Citation
The Relative Influences of Predation and Prey Availability on Ardeid Breeding Colony Site Selection

Material Information

Title:
The Relative Influences of Predation and Prey Availability on Ardeid Breeding Colony Site Selection
Copyright Date:
2008

Subjects

Subjects / Keywords:
Bird nesting ( jstor )
Breeding ( jstor )
Everglades ( jstor )
Foraging ( jstor )
Modeling ( jstor )
Polygons ( jstor )
Statistical models ( jstor )
Vegetation ( jstor )
Wading birds ( jstor )
Water depth ( jstor )
The Everglades ( local )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Embargo Date:
5/1/2005

Downloads

This item has the following downloads:


Full Text












THE RELATIVE INFLUENCES OF PREDATION AND PREY AVAILABILITY
ON ARDEID BREEDING COLONY SITE SELECTION














By

MATTHEW JOHN BOKACH


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE

UNIVERSITY OF FLORIDA


2005

































Copyright 2005

by

Matthew John Bokach


























This thesis is dedicated to my parents Mark and Cathy, who have
weathered the many twists and turns of my life's last decade with unconditional
love and support, leavened with just a touch of legitimate bemusement.















ACKNOWLEDGMENTS

I could not have conducted this research or even made it through graduate school

without the support and assistance of a humbling number of other people. Dr. Peter

Frederick has been the best advisor I could have asked for. Without his support, patience,

and humor, I would not have finished this degree. My data on the wading bird colony

sites was collected by a long series of his former students and technicians, and this

research is therefore as much a product of their work as mine. Dr. Graeme Cumming and

Dr. Jane Southworth, my other two committee members, gave me invaluable advice

through nearly every stage of the research. My initial explorations into the daunting task

of analyzing these data benefited immensely from Dr. Tim Fik's statistical insights and

encouragement. I gratefully acknowledge Dr. Carl Fitz, Mr. Jason Godin, and Mr. Ken

Rutchey of the South Florida Water Management District; and, Mr. Troy Mullins of the

National Park Service, for providing me with the hydrology and vegetation data used

herein. Lorraine Heisler, Les Vilchek, and Shawn Komlos all provided insights

concerning the development and appropriate uses of these data. Thanks to Theresa

Burcsu and Desiree Price, I was allowed constant access to the computer labs of the

University of Florida Geography Department, even when not enrolled in one of their

classes; I will always think of Turlington Hall as the "home" of this research. Lin

Cassidy, Andres Guhl, and especially Matt Marsik all gave me valuable assistance with

GIS questions. I blame Carl Evans for both complicating and vastly simplifying my life

by introducing me to Matlab. The patience and support of my boyfriend Rodney Brown









was a steadying influence through most of this long process. Finally, I thank my family

for all of their love and support through the years.

I was supported financially for nearly my entire time at the University of Florida by

a research assistantship through the College (now School) of Natural Resources and

Environment. I extend my most sincere thanks to Dr. Steve Humphrey, Cathy Ritchie,

and Meisha Wade for giving me the opportunity to pursue this degree, and for extending

my funding for an extra semester.
















TABLE OF CONTENTS



A C K N O W L E D G M E N T S ................................................................................................. iv

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

LIST OF FIGURES ........................................... ............................ viii

A B S T R A C T ........................................................................................................ ........... ix

CHAPTER

1 IN T R O D U C T IO N ................................................. .............................................. .

2 M E T H O D S .......................................................................................................... ... 5

S tu dy A rea ....................................................................... ................................. . 5
D ata Sources ................................................................................. . .. ...............6
W ater D epth G rids ... ... ......................................... ....................... . .......... 6
V egetation M aps... ....................................................................... . .......... 6
W ading Bird Colony Locations.................. ..................................................8...
D istributional R relationships ......................................... ........................ ...............9...
C alculation of V ariables ............. .. ............... ..............................................9... .
M u ltiv ariate A n aly ses .... ... ......................................... ....................... ............... 11

3 R E S U L T S ................................................................................................................. .. 1 7

D distribution of C olony Sites .................................................................. ............... 17
Bootstrap of Foraging H habitat Calculation............................................ ............... 17
M ultiv ariate A naly ses .... ... ......................................... ....................... ............... 18

4 D ISCU SSION ............................................................................ ....... ........ ............... 38

E effects of Site A availability ......................................... ........................ ................ 38
M odel Perform ance ............................. ..... ... ............... 38
R esponses to H ydrological V ariability .................................................. ................ 39
M anagem ent Im plications ................. .............................................................. 40

L IT E R A T U R E C IT E D ................................................... ............................................. 42

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















LIST OF TABLES


Table page

1-1 V ariables and m ethods of calculation ................................................... ...............4...

1-2 Weeks of nest initiation and duration of breeding period for three focal species......4

3-1 Number of colonies inhabited by each species, 1993-2000 ...............................19

3-2 Values of FOR calculated in 30 randomly-chosen cells using constant
proportions, compared to approximate 95% confidence intervals from a
bootstrap analysis involving 1000 iterations where proportions were allowed to
ran d o m ly v ary ......................................................................................................... 2 0

3-3 Logistic regression models tested for Great Blue Herons and associated
statistical v alu e s........................................................................................................ 2 1

3-4 Logistic regression models tested for Great Egrets and associated statistical
v alu e s ...................................................................................................... ........ .. 2 1

3-5 Logistic regression models tested for Tricolored Herons and associated
statistical v alu es........................................................................................................ 2 2

3-6 Best models for each species and associated measures of classification
p erfo rm an ce .............................................................................................................. 2 3

3-7 Relative importance of variables to each species................................................23















LIST OF FIGURES


Figure page

2-1 Southeastern Florida showing the location of WCA3 within the larger
Everglades ecosystem .............. ................. ................................................ 14

2-2 Everglades Landscape M odel cells in W CA3 ..................................... ................ 15

2-3 Algorithm for determining which models to test for each species........................ 16

3-1 D distribution of ardeid colonies in W CA 3 ........................................... ................ 24

3-2 Relationship of Great Blue Heron colonies relative to all available sites in
WCA3 as shown by linearized Ripley's K (l[d]) graphs ....................................26

3-3 Relationship of Great Egret colonies relative to all available sites in WCA3 as
shown by linearized Ripley's K (l[d]) graphs..................................... ................ 30

3-4 Relationship of Tricolored Heron colonies relative to all available sites in
WCA3 as shown by linearized Ripley's K (l[d]) graphs ....................................34















Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science

THE RELATIVE INFLUENCES OF PREDATION AND PREY AVAILABILITY ON
ARDEID BREEDING COLONY SITE SELECTION

By

Matthew John Bokach

May 2005

Chair: Peter Frederick
Major Department: Natural Resources and Environment

Nest predation and prey availability are two of the most important factors affecting

breeding success in long-legged wading birds (Ciconiiformes). I investigated whether the

breeding colony site selection of three species of herons and egrets (family Ardeidae) in

the Florida Everglades was influenced by environmental characteristics that mediated

these two factors. These characteristics were 1) likelihood that sites remained inundated

throughout the breeding period; 2) amount of foraging habitat around sites; 3) average

weekly proportional decline in water depths around sites throughout the breeding period;

4) spatial variation in water depths around sites at the time of nesting. Variables were

calculated within a geographic information system using both raster and vector inputs. I

used measures derived from the Akaike information criterion to select the best logistic

regression model and to evaluate the relative importance of these four variables for

colony site selection in each species.









Amount of foraging habitat and likelihood of remaining inundated were the most

important variables influencing colony site selection by all three species. Great Blue

Herons (Ardea herodius) also selected sites with high rates of average weekly

proportional declines in water depth, but it is likely this variable was a proxy for deep

water for this species rather than a reflection of prey availability. Overall, these species

seemed to favor stable (rather than variable) hydrological conditions. This might also

indicate that their colony site selection is based on conditions at the time of nesting rather

than an attempt to predict future conditions. These results confirm the importance of

managing the Everglades to maximize the extent of slough habitats, if the goal is to

increase breeding populations of wading birds therein.














CHAPTER 1
INTRODUCTION

Two of the most important factors affecting the breeding success of long-legged

wading birds (Ciconiiformes) are egg and chick predation, and the availability of

adequate food to raise chicks to fledging (Taylor & Michael 1971; Frederick & Collopy

1989a; Frederick & Spalding 1994; Frederick 2002). Although wading birds cannot

control the behaviors of predators or their prey, it is likely that selection has favored the

recognition of breeding colony sites with beneficial environmental characteristics that

mediate or constrain those behaviors.

For example, wading birds nearly always nest on islands or in trees and/or shrubs

that are inundated at their base, a pattern that has most often been interpreted as a strategy

for deterring mammalian predators (Rodgers 1987; Bancroft et al. 1988; Frederick &

Collopy 1989b; Smith & Collopy 1995). However, whether this pattern represents actual

selection for this characteristic has never been quantitatively studied. With regard to prey

availability, although previous studies (Gibbs et al. 1987; Gibbs 1991; Gibbs & Kinkel

1997; Baxter & Fairweather 1998; Bancroft et al. 2002) have shown that wading birds

select colony sites that maximize the amount of wetland habitats within a reasonable

foraging range, other environmental characteristics (e.g., water depth) are better

determinants of wading birds' ability to capture prey.

If wading bird colony site selection represents an attempt to deter predators and/or

maximize prey availability, then used sites should be measurably different from available

unused sites with relation to environmental characteristics that affect these two factors. I









identified one characteristic that deters predators from reaching nests and three

characteristics that are likely to influence prey availability around sites (Table 1-1) and

make the following predictions.

Prediction 1: Used sites will have a greater likelihood of remaining inundated

throughout the breeding period than unused sites. This characteristic seems to be a

good measure of a site's ability to deter mammalian predators (Frederick & Collopy

1989b; Smith & Collopy 1995), which are generally the most destructive in their effects

on wading bird colonies (Rodgers 1987; Post 1990; Smith & Collopy 1995).

Prediction 2: Used sites will be surrounded by more open or sparsely

vegetated habitats than unused sites. Not all wetland habitats are appropriate foraging

grounds for wading birds. In particular, they are known to avoid dense vegetation (such

as monospecific stands of sawgrass [Cladiumjamaicense] or cattail [Typha latifolia]),

which interferes with their visual or tactile hunting techniques, provides prey animals

with more hiding places, and could serve as a hiding place for predators of the birds

themselves (Hoffman et al. 1994; Smith et al. 1995; Surdick 1998).

Prediction 3: Used sites will be located in areas that experience greater

proportional declines in water depth throughout the breeding season than unused

sites. Wading birds are generally limited to foraging in water that is shallower than the

length of their bills or legs (Custer & Osborn 1978; Powell 1987). They are often

attracted to areas of declining water depth where prey have been concentrated into pools

or depressions that remain inundated late in the season (Kushlan 1976; Bancroft 1989;

Frederick & Collopy 1989a; Smith 1995; Gawlik 2002; but see Frederick & Spalding

1994 for a critique of this idea).









Prediction 4: Variability in water depths at the time of nesting will be greater

around used sites than around unused sites. Water levels are dynamic and

unpredictable in most wetlands. Heterogeneous topography around a colony site could

therefore provide a foraging advantage to wading birds as this would provide the most

diversity of potential foraging sites at almost any time or water condition (Kahl 1964;

Bancroft et al. 2002).

I used logistic regression models to assess the relative influence of these four

variables on the colony site selection of three wading bird species in the ciconiiform

family Ardeidae (Table 1-2) over a period of 8 years in the Florida Everglades. Measures

derived from the models' Akaike information criteria (AIC) (Akaike 1973; Burnham &

Anderson 2002) were used to choose the best model and determine the relative influence

of the variables on the colony site selection for each species.









Table 1-1. Variables and methods of calculz
Variable Abbreviation
Affecting predator deterrence
Likelihood that a
cell will remain
inundated for
the duration of
the breeding
period
Affecting prey availability
Amount of foraging
vegetation around FOR
sites


Average rate of
proportional
decline in depths
around a site for
the duration of a
species' breeding
period


FWI


Method of calculation

Used GREATERTHAN function to
output number of weeks over duration
of a species' breeding period that
depths remained above 0; divided by
number of weeks to yield values
between 0.0 and 1.0


See text for details


Calculated proportional decline in water
depths for every week during a
species' breeding period (rises in depth
expressed as 0); averaged these values
over all weeks; used the
FOCALMEAN function to average
these cell averages over 3x3 cell
neighborhoods, yielding values
between 0.0 and 0.5


Spatial variation in
water depths Used FOCALSTD function to calculate
around a site at WV standard deviation of depths in 3x3
week of nest cell neighborhoods
initiation
All calculations carried out in ESRI Arc/INFO workstation v.8.3 except FOR, which was
calculated in ESRI ArcMap v.8.3. Weeks of nest initiation and duration of breeding
periods are listed in Table 1-2.

Table 1-2. Weeks of nest initiation and duration of breeding period for three focal species
Week of nest
Species initiationa Duration Source for duration
initiation


Great Blue Heron
(Ardea herodius)
Great Egret 9th
(Ardea alba)
Tricolored Heron
(Egretta tricolor)
aSource is Frederick (pers. comm.)


14 weeks Butler 1992
10 weeks McCrimmon et al. 2001

11 weeks Frederick 1997














CHAPTER 2
METHODS

Study Area

I studied ardeid colony site selection within Water Conservation Area (WCA) 3, a

2,350-km2 human-made impoundment in the central Everglades (Figure 2-1). Most of the

wading birds that breed within the Everglades ecosystem have selected colony sites

within this impoundment since the early 1970s (Ogden 1994; Frederick & Ogden 2001).

The annual hydrology of WCA3 is characterized by a decline in water depths during the

dry season between November and April, and most annual rainfall occurring during a

subtropical wet season between June and November. The northern end of the

impoundment is shallow and quick to dry, while the southern end is almost permanently

inundated. This same gradient exists to a lesser extent from west (where flow of water

into the adjoining Big Cypress National Preserve is unimpeded) to east (the enclosed sub-

impoundment WCA3B created by the L-67 canals, Figure 2-1). Wading birds therefore

have a wide range of hydrological conditions within which to select sites in WCA3.

The vegetation of WCA3 is characterized by open "wet prairie" communities in

deeper areas interspersed between narrow ridges covered in cattail and/or sawgrass

(Gunderson 1994). The patches of woody or shrubby vegetation that serve as potential

nesting sites for wading birds are scattered throughout. Large areas in the northern end of

the impoundment are dominated by an almost complete monoculture of cattails.









Data Sources

Water Depth Grids

I derived hydrological variables using estimated water depths from the Everglades

Landscape Model (ELM; Fitz et al. 2004), which simulates the flows and stages of water

across the entire Everglades ecosystem at a resolution of 1 km2. Model output was

available for the period 1993-2000, so these 8 years comprised the duration of my study.

Each output layer is an Arc/Info grid (Environmental Systems Research Institute 2002)

whose values are the estimated weekly average above-ground depth of water in each cell.

The model generally performs well both spatially and temporally when compared to

actual depths measured throughout the landscape (Fitz et al. 2004). The largest errors

occur in cells containing canals, as these contain the most intracellular variation in water

depths. ELM 1 km2 cells, rather than discrete colony sites, were my units of analysis as

these were the coarsest in resolution of all the data.

Vegetation Maps

I used a polygon vector geographic information system (GIS) layer showing the

distribution of vegetation in the central and southern Everglades and based on aerial

infrared photography acquired in 1995 over the entire area of WCA3, Everglades

National Park, and Big Cypress National Preserve. The latter two areas were included in

my analysis since they are within the typical foraging range of birds nesting near the

boundaries of WCA3 (Bancroft et al. 1994). The distinctive reflectance signatures of

different types of vegetation within the aerial photography were used to digitize polygons

representing discrete patches. These could contain up to three specific vegetative classes,

assigned hierarchically. After comparing the polygon layer to 1999 United States

Geological Survey (USGS) digital orthophoto quarter quads (DOQQs), I was satisfied









that this layer was an accurate depiction of the vegetation in WCA3 for the entire time

period of this study.

Since all reports of nesting by ardeid wading birds are on trees or shrubs in WCA3,

I defined potential colony sites as polygons that contained at least one class of woody

vegetation. Among the classifications used in the vegetation map these included: all

variants of forest, all variants of shrublands, Melaleuca, Brazilian pepper, disturbed fish

camp sites, spoil areas, and artificial "deer islands." I aggregated contiguous polygons

that contained at least one of these vegetation types into single discrete patches, which I

hereafter refer to interchangeably as "patches of woody vegetation" or "potential colony

sites."

The locations of wading bird colonies (see below) revealed that a small proportion

(-1.1%) of the potential colony sites did not occur within a reasonable distance from a

patch of woody vegetation. In October 2002 I visited all such sites in WCA3B and in

WCA3A north of Highway 1-75 (Figure 2-1), and determined that in most cases either

some woody vegetation had not been included in an existing polygon, or that a single tree

or small patch of shrubs was present and should have itself been digitized as a polygon.

In these cases, I therefore either added a woody vegetation classification to the second or

third tier of the existing polygon, or digitized the patch myself using the 1999 USGS

DOQQs.

Water model cells that either contained no potential colony sites, or contained only

the "clipped" edges of sites whose bulk were in an adjacent cell, were excluded from

further analysis since wading birds could not have nested in them. At the scale at which I









calculated the variables (see below), this removed less than 3% of the total area of WCA3

from analysis (Figure 2-2).

Wading Bird Colony Locations

Locations, species composition, and size of wading bird colonies were determined

using both aerial and ground survey techniques (Frederick et al. 1996). Aerial surveys

were designed such that the entire ground surface of the study area was visible along at

least one east-west transect. These were spaced 2.9 km apart, flown at 245 m altitude, and

were conducted monthly between January and June of each year. Dark-colored species

are often not visible in aerial surveys (Frederick et al. 1996), and for these species

ground-based airboat searches were conducted. All tree islands in the study area were

approached and searched for signs of nesting in the middle to late part of the nesting

season (April through May). Colony site coordinates were determined by handheld GPS

receivers, and populations of every species present were estimated by counting nests

and/or number of adults that flushed from the site. In cases where a colony was visited

more than once per season, I used the peak count of nests at that colony for a given

species.

I created point vector GIS layers representing the colony locations for each of the

three focal species for every year. Given that point coordinates reflected the location of

the airboat or airplane at the time of data collection, rather than the actual location of the

colony, I used a spatial join operation to shift points to the centroid of the nearest patch of

woody vegetation. In general, these shifts were not more than 100 m for points collected

on the ground, or 300 m for points collected during aerial surveys. I then overlaid the

water model grid and classified all cells as either used or unused by each of the species

for every year.









Distributional Relationships

To rule out the possibility that the distribution of these species' colonies was

simply a reflection of the distribution of available sites, I calculated Ripley's K statistic

for each species-year pattern of colony sites as well as for the parent pattern of all

available sites, at 500 m intervals from 500 to 20,000 m. This statistic is a

scale-dependent measure of spatial dispersion within a defined area (Fotheringham et al.

2000). However, it can also be used to assess whether two patterns within the same area

are spatially similar, especially when one is a subset of the other. Differences between

two patterns' K statistics (i.e., the linearized Ripley's K, or l[d]) indicate that they are

spatially dissimilar, with positive values of 1(d) indicating a dispersed pattern relative to

the parent pattern, and negative values of 1(d) indicating a clustered pattern relative to the

parent pattern. Since there is no significance test to determine how different from 0 an

1(d) value must be to indicate that the patterns are different, I used bootstrapping to create

sample-size-dependent 95% confidence intervals against which to compare each species-

year point pattern (Fotheringham et al. 2000).

Calculation of Variables

All variables were calculated at the scale of a 3x3 neighborhood of cells (i.e., 3 km

x 3 km) centered on the cell in question except INUN, which was calculated at the scale

of the cell itself. I chose these small scales to maximize the spatial independence

between samples, since larger neighborhoods would have increased the distance required

between two cells before their neighborhoods did not overlap. The calculation of the

three hydrological variables is described in Table 1-1.

To calculate FOR I first identified all polygons in the vegetation map that contained

any of the foraging habitat types. Within the map's classification system these included:









all variations of savanna; spike rush and maidencane-spike rush prairies; all variations of

non-graminoid emergent marsh; open water; mixed mangrove scrub; and, buttonwood

and saw palmetto scrubs.

Vegetation classes were hierarchically assigned to polygons qualitatively (e.g., a

large area of wet prairie might contain some scattered shrubs), yet I needed a quantitative

estimate of the areal extent of each class within polygons. Since vegetation classes could

be assigned to any of the tiers within a polygon, with adjoining polygons often having the

same classes in reversed order, the least biased method of estimating these extents was to

allocate the same proportion to a given tier in every polygon. In polygons containing two

classes of vegetation, I allocated 70% of the area to the dominant vegetation type and

30% to the secondary; in three-class polygons, I allocated 50, 30, and 20% of the area to

the three tiers respectively. After allocating a polygon's area between its vegetation

classes, I calculated the total extent of foraging habitat present within the polygons. I

intersected these polygons with a square grid polygon layer corresponding to the ELM

grid cell boundaries to assign the vegetation polygons to the cell they inhabited, and

summed the areas of foraging vegetation present within every cell. I recalculated the area

of foraging vegetation in polygons that had been split by cell boundaries by multiplying

their new area by the same proportion of foraging vegetation calculated for the parent

polygon. Finally, I calculated the 3x3 neighborhood sums for every cell in WCA3.

To ensure that the proportions I chose did not bias the ultimate values of FOR, I

randomly chose 30 cells and performed a bootstrap analysis whereby the proportions

allocated to vegetation classes within their polygons were allowed to randomly vary.

Since the vegetation classes were originally assigned hierarchically, I constrained the









proportions such that the dominant class proportion was always greater than the

secondary, and this was always greater than the tertiary. I calculated FOR for each cell

1000 times based on these randomly-determined proportions and used these values to

create an approximate 95% confidence interval for the "real" value of FOR in each cell. I

then determined how many of the values of FOR calculated in the 30 cells using the

proportions described above fell within these confidence intervals.

Multivariate Analyses

I used the Number Cruncher Statistical Systems software (NCSS, Hintze 2001) to

evaluate logistic regression models where the response variable was the use (presence of

at least one colony) or non-use of a water model cell by a species. Model inputs consisted

of the values for all cells used by a given species in all 8 years combined with an equal

number of randomly-selected unused cells from each year. I bootstrapped these

randomly-selected sets for every species-year to ensure that their mean values for all four

variables were within the 95% confidence interval suggested by the bootstrap.

Results of the logistic regression analyses were interpreted within the information-

theoretic approach of Burnham and Anderson (2002). Multiple models were evaluated

and the Akaike information criterion (AIC; Akaike 1973) calculated for each one thus:

AIC = -2(log-likelihood ratio) + 2K (2-1)

where K was the number of parameters estimated in the model. The model with the

lowest AIC was the "best" model among all those evaluated. The last term in Eq. 2-1

therefore represented a "penalty" against a model's AIC value for every variable added to

the model.









The performance of a given model relative to the best one in the set was expressed

using three values derived from the AIC values. The AIC difference (A,) for model i was

calculated:

A, = AIC, AICm,, (2-2)

This was converted to "Akaike weights" (w,):

R
w, = exp(-1A,)/ exp(- A,) (2-3)
r=l

where the denominator in Eq. 2-3 was the sum of all models' A, values. The w, for model

i was a measure of the likelihood (often expressed as a percent) that it is the best model.

An analogous quantity is the evidence ratio for model i:

evidence ratio, = w, / wj (2-4)

where w, was the w, value for the best model in the set. The evidence ratio was a measure

of "how much" better the best model was than model i. Because they relied on the sum of

all models' A, values, the w, and evidence ratios were recalculated every time a new

model entered the set.

I followed the algorithm shown in Figure 2-3 to determine the models tested for

each species. Upon identifying the best model for each species, I report three measures of

its classification performance (Fielding & Bell 1997): 1) the number of used and unused

cells correctly classified; 2) the Kappa statistic (K), which measures the extent to which a

model's predictions are correct above the proportion expected simply by chance; and, 3)

the area under the model's ROC curve, which measures the likelihood that the value

assigned by the final model to a randomly-selected "used" cell will be greater than that of

a randomly-selected "unused" cell. I also summarized the relative importance of the






13


variables tested for each species by adding the wi values for all models in which a given

variable is included (Burnham & Anderson 2002: 167-168).




















MIiccosukee
Indian
Reservation


t Palm
11


Iort
Lauderdale


Miami


K lometers
0 10 20 30 40


Figure 2-1. Southeastern Florida showing the location of WCA3 within the larger
Everglades ecosystem. The dotted areas along the east coast show the extent
of urban landscapes and their proximity to what is left of the Everglades.












**.** **- ********** ****o****** *







*094 9 0*
****** *** :*:*::* : *******:
^ ***** ^:******* :* :** ** *** *





9*0*0*9I 0 9099EEEEEEE a
0||||||||||||||||||||||0,E E E
******* 0** *060* *
9999 0*0*0 000 R IM 99


Kilometers
0 5 10 15


Figure 2-2. Everglades Landscape Model cells in WCA3. Black dots indicate cells that
contained potential colony sites and were included in multivariate analysis.
The darker shaded area indicates all cells that were within a 3x3 cell
neighborhood of analyzed cells and whose characteristics were therefore
included in the calculation of variables. Less than 3% of the area of WCA3,
represented by the unshaded areas, was excluded from analysis.











Test saturated model (all 4
variables) and all 3-term models.


Test all 2-term models
nested within best model


Test all 1-term models
nested within best model


BEST MODEL


Test all models comprised of best model
+ single second-order interaction term
between component variables.


Figure 2-3. Algorithm for determining which models to test for each species. At every
step the current best model competed with lower-term models comprised of its
component variables. Once the best single-order model was identified, models
that added a single second-order interaction term were tested.


C'J














CHAPTER 3
RESULTS

Distribution of Colony Sites

Table 3-1 summarizes the number of colonies and water model cells inhabited by

each species in each year, and Figure 3-1 presents the annual maps of these distributions.

Figures 3-2 through 3-4 show the l(d) graphs for each species-year combination, along

with their corresponding approximate 95% confidence intervals from the bootstrap

simulation. Great Blue Heron colonies were extremely clustered relative to the parent

pattern of all potential colony sites in all years except 1998, when the l(d) values were

closer to the lower bootstrap limit. Great Egret colonies were dispersed relative to the

parent pattern at all scales in 1993, 1994, and 1995; at small scales in 1996, and at large

scales in 1998. Their colony patterns were mostly indistinguishable from the parent

pattern at practically all scales in 1997, 1999, and 2000. Tricolored Heron colony patterns

showed the most variability: they were highly dispersed relative to the parent pattern at

medium to large scales in 1993; highly clustered at all scales in 1996; slightly clustered at

small scales in 1994 and medium scales in 1999 and 2000; and, otherwise mostly

indistinguishable from the parent pattern in 1995, 1997 and 1998.

Bootstrap of Foraging Habitat Calculation

Table 3-2 lists the values of FOR calculated for 30 randomly-chosen cells using the

constant proportions described above and the upper and lower values of the approximate

95% confidence interval from the bootstrap analysis. Four of the 30 cells contained only

polygons that did not vary in their values of FOR. In only one of the remaining cells was









the value of FOR calculated using the constant proportions outside the 95% confidence

interval from the bootstrap; however, this cell's value was within the 99% confidence

interval. This is strong evidence that the constant proportions used to calculate this

variable did not bias the analysis.

Multivariate Analyses

I tested 11 models for Great Blue and Tricolored Herons, and 12 for Great Egrets

(Tables 3-3 through 3-5). A single best model (w, > 98%) was found for the first two of

these species, while the best two models for Great Egrets were nearly indistinguishable

(w, values of 29.20% and 23.18%, respectively). Table 3-6 lists the best models) for each

species and its associated measures of classification performance. Best models correctly

classified between 68.8% (Great Egrets, model 1) and 78.9% (Great Blue Herons) of the

used sites, and between 36.7% (Great Egrets, model 1) and 56.7% (Great Blue Herons) of

the unused sites. Kappa statistics were routinely low, ranging between 0.055 (Great

Egrets, model 1) and 0.357 (Great Blue Herons). The areas under the ROC curves also

indicated relatively poor model performance; these ranged between 0.524 (Great Egrets,

model 1) and 0.745 (Great Blue Herons).

The amount of foraging habitat (FOR) around potential sites was always the most

important variable, with all Ew, values greater than 98% (Table 3-7). It was followed by

the likelihood that a site would remain inundated (INUN), with Yw, values ranging

between 83.88 and 99.99%. As predicted, all three species responded to both of these

variables positively, i.e., used cells had higher values overall than unused cells. The

pattern for the remaining variables was less consistent across species, with Great Blue

Herons responding strongly and positively to FWI, and Great Egrets responding weakly

and negatively to WV (Table 3-7).






19

Table 3-1. Number of colonies inhabited by each species, 1993-2000. Number of water
model cells inhabited in parentheses.
Great Blue Heron Great Egret Tricolored Heron
1993 156(127) 32(32) 11 (11)
1994 200 (153) 45(44) 46(43)
1995 298(229) 39(39) 40(39)
1996 167(136) 45(43) 52(42)
1997 101 (90) 34 (33) 8 (8)
1998 110(97) 44(41) 23(23)
1999 258(206) 77(71) 66(63)
2000 309 (235) 71 (62) 57 (53)









Table 3-2. Values of FOR calculated in 30 randomly-chosen cells using constant
proportions, compared to approximate 95% confidence intervals from a
bootstrap analysis involving 1000 iterations where proportions were allowed
to randomly vary.
Lower limit of Value of FOR Upper limit of Observed within
95% confidence calculated using 95% confidence confidence
interval constant proportions interval interval?
75 75 75 Constant value cell
251 339 644 Yes
750 750 750 Constant value cell
118 887 1,155 Yes
149 1,742 2,389 Yes
4,321 4,321 4,321 Constant value cell
5,925 5,925 5,925 Constant value cell
9,763 10,296 12,057 Yes
1,526 20,539 29,209 Yes
24,992 25,858 26,380 Yes
20,804 35,765 42,190 Yes
74,514 80,693 84,326 Yes
101,367 103,625 119,108 Yes
146,646 149,072 150,791 Yes
172,177 218,953 244,864 Yes
163,585 256,693 339,316 Yes
281,378 286,925 305,647 Yes
28,983 289,828 328,472 Yes
337,236 352,078 356,356 Yes
115,366 360,059 418,467 Yes
77,850 376,847 504,702 Yes
92,661 382,298 501,854 Yes
373,516 445,018 635,874 Yes
220,282 470,162 569,560 Yes
413,803 477,840 597,608 Yes
271,067 504,240 587,148 Yes
307,407 507,071 567,364 Yes
385,418 532,264 559,474 Yes

504,530 607,378 604,063 No, 99%i
confidence interval
531,142 638,535 646,606 Yes











Table 3-3. Logistic regression models tested for Great Blue Herons and associated statistical values. Italicized models are the best
from their subset, while the best overall model is shown in boldface.


Model Log-Likelihood
FOR+WV+FWI+INUN -1492.38
FOR+WV+FWI -1495.94
FOR+ FWI+ INUN -1493.21
FOR+WV+INUN -1495.21
WV+FWI+INUN -1721.50
FOR+FWI -1496.75
FOR+INUN -1496.40
FWI+INUN -1745.57
FOR+FWI+INUN+FOR*FWI -1492.98
FOR+FWI+INUN+FOR*INUN -1493.15
FOR+FWI+INUN+FWI*INUN -1484.30
a,"-, indicates a w, value < 0.01%; b "-,, indicates


K AIC AAIC
5 2994.76 16.15
4 2999.88 21.28
4 2994.42 15.81
4 2998.42 19.81
4 3451.00 472.40
3 2999.51 20.90
3 2998.81 20.21
3 3497.14 518.54
5 2995.96 17.36
5 2996.29 17.69
5 2978.60 0.00
an evidence ratio > 100


w1a Evidence Ratiosb


0.04%


0.02%
0.01%
99.89%


1.00


Table 3-4. Logistic regression models tested for Great Egrets and associated statistical values. Italicized models are the best from their
subset, while the best overall models are shown in boldface.


Model Log-Likelihood
FOR+WV+FWI+INUN -498.86
FOR+WV+FWI -499.97
FOR+FWI+INUN -499.59
FOR+ WV+ INUN -498.86
WV+FWI+INUN -502.60
FOR+WV -500.48
FOR +INUN -499.61
WV+INUN -502.69
FOR -501.08
INUN -503.99
FOR+INUN+FOR*INUN -497.75
FOR+WV+INUN+FOR*INUN -496.98
b "-" indicates an evidence ratio > 100


AIC
1007.73
1007.94
1007.17
1005.73
1013.20
1006.96
1005.21
1011.38
1006.16
1011.99
1003.51
1003.97


AAIC
4.22
4.44
3.67
2.22
9.69
3.46
1.71
7.88
2.65
8.48
0.00
0.46


w,
3.54%
3.17%
4.67%
9.62%0
0.23%
5.18%
12.45%o
0.57%
7.76%
0.42%
29.20%
23.18%


Evidence Ratiosa
8.25
9.20
6.26
3.03

5.63
2.35
51.35
3.76
69.43
1.00
1.26


Rank
8
9
7
4
12
6
3
10
5
11
1
2


Rank
3
9
2
6
10












Table 3-5. Logistic regression models tested for Tricolored Herons and associated statistical values. Italicized models are the best from
their subset, while the best overall model is shown in boldface.


Model
FOR+WV+FWI+INUN
FOR+WV+FWI
FOR+FWI+INUN
FOR WV+ INUN
WV+FWI+INUN
FOR+WV
FOR +INUN
WV+INUN
FOR
INUNUN+FOR*INUN
FOR+INUN+FOR*INUN


Log-Likelihood
-363.45
-364.24
-363.83
-363.50
-387.66
-364.70
-363.92
-387.67
-365.35
-390.82
-357.38


AIC
736.90
736.48
735.66
734.99
783.33
735.40
733.84
781.33
734.70
785.64
722.76


A AIC
14.14
13.72
12.90
12.23
60.57
12.64
11.07
58.57
11.94
62.88
0.00


wa
0.08%
0.10%
0.16%
0.22%

0.18%
0.39%

0.25%

98.62%


Evidence Ratiosb


Rank
8
7


1.00


a" indicates a w, value < 0.01%; o "-" indicates an evidence ratio > 100









Table 3-6. Best models for each species and associated measures of classification
performance.
Great Blue Herons


Model
n, % Correct (Used)
n, % Correct (Unused)
K Statistic
Area Under ROC Curve
Great Egrets
Model 1
n, % Correct (Used)
n, % Correct (Unused)
K Statistic
Area Under ROC Curve
Model 2
n, % Correct (Used)
n, % Correct (Unused)
K Statistic
Area Under ROC Curve
Tricolored Herons
Model
n, % Correct (Used)
n, % Correct (Unused)
K Statistic
Area Under ROC Curve


FOR + FWI + INUN FWI*INUN
1005, 78.9%
722, 56.7%
0.357
0.745

FOR + INUN + FOR*INUN
251, 68.8%
134, 36.7%
0.055
0.524
FOR WV + INUN + FOR*INUN
274, 75.1%
152, 41.6%
0.167
0.559

FOR + INUN + FOR*INUN
218, 77.3%
137, 48.6%
0.259
0.654


Table 3-7. Relative importance of variables to each species.


FOR
WV
FWI
INUN
FOR*FWI
FOR*INUN
FWI*INUN


Great Blue Heron
100.00%
0.04%
99.99%
99.99%
0.02%
0.01%
99.89%


Great Egret
98.78%
45.50%
11.61%
83.88%

52.38%


Tricolored Heron
100.00%
0.58%
0.34%
99.47%

98.62%












A B











\


4v
v T v



v D
vv IT T 0
V v T v *




v 7 v v V


















V / **,* *
v v *
T *V



19*. If B u
C D
\ V\









Fig b o a o. B 1 C 1 D








Heron... c ic G Eto, ad c i






T rivcoo Ire Heo coones
| N 5 10 1 |






























*














v*
v v v


T v

v V


Figure 3-1. Continued


F


*\

\




















0 *
S v v
* *











V V i


*





*



*


v
V *
* .


*


10 15














0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1


0 5000


0.6
0.4
0.2
0
-0.2
-0.4

-0.6
-0.8
-1
-1.2


5000


Figure 3-2. Relationship of Great Blue Heron colonies relative to all available sites in
WCA3 as shown by linearized Ripley's K (l[d]) graphs. A) 1993. B) 1994. C)
1995. D) 1996. E) 1997. F) 1998. G) 1999. H) 2000. Sample sizes for each
year listed in Table 3-1. Dashed lines: 95% confidence intervals from Monte
Carlo simulations of 999 identically-sized, randomly-selected samples from
each year; solid line: 1(d) graph for used sites.


10000


15000


Scale (m)


20000


-
p
I
I


10000


15000


Scale (m)


20000












0.4
0.2
0
-0.2
-0.4
-0.6
-0.8


5000


0.6
0.4
0.2
0
S-0.2
-0.4
-0.6
-0.8
-1
-1.2


5000


10000
Scale (m)


10000
Scale (m)


Figure 3-2. Continued


-- .~ ~. -

-




2>


15000


20000


-- ------


15000


20000















1

0.5

0

-0.5

-1

-1.5

-2


5000


0.8
0.6
0.4
0.2
0
-0.2
-- -0.4
-0.6
-0.8
-1
-1.2
-1.4


5000


10000

Scale (m)


10000

Scale (m)


Figure 3-2. Continued


-

-
I
I
A
I


15000


20000


-


15000


20000














0.4

0.2

0

-0.2

-0.4

-0.6

-0.8


5000


10000


15000


Scale (m)



H


5000


10000


15000


Scale (m)


Figure 3-2. Continued


-
-

-
I


7


20000


0.4
0.3
0.2
0.1
0
-0.1
- -0.2
-0.3
-0.4
-0.5
-0.6


-
-

-
-
I


20000














2
1.5
1
0.5
S 0
-0.5
-1
-1.5
-2


0 5000 10000 15000


Scale (m)



B


2
1.5
1
0.5
S0
-0.5
-1
-1 -
-1.5
-2


5000


10000


15000


Scale (m)


Figure 3-3. Relationship of Great Egret colonies relative to all available sites in WCA3 as
shown by linearized Ripley's K (l[d]) graphs. A) 1993. B) 1994. C) 1995. D)
1996. E) 1997. F) 1998. G) 1999. H) 2000. Sample sizes for each year listed
in Table 3-1. Dashed lines: 95% confidence intervals from Monte Carlo
simulations of 999 identically-sized, randomly-selected samples from each
year; solid line: 1(d) graph for used sites.


20000


I -
' I
I t
'I


20000














1.5

1

0.5

0

-0.5

-1
-1 -

-1.5


5000


10000


15000


Scale (m)



D


5000


10000


15000


Scale (m)


Figure 3-3. Continued


I # -~
' /


20000


1.5

1-

0.5

0

-0.5

-1 -

-1.5


I
1




S- i

i J
\ t
iO


20000














1.5

1

0.5

0

-0.5

-1

-1.5
-2


5000 10000 15000


Scale (m)



F


0 5000 10000 15000


Scale (m)


Figure 3-3. Continued


- - - B-


20000


20000













1

0.5

0

-0.5

-1

-1.5









1

0.5

0

-0.5

-1

-1.5


0 5000 10000 15000

Scale (m)


0 5000 10000 15000

Scale (m)


Figure 3-3. Continued


20000


20000














1.5

1

0.5
S0

-0.5
-1

-1.5
-2


5000


10000


Scale (m)



B


10000


Scale (m)


Figure 3-4. Relationship of Tricolored Heron colonies relative to all available sites in
WCA3 as shown by linearized Ripley's K (l[d]) graphs. A) 1993. B) 1994. C)
1995. D) 1996. E) 1997. F) 1998. G) 1999. H) 2000. Sample sizes for each
year listed in Table 3-1. Dashed lines: 95% confidence intervals from Monte
Carlo simulations of 999 identically-sized, randomly-selected samples from
each year; solid line: 1(d) graph for used sites.


- -d
% s,
i


15000


20000


1.5

1

0.5

0

-0.5

-1 -

-1.5


.



1 -


*


5000


15000


20000















2.5

1.5

o 0.5
*0
- -0.5

-1.5

-2.5


5000


1.5
1

0.5
- 0-
0

S-0.5

-1

-1.5
-2


10000

Scale (m)



D


5000


10000

Scale (m)


Figure 3-4. Continued


1

i- -- %_.- ..-.---------------..........
J m m

-
I


15000


20000


t
-


J.. .- ... _me i m e eJ


! !

.


15000


20000















4.5

3.5

2.5

. 1.5

^ 0.5

-0.5

-1.5

-2.5


5000


10000


15000


Scale (m)



F


5000


10000


15000


Scale (m)


Figure 3-4. Continued


t
- <.

V --


-t cf"^^
-* -

---.-
~t *. me- e mm--


20000


1~ ~


2.5
2
1.5
1
S0.5
g 0-

-0.5
-1
-1 -
-1.5
-2


20000














1.5

1

0.5

S0

-0.5

-1
-1 -

-1.5


5000


1.5

1

0.5

S0

-0.5
-1
-1 -

-1.5


5000


10000

Scale (m)


10000

Scale (m)


Figure 3-4. Continued


- -


15000


20000


-
-



'5
V


15000


20000














CHAPTER 4
DISCUSSION

Effects of Site Availability

The spatial patterns of colony sites used by these three species during the study

period were in general dissimilar from the overall pattern of available sites (Figures 3-2

through 3-4). This was always true for Great Blue Herons, and was true at most scales for

Great Egrets and Tricolored Herons in 4 and 5 of the 8 years, respectively. These two

species exhibited both clustered and dispersed patterns, which likely reflects responses by

the birds to different hydrological conditions in each year. These results indicate that site

selection by these three species was probably independent of overall site availability or

distribution.

Model Performance

Although there was very strong evidence supporting the best models for Great Blue

and Tricolored Herons, relative to the total set of models I tested, model classification

performance was poor overall. None of the best models achieved a Kappa statistic of 0.4,

which is often considered to be the minimum value for "good" performance (Fielding &

Bell 1997), and only the best model for Great Blue Herons had an area under its ROC

curve of at least 0.7. Models always did a better job of classifying used than unused cells

(Table 3-6), which suggests that there were more appropriate sites available than were

used (Fielding & Bell 1997). A major limitation of this analysis was the coarse spatial

resolution of the hydrological data, which prevented me from including variables that

operate at the scale of the sites themselves such as vegetation type or patch size. Finally,









simply classifying cells as used or unused ignored the enormous variation in the number

of Great Egret pairs nesting at sites within WCA3, which is probably why the models for

this species performed the worst.

Despite these limitations, these results clearly show that the amount of foraging

habitat around sites, and the likelihood that they would remain inundated throughout the

breeding season, were important influences on the colony site selection for these three

species. This is the first study to quantitatively establish the importance of deterring

mammalian predators as an influence on colony site selection in this group. These results

also establish that the distribution of foraging habitats is as important to wading birds in

an enormous wetland-dominated landscape such as the Everglades, as it is in more typical

terrestrial landscapes where wetlands are part of a habitat mosaic (Gibbs et al. 1987;

Gibbs 1991; Gibbs & Kinkel 1997).

Responses to Hydrological Variability

In general, neither the average weekly proportional decline in water depths (FWI)

nor the spatial variation in water depths (WV) were important influences on colony site

selection in these three species. WV was weakly selected against by Great Egrets, and not

included in the best models for either of the other two species. Only Great Blue Herons

selected cells with large values of FWI. However, areas that began the breeding season

with deeper water were likely to have the largest values of for this variable, since they

could experience proportional declines for much longer before going dry. Thus it is

unclear whether FWI actually represented prey availability for Great Blue Herons or was

instead a proxy for deep water. Great Blues also selected against cells that had both rapid

declines in water depths and a high likelihood of remaining inundated (i.e., the

INUN*FWI interaction term). These could represent either permanently shallow areas









(such as on the west side of WCA3, where water flow westward into Big Cypress

National Preserve is unimpeded), and/or areas near canals, both of which were mostly

avoided by this species (Figure 3-1).

These results suggest that ardeid wading birds are selecting colony sites that are

characterized by stable, rather than variable, water depths. Another possible interpretation

is that colony site selection by these species does not reflect an attempt to predict future

conditions (i.e., a "bet hedging" strategy), but is rather simply a response to conditions at

the time of nesting (Bancroft et al. 1994). These three species are known to be less

dependent on concentrated patches of prey created by declining water depths than

tactile/social foragers such as Wood Storks (Mycteria americana), White Ibises

(Eudocimus albus), and Snowy Egrets (Egretta thula) (Kahl 1964; Gawlik 2002), so it is

also possible that they have not experienced any selective pressure to recognize or

respond to this environmental characteristic.

Management Implications

Previous authors have speculated that the size of wading bird breeding populations

in the Everglades is dependent on the extent of open slough habitat available to them

(Bancroft et al. 1994; Bancroft et al. 2002), and the results of this study establish that this

is the most important environmental characteristic determining where the birds nest.

Hydroperiod and fire are the two largest factors controlling the distribution of wetland

vegetation (Gunderson 1994), so water management can play an important role in

maximizing the extent of open slough habitat within the Everglades. Water depths should

be maintained at a deep enough level to prevent most breeding sites from going dry, but

shallow enough for wading birds to forage. Using these guidelines could make a major






41


contribution to re-establishing historical breeding populations of wading birds in the

Everglades.















LITERATURE CITED


Akaike, H. 1973. Information theory as an extension of the maximum likelihood
principle. Pages 267-281 in B. N. Petrov and F. Csaki, editors. Second International
Symposium on Information Theory. Akademiai Kiado, Budapest.

Bancroft, G. T. 1989. Status and conservation of wading birds in the Everglades.
American Birds 43:1258-1265.

Bancroft, G. T., D. E. Gawlik, and K. Rutchey. 2002. Distribution of wading birds
relative to vegetation and water depths in the northern Everglades of Florida, USA.
Waterbirds 25:263-277.

Bancroft, G. T., J. C. Ogden, and B. W. Patty. 1988. Wading bird colony formation and
turnover relative to rainfall in the Corkscrew Swamp area of Florida during 1982
through 1985. Wilson Bulletin 100:50-59.

Bancroft, G. T., A. M. Strong, R. J. Sawicki, W. Hoffman, and S. D. Jewell. 1994.
Relationships among wading bird foraging patterns, colony locations, and
hydrology in the Everglades. Pages 615-657 in S. Davis and J. Ogden, editors.
Everglades: the ecosystem and its restoration. St. Lucie Press, Delray Beach,
Florida, USA.

Baxter, G. S., and P. G. Fairweather. 1998. Does available foraging area, location or
colony character control the size of multispecies egret colonies? Wildlife Research
25:23-32.

Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference: a
practical information-theoretic approach. Second edition. Springer-Verlag, New
York, New York, USA.

Butler, R. W. 1992. Great Blue Heron (Ardeas herodias). No. 25 in A. Poole and F. Gill,
editors. The Birds of North America. The Academy of Natural Sciences,
Philadelphia, Pennsylvania, USA; The American Ornithologists' Union,
Washington, D.C., USA.

Custer, T. W., and R. G. Osborn. 1978. Feeding-site description of three heron species
near Beaufort, North Carolina. Pages 355-360 in A. Sprunt IV, J. C. Ogden, and S.
Winckler, editors. Wading Birds, Research Report No. 7. National Audubon
Society, New York, New York, USA.









Environmental Systems Research Institute, 2002. Arc/Info Workstation. Version 8.3.
Redlands, California, USA.

Fielding, A. H., and J. F. Bell. 1997. A review of methods for the assessment of
prediction errors in conservation presence/absence models. Environmental
Conservation 24:38-49.

Fitz, H. C., J. Godin, B. Trimble, and N. Wang. 2004. Everglades Landscape Model
documentation: ELM v2.2. South Florida Water Management District, West Palm
Beach, Florida, USA.

Fotheringham, A. S., C. Brunsdon, and M. Charlton. 2000. Quantitative geography:
perspectives on spatial data analysis. Sage Publications, London, UK.

Frederick, P. C. 1997. Tricolored heron (Egretta tricolor). No. 306 in A. Poole and F.
Gill, editors. The Birds of North America. The Academy of Natural Sciences,
Philadelphia, Pennsylvania, USA; The American Ornithologists' Union,
Washington, D.C., USA.

Frederick, P. C. 2002. Wading birds in the marine environment. Pages 617-655 in E. A.
Schreiber and J. Burger, editors. Biology of marine birds. CRC Press, Boca Raton,
Florida, USA.

Frederick, P. C., and M. W. Collopy. 1989a. Nesting success of five ciconiiform species
in relation to water conditions in the Florida Everglades. Auk 106:625-634.

Frederick, P. C., and M. W. Collopy. 1989b. The role of predation in determining
reproductive success of colonially nesting wading birds in the Florida Everglades.
Condor 91:860-867.

Frederick, P. C., and J. C. Ogden. 2001. Pulsed breeding of long-legged wading birds and
the importance of infrequent severe drought conditions in the Florida Everglades.
Wetlands 21:484-491.

Frederick, P.C., and M. G. Spalding. 1994. Factors affecting reproductive success of
wading birds (Ciconiiformes) in the Everglades ecosystem. Pages 659-691 in S.
Davis and J. Ogden, editors. Everglades: the ecosystem and its restoration. St.
Lucie Press, Delray Beach, Florida, USA.

Frederick, P.C., T. Towles, R. J. Sawicki, and G. T. Bancroft. 1996. Comparison of aerial
and ground techniques for discovery and census of wading bird (Ciconiiformes)
nesting colonies. Condor 98: 837-840.

Gawlik, D. E. 2002. The effects of prey availability on the numerical response of wading
birds. Ecological Monographs 72:329-346.

Gibbs, J. P. 1991. Spatial relationships between nesting colonies and foraging areas of
Great Blue Herons. Auk 108:764-770.









Gibbs, J. P., and L. K. Kinkel. 1997. Determinants of the size and location of Great Blue
Heron colonies. Colonial Waterbirds 20:1-7.

Gibbs, J. P., S. Woodward, M. L. Hunter, and A. E. Hutchinson. 1987. Determinants of
Great Blue Heron colony distribution in coastal Maine. Auk 104:38-47.

Gunderson, L. H. 1994. Vegetation of the Everglades: determinants of community
composition. Pages 323-340 in S. Davis and J. Ogden, editors. Everglades: the
ecosystem and its restoration. St. Lucie Press, Delray Beach, Florida, USA.

Hintze, J. 2001. NCSS and PASS. Number Cruncher Statistical Systems. Kaysville, Utah,
USA.

Hoffman, W., G. T. Bancroft, and R. J. Sawicki. 1994. Foraging habitat of wading birds
in the Water Conservation Areas of the Everglades. Pages 585-614 in S. Davis and
J. Ogden, editors. Everglades: the ecosystem and its restoration. St. Lucie Press,
Delray Beach, Florida, USA.

Kahl, M. P., Jr. 1964. Food ecology of the Wood Stork (Mycteria americana) in Florida.
Ecological Monographs 34:97-117.

Kushlan, J. A. 1976. Wading bird predation in a seasonally fluctuating pond. Auk
93:464-476.

Kushlan, J. A. 1987. External threats and internal management: the hydrologic regulation
of the Everglades, Florida, USA. Environmental Management 11:109-119.

McCrimmon, D. A., Jr., J. C. Ogden, and G. T. Bancroft. 2001. Great Egret (Ardea alba).
No. 570 in A. Poole and F. Gill, editors. The Birds of North America. The
Academy of Natural Sciences, Philadelphia, Pennsylvania, USA; The American
Ornithologists' Union, Washington, D.C., USA.

Ogden, J. C. 1994. A comparison of wading bird nesting colony dynamics (1931-1946
and 1974-1989) as an indication of ecosystem conditions in the southern
Everglades. Pages 533-570 in S. Davis and J. Ogden, editors. Everglades: the
ecosystem and its restoration. St. Lucie Press, Delray Beach, Florida, USA.

Post, W. 1990. Nest survival in a large ibis-heron colony during a three-year decline to
extinction. Colonial Waterbirds 13:50-61.

Powell, G. V. N. 1987. Habitat use by wading birds in a subtropical estuary: implications
of hydrography. Auk 104:740-749.

Rodgers, J. A. 1987. On the antipredator advantages of coloniality: a word of caution.
Wilson Bulletin 99:269-270.

Rutchey, K., L. Vilchek, and M. Love, in review. Development of a vegetation map for
Water Conservation Area 3.









Smith, J. P. 1995. Foraging flights and habitat use of nesting wading birds
(Ciconiiformes) at Lake Okeechobee, Florida. Colonial Waterbirds 18:139-158.

Smith, J. P., and Collopy, M. W. 1995. Colony turnover, nest success and productivity,
and causes of nest failure among wading birds (Ciconiiformes) at Lake
Okeechobee, Florida (1989-1992). Pages 287-316 in N. G. Aumen and R. G.
Wetzel, editors. Ecological studies on the littoral and pelagic systems of Lake
Okeechobee, Florida (USA). E. Schweizerbart'sche Verlagsbuchhandlung,
Stuttgart, Germany.

Smith, J. P., J. R. Richardson, and M. W. Collopy. 1995. Foraging habitat selection
among wading birds (Ciconiiformes) at Lake Okeechobee, Florida, in relation to
hydrology and vegetative cover. Pages 247-285 in N. G. Aumen and R. G. Wetzel,
editors. Ecological studies on the littoral and pelagic systems of Lake Okeechobee,
Florida (USA). E. Schweizerbart'sche Verlagsbuchhandlung, Stuttgart, Germany.

Surdick, J. A., Jr. 1998. Biotic and abiotic indicators of foraging site selection and
foraging success of four ciconiiform species in the freshwater Everglades of
Florida. Thesis. University of Florida, Gainesville, Florida, USA.

Taylor, R. J., and E. D. Michael. 1971. Predation on an inland heronry in eastern Texas.
Wilson Bulletin 83:172-177.















BIOGRAPHICAL SKETCH

Matthew John Bokach completed his Bachelor of Science degree with a double

major in biology and chemistry at Adrian College (Adrian, Michigan) in 1994. He fled

the sciences for 2 years in a master's program in student affairs administration at

Michigan State University, then returned to them as a U.S. Peace Corps Volunteer at

Lukosi Government School in Hwange District, Zimbabwe. While teaching general

science at Lukosi, Matthew fell in love with the southern African avifauna and spent an

additional 2 years in Zimbabwe, mostly so he could continue watching them. He returned

to the United States in 2000 and began a master's degree in interdisciplinary ecology at

the University of Florida in 2002. He plans to move to the San Francisco Bay area,

pursue a career in GIS application development, and eventually flee the sciences once

again to pursue a career in popular music.




Full Text

PAGE 1

THE RELATIVE INFLUENCES OF PR EDATION AND PREY AVAILABILITY ON ARDEID BREEDING COLONY SITE SELECTION By MATTHEW JOHN BOKACH A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Matthew John Bokach

PAGE 3

This thesis is dedicated to my parents Mark and Cathy, who have weathered the many twists and turns of my lifes last decade with unconditional love and support, leavened with just a touch of legitimate bemusement.

PAGE 4

ACKNOWLEDGMENTS I could not have conducted this research or even made it through graduate school without the support and assistance of a humbling number of other people. Dr. Peter Frederick has been the best advisor I could have asked for. Without his support, patience, and humor, I would not have finished this degree. My data on the wading bird colony sites was collected by a long series of his former students and technicians, and this research is therefore as much a product of their work as mine. Dr. Graeme Cumming and Dr. Jane Southworth, my other two committee members, gave me invaluable advice through nearly every stage of the research. My initial explorations into the daunting task of analyzing these data benefited immensely from Dr. Tim Fiks statistical insights and encouragement. I gratefully acknowledge Dr. Carl Fitz, Mr. Jason Godin, and Mr. Ken Rutchey of the South Florida Water Management District; and, Mr. Troy Mullins of the National Park Service, for providing me with the hydrology and vegetation data used herein. Lorraine Heisler, Les Vilchek, and Shawn Komlos all provided insights concerning the development and appropriate uses of these data. Thanks to Theresa Burcsu and Desiree Price, I was allowed constant access to the computer labs of the University of Florida Geography Department, even when not enrolled in one of their classes; I will always think of Turlington Hall as the home of this research. Lin Cassidy, Andres Guhl, and especially Matt Marsik all gave me valuable assistance with GIS questions. I blame Carl Evans for both complicating and vastly simplifying my life by introducing me to Matlab. The patience and support of my boyfriend Rodney Brown iv

PAGE 5

was a steadying influence through most of this long process. Finally, I thank my family for all of their love and support through the years. I was supported financially for nearly my entire time at the University of Florida by a research assistantship through the College (now School) of Natural Resources and Environment. I extend my most sincere thanks to Dr. Steve Humphrey, Cathy Ritchie, and Meisha Wade for giving me the opportunity to pursue this degree, and for extending my funding for an extra semester. v

PAGE 6

TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES............................................................................................................vii LIST OF FIGURES.........................................................................................................viii ABSTRACT.......................................................................................................................ix CHAPTER 1 INTRODUCTION........................................................................................................1 2 METHODS...................................................................................................................5 Study Area....................................................................................................................5 Data Sources.................................................................................................................6 Water Depth Grids.................................................................................................6 Vegetation Maps....................................................................................................6 Wading Bird Colony Locations.............................................................................8 Distributional Relationships.........................................................................................9 Calculation of Variables...............................................................................................9 Multivariate Analyses.................................................................................................11 3 RESULTS...................................................................................................................17 Distribution of Colony Sites.......................................................................................17 Bootstrap of Foraging Habitat Calculation.................................................................17 Multivariate Analyses.................................................................................................18 4 DISCUSSION.............................................................................................................38 Effects of Site Availability.........................................................................................38 Model Performance....................................................................................................38 Responses to Hydrological Variability.......................................................................39 Management Implications..........................................................................................40 LITERATURE CITED......................................................................................................42 BIOGRAPHICAL SKETCH.............................................................................................46 vi

PAGE 7

LIST OF TABLES Table page 1-1 Variables and methods of calculation........................................................................4 1-2 Weeks of nest initiation and duration of breeding period for three focal species......4 3-1 Number of colonies inhabited by each species, 1993-2000.....................................19 3-2 Values of FOR calculated in 30 randomly-chosen cells using constant proportions, compared to approximate 95% confidence intervals from a bootstrap analysis involving 1000 iterations where proportions were allowed to randomly vary..........................................................................................................20 3-3 Logistic regression models tested for Great Blue Herons and associated statistical values........................................................................................................21 3-4 Logistic regression models tested for Great Egrets and associated statistical values........................................................................................................................21 3-5 Logistic regression models tested for Tricolored Herons and associated statistical values........................................................................................................22 3-6 Best models for each species and associated measures of classification performance..............................................................................................................23 3-7 Relative importance of variables to each species.....................................................23 vii

PAGE 8

LIST OF FIGURES Figure page 2-1 Southeastern Florida showing the location of WCA3 within the larger Everglades ecosystem..............................................................................................14 2-2 Everglades Landscape Model cells in WCA3..........................................................15 2-3 Algorithm for determining which models to test for each species...........................16 3-1 Distribution of ardeid colonies in WCA3................................................................24 3-2 Relationship of Great Blue Heron colonies relative to all available sites in WCA3 as shown by linearized Ripleys K (l[d]) graphs.........................................26 3-3 Relationship of Great Egret colonies relative to all available sites in WCA3 as shown by linearized Ripleys K (l[d]) graphs..........................................................30 3-4 Relationship of Tricolored Heron colonies relative to all available sites in WCA3 as shown by linearized Ripleys K (l[d]) graphs.........................................34 viii

PAGE 9

Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science THE RELATIVE INFLUENCES OF PREDATION AND PREY AVAILABILITY ON ARDEID BREEDING COLONY SITE SELECTION By Matthew John Bokach May 2005 Chair: Peter Frederick Major Department: Natural Resources and Environment Nest predation and prey availability are two of the most important factors affecting breeding success in long-legged wading birds (Ciconiiformes). I investigated whether the breeding colony site selection of three species of herons and egrets (family Ardeidae) in the Florida Everglades was influenced by environmental characteristics that mediated these two factors. These characteristics were 1) likelihood that sites remained inundated throughout the breeding period; 2) amount of foraging habitat around sites; 3) average weekly proportional decline in water depths around sites throughout the breeding period; 4) spatial variation in water depths around sites at the time of nesting. Variables were calculated within a geographic information system using both raster and vector inputs. I used measures derived from the Akaike information criterion to select the best logistic regression model and to evaluate the relative importance of these four variables for colony site selection in each species. ix

PAGE 10

Amount of foraging habitat and likelihood of remaining inundated were the most important variables influencing colony site selection by all three species. Great Blue Herons (Ardea herodius) also selected sites with high rates of average weekly proportional declines in water depth, but it is likely this variable was a proxy for deep water for this species rather than a reflection of prey availability. Overall, these species seemed to favor stable (rather than variable) hydrological conditions. This might also indicate that their colony site selection is based on conditions at the time of nesting rather than an attempt to predict future conditions. These results confirm the importance of managing the Everglades to maximize the extent of slough habitats, if the goal is to increase breeding populations of wading birds therein. x

PAGE 11

CHAPTER 1 INTRODUCTION Two of the most important factors affecting the breeding success of long-legged wading birds (Ciconiiformes) are egg and chick predation, and the availability of adequate food to raise chicks to fledging (Taylor & Michael 1971; Frederick & Collopy 1989a; Frederick & Spalding 1994; Frederick 2002). Although wading birds cannot control the behaviors of predators or their prey, it is likely that selection has favored the recognition of breeding colony sites with beneficial environmental characteristics that mediate or constrain those behaviors. For example, wading birds nearly always nest on islands or in trees and/or shrubs that are inundated at their base, a pattern that has most often been interpreted as a strategy for deterring mammalian predators (Rodgers 1987; Bancroft et al. 1988; Frederick & Collopy 1989b; Smith & Collopy 1995). However, whether this pattern represents actual selection for this characteristic has never been quantitatively studied. With regard to prey availability, although previous studies (Gibbs et al. 1987; Gibbs 1991; Gibbs & Kinkel 1997; Baxter & Fairweather 1998; Bancroft et al. 2002) have shown that wading birds select colony sites that maximize the amount of wetland habitats within a reasonable foraging range, other environmental characteristics (e.g., water depth) are better determinants of wading birds ability to capture prey. If wading bird colony site selection represents an attempt to deter predators and/or maximize prey availability, then used sites should be measurably different from available unused sites with relation to environmental characteristics that affect these two factors. I 1

PAGE 12

2 identified one characteristic that deters predators from reaching nests and three characteristics that are likely to influence prey availability around sites (Table 1-1) and make the following predictions. Prediction 1: Used sites will have a greater likelihood of remaining inundated throughout the breeding period than unused sites. This characteristic seems to be a good measure of a sites ability to deter mammalian predators (Frederick & Collopy 1989b; Smith & Collopy 1995), which are generally the most destructive in their effects on wading bird colonies (Rodgers 1987; Post 1990; Smith & Collopy 1995). Prediction 2: Used sites will be surrounded by more open or sparsely vegetated habitats than unused sites. Not all wetland habitats are appropriate foraging grounds for wading birds. In particular, they are known to avoid dense vegetation (such as monospecific stands of sawgrass [Cladium jamaicense] or cattail [Typha latifolia]), which interferes with their visual or tactile hunting techniques, provides prey animals with more hiding places, and could serve as a hiding place for predators of the birds themselves (Hoffman et al. 1994; Smith et al. 1995; Surdick 1998). Prediction 3: Used sites will be located in areas that experience greater proportional declines in water depth throughout the breeding season than unused sites. Wading birds are generally limited to foraging in water that is shallower than the length of their bills or legs (Custer & Osborn 1978; Powell 1987). They are often attracted to areas of declining water depth where prey have been concentrated into pools or depressions that remain inundated late in the season (Kushlan 1976; Bancroft 1989; Frederick & Collopy 1989a; Smith 1995; Gawlik 2002; but see Frederick & Spalding 1994 for a critique of this idea).

PAGE 13

3 Prediction 4: Variability in water depths at the time of nesting will be greater around used sites than around unused sites. Water levels are dynamic and unpredictable in most wetlands. Heterogeneous topography around a colony site could therefore provide a foraging advantage to wading birds as this would provide the most diversity of potential foraging sites at almost any time or water condition (Kahl 1964; Bancroft et al. 2002). I used logistic regression models to assess the relative influence of these four variables on the colony site selection of three wading bird species in the ciconiiform family Ardeidae (Table 1-2) over a period of 8 years in the Florida Everglades. Measures derived from the models Akaike information criteria (AIC) (Akaike 1973; Burnham & Anderson 2002) were used to choose the best model and determine the relative influence of the variables on the colony site selection for each species.

PAGE 14

4 Table 1-1. Variables and methods of calculation. Variable Abbreviation Method of calculation Affecting predator deterrence Likelihood that a cell will remain inundated for the duration of the breeding period INUN Used GREATERTHAN function to output number of weeks over duration of a species' breeding period that depths remained above 0; divided by number of weeks to yield values between 0.0 and 1.0 Affecting prey availability Amount of foraging vegetation around sites FOR See text for details Average rate of proportional decline in depths around a site for the duration of a species' breeding period FWI Calculated proportional decline in water depths for every week during a species' breeding period (rises in depth expressed as 0); averaged these values over all weeks; used the FOCALMEAN function to average these cell averages over 3x3 cell neighborhoods, yielding values between 0.0 and 0.5 Spatial variation in water depths around a site at week of nest initiation WV Used FOCALSTD func tion to calculate standard deviation of depths in 3x3 cell neighborhoods All calculations carried out in ESRI Arc/INFO workstation v.8.3 except FOR, which was calculated in ESRI ArcMap v.8.3. Weeks of nest initiation and duration of breeding periods are listed in Table 1-2. Table 1-2. Weeks of nest init iation and duration of breeding period for three focal species Species Week of nest initiationa Duration Source for duration Great Blue Heron ( Ardea herodius ) 9th 14 weeks Butler 1992 Great Egret ( Ardea alba ) 9th 10 weeks McCrimmon et al. 2001 Tricolored Heron ( Egretta tricolor ) 12th 11 weeks Frederick 1997 aSource is Frederick (pers. comm.)

PAGE 15

CHAPTER 2 METHODS Study Area I studied ardeid colony site selection within Water Conservation Area (WCA) 3, a 2,350-km2 human-made impoundment in the central Everglades (Figure 2-1). Most of the wading birds that breed within the Everglades ecosystem have selected colony sites within this impoundment since the early 1970s (Ogden 1994; Frederick & Ogden 2001). The annual hydrology of WCA3 is characterized by a decline in water depths during the dry season between November and April, and most annual rainfall occurring during a subtropical wet season between June and November. The northern end of the impoundment is shallow and quick to dry, while the southern end is almost permanently inundated. This same gradient exists to a lesser extent from west (where flow of water into the adjoining Big Cypress National Preserve is unimpeded) to east (the enclosed sub-impoundment WCA3B created by the L-67 canals, Figure 2-1). Wading birds therefore have a wide range of hydrological conditions within which to select sites in WCA3. The vegetation of WCA3 is characterized by open wet prairie communities in deeper areas interspersed between narrow ridges covered in cattail and/or sawgrass (Gunderson 1994). The patches of woody or shrubby vegetation that serve as potential nesting sites for wading birds are scattered throughout. Large areas in the northern end of the impoundment are dominated by an almost complete monoculture of cattails. 5

PAGE 16

6 Data Sources Water Depth Grids I derived hydrological variables using estimated water depths from the Everglades Landscape Model (ELM; Fitz et al. 2004), which simulates the flows and stages of water across the entire Everglades ecosystem at a resolution of 1 km2. Model output was available for the period 1993, so these 8 years comprised the duration of my study. Each output layer is an Arc/Info grid (Environmental Systems Research Institute 2002) whose values are the estimated weekly average above-ground depth of water in each cell. The model generally performs well both spatially and temporally when compared to actual depths measured throughout the landscape (Fitz et al. 2004). The largest errors occur in cells containing canals, as these contain the most intracellular variation in water depths. ELM 1 km2 cells, rather than discrete colony sites, were my units of analysis as these were the coarsest in resolution of all the data. Vegetation Maps I used a polygon vector geographic information system (GIS) layer showing the distribution of vegetation in the central and southern Everglades and based on aerial infrared photography acquired in 1995 over the entire area of WCA3, Everglades National Park, and Big Cypress National Preserve. The latter two areas were included in my analysis since they are within the typical foraging range of birds nesting near the boundaries of WCA3 (Bancroft et al. 1994). The distinctive reflectance signatures of different types of vegetation within the aerial photography were used to digitize polygons representing discrete patches. These could contain up to three specific vegetative classes, assigned hierarchically. After comparing the polygon layer to 1999 United States Geological Survey (USGS) digital orthophoto quarter quads (DOQQs), I was satisfied

PAGE 17

7 that this layer was an accurate depiction of the vegetation in WCA3 for the entire time period of this study. Since all reports of nesting by ardeid wading birds are on trees or shrubs in WCA3, I defined potential colony sites as polygons that contained at least one class of woody vegetation. Among the classifications used in the vegetation map these included: all variants of forest, all variants of shrublands, Melaleuca, Brazilian pepper, disturbed fish camp sites, spoil areas, and artificial deer islands. I aggregated contiguous polygons that contained at least one of these vegetation types into single discrete patches, which I hereafter refer to interchangeably as patches of woody vegetation or potential colony sites. The locations of wading bird colonies (see below) revealed that a small proportion (~1.1%) of the potential colony sites did not occur within a reasonable distance from a patch of woody vegetation. In October 2002 I visited all such sites in WCA3B and in WCA3A north of Highway I-75 (Figure 2-1), and determined that in most cases either some woody vegetation had not been included in an existing polygon, or that a single tree or small patch of shrubs was present and should have itself been digitized as a polygon. In these cases, I therefore either added a woody vegetation classification to the second or third tier of the existing polygon, or digitized the patch myself using the 1999 USGS DOQQs. Water model cells that either contained no potential colony sites, or contained only the clipped edges of sites whose bulk were in an adjacent cell, were excluded from further analysis since wading birds could not have nested in them. At the scale at which I

PAGE 18

8 calculated the variables (see below), this removed less than 3% of the total area of WCA3 from analysis (Figure 2-2). Wading Bird Colony Locations Locations, species composition, and size of wading bird colonies were determined using both aerial and ground survey techniques (Frederick et al. 1996). Aerial surveys were designed such that the entire ground surface of the study area was visible along at least one east-west transect. These were spaced 2.9 km apart, flown at 245 m altitude, and were conducted monthly between January and June of each year. Dark-colored species are often not visible in aerial surveys (Frederick et al. 1996), and for these species ground-based airboat searches were conducted. All tree islands in the study area were approached and searched for signs of nesting in the middle to late part of the nesting season (April through May). Colony site coordinates were determined by handheld GPS receivers, and populations of every species present were estimated by counting nests and/or number of adults that flushed from the site. In cases where a colony was visited more than once per season, I used the peak count of nests at that colony for a given species. I created point vector GIS layers representing the colony locations for each of the three focal species for every year. Given that point coordinates reflected the location of the airboat or airplane at the time of data collection, rather than the actual location of the colony, I used a spatial join operation to shift points to the centroid of the nearest patch of woody vegetation. In general, these shifts were not more than 100 m for points collected on the ground, or 300 m for points collected during aerial surveys. I then overlaid the water model grid and classified all cells as either used or unused by each of the species for every year.

PAGE 19

9 Distributional Relationships To rule out the possibility that the distribution of these species colonies was simply a reflection of the distribution of available sites, I calculated Ripleys K statistic for each species-year pattern of colony sites as well as for the parent pattern of all available sites, at 500 m intervals from 500 to 20,000 m. This statistic is a scale-dependent measure of spatial dispersion within a defined area (Fotheringham et al. 2000). However, it can also be used to assess whether two patterns within the same area are spatially similar, especially when one is a subset of the other. Differences between two patterns K statistics (i.e., the linearized Ripleys K, or l[d]) indicate that they are spatially dissimilar, with positive values of l(d) indicating a dispersed pattern relative to the parent pattern, and negative values of l(d) indicating a clustered pattern relative to the parent pattern. Since there is no significance test to determine how different from 0 an l(d) value must be to indicate that the patterns are different, I used bootstrapping to create sample-size-dependent 95% confidence intervals against which to compare each species-year point pattern (Fotheringham et al. 2000). Calculation of Variables All variables were calculated at the scale of a 3x3 neighborhood of cells (i.e., 3 km x 3 km) centered on the cell in question except INUN, which was calculated at the scale of the cell itself. I chose these small scales to maximize the spatial independence between samples, since larger neighborhoods would have increased the distance required between two cells before their neighborhoods did not overlap. The calculation of the three hydrological variables is described in Table 1-1. To calculate FOR I first identified all polygons in the vegetation map that contained any of the foraging habitat types. Within the maps classification system these included:

PAGE 20

10 all variations of savanna; spike rush and maidencane-spike rush prairies; all variations of non-graminoid emergent marsh; open water; mixed mangrove scrub; and, buttonwood and saw palmetto scrubs. Vegetation classes were hierarchically assigned to polygons qualitatively (e.g., a large area of wet prairie might contain some scattered shrubs), yet I needed a quantitative estimate of the areal extent of each class within polygons. Since vegetation classes could be assigned to any of the tiers within a polygon, with adjoining polygons often having the same classes in reversed order, the least biased method of estimating these extents was to allocate the same proportion to a given tier in every polygon. In polygons containing two classes of vegetation, I allocated 70% of the area to the dominant vegetation type and 30% to the secondary; in three-class polygons, I allocated 50, 30, and 20% of the area to the three tiers respectively. After allocating a polygons area between its vegetation classes, I calculated the total extent of foraging habitat present within the polygons. I intersected these polygons with a square grid polygon layer corresponding to the ELM grid cell boundaries to assign the vegetation polygons to the cell they inhabited, and summed the areas of foraging vegetation present within every cell. I recalculated the area of foraging vegetation in polygons that had been split by cell boundaries by multiplying their new area by the same proportion of foraging vegetation calculated for the parent polygon. Finally, I calculated the 3x3 neighborhood sums for every cell in WCA3. To ensure that the proportions I chose did not bias the ultimate values of FOR, I randomly chose 30 cells and performed a bootstrap analysis whereby the proportions allocated to vegetation classes within their polygons were allowed to randomly vary. Since the vegetation classes were originally assigned hierarchically, I constrained the

PAGE 21

11 proportions such that the dominant class proportion was always greater than the secondary, and this was always greater than the tertiary. I calculated FOR for each cell 1000 times based on these randomly-determined proportions and used these values to create an approximate 95% confidence interval for the real value of FOR in each cell. I then determined how many of the values of FOR calculated in the 30 cells using the proportions described above fell within these confidence intervals. Multivariate Analyses I used the Number Cruncher Statistical Systems software (NCSS, Hintze 2001) to evaluate logistic regression models where the response variable was the use (presence of at least one colony) or non-use of a water model cell by a species. Model inputs consisted of the values for all cells used by a given species in all 8 years combined with an equal number of randomly-selected unused cells from each year. I bootstrapped these randomly-selected sets for every species-year to ensure that their mean values for all four variables were within the 95% confidence interval suggested by the bootstrap. Results of the logistic regression analyses were interpreted within the information-theoretic approach of Burnham and Anderson (2002). Multiple models were evaluated and the Akaike information criterion (AIC; Akaike 1973) calculated for each one thus: AIC = -2(log-likelihood ratio) + 2K (2-1) where K was the number of parameters estimated in the model. The model with the lowest AIC was the best model among all those evaluated. The last term in Eq. 2-1 therefore represented a penalty against a models AIC value for every variable added to the model.

PAGE 22

12 The performance of a given model relative to the best one in the set was expressed using three values derived from the AIC values. The AIC difference (i) for model i was calculated: i = AICi AICmin (2-2) This was converted to Akaike weights (wi): wi = Rrri12121)exp(/)exp( (2-3) where the denominator in Eq. 2-3 was the sum of all models i values. The wi for model i was a measure of the likelihood (often expressed as a percent) that it is the best model. An analogous quantity is the evidence ratio for model i: evidence ratioi = wi / wj (2-4) where wj was the wi value for the best model in the set. The evidence ratio was a measure of how much better the best model was than model i. Because they relied on the sum of all models i values, the wi and evidence ratios were recalculated every time a new model entered the set. I followed the algorithm shown in Figure 2-3 to determine the models tested for each species. Upon identifying the best model for each species, I report three measures of its classification performance (Fielding & Bell 1997): 1) the number of used and unused cells correctly classified; 2) the Kappa statistic (), which measures the extent to which a models predictions are correct above the proportion expected simply by chance; and, 3) the area under the models ROC curve, which measures the likelihood that the value assigned by the final model to a randomly-selected used cell will be greater than that of a randomly-selected unused cell. I also summarized the relative importance of the

PAGE 23

13 variables tested for each species by adding the wi values for all models in which a given variable is included (Burnham & Anderson 2002: 167-168).

PAGE 24

14 Figure 2-1. Southeastern Florida showing the location of WCA3 within the larger Everglades ecosystem. The dotted areas along the east coast show the extent of urban landscapes and their proximity to what is left of the Everglades.

PAGE 25

15 Figure 2-2. Everglades Landscape Model cells in WCA3. Black dots indicate cells that contained potential colony sites and were included in multivariate analysis. The darker shaded area indicates all cells that were within a 3x3 cell neighborhood of analyzed cells and whose characteristics were therefore included in the calculation of variables. Less than 3% of the area of WCA3, represented by the unshaded areas, was excluded from analysis.

PAGE 26

16 Test saturated model (all 4 variables ) and all 3-term models. Best model has 3 terms Test all 2-term models nested within best model Figure 2-3. Algorithm for determining which models to test for each species. At every step the current best model competed with lower-term models comprised of its component variables. Once the best single-order model was identified, models that added a single second-order interaction term were tested. Test all 1-term models nested within best model Test all models comprised of best model + single second-order interaction term between component variables. BEST MODEL Best model has 4 terms Best model has 2 terms Best model has 3 terms Best model has 2 terms Best model has 1 term

PAGE 27

CHAPTER 3 RESULTS Distribution of Colony Sites Table 3-1 summarizes the number of colonies and water model cells inhabited by each species in each year, and Figure 3-1 presents the annual maps of these distributions. Figures 3-2 through 3-4 show the l(d) graphs for each species-year combination, along with their corresponding approximate 95% confidence intervals from the bootstrap simulation. Great Blue Heron colonies were extremely clustered relative to the parent pattern of all potential colony sites in all years except 1998, when the l(d) values were closer to the lower bootstrap limit. Great Egret colonies were dispersed relative to the parent pattern at all scales in 1993, 1994, and 1995; at small scales in 1996, and at large scales in 1998. Their colony patterns were mostly indistinguishable from the parent pattern at practically all scales in 1997, 1999, and 2000. Tricolored Heron colony patterns showed the most variability: they were highly dispersed relative to the parent pattern at medium to large scales in 1993; highly clustered at all scales in 1996; slightly clustered at small scales in 1994 and medium scales in 1999 and 2000; and, otherwise mostly indistinguishable from the parent pattern in 1995, 1997 and 1998. Bootstrap of Foraging Habitat Calculation Table 3-2 lists the values of FOR calculated for 30 randomly-chosen cells using the constant proportions described above and the upper and lower values of the approximate 95% confidence interval from the bootstrap analysis. Four of the 30 cells contained only polygons that did not vary in their values of FOR. In only one of the remaining cells was 17

PAGE 28

18 the value of FOR calculated using the constant proportions outside the 95% confidence interval from the bootstrap; however, this cells value was within the 99% confidence interval. This is strong evidence that the constant proportions used to calculate this variable did not bias the analysis. Multivariate Analyses I tested 11 models for Great Blue and Tricolored Herons, and 12 for Great Egrets (Tables 3-3 through 3-5). A single best model (wi > 98%) was found for the first two of these species, while the best two models for Great Egrets were nearly indistinguishable (wi values of 29.20% and 23.18%, respectively). Table 3-6 lists the best model(s) for each species and its associated measures of classification performance. Best models correctly classified between 68.8% (Great Egrets, model 1) and 78.9% (Great Blue Herons) of the used sites, and between 36.7% (Great Egrets, model 1) and 56.7% (Great Blue Herons) of the unused sites. Kappa statistics were routinely low, ranging between 0.055 (Great Egrets, model 1) and 0.357 (Great Blue Herons). The areas under the ROC curves also indicated relatively poor model performance; these ranged between 0.524 (Great Egrets, model 1) and 0.745 (Great Blue Herons). The amount of foraging habitat (FOR) around potential sites was always the most important variable, with all wi values greater than 98% (Table 3-7). It was followed by the likelihood that a site would remain inundated (INUN), with wi values ranging between 83.88 and 99.99%. As predicted, all three species responded to both of these variables positively, i.e., used cells had higher values overall than unused cells. The pattern for the remaining variables was less consistent across species, with Great Blue Herons responding strongly and positively to FWI, and Great Egrets responding weakly and negatively to WV (Table 3-7).

PAGE 29

19 Table 3-1. Number of colonies inhabited by each species, 1993-2000. Number of water model cells inhabited in parentheses. Great Blue Heron Great Egret Tricolored Heron 1993 156 (127) 32 (32) 11 (11) 1994 200 (153) 45 (44) 46 (43) 1995 298 (229) 39 (39) 40 (39) 1996 167 (136) 45 (43) 52 (42) 1997 101 (90) 34 (33) 8 (8) 1998 110 (97) 44 (41) 23 (23) 1999 258 (206) 77 (71) 66 (63) 2000 309 (235) 71 (62) 57 (53)

PAGE 30

20 Table 3-2. Values of FOR calculated in 30 randomly-chosen cells using constant proportions, compared to approximate 95% confidence intervals from a bootstrap analysis involving 1000 iterations where proportions were allowed to randomly vary. Lower limit of 95% confidence interval Value of FOR calculated using constant proportions Upper limit of 95% confidence interval Observed within confidence interval? 75 75 75 Constant value cell 251 339 644 Yes 750 750 750 Constant value cell 118 887 1,155 Yes 149 1,742 2,389 Yes 4,321 4,321 4,321 Constant value cell 5,925 5,925 5,925 Constant value cell 9,763 10,296 12,057 Yes 1,526 20,539 29,209 Yes 24,992 25,858 26,380 Yes 20,804 35,765 42,190 Yes 74,514 80,693 84,326 Yes 101,367 103,625 119,108 Yes 146,646 149,072 150,791 Yes 172,177 218,953 244,864 Yes 163,585 256,693 339,316 Yes 281,378 286,925 305,647 Yes 28,983 289,828 328,472 Yes 337,236 352,078 356,356 Yes 115,366 360,059 418,467 Yes 77,850 376,847 504,702 Yes 92,661 382,298 501,854 Yes 373,516 445,018 635,874 Yes 220,282 470,162 569,560 Yes 413,803 477,840 597,608 Yes 271,067 504,240 587,148 Yes 307,407 507,071 567,364 Yes 385,418 532,264 559,474 Yes 504,530 607,378 604,063 No, 99% confidence interval 531,142 638,535 646,606 Yes

PAGE 31

21 Table 3-3. Logistic regression models tested for Great Blue Herons and associated statistical values. Italicized models are the best from their subset, while the best overall model is shown in boldface. Model Log-Likelihood K AIC AIC wia Evidence Ratiosb Rank FOR+WV+FWI+INUN -1492.38 5 2994.76 16.15 3 FOR+WV+FWI -1495.94 2999.88 4 21.28 9 FOR+FWI+INUN -1493.21 2994.42 4 15.81 0.04% 2 FOR+WV+INUN -1495.21 2998.42 4 19.81 6 WV+FWI+INUN -1721.50 3451.00 4 472.40 10 FOR+FWI -1496.75 2999.51 3 20.90 8 FOR+INUN -1496.40 2998.81 3 20.21 7 FWI+INUN -1745.57 3497.14 3 518.54 11 FOR+FWI+INUN+FOR*FWI -1492.98 5 2995.96 17.36 0.02% 4 FOR+FWI+INUN+FOR*INUN -1493.15 5 2996.29 17.69 0.01% 5 FOR+FWI+INUN+FWI*INUN -1484.30 5 2978.60 0.00 99.89% 1.00 1 a indicates a wi value < 0.01%; b indicates an evidence ratio > 100 Table 3-4. Logistic regression models tested for Great Egrets and associated statistical values. Italicized models are the best from their subset, while the best overall models are shown in boldface. Model Log-Likelihood K AIC AIC wi Evidence Ratiosa Rank FOR+WV+FWI+INUN -498.86 5 1007.73 4.22 3.54% 8.25 8 FOR+WV+FWI -499.97 4 1007.94 4.44 3.17% 9.20 9 FOR+FWI+INUN -499.59 4 1007.17 3.67 4.67% 6.26 7 FOR+WV+INUN -498.86 4 1005.73 2.22 9.62% 3.03 4 WV+FWI+INUN -502.60 4 1013.20 9.69 0.23% 12 FOR+WV -500.48 1006.96 3 5.18% 3.46 5.63 6 FOR+INUN -499.61 1005.21 3 12.45% 1.71 2.35 3 WV+INUN -502.69 1011.38 3 0.57% 7.88 51.35 10 FOR -501.08 1006.16 2 2.65 7.76% 3.76 5 INUN -503.99 1011.99 2 8.48 0.42% 69.43 11 FOR+INUN+FOR*INUN -497.75 4 1003.51 0.00 29.20% 1.00 1 FOR+WV+INUN+FOR*INUN -496.98 5 1003.97 0.46 23.18% 1.26 2 b indicates an evidence ratio > 100

PAGE 32

22 Table 3-5. Logistic regression models tested for Tricolored Herons and associated statistical values. Italicized models are the best from their subset, while the best overall model is shown in boldface. Model Log-Likelihood K AIC AIC wia Evidence Ratiosb Rank FOR+WV+FWI+INUN -363.45 5 736.90 14.14 0.08% 8 FOR+WV+FWI -364.24 4 736.48 13.72 0.10% 7 FOR+FWI+INUN -363.83 4 735.66 12.90 0.16% 6 FOR+WV+INUN -363.50 4 734.99 12.23 0.22% 4 WV+FWI+INUN -387.66 4 783.33 60.57 10 FOR+WV -364.70 735.40 3 12.64 0.18% 5 FOR+INUN -363.92 733.84 3 11.07 0.39% 2 WV+INUN -387.67 781.33 3 58.57 9 FOR -365.35 734.70 2 11.94 0.25% 3 INUN -390.82 785.64 2 62.88 11 FOR+INUN+FOR*INUN -357.38 4 722.76 0.00 98.62% 1.00 1 a indicates a wi value < 0.01%; b indicates an evidence ratio > 100

PAGE 33

23 Table 3-6. Best models for each species and associated measures of classification performance. Great Blue Herons Model FOR + FWI + INUN FWI*INUN n, % Correct (Used) 1005, 78.9% n, % Correct (Unused) 722, 56.7% Statistic 0.357 Area Under ROC Curve 0.745 Great Egrets Model 1 FOR + INUN + FOR*INUN n, % Correct (Used) 251, 68.8% n, % Correct (Unused) 134, 36.7% Statistic 0.055 Area Under ROC Curve 0.524 Model 2 FOR WV + INUN + FOR*INUN n, % Correct (Used) 274, 75.1% n, % Correct (Unused) 152, 41.6% Statistic 0.167 Area Under ROC Curve 0.559 Tricolored Herons Model FOR + INUN + FOR*INUN n, % Correct (Used) 218, 77.3% n, % Correct (Unused) 137, 48.6% Statistic 0.259 Area Under ROC Curve 0.654 Table 3-7. Relative importance of variables to each species. Great Blue Heron Great Egret Tricolored Heron FOR 100.00% 98.78% 100.00% WV 0.04% 45.50% 0.58% FWI 99.99% 11.61% 0.34% INUN 99.99% 83.88% 99.47% FOR*FWI 0.02% FOR*INUN 0.01% 52.38% 98.62% FWI*INUN 99.89%

PAGE 34

24 Figure 3-1. Distribution of ardeid colonies in WCA3. A) 1993. B) 1994. C) 1995. D) 1996. E) 1997. F) 1998. G) 1999. H) 2000. Triangles indicate Great Blue Heron colonies, stars indicate Great Egret colonies, and circles indicate Tricolored Heron colonies.

PAGE 35

25 Figure 3-1. Continued

PAGE 36

26 A-1-0.8-0.6-0.4-0.200.20.40.605000100001500020000Scale (m)l(d) B-1.2-1-0.8-0.6-0.4-0.200.20.40.605000100001500020000Scale (m)l(d) Figure 3-2. Relationship of Great Blue Heron colonies relative to all available sites in WCA3 as shown by linearized Ripleys K (l[d]) graphs. A) 1993. B) 1994. C) 1995. D) 1996. E) 1997. F) 1998. G) 1999. H) 2000. Sample sizes for each year listed in Table 3-1. Dashed lines: 95% confidence intervals from Monte Carlo simulations of 999 identically-sized, randomly-selected samples from each year; solid line: l(d) graph for used sites.

PAGE 37

27 C-0.8-0.6-0.4-0.200.20.405000100001500020000Scale (m)l(d) D-1.2-1-0.8-0.6-0.4-0.200.20.40.605000100001500020000Scale (m)l(d) Figure 3-2. Continued

PAGE 38

28 E-2-1.5-1-0.500.5105000100001500020000Scale (m)l(d) F-1.4-1.2-1-0.8-0.6-0.4-0.200.20.40.60.805000100001500020000Scale (m)l(d) Figure 3-2. Continued

PAGE 39

29 G-0.8-0.6-0.4-0.200.20.405000100001500020000Scale (m)l(d) H-0.6-0.5-0.4-0.3-0.2-0.100.10.20.30.405000100001500020000Scale (m)l(d) Figure 3-2. Continued

PAGE 40

30 A-2-1.5-1-0.500.511.5205000100001500020000Scale (m)l(d) B-2-1.5-1-0.500.511.5205000100001500020000Scale (m)l(d) Figure 3-3. Relationship of Great Egret colonies relative to all available sites in WCA3 as shown by linearized Ripleys K (l[d]) graphs. A) 1993. B) 1994. C) 1995. D) 1996. E) 1997. F) 1998. G) 1999. H) 2000. Sample sizes for each year listed in Table 3-1. Dashed lines: 95% confidence intervals from Monte Carlo simulations of 999 identically-sized, randomly-selected samples from each year; solid line: l(d) graph for used sites.

PAGE 41

31 C-1.5-1-0.500.511.505000100001500020000Scale (m)l(d) D-1.5-1-0.500.511.505000100001500020000Scale (m)l(d) Figure 3-3. Continued

PAGE 42

32 E-2-1.5-1-0.500.511.505000100001500020000Scale (m)l(d) F-1.5-1-0.500.511.505000100001500020000Scale (m)l(d) Figure 3-3. Continued

PAGE 43

33 G-1.5-1-0.500.5105000100001500020000Scale (m)l(d) H-1.5-1-0.500.5105000100001500020000Scale (m)l(d) Figure 3-3. Continued

PAGE 44

34 A-2-1.5-1-0.500.511.505000100001500020000Scale (m)l(d) B-1.5-1-0.500.511.505000100001500020000Scale (m)l(d) Figure 3-4. Relationship of Tricolored Heron colonies relative to all available sites in WCA3 as shown by linearized Ripleys K (l[d]) graphs. A) 1993. B) 1994. C) 1995. D) 1996. E) 1997. F) 1998. G) 1999. H) 2000. Sample sizes for each year listed in Table 3-1. Dashed lines: 95% confidence intervals from Monte Carlo simulations of 999 identically-sized, randomly-selected samples from each year; solid line: l(d) graph for used sites.

PAGE 45

35 C-2.5-1.5-0.50.51.52.505000100001500020000Scale (m)l(d) D-2-1.5-1-0.500.511.505000100001500020000Scale (m)l(d) Figure 3-4. Continued

PAGE 46

36 E-2.5-1.5-0.50.51.52.53.54.505000100001500020000Scale (m)l(d) F-2-1.5-1-0.500.511.522.505000100001500020000Scale (m)l(d) Figure 3-4. Continued

PAGE 47

37 G-1.5-1-0.500.511.505000100001500020000Scale (m)l(d) H-1.5-1-0.500.511.505000100001500020000Scale (m)l(d) Figure 3-4. Continued

PAGE 48

CHAPTER 4 DISCUSSION Effects of Site Availability The spatial patterns of colony sites used by these three species during the study period were in general dissimilar from the overall pattern of available sites (Figures 3-2 through 3-4). This was always true for Great Blue Herons, and was true at most scales for Great Egrets and Tricolored Herons in 4 and 5 of the 8 years, respectively. These two species exhibited both clustered and dispersed patterns, which likely reflects responses by the birds to different hydrological conditions in each year. These results indicate that site selection by these three species was probably independent of overall site availability or distribution. Model Performance Although there was very strong evidence supporting the best models for Great Blue and Tricolored Herons, relative to the total set of models I tested, model classification performance was poor overall. None of the best models achieved a Kappa statistic of 0.4, which is often considered to be the minimum value for good performance (Fielding & Bell 1997), and only the best model for Great Blue Herons had an area under its ROC curve of at least 0.7. Models always did a better job of classifying used than unused cells (Table 3-6), which suggests that there were more appropriate sites available than were used (Fielding & Bell 1997). A major limitation of this analysis was the coarse spatial resolution of the hydrological data, which prevented me from including variables that operate at the scale of the sites themselves such as vegetation type or patch size. Finally, 38

PAGE 49

39 simply classifying cells as used or unused ignored the enormous variation in the number of Great Egret pairs nesting at sites within WCA3, which is probably why the models for this species performed the worst. Despite these limitations, these results clearly show that the amount of foraging habitat around sites, and the likelihood that they would remain inundated throughout the breeding season, were important influences on the colony site selection for these three species. This is the first study to quantitatively establish the importance of deterring mammalian predators as an influence on colony site selection in this group. These results also establish that the distribution of foraging habitats is as important to wading birds in an enormous wetland-dominated landscape such as the Everglades, as it is in more typical terrestrial landscapes where wetlands are part of a habitat mosaic (Gibbs et al. 1987; Gibbs 1991; Gibbs & Kinkel 1997). Responses to Hydrological Variability In general, neither the average weekly proportional decline in water depths (FWI) nor the spatial variation in water depths (WV) were important influences on colony site selection in these three species. WV was weakly selected against by Great Egrets, and not included in the best models for either of the other two species. Only Great Blue Herons selected cells with large values of FWI. However, areas that began the breeding season with deeper water were likely to have the largest values of for this variable, since they could experience proportional declines for much longer before going dry. Thus it is unclear whether FWI actually represented prey availability for Great Blue Herons or was instead a proxy for deep water. Great Blues also selected against cells that had both rapid declines in water depths and a high likelihood of remaining inundated (i.e., the INUN*FWI interaction term). These could represent either permanently shallow areas

PAGE 50

40 (such as on the west side of WCA3, where water flow westward into Big Cypress National Preserve is unimpeded), and/or areas near canals, both of which were mostly avoided by this species (Figure 3-1). These results suggest that ardeid wading birds are selecting colony sites that are characterized by stable, rather than variable, water depths. Another possible interpretation is that colony site selection by these species does not reflect an attempt to predict future conditions (i.e., a bet hedging strategy), but is rather simply a response to conditions at the time of nesting (Bancroft et al. 1994). These three species are known to be less dependent on concentrated patches of prey created by declining water depths than tactile/social foragers such as Wood Storks (Mycteria americana), White Ibises (Eudocimus albus), and Snowy Egrets (Egretta thula) (Kahl 1964; Gawlik 2002), so it is also possible that they have not experienced any selective pressure to recognize or respond to this environmental characteristic. Management Implications Previous authors have speculated that the size of wading bird breeding populations in the Everglades is dependent on the extent of open slough habitat available to them (Bancroft et al. 1994; Bancroft et al. 2002), and the results of this study establish that this is the most important environmental characteristic determining where the birds nest. Hydroperiod and fire are the two largest factors controlling the distribution of wetland vegetation (Gunderson 1994), so water management can play an important role in maximizing the extent of open slough habitat within the Everglades. Water depths should be maintained at a deep enough level to prevent most breeding sites from going dry, but shallow enough for wading birds to forage. Using these guidelines could make a major

PAGE 51

41 contribution to re-establishing historical breeding populations of wading birds in the Everglades.

PAGE 52

LITERATURE CITED Akaike, H. 1973. Information theory as an extension of the maximum likelihood principle. Pages 267-281 in B. N. Petrov and F. Csaki, editors. Second International Symposium on Information Theory. Akademiai Kiado, Budapest. Bancroft, G. T. 1989. Status and conservation of wading birds in the Everglades. American Birds 43:1258. Bancroft, G. T., D. E. Gawlik, and K. Rutchey. 2002. Distribution of wading birds relative to vegetation and water depths in the northern Everglades of Florida, USA. Waterbirds 25:263. Bancroft, G. T., J. C. Ogden, and B. W. Patty. 1988. Wading bird colony formation and turnover relative to rainfall in the Corkscrew Swamp area of Florida during 1982 through 1985. Wilson Bulletin 100:50. Bancroft, G. T., A. M. Strong, R. J. Sawicki, W. Hoffman, and S. D. Jewell. 1994. Relationships among wading bird foraging patterns, colony locations, and hydrology in the Everglades. Pages 615 in S. Davis and J. Ogden, editors. Everglades: the ecosystem and its restoration. St. Lucie Press, Delray Beach, Florida, USA. Baxter, G. S., and P. G. Fairweather. 1998. Does available foraging area, location or colony character control the size of multispecies egret colonies? Wildlife Research 25:23. Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference: a practical information-theoretic approach. Second edition. Springer-Verlag, New York, New York, USA. Butler, R. W. 1992. Great Blue Heron (Ardeas herodias). No. 25 in A. Poole and F. Gill, editors. The Birds of North America. The Academy of Natural Sciences, Philadelphia, Pennsylvania, USA; The American Ornithologists Union, Washington, D.C., USA. Custer, T. W., and R. G. Osborn. 1978. Feeding-site description of three heron species near Beaufort, North Carolina. Pages 355 in A. Sprunt IV, J. C. Ogden, and S. Winckler, editors. Wading Birds, Research Report No. 7. National Audubon Society, New York, New York, USA. 42

PAGE 53

43 Environmental Systems Research Institute, 2002. Arc/Info Workstation. Version 8.3. Redlands, California, USA. Fielding, A. H., and J. F. Bell. 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environmental Conservation 24:38. Fitz, H. C., J. Godin, B. Trimble, and N. Wang. 2004. Everglades Landscape Model documentation: ELM v2.2. South Florida Water Management District, West Palm Beach, Florida, USA. Fotheringham, A. S., C. Brunsdon, and M. Charlton. 2000. Quantitative geography: perspectives on spatial data analysis. Sage Publications, London, UK. Frederick, P. C. 1997. Tricolored heron (Egretta tricolor). No. 306 in A. Poole and F. Gill, editors. The Birds of North America. The Academy of Natural Sciences, Philadelphia, Pennsylvania, USA; The American Ornithologists Union, Washington, D.C., USA. Frederick, P. C. 2002. Wading birds in the marine environment. Pages 617 in E. A. Schreiber and J. Burger, editors. Biology of marine birds. CRC Press, Boca Raton, Florida, USA. Frederick, P. C., and M. W. Collopy. 1989a. Nesting success of five ciconiiform species in relation to water conditions in the Florida Everglades. Auk 106:625. Frederick, P. C., and M. W. Collopy. 1989b. The role of predation in determining reproductive success of colonially nesting wading birds in the Florida Everglades. Condor 91:860. Frederick, P. C., and J. C. Ogden. 2001. Pulsed breeding of long-legged wading birds and the importance of infrequent severe drought conditions in the Florida Everglades. Wetlands 21:484. Frederick, P.C., and M. G. Spalding. 1994. Factors affecting reproductive success of wading birds (Ciconiiformes) in the Everglades ecosystem. Pages 659 in S. Davis and J. Ogden, editors. Everglades: the ecosystem and its restoration. St. Lucie Press, Delray Beach, Florida, USA. Frederick, P.C., T. Towles, R. J. Sawicki, and G. T. Bancroft. 1996. Comparison of aerial and ground techniques for discovery and census of wading bird (Ciconiiformes) nesting colonies. Condor 98: 837. Gawlik, D. E. 2002. The effects of prey availability on the numerical response of wading birds. Ecological Monographs 72:329. Gibbs, J. P. 1991. Spatial relationships between nesting colonies and foraging areas of Great Blue Herons. Auk 108:764.

PAGE 54

44 Gibbs, J. P., and L. K. Kinkel. 1997. Determinants of the size and location of Great Blue Heron colonies. Colonial Waterbirds 20:1. Gibbs, J. P., S. Woodward, M. L. Hunter, and A. E. Hutchinson. 1987. Determinants of Great Blue Heron colony distribution in coastal Maine. Auk 104:38. Gunderson, L. H. 1994. Vegetation of the Everglades: determinants of community composition. Pages 323 in S. Davis and J. Ogden, editors. Everglades: the ecosystem and its restoration. St. Lucie Press, Delray Beach, Florida, USA. Hintze, J. 2001. NCSS and PASS. Number Cruncher Statistical Systems. Kaysville, Utah, USA. Hoffman, W., G. T. Bancroft, and R. J. Sawicki. 1994. Foraging habitat of wading birds in the Water Conservation Areas of the Everglades. Pages 585 in S. Davis and J. Ogden, editors. Everglades: the ecosystem and its restoration. St. Lucie Press, Delray Beach, Florida, USA. Kahl, M. P., Jr. 1964. Food ecology of the Wood Stork (Mycteria americana) in Florida. Ecological Monographs 34:97. Kushlan, J. A. 1976. Wading bird predation in a seasonally fluctuating pond. Auk 93:464. Kushlan, J. A. 1987. External threats and internal management: the hydrologic regulation of the Everglades, Florida, USA. Environmental Management 11:109. McCrimmon, D. A., Jr., J. C. Ogden, and G. T. Bancroft. 2001. Great Egret (Ardea alba). No. 570 in A. Poole and F. Gill, editors. The Birds of North America. The Academy of Natural Sciences, Philadelphia, Pennsylvania, USA; The American Ornithologists Union, Washington, D.C., USA. Ogden, J. C. 1994. A comparison of wading bird nesting colony dynamics (1931 and 1974) as an indication of ecosystem conditions in the southern Everglades. Pages 533 in S. Davis and J. Ogden, editors. Everglades: the ecosystem and its restoration. St. Lucie Press, Delray Beach, Florida, USA. Post, W. 1990. Nest survival in a large ibis-heron colony during a three-year decline to extinction. Colonial Waterbirds 13:50. Powell, G. V. N. 1987. Habitat use by wading birds in a subtropical estuary: implications of hydrography. Auk 104:740. Rodgers, J. A. 1987. On the antipredator advantages of coloniality: a word of caution. Wilson Bulletin 99:269. Rutchey, K., L. Vilchek, and M. Love, in review. Development of a vegetation map for Water Conservation Area 3.

PAGE 55

45 Smith, J. P. 1995. Foraging flights and habitat use of nesting wading birds (Ciconiiformes) at Lake Okeechobee, Florida. Colonial Waterbirds 18:139. Smith, J. P., and Collopy, M. W. 1995. Colony turnover, nest success and productivity, and causes of nest failure among wading birds (Ciconiiformes) at Lake Okeechobee, Florida (1989). Pages 287 in N. G. Aumen and R. G. Wetzel, editors. Ecological studies on the littoral and pelagic systems of Lake Okeechobee, Florida (USA). E. Schweizerbart'sche Verlagsbuchhandlung, Stuttgart, Germany. Smith, J. P., J. R. Richardson, and M. W. Collopy. 1995. Foraging habitat selection among wading birds (Ciconiiformes) at Lake Okeechobee, Florida, in relation to hydrology and vegetative cover. Pages 247 in N. G. Aumen and R. G. Wetzel, editors. Ecological studies on the littoral and pelagic systems of Lake Okeechobee, Florida (USA). E. Schweizerbart'sche Verlagsbuchhandlung, Stuttgart, Germany. Surdick, J. A., Jr. 1998. Biotic and abiotic indicators of foraging site selection and foraging success of four ciconiiform species in the freshwater Everglades of Florida. Thesis. University of Florida, Gainesville, Florida, USA. Taylor, R. J., and E. D. Michael. 1971. Predation on an inland heronry in eastern Texas. Wilson Bulletin 83:172.

PAGE 56

BIOGRAPHICAL SKETCH Matthew John Bokach completed his Bachelor of Science degree with a double major in biology and chemistry at Adrian College (Adrian, Michigan) in 1994. He fled the sciences for 2 years in a masters program in student affairs administration at Michigan State University, then returned to them as a U.S. Peace Corps Volunteer at Lukosi Government School in Hwange District, Zimbabwe. While teaching general science at Lukosi, Matthew fell in love with the southern African avifauna and spent an additional 2 years in Zimbabwe, mostly so he could continue watching them. He returned to the United States in 2000 and began a masters degree in interdisciplinary ecology at the University of Florida in 2002. He plans to move to the San Francisco Bay area, pursue a career in GIS application development, and eventually flee the sciences once again to pursue a career in popular music. 46