Avian body-size clumps and the response of birds to scale-dependent landscape structure in suburban habitats

MISSING IMAGE

Material Information

Title:
Avian body-size clumps and the response of birds to scale-dependent landscape structure in suburban habitats
Physical Description:
xiii, 246 leaves : ill. ; 29 cm.
Language:
English
Creator:
Hostetler, Mark
Publication Date:

Subjects

Subjects / Keywords:
Zoology thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Zoology -- UF   ( lcsh )
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1997.
Bibliography:
Includes bibliographical references (leaves 225-245).
Statement of Responsibility:
by Mark E. Hostetler.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 028024209
oclc - 37860074
System ID:
AA00012938:00001


This item is only available as the following downloads:


Full Text









AVIAN BODY-SIZE CLUMPS AND THE RESPONSE OF BIRDS TO SCALE-
DEPENDENT LANDSCAPE STRUCTURE IN SUBURBAN HABITATS










By


MARK E. HOSTETLER


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA


1997












ACKNOWLEDGMENTS


"What a strange (but wonderful) trip this has been." I thank
my advisor, Dr. C. S. "Buzz" Holling, for his guidance and support
throughout this study. His vision and expertise allowed me to ask
questions and to conduct research at large scales. Making the jump
from the certainty of controlled, small-scaled experimental research
to research at large scales that does not lend itself to experiments
was "gut-wrenching" and caused much hair loss during the course of
my study. However, this brand of research permitted me to ask
interesting and meaningful questions. Buzz truly taught this
experimental biologist to "embrace uncertainty." I also thank the
members of my committee, Drs. C. Lanciani, D. Levey, P. Frederick,
and R. Kiltie for their helpful comments and suggestions in designing
the study, analyzing the data, and editing the manuscript.
I also thank a whole host of scientists that provided me with
suburban bird data to include in my analyses: S. T. Penland (Seattle,
WA); R. M. Degraaf (Amherst and Springfield, MA); C. W. Sexton
(Austin, TX); V. J. Lucid (Blacksburg, VA); R. W. Guth (Chicago, IL); R.
K. Lancaster and W. C. Weber (Vancouver, B.C.); and G. S. Mills, J. B. J.
Dunning and J. M. Bates (Tucson, AZ). In addition, I thank several
local bird experts that helped me refine my regional bird lists: Jamie
Smith, Richard Cannings, and Christine Adkins of Vancouver; Steve








Russell of Tucson;.-Curtis Adkisson of Blacksburg; Jeff Skriletz of
Seattle; and H. D. Bohlen of Chicago.
I am grateful to my girlfriend, Meryl Klein, for her support and
assistance during this project. Her eternal optimism and confidence
helped me through many of the rough spots. Many thanks go to my
colleagues in Holling's lab, P. Marples, J. Sendzimir, G. Peterson, and L.
Gunderson for their constructive comments, suggestions, and for the
many fruitful interactions that I have had with them. Special thanks
goes to my good friend Paul Marples who not only developed the
statistical clump analyses used in my research, but the many beer
sessions allowed me to bounce ideas off of him, to vent my
frustrations, and to generally keep chugging on down road.
Additional thanks to the graduate students, faculty, and staff in the
zoology department. The interactions that I have had with them
helped me become a mature and productive scientist.
I especially thank my friends and my family for their support
and encouragement. My parents, grandparents, and my brother
helped me keep things in perspective throughout this process.













TABLE OF CONTENTS


ACKNOWLEDGEMENTS .......................................................... ..............................ii

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

LIST OF FIGU RES................................................................... ................................ x

AB STRACT....................................................... .................................... xiii

CHAPTERS

1 INTRODUCTION AND OVERVIEW.......................................................... 1

2 DETECTION OF HIERARCHICAL LANDSCAPE STRUCTURE USING
DISCONTINUITIES IN AVIAN BODY-SIZE DISTRIBUTIONS

Introduction.......................................................................................................... 1 1
M ethods.................................................................... ................................ 1 8
Methodology of Clump Analysis ..................................... ............. 1 8
Clump Analyses of North American Temperate Forest
Avian Communities .................................................................................. 29
Comparison Between the Clumps of a Desert and Several
Temperate Forest Avian Communities........................................3 3
Densities of Species in Body-size Clumps of Each
Ecoregion................................. ................................................ 34
R esults...... .................................................................. ..................................... 3 5
Body-size Clumps of the Temperate Forest Biome .................... 35
Amherst (Temperate Forest Biome) vs. Tucson (Desert
B iom e) ............................................................... ............................... 3 8
Species Densities in Each Body-size Clump................................... 4 1
D iscussion....................................... ........... ............................................ 4 2
Body-size Clump Matches........................... ................................. 43
Quantity of Species in Each Clump.......................................... 49
Revisiting Body-size Clump Patterns & Future Research............5 1
Sum m ary.......................................................................................................... 5 5









3 BODY-SIZE CLUMPS OF NEW WORLD AND OLD WORLD AVIAN
COMMUNITIES

Introduction.................................................................. ............................ 5 8
M eth o d s.................................................................................................................. 6 2
B etw een C ontinents.......................................... .................................... 6 6
Within Continents...................................................... 67
R esu lts..................................................................................................................... 6 7
B etw een C ontinents................................................ .............................. 6 7
Within Each Continent............................. ............................... 72
Trends in Body-size Clump Distributions...................... ............. 7 2
D iscussion.............................................................. ............................................ 7 6
Cross Continental Body-size Clump Comparison........................... 77
Within Continental Body-size Clump Comparison..................... 83
Future Research ..................................... ...... .............................. 85
Summary and Conclusions.......................... ................................. 86

4 RESPONSES OF BIRDS TO SCALE-DEPENDENT LANDSCAPE
STRUCTURE

Introduction............................................................... .................................... 8 8
M eth o d s.................................................................................................................. 9 6
O overview ............................................................................... 96
Suburban Bird Data Sets............................ ............................ 98
Body-size Clumps .................................................................................... 102
Measurement of Canopy Patches from Aerial
Photographs............................................................................................ 10 2
Exploring the Spatial Areas and Patch Sizes of Birds in
Different Size Categories............................................................. 07
R esu lts ..................................................... ............................ ........... 1 1 0
Bird Counts for Suburban Sites ......................................................... 110
Percent Tree Canopy Cover of Each Patch Size Category........... 111
Spring Surveys: Principle Component Analyses (PCA) and
Multiple Regression Analyses.................. ............................ 116
Summer Surveys: Principle Component Analyses (PCA)
and Multiple Regression Analyses................................................12 9
D iscussion................................................................. .............................. 139
Trends in the Spring and Summer..............................................140
Avian Conservation and Metapopulation Dynamics.................145
Problems with Measuring the Scales Relevant to Avian
Species................................ ............... ....................................14 9








Bird Clump Analyses and Generalizations to Other Avian
C om m unities.......................................... ........................................... 55
Future Research on Determining Scales Relevant to Avian
Species................................ ................................................ 60
Summary and Conclusions......................... .......................... 62

5 AVIAN BODY-SIZE CLUMPS AND THE IMPACT OF SUBURBAN
LANDSCAPES ON AVIAN COMMUNITIES: A SYNTHESIS

Introduction........................................................................................................ 6 5
Avifauna Studies in Suburban Environments................................ 67
Avian Body-Size Clump Structure and Suburban
L andscapes.... ................................................................................... 170

APPENDIX A
Avifauna of various ecoregions of North America..........................177

APPENDIX B
New world and old world avifauna of the temperate biome........1 82

APPENDIX C
New world and old world avifauna of the boreal biome......1......189

APPENDIX D
Sites and dates of surveys used in the spring and summer
data lists............................................ ............................. ........ ........ 1 9 4

APPENDIX E
Suburban spring bird surveys........................ ................................. 96

APPENDIX F
Suburban summer bird surveys........................ .............................203

APPENDIX G
Percent tree canopy cover for spring survey sites.....................2..... 1 6

APPENDIX H
Percent tree canopy cover for summer survey sites........................ 221

LIST OF REFERENCES............................................. ............ 2 5

BIOGRAPHICAL SKETCH.................................................................................2 46









LIST OF TABLES


Table pag

2-1. A qualitative vegetative and climatic description of six
different ecoregions of North America ............................................... 32

2-2. Number of body-size clumps detected for each ecoregion from
the Gap Rarity Index.......................... ...... ............................ 3 6

2-3. Size range for each of the 10 avian body-size clumps
determined from North American temperate forest avifauna
lists ..................................................... ............................................................... 3 9

2-4. Number of species recorded for each bird-size category from
North American avifauna lists.......................... ................................... 42

3-1. The size ranges of European and eastern North American
temperate forest avian body-size clumps. The number of species
in each clump is also listed............................................ ........... .............. 7 0

3-2. The size ranges of temperate forest avian body-size clumps
determined from Amherst, Austin, and Vancouver avifauna lists
(Chapter 2) compared to those from eastern North American
temperate forest avifauna lists........................................ .................. 1

3-3. The size ranges of European and eastern North American
boreal forest avian body-size clumps. The number of species in
each clump is also listed................................. ..... ................7 4

4-1. Tree canopy patch categories for each of the four spatial areas
(0.2 km2 and 1.5 km2, 25.0 km2, and 85.0 km2)..............................106

4-2. Bird counts (#birds/hour) of different size categories from
Spring bird censuses. All birds characterized as primarily tree
canopy users*........................................................ ...................................... 12

4-3. Bird counts (#birds/hour) of different size categories from
Summer bird censuses. All birds characterized as primarily tree
canopy users*................................... ........................................................... 1 14

4-4. Squared correlation coefficients (r2) for bird-size categories
from multiple regression equations between Spring bird counts








(#birds/hr) and percent tree canopy coverage at several different
scales.............................................................................................................. 1 8

4-5. Standardized partial regression coefficients (b) for bird-size
categories from multiple regression analyses of Spring bird counts
(#birds/hr) and percent tree canopy coverage at 1:2400 (0.2
km 2) ................................................ ....... ......... ....................................... 1 9

4-6. Standardized partial regression coefficients (b) for bird-size
categories from multiple regression analyses of Spring bird counts
(#birds/hr) and percent tree canopy coverage at 1:2400 (1.5
k m 2) ........................................ ............. .................. .............................. 1 2 0

4-7. Standardized partial regression coefficients (b) for bird-size
categories from multiple regression analyses of Spring bird counts
(#birds/hr) and percent tree canopy coverage at 1:40,000 (85.0
k m 2) ........................................... ............... .................................................... 12 1

4-8. Standardized regression coefficient (b) for the bird-size
category of > 185.0 g from multiple regression analysis of Spring
bird counts (#birds/hr) and percent tree canopy coverage at
1:24,000 (25.0 km 2) ...................................... ................................... 121

4-9. Squared correlation coefficients (r2) and standardized partial
regression coefficients (b) from multiple regression of PCA axes
that gave the highest partial regression coefficients at each of the
spatial areas during the Spring............................................................. 124

4-10. Squared correlation coefficient (r2) and standardized partial
regression coefficients (b) from multiple regression of PCA axes
that gave the highest partial regression coefficients at each of the
spatial areas during the Spring.............................................................. 24

4-11. Summary of how birds respond to canopy cover during the
spring season................................................... ..................................... 12 8

4-12. Squared correlation coefficients for bird-size categories from
multiple regression equations between Summer bird counts
(#birds/hr) and percent tree canopy coverage at several different
scales............................... ......... ......... ... ...................................... 1

4-13. Standardized partial regression coefficients (b) for bird-size
categories from multiple regression analyses of Summer bird
counts (#birds/hr) and percent tree canopy coverage at 1:2400
(0.2 km 2) ............................... .............. .............................................. 132







4-14. Standardized partial regression coefficients (b) for bird-size
categories from multiple regression analyses of Summer bird
counts (#birds/hr) and percent tree canopy coverage at 1:2400
(1.5 km 2) ................................................ ................................................ 13 3

4-15. Squared correlation coefficients (r2) and standardized partial
regression coefficients (b) from multiple regression of PCA axes
that gave the highest partial regression coefficients at each of the
spatial areas during the Spring............................................................... 134

4-16. Summary of how birds responded to canopy cover during the
sum m er season.................................................. .................................. 1 3 8













LIST OF FIGURES


Figurage

2-1. A hypothetical data set fitted with a 2 mode kernel
estimate. Each clump is defined by inflection points that
give maximal and minimal slope.............................................................. 24

2-2. Body-size clump structure of three different ecoregions of the
North American temperate forest biome.......................................... 37

2-3. Avian body-size clump structure of two different ecoregions;
Tucson avifauna regional list is in the desert biome and the
Amherst avifauna regional list is in the temperate forest
biom e ................................. .. .................................................. 4 0

3-1. Outline of the boreal forest biome in eastern North America
and Europe ................................................................ .............................. 6 3

3-2. Outline of the temperate forest biome in eastern North
America and Europe..................................................... ........................... 64

3-3. Avian body-size clump structure of the eastern North
American and the European temperate forest biome. North
American body-size clump structure has 10 clumps and the
European body-size clump structure has 7 clumps .......................... 69

3-4. Avian body-size clump structure of the eastern North
American and the European boreal forest biome. North American
body-size clump structure has 10 clumps and the European body-
size clump structure has 9 clumps ..................................... .............. 73

3-5. Avian body-size clump structure of the eastern North
American and the European boreal and temperate forest biome.75

4-1. Different scale-dependent landscape structures perceived by a
wren and a hawk.................................... ................ .............................. 89








4-2. Plot of squared correlation coefficients from simple regressions
of bird size category 2 and patch size categories from Spring
m easurem ents ..............................................................................................125

4-3. Plot of squared correlation coefficients from simple
regressions of each bird-size category and patch size category
from Spring measurements ............................................................... 126

4-4. A plot of the relationship between bird-size category and the
upper limit of best patch size during the spring..............................27

4-5. Plot of squared correlation coefficients from simple
regressions of each bird-size category and patch size category
from Summer measurements .......................................................... 136

4-6. Plot of squared correlation coefficients from simple
regressions of each bird-size category and patch size category
from Summer measurements .......................................................... 137

4-7. Three hypothetical areas where birds were censused.............1 52

4-8. Representation of two possible perceptions of the landscape by
birds: A) a group of patches that actually are viewed as one big
patch, and B) one big patch that actually is viewed as one little
p atch ...................................................... ......................................................... 1 5 4













Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree Doctor of Philosophy


AVIAN BODY-SIZE CLUMPS AND THE RESPONSE OF BIRDS TO SCALE-
DEPENDENT LANDSCAPE STRUCTURE IN SUBURBAN HABITATS

By

Mark E. Hostetler

August, 1997




Chairman: C. S. Holling
Major Department: Zoology


Habitat selection is an important element in the spatial
distribution and population dynamics of avian communities in
heterogeneous landscapes. However, little empirical information
exists about how birds respond to landscape structure at large scales.
Previous research has shown that avian body-size clumps may
reflect the hierarchical structure of landscapes. Body-size clumps are
essentially groups of similar-sized adult birds that theoretically
respond to structure at the same scale. The objective of this project
was to test whether clumps reflect hierarchical landscape structure
and whether they could be used to reveal at what scale different
bird species are responding to suburban landscape structure.








Chapters 2 and 3 use a new statistical technique to determine
whether avian body-size clumps exist for chosen faunal lists. In
comparisons of avifauna lists derived from the same biome
(temperate forest and boreal), results indicated that body-size
clumps occurred at similar body-size ranges. This was interpreted as
evidence that body-size clumps reflect the hierarchical structure of
landscapes. In cross-continental comparisons, fewer European than
eastern North American medium-sized bird species were present.
European forests were historically much more fragmented than
eastern North American forests. It was hypothesized that the
fragmented European landscape contained disproportionately less
resources for medium-sized birds.
In chapter 4, I explored at what scale birds in different body-
size clumps responded to landscape structure during the summer
and spring. From several North American suburban landscapes, I
used the amount of tree canopy patches at four different spatial
scales (0.2 km2, 1.5 km2, 25.0 km2, and 85.0 km2) to predict the
variation of bird counts in these landscapes. Birds were grouped into
body-size clumps determined for the temperate forest biome
(Chapters 2 & 3). From multiple regression analyses, it was
determined that birds of each size clump responded to tree canopy
cover at different scales. The results provide a framework for
researchers to conduct future research on how large-scale changes in
the landscape affect different bird species. Also, the results can
serve as guidelines to design suburban landscapes to attract a
greater diversity of birds.











CHAPTER 1


INTRODUCTION AND OVERVIEW


Birds must respond to landscape structure at a variety of
spatial scales (e.g., 1.0 m2 to 1.0 km2). The range of scales sampled
by a bird is defined by the spatial extent, or largest area that is
sampled by each bird, and the spatial grain, or minimum size of
objects sampled by each bird (Kotliar and Wiens 1990, Wiens 1990).
Throughout this dissertation, the term "response" is defined as the
ability of a bird to "perceive" or utilize structure (e.g., tree canopy
patches) in a landscape. Operationally, response is measured by the
quantity of birds that occur in a particular area. The quantity of
birds in a given area depends on whether sufficient structure (e.g.,
distribution of tree canopy patches, natural ground cover patches, or
other vegetative features) exists at a scale range relevant to a
particular species. Wiens (1989b) proposed that the range of scales
pertinent to a particular bird species is dependent on the size of the
bird. Theoretically, smaller birds respond to structure at smaller
scales than larger birds. For example, to select a home range, a wren
probably samples the distribution of individual bushes and trees
whereas a hawk samples much larger groupings of trees (e.g.,
riparian habitat). However, exactly how large an area and what sizes
of objects (e.g., trees and bushes) are sampled by different species is
largely unknown. Few empirical studies address the range of scales
relevant to different sizes of birds (Wiens 1995, Opdam 1991,








Harrison 1992). In particular, empirical studies that address how
birds or other organisms respond to structure at large scales (e.g.,
above the level of the home range) are virtually nonexistent (Lima
and Zollner 1996).
The range of scales relevant to different organisms is difficult
to measure. Hierarchy theory, though, could provide a framework to
simplify such analyses. In hierarchy theory, complex ecosystems are
decomposed into discrete hierarchical levels. Each level represents a
scale at which biotic and abiotic factors interact much more strongly
with each other than with other factors at other levels (Allen and
Starr 1982, O'Neill et al. 1986). Hierarchical levels in an ecosystem
can be defined by a discontinuous frequency distribution of process
rates (Allen and Starr 1982, O'Neill et al. 1986, Holling 1992a) or by
a discontinuous frequency distribution of object sizes in a landscape
(O'Neill et al. 1986, Holling 1992b). Process rates (e.g., plant growth,
fire regimes, erosion) are considered to be discontinuous when they
operate at distinctly different temporal speeds (e.g., years, decades,
centuries). Likewise, sizes of landscape objects (e.g., leaves, crown
volume, stands of trees) are considered to be discontinuous when
several modes appear in a frequency distribution of their sizes.
Thus, each hierarchical level contains a specific category of processes
and/or a defined size range of landscape objects.
A hierarchical organization of ecosystems has been inferred
from empirical studies of landscapes and from models that captured
the dynamic behavior of managed ecosystems (Allen and Starr 1982,
O'Neill et al. 1986, Holling 1992a, Kent and Wong 1982, Bradbury and
Reichelt 1983, Holling 1986). In some empirical studies of







landscapes, fractal dimensions have been used to detect hierarchical
levels (e.g., Kent and Wong 1982, Bradbury and Reichelt 1983). A
fractal dimension measures the degree of complexity in a landscape,
and significantly different fractal dimensions theoretically represent
separate hierarchical levels. For example, Kent and Wong (1982)
calculated two different fractal dimensions of lake shorelines. The
fractal dimensions corresponded to glacial processes that shape lake
shorelines at large scales and to erosional processes at smaller scales.
In modelling studies, models were able to describe the dynamic
behavior of several managed ecosystems with only 3-4 sets of
variables (see review in Holling 1986). For example, boreal forest-
insect dynamics could be described by the interaction of four
dominant cycles (Holling 1992a). These cycles include a 3-5 yr cycle
controlled by the interaction of insects, parasites, and needles; a 10-
15 yr cycle caused by the interaction between insect defoliation and
the regeneration of crown foliage that results in insect outbreaks
within limited areas; a 35-40 yr cycle represented by the gradual
growth of trees and insect outbreaks that cause tree mortality over
an extensive area; and a > 80 yr cycle represented by a long
successional period and tree longevity.
The above studies suggest that ecosystems could be
hierarchically structured, but more independent tests are needed. If
ecosystems were hierarchically organized, then separate groups of
species may respond to structure primarily at different hierarchical
levels (Holling 1992b). If so, Holling (1992b) hypothesized that
animal masses would be grouped into several distinct "clumps."
Clumps are essentially groups of similar-sized adult animals that are







separated by significant gaps in a body-mass distribution. Body-size
clumps are hypothesized to reflect both qualitative and quantitative
structural properties of ecosystems. Different ecosystems would
presumably have qualitatively dissimilar animal body-size clumps
(i.e., the clumps would not match); differences in the clump
structures theoretically reflect the unique landscape structure in a
ecosystem. Also, the number of species in each clump could be
dissimilar between different ecosystems; more species in a body-size
clump is hypothesized to reflect a greater diversity of resources at a
specific scale (Holling 1992b). A growing body of studies indicates
that animal body-size clumps do reflect hierarchical structure in
landscapes (see Holling et al. 1994). Thus, animal body-size clumps
could be used as an "ecoassay" to measure both qualitative and
quantitative landscape properties at a variety of scales.
Ecological systems to the human eye seem extraordinarily
complex (Holling 1987), and it is a daunting task for ecologists to
understand and measure ecological phenomena across multiple
scales. As mentioned previously, it is especially difficult to
determine the scales at which different species respond to landscape
structure. For example, at what scale does the spatial distribution of
trees determines whether a species is in a given area? One goal in
this dissertation is to determine whether body-size clumps occur in
avian body-size distributions. A second goal is to determine whether
these clumps can be used to explore at what scales birds respond to
structure in an environment. A third goal is to evaluate how body-
size clumps can be used to measure the effect of suburban








environments on avian species diversity and abundance. To
accomplish these goals, I will take several steps:


Step (1) Chapter 2 uses a new statistical technique to detect
clumps in avifauna lists from different regions in
North America. The body-size clumps determined in
this step hypothetically represent the underlying
hierarchical structure of the temperate forest biome.
Step (2) In Chapter 3, avifauna lists from the boreal and
temperate forest biome are compared between
eastern North America and Europe. I hypothesize
that within each biome, the body-size clumps should
match if clumps reflect the hierarchical structure of
landscapes. This step also provides further tests for
the presence of body-size clumps of the temperate
forest biome.
Step (3) In chapter 4, the size limits of the body-size clumps
detected for the temperate forest biome (chapters 2
and 3) are used to place species into size categories.
Each category theoretically represents a group of
species that respond to landscape structure within the
same range of scales. In suburban sites, variation in
the percent of tree canopy cover (at four different
scales) is used to predict the number of birds/hr
censused for each body-size clump. Ultimately, for
each clump, this analysis determines the scale(s) at
which birds are responding to landscape structure.







Step (4) In chapter 5, a review is given of ecological studies
conducted in suburban environments. Then, the
results of chapter 4 are used to suggest how suburban
landscapes could be designed to attract more species.
Finally, I suggest how future empirical studies should
be conducted to determine the scales at which birds
respond to landscape structure.


Chapter 2 begins with a description of a new technique used to
determine avian body-size clumps that theoretically represent the
hierarchical structure of the North American temperate forest biome.
The objectives of chapter 2 are 1) to determine whether avian body-
size clumps exist for the North American temperate forest biome, 2)
to determine whether body-size clumps match between a desert
avian community and several temperate forest avian communities,
and 3) to determine whether temperate forest avian communities
and a desert avian community differ with respect to the number of
species in each body-size clump. If body-size clumps reflect the
hierarchical structure of landscapes, then body-size clumps should
match between temperate forest avifauna lists whereas body-size
clumps should not match between desert and temperate forest
avifauna lists. The methodology and results from chapter 2 are
basically the foundation for all other chapters. For example, the
temperate forest body-size clumps discovered in chapter 2 were
used to place censused suburban bird species (Chapter 4) into size
categories (i.e., clumps) that theoretically represent groups of species
responding to patches of canopy cover at the same range of scales.








However, the body-size clumps found in chapter 2 may, in part,
be due to the degree of species overlap that inevitably occurred
when I compared avian communities on the same continent (North
America). Therefore in chapter 3, I compared the body-size clumps
of New World and Old World avifauna from both temperate and
boreal biomes. One objective of this chapter was to determine
whether avifauna lists from similar biomes (but composed of entirely
different species) had qualitatively similar body-size clumps. In
other words, do the body-size clumps match between the avifauna
lists from different continents? If body-size clumps reflect the
hierarchical structure of landscapes, I predicted that body-size
clumps should match between North American and European
avifaunas (from the same biome).
While the avian body-size clumps should show an overall
match, quantitative differences (i.e., the number of species in each
clump) and qualitative differences (i.e., the position and number of
clumps) probably exist between the North American and European
avifaunas. Historically, changes in the pattern of North American
and European landscapes have been quite different. Since the
Pleistocene, European forests have been quite fragmented whereas
North American forests have been quite extensive and continuous
(Monkkonen and Welsh 1994). Because of this historical difference, I
predicted that 1) fewer medium-sized birds would exist in Europe
versus eastern North America, and 2) more clumps in the middle size
range would appear in eastern North America than in Europe. I also
compared the avian body-size clumps between the two biomes on
the same continent. I predicted that body-size clumps within a small








body-size range would display a better match than within a larger
size range. The above predictions were based on the propositions
that fragmentation would affect primarily medium-sized organisms
(Morton 1990) and that different landscapes would be most similar
at small scales (Holling 1992b).
In chapter 4, I explored at what scales birds (in different body-
size clumps) responded to large-scale landscape structure when they
choose home range areas during the summer and
stopover/dispersing sites during the spring. Avian biologists have
studied how birds respond to structure primarily at small scales (e.g.,
foraging decisions within a food patch); however, the spatial
distribution of objects (such as tree patches) at large scales probably
determines whether a bird occurs in a given area. Many ecologists
have expressed the need for empirical studies of how species
respond to structure at large scales (Lima and Zollner 1996, Wiens
1990, Wiens 1995, Opdam 1991, Harrison 1992). To simplify
empirical analyses in chapter 4, I grouped birds into the temperate
forest body-size clumps determined in Chapters 2 and 3. These
body-size clumps are useful because birds in each body-size clump
are theoretically responding to the same sizes of structural objects in
a landscape within the same range of scales (Holling 1992b). I
limited the scope of the study to only birds that forage or nest in the
tree canopy. The objectives of chapter 4 are (1) to ascertain whether
body-size clumps could be used to determine the scale(s) at which
different sizes of birds make landscape-level decisions during the
spring and summer, (2) to develop hypotheses about the spatial
areas at which birds respond to large-scale landscape structure, and







(3) to develop hypotheses about the patch sizes within these areas
that are sampled by birds of different sizes.
Chapter 5 is one of synthesis and speculation. The future
survival of a vast range of animal species is intricately tied to the
activities of humans because we are such a primary force of change
in the environment. Suburban habitats are one way that humans
have drastically altered the environment, and I emphasize the
importance of conducting ecological research in these areas. I
suggest that the size of a bird could be used to approximate the
range of scales at which it responds to landscape structure. How
different species of birds respond to habitat fragmentation is
dependent on the range of scales relevant to each species (Wiens
1990, Kotliar and Wiens 1990). Many species may respond to
structure within the same range of scales. Body-size clumps may
represent categories of birds that respond to structure at similar
scales. At large scales, once the spatial areas and patch sizes of birds
are determined, then researchers may be able to predict which
species will be most affected by human modifications of landscapes
in suburban environments.
Further, body-size clumps could be used to determine the
hierarchical structure of different ecosystems and amount of
structure available to birds at different scales. This would be useful
to urban wildlife managers. For example, if one knew that a majority
of the species in an avian community occurred in one body-size
clump, then a development could be designed to retain the scale-
dependent landscape features pertinent to these sizes of birds.
Overall, body-size clump analysis could guide researchers to the type





10

of structure (at different scales) available to a particular size range of
birds in a specific area. This would help researchers to design or
restore features in suburban areas to increase the abundance and
diversity of birds in these impacted landscapes.











CHAPTER 2


DETECTION OF HIERARCHICAL LANDSCAPE STRUCTURE USING
DISCONTINUITIES IN AVIAN BODY-SIZE DISTRIBUTIONS


Introduction


A central question (and problem) in ecology today is the
identification of the scales at which landscape patterns change in
ecological systems (Levin 1992, Wiens 1989b, Pickett and White
1985, Dayton and Tegner 1984). Hierarchy theory may simplify
studies that address this question because it decomposes complex
ecosystems into discrete hierarchical levels (Allen and Starr 1982,
O'Neill et al. 1986, Holling 1992a). A hierarchical level represents a
scale at which biotic and abiotic factors interact much more strongly
with each other than with other factors at other levels (Allen and
Starr 1982, O'Neill et al. 1986). For example, a familiar ecological
hierarchy is the organization of biological systems into different
levels (e.g., cell, organism, population community, and ecosystem).
Ecosystems can also be hierarchically organized. Hierarchical levels
in an ecosystem are defined by discontinuous process rates (Allen
and Starr 1982, O'Neill et al. 1986, Holling 1992a) or by a
discontinuous frequency distribution of object sizes (e.g., tree
patches) in a landscape (O'Neill et al. 1986, Holling 1992b). Process
rates (e.g., plant growth, fire regimes, erosion) are considered to be
discontinuous when they operate at distinctly different temporal







speeds (e.g., years, decades, centuries). Likewise, from small to large
spatial scales, sizes of landscape objects (e.g., leaves, crown volume,
stands of trees) are considered to be discontinuous when several
modes appear in a frequency distribution of their sizes. Thus, each
hierarchical level contains a specific category of processes and may
contain a defined size range of landscape objects.
A hierarchical organization of ecosystems has been inferred
from empirical studies of ecosystems and from models that captured
the dynamic behavior of managed ecosystems (Allen and Starr 1982,
O'Neill et al. 1986, Holling 1992a, Kent and Wong 1982, Bradbury and
Reichelt 1983, Holling 1986). In some empirical studies of
landscapes, fractal dimensions have been used to detect hierarchical
levels (e.g., Kent and Wong 1982, Bradbury and Reichelt 1983). A
fractal dimension basically measures the degree of complexity in a
landscape, and significantly different fractal dimensions theoretically
represent separate hierarchical levels. For example, Kent and Wong
(1982) calculated two different fractal dimensions of lake shorelines.
The fractal dimensions corresponded to glacial processes that shape
lake shorelines at large scales and to erosional processes at smaller
scales. In modelling studies, models were used to describe the
dynamic behavior of several managed ecosystems using only 3-4
sets of variables (see review in Holling 1986). However, these
models and empirical studies may be a result of the way researchers
make decisions or measure variables, not how ecosystems are
actually structured. More independent tests are needed to establish
whether ecosystems are hierarchically structured.








One independent test, proposed by Holling (1992b), states that
if ecosystems were hierarchically organized, then gaps should appear
in a distribution of animal body masses. Holling hypothesized that
animal masses would be grouped into several distinct "clumps",
reflecting the hierarchical structure of landscapes. Clumps are
essentially groups of similar-sized adult animals that are separated
by significant gaps in a body mass distribution. Animals in each
body size clump are theoretically responding to the same sizes of
structural objects in a landscape within the same range of scales
because the size of an animal constrains the sizes of objects available
to it (Holling 1992b). At this stage, the identity of the physical
structure that a bird species is responding to is largely unknown.
However, natural life history characteristics and the size of a species
permits one to imagine what a species is probably responding to in a
given landscape. Intuitively, different sizes of birds are responding
to different types of structure in a landscape. For example, to select
a home range, a wren probably samples the distribution of individual
bushes and trees whereas a hawk samples much larger groupings of
trees (e.g., riparian habitat).
Body-size clumps are hypothesized to reflect both qualitative
and quantitative structural properties of landscapes. Different
landscapes would have qualitatively dissimilar animal body-size
clumps (i.e., the mass ranges of clumps from different landscapes
would not match); differences in the clump structures theoretically
reflect the unique hierarchical organization in each ecosystem. Also,
the number of species may be dissimilar in clumps of different
ecosystems; more species in a body-size clump are hypothesized to








reflect a greater diversity of resources at a specific scale (Holling
1992b). Holling (1992b) formally stated this theory as:


"The Textural-Discontinuity Hypothesis: Animals should
demonstrate the existence of a hierarchical structure and of
the discontinuous texture of the landscape they inhabit by
having a discontinuous distribution of their sizes, searching
scales, and behavioral choices. Landscapes with different
hierarchical structures should have corresponding differences
in the clumps identified by such a bioassay."


Testing this hypothesis, Holling (1992b) found that bird faunas
of the boreal forest, of the short-grass prairie, and pelagic birds of
the pacific northwest were grouped into several statistically distinct
body-size clumps. Birds in these different ecosystems had dissimilar
clump structures; the differences in the clump structures seemed to
reflect qualitative and quantitative landscape differences in each
ecosystem. For example, the smallest clump category was absent in
the short-grass prairie, but it was present in the boreal forest
(qualitative difference). Also, fewer species were present in the
smaller-sized clumps of the short-grass prairie than in the smaller-
sized clumps of the boreal forest (quantitative difference). These
differences probably reflect the greater diversity of fine textures
distributed through the boreal forest tree canopy, whereas the
prairie has a much smaller amount of fine textures distributed just
above the ground.







Although the above results are consistent with the textural-
discontinuity hypothesis, other alternative hypotheses exist to
explain the observed avian body-size clumps (see Holling 1992b for
review). Other hypotheses include 1) "The "Urtier Historical
Hypothesis": current animal sizes could be historically constrained by
a limited number of ancestral forms, and these ancestral forms were
already in discrete size ranges independent of landscape structure;
2) "The Limited-Morph Hypothesis": the number of life-forms (such
as the hovering ability of a hummingbird vs. the soaring ability of a
vulture) are limited, and this limitation constrained the size ranges of
animals; and 3) "The Trophic-Trough Hypothesis": trophic
interactions (e.g., predator/prey interactions) may maintain
discontinuities in body masses because there is a relationship
between the size of an organism and what it eats. Holling (1992b)
tested these hypotheses by comparing body-mass clumps in
different ecosystems and by comparing different animal taxa (birds
vs. mammals) or animals with different feeding strategies
(carnivores vs. herbivores). From these analyses, only the textural-
discontinuity hypothesis resisted disproof. Thus, animal body-size
clumps could be used as an "ecoassay" to measure both qualitative
and quantitative landscape attributes relevant to birds at a variety
of scales. This possibility is explored for North American temperate
forest avifauna in this chapter.
Holling's original tests for animal body-size clumps did
demonstrate that different animal communities displayed
discontinuities in body-size distributions, but these analyses were
based on visualization techniques that were somewhat subjective








(e.g., eye-balling how clumps matched between animal lists) and
have been criticized in the literature (Manly 1996). Although
visualization can be as valid as statistical inference (Cleveland 1993),
formal statistical tests are more accepted in the scientific community.
Also, discontinuities in body-size distributions could be caused by
chance alone, and this issue was not adequately addressed in
Holling's original paper. A goal of chapter 2 was to utilize and
describe a new statistical technique and methodology to determine
clumps in animal body-size distributions. A second goal was to use
additional faunal lists to further test the textural-discontinuity
hypothesis. The objectives of this goal were 1) to determine whether
avian body-size clumps exist for the North American temperate
forest biome, (2) to ascertain whether a desert avian community and
several temperate forest avian communities have similar body-size
clumps, and 3) to determine whether temperate forest avian
communities and a desert avian community differ with respect to the
number of species in each body-size clump.
If body-size clumps reflect (qualitatively) the hierarchical
structure of landscapes, then body-size clumps should match
between temperate forest avifauna lists. However, between desert
and temperate forest avifauna lists, body-size clumps should not
match at all. In all of the above comparisons, I hypothesized that
most of the mismatch between body-size clumps would be present at
the medium to large size range. Small-scale structure is probably
geometrically the same regardless of the landscape concerned
(Holling 1992b). For example, bushes, trees, or small patches of
herbaceous ground cover may have the same geometric attributes








from one landscape to the next. Thus, I predicted that the number
and location of body-size clumps at the smaller size range will be
similar even when the temperate forest and desert avifauna are
compared.
Further, if body-size clumps reflect (quantitatively) the
diversity of resources at different hierarchical levels, then the desert
avifauna (compared to the temperate forest avifauna) should contain
relatively fewer species in the smaller size range. This prediction is
based on previous results that structurally simpler landscapes had
fewer small species (Holling 1992b). The rationale is that the
vertical height diversity is severely limited in deserts relative to
temperate forest regions. This would decrease the diversity of
resources available to small birds, resulting in fewer desert species.
The above predictions are contingent on several underlying
assumptions. First, in a avifauna list, birds are considered to
primarily utilize structure within the boundaries of a chosen
landscape (e.g., a temperate versus a desert biome). Second, the
"boundaries" of a landscape are assumed to be accurately and
consistently defined in the literature. Third, when one compares
avifauna lists derived from different areas that are assumed to be
similar (e.g., different regions within the temperate forest biome), it
is assumed that the selected areas are "truly" similar in terms of the
physical landscape structure. Fourth, reported masses of individual
bird species is assumed to represent the true size of a species. If any
of these assumptions are violated, then the outcome of an analysis
could be incorrect. Thus, care should be taken in selecting a
landscape and assembling a avifauna list.







Methods

In this section, I first describe a methodology and statistical
technique for identifying significant clumps in animal body-size
distributions. Second, I applied this methodology to several avifauna
lists of the temperate forest biome and of the desert biome. This
second step tests whether the body-size clumps theoretically
represent a hierarchical structure that is characteristic of the
temperate forest biome. Third, I compared the number of species
(within certain size ranges) among different ecoregions. This third
step addresses whether the diversity of resources are different
between different ecoregions of North America.

Methodology of Clump Analysis

If body-size clumps reflect the underlying hierarchical
structure of landscapes, then I predicted that


(1) the distribution of body sizes in faunal lists should
contain body-size clumps rather than be continuous,
(2) faunal lists from similar landscapes should have body-
size clumps in the same size ranges, and
(3) faunal lists from different landscapes should have body-
size clumps in different size ranges.


But how does one first objectively determine whether clumps exist in
a body-size distribution (prediction 1)? Then, how does one test
predictions 2 and 3? Current statistical techniques cannot address







prediction 1 effectively because available clustering methods cannot
reliably identify the correct number of clusters (i.e., clumps) in data
sets (see reviews in Milligan and Cooper 1985, Bock 1985). Further,
no techniques have been developed to address predictions 2 and 3.
A new technique had to be developed. The steps outlined below are
a result of a four year process of trial and error with examining
various data sets, using different statistical techniques, and computer
simulation. Much of the work was conducted by Paul Marples who
tenaciously worked on this problem and developed the procedure
outlined below.
This procedure uses inferential statistics, but it is not
conventional. It does answer the question whether chosen faunal
lists contain body-size clumps, and whether faunal lists have similar
or different body-size clumps. However, it is much more problematic
to estimate the "true" number of clumps that theoretically reflect the
underlying hierarchical structure of a particular landscape. In this
case, the method is meant more as a heuristic approach to guide
researchers to what the "true" number of clumps is in a body mass
distribution. To gain more confidence in a determination of the
number of clumps for a particular landscape type (e.g., deserts),
multiple data sets need to be analyzed. If multiple data sets reveal a
consistent number of clumps, then one can be more confident what
number of clumps theoretically reflect the hierarchical organization
of a particular landscape type.
Step 1. The purpose of this step is derive faunal lists that can
be used to address predictions 1, 2 and 3 (above). If one wants to
test whether faunal lists from similar landscapes have similar body-








size clumps, then faunal lists should be derived from regions that
theoretically would have the same landscape structure. Conversely,
if one wants to test whether faunal lists from different landscapes
have different body-size clumps, then faunal lists should be derived
from regions that theoretically would have different landscape
structure.

But what scale should represent a landscape and its
accompanying faunal list? Landscapes at the biome level (e.g., desert
vs. forest) are the most drastically different (at least to our eye); thus
it is reasonable to compare faunal lists that represent a particular
biome. However, for any comparison, the fewer number of species
shared between the faunal lists, the better. A large degree of species
overlap may bias the analyses to find the same clumps because they
share too many species. This is no problem when comparing faunal
lists from different biomes (i.e., testing prediction 3), but it is a
problem when comparing faunal lists that represent the same biome
(i.e., testing prediction 2). One way to minimize species overlap is to
compare faunal lists from different continents (but the same biome).
Another way is to compare faunal lists derived from different
ecoregions within a biome. The basic difference between biomes and
ecoregions is a matter of scale. At the largest scale (i.e., biome level),
classification is largely based on climatic factors (e.g., annual rainfall).
At smaller scales (e.g., ecoregion), the classification is based on
vegetational macrofeatures along with climatic factors (Baily 1995,
Omernik 1987). Ecoregions within the same biome theoretically
share similar hierarchical properties specific to that biome. I
therefore hypothesize that avian body-mass clumps will be similar







among ecoregions within a biome. A comparison between ecoregions
from the same biome could be used to test prediction 2.
Step 2. The purpose of this step is to establish whether chosen
faunal lists are "clumpier" than expected by chance alone. This test,
called the Gap Rarity Index (GRI) does not estimate the true number
of lumps because it relies on avoidance of Type I errors (i.e., alpha =
0.05). The true number of clumps may actually be more if power
was taken into account. The number of clumps determined in this
step are regarded as a minimum number of clumps that are likely
for a given faunal list.
First, a faunal list is arranged from the smallest species to the
largest species, and the log mass is calculated for each of the species.
All bird masses (in grams) are derived from Dunning's (1992) book
of avian body masses; the mass for each species is the mean mass of
reported male and female averages. Between each consecutive
species, the size of the "gap" is calculated by taking the mass of
species (i + 1) species (i). One obtains a variety of observed gap
sizes where larger gaps may indicate divisions between clumps. But
which gap sizes are larger than what would occur by chance alone?
To answer this, simulated data sets are randomly sampled from
a unimodal distribution (the sample size is kept constant between
the observed and the simulated data sets). A unimodal distribution
(i.e., lognormal distribution) is calculated by taking a kernel estimate
for an observed data set. A kernel estimate is simply a smoothed
histogram (Silverman 1986). A single parameter (h), similar to the
bin-width of a histogram, controls the amount of smoothing in a data
set. This parameter is decreased just enough to give one mode (i.e.,








unimodal) and no. more (Silverman 1986, Manly 1996). This
unimodal distribution represents a null model where no
discontinuities exist in a body-size distribution.
From this unimodal distribution, 10,000 simulated data sets are
generated. A distribution of expected gap sizes are calculated from
the simulated data sets. Thus for each observed gap, 10,000
simulated gaps are generated. These simulated gaps represent a
range of gap sizes that could occur by chance alone. An observed gap
is considered to be significant when a large number of the simulated
gap sizes is less than the observed gap size. Conventionally, a
significant number is where 9,499 of the simulated gap sizes are less
than the observed gap size (i.e., alpha = 0.05). For each observed
gap, one can determine its significance by comparing it to the
simulated gap sizes.
Stei 3. The purpose of this step is to determine whether
chosen faunal lists have similar body-size clumps (i.e., do the size
intervals match?). This step is used to further test the textural-
discontinuity hypothesis and to guide researchers to the number of
clumps that theoretically reflect the underlying hierarchical
landscape structure. If body-size clumps reflect the hierarchical
organization of different landscapes, then similar landscapes should
have similar body-size clumps and different landscapes should have
different body-size clumps. One compares a variety of combinations
of numbers of clumps in each data set above or equal to the
minimum number of clumps determined in step 2. The question
addressed is whether a significant match occurs when a variety of
combinations are compared (e.g., 7 vs. 8, 8 vs. 10, etc.). Each








comparison represents a proposed model that could result in a
significant "match" (if at all). A significant match (or matches)
indicates that the data sets are similar in terms of where the clumps
appear in a body-mass distribution. Below, the procedure to test
whether chosen faunal lists match is outlined.
For simplification, this procedure will be discussed in terms of
a pair-wise comparison, but a three-way or n-way comparison can
be done. As mentioned in step 2, each faunal list is arranged from
the smallest species to the largest species, and the log mass is
calculated for each of the species. A kernel estimate (Silverman
1986) is then used to calculate a variety of frequency distributions
for each faunal list (e.g., 2 modes, 3 modes, 4 modes, etc.). A
specified number of modes are made by varying the smoothing
parameter (h) to one particular value; the parameter is decreased
just enough to give the specified number of modes and no more
(Silverman 1986, Manly 1996). We found that frequency
distributions of 2 to 15 modes (i.e., clumps) are a sufficient range to
explore for most data sets.
Once the frequency distributions have been calculated, these
distributions are then defined in terms of clumps and gaps. Clumps
are the interval around each mode in a distribution defined by the
inflection points that give maximal and minimal slope (Figure 2-1).
All other intervals in a distribution are defined as gaps.
Across the chosen faunal lists, all combinations of clumps
(equal to or above the lower limit determined in step 2) are
compared to determine the degree of mismatch (m) for each
comparison. The number of clumps determined in step 2 are











0.4,


(0
E
= 0.2
U)


a)


clump gap clump
i I 1 r I


1 3 5
size (log g)


Figure 2-1. A hypothetical data set fitted with a 2 mode kernel
estimate (the units for the density estimate is the percent area under
the curve within a defined mass range). Each clump is defined by
inflection points that give maximal and minimal slope. Maximal and
minimal slopes are mathematically calculated.




considered to be a conservative estimate (i.e., controlling for Type 1
error), and more modes may actually exist in a distribution. For each
comparison, the degree of mismatch (m) is calculated by adding the
proportion of gaps filled by lumps (Pgl) and the proportion of lumps
filled by gaps (Plg):


m = Pgl + Pig







Because m is a summation of two proportions, m can vary between 0
and 2 where 0 is a perfect match and 2 is complete mismatch. This
technique is similar to a phi correlation index (see Wherry 1984).
Unlike the phi correlation index, though, the degree of mismatch (m)
is better at detecting significant matches between data sets. For each
model, the observed mismatch is set aside to be compared to
mismatches calculated from models of simulated data sets drawn
from a unimodal distribution.
To distinguish between significant matches and spurious ones,
a distribution of expected mismatches (drawn from a unimodal
distribution) for each proposed model is computed (see paragraph
below). In other words, what is the probability that each observed
mismatch could have occurred by chance alone? Expected
mismatches represent the degree of mismatch expected when
several data sets (each drawn from a unimodal distribution) are
compared to each other for a given model. A significant match
occurs in the real data set when a large number of the expected
mismatches is more than the observed mismatch of a particular
model. The outline below describes how to generate the expected
mismatch values for each proposed model.
First, a unimodal distribution is calculated with a kernel
estimate that gives one mode for each observed data set. This
unimodal represents a null model where no discontinuities exist in a
body-size distribution. From this unimodal distribution, simulated
data sets are randomly sampled (the sample size is kept constant
between the observed and the simulated data sets). For these
simulated data sets, expected mismatch values are calculated for







each model (e.g., 7 vs. 7 clumps, 8 vs. 9 clumps, etc.) as explained
above. At least 10,000 potential mismatches are calculated for each
model by randomly comparing the simulated data sets 10,000
different times. For a given model's distribution of expected
mismatches, a critical value can be determined. For example, at 0.05
(i.e., alpha = 0.05), a critical value is where 9,499 expected mismatch
values are greater than this value. Thus, for each proposed model,
we can determine the significance of its observed mismatch value by
comparing it to the critical mismatch value. A null hypothesis states
that the observed mismatch value of a model is no different from the
critical mismatch value generated from a unimodal distribution. A
model that gives an observed mismatch value at or below the critical
mismatch value (alpha = 0.05) is considered to be significant.
Step 3 can be used as a heuristic tool to determine what the
"true" number of clumps is for a given landscape (i.e., reflecting the
underlying hierarchical landscape structure). It should be
emphasized that the only way to determine the "true" number of
clumps for a given landscape type is to look at the matching between
faunal lists that theoretically represent the same underlying
hierarchical organization. If only one match resulted from comparing
faunal lists derived from the same landscape, this match may
represent the underlying hierarchical landscape structure. However,
for some comparisons, more than one model may give a significant
match. When several different models match, one can tentatively
select the one model that has the most equitable number of clumps
(between the data sets). The rationale is that faunal lists from the
same landscape type should have a similar number of clumps







because each list is a theoretical representation of the same
hierarchical landscape structure. Even if a model is selected, this
selection is tentative and more independent evidence is needed.
Additional faunal lists (representing the same landscape type) should
be compared to determine whether a similar match exists.
One additional point should be mentioned when comparing
faunal lists derived from the same landscape type. Recall that the
idea in these comparisons is to ascertain whether body-size clumps
reflect the hierarchical organization of a landscape. Chosen faunal
lists usually represent landscapes at the biome or ecoregion level. At
such large scales, some landscape variation exists between virtually
any comparison. This holds true for faunal lists derived from the
same biome on different continents and for faunal lists taken from
ecoregions within the same biome. Even though this variation exists,
landscapes from the same biome probably contain similar
hierarchical organization at the biome level. A significant overall
body-size clump match would theoretically be a reflection of this
biome-level organization. However, within some size ranges, the
body-size clumps may not line-up as well as in other places. Where
clumps do not line-up as well may be an indication of the unique
landscape structure of a particular ecoregion or biome. Thus,
comparing faunal lists representing the same landscape type will not
only reveal (theoretically) hierarchical similarities but also
differences. Where differences occur theoretically reflect the scales
at which structural differences exist between landscapes.
The above interpretations presume that none of the underlying
assumptions have been violated. As stated earlier, these








assumptions are 1) birds of each list are considered to primarily
utilize structure within the boundaries of a chosen landscape, 2) the
"boundaries" of a landscape (e.g., temperate forest biome) are
accurately and consistently defined in the literature, 3) areas that
are selected to be similar are "truly" similar in terms of the physical
landscape structure, and 4) reported masses of individual bird
species is assumed to represent the true size of a species. If any of
these assumptions are violated, then the interpretation of an analysis
could be incorrect.
Step 4. The purpose of this step is to group species into their
appropriate clumps. If a significant match is determined and one
particular model was selected (step 3), then the size range of each
body-size clump is defined by the inflection points that give maximal
and minimal slopes (i.e., edges: see step 3). However, some species
occur outside these intervals (i.e., in the gaps). One usually wants to
compare the quantity of species in each clump using the full species
list. It is useful to group species in each gap interval into either the
upper or lower clump. To do this, the midpoint of the largest gap in
the body masses between two species (in this gap interval) is used as
the dividing line; species to the left of this midpoint are placed in the
lower clump and species to the right of this midpoint are placed in
the upper clump. Thus, the upper and lower masses of a reported
body-size clump are not the "statistical" edges (those are the
inflection points), but they theoretically represent a size range of
species that respond to structure within the same range of scales.
In addition, even when the body-size clumps of faunal lists
match statistically, the upper and lower masses of these clumps







usually do not line-up exactly; thus, it is somewhat problematic to
construct body-size clumps that represent all of the analyzed data
sets. The reason why clumps do not line-up exactly between faunal
lists (assuming that each faunal list represents the same underlying
hierarchical landscape structure) is that each faunal list represents
an approximation of underlying landscape structure at the biome
level. As explained earlier, even landscapes from within the same
biome would contain some structural differences. Given this, one
should not expect an exact match between different faunal lists.
To construct a representative body-size clump pattern for all
faunal lists, the following is done. First, the body-size clumps of each
faunal list are arranged from the smallest to the largest clump (i.e.,
clump 1, clump 2, etc.). Then, the same clump (e.g., clump 1) is
compared across the data sets. If two or more data sets had clumps
that lined-up exactly, then this mass range marks the upper and
lower masses of this clump. When none of the upper and lower
masses of a specific clump lined-up, then an average of the upper
and lower masses (across all faunal lists) is used to delineate the
"size range" of this clump.


Clump Analyses of North American Temperate Forest Avian
Communities

An hypothesis tested here was whether a match occurred
between body-size clumps from different ecoregions of the
temperate forest biome. A significant match would be consistent
with the hypothesis that body-size clumps reflect the hierarchical
organization of a particular biome (in this case, the temperate forest







biome). Also, even if an overall significant match resulted between
the different ecoregions, it was hypothesized that most of the
mismatch would be present at the medium to large size range. As
mentioned previously, small-scale structure is probably
geometrically the same regardless of the landscape concerned; body-
size clumps at the smaller size range will be similar when avifauna
from different ecoregions are compared. I followed the outline (see
above methodology of clump analysis) to determine whether
matching occurred between body-size clumps of different avifauna
lists of the North American temperate forest biome.
I could have selected any ecoregion in the North American
temperate forest biome, but I selected areas where researchers had
conducted suburban bird surveys in several cities. The main reason
I selected areas based on the location of suburban surveys is that in
Chapter 4, I use these suburban surveys to estimate the scales at
which birds in different body-size clumps respond to landscape
structure. However, I first had to determine whether avian
communities in these different suburban areas could be grouped into
similar body-size clumps. These body-size clumps were determined
in this chapter.
I constructed bird faunal lists based on the geographic location
of each city (hereafter called regional bird lists). These regional bird
lists were all landbird species, excluding exotics, that could
hypothetically breed within each suburban area (i.e., their breeding
ranges overlapped with the exact locations of surveys conducted in a
suburban area). These regional bird lists were used to establish a
body-size clump pattern that hypothetically reflects structure of a







landscape before a human-caused transformation occurred. It
should be noted that all birds that historically could breed in a given
area were included in the avifauna lists, even if the birds had been
locally extirpated due to human activity. All bird masses (in grams)
were derived from Dunning's (1992) book of avian body masses; for
each species, the mass used was the mean mass of both reported
male and female averages. Sites selected were Amherst &
Springfield, MA (DeGraaf and Wentworth 1981); Austin, TX (Sexton
1987); Blacksburg, VA (Lucid 1974); Chicago, IL (Guth 1980); Seattle,
WA (Penland 1984); and Vancouver, B.C. (Lancaster 1976; Weber
1972). Based on Odum's (1971) map of the major biomes of the
world, all of the cities are located in the temperate forest biome. The
cities are located in five different ecoregions of this biome: Amherst
& Springfield Northeastern Highlands; Austin Texas Blackland
Prairies and Central Texas Plateau; Blacksburg Central Appalachian
Ridges and Valleys; Chicago Central Corn Belt Plains; and Seattle &
Vancouver Puget Lowland (Omernik 1987). A qualitative
vegetative and climatic description of each ecoregion is given in
Table 2-1.
Only the regional avifauna lists that had minimal species
overlap were used to determine whether body-size clumps matched
for regions in the temperate forest biome. Several of the regional
avifauna lists shared a considerable amount of bird species (70% or
higher overlap). However, the Amherst, Austin, and Vancouver
regional lists had a species overlap of 50% or less. Only these three
data sets were used in the subsequent clump analysis. Steps 2 4
were followed as described above (see clump analysis).










Table 2-1. A qualitative vegetative and climatic description of six
different ecoregions of North America. Source: Bailey, 1995 and
Omernik 1987.

City Ecoregion Avg. Vegetation
annual
rainfall


Northeastern 35 60
Highlands inches


Austin Texas
Blackland
Prairies and
Central Texas
Plateau


Blacksburg Central
Appalachian
Ridges and
Valleys


Deciduous forest dominated by
tall broadleaf trees; high flora
diversity; high vertical height
diversity


10- 30 Semi-arid and subtropical
inches savanna woodland; pinyon-
juniper woodland; scrub oak;
mesquite; medium low flora
diversity; medium low
vertical height diversity

35- 60 Deciduous forest dominated by
inches tall broadleaf trees; high flora
diversity; high vertical height
diversity


Chicago Central Corn 30- 40 Deciduous forest dominated by
Belt Plains inches oak-hickory in dry areas and
beech-sugar maple in wet
areas; medium-high flora
diversity; medium-high
vertical height diversity


Seattle,
Vancouver


Puget Lowland


Tucson Southern Basin
and Range


30- 50 Coniferous lowland mixed with
inches deciduous trees (big-leaf
maple, oregon ash, douglas
fir); medium-high flora
diversity; medium-high
vertical height diversity

2 10 Arid landscape, sparse
inches vegetation consisting of cacti,
shrubs, creosote bush, open
stands of trees; low flora
diversity; low vertical height
diversity


Amherst,
Spring-
field









Comparison Between the Clumps of a Desert and Several Temperate
Forest Avian Communities

The hypothesis tested here was whether avian body-size
clumps from different biomes would match. It was predicted that a
clump analysis between desert and temperate forest avifauna would
reveal no significant match. No match would be evidence in support
of the hypothesis that body-size clumps do reflect the hierarchical
"landscape signature" of biomes. However, even though an overall
mismatch may occur, it was predicted that body-size clumps at the
smaller size range will match between a desert and a forest avian
community. As mentioned previously, this is because small scale
structure is probably geometrically similar even between different
landscapes.
A regional avifauna list from Tucson, AZ was compared to the
Amherst, Austin, and Vancouver regional avifauna lists. I selected
Tucson because a suburban survey was also conducted in this city
(Mills et al. 1989). For the purposes of this study, I could have
selected several different areas in the Southwest region of the U.S.,
but I selected an area where a suburban study was conducted. This
was done to gather data the same way it was gathered for the
temperate lists. Essentially the Tucson avifauna list represents an
ecoregion of the desert biome and the Amherst, Austin, and
Vancouver avifauna lists represent ecoregions of the temperate
forest biome. Again, Steps 2 4 were followed as described above
(see methodology of clump analysis).









Densities of Species in Body-size Clumps of Each Ecoregion

The above five ecoregions of the temperate forest biome and
the one ecoregion of the desert biome may contain different amounts
of resources at different scales. The body-size clumps determined
from the above analysis are hypothesized to represent groups of
birds that respond to the same sizes of landscape attributes within
the same range of scales. Therefore, the number of species occurring
in each body-size clump of a region is theoretically reflecting the
diversity of resources available to birds at certain scales. I compared
how similar the number of species in each body-size clump was
among the ecoregions of the temperate forest biome. I also
compared the Tucson avifauna to the other temperate forest
avifauna. I first determined how many Tucson species fit into each
body-size clump determined for the temperate forest biome. I did
this because I had to use similar size ranges in order to compare the
number of species in the Tucson region versus the other ecoregions.
G-tests were used to detect significant differences (alpha = 0.05).
Based on the vegetative and climatic characteristics (Table 2-
1), I hypothesized that the Tucson and Austin avifauna list would
have relatively fewer number of small birds than the other
ecoregions. I made these predictions based on previous results that
structurally simpler landscapes had fewer small species (Holling
1992b), and on earlier work that found bird diversity increased with
an increase in foliage height diversity (MacArthur 1972, Recher
1969). The rationale is that the vertical height diversity is severely








limited in Tucson and somewhat limited in the Austin area (relative
to the other ecoregions). This would decrease the diversity of fine-
texture resources available to small birds, resulting in a fewer
number of small birds in the Tucson and Austin area.



Results


Body-size Clumps of the Temperate Forest Biome

From the Gap Rarity Index (GRI), all of the ecoregions displayed
a significant number of clumps (Table 2-2). Looking at all
combinations of 8 clumps and above (the minimum detected for each
ecoregion), the Amherst, Austin, and Vancouver comparisons
resulted in several significant models: a 14-9-8, 14-10-8, 14-10-9,
and 10-10-10 (clumps in Amherst, Austin, and Vancouver,
respectively) displayed significant matches (P < 0.05). The 10-10-10
model was chosen to represent the underlying hierarchical landscape
structure (because it was the most equitable). The clump and gap
intervals are displayed in Figure 2-2. Comparing the position of the
clumps (by eye) for the 10-10-10 model, most of the mismatch
occurred in the medium to large size range from 101.3 to 103.5 grams
(Figure 2-2). At 101.5 grams, the Amherst faunal list did not have a
clump whereas the Austin and Vancouver lists did at this mass.
Also, the Amherst faunal list had a clump at around 102.5 grams
whereas the Austin and Vancouver lists did not have a clump at this
mass. Between 102.5 grams and 103.5 grams, a minor amount of



















Table 2-2. Number of body-size clumps detected for each ecoregion
from the Gap Rarity Index.




Ecoregion Number of Clumps

Amherst 9 *
Vancouver 8 *
Austin 9
Tucson 8

* P < 0.05





Clump and gap intervals of bird body-mass data (10 lump model)


1 1111



1 1111


111


II II


I III


Figure 2-2. Body-size clump structure of three different ecoregions of the North American
temperate forest biome (Austin, TX; Amherst, MA; Vancouver, B. C.).


Austin

Amherst

Vancouver


0 0.5 1 1.5 2 2.5 3 3.5 4
log body-mass (g)








mismatch also occurred; Austin had a large clump around 103.2
grams, Amherst had a smaller clump, and Vancouver did not have a
clump. Body-size clumps at 101.3 grams and below seemed to match
among the three faunal lists.
For the 10-10-10 model, the size range of each clump (defined
by the upper and lower masses) are listed in Table 2-3. For ease
of reporting the results, the 10 size categories were assigned
numbers in ascending order starting with 1 for the smallest size
category (0.0 6.9 grams) to 10 for the largest size category (greater
than 2080.5 grams). This 10-10-10 model may reflect the
underlying hierarchical landscape structure of the temperate forest
biome. Chicago, Seattle, and Blacksburg are located in the temperate
forest biome and they had quite a large degree of species overlap
with the Vancouver and Amherst avifauna list. Thus, their avifauna
regional lists are also proposed to have 10 body-size clumps.

Amherst. Vancouver. Austin (Temperate Forest Biome) vs. Tucson
(Desert Biome)

Looking at all comparisons of 8 clumps and above, no
significant match was found for the Austin vs. Tucson and Vancouver
vs. Tucson comparisons (P > 0.05). However, several significant
matches were found for the Amherst vs. Tucson comparison: 8-8, 9-
9, and 10-10 (P < 0.05). For the 10-10 model, the clumps and gaps
are displayed in Figure 2-3. Among the four avifauna lists, body-size
clumps at 101.3 grams and below seemed to match quite well, but a
larger degree of mismatch occurred at clumps above 101.3 grams


















Table 2-3. Size ranges for each of the 10 avian body-size clumps
determined from the Amherst, Vancouver, and Austin comparison.


riB d-size Category


Size Ranae (a)


0-6.9
7.0-16.5
16.6-21.6
21.61-33.75
33.76-61.6
61.7-115.0
115.1-184.0
185.0 576.5
577.0-2080.5
> 2080.5







Lump and gap intervals of bird body-mass data
(10 lump model)


I 1111
IIII


IIH I


II II


I 1111111 I


0 0.5


1.5 2 2.5


I

I

U


3.5 4


I log body mass (g)
Figure 2-3. Avian body-size clump structure of four different ecoregions; Tucson avifauna
regional list is in the desert biome and the Amherst, Vancouver, and Austin avifauna regional
list are in the temperate forest biome.


Austin

Amherst

Vancouver

Tucson








(especially for the-Tucson vs. Vancouver and the Tucson vs. Austin
comparisons).

Species Densities in Each Body-size Clump

Using the 10 clump model, the number of species in each body-
size clump was quite similar for the ecoregions of the temperate
forest biome. However, the Austin ecoregion, as predicted, had the
fewest number of category 2 species than the other temperate forest
sites (Table 2-4), but this was not statistically significant (P > 0.05).
In the Tucson ecoregion (desert biome), the number of species in
bird-size category 2 was significantly less than in the other
temperate forest ecoregions (P < 0.05), except for the Tucson vs.
Austin comparison (P > 0.05). In addition, the Tucson ecoregion had
more species in bird-size category 5 when compared to the
temperate forest ecoregions (Table 2-3), but this was only
statistically significant for the Seattle/Vancouver vs. Tucson
comparison (P < 0.05).








Table 2-4. Number of species recorded for each bird-size category
from North American avifauna lists.

Bird-
size
Category Size Range (g) Seat Vanc Amh Blac Chic Aus Tuc

1 0.0 6.9 5 5 3 2 2 3 6
2 7.0 16.5 26 25 29 21 25 17 13
3 16.6 21.6 9 9 16 12 17 12 7
4 21.61 33.75 10 8 14 11 13 13 1 1
5 33.76 61.6 9 9 13 12 13 11 20
6 61.7 115.0 11 10 11 11 14 11 9
7 115.1 184.0 7 7 6 5 7 8 9
8 185.0 576.5 9 8 6 9 12 4 6
9 577.0 2080.5 10 10 5 4 4 5 10
10 > 2080.5 1 1 1 2 1 2 1



Seat = Seattle, WA; Vanc = Vancouver, B.C.; Amh = Amherst, MA; Blac =
Blacksburg, VA; Chic = Chicago, IL; Aus = Austin, TX; and Tuc = Tucson, AZ.




Discussion



Bird body-size distributions were shown to be discontinuous

(i.e., clumpyy") by the Gap Rarity Index. Further, the Amherst,

Vancouver, and Austin avifauna lists (temperate forest biome) had

body-size clumps at similar body-size ranges. If the clumps had not

matched, then there would be little support for the textural-

discontinuity hypothesis (i.e., similar landscapes having similar

body-size clump patterns). Further, the body-size clumps between

Tucson (desert biome) and Amherst, Austin, and Vancouver








(temperate forest biome) demonstrated that two of the three
comparisons did not match. The Vancouver/Austin vs. Tucson
comparison did not show a significant match, but the Amherst vs.
Tucson comparison did. The mismatch between at least two of the
comparisons between different biomes is consistent with the

textural-discontinuity hypothesis (i.e., different landscapes having
different body-size clumps). However, the one significant match
between the avifauna of different biomes suggests that 1) body-size
clumps may not reflect the hierarchical structure of landscapes, 2)
the hierarchical structure of biomes may be quite similar at a variety
of scales, or 3) errors occurred in the construction of the avifauna
lists (e.g., reported masses were incorrect or the avifauna lists did
not accurately contain only those species that breed in each
respective area). In addition, other alternative hypotheses (as
mentioned in the intro.) may explain body-size clumps (see Holling
1992b for review). To further address the textural-discontinuity
hypothesis, future studies should continue comparing faunal lists
from different biomes to see whether significant or non-significant
matches occur.


Body-size Clump Matches

The 10 clump model found in the Amherst, Austin, Vancouver
comparison may represent the hierarchical landscape "signature"
characteristic of the temperate forest biome. However, it is still
tentative whether body-size clumps reflect hierarchical landscape
structure. Nevertheless, if body-size clumps do reflect landscape
structure, then where clumps do not match may be an indication of







the scales at which structural differences exist. The 10 clump model
was used to see how well the body-size clumps matched between the
different ecoregions.
The Amherst, Vancouver, and Austin comparison did reveal, as
predicted, that the body-size clumps at the smaller size range (101.3
grams and below) were the most similar. Even the Amherst/
Vancouver/Austin versus Tucson comparison showed that the body-
size clumps at the smaller size range matched. These results suggest
that to some degree, small-scale structure from landscape to
landscape may be geometrically similar. That is, the frequency
distribution of landscape objects at small scales could be the same
regardless of the landscape type. For example, perhaps the
physiognomy and the spatial distribution of small bushes, trees, and
herbaceous ground is similar. This is only one example, though;
many more avifauna body-size distributions need to be compared to
see whether body-size clumps at the smaller size range are
consistently the most similar.
Although there was an overall significant match between the
Amherst, Vancouver, and Austin avifauna lists, the medium to large
size range (101.3 to 103.5 grams) displayed some degree of mismatch.
One explanation may simply be that this mismatch is a result of
sampling error. The measurements of body mass or the construction
of the avifauna lists may have introduced enough variability to cause
the body-size clumps to show some degree of mismatch. However,
this mismatch may alternatively be an indication of different
hierarchical landscape structure at medium to large scales between
the three ecoregions. The three avifauna lists are derived from three








different ecoregions (Amherst Northeastern Highlands; Austin -
Texas Blackland Prairies and Central Texas Plateau; and Vancouver -
Puget Lowland [Omernik] 1987). The mismatch of body-size clumps
within this medium to large size range could be caused by different
processes in each ecoregion producing a unique frequency
distribution of landscape objects at medium to large scales (e.g.,
patches of trees). What actual structural difference (in terms of the
spatial geometry) exists at medium to large scales is difficult to say,
but this analysis indicates where researchers can begin to look. A
detailed analysis of structure relevant to medium to large birds of
each ecoregion may reveal some landscape features that are present
in one ecoregion but absent in another. For example, if one knew the
scale(s) at which medium to large birds respond to landscape
structure, then the spatial distribution of the vegetation (relevant to
birds in their respective ecoregions) could be compared between the
two regions. This type of study may not only reveal structural
differences between different landscapes, but may shed light on
what types of processes are unique to different regions.
Each body-size clump, though, does not represent a group of
species that respond to only landscape structure at one hierarchical
level. I propose that birds in each body-size category respond to
objects produced at several different hierarchical levels. For
example, a small wren not only searches for prey items among fine
textured objects at small scales (e.g., leaves and twigs), but it also
establishes a home range based on large texture objects at large
scales (e.g., tree and plant densities). However, the size of an
organism dictates the size range of objects that it responds to in a







landscape; small birds encounter landscape structure within smaller
scale ranges than do larger birds. Small birds use objects at a scale
range from millimeters when foraging for food (e.g., deciduous litter
or pine needles) to kilometers when selecting territories (e.g., small
areas of forest) whereas large birds use objects in a range from tens
of centimeters (e.g., tree crown volume) to thousands of kilometers
(e.g., a large forest tract) for the same choices. Thus, hypothetically,
each body-size clump represents a group of species that respond to
landscape structure within the same scale range, but some degree of
overlap must exist when species in different clumps are sampling
objects in a landscape.
But how can landscape structure, over evolutionary time,
influence the size of an organism? An animal's size presents it with a
trade-off between its ambit (or distance it can travel) and the size of
objects that it samples. On the one hand, in a textured environment,
smaller animals are exposed to more resources per unit area and do
not have to travel long distances to obtain sufficient resources.
Larger animals, though, are exposed to less resources per unit area
because of the coarse grain of their choices; they gather enough
resources by their ability to travel greater distances. Therefore, each
body-size clump theoretically represents sizes of animals that satisfy
this trade-off between the sizes of landscape objects available to an
animal and how far an animal can travel to obtain resources (Holling
1992b).
As indicated earlier, though, birds sample landscape objects
through a range of scales. At what scale is the distribution of
landscape objects playing a dominant role in determining (over








evolutionary time) the body-size of a bird? Most likely, the spatial
distribution of landscape objects at large scales determines the body
size of a bird (e.g., home range area). Although birds are sampling
objects at smaller scales, these smaller objects are contained within
objects at larger scales. For example, tree branches and leaves (small
scales) that arboreal birds forage in are contained within patches of
forest (large scales). The spatial distribution of forest patches
probably dictates which sizes of birds are in a landscape because
these patches contain all of the necessary structure for foraging,
nesting, and protection from predators.
But what ultimately causes the gaps to appear in a body-size
distribution? Theoretically, hierarchically structured landscapes
produce a discontinuous frequency distribution of object sizes in an
environment (Holling 1992b). The gaps that do occur in body-size
distributions could be a result of two mechanisms. First, the gaps
could represent "zones of exclusion" where animal sizes that are in
these gaps are maladaptive in exploiting objects of either the upper
or lower range of scales (Holling 1992b). Animals with body sizes
that fall in these gaps are not able to satisfy the trade-off between
the sizes of landscape objects available to it and how far it can travel
to obtain resources. Second, the gaps could occur because animal
sizes are "clustering" around the abundance of certain sizes of
resources in an environment. These two mechanisms may not be
mutually exclusive, both mechanisms combined could cause gaps to
appear in animal body-size distributions.
Addressing the "zones of exclusion" hypothesis, gaps in body-
size distributions may represent size ranges where birds are at a








selective disadvantage. In some cases, the gaps between body-size
clumps sometimes span only a few grams. Do a few grams really
make that much of a difference to cause birds that lie in these gaps
to be at a selective disadvantage? It probably does, especially
during "ecological crunches" or perhaps under disruptive selection.
Studies have shown that variation in morphological characters
permit organisms to exploit different resources in a landscape (e.g.,
Lack 1971, Karr and James 1975, Fitzpatrick 1985, Grant 1986). In
some cases, minute differences in morphological characters within or
between species can result in some individuals having greater
reproductive success than others. One example of this was in a study
on "beak size" of an African finch in Cameroon, West Africa (Smith
1987, Smith 1990). It was demonstrated that lower bill width was
bimodal for juvenile finches; the mean at one peak was 12.7 mm and
the other peak was 15.7 mm. The abundance of soft- and hard sedge
seeds, which are scarce during the dry season, was proposed to cause
the selection for bimodality in bill sizes. Those finches that had
extreme or intermediate bill sizes between the two peaks had a
lower survivorship than birds with beak sizes near the peaks. It was
hypothesized that this differential survivorship was due to feeding
efficiency. Finches with large bills could feed efficiently on the hard
seeds and finches with small bills could feed efficiently on the soft
seeds (Smith 1987, Smith 1990). Along these lines, body sizes that
lie in the gaps may not be able to obtain resources efficiently, and
this may cause the clumps to appear in size distributions.
I propose that the 10 body-size clumps determined in this
study represent suites of species that respond to similar sizes of








objects in a landscape within the same range of scales. These
landscape objects may be of similar sizes, but they can be
qualitatively different. A wide variety of objects can be found at
each hierarchical level in a landscape, and birds have developed a
wide variety of natural life history strategies to exploit all objects at
all scales. For example, an arboreal bird species exploits tree needles
and leaves (small-scale objects), large trees (medium-scale objects),
and patches of trees (large-scale objects). A similar-sized ground
bird species exploits grasses (small-scale objects), small patches of
grasses (medium-scale objects), and large meadows (large-scale
objects). Both birds are measuring the same sizes of objects at each
scale, but the objects are qualitatively different. The quantity of
scale-dependent landscape objects (specific to each species) dictates
the presence or absence of a species in a given area. If the sizes of
objects utilized by species in one body-size clump were known, these
sizes are probably relevant to all species that fall in this size
category. If body-size clumps represent groups of species that
respond to landscape structure within the same range of scales, then
the 10 body-size clumps found in this study sets the stage for
exploring the sizes of objects (within a particular scale range) of a
wide variety of species.

Quantity of Species in Each Clump

In this chapter, several of the different ecoregions differed
with respect to the number of species in certain body-size clumps.
Recall that I hypothesized that the number of species in each clump
represents the diversity of resources within a particular scale range.








It was predicted that the Tucson and Austin avifauna lists would
have the fewest number of small birds. Indeed, the Tucson region
(of the desert biome) had the fewest number of small-sized species
(category 2) than any of the other temperate forest regions. One
interpretation of this disparity is that the interaction of structuring
processes in the temperate forest regions produced a greater
diversity of structures (e.g., more tree species) available for category
2 birds. The fewer number of category 2 birds in the Tucson area
may be a result of the limited amount of fine textures in deserts
when compared to forests. The vertical height diversity in deserts is
limited, most of the fine texture is distributed over a layer just above
the ground. In contrast, the large volume of canopy cover in the
forested ecosystems would contain a diversity of fine-texture
resources exploitable by birds, theoretically resulting in more
category 2 birds in temperate forests than in deserts. The Austin
avifauna list also had a smaller number of species in category 2 than
the other temperate forest ecoregions (not significant though).
Austin is located in Texas Blackland Prairies and Central Texas
Plateau (Omernik 1987), and this ecoregion has less rainfall and
more sparsely distributed vegetation than the other temperate forest
ecoregions (Bailey 1995). A decreased amount of rainfall and open
stands of trees would result in a decreased amount of fine landscape
texture (e.g., tree crown volume) that small birds could exploit.
For category 5 birds, the Tucson region (of the desert biome)
had more species than any of the other ecoregions in the temperate
forest biome (although only significant between the
Seattle/Vancouver and Tucson comparison). The greater amount of








category 5 birds in the Tucson region is more difficult to explain.
One would think that forests would have more resources than
deserts even at medium scale ranges. However, deserts have much
more open areas than forests; the combination of open ground cover
and of scattered vegetation patches may provide more types of
resources for category 5 birds. Many of Tucson's category 5 birds
are ground foragers or hunt in open areas (e.g., Pyrrhuloxia, Elf Owl,
Canyon Towhee, Abert's Towhee, and Inca Dove) (Ehrlich et al. 1988).
A combination of open ground and patches of trees or bushes is used
by many birds in this size category. Perhaps at scales relevant to
category 5 birds, the landscape mosaic of temperate forest ecoregions
provides limited opportunities. It would be interesting to see if this
pattern holds true for other desert and temperate avian communities
on other continents.

Revisiting Body-size Clump Patterns & Future Research

The results in this study are consistent with the hypothesis
that body-size clumps reflect both the qualitative and quantitative
properties of hierarchically structured landscapes. However, this
type of study does not lend itself to the traditional experimental
approach where variables are manipulated to understand the
underlying mechanisms that explain a given pattern. Typically, the
experimental approach is only applicable to ecological questions at
small scales that enable the researcher to tightly design an
experiment. This type of approach is not practical for testing clump
theory; testing of this theory lies under the construct of confirmation.
Confirmation uses multiple lines of evidence to confirm or reject the








generality of a theory (Pickett et al. 1994). Confirmation of a theory
inherently requires a variety of empirical cases that support the
claims of a theory. So far, the empirical patterns detected in this
study are generally consistent with the expectations of clump theory
(although the Amherst vs. Tucson comparison does not). However,
many more independent examples are needed to determine whether
body-size clumps truly reflect the hierarchical structure of
landscapes. Such studies are currently underway in the lab of C. S.
Holling.
A hierarchical framework has a wide application to studying
many important aspects of ecology, such as optimal foraging theory,
habitat selection, population dynamics, and the impact of habitat
fragmentation on animal communities (Kotliar and Wiens 1990). At
this stage, it appears that body-size clumps may be useful as a tool to
explore both the hierarchical structures of landscapes and at what
scales animals respond to structure over a range of scales. The
important theoretical concepts of this technique for researchers are
(1) that body-size clumps reflect the unique hierarchical structure of
a landscape, (2) that birds in each body-size clump respond to
similar sizes of landscape objects within the same range of scales,
and (3) that the number of bird species in a clump reflect the
diversity of resources within a scale range.
Understanding how changes in the landscape affect the
distribution and abundance of animal populations is an important
issue in ecology (e.g., Levin 1992, Gardner et al. 1993, Rahel 1990,
Kotliar and Wiens 1990). Many cross-scale studies have shown that
landscape heterogeneity and patchiness can be found at almost any








scale (Powell 1989, Levin 1992). Ecologists have developed different
techniques and models that attempt to capture and describe patterns
in ecological systems (Gardner et al. 1987, Milne 1988, Fahrig 1988).
Some of these techniques have been used to determine the
hierarchical structure of landscapes. For example, fractal analyses
have been highly regarded as a tool in ecology to describe landscapes
and landscape changes across a variety of scales (Milne 1988,
Mandelbrot 1977, Bradbury and Reichelt 1983). In a hierarchical
analysis, fractal breaks are hypothesized to represent the transitions
from one hierarchical level to another (e.g., Milne 1988). These
analyses, though, are based on how we measure the landscape and
not whether animals measure hierarchical structure in landscapes.
Using animal body-size clumps removes the subjectivity that
inevitably occurs when humans measure landscape structure at
different scales. Although our eye tells us that the landscape
structure is quite different from one ecosystem to the next (or from
one scale to the next), the landscape differences that we perceive (or
measure) may be quite different than what animals "perceive". For
example, the Tucson avian community (desert biome) had more
category 5 birds than avian communities of the temperate forest
biome. This indicates that desert regions may have more resources
for category 5 birds than temperate forest regions. Intuitively, one
would have hypothesized that temperate forest landscapes would
have more resources (at any scale) because these ecosystems are
much more structurally diverse than deserts.
If ecosystems were hierarchically structure, then the structure
and function of ecosystems may be controlled by a small number of








processes at different scales (Holling 1992b, O'Neill 1986). Each
hierarchical level would be defined by its own spatial textures and
set of processes. This would greatly simplify studying cross-scale
dynamics of ecosystems. Particularly, it would simplify models that
attempt to predict the behavior of ecosystems across space and time.
Examples of such models include the spruce/budworm system of
eastern North America (Clark and Holling 1979) and the Everglades
system of Florida (Walters et al. 1992). Body-size clumps may
provide additional evidence of the hierarchical organization in
ecosystems. Further, body-size clumps may help researchers to
relate scale-dependent landscape structure to the different species in
a community; this has been an elusive problem to solve in ecology
(Levin 1992, Kotliar and Wiens 1990). Body-size clumps will greatly
simplify this endeavor because theoretically the species in each
body-size clump respond to structure within the same range of
scales.

The proposed 10 body-size clump model represents groups of
birds that theoretically respond to similar sizes of landscape objects
within the same range of scales. The next step is to relate the body-
size clump structure to actual scale-dependent landscape structure.
But how does one determine what range of scales and sizes of
landscape objects are relevant to species in each body-size clump? I
address this question in chapter 4 by correlating the counts of birds
in each clump to the amount of tree canopy cover remaining (at
different scales) in suburban areas mentioned in this study.
Fragmentation of suburban habitats affects a variety of species, but
different sizes of animals respond to different scales of








fragmentation. For example, in suburban habitats, a wren probably
would respond to the amount of small trees and shrubs in your
backyard, but a hawk would respond to the distribution of trees in
your whole neighborhood. Knowing the scale-dependent objects
important to birds in each body-size clump would greatly enhance
our understanding of how human-modified landscapes affect avian
communities.
An additional question is whether the 10 body-size clump
model truly represents the unique hierarchical signature of
temperate forest landscapes. This model was chosen even though a
few other models gave a significant match. Also, this model may be
a result of the large degree of species overlap between the
temperate forest avifauna lists. If the 10 body-size clump model
truly represented the hierarchical structure of the temperate forest
biome, then a comparison of temperate forest avian communities of
two different continents should reveal the same clumps. In chapter
3, I compare the body-size clumps of the temperate avian
communities of North America and Europe. If the 10 body-size
clump model reflects the hierarchical structure of temperate forest
landscapes, then the avian clump structure of New World and Old
World temperate forest biome should match.

Summary

Most of the results were consistent with the predictions that
were derived from the textural-discontinuity hypothesis. The
temperate forest avifauna comparisons did show matching in the
location of their clumps whereas two of the three desert versus








temperate forest comparisons did not match. The results in this
chapter suggest that avian communities in temperate forest
landscapes display roughly 10 body-size clumps. Further, within the
smallest size range, the body-size clumps matched quite well across
all avifauna lists, even between the Amherst and Tucson lists that
represented drastically different landscapes. This may indicate that
structure at small scales is geometrically similar across all types of
landscapes. In terms of the quantity of species in each clump, the
avifauna lists from landscapes with theoretically more fine-textured
resources had more small-sized species. In addition, the Tucson
avifauna (desert landscape) had more category 5 birds than the
other temperate forest avifauna; perhaps a combination of more
open areas with patches of vegetation provide more resources for
larger birds.

The above findings are at least consistent with the hypothesis
that avian body-size clumps seem to reflect both qualitative and
quantitative aspects of the hierarchical structure of landscapes. This
would be extremely useful for ecologists studying how scale-
dependent changes in a landscape affect the distribution and
abundance of animal populations. For example, the loss of species in
a certain body-size clump would indicate at what scale humans have
altered the landscape (of course one has to determine the scale at
which birds respond to structure). Detecting the hierarchical
structure of landscapes with body-size clumps is unlike previous
ecological techniques (e.g., fractal analysis) that measure landscape
structure at different scales. This is an important distinction
between the two techniques because the way animals measure





57

discontinuous landscape structure may be completely different than
the way ecologists measure them. Although the results were
consistent with the textural-discontinuity hypothesis, more tests are
needed to become more confident that animal body-size clumps
reflect the hierarchical structure of landscapes.











CHAPTER 3


BODY-SIZE CLUMPS OF NEW WORLD AND OLD WORLD AVIAN
COMMUNITIES


Introduction

Body-size clumps are hypothesized to reflect the hierarchical
structure of landscapes. Where the clumps appear in a body-size
distribution theoretically reflects the unique hierarchical structure of
a landscape, and the number of species in each clump reflects the
diversity of resources available to species in each clump at a limited
range of scales (Holling 1992b). This was formally stated as the
textural-discontinuity hypothesis (Holling 1992b). Testing this
hypothesis, Holling (1992b) found that the bird fauna of the boreal
forest, of the short-grass prairie, and pelagic birds of the pacific
northwest were grouped into several distinct body-size clumps. The
avifauna lists were dissimilar in terms of the number of species in
each clump and where the clumps were located in a size distribution.
The differences seem to reflect qualitative and quantitative
differences in the physical architecture in each ecosystem. These
initial analyses revealed that body-size clumps may reflect an
ecosystem's unique hierarchical landscape structure, but only a few
data sets were analyzed and additional tests are needed.
In chapter 2, several bird faunal lists of North America were
used to determine whether body-size clumps exist. Ten body-size
clumps were proposed to reflect the hierarchical landscape structure








of the temperate forest. However, about 50% overlap in bird species
occurred among the data sets that were compared. Thus, there is a
possibility that the body-size clumps may, in part, be determined by
the phylogenetic similarity of the faunal lists. Further, this was only
one example, and additional tests are needed. Thus, an independent
analysis should be conducted between avifauna lists that share very
few species. A good comparison would be between eastern North
American and European avian communities because very little
species and genera overlap occur between the two continents (Helle
and Monkkonen 1990). For example, eastern North America and
Europe do not have many bird genera in common for birds that
winter south of U.S. and Europe (Monkkonen and Welsh 1994). If
body-size clumps reflect the hierarchical structures of landscapes,
then the edges of body-size clumps should match between North
American and European avian communities of the same biome. Such
a comparison further tests the textural-discontinuity hypothesis.
Historically, though, the landscape patterns of North American
and European continents have been quite different. In Europe,
landscapes were much more fragmented than landscapes in North
America. European forests were almost completely fragmented (e.g.,
Kurten 1972) because of Pleistocene glaciations in the Old World.
Most of Western and Central Europe were covered by some type of
tundra (Peterson et al. 1979), and forests persisted in small
fragments along European mountains in the south (Huntley 1993)
with a few small remnants in the north. In contrast, North American
Pleistocene glaciations did not result in extensive fragmentation
(Monkkonen and Welsh 1994). Glaciation shifted the north to south








boundaries of temperate and boreal regions, but extensive deciduous
and coniferous forests existed south of the glaciers (CLIMAP 1976,
Webb 1988, Pielou 1991). Thus, European forests during the
Pleistocene were quite fragmented whereas North American forests
were quite extensive and continuous.
In addition, after the Pleistocene glaciations, the scale of human
impact was much larger and happened much earlier in Europe than
in North America. Humans in the Old World had developed an
advanced form of agriculture, domesticated much of our barnyard
animals, and built many cities long before peoples of the New World
did so (Crosby 1986). Major landscape changes in temperate regions
of Europe began to take place 5000 years ago with much of the
landscape being converted into cultivated areas (Chambers 1993). It
has been estimated that present European temperate landscapes
already existed 1000 years ago in certain parts of Europe (Rackman
1980, Thirgood 1981). In eastern North America, large scale human
impact of temperate regions did not begin until 300 years ago with
greatest changes occurring within the past 100 years (Williams
1989). Within the boreal zone, human impact occurred much more
recently in both continents, but European regions were impacted
much earlier than North American regions (Monkkonen and Welsh
1994). For example, large scale forest harvesting occurred in Finland
between 1500 and 1900 (Darby 1956), but large scale forest clearing
did not begin in the Canadian boreal forest until the 1950s.
Fragmented European forests limited the amount of area
available to forest taxa, ultimately decreasing the diversity of
European forest flora (see Huntley 1993). Also, fragmented forests







are characterized by a patchy distribution of forested areas. I
hypothesize that fragmentation may primarily reduce the diversity
of resources that could be exploited by medium-sized birds than
small or large birds. Small birds theoretically could still find
sufficient resources within each small patch of forest because their
small size exposes them to more resources per unit area than larger
birds. Also, large birds could find sufficient resources in a
fragmented landscape because their ambit is large enough to travel
between patches. However, medium-sized birds could neither find
enough resources in each small patch nor travel between the patches.
If body-size clumps reflect landscape structure, then fewer clumps
within certain size ranges would indicate a limited diversity of
resources at certain scales. Because of the greater extent of European
forest fragmentation, I hypothesize that there would be fewer body-
size clumps in the medium size range of the eastern European
avifauna list than of the North American avifauna list.
Furthermore, the negative effects of fragmented European
forests on medium-sized birds would ultimately decrease the
number of medium-sized European species. It was hypothesized that
the number of medium-sized species would be dramatically less in
Europe than in eastern North America. Fragmentation of forested
landscapes would also cause less fine textured resources to be
available to small birds because of decreased diversity of forest taxa.
Remnant fragments, theoretically, could support small birds but the
diversity of forest taxa would be reduced. Thus, I hypothesize that
the number of small species would be less in Europe than in eastern
North America.








To test the above predictions, I compared eastern North
American and European body-size distributions of avian
communities from the boreal and temperate forest biomes. All of the
above predictions were based on the hypothesis that body-size
clumps reflect both quantitative and qualitative properties of
hierarchically structured landscapes The main objectives were 1) to
determine whether the avian body-size clunips of the same biome
matched between the two continents, 2) to determine whether the
number of birds in each body-size clump differed between the two
continents, and 3) to determine whether the number of clumps
differed between the avifauna lists of the two continents. A fourth
objective was to compare the avian body-size clumps between the
two biomes (temperate and boreal) on the same continent. I
hypothesize that the body-size clumps within a small body-size
range would display a better match than a larger size range. This
hypothesis was based on results of chapter 2 that showed the best
match occurred within the small body-size range between avifauna
of the desert and temperate forest biome.



Methods

The first step was to determine the geographic boundaries for
the temperate and boreal biomes on both continents. These
boundaries were derived from Odum's (1971) map of the major
biomes of the world and are displayed in Figures 3-1 and 3-2. The
next step was to construct avifauna lists of each continental biome.
























































Figure 3-1. Outline of the boreal forest biome in
eastern North America and Europe.


LEGEND
- Boreal Foresi



















































Figure 3-2. Outline of the temperate forest biome in
eastern North America and Europe.







These avifauna lists consisted of all landbirds, excluding exotics in
each respective continent, with breeding ranges overlapping the
geographic boundary of the temperate and boreal biome. Breeding
ranges of European birds were determined from Jim Flegg's field
guide (1990) whereas breeding ranges of eastern North American
birds were determined from the Peterson Field Guide (1980). All
bird masses (in grams) were derived from Dunning's (1992) book of
avian body masses; for each species, the mass used was the mean
mass of both reported male and female averages. Some of the
breeding ranges of certain species barely overlapped with the chosen
geographical regions of eastern North America and Europe. In these
cases, the field guides were consulted to ascertain which type of
biome these birds normally breed in. For example, if a species had
part of its breeding range in the boreal forest, but it was
characterized as breeding in the tundra, this species was not included
in the boreal forest bird list.
The body-size clumps of eastern North American and European
avian communities (boreal and temperate zones) were determined as
outlined in the methods section of chapter 2. Avifauna lists of
eastern North American boreal and temperate biomes were
compared to their respective European avifauna lists. Only body-size
clumps that gave a significant match (alpha = 0.05) and had the most
equitable number of clumps between the lists were considered as the
body-size clump pattern representing each biome.
If one body-size clump model was determined for the boreal
and temperate avifauna lists of each continent, the clumps were then
compared within each biome between the continents and between








each biome within a continent. The purpose of this comparison was
to determine whether body-size clumps of each biome showed
similar trends in terms of the number of clumps and the number of
species within certain size ranges.


Between Continents

Because European landscapes have been more fragmented than
eastern North American landscapes, I predicted that temperate and
boreal eastern North American avifauna lists would display more
body-size clumps than European avifauna lists. These additional
clumps would appear in the medium size range because the
relatively unfragmented eastern North American forests provide a
greater diversity of landscape resources that could be exploited by
medium-sized birds. In addition, the body-size clumps of the
temperate eastern North American avifauna lists should closely
match the clumps determined for the temperate forest biome of
chapter 2.
The increased amount of fragmentation in Europe would also
affect the number of species within certain size ranges. I predicted
that European avifauna lists would contain dramatically fewer
medium-sized birds than eastern North American avifauna lists.
This is because medium-sized birds (especially forest species) would
not be able to obtain enough resources in fragmented European
landscapes. European avifauna lists were also predicted to have
fewer small-sized birds than eastern North American avifauna lists.
Fragmented European landscapes would contain less fine textured
resources (due to decreased forest taxa diversity) than eastern North








American landscapes. G-tests (alpha=0.05) were used to detect
differences in the number of species in each clump between the
biomes.


Within Continents

Because fine textured resources are hypothesized to be
geometrically similar between different landscape types, I
hypothesized that body-size clumps within a small size range
(around 1.3 log g and below) would be similar between the
temperate and boreal biomes on each continent. The clumps within
this small size range would be more similar than larger size ranges.
This size range was chosen because body-size clumps lined-up within
this size range between different landscapes of North America
(chapter 2).



Results

The avifauna lists for the eastern North American and
European temperate and boreal biome are displayed in Appendix B
and C. Less than 10% species overlap occurred between North
American and European avifauna lists (both biomes). Below, the
results of the avifauna clump analysis are given.

Between Continents

Temperate forest biome. The Gap Rarity Index found 9 clumps
in the eastern North American avifauna and 7 clumps in the
European avifauna. Match analyses (7 to 15 clumps) of the two







temperate forest avian communities resulted in only one significant
match (P < 0.05) with 10 body-size clumps in eastern North America
and 7 body-size clumps in Europe (Figure 3-3). The body-size clump
pattern in eastern North America had 3 extra clumps between 16.0 g
and 52.0 g (1.2 log g and 1.7 log g: Figure 3-3). The body-size clumps
defined by the upper and lower masses are shown in Table 3-1.
These body-size clumps are similar to the body-size clumps of
temperate forest ecoregions discovered in Chapter 2 (Table 3-2).
The number of species in certain size ranges for the two
continents is displayed in Table 3-1. Within the smaller size range of
7.0 g 32.7 g, eastern North America had more species than Europe
(91 vs. 84), but this was not significant (P > 0.05); within the medium
size range of 32.8 g 105.0 g, eastern North America had more
species than Europe (43 vs. 27), but this was not significant at
conventional levels (P = 0.055); and within the large size range of
105.1 2085.0 g, Europe had more species than eastern North
America (51 vs. 43) but this was not significant (P > 0.05).
Boreal forest biome. The Gap Rarity Index found 8 clumps for
both the eastern North American and European avifauna lists. Match
analyses (8 to 15 clumps) of boreal forest avian communities
resulted in several significant matches (P < 0.05): comparing eastern
North America vs. Europe, 10 vs. 9, 13 vs. 9, 14 vs. 9, and 15 vs. 9
models were found to be significant. Of these models, the 10 vs. 9
model was chosen to represent the landscape structure of the boreal
forest biome because it had the most equitable number of clumps
(Figure 3-4). Using this 10-9 model, eastern N. America had 1 extra
clump between 27.4 g and 39.6 g (1.4 log g and 1.6 log g: Figure 3-4).








TEMPERATE FOREST BIOME


Clump and gap intervals of bird body-mass data




Europe
TemperateI





N. America1 111
Temperate


I I....I..... ... ....
0 0.5 1 1.5 2 2.5 3 3.5 4

Figure 3-3. Avian body-size clump structure of the eastern North American and the European
temperate forest biome. North American body-size clump structure has 10 clumps and the
European body-size clump structure has 7 clumps. The asterisks indicate 3 extra clumps in the
North American body-size clump structure between 1.2 and 1.7 log grams.





70

Table 3-1. The size ranges of European and eastern North American
temperate forest avian body-size clumps. The number of species in
each clump is also listed.


Europe Eastern North America

Size range(g) # of species Size range (g) # of species
0.0 6.9 2 0.0 6.9 5
7.0 32.7 84 7.0 16.2 56
16.3 21.5 16
21.6 32.7 19
32.8 102.0 27 32.8 52.0 20
52.1 105.0 23
103.0 188.0 16 105.1 186.0 11
189.0 570.0 22 187.1 576.5 18
571.0 2080 13 577.0 2080.5 14
> 2080.0 5 > 2080.5 2









Table 3-2. The size ranges of temperate forest avian body-size
clumps determined from Amherst, Austin, and Vancouver avifauna
lists (Chapter 2) compared to those from eastern North American
temperate forest avifauna lists (this chapter).


Amherst, Austin,
Vancouver regions Eastern North America

Size range (g) Size range (g)


0.0 6.9 0.0 6.9
7.0 16.5 7.0 16.2
16.6 21.6 16.3 21.5
21.7 33.75 21.6 32.7
33.76 61.6 32.8 52.0
61.7 115.0 52.1 105.0
115.0 184.0 105.1 186.0
185.0 576.5 187.1 576.5
577.0 2080.5 577.0 2080.5
> 2080.5 > 2080.5







The body-size clumps defined by the upper and lower masses are
shown in Table 3-3.
Body-size clumps defined by the upper and lower masses of
each clump resulted in slightly different size ranges between the two
continents (Table 3-3). Thus, to compare continental differences in
the number of species within each body-size clump, I arbitrarily
used the European size ranges. Within the smaller size range of 7.0 g
33.5 g, eastern North America had more species than Europe (71 vs.
56), but this was not significant (P > 0.05); within the medium size
range of 33.6 g 82.3 g, eastern North America had significantly
more species than Europe (25 vs. 13, P < 0.05); and within the large
size range of 82.4 g 2204.0 g, Europe had significantly more species
than eastern North America (46 vs. 29, P < 0.05).

Within Each Continent

Looking at each continent separately, the body-size clumps of
the temperate and boreal biome matched within smaller and larger
size ranges. However, the medium size range (16.3 g 316.0 g; 1.2
log g and 2.5 log g) had the most mismatch (Figure 3-5).

Trends in Body-size Clump Distributions

Here, the results are summarized in terms of whether both the
temperate forest and boreal biome showed some similar trends in
the results mentioned previously.
Between the two continents. First, both eastern North
American biomes had more species in the smaller size range (around
7.0 g to 33.0 g) and the medium size range (temperate forest: 34.0 g






BOREAL FOREST BIOME


Clump and gap Intervals of bird body-mass data


Europe Boreal



N. America Boreal


1111111


1 1111111 1 I
0 0.5 1 1.5 2 2.5 3 3.5
log body-mass (g)


Figure 3-4. Avian body-size clump structure of the eastern North American and the European
boreal forest biome. North American body-size clump structure has 10 clumps and the
European body-size clump structure has 9 clumps. The asterisk indicates 1 extra clump in the
North American body-size clump structure between 1.4 and 1.6 log grams.







Table 3-3. The size ranges of European and eastern North American
boreal forest avian body-size clumps. The number of species in each
clump is also listed.


Europe


Eastern North America


Size range (g) # of species Size range (g) # of species
0.0 6.9 1 0.0 6.9 4
7.0 16.6 31 7.0 15.2 39
16.7 33.5 25 15.3 27.3 23
33.6 82.3 13 27.4 39.6 15
39.7 89.0 22
82.4 185.0 15 89.1 185.5 11
186.0 365.0 11 186.0 365.0 3
366.0 570.0 8 366.0 577.0 8
571.0 2204.0 12 578.0 2354.5 6
> 2204.0 4 > 23545.0 2





TEMPERATE vs. BOREAL
Lump and gap Intervals of bird body-mass data


111

I lIl


I

II

II

II


Europe
Temperate
Europe
Boreal
N. America
Temperate
N. America
Boreal


Ill

II I

II I

I I I


0 0.5 1 1.5 2 2.5 3 3.5 4
log body-mass (g)
Figure 3-5. Avian body-size clump structure of the eastern North American and the European
boreal and temperate forest biome. The box C3 outlines a size range (log 1.2 g and 2.5 g) with
the most mismatch between the temperate and boreal zones on each continent.


11111


I








to 105.0 g; boreal forest: 34.0 g to 82.0 g) than their respective
European biomes. However, in the larger size range (temperate
forest: around 105.0 g 2080.0 g; boreal forest: around 82.4 g -
2204.0 g) fewer species occurred in both of the eastern North
America biomes than in Europe. Second, more body-size clumps
occurred in eastern North America than in Europe for both biomes.
The additional clumps appeared in the size range of 16.0 g 52.0 g
(1.2 log g and 1.7 log g: Figures 3-3 and 3-4).
Within each continent. On both continents, most of the
mismatch between the biomes occurred in the medium size range
(16.3 g 316.0 g; 1.2 log g and 2.5 log g) (Figure 3-5).



Discussion

Many new insights about the hierarchical landscape structure
of eastern North American and European temperate and boreal
biome were revealed by the clump analyses. In the following
discussion, I will (1) discuss the similarities and differences between
eastern North American and European avian body-size clumps, both
in terms of the number of clumps and the number of species within
certain size ranges, and (2) discuss the similarities and differences
between temperate and boreal avian body-size clumps on the same
continent in terms of the number of clumps within certain size
ranges.









Cross Continental Body-size Clump Comparison

How well did they match? The temperate and boreal forest
avian body-size clumps between eastern North America and Europe
both showed significant matches. New World and Old World
avifauna lists (from the same biome) had very little species or
genera overlap, but they had similar body-size clump structures.
This is evidence that body-size clump matches are not due to species
overlap. I propose that structuring processes in the temperate forest
biome produce a hierarchical landscape structure that is consistent
from one continent to the next. Essentially, over evolutionary time,
this hierarchical landscape structure is a "template" that is favorable
for only certain animal body sizes. Holling (1992b) proposed that
body-size clumps could be used to reflect existing hierarchical
structure of landscapes, and results from this study do not refute this
idea. In addition, the 10 temperate forest body-size clumps of
eastern North America in this study matched quite well with the 10
body-size clumps of temperate forest regions of North America
shown in Chapter 2. As mentioned previously, birds in each body-
size clump are theoretically responding to structure within the same
range of scales. If true, this match sets the stage to explore whether
birds in each clump respond to landscape structure within the same
range of scales.
Although body-size clumps matched between the two
continents, more clumps occurred in the medium size range in both
temperate and boreal biomes of eastern North America than of








Europe. The temperate forest biome of eastern North America had 3
extra clumps between 16.0 g and 52.0 g, and using the 10 vs. 9
model, the boreal forest biome of eastern North America had one
extra clump between 27.4 g and 39.6 g. These additional clumps
could be a result of the unique histories on these two continents. As
mentioned before, European forests (since the Pleistocene) have been
quite fragmented whereas North American forests have been
relatively extensive and continuous (Monkkonen and Welsh 1994).
This fragmentation difference between the two continents led to
marked differences in the floral diversity. In the temperate zone,
several studies have reported the low taxonomic diversity of
European forest flora, compared to North American flora (Kurten
1972, Sauer 1988). Many of the forest taxa became extinct in Europe
during the Pleistocene glaciations, and Huntley (1993) proposed that
the low taxonomic diversity of European forests was primarily due to
the limited amount of area available to forest taxa. The continuous
forest tracts of North America may have provided more diverse
landscape resources that could be exploited by medium-sized birds.
The additional clumps in eastern North America possibly reflect
these additional resources in N. American landscapes.
However, the relatively continuous forest tracts of North
America should provide more types of resources for all sizes of birds.
Why were there not additional clumps in the eastern North American
small and large size ranges when compared to the European small
and large size ranges? Probably, the hierarchical landscape structure
at small and large scales was preserved in European landscapes,
despite the high degree of forest fragmentation. Most likely, the








diversity of resources at small and large scales had been changed;
this would primarily affect the quantity of species at small and large
size ranges (see section below), not the position or the number of
clumps within these size ranges. At scales pertinent to medium-
sized birds, the landscape structure (e.g., forest cover) probably
changed the most. As landscapes become fragmented, I suggest that
medium-sized species theoretically would not be able to find enough
resources within remnant patches of forest nor could they travel
between patches. Thus, many forest patches may not be available to
medium-sized birds in fragmented landscapes.
Ouantity of species in each body-size clump. In both the
temperate and boreal biomes respectively, eastern North America
(compared to Europe) had more species in the medium size range.
This difference was significant for the boreal biome, but not
significant for the temperate biome at conventional statistical levels.
One explanation for this is that a greater diversity of resources is
available to medium-sized animals in eastern North America than in
Europe. The high degree of fragmentation in Europe left a number of
forest patches separated by some distance from one patch to the
next. Fragmentation probably selected against European medium-
sized animals because they could neither acquire enough resources
within each patch nor, because of their ambit, could they travel
between patches to gather enough resources.
This explanation is similar to what Morton (1990) proposed as
the mechanism of extinction of many middle-sized Australian
mammals. Changed fire regimes and the introduction of rabbits
caused the distribution of productive patches to become steadily







smaller and to be spaced steadily farther apart. Small animals could
still find enough resources in the remaining small patches whereas
large animals had an ambit large enough to travel between the
patches. Middle-sized animals went extinct because they could
neither find enough resources in each small patch nor travel between
patches. A similar scenario may have happened to medium-sized
European birds since the Pleistocene.
In the smaller size range, the slightly larger number of eastern
North American than European birds could also be attributed to the
greater amount of fragmentation in Europe than in eastern North
America. As mentioned earlier, the fragmented forests of Europe
have a low taxonomic diversity of forest flora due to limited amount
of areas available to forest taxa (Huntley 1993). Limited forest cover
also means decreased vertical height diversity across European
landscapes. Perhaps these two factors reduced the number of
smaller bird species that could specialize on forest vegetation in
Europe. On the other hand, extensive forests in eastern North
America may have provided more unique opportunities for small
bird species to specialize on specific types of vegetation at small
scales. This resulted in more small-sized species in North America.
It is interesting to note that eastern North America contains a
large number of neotropical migrants whereas only a few long-
distant migrants (that winter in Africa) occur in Europe (Monkkonen
and Welsh 1994, Rappole 1995). Most of the European migrants are
primarily early successional species whereas many N. American
migrants are found most often in mature forest (Monkkonen and
Helle 1989). These natural life history differences of eastern N.







American and European birds are an indication of how fragmented
landscapes may affect the natural life history characteristics of birds.
On possible explanation is that the extensive eastern North American
forests (compared to the fragmented European forests) provided a
greater amount of opportunities for small to medium sized birds to
specialize on forest vegetation.
Finally, the greater number of European than eastern N.
American birds in the larger size range may also be a result of the
higher degree of fragmentation in Europe than in eastern North
America. Perhaps large amounts of open areas, combined with
patches of forest, are suitable for larger bird species. The species
lists of both continents indicate that a majority of the larger birds are
raptors (Appendix B and C). Raptors generally prefer to nest and
forage in semi-open country (Ehrlich et al. 1988), and the
fragmented European landscape historically had more semi-open
areas than the landscape in eastern N. America. Perhaps a landscape
mosaic that contains a combination of open areas with patches of
vegetation (e.g., trees) provides more opportunities for large species
than a landscape that is completely forested or open. Thus,
fragmented European landscapes may have had more opportunities
for large birds to specialize on different aspects of this landscape
mosaic. This ultimately would allow for the presence of a greater
number of European than eastern N. American large-sized species.
Both biomes showed that fewer small- and medium-sized birds
and more large birds exist in Europe than in eastern North America.
These results were consistent for both biomes, but the trends were
not statistically significant (except for the boreal biome: medium-







sized birds). Also, it should be noted that the trends found in this
study could be explained by historical differences in landscape
structure on the wintering grounds (e.g., South America and Africa).
Perhaps the South American landscape provided more opportunities
for the speciation of small to medium-sized migrants, and this is the
reason for the increased number of birds in eastern North America
than in Europe. To address whether fragmentation disproportionally
affects small to medium-sized species, more comparisons are needed
between fragmented and unfragmented forested landscapes.
Multiple examples of the trends noted in this study will provide
more confidence that these trends are real.
Nevertheless, the cross continental comparisons in this study
suggest that eastern N. American birds in the small to medium size
ranges (7.0 g 105.0 g) are most susceptible to human-caused
fragmentation of forested landscapes. The European body-size
clumps, which consist of species that have experienced fragmentation
for a long period of time, have fewer species in the medium and
small size ranges than their eastern N. American counterpart. This
suggests that if eastern North America became as fragmented as
Europe, one would expect a dramatic population decline or extinction
of small to medium sized bird species. This may already be
happening in North America. Several eastern N. American breeding
passerine populations are hypothesized to be declining (e.g.,
Whitcomb et al. 1981, Terborgh 1989, Rappole 1995, although see
James et al. 1992) whereas most European passerine populations are
not declining (Opdam et al. 1985, Haila 1986). Monkkonen and
Welsh (1994) suggest that the evolutionary history of European







passerines resulted in an avifauna that is adapted to fragmentation.
European passerines are not sensitive to fragmentation because
European landscapes have been fragmented for a long period of time.
However, eastern N. American passerines have evolved in a
botanically more diverse and extensively forested landscape, and
some species (especially interior forest specialists) are particularly
sensitive to fragmentation.
North American passerines that are hypothesized to be
declining, such as the Wood Thrush (Hylocichla mustelina), Yellow-
throated vireo (Vireo flavifrons), and Hooded Warbler (Wilsonia
citrina), all have sizes that fall in the small to medium size range (7.0
g 105.0 g). Fragmentation of the breeding grounds may be only
one cause for the decline of some North American passerines, other
causes include tropical deforestation and the habitat decline of
stopover sites (Rappole 1995, Terborgh 1989, Hagan and Johnston
1992). All three causes probably play a role in the proposed decline
of some species. The results in this chapter are consistent with the
hypothesis that small to medium sized interior forest specialists will
be most affected in the near future if fragmentation of North
American breeding grounds reach proportions similar to Europe.

Within Continental Body-size Clump Comparison

As predicted, the avian body-size clumps of the boreal and
temperate biome matched within the smaller size range (0.0 g 16.3
g). However, the body-size clumps at the larger size range (> 185.0
g) also matched. The medium size range (16.3 g 185.0 g) had the
most mismatch when body-size clumps of temperate and boreal







biomes were compared to each other on the same continent. This
suggests that landscape structure at small and large scales is similar
between boreal and temperate zones, but structure at medium scales
is different between the two biomes. This of course depends on
whether the position of clumps at the medium size ranges is a true
reflection of the hierarchical structure in each biome. But why is
there a mismatch between clumps in the medium size range and a
match at the smaller and the larger size range?
Most likely, mesoscale disturbance processes would have the
most appreciable effect on animal body-size clumps. Many
mesoscale disturbances alter the spatial distribution of resources at
scales relevant to most organisms. For example, mesoscale
disturbance processes, such as fires, insect outbreaks, storm damage
and diseases, alter the spatial distribution and abundance of trees
and plants at a scale from 0.03 100.0 km (Holling 1992b) If clumps
reflect structure in landscapes, then landscapes with different
disturbance regimes should show some differences in body-size
clump patterns. This probably will not occur at all scales. Where
body-size clumps match between different landscapes would indicate
the range of scales where structure is geometrically similar. Where
there is a mismatch would be an indication of the range of scales
where landscape structure is geometrically different.
Thus, the match of boreal and temperate avian body-size
clumps at the small and large size ranges and the mismatch at the
medium size range suggests that structure is similar at small and
large scales but different at middle scales. Boreal and temperate
mesoscale processes may produce geometrically different types of







structure relevant .to medium-sized birds, but similar structure
relevant to small and large birds. One possible mesoscale process
that affects landscape structure at scales relevant to avian
communities is fire. Differences in the boreal and temperate fire
regime may explain the apparent mismatch in the avian clump
structure (at the medium size range) between the two biomes. In
the boreal forest, the fire regime burns a site approximately every
50 to 100 years (Heinselman 1973, Zackrisson 1977), and each fire
typically burns hundreds of thousands of square kilometers (Wein
and MacLean 1983). In contrast, the fire regime of deciduous forests
rarely burn at such a large scale; most of the disturbance in
temperate forests is controlled by single and multiple treefalls
(Oliver and Stephens 1977). As a result, the age structure of
deciduous forests is uneven whereas large tracts of boreal forest
tend to be even aged (Cayford and McRae 1983). In part, the
different fire regimes cause structural differences between
temperate and boreal forests (another possible disturbance includes
insect outbreaks see Holling 1992a). Temperate forests consist of a
variety of patches with different tree species whereas boreal forests
consist of primarily large tracts of forest dominated by a few species.
Perhaps these structural differences between the boreal and
temperate forest are primarily relevant to medium-sized birds.

Future Research

All of the above results are consistent with the hypothesis that
avian body-size clumps reflect the hierarchical structure of biomes.
These results provide more confirmation to the textural-







discontinuity hypothesis, but more examples are needed. Future
studies should compare the body-size distributions of avifauna from
other areas in the temperate and boreal biome. For example, boreal
and temperate avifauna lists from the eastern Palearctic
biogeographic realm could be compared. In this comparison, do the
same body-size clumps appear as what was revealed in this study?
What about comparisons between avifauna in fragmented and
unfragmented landscapes; do the medium-sized species tend to be
the most affected in fragmented landscapes? Additional research in
this area will be fruitful to determine the generality of clumps.

Summary and Conclusions

The match between the body-size clump structure of eastern
North American and European boreal and temperate avian body-size
clumps suggests that the hierarchical landscape structure of each
biome is similar between the two continents. Comparing the number
of species in each clump, fewer European than eastern North
American bird species were present in the small and medium size
ranges, but more European than eastern North American bird species
were present in the large size range. I hypothesized that the cross-
continental difference in the number of species at the different size
ranges was a result of historical landscape changes in each continent.
The highly fragmented European landscape contained less resources
for small- and medium-sized birds, but more resources for large
birds. The larger number of body-size clumps found in eastern
North American avifauna lists (both biomes) was also attributed to
the higher degree of landscape fragmentation in Europe than in





87

eastern North America. Further, comparisons within a continent
revealed that the avian body-size clump structures of the boreal and
temperate zones matched at the small and large size ranges, but
some mismatch occurred at the medium size range. This was
interpreted as an indication that landscape structure between the
two biomes are similar at small and large scales, but different at
middle scales. Overall, these results provide more evidence that
animal body-size clumps may reflect the hierarchical structure in
landscapes.




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EMF1YD99Q_EZGSMT INGEST_TIME 2013-01-23T14:20:49Z PACKAGE AA00012938_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES