

2-km segments 0.] 1 0.49 0.53
Sinuosity

500-m segments 0.41 0.03 -0.37

l-km segments 0.42 -0. 13 0.34

2-km segments 0.31 -0.25 0.48
Fractal dimension

500-m segments 0.40 -0.04 -0.34

l-km segments 0.41 -0.02 -0.03

2-km segments 0.44 -0.09 -0.21

 

122

BIBLIOGRAPHY

123

 

Ll

 

Anderson, M. G., and R. D. Titman. 1992. Spacing Patterns. Pages 251-289 in B. D. J.
Batt, A. D. Afton, M. G. Anderson, C. D. Ankney, D. H. Johnson, J. A. Kaldec, and
G. L. Krapu, eds, Ecology and management of breeding waterfowl. University of
Minnesota Press, Minneapolis, Minnesota, USA.

Andren, H. 1995. Effects of landscape composition on predation rates at habitat edges.
Pages 225-255 in L. Hansson, L. Farhig, and G. Merriam, eds. Mosaic landscapes
and ecological processes. Chapman and Hall, London.

Bayne, E. R. and K. A. Hobson. 1997. Comparing the effects of landscape
fragmentation by forestry and agriculture on predation of artificial nests.
Conservation Biology 1 1:1418-1429.

Bedford, K. W. 1992. The physical effects of the Great Lakes on tributaries and
wetlands. Journal of Great Lakes Research 18:571-589.

Bensch, S., and D. Hasselquist. 1991. Territory infidelity in the polygynous great reed
warbler Arcocephalus arundinaceus: the effect of variation in territory
attractiveness. Journal of Animal Ecology 60:857-871.

Bergin, T. M., L. B. Best, and K. E. Freemarck. 1997. An experimental study of

predation on artificial nests in roadsides adjacent to agricultural habitats in Iowa.
Wilson Bulletin 109:437-448.

Blair, R. B. 1996. Landuse and avian species diversity along an urban gradient.
Ecological Applications 6:506-519.

Blake, J. G. and J. R. Karr. 1987. Breeding birds of isolated woodlots: area and habitat
relationships. Ecology 68: 1724-1734.

Blancher, P. J. and R. J. Robertson. 1985. Site consistency in kingbird breeding
performance: implications for site fidelity. Journal of Animal Ecology 54:1017-
1027.

Bloutin, C. and B. Jobin. 1998. Intensity of agricultural practices and effects on adjacent
habitats. Ecological Applications 82544-557.

Bock, C. E., J. H. Bock, and B. C. Bennett. 1999. Songbird abundance in grasslands at a
suburban interface on the Colorado high plains. Studies in Avian Biology 19:131-
136.

Bolger, D. T., T. A. Scott, and J. T. Rotenberry. 1997. Breeding bird abundance in an

urbanizing landscape in coastal southern California. Conservation Biology 11:406-
421.

124

 

 

 

Bowers, M. A. and B. Breland. 1996. Foraging of gray squirrels on and urban-rural
gradient: use of the GUD to assess anthropogenic impact. Ecological Applications
621135-1142.

Brewer, R., G. A. McPeek, and R. J. Adams. 1991. The atlas of breeding birds of
Michigan. Michigan State University Press, East Lansing, Michigan, USA.

Brown, M. and J. J. Dinsmore. 1991. Area-dependent changes in bird densities in Iowa
marshes. Journal of the Iowa Academy of Sciences. 98:124-126.

Brown, M. and J. J. Dinsmore. 1986. Implications of marsh size and isolation for marsh
bird management. Journal of Wildlife Management 50:392-3 97.

Burhans, D. E. and F. R. Thompson 111. 1999. Habitat patch size and nesting success of
Yellow-breasted Chats. Wilson Bulletin 111:210-215.

Burton, T. M. 1985. The effects of water level fluctuations on Great Lakes coastal
marshes. Pages 3-14 in Prince, H. H. and F. M. D'Itri eds, Coastal wetlands.
Lewis Publishers, Inc., Chelsea, Michigan.

Cantero, J. J ., M. Partel, and M. Zobel. 1999. Is species richness dependent on the

neighboring stands? An analysis of the community patterns in mountain grasslands
of central Argentina. Oikos 87:346-354.

Conway, C. J., G. V. Powell, and J. D. Nichols. 1995. Overwinter survival of
Neotropical migratory birds in early-successional and mature tropical forests.
Conservation Biology 9:855-864.

Craig, R. J. and K. G. Beal. 1992. The influence of habitat variables on marsh bird
communities of the Connecticut River estuary. Wilson Bulletin 104:295-311.

Diamond, J. M. and R. M. May. 1977. Species turnover rates on islands: dependence on
census interval. Science 197:266-270.

Donovan, T. M., R. H. Lamberson, A. Kimber, F. R. Thompson 111, and J. Faaborg.
1995. Modeling the effects of habitat fragmentation on source and sink
demography of Neotropical migrant birds. Conservation Biology 9: 1396-1407.

Drolet, B., A. Desrochers, and M. Fortin. 1999. Effects of landscape structure on nesting
songbird distributions in a harvested boreal forest. Condor 101:699-704.

Dunning, J. B. Jr., R. Borgellas, Jr., K. Clements, and G. K. Meffe. 1995. Patch

isolation, corridor effects, and colonization by a resident sparrow in a managed pine
woodland. Conservation Biology 92542-550.

125

 

 

Estandes, C. F. and S. A. Temple. 1999. Deciduous-forest bird communities in a
fragmented landscape dominated by exotic pine plantations. Ecological
Applications 91573-585.

Faaborg, J. M. Brittingham, T. Donovan, and J. Blake. 1995. Pages 357-380 in Martin,
T. E. and D. M. Finch eds, Ecology and management of Neotropical migrant birds.
Oxford University Press, New York.

Flather, C. H. and J. R. Sauer. 1996. Using landscape ecology to test hypotheses about
large-scale abundance patterns in migratory birds. Ecology 77 :28-35.

Forman, R. T. T. 1995. Land mosaics: the ecology of landscapes and regions.
Cambridge University Press, New York, New York.

Forman, R. T. T., and L. E. Alexander. 1998. Roads and their major ecological effects.
Annual Review of Ecology and Systematics 29:207-231.

Forman, R. T. T. and M. Godron. 1986. Landscape Ecology. John Wiley and Sons,
New York.

Forys, E. and S. R. Humphrey. 1999. The importance of patch attributes and context to
the management and recovery of an endangered lagomorph. Landscape Ecology
14:177-185.

Freemark, K. E., J. B. Dunning, S. J. Hejl, and J. R. Probst. 1995. A landscape ecology
perspective for research, conservation and management. Pages 381-427 in Martin,
T. E. and D. M. Finch eds, Ecology and management of Neotropical migrant birds.
Oxford University Press, New York.

Friesen, L. E., P. F. Eagles, and R. J. MacKay. 1995. Effects of residential development
on forest-dwelling Neotropical migrant songbirds. Conservation Biology 9: 1408-
1414.

Galli, A. E., C. F. Leck, and R. T. T. Forman. 1976. Avian distribution patterns in forest
islands of different sizes in central New Jersey. Auk 93:356-364.

Gathman, J.P., T.M. Burton, and B]. Armitage. 1999. Coastal wetlands of the Upper
Great Lakes: distribution of invertebrate communities in response to environmental
variation. In: Batzer, D.P., RB. Rader, S.A. Wissinger (eds). Invertebrates in
freshwater wetlands of North America. John Wiley and Sons, Inc., New York, New
York, USA.

Gibbs, J. P. 1993. Importance of small wetlands for the persistence of local populations
of wetland-associated animals. Wetlands 13:25-31.

126

-l

 

 

Gibbs, J. P. and S. M. Melvin. 1993. Call-response surveys for monitoring breeding
waterbirds. Journal of Wildlife Management 57:27-34.

Gill, F. B. 1995. Ornithology. W. H. Freeman and Company, New York, New York.

Gutzwiller, K. J. and S. H. Anderson. 1987a. Multi-scale associations between cavity-
nesting birds and features of Wyoming streamside woodlands. Condor 89:534-
548.

Gutzwiller, K. J. and S. H. Anderson. 1987b. Short-term dynamics of cavity-nesting
bird communities in disjunct floodplain habitats. Condor 892710-720

Gutzwiller, K. J. and S. H. Anderson. 1992. Interception of moving organisms:
influences of patch shape, size and orientation on community structure. Landscape
Ecology 6:293-303

Haig, S. M., D. W. Mehlman, and L. W. Oring. 1998. Avain movements and wetland
connectivity in landscape conservation. Conservation Biology 12:749-758.

Hansen, A. J. and D. L. Urban. 1992. Avian response to landscape pattern: the role of
species' life histories. Landscape Ecology 7: 163-180.

Hauber, M. E. and S. A Russo. 2000. Perch proximity correlates with higher rates of
cowbird parasitism of ground nesting Song Sparrows. Wilson Bulletin 112:150-
153.

Helzer, C. J. and D. E. Jelinski. 1999. The relative importance of patch area and
perimeter-area ratio to grassland breeding birds. Ecological Applications 9: 1448-
1458.

Herdendorf, C. E. 1992. Lake Erie coastal wetlands: and overview. Journal of Great
Lakes Research 18:533-551.

Herkert, J. R. 1994. The effects of habitat fragmentation on midwestem grassland bird
communities. Ecological Applications 4:461-471 .

Hoover, J. P., M. C. Brittingham, and L. J. Goodrich. 1995. Effects of forest patch size
on nesting success of wood thrushes. Auk 1 12: 146-155.

Hosmer, D. W. and S. Lemeshow. 1989. Applied logistic regression. John Wiley and
Sons, New York, New York, USA.

Jobin, B. and J. Picman. 1997. Factors affecting predation on artificial nests in marshes.
Journal of Wildlife Management. 61:792-800.

Kaldec, J. A. and L. M. Smith. 1992. Habitat management for breeding areas. Pages
590-610 in B. D. J. Batt, A. D. Afton, M. G. Anderson, C. D. Ankney, D. H.

127

 

‘

 

 

Johnson, J. A. Kaldec, and G. L. Krapu, eds, Ecology and management of breeding
waterfowl. University of Minnesota Press, Minneapolis, Minnesota, USA.

Kent, C. and J. Wong. 1982. An index of littoral zone complexity and its measurement.
Canadian Journal of Fisheries and Aquatic Science. 39:847-853.

Kiester, A. R. 1971. Species diversity of North American amphibians and reptiles.
Systematic Zoology 20: 127-137.

Knopf, F. A. and J. A. Sedgewick. 1992. An experimental study of nest-site selection by
yellow warblers. Condor 94:734-742.

Lam, N. S. and L. De Cola. 1993. Fractals in geometry. Prentice Hall, Englewood
Cliffs, California, USA.

Liu, J. and P. S. Ashton. 1999. Simulating effects of landscape context and timber
harvest on tree species diversity. Ecological Applications 9: 1 86-201.

MacArthur, R H. and E. O. Wilson. 1967. The theory of island biogeography.
Princeton Universtiy Press, Princeton, New York.

Martin, T. E. and D. M. Finch. 1995. Ecology and management of Neotropical migrant
birds. Oxford University Press, New York, USA.

 

Milne, B. T. and R. T. T. Forman. 1986. Peninsulas in Maine: woody plant diversity,
distance, and environmental patterns. Ecology 67:967-974.

Minc, L. D. and D. A. Albert. 1998. Great Lakes coastal wetlands: abiotic and floristic
characterization. Michigan Natural Features Inventory Report, Lansing, Michigan.

Mitsch, W. J. and J. G. Gosselink. 1993. Wetlands. Van Nostrand Reinhold, New York,
New Yorlg USA.

Mladenoff, D. J ., M. A. White, T. R. Crow, and J. Pastor. 1994. Applying principles of
landscape design and management to integrate old-growth forest enhancement and
commodity use. Conservation Biology 82752-762.

Morrison, M. L., B. G. Marcot, and R. W. Mannan. 1998. Wildlife-habitat relationships:
concepts and applications. University of Wisconsin Press, Madison, Wisconsin,
USA.

Myers, R. H. 1989. Classical and modern regression with applications. PWS-Kent
Publishing Company, Boston, Massachusetts, USA.

128

 

Naugle, D. E., K. F. Higgins, S. M. Nusser, and W. C. Johnson. 1999. Scale-dependent
habitat use in three species of prairie wetland birds. Landscape Ecology 14:267-
276.

Newman, S., J. Schuette, J. B. Grace, K. Rutchey, T. Fontaine, K. R. Reddy, and M.
Pietrucha. 1998. Factors influencing cattail abundance in the northern Everglades.
Aquatic Botany 60:265-280.

Nilsson, L. 1978. Breeding waterfowl in eutrophicated lakes in south Sweden.
Wildfowl 29:101-110.

Orians, G. H. 1980. Some adaptations of marsh-nesting blackbirds. Monographs in
population biology #14. Princeton University Press, Princeton, New Jersey, USA.

Ott, L. 1988. An introduction to statistical analysis. PWS-Kent Publishing Company,
Boston, Massachusetts.

Paton, P. W. C. 1994. The effect of edge on avian nest success: how strong is the
' evidence? Conservation Biology 8: 17-26.

Patten, M. A. and J. T. Rotenberry. 1998. Post-disturbance changes in a desert breeding
bird community. Journal of Field Ornithology 69:6‘4-625.

Pearson, S. M. 1993. The spatial extent and relative influence of landscape-level factors
on wintering bird populations. Landscape Ecology 823-18.

 

Picman, J. 1988. Experimental study of predation on eggs of ground-nesting birds:
effects of habitat and nest distribution. Condor 90:124-131.

ALI

Prince, H. H. and C. S. Flegel. 1995. Breeding avifauna of Lake Huron. Pages 247-272
in M. Munawar, T. Edsall, and J. Leach, eds, The Lake Huron ecosystem: ecology,
fisheries and management. SPB Academic Publishing, Amsterdam, The
Netherlands.

Prince, H. H., P. I. Padding, and R. W. Knapton. 1992. Waterfowl use of the Laurentian
Great Lakes. Journal of Great Lakes Research 18:673-699.

Pulliam, H. R. 1988. sources, sinks and population regulation. American Naturalist
132:652-661.

Rahbek, C. 1997. The relationship among area, elevation, and regional species richness
in neotropical birds. American Naturalist 149:875-902.

 

Rencher, A. C. 1992. Interpretation of canonical discriminant functions, canonical
variates, and principal components. The American Statistician 46:217-225.

129

Rice, J., R. D. Ohmart, and B. A. Anderson. 1983. Turnovers in species composition of
avian communities in contiguous riparian habitats. Ecology 64: 1444-1455.

Riffell, S. K., K. J. Gutzwiller, and S. H. Anderson. 1996. Does repeated human
intrusion cause cumulative declines in avian richness and abundance. Ecological
Applications 62492-505.

Risch, S. J ., D. Andow, and M. A. Altieri. 1983. Agroecosystem diversity and pest
control: data, tentative conclusions, and new research direction. Environmental
Entomology 12:625-629.

Robbins, C. S. 1981. Bird activity levels related to weather. Studies in Avian Biology
62301-310.

Robinson, S. K., F. R. Thompson, T. M. Donovan. D. R. Whitehead, and J. Faaborg.
1995. Regional forest fragmentation and nesting success of migratory birds.
Science 267: 1987-1990.

Rosenberg, K. V., J. D. Lowe, and A. A. Dhondt. 1999. Effects of forest fragmentation
on breeding tanagers: a continental perspective. Conservation Biology 13:568-583.

Rotenberry, J. T. and J. A. Wiens. 1980 Habitat structure, patchiness, and avian
communities in North American steppe vegetation: a multivariate analysis.
Ecology 61:1228-1250.

Russell, G. J ., J. M. Diamond, S. L. Pimm, and T. M. Reed. 1995. A century of
turnover: community dynamics at three timescales. Journal of Animal Ecology
64:628-641.

Saab, V. A. 1999. Importance of spatial scale to habitat use by breeding birds in riparian
forests: a hierarchical analysis. Ecological Applications 9: 135-151.

SAS Institute, Inc. 1989. SAS/STAT user's guide, version 6, 4th edition, volume 2.
SAS Institute, Inc., Cary, North Carolina, USA.

SAS Institute, Inc. 1990. SAS/STAT software: changes and enhancements through
release 6.12. SAS Institute, Inc., Cary, North Carolina, USA.

Savidge, J. 1987. Extinction of an island avifauna by an introduced snake. Ecology
68:660-668.

Schwarz, M. W. 1999. Choosing the appropriate scale of reserves for conservation.
Annual Review of Ecology and Systematics. 30:83-108.

Semlitsch, R. D. and J. R. Bodie. 1998. Are small, isolated wetlands expendable?
Conservation Biology 12: 1 129-1 133.

130

.....‘

A;

 

Skagen, S. K. and F. L. Knopf. 1994. Migrating shorebirds and habitat dynamics at a
prairie wetland complex. Wilson Bulletin 106:91-105.

Szaro, R. C. and M. D. Jackle. 1985. Avian use of a desert riparian island and its
adjacent scrub habitat. Condor 87:511-519.

Taylor, P. D. 1993. Connectivity is a vital element of landscape structure. Oikos
68:571-573.

Taylor, R. J. 1987. Geometry of colonization: 2. Peninsulas. Oikos 48:232-237.

Taylor, R. J, and P. J. Regal. 1978. The peninsular effect on species diversity and the
biogeography of Baja California. American Naturalist 112:583-593.

Terres, J. K. 1980. The Audubon Society encyclopedia of North American birds. Alfred
A. Knopf, New York, New York.

Tilton, D. L. and B. R. Schwegler. 1978. The values of wetland habitat in the Great
Lakes basin. Pages 267-277 in P. E. Greeson, J. R. Clark, and J. E. Clark, eds,
Wetland function and values: the state of our understanding. American Water
Resources Association, Minneapolis, Minnesota, USA.

Trexler, J. C. and J. Travis. 1993. Nontraditional regression analysis. Ecology 74:1629-
1637.

van Dorp, D. and P. F. M. Opdam. 1987. Effects of patch size, isolation and regional
abundance on forest bird communities. Landscape Ecology 1:59-73.

Verner, J. 1985. Assessment of counting techniques. Pages 247-302 in R. F. Johnston,
editor. Current Ornithology, Volume 2. Plenum Press, New York, New York,
USA.

Vemer, J. and M. F. Willson. 1966. The influence of habitats on mating systems of
North American passerine birds. Ecology 47: 143-147

Vickery, P. D., M. L. Hunter, and S. M. Melvin. 1994. Effects of habitat area on the
distribution of grassland birds in Maine. Conservation Biology 8:1087-1097.

Vickery, P. D., P. L. Tubaro, J. M. Cardoso da Silva, B. G. Peterjohn, J. R. Herkert, and
R. B. Cavalcanti. 1999. Conservation of grassland birds in the western
hemishpere. Studies in Avian Biology 19:2-26.

Villard, M. A., M. K. Trzcinski, and G. Merriam. 1999. Fragmentation effects on forest
birds: relative influence of woodland cover and configuration on landscape
occupancy. Conservation Biology 13:774-783.

131

 

 

 

 

Waide, R. B. and P. M. Naims. 1988. Tropical forest bird counts and the effect of sound
attenuation. Auk 105:296-302.

Webb, N. R. and P. J. Hopkins. 1984. Invertebrate diversity on a fragmented Calluna
heathland. Journal of Applied Ecology 21:921-933.

Wiens, J. A. 1981. Single-sample surveys of communities: are the revealed patterns
real? American Naturalist 117:90-98.

Wiens, J. A. 1996. Wildlife in patchy environments: metapopulations, mosaics, and
management. Pages 53-84 in D. R. McCullough, ed. Metapopulations and wildlife
conservation. Island Press, Washingtion, D. C., USA.

Weller, M. W. 1994. Bird-habitat relationships in a Texas estuarine marsh during
summer. Wetlands 14:293-300.

Weller, M. W. 1999. Wetland birds. Cambridge University Press, Cambridge, United
Kingdom.

Westmoreland, D., and L. B. Best. 1985. The effect of disturbance on Mourning Dove
nesting success. Auk 102:774-780.

Zar, J. H. 1984. Biostatistical analysis. Second edition. Prentice-Hall, Englewood
Cliffs, California.

132

 

 

"111111111111111117111“