DEVELOPMENT, EVALUATION, AND APPLICATION OF SPATIO-TEMPORAL

WADING BIRD FORAGING MODELS TO GUIDE EVERGLADES RESTORATION

by

James M. Beerens

A Dissertation Submitted to the Faculty of

The Charles E. Schmidt College of Science

in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

Florida Atlantic University

Boca Raton, FL

May 2014

ii

ACKNOWLEDGMENTS

Funding for this entire research project was provided by the U.S. Army Corps of

Engineers. I was also supported through a scholarship from The Everglades Foundation and the Presidential Fellowship and Newell Doctoral Fellowship from Florida Atlantic

University.

First and foremost, I thank my advisor, Dr. Erik Noonburg, for the countless hours he contributed to the ideas underlying this dissertation. His availability and support helped provide a platform for the project to be a success. Dr. Dale Gawlik played an early and influential role in my development as an avian ecologist in the Everglades and provided the opportunity to continue the work that is my passion. Further, his appreciation for conservation and the natural world continue to provide a model for a good mentor. I also thank Dr. Brian Benscoter and Dr. Ed Proffitt for their contribution to my dissertation as committee members.

Some of the most important people in this entire process are fellow scientists that have challenged my ideas and provided valuable feedback along my journey. I especially want to thank Mark Barrett, Bryan Botson, Mark Cook, and Peter Frederick for the time and effort they contributed to improving my quality of work. I also am grateful to Doug

Donalson and Andy Loschiavo at the U.S. Army Corps of Engineers for their dedication to this project and willingness to trust my vision in representing how wading birds interact with a changing world.

iii

I thank Thomas Bancroft, Sonny Bass, Marty Fleming, Wayne Hoffman, Dave

Nelson, Jim Shortemeyer, and all the other people who spent numerous hours in small planes collecting valuable wading bird data. I also am indebted to Paul Conrads, Heather

Henkel, Aaron Higer, Pamela Telis, and the Everglades Depth Estimation Network

(EDEN) team for their dedication to provide a high resolution hydrology model for the

Everglades. Leonard Pearlstine was instrumental in providing the appropriate hydrology data from the Regional System Model (RSM) and South Florida Water Management

Model (SFWMM) that contributed to the applications in this project.

I am sincerely thankful to my family for their love and the value system they instilled in me from a young age. They called me to continually look out into the world and speak for those without a voice. Under increasing pressure from human development, countless plant and animal species have precipitously declined. It is our call as environmental scientists to accurately portray this change to plan for a sustainable future. Consequently, I also thank the birds for their incredible ability to adapt to human- created challenges and hold on as we correct some of the ecological mistakes from our past.

Finally, I thank my wife Maria for enduring the long hours of a graduate student, and for being a sounding board and skillful editor. She has stood by me through my hardest challenges and I am forever grateful for her inspiration in my life. This dissertation has greatly benefited from her selflessness, grace, and humor.

iv

ABSTRACT

Author: James M. Beerens

Title: Development, Evaluation, and Application of Spatio-Temporal Wading Bird Foraging Models to Guide Everglades Restoration

Institution: Florida Atlantic University

Degree: Doctor of Philosophy

Dissertation Advisor: Dr. Erik. Noonburg

Year: 2014

In south Florida, the Greater Everglades ecosystem supports sixteen species of wading birds. Wading birds serve as important indicator species because they are highly mobile, demonstrate flexible habitat selection, and respond quickly to changes in habitat quality. Models that establish habitat relationships from distribution patterns of wading birds can be used to predict changes in habitat quality that may result from restoration and climate change. I developed spatio-temporal species distribution models for the

Great Egret, White Ibis, and Wood Stork over a decadal gradient of environmental conditions to identify factors that link habitat availability to habitat use (i.e., habitat selection), habitat use to species abundance, and species abundance (over multiple scales) to nesting effort and success. Hydrological variables (depth, recession rate, days since drydown, reversal, and hydroperiod) over multiple temporal scales and with existing links to wading bird responses were used as proxies for landscape processes that

v

influence prey availability (i.e., resources). In temporal foraging conditions (TFC) models, species demonstrated conditional preferences for resources based on resource levels at differing temporal scales. Wading bird abundance was highest when prey production from optimal periods of wetland inundation was concentrated in shallow depths. Similar responses were observed in spatial foraging conditions (SFC) models predicting spatial occurrence over time, accounting for spatial autocorrelation. The TFC index represents conditions within suitable depths that change daily and reflects patch quality, whereas the SFC index spatially represents suitability of all cells and reflects daily landscape patch abundance. I linked these indices to responses at the nest initiation and nest provisioning breeding phases from 1993-2013. The timing of increases and overall magnitude of resource pulses predicted by the TFC in March and April were strongly linked to breeding responses by all species. Great Egret nesting effort and success were higher with increases in conspecific attraction (i.e., clustering). Wood Stork nesting effort was closely related to timing of concurrently high levels of patch quality

(regional scale) and abundance (400-m scale), indicating the importance of a multi-scaled approach. The models helped identify positive and negative changes to multi-annual resource pulses from hydrological restoration and climate change scenarios, respectively.

vi

DEVELOPMENT, EVALUATION, AND APPLICATION OF SPATIO-TEMPORAL

WADING BIRD FORAGING MODELS TO GUIDE EVERGLADES RESTORATION

LIST OF TABLES ...... x LIST OF FIGURES ...... xii CHAPTER 1: INTRODUCTION ...... 1 INTRODUCTION ...... 1 LITERATURE CITED ...... 7 CHAPTER 2: ANALYSIS OF HABITAT USE ...... 10 METHODS ...... 10 RESULTS AND DISCUSSION ...... 12 Depth ...... 12 Recession Rate ...... 16 Prey Production and Days since Drydown ...... 20 LITERATURE CITED ...... 21 CHAPTER 3: MODELING SPATIO-TEMPORAL RESPONSES OF WADING BIRD INDICATOR SPECIES ACROSS RESOURCE GRADIENTS FOR EVERGLADES RESTORATION...... 25 ABSTRACT ...... 25 INTRODUCTION ...... 26 METHODS ...... 32 Data and Variables ...... 32 Identifying Available Habitat ...... 34 Temporal Foraging Conditions (TFC; Patch Quality) ...... 34 Spatial Foraging Conditions (SFC; Patch Abundance) ...... 36 RESULTS ...... 37 Identifying Available Habitat ...... 37

vii

Depth Use ...... 38 Recession Rate Use ...... 39 Days since Drydown Use ...... 40 Temporal Foraging Conditions (Patch Quality) ...... 41 Spatial Foraging Conditions (Patch Abundance) ...... 42 DISCUSSION ...... 44 Water Dynamics and Prey ...... 44 Temporal Abundance ...... 46 Spatial Abundance ...... 49 Application ...... 51 LITERATURE CITED ...... 53 CHAPTER 4: HABITAT EVALUATION OVER MULTIPLE PHASES OF BREEDING USING DYNAMIC SPATIO-TEMPORAL HABITAT SUITABILITY INDICES ...... 72 ABSTRACT ...... 72 INTRODUCTION ...... 73 METHODS ...... 77 Study Area ...... 77 Wading Bird Distribution Evaluation Models (WADEM) ...... 78 Nesting Effort ...... 81 Nesting Success ...... 81 Variables & Statistical Methods ...... 82 RESULTS ...... 84 Great Egret ...... 84 White Ibis ...... 85 Wood Stork ...... 86 DISCUSSION ...... 86 Application ...... 91 LITERATURE CITED ...... 93

viii

CHAPTER 5: CENTRAL EVERGLADES PLANNING PROJECT APPLICATION ...... 108 INTRODUCTION ...... 108 METHODS ...... 109 Alternative Restoration Scenarios ...... 109 Wading Bird Distribution Evaluation Models (WADEM) ...... 109 RESULTS ...... 110 Great Egret ...... 110 White Ibis ...... 110 Wood Stork ...... 111 Summary ...... 111 LITERATURE CITED ...... 113 CHAPTER 6: CLIMATE CHANGE APPLICATION ...... 124 INTRODUCTION ...... 124 METHODS ...... 124 Alternative Climate Scenarios ...... 124 Preparation of the SFWMM scenarios ...... 125 Wading Bird Distribution Evaluation Models (WADEM) ...... 125 RESULTS ...... 126 LITERATURE CITED ...... 128

ix

TABLES

Table 3.1. Ranking of candidate models describing variables influencing daily mean

depth use of Great Egrets, White Ibises, and Wood Storks in the Florida

Everglades (Proc Mixed)...... 66

Table 3.2. Ranking of candidate models describing variables influencing daily mean

2-week recession rate use of Great Egrets, White Ibises, and Wood Storks in

the Florida Everglades (Proc Mixed)...... 67

Table 3.3. Ranking of candidate models describing variables influencing daily mean

days since drydown (DSD) use of Great Egrets, White Ibises, and Wood Storks

in the Florida Everglades (Proc Mixed)...... 68

Table 3.4. Ranking of candidate models describing variables influencing daily mean

flock abundance of Great Egrets, White Ibises, and Wood Storks in the Florida

Everglades (Proc Mixed)...... 69

Table 3.5. Ranking of candidate models describing variables influencing daily

individual abundance of Great Egrets, White Ibises, and Wood Storks in the

Florida Everglades (Proc Mixed)...... 70

Table 3.6. Ranking of candidate models describing variables influencing frequency

of cell use (i.e., spatial occurrence) over the study period for the Great Egret,

White Ibis, and Wood Stork (Proc Glimmix)...... 71

x

Table 4.1. Ranking of candidate models describing variables influencing nesting

effort (i.e., max nesting pairs) of Great Egrets, White Ibises, and Wood Storks

in the Florida Everglades (Proc Mixed)...... 106

Table 4.2. Ranking of candidate models describing variables influencing nesting

success (Mayfield method) of Great Egrets and White Ibises in the Florida

Everglades (Proc GLM)...... 107

xi

FIGURES

Figure 2.1. Mean Everglades Depth Estimation Network (EDEN) Dry Season

Water Depth (±SE) for 2000-2009...... 24

Figure 3.1. South Florida study system displaying Everglades hydrological basins

(regions) and Systematic Reconnaissance Flight (SRF) survey extent...... 61

Figure 3.2. Daily mean landscape flocks and individuals (fourth-root transformed)

predicted by the model-averaged terms for the Great Egret...... 62

Figure 3.3. Daily mean landscape flocks and individuals (fourth-root transformed)

predicted by the model-averaged terms for the White Ibis...... 63

Figure 3.4. Daily mean landscape flocks and individuals (fourth-root transformed)

predicted by the model-averaged terms for the Wood Stork ...... 64

Figure 3.5. Map displaying XY parameter estimates accounting for residual spatial

correlation of Great Egret frequency of use...... 65

Figure 4.1. South Florida study system displaying Everglades hydrological basins

(regions)...... 101

Figure 4.2. Hydrograph depicting Everglades Depth Estimation Network (EDEN)

mean water depths (cm) during 1993–1999 and 2000–2013...... 102

Figure 4.3. Great Egret nest effort (i.e., max nesting pairs) and success estimates

(± SD) from 1993–2013...... 103

xii

Figure 4.4. White Ibis nest effort (i.e., max nesting pairs) and success (± SD)

estimates from 1993–2013...... 104

Figure 4.5. Wood Stork nest effort (i.e., max nesting pairs) and success (± SD)

estimates from 1993–2013...... 105

Figure 5.1. Cumulative mean percent change in Great Egret temporal foraging

conditions (TFC) for the tentatively selected plan (TSP), relative to the baseline

(future without restoration; FWO) during the breeding months of Jan-May,

1967-2004...... 114

Figure 5.2. Cumulative mean percent change in Great Egret, White Ibis, and Wood

Stork spatial foraging conditions (SFC) for the tentatively selected plan (TSP),

relative to the baseline (future without restoration; FWO) during the breeding

months of Jan-May, 1967-2004...... 115

Figure 5.3. The coloration in the map represents the mean percent change in Great

Egret spatial foraging conditions (Jan – May, 1967-2004) for the tentatively

selected plan (Alt 4r2) relative to future without restoration (FWO)...... 116

Figure 5.4. Mean Great Egret temporal foraging conditions (TFC) for the

tentatively selected plan (A4r2) and baseline (future without restoration; FWO)

during the breeding months of Jan-May, 1967-2004...... 117

Figure 5.5. Cumulative mean percent change in White Ibis temporal foraging

conditions (TFC) for the tentatively selected plan (TSP), relative to the baseline

(future without restoration; FWO) during the breeding months of Jan-May,

1967-2004...... 118

xiii

Figure 5.6. Mean White Ibis temporal foraging conditions (TFC) for the tentatively

selected plan (A4r2) and baseline (future without restoration; FWO) during the

breeding months of Jan-May, 1967-2004...... 119

Figure 5.7. The coloration in the maps represents the mean percent change in White

Ibis spatial foraging conditions (Jan – May, 1967-2004) for the tentatively

selected plan (Alt 4r2) relative to future without restoration (FWO)...... 120

Figure 5.8. Cumulative mean percent change in Wood Stork temporal foraging

conditions (TFC) for the tentatively selected plan (TSP), relative to the baseline

(future without restoration; FWO) during the breeding months of Jan-May,

1967-2004...... 121

Figure 5.9. The coloration in the maps represents the mean percent change in Wood

Stork spatial foraging conditions (Jan – May, 1967-2004) for the tentatively

selected plan (A4r2) relative to future without restoration (FWO)...... 122

Figure 5.10. Mean Wood Stork temporal foraging conditions (TFC) for the

tentatively selected plan (A4r2) and baseline (future without restoration; FWO)

during the breeding months of Jan-May, 1967-2004...... 123

Figure 6.1. Cumulative mean percent change in Great Egret (GE), White Ibis (WI),

and Wood Stork (WS) cell use for the simulations +RF+ET, +ET, and -RF+ET,

relative to the baseline during the breeding months of Jan-May, 1967-2005...... 129

Figure 6.2. Predicted mean Great Egret habitat suitability maps (1967-2005) for 4

climate scenarios (clockwise: BASE, +RF+ET, -RF+ET, +ET)...... 130

Figure 6.3. Predicted mean White Ibis habitat suitability maps (1967-2005) for 4

climate scenarios (clockwise: BASE, +RF+ET, -RF+ET, +ET)...... 131 xiv

Figure 6.4. Predicted mean Wood Stork habitat suitability maps (1967-2005) for 4

climate scenarios (clockwise: BASE, +RF+ET, -RF+ET, +ET)...... 132

xv

CHAPTER 1: INTRODUCTION

INTRODUCTION

In south Florida, the Greater Everglades ecosystem supports sixteen species of wading birds. Plentiful food and habitat provided by this vast system have historically supported large breeding populations of wading birds. John James Audubon wrote during a visit to south Florida in the early 1830s: "We observed great flocks of wading birds flying overhead toward their evening roosts .... They appeared in such numbers to actually block out the light from the sun for some time" (Audubon, 1832). In a system defined by hydrological variability and intermittent resource pulses, wading birds thrived.

During peak years, the number of White Ibises (Eudocimus albus) nesting in south

Florida in the 1930-1940s was estimated at 100,000-200,000 pairs and the number of

Wood Storks (Mycteria americana) nesting reached 5,000-20,000 (Crozier and Gawlik

2003). Due to human alteration of the south Florida landscape over the last 70 years, the size of breeding populations have declined by an estimated 90% (Frederick and Spalding

1994).

Significant alteration of the Everglades landscape began as reaction to flooding of human settlements in 1926, 1928 and 1947; dikes, levees, and canals were created to divert water flow for flood protection, agriculture, and human use (Light and Dineen

1

1994). Further drainage was prompted by a population boom after World War II leading to a disproportionate reduction (85%) in short hydroperiod wetlands that served as wading bird habitat (Fleming et al. 1994). In the remaining Everglades, the hydrological properties have been severely altered by regulating water flows through discrete management units via man-made structures. These practices have changed hydroperiods

(i.e., duration of inundation), water flow, water recession during the dry season, and water depths; thus altering the location, seasonal timing, and magnitude of prey concentrations relative to pre-drainage colony locations (Ogden 2005). Wading birds have reacted to such changes by nesting in smaller colonies in the interior Everglades. In response to the loss of ecological productivity, the Comprehensive Everglades

Restoration Plan (CERP) was enacted in 2000 to restore portions of the Everglades. The

CERP designated wading birds as a defining feature of the pre-drainage system because they clearly respond to the seasonal pulses of secondary productivity provided by the

Everglades ecosystem (RECOVER 2006).

The Florida Everglades is a dynamic subtropical wetland subject to rapid seasonal spatial and temporal resource pulses (Ogden 2005). In this shallow wetland system, length of inundation over periods of months and years increases the density of the prey base (Loftus and Eklund 1994, Trexler 2010), whereas dry season drying and ponding of water over periods of weeks concentrates prey and is linked to wading bird foraging density (Russell et al. 2002). Optimal management of the Everglades system builds low trophic level productivity through ponding of wet season rainfall and releases the concentrated productivity to higher-level trophic predators through dry season water recession. However, wading bird populations are affected by climatic variation that 2

causes fluctuations in the production and availability of prey (Gawlik 2002), and from management strategies to protect human and ecological interests from flooding and droughts that constrain these pulses (Light and Dineen 1994). Thus, there exists an ecological trade-off between flooding of the Everglades to increase prey production and drying of the Everglades to increase prey availability. As drying resets production and alters prey community dynamics, these pulses of productivity are typically released over multi-annual cycles.

Changes to these cycles and the disproportional loss of short hydroperiod wetlands have disparately affected wading bird species with a more constrained niche

(i.e., specialists; Herring et al. 2010, Beerens et al. 2011). Populations of wading bird species that are tactile foragers and require higher prey concentration (e.g., White Ibis and Wood Stork) have disproportionally decreased from the 1930s to 2001 across the

Everglades when compared with populations of visual foragers that favor deeper water

(e.g., Great Egret; Crozier and Gawlik 2003). In addition, the White Ibis and Wood

Stork, while similar in foraging strategy, differ in other traits such as prey size selection, foraging flight distance, nest initiation date, and nest cycle length (Frederick and Ogden

1997) and thus may serve unique functions as indicators.

As highly mobile top predators, wading bird populations integrate productivity across trophic levels and over a large landscape scale (Frederick et al. 2009). Changes in their distributions and populations have served as an initial indicator of the consequences of habitat alteration. In a similar manner, models of wading bird-habitat relationships can predict changes that result from habitat restoration. However, understanding and predicting wading bird responses to restoration are hindered by their complex behavior. 3

Wading birds have a proximate adaptive response to poor foraging conditions that is difficult to capture over long-term periods with traditional habitat selection models

(Beerens et al. 2011). When wet season conditions result in less prey production, individuals select foraging habitat with a more rapid drying process to mitigate the loss of productive foraging habitat. In contrast, individuals select for higher prey production

(but not the drying process) with better wet season conditions (Beerens et al. 2011).

Clearly, long- and short-term processes both limit prey availability to wading birds, creating a hierarchy of nested temporal landscape processes. Wading birds also depend on habitat characteristics at multiple spatial scales incorporating individual patches and their spatial distribution. Understanding the interaction between these spatio-temporal processes and the response of species distributions requires a dataset that covers a large spatial extent with varying combinations of resource levels over multiple years.

From 1985-2012, a standard protocol of Systematic Reconnaissance Flights (SRF) has been used to document the monthly dry season abundance, flock composition, and distribution of foraging wading birds across the Greater Everglades system, specifically in the Water Conservation Areas, Big Cypress National Park and Everglades National

Park (Bancroft et al. 1994, Russell et al. 2002). This long-term and comprehensive dataset is ideal for quantifying the wading bird response to environmental conditions because it spans the extant Everglades and represents a decadal gradient of hydrological variation. SRF data were linked with daily hydrological variables calculated from a 400- m grid network of water depths generated by the Everglades Depth Estimation Network

(EDEN; Telis 2006). EDEN is a vital tool for creating high resolution predictors at

4

multiple spatio-temporal scales, thus representing the hierarchy of processes that provide resources to foraging wading birds.

This dissertation examines the habitat preferences of wading birds using SRF and

EDEN data to predict how changes from restoration and climate change in the Everglades will impact wading bird distributions and population trends. Long-term observational records combined with environmental predictors of high temporal resolution were used specifically to capture behavioral plasticity in response to environmental gradients.

Chapter 2 quantifies the water depth and recession ranges of Great Egrets, White Ibises, and Wood Storks to compare to ranges in published reports. The water depth ranges were also used to define the area of Everglades landscape available for foraging to develop species distribution models (SDM) in Chapter 3.

Chapter 3 describes the Wading Bird Distribution Evaluation Models (WADEM), developed as a framework for isolating and modeling the spatial and temporal responses of wading bird distributions over a decade of environmental variation. A spatial foraging conditions model (SFC) predicted wading bird abundance over time at a fixed 400-m spatial scale and a temporal foraging conditions model (TFC) predicted abundance across space at a fixed daily temporal scale. These two approaches were necessary because wading bird occurrence patterns are highly variable in time and space as individuals track changing seasonal and annual locations of high quality foraging patches. In addition, the indices resulting from the two models represent proxies for different components of patch dynamics. Patch quality within suitable depths was reflected by TFC and landscape patch abundance by SFC. The product of these two indices (quality × area; or foraging index [FI]) provides a metric to account for both processes. WADEM also represents 5

conspecific attraction (i.e., aggregation) by modeling the ratio of the number of individuals per flock to indicate clustering trends throughout the dry season.

In the Everglades, it is important to establish a direct relationship between shifting patterns of observed use and measures of fitness because management recommendations from habitat selection studies guide long-term (30 yr) and large-scale (cf 4,000 km2) restoration projects. Many studies report selection and preference from observations of use (Chapter 3), but few have linked selection of specific resources to fitness (Whitham

1980, Hollander et al. 2011), especially in dynamic environments where resources change daily. Chapter 4 links SDM output from WADEM to annual nesting effort and success

(i.e., fitness) over a period of 21 and 13 years, respectively. Specifically, these models determined key periods of the breeding cycle by identifying when foraging conditions from WADEM most affected nesting effort and success. This is critical because a shift in the timing of good foraging conditions has been linked to delayed nest initiations and resulted in poorer nesting success by shortening the nesting cycle (Ogden 1994, Ogden

2006). Dispersion of foraging individuals was also an important consideration, because high patch quality might result in strong clustering of individuals (individuals/flock) in preferred habitat, whereas large numbers of patches might result in a low ratio. This clustering is likely a better representation of habitat quality than abundance alone for less social species, such as the Great Egret.

This dissertation also describes the multiple applications to which this framework can be applied: to evaluate proposed restoration alternatives for the Central Everglades

Planning Project (CEPP; Chapter 5), and examine potential changes in wading bird habitat under climate change scenarios (Chapter 6). 6

LITERATURE CITED

Bancroft, G. T., A. M. Stong, 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 Everglades, the Ecosystem and its

Restoration (S. M. Davis and J. C. Ogden, Eds.). St. Lucie Press, Delray Beach,

FL.

Beerens, J. M., D. E. Gawlik, G. Herring, and M. I. Cook. 2011. Dynamic habitat

selection by two wading bird species with divergent foraging strategies in a

seasonally fluctuating wetland. The Auk 128:651–662.

Crozier, G. E., and D. E. Gawlik. 2003. Wading bird nesting effort as an index to wetland

ecosystem integrity. Waterbirds 26: 303–324.

Fleming, D., W. Wolff, and D. DeAngelis. 1994. Importance of landscape heterogeneity

to wood storks in Florida Everglades. Environmental Management 18:743–757.

Frederick, P. C., D. E. Gawlik, J. C. Ogden, M. I. Cook, and M. Lusk. 2009. The White

Ibis and Wood Stork as indicators for restoration of the everglades ecosystem.

Ecological Indicators 9:S83–S95.

Frederick, P. C., and J. C. Ogden. 1997. Philopatry and nomadism : Contrasting long-

term movement behavior and population dynamics of White Ibises and Wood

Storks. Colonial Waterbirds 20:316–323.

7

Frederick, P. C., and M. G. Spalding. 1994. Factors affecting reproductive success of

wading birds (Ciconiiformes) in the Everglades ecosystem. Pages 659–691 in

Everglades, the Ecosystem and its Restoration (S. M. Davis and J. C. Ogden,

Eds.). St. Lucie Press, Delray Beach, FL.

Gawlik, D. E. 2002. The effects of prey availability on the numerical response of wading

birds. Ecological Monographs 72:329–346.

Herring, G., D. E. Gawlik, M. I. Cook, and J. M. Beerens. 2010. Sensitivity of nesting

Great Egrets (Ardea alba) and White Ibises (Eudocimus albus) to reduced prey

availability. The Auk 127:660–670.

Hollander, F. a, H. Van Dyck, G. San Martin, and N. Titeux. 2011. Maladaptive habitat

selection of a migratory passerine bird in a human-modified landscape. PloS one

6:e25703.

Light, S. S. and J. W. Dineen. 1994. Water control in the Everglades: A historical

perspective. Pages 47–84 in Everglades: the Ecosystem and its Restoration. S. M.

Davis and J. C. Ogden (Eds.). St. Lucie Press, Delray Beach, Florida.

Loftus, W. F., and A. M. Eklund. 1994. Long-term dynamics of an Everglades small-fish

assemblage. Pages 461–483 in Everglades, the Ecosystem and its Restoration (S.

M. Davis and J. C. Ogden, Eds.). St. Lucie Press, Delray Beach, FL.

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 Everglades, the Ecosystem and its Restoration (S.

M. Davis and J. C. Ogden, Eds.). St. Lucie Press, Delray Beach, FL.

8

Ogden, J. C. 2005. Everglades ridge and slough conceptual ecological model. Wetlands

25:810–820.

RECOVER, 2006. System-wide Performance Measures. REstoration COordination and

VERification Program, c/o Jacksonville District, United States Army Corps of

Engineers, Jacksonville, Florida.

Russell, G. J., O. L. Bass, and S. L. Pimm. 2002. The effect of hydrological patterns and

breeding-season flooding on the numbers and distribution of wading birds in

Everglades National Park. Animal Conservation 5:185–199.

Telis, P. A. 2006. The Everglades Depth Estimation Network (EDEN) for support of

ecological and biological assessments. U.S. Geological Survey Fact Sheet 2006-

3087.

Trexler, J. C. 2010. Greater Everglades aquatic trophic levels DRAFT small-sized

freshwater fish density performance measure. Draft DECOMP Performance

Measure Documentation Sheet submitted to the Army Corp of Engineers.

Whitham, T. 1980. The theory of habitat selection: examined and extended using

Pemphigus aphids. American Naturalist 115:449–466.

9

CHAPTER 2: ANALYSIS OF HABITAT USE

METHODS

Ranges of species-specific habitat use were quantified to compare the differences among ecological boundaries between species and to identify their respective water depth ranges suitable for foraging. Data used to identify ranges of habitat use came from aerial wading bird Systematic Reconnaissance Flights (SRF) that document the abundance, flock composition, and spatiotemporal distribution of foraging wading birds across the

Greater Everglades system (Water Conservation Areas, Big Cypress National Park and

Everglades National Park). The surveys are conducted monthly during most of the dry season through the end of the breeding season (January-June). Flock presence was defined as one or more birds of the target species (Great Egret, White Ibis and Wood

Stork) detected in a given Everglades Depth Estimation Network (EDEN) cell.

SRF data from 2000-2009 were used to quantify species-specific water depth use, recession rate use, and days since drydown (DSD) use. This long-term and comprehensive dataset is ideal for quantifying wading bird habitat use because it spans the extant Everglades and demonstrates the species-specific foraging response to hydrological variation over multiple years. EDEN was used to pair foraging observations

10

from the SRF with depth values and calculated recession rate and DSD values that corresponded to the date and cell of use. Daily recession rate was obtained by subtracting the water depth in a cell on a given day from the water depth two weeks prior and dividing by 14 days (Beerens et al. 2011). The 14-day recession rate variable also predicts dry season prey biomass when incorporated with wet season biomass or days since drydown (Botson and Gawlik 2010). DSD was obtained by counting the number of consecutive days that a cell had a water depth of greater than zero. Water depth use and two-week recession rate use were also summarized by year (from 1 January to the onset of the rainy season) and compared with mean landscape water depths and mean landscape recession rates during the dry seasons of 2000-2009 using linear regression.

For all species, EDEN water depths at foraging observations were classified as

“suboptimal dry” (from the 10% quantile to the 25% quantile), “optimal” (from the 25% quantile to the 75% quantile) and “suboptimal wet” (from the 75% quantile to the 90% quantile). A suitable depth range for Great Egrets, White Ibises, and Wood Storks was established using the 10% to 90% quantiles of depth use (see Beerens and Cook, 2010), which is used to define available recession rates. Species-specific depth ranges were used to classify habitat considered available to foraging birds. Mean available water depth was also summarized within the optimal depth range for each species.

Mean landscape recession rate was summarized for each year and within the optimal depth ranges for each species (i.e., available recession). For all species, recession rates were classified as “suboptimal slow/reversal” (from the mean to the 25% quantile),

“optimal” (from the mean to the 75% quantile) and “suboptimal rapid” (from the 75% quantile to the 90% quantile). This decision to classify recession rates differently from 11

depths was based on several lines of evidence. As receding water functions to provide new foraging patches, I decided to clip the suboptimal range at the 25% quantile, where in all three species the recession rate value was near 0 cm/day. Also, studies indicate that a recession rate of 0.5 cm/day is ideal for wading bird habitat preference (Beerens et al.

2011), nest survival (Herring et al. 2010), and nesting success (Frederick and Collopy

1989). While recession rates greater than 0.5 cm/day may still be beneficial to birds, a landscape drawdown this rapid would likely produce drought conditions, limit foraging opportunities late in the breeding season, and decrease days since drydown in future years

(Herring et al. 2010, Frederick et al. 2009). The 75% quantile in all three species corresponded to ~0.5 cm/day; therefore, the mean to75% quantile range was used to reflect the range of optimal recession rates.

RESULTS AND DISCUSSION

Depth

Great Egrets

The wettest dry season of the decade was in 2003 (mean = 19.88 cm ± 0.01 SE) and the driest dry season was in 2001 (mean = -7.63 cm ± 0.01 SE; Fig. 2.1). Great

Egrets used a mean water depth of 19.68 (±0.07 SE) cm. The maximum mean depth use of 26.85 (± 0.24 SE) cm was observed in 2000. The minimum mean depth use of 13.40

(± 0.24 SE) cm was observed the following year. Thus, the range of the means is 13.45 cm. A linear regression shows that as mean dry season landscape depths increase, Great

Egret mean water depth use increases (n = 10, r2 = 0.74, P = 0.001). The optimal range of water depth (25% to 75% quantile) is 8.71 – 31.04 cm. This is the largest range of

12

depths of the three species and supports the Great Egret’s tolerance of a wider range of water depths (Gawlik 2002, Beerens et al. 2011). The suboptimal dry category includes the water depths from -1.78 – 8.71 cm and the suboptimal wet category includes the water depths from 31.04 – 1.63 cm. Thus, the Great Egret depth range used to define available recession rates is -1.78 – 41.63 cm (10% – 90% quantile).

The mean water depth use reported here is within 1.2 cm of the average depth taken from direct observations measured against leg length (Frederick et al. 1996), and

0.3 cm of measured depths at precise locations of feeding Great Egrets obtained from

GPS locations and aerial photographs (Beerens 2008). The optimal range in this study

(8.71 – 31.04 cm) overlaps considerably with a foraging conditions index model for long- legged wading birds (5 – 35 cm; Curnutt et al. 2000). In a set of regression models,

Bancroft et al. (2002) reported a quadratic relationship between the predicted number of birds per cell and water depth in an average year. In the following dry year, the predicted number of birds per cell increased with increasing water depth because the deeper threshold of depths was not observed in the landscape. In both years, predicted use was highest between 0 – 50 cm, slightly broader than the range reported here. However, including suboptimal categories (-1.78 – 41.63 cm) yields a closer approximation to values reported by Bancroft et al. (2002). Similarly, Frederick et al. (2009) reported a deep water range of 25 – 45 cm for Great Egrets. A study of radio-tagged Great Egrets in

2006 revealed 99% of foraging observations were in water depths (EDEN) less than

44.34 cm, and measured depths at a subset of observations reveal a deep water threshold of 49.50 cm (Beerens 2008).

13

White Ibises

White Ibises used a mean water depth of 13.40 cm (±0.09 SE) during 2000-2009.

The maximum mean depth use of 16.78 cm (±0.29 SE) was observed in 2004, whereas the minimum mean depth use of 7.17 cm (±0.50 SE) was observed in 2009. The range of the means is 9.61 cm. A linear regression shows that as mean dry season landscape depths increase, White Ibis water depth use does not significantly increase (n = 10, r2 =

0.31, P = 0.10). The optimal range of water depth (25% to 75% quantile) is 3.87 – 22.82 cm. The suboptimal dry category includes the water depths from -4.95 – 3.87 cm and the suboptimal wet category includes the water depths from 22.82 – 32.62 cm. Thus, the depth range for the White Ibis used to define available recession rates is -4.95 – 32.62 cm

(10% – 90% quantile).

The mean water depth use reported here is within 1 cm of the average depth taken from direct observations measured against leg length (Frederick et al. 1996) and within

0.8 cm of measured depths at precise locations of feeding White Ibises obtained from

GPS and aerial photographs (Beerens 2008). The response curves to water depth (see

Bancroft et al. 2002) were similar to the Great Egret; however, the White Ibis depth range was more restricted with the predicted numbers of birds per cell highest between 0 -25 cm (compared to 0-50 cm for Great Egrets). Optimal depths for White Ibis in this study

(3.87 – 22.82 cm) fall within that range and are similar to the range used by Curnutt et al.

(2000) in a foraging conditions index model for short-legged wading birds (0 – 20 cm).

Functions from Gawlik et al. (2004) show the highest suitability index between 0.0 –

15.25 cm and a declining suitability index through 24.38 cm. Interestingly, there is also a declining index from 0.0 cm to -9.14 cm (Gawlik et al. 2004). Use of cells with negative 14

modeled depths is supported by results from this study (suboptimal dry; -4.95 – 3.87 cm).

In 2007, 90% of radio-tagged White Ibis used EDEN depths greater than -14.50 cm

(Beerens 2008); nonetheless, measured depths from a subset of those observations do not show depth use less than 5.0 cm, the lower threshold reported by Kushlan (1979, 1986) and Frederick et al. (2009). Thus, water depth use is scale dependent contingent on whether depth is obtained from a model at a cell scale (400-m) or measured at the local scale. A similar relationship exists for deep water depths; in this study, 15% of water depth use was recorded from 22.82 – 32.62 cm. Similarly, 15% of radio-tagged White

Ibis locations were in depths from 19.19 – 29.86 cm (Beerens et al. 2011). To date, depth use greater than 30 cm has not been reported in the literature; however, it appears that

White Ibis are still able to find suitable water depths within an EDEN cell up to 33 cm.

The use of both negative and deep water depths (i.e., broader range) helps explain why modeled and measured mean depths are still similar.

Wood Storks

Wood Storks used a mean water depth of 12.86 (±0.29 SE) cm. The maximum mean depth use of 22.44 cm (±0.90 SE) was observed in 2004, an average dry season

(Fig. 2.1). The minimum mean depth use of 5.60 cm (±0.77 SE) was observed in 2008, also an average dry season (Fig. 2.1). The difference translates to a mean range of 16.84 cm. A linear regression shows a significant relationship between Wood Stork depth use and mean dry season landscape depth (n = 10, r2 = 0.42, P = 0.04). The optimal range of water depth (25% - 75% quantile) is 2.75 – 24.21 cm. The suboptimal dry category includes the water depths from -8.73 – 2.75 cm and the suboptimal wet category includes

15

the water depths from 24.21 – 35.77 cm. Thus, the depth range for the Wood Stork used to define available recession rates is -8.73 – 35.77 cm (10% – 90% quantile).

The mean water depth use reported here is within 11.4 cm of the average depth taken from direct observations measured against leg length (Frederick et al. 1996), the greatest disparity among the three species. Herring and Gawlik (2011) report a mean

EDEN depth use of 7.93 cm ± 20.59 SD in 2006. The response curves to water depth

(see Bancroft et al. 2002) were similar to the White Ibis in form and depth range – the predicted numbers of birds per cell was highest between 0 -25 cm. Optimal depths for

Wood Storks in this study (2.75 – 24.21 cm) approximate that range. Kahl (1964) reports depth use from 15 – 50 cm, whereas Kushlan (1986) reports a maximum depth use of 40 cm. Results from this study show 90% of Wood Stork observations were observed in an

EDEN depth less than 35.77 cm. The water depth suitability index from Gawlik et al.

(2004) begins increasing from zero at a depth of -9.14 cm, and is concurrent with validation results (8.73 – 2.75 cm; suboptimal dry).

Recession Rate

Great Egrets

The dry season of 2009 had the most rapid mean recession rate (0.46 cm/day), whereas the dry season of 2001 had the slowest mean recession rate (-0.02 cm/day).

Great Egrets used a mean recession rate of 0.20 cm/day. The maximum mean recession rate use of 0.48 cm/day was observed in 2009, a year with the highest recession rate within depths considered suitable for Great Egrets (0.46 cm/day) and record nest effort

(Cook and Kobza 2009). Similar to the linear water depth response, an increase in recession rate use corresponded with an increase in available recession rates (10% - 90% 16

quantiles of use; n = 10, r2 = 0.86, P < 0.001). The optimal range of recession rates

(mean – 75% quantile) is 0.20 – 0.45 cm/day. The suboptimal slow/reversal category includes recession rates from -0.01 – 0.20 cm/day and the suboptimal rapid category includes recession rates from 0.45 – 0.66 cm/day.

Two prior studies (Russell et al. 2002 and Beerens et al. 2011) found a positive linear response between Great Egret foraging and increasing recession rate, although these estimates were computed differently. Russell et al. (2002) used the slope of a fitted regression of dry season water depths, whereas Beerens et al. (2011) used the mean water depth change in a cell over a 14-day period. In a foraging conditions index model for long-legged wading birds, Curnutt et al. (2000) specified that for a cell to be suitable there must be a daily falling hydrograph for a 90-day cycle. The response of radio-tagged

Great Egret and White Ibis to receding water in 2006 and 2007 (Beerens et al. 2011) was used to develop an optimal recession index for the South Florida Water Management

District. White Ibises used the lowest recession rates (0.25 cm/day) in 2006 (good year) and Great Egrets used the highest recession rates (0.58 cm/day) in 2007 (poor year). The

95% confidence intervals were then incorporated to yield a range of 0.19 – 0.66 cm/day.

Compared to results from this study, the SFWMD range is closely aligned with the combined optimal and suboptimal rapid category for Great Egrets (0.20 – 0.66 cm/day).

In contrast, Herring et al. (2010) reports a quadratic relationship between Great Egret daily nest survival and recession rate, with highest daily nest survival occurring at a recession rate of 0.5 cm/day; however, recession rate was calculated as the mean water depth change in a cell over a 7-day period. It was a goal of the current study to determine the recession rate time interval that best predicts the Great Egret foraging response, as 17

differing response variables may correspond to different recession time intervals. In addition, consideration was given to modeling recession as a conditional response of antecedent hydrology, rather than solely a static range.

White Ibis

White Ibises used a mean recession rate of 0.21 cm/day. The maximum mean recession rate used (0.50 cm/day) was observed in 2009, a year with the highest recession rates within depths considered suitable for White Ibises (0.57 cm/day) and record nest effort. The relationship between recession rate use and available recession rates was significant (n = 10, r2 = 0.89, P < 0.001). The optimal range of recession rates (mean –

75% quantile) is 0.21 – 0.47 cm/day. The suboptimal slow/reversal category includes recession rates from -0.01 – 0.21 cm/day and the suboptimal rapid category includes recession rates from 0.47 – 0.72 cm/day.

Russell et al. (2002) reports recession rate as the best predictor of White Ibis SRF abundance in May. Beerens et al. (2011) showed that the White Ibis selected high recession rates only in a year with low prey availability (i.e., recession selectivity model) and that the response was weaker than for Great Egrets. Kushlan (1986) found that recession rate had no effect on the distribution of White Ibis foraging, whereas Curnutt et al. (2000) states that there must be a falling hydrograph for a site to be suitable; however, the rate is not specified. Gawlik et al. (2004) built a recession rate habitat suitability index for the White Ibis, Wood Stork, and Snowy Egret based on the positive relationship between recession rate and the timing and numbers of White Ibis nesting attempts

(Frederick and Spaulding 1994). The optimal range is reported as 0.22 – 0.70 cm/day with a sharply declining index from 0.70 – 2.61 cm/day and a slowly declining index 18

from -0.17 – 0.22 cm/day. There is strong agreement between the Gawlik et al. (2004) range, the SFWMD index for Great Egrets and White Ibises (0.19 – 0.66 cm/day), and validation results from this study (0.21 - 0.72 cm/day; optimal and suboptimal rapid).

Wood Stork

Wood Storks used a mean recession rate of 0.27 cm/day. The maximum mean recession rate use (0.51 cm/day) was observed in 2009, a year with the highest recession rate within depths considered suitable for Wood Storks (0.57 cm/day) and record nesting effort. Increased recession rate use was positively correlated with available recession rates (n = 10, r2 = 0.39, P = 0.05). The optimal range of recession rates (mean – 75% quantile) is 0.27 – 0.46 cm/day. The suboptimal slow/reversal category includes recession rates from 0.08 – 0.27 cm/day and the suboptimal rapid category includes recession rates from 0.46 – 0.70 cm/day. Wood Storks had the lowest use of rising water (recession rate

< 0 cm/day), supporting the idea that the Wood Stork is the most sensitive to reversals

(Frederick et al. 2009). Furthermore, Wood Storks had low use of sites that had dried and rewet within a dry season (14%), compared with White Ibises (18%).

Russell (2002) reports the linear form of recession rate to be the best predictor of

SRF Wood Stork abundance in May. Similarly, Kushlan (1986) describes recession rate as being “critical for Wood Storks, although not for White Ibis.” Wood Storks that were followed in 2006 used locations with similar recession rates (0.36 cm/day) to those available (0.41 cm/day) and therefore no selection was inferred (Herring and Gawlik

2011). Indeed, in 2006 the validation shows that Wood Storks were using cells with recession rates (0.21 cm/day) similar to those available in the landscape (0.19 cm/day).

Of the three species, the Wood Stork is the only species with over 75% of observations in 19

receding water. The combined optimal and suboptimal rapid ranges (0.27 – 0.70 cm/day) are in close proximity to the optimal range reported by Gawlik et al. (2004; 0.22 – 0.70 cm/day).

Prey Production and Days since Drydown

Several metrics have been used in the literature as proxies for prey production.

Curnutt et al. (2000) utilized a function for fish density based on the previous year’s water depths for long and short-legged wading birds. Both Trexler et al. (2009) and

Botson and Gawlik (2010) have emphasized days since drydown as an important variable in the production of wet season prey biomass and availability of dry season prey biomass, respectively. A three-year running summary of days since drydown and the ten-year mean hydroperiod was calculated by Beerens et al. (2011) to determine Great Egret and

White Ibis habitat selection. Both species selected cells with higher days since drydown than those available in the landscape in the combined year models. The validation suggests White Ibis and Wood Storks have similar responses to days since drydown, using cells that had been wet for an average of 372 (± 1.8 SE) and 372 (± 4.6 SE) days, respectively. Great Egrets, however, use cells with a higher days since drydown average

(406 days ± 1.2 SE).

20

LITERATURE CITED

Bancroft, G. T., D. E. Gawlik, and K. Rutchey. 2002. Distribution of wading dirds

relative to vegetation and water depths in the northern Everglades of Florida,

USA. Waterbirds 25:265-391.

Beerens, J. M., D. E. Gawlik, G. Herring, and M. I. Cook. 2011. Dynamic habitat

selection by two wading bird species with divergent foraging strategies in a

seasonally fluctuating wetland. The Auk 128:651-662.

Beerens, J. M. and M. Cook. 2010. Using Wood Stork distribution data to develop water

management guidelines. Appendix B in USFWS Multi-Species Transition

Strategy for Water Conservation Area 3A, U.S. Fish and Wildlife Services, South

Florida Ecosystem Services Office, Vero Beach FL.

Beerens, J. M. 2008. Hierarchical resource selection and movements of two wading Bird

species with divergent foraging strategies in the Florida Everglades. M.S. Thesis,

Florida Atlantic University, Boca Raton.

Botson, B. and D. E. Gawlik. 2010. Aquatic fauna seasonal concentrations. Final Report

for the South Florida Water Management District. Florida Atlantic University

Cook, M. I. and R. M. Kobza. 2009. South Florida Wading Bird Report, vol. 14. South

Florida Water Management District, West Palm Beach.

21

Curnutt, J. L., J. Comiskey, M. P. Nott, and L. J. Gross. 2000. Landscape-based spatially

explicit index models for Everglades restoration. Ecological Applications

10:1849-1860.

Frederick, P., D. E. Gawlik, J. C. Ogden, M. I. Cook, and M. Lusk. 2009. The White Ibis

and Wood Stork as indicators for restoration of the Everglades ecosystem.

Ecological Indicators 9S:S83-S95.

Frederick, P. C., J. Salatas, and J. Surdick. 1996. Monitoring and research on wading

birds in the Water Conservation Area of the Everglades: The 1996 nesting season.

Final Report for U.S. Corps of Engineers. University of Florida.

Frederick, P. C. and M. G. Spalding. 1994. Factors affecting reproductive success of

wading birds (Ciconiiformes) in the Everglades ecosystem. Pages 659-692 in

Everglades, the Ecosystem and its Restoration (S. M. Davis and J. C. Ogden,

Eds.). St. Lucie Press, Delray Beach, FL.

Frederick, P. C. and M. W. Collopy. 1989. Nesting success of five Ciconiiform species in

relation To water conditions in the Florida Everglades. The Auk 106:625–634.

Gawlik, D. E., G. Crozier, and K. H. Tarboton. 2004. Wading bird habitat suitability

index. Pages 111-127 in K. C. Tarboton, M. M. Irizarry-Ortiz, D. P. Loucks, S.

M. Davis, and J. T. Obeysekera. Habitat suitability indices for evaluation water

management alternatives. Technical Report, South Florida Water Management

District, West Palm Beach, FL.

Gawlik, D. E. 2002. The effects of prey availability on the numerical response of wading

birds. Ecological Monographs 72:329–346.

22

Herring, G., D. E. Gawlik, M. I. Cook, and J. M. Beerens. 2010. Sensitivity of nesting

Great Egrets (Ardea Alba) and White Ibises (Eudocimus Albus) to reduced prey

availability. The Auk 127:660-670.

Herring, H. K. and D. E. Gawlik. 2011. Resource selection functions for Wood Stork

foraging habitat in Everglades National Park. Waterbirds 34:133-142.

Kahl, M. P., Jr. 1964. Food ecology of the Wood Stork (Mycteria Americana) in Florida.

Ecological Monographs 34:97–117.

Kushlan, J. A. 1986. Response of wading Birds to seasonally fluctuating water levels:

Strategies and their limits. Colonial Waterbirds 9:155–162.

Kushlan, J. 1976. Site selection for nesting colonies by the American White Ibis

Eudocimus albus in Florida. Ibis 118:590–593.

Russell, G. J., O. L. Bass, Jr., and S. L. Pimm. 2002. The effect of hydrological patterns

and breeding-season flooding on the numbers and distribution of wading birds in

Everglades National Park. Animal Conservation 5:185–199.

Trexler, J. C. and C. W. Goss. 2009. Aquatic fauna as indicators for Everglades

restoration: Applying dynamic targets in assessments. Ecological Indicators

9S:S108-S119.

23

Figure 2.1. Mean Everglades Depth Estimation Network (EDEN) Dry Season Water

Depth (±SE) for 2000-2009.

24

CHAPTER 3: MODELING SPATIO-TEMPORAL RESPONSES OF WADING BIRD

INDICATOR SPECIES ACROSS RESOURCE GRADIENTS FOR EVERGLADES

RESTORATION

ABSTRACT

Species distribution models (SDM) link species occurrence with a suite of environmental predictors and provide an estimate of habitat quality when the variable set captures the biological requirements of the species. SDMs are inherently more complex when they include components of a species’ ecology such as conspecific attraction and behavioral flexibility to exploit dynamic spatio-temporal resources. Wading birds are highly mobile, demonstrate flexible habitat selection, and respond quickly to changes in habitat quality; thus serving as important indicator species for the Everglades. I developed a spatio-temporal, multi-SDM framework using Great Egret (Ardea alba),

White Ibis (Eudocimus albus), and Wood Stork (Mycteria americana) distributions over a decadal gradient of environmental conditions to predict species-specific abundance over space and spatial occurrence over time. Models jointly accounted for flexible habitat selection of resources within and among temporal scales, responses to environmental gradients, conspecific attraction, and spatial autocorrelation. In temporal models, species

25

demonstrated conditional preferences for resources based on resource levels at differing temporal scales. Wading bird abundance was highest when prey production from optimal periods of inundation was concentrated in shallow depths. Similar responses were observed in models predicting spatial occurrence over time, accounting for spatial autocorrelation. Species clustered in response to differing habitat conditions, indicating that social attraction can co-vary with foraging strategy, water-level changes, and habitat quality. This modeling framework can be applied to evaluate the multi-annual resource pulses of real-time conditions, climate change scenarios, or restorative hydrological regimes by tracking changing seasonal and annual distribution and abundance of high quality foraging patches.

INTRODUCTION

Species distribution models (SDM) link species occurrence with a suite of environmental predictors and have a wide range of applications in wildlife science and management by predicting species distributions across landscapes. They provide a powerful tool for land managers when the variable set captures the biological requirements of the species and can be manipulated to account for various management activities (Rittenhouse et al. 2010). The resulting suitability metrics are most relevant to population dynamics in disturbed ecosystems when linked with measures of reproductive success and applied to a suite of restoration alternatives (Garshelis 2000).

The efficacy of a SDM is impacted by the availability and choice of predictors and scale, the modeling method, and the degree of spatial and temporal extrapolation

(Elith and Leathwick 2009). Moreover, many issues remain relatively unexplored in the

26

SDM field concerning how a species’ ecology can affect model building and evaluation such as factors of conspecific attraction, phenotypic and behavioral plasticity, and response to environmental gradients. Recent reviews suggest that ecological theory is rarely considered in SDMs (Austin 2007, Elith and Leathwick 2009, McLoughlin et al.

2010). Habitat preference, for example, has in large part been assumed to remain unaltered as a function of habitat availability, which can change daily in dynamic habitats.

To address dynamic habitat selection, discrete-choice resource selection function

(RSF) models were developed to model a series of choices with a discrete set of resources over time (Arthur et al. 1996). A weakness of this approach is that static resource selection models are still produced and only reflect average preference within the range of habitat conditions encountered during the study period (Beyer et al. 2010).

Therefore, an important consideration must be given to modeling the response curve of habitat selection by quantifying animal use over a wide-range of habitat availabilities

(i.e., the functional response to habitat selection; Mysterud and Ims 1998) and interpreting the results in the context of the species’ ecology (Austin 2007).

While some authors have attempted to capture changing preference in response to the average change in a continuous variable (Gillies et al. 2006, Hebblewhite and Merrill

2008), these measures did not reflect the area of available habitat encountered. In fact, quantifying the functional response to a continuous gradient of available resources has proven difficult (Beyer et al. 2010). Studies have commonly examined an animal’s selection of food resources over a spatial gradient of exposure to predators or anthropogenic disturbance (Godvik et al. 2009, Hebblewhite and Merrill 2008); however, 27

in dynamic ecosystems the gradient could just as easily be across the temporal scale at which prey are produced and become available. For example, food may be produced from longer-term processes that are selected for when available; however, selection for short-term food resources may occur when food production is low. Furthermore, functional responses in habitat selection are expected when multiple limiting resources

(over distinct temporal scales) have been depleted by human activities (Mysterud and Ims

1998, Frid and Dill 2002, Hebblewhite and Merrill 2008). A non-linear functional curve will result when decreased availability of a resource in the landscape results in increased selection for another (Mysterud and Ims 1998).

Animals selecting habitats evaluate the state of a resource that has been subject to scale-dependent spatial and temporal processes (Johnson 1980, Orians and Wittenberger

1991). While the sensitivity of the animal’s response to these different scales is debatable, the perceived costs and benefits shape behavior. An animal’s access to the prey base (i.e., prey availability), can fluctuate in some aquatic ecosystems that are subject to seasonal cycles of drying and flooding (Gawlik 2002, Frederick et al. 2009,

Mazzotti et al. 2009). In these habitats, distinct temporal processes regulate the spatial immigration and population dynamics of the prey base over long periods of inundation

(i.e., prey production; Loftus and Eklund 1994) versus the spatial and temporal concentration of prey animals in intermediate depths over the shorter period of drying

(i.e., prey concentration; (Russell et al. 2002). When wet season conditions result in less prey production, mobile predator species have shown selection for the shorter-term drying process to mitigate the loss of productive foraging habitat (Beerens et al. 2011).

In this case, selection increases for one process when another is limiting. These two 28

identifiable and spatially segregated resources (i.e., prey production and prey concentration) can also converge to create high quality habitat depending on spatio- temporal landscape conditions. Thus, animals also depend on the habitat characteristics at multiple spatial scales; incorporating individual patches and their spatial distribution.

A challenge in interpreting the results of spatial SDMs can occur when spatial autocorrelation is prevalent in that resources closer in distance are more similar. Spatial models that incorporate spatial autocorrelation are more likely to avoid consequences of pseudoreplication and represent biological processes unaccounted for by the environmental covariates (Dormann 2007). Therefore, selecting the appropriate scale (or observational unit) is essential to matching ecological trade-offs with their respective spatial and temporal processes.

The Florida Everglades is a dynamic subtropical wetland subject to seasonal resource pulses. A decline in the productivity generated by these pulses, across all trophic levels, resulted from historic water management practices and loss of spatial extent (Kahl 1964, Ogden 1994). The desired condition of Everglades’ restoration is the reintroduction of historic water flow patterns to recover and sustain the defining characteristics of the greater Everglades (RECOVER 2004). Principally, historical hydrology has been associated with an abundant and stable population of wading birds, important indicator species for the Greater Everglades ecosystem (Frederick et al. 2009).

Because wading birds are highly mobile top predators, their populations integrate productivity across trophic levels and over a large landscape scale. However, understanding and predicting wading bird responses to environmental restoration are hindered by their changing preference for hydrological characteristics depending on the 29

production and concentration of prey, which occur over different temporal scales.

Multiple hydrological variables influence habitat selection and conspecific attraction that can oscillate in strength depending on resource availability and foraging strategy of wading birds (Gawlik and Crozier 2007, Beerens et al. 2011, Lantz et al. 2011).

Limitations in the spatial extent of the Everglades from both cumulative losses of short- hydroperiod wetland and ponding in long hydroperiod wetland now restrict the spatio- temporal availability of high quality patches to a narrow range of hydrological conditions

(Gawlik 2002, Ogden 2005). Further, changes in timing within this narrow range have altered pre-breeding physiology, nesting initiations, chick provisioning, and nesting success (Fleming et al. 1994, Frederick and Spalding 1994, Herring et al. 2010). An abundant and spatio-temporally available prey base is required to support a large breeding population of wading birds through the end of the dry season when chicks fledge.

Additionally, increasing evidence suggests that changes in long-term habitat quality and prey availability have disparately affected wading bird species with a more constrained niche (i.e., specialists; Herring et al. 2010, Beerens et al. 2011). Across the

Everglades, populations of wading bird species that require higher prey concentration, such as tactile foragers (i.e., White Ibis and Wood Stork; hereafter ibises and storks), have disproportionally decreased from the 1930s to 2001 when compared with populations of visual foragers that favor relatively deeper water (i.e., Great Egret; hereafter egrets). This pattern likely indicates an overall decline in prey availability

(Frederick et al. 2009). In addition, the ibis and stork, while similar in foraging strategy, differ in other traits such as prey size selection, foraging flight distance, nest initiation 30

date, and nest cycle length (Frederick and Ogden 1997) and thus may serve unique functions as indicators.

Focusing on three representative species of wading birds (egrets, ibises, and storks), I developed a spatio-temporal modeling framework utilizing long-term observational records combined with high temporal-resolution environmental predictors.

This approach served three purposes: (1) to capture changing preference of wading birds across resource gradients, (2) to determine the effect of conspecific attraction on foraging distributions, and (3) to evaluate species abundance in terms of the ecological trade-offs associated with the quantity, timing, and distribution of water. Using a multi-model approach, I aggregated wading bird foraging distributions over space and then time to reduce the noise associated with individual locations and isolate variables occurring at independent spatio-temporal scales. A spatial foraging conditions model (SFC) examined spatial and hydrological dynamics at a fixed spatial scale (i.e., cell) to predict wading bird frequency of use over time, whereas a temporal foraging conditions model

(TFC) predicted flock and individual abundance across the landscape from daily selection of three distinct temporal processes (i.e., available resources) at a fixed temporal scale.

Indices resulting from the two models represent proxies for different components of patch dynamics. The TFC index represents conditions within a suitable set of depths that change daily and reflects patch quality, whereas the SFC index is a spatial representation of the suitability of all cells and reflects landscape patch abundance. In a seasonal wetland, I expect patch abundance to increase to a maximum when the greatest area is within a species’ suitable depth range (i.e., high heterogeneity), and decrease as the landscape dries. In contrast, I expect patch quality to continually increase as patches with 31

longer hydroperiods, and thus higher prey density, become available within suitable depths. The product of these two indices (area × quality; or foraging index [FI]) provides a metric to account for both processes.

METHODS

Data and Variables

Egret, ibis, and stork locality data were produced from Systematic

Reconnaissance Flights (SRF) that document the abundance, flock composition, and spatio-temporal distribution of foraging wading birds across the Greater Everglades system (Water Conservation Areas, Big Cypress National Park and Everglades National

Park; Fig. 3.1). From January–June, during low altitude (61m) flights, observers estimated numbers and species of birds in belt transects spaced at 2-km intervals

(Bancroft et al. 1994). This long-term and comprehensive dataset was ideal for quantifying habitat use by wading birds because it spans the extant Everglades and demonstrates the species-specific foraging response to hydrological variation over multiple years.

SRF data from 2000 – 2009 were overlaid on water depth data for the same dates that were derived from the Everglades Depth Estimation Network (EDEN). The EDEN is a nearly real-time hydrological model that calculates daily water depth (within ± 5 cm) in 400 m × 400 m grid cells accounting for evapotranspiration, rainfall, and sheet flow

(Liu et al. 2009). From EDEN, hydrological variables over multiple temporal scales and with existing links to wading bird responses were calculated as proxies for landscape processes that influence prey availability (i.e., resources). Days since drydown (DSD)

32

quantified dynamics of long-term prey production by counting the number of consecutive days over 3 years that a cell had a water depth of greater than zero. DSD is an important variable that accounts for prey biomass production in the wet season and availability in the dry season (Trexler and Goss 2009, Botson and Gawlik 2010). Furthermore, egrets and ibises selected for DSD in a year with good, but not poor, habitat conditions indicating it was likely a limiting resource (Beerens et al. 2011). Recession rate quantified dynamics of 2-week prey concentration by subtracting the water depth in a cell on a given day from the water depth 2 weeks prior and dividing by 14 days. The 2-week recession rate variable also predicts dry season biomass when incorporated with wet season biomass or days since drydown (Botson and Gawlik 2010). Described as the recession selectivity model, daily recession rate was selected by egrets and ibises in a year with poor habitat conditions, an adaptation to access new patches when prey was limiting (Beerens et al. 2011). Daily water depth quantified short-term availability of prey. Water depth is an important predictor of differential habitat selection in wading birds (Gawlik 2002) and of relative reproductive success among species (Frederick and

Spalding 1994, Herring et al. 2010). Using RSFs, the selection of water depths has been described for all species in this study (Herring et al. 2010, Beerens et al. 2011). I also calculated a dry to wet reversal variable to estimate the percentage of the availability window that had dried below the 10% quantile of depth use and rewet for each species.

This variable was necessary in predicting species distributions because cells could return to the suitable depth range after going dry, but would have severely depleted prey populations (Trexler et al. 2002). Hydroperiod was defined as the mean number of days per year (during the study period) that water depth of an EDEN cell was greater than 33

zero. In addition to affecting fish density, hydroperiod affects wading bird distributions by influencing long-term changes in microtopography and vegetation communities

(Gunderson 1994).

Identifying Available Habitat

Available habitat was quantified by identifying species-specific water depth ranges suitable for foraging. For egrets, ibises and storks, EDEN water depths at foraging observations were classified as “suboptimal dry” (from the 10% quantile to the

25% quantile), “optimal” (from the 25% quantile to the 75% quantile) and “suboptimal wet” (from the 75% quantile to the 90% quantile). The 10% to 90% quantiles of suitable depth use (e.g., Beerens and Cook 2010) were used to classify available habitat and average resource levels considered available to foraging birds for the TFC model.

Temporal Foraging Conditions (TFC; Patch Quality)

The SRF surveys from days in Jan – May, 2000-2009 (n=243) were used to determine the daily abundance of foraging flocks and total individuals in the Everglades landscape. Flock presence was defined as one or more birds of the target species (egret, ibis and stork) detected in a given EDEN cell, whereas abundance of individual species considered the total number of birds present. For each species within available habitat of the daily SRF survey extent, I calculated the daily average of three temporally-specific

EDEN hydrological variables (depth, recession rate, and days since drydown) as a proxy for prey dynamics. These resource levels and their spatial heterogeneity (SD) were employed to predict values of resource use by wading birds (i.e., selection), which were then used to predict daily flock and individual abundance.

34

Mean daily use values by wading birds were plotted as a function of mean daily availability of resources to determine the species’ selection response to a gradient of their environmental conditions. In addition, an a priori set of candidate models was tested in

Proc Mixed (SAS Institute 2010a) to determine if other hydrological variables were contributing to depth, recession and DSD use. Models consisting of unique combinations of terms (available Depth, Depth SD, Depth2; available Recession, Recession SD,

Recession2; available DSD, DSD SD, DSD2), the interactions between the three resource terms, and interaction between depth availability and depth use (i.e., depth selection; for

DSD use only) were evaluated for parsimony using Akaike’s Information Criterion for small sample sizes (AICc; Burnham and Anderson 2002). The random term Month was included to determine temporal effects of resource selection independent of hydrology.

Models were retained if ΔAICc < 4 (Burnham and Anderson 2002). Model-averaged coefficients and standard errors (SE) were calculated for each fixed and random parameter by averaging all models containing the variable in proportion to the AIC weight. The importance of a specific term was determined by summing the weights of all models containing that term.

The daily abundance of flocks is correlated to the abundance of wading birds utilizing the landscape on any given day (mean R2 = 0.77). Both individual and flock responses were modeled because wading birds are highly social and select foraging habitat based in part on the presence of conspecifics, a process that may increase or decrease individual fitness (Campomizzi et al. 2008). Data for individuals and flocks across the landscape were all fourth-root transformed to meet assumptions of normality, with the exception of stork flocks, which were inverse fourth-root transformed because of 35

the high occurrence of single flocks within SRF surveys. An a priori set of candidate models was tested in Proc Mixed (SAS Institute 2010a) to determine how hydrological variables used by wading birds and the month of the dry season contributed to the flock and individual response across the landscape. Models consisting of unique combinations of terms (Depth use, Depth use2, available Depth SD; Recession use, Recession use2, available Recession SD; DSD use, DSD use2, available DSD SD; available dry to wet

Reversal), the interactions between the three resources, and interaction between depth availability and depth use were evaluated for parsimony using Akaike’s Information

Criterion for small sample sizes (AICc; Burnham and Anderson 2002). Models were evaluated similar to analyses mentioned above.

Spatial Foraging Conditions (SFC; Patch Abundance)

Throughout the greater Everglades, SRF data from 2000-2009 were used to identify the number of instances over the period of record that a species used a given cell

(i.e., frequency of use). Water depth, recession rate, DSD, dry to wet reversal, and hydroperiod use were then averaged over each instance of use, with the expectation that hydrological variables would converge on optimal values the more a cell was frequented.

Frequencies of cell use were all fourth-root transformed to meet assumptions of normality, with the exception of storks, which were inverse square-root transformed because of many single occurrences in a cell over the decade. An a priori set of candidate models was tested in Proc Glimmix (SAS Institute 2010b) to determine how the aforementioned hydrological variables were contributing to frequency of cell use (i.e., spatial occurrence). Models consisting of unique combinations of terms (Depth, Depth2,

Recession, Recession2, DSD, DSD2, Reversal, Hydroperiod (HP) and HP2); the 36

interactions among depth, recession, DSD; and the random effect Region (WCA, BCNP, and ENP; see Fig. 3.1) were evaluated for parsimony using Akaike’s Information

Criterion for small sample sizes (AICc; Burnham and Anderson 2002). Proc Glimmix was utilized specifically because it can account for residual spatial correlation using a non- parametric radial smoother. It uses a random effect to model complicated trend surfaces

(such as a spatial grid) without having to specify a particular parametric functional form. The net effect of including a random effect for this spatial correlation is a semi- parametric model where the residuals in a landscape are mostly free of spatial correlation

(McCarter and Burris 2010). By capturing patterns in the spatial variation of the landscape through radial smoothing, the noise independent of the hydrologic parameters can be reduced to better capture the species-specific behavioral response to rapidly changing habitat conditions (Dormann 2007). Models were retained if ΔAICc < 4

(Burnham and Anderson 2002). Model-averaged coefficients and SE were calculated for each fixed and random parameter by averaging all models containing the variable in proportion to the AIC weight. The importance of a specific term was determined by summing the weights of all models containing that term.

RESULTS

Identifying Available Habitat

Egrets used a mean water depth of 19.68 (±0.07 SE) cm with an optimal range

(25% to 75% quantile) of 8.71 – 31.04 cm. This is the largest range of depths of the three species and supports the egret’s tolerance of a wider range of water depths (Gawlik 2002,

Beerens et al. 2011). Water depths for suboptimal dry and suboptimal wet categories

37

were -1.78 – 8.71 cm and 31.04 – 41.63 cm, respectively; thus, restricting available resource levels to the depth range of -1.78 – 41.63 cm (10% – 90% quantile). Ibises used a mean water depth of 13.40 cm (±0.09 SE) with an optimal range (25% to 75% quantile) of 3.87 – 22.82 cm. Water depths for suboptimal dry and suboptimal wet categories were

-4.95 – 3.87 cm and 22.82 – 32.62 cm, respectively; thus, restricting available habitat to a depth range of -4.95 – 32.62 cm (10% – 90% quantile). Storks used a mean water depth of 12.86 (±0.29 SE) cm with an optimal range (25% - 75% quantile) of 2.75 – 24.21 cm.

Water depths for suboptimal dry and suboptimal wet categories were -8.73 – 2.75 cm and

24.21 – 35.77 cm, respectively; thus, restricting available habitat to the depth range of -

8.73 – 35.77 cm (10% – 90% quantile).

Depth Use

The model that best explained egret depth use included the terms for Depth,

Depth SD, Depth2 and Recession (Table 3.1). Depth use decreased (i.e., egrets use shallower depths) in a dryer landscape when less heterogeneity occurs within their depth range. Also, egrets used shallower depths with increasing recession rates. The top model accounted for 81% of the variation in depth use with all variables exhibiting high importance (>70%; Table 3.1). The most parsimonious model describing depth use by ibis included the terms for Depth, Depth SD, and the Recession*DSD interaction (Table

3.1). Similar to the egret, ibis used shallower depths as the landscape dried and variability decreased. The interaction term indicated a greater selection for shallow depths as recession rates increased when DSD availability was high, as ibis targeted sites with highly concentrated prey. Ibis used deeper water in January and February and shallower water in April and May. The top model explained 79% of the variation in depth 38

use with all of the terms exhibiting high importance (>70%; Table 3.1). Stork depth use was best explained by Depth, Depth2, DSD SD, and the Depth*DSD interaction, with selection for shallower depths when DSD availability and heterogeneity were high. The top model accounted for 78% of the variation in depth use and all terms were of high importance (Table 3.1).

Recession Rate Use

The model that best explained egret recession rate use included the terms for

Recession, Recession SD, Depth SD, and the Depth*Recess interaction (Table 3.2).

Egret recession rate use increased with increasing recession rate availability and heterogeneity, and a decrease in depth heterogeneity (as the landscape dries). Higher recession rates were selected for in shallow depths. This model explained 95% of the variation in recession rate use. The most parsimonious model describing recession rate use by ibis included the terms Recession, Recession SD, and the Depth*Recess interaction (Table 3.2). A competing model with similar weight also included Recess2; however, the parameter estimate plus the SE overlaps zero suggesting a weak effect. Ibis recession rate use also increased with increasing recession rate availability and heterogeneity with a selection for higher rates in a dryer landscape. The top model accounted for 95% of the variation in recession rate use. Recession rate use by storks was best explained by the terms Recession, Recession SD, Depth SD, and

Depth*Recession. The model explained 85% of the variation in recession rate use (Table

3.2). Recession rate use increased with increasing recession rate availability and heterogeneity and decreasing depth heterogeneity. Similar to the other species, selection

39

for more rapid rates occurred in shallow landscape depths. No variation was evident in recession rate use across months, so the term was removed.

Days since Drydown Use

The model that best explained DSD use by egrets included the terms for Depth,

Depth SD, Depth2, DSD, DSD SD, Depth*Depth Use, Depth*DSD and Recession*DSD

(Table 3.3). Accounting for DSD availability, DSD selection increased with increasing landscape depths and lower depth heterogeneity. Longer DSD use was evident when egrets were using deeper water, but primarily in mean available depths > 15-cm when

DSD selection was high. DSD selection was also higher as recession rates slowed and

DSD use increased with higher DSD heterogeneity. The top model explained 92% of the variation in DSD use. The most parsimonious model describing ibis DSD use included the terms for Depth, Depth2, DSD, DSD SD, Depth*Depth Use, and Depth*DSD (Table

3.3). Accounting for DSD availability, DSD selection increased in a wetter landscape, an effect strengthened when ibis used deeper water. Similar to egrets, in mean available depths > 10-cm, there was greater selection for long DSD cells. DSD use also increased as available DSD and heterogeneity increased. The top model accounted for 92% of the variation in DSD use. DSD use by storks was best explained by the terms Depth, Depth2,

DSD, DSD SD and Depth Use*Depth (Table 3.3). Storks used longer DSD cells when

DSD availability and heterogeneity were high and in a wetter landscape when feeding in deeper water. No variation was evident in DSD use across months, so the term was removed.

40

Temporal Foraging Conditions (Patch Quality)

Flocks of egrets across the landscape are best explained by Depth SD, Depth2,

DSD SD, DSD2, Reversal, and Depth Use*Depth. The second best model included the term for Depth*DSD, but this was a term with low importance (Table 3.4). The top model for landscape abundance included the terms Depth availability, Depth SD, Depth2,

DSD SD, Reversal, and Depth Use*Depth (Table 3.5). Egret flock and individual abundance increased with increasing DSD use (Fig. 3.2) and heterogeneity, in a similar form to the response of small fish density to days since last drydown (Trexler 2010). The depth terms indicated higher flock and individual abundance when depth heterogeneity was high and when egrets were selecting deeper water. Furthermore, individual abundance increased as the landscape dried, whereas this term was not present in the flocks model (Fig. 3.2). Flocks and individuals were negatively impacted by dry to wet reversals. Ibis landscape flocks were best explained by Depth SD, DSD, DSD SD, DSD2, and Depth*DSD (Table 3.4). A competing model with less support included the term

Depth; however, with low variable importance, this term was not considered. The top model for individual abundance included the terms Depth availability, Depth SD, Depth2,

Recess SD, DSD, DSD2, Reversal, and Depth Use*Depth (Table 3.5). Similar to the egret, ibis flock and individual abundance increased with increasing DSD use, but started to decline more rapidly with DSD above 450 days (Fig. 3.3), and this was pronounced in shallow water. Also, flocks and individuals increased as depth heterogeneity increased, occurring at intermediate landscape depths (Fig. 3.3). Individual abundance declined with increasing reversals, whereas this variable had no effect on flock abundance. Stork flocks across the landscape were best explained by Depth availability, Depth SD, Depth2, 41

Recession2, Reversal, and Depth use*Depth (Table 3.4). The best model for stork individuals included the terms Depth availability, Depth SD, Depth2, Recession SD,

Recession2, Reversal, and Depth use*Depth (Table 3.5). Predicted flock and individual abundance of storks increased with a drying landscape (Fig. 3.4) and increasing depth heterogeneity, which was highest at intermediate landscape depths. Similar to the egret, stork flocks and individuals increased when they were selecting deeper water. Flock and individual abundance also increased at intermediate recession rate use (peaking at ~0.5 cm/day; Fig. 3.4) and with fewer reversals.

Spatial Foraging Conditions (Patch Abundance)

The model that best explained the frequency of cell use (i.e., spatial occurrence) by egrets contained the terms Depth, Depth2, DSD, DSD2, HP, Depth*DSD, and

Recession*DSD (Table 3.6). The quadratic depth terms indicated that spatial occurrence was highest at intermediate values. Spatial occurrence increased with increasing hydroperiod and peaked at an intermediate DSD, again suggesting an optimal DSD of

~500 days. There was a positive effect of shallow depth use on spatial occurrence only when DSD use was high, as shallow depths are only beneficial to egrets when prey is highly concentrated. Spatial occurrence was also more positively influenced by higher recession rate use in low versus high DSD cells. Taken together, these results suggest that the frequency of cell use increased more rapidly with increasing recession in cells with fewer DSD, yet these cells were still used at a lower mean frequency than cells where egrets were using greater depths and long DSD. Because recession rate use was strongly related to landscape recession rates (Table 3.2), rapid landscape recession rates likely increased spatial occurrence (and likely prey density) when DSD use was low to a 42

level similar to that when DSD was high. Accounting for residual spatial correlation (see

Fig. 3.5), the spatial occurrence of egrets was higher in the ENP region and slightly lower in the BCNP region.

The most parsimonious ibis model explaining spatial occurrence included the terms Depth2, Recession, Recession2, Depth*DSD, and Depth*Recession (Table 3.6).

The quadratic depth term revealed predicted cell use frequency was highest at depths that were shallower than the optimum for egrets, but ibis frequency still declined slightly if cells were too dry. The squared recession term again indicated high spatial occurrence at intermediate recession rates (~0.5 cm/day). The DSD function had a similar shape to that of the egrets, showing a peak in predicted cell use frequency at ~500 DSD, an effect strengthened in shallow water. Radial spatial (x,y) smoothing was applied to the landscape, but no additional differences were found among regions.

The spatial occurrence of storks was best predicted by the terms Depth, Depth2,

Recession2, DSD, DSD2, HP, HP2, and Depth* DSD (Table 3.6). The form of the depth response was similar to the ibis with a peak at shallow positive values of depth.

Quadratic DSD and hydroperiod terms showed a peak in spatial occurrence at ~450 DSD, a sharp decline thereafter, and an increase in spatial occurrence at longer hydroperiods.

Finally, differences in DSD primarily affected spatial occurrence in shallow water, with occurrence increasing under high DSD and decreasing under low DSD. This corroborates results from the egret and ibis model that higher DSD use was associated with many more birds when the associated high prey density was concentrated in shallow water. After accounting for residual spatial correlation, the spatial occurrence of storks was higher in ENP and slightly lower in BCNP. 43

DISCUSSION

Water Dynamics and Prey

Depth

Wading bird foraging is strongly predicted by water depth, which changes daily, rapidly altering habitat suitability that varies with each species (Gawlik 2002, Beerens et al. 2011). I restricted available depths to expected wading bird foraging ranges, and all species selected shallower depths when available landscape depths were at the high and low end of the hydrological gradient. In the early dry season when most of the landscape is wet, the few shallow areas are targeted by wading birds. In contrast, at the end the dry season, birds select shallower water to exploit concentrated prey resources. These results suggest that selection for water depth can occur in response to both limitation and exploitation over the gradient of landscape conditions. This idea is supported by wading bird resource selection functions, demonstrating that selection of water depths can change given available conditions (Beerens et al. 2011).

In addition, egrets and ibises selected for shallower water when prey were being concentrated by high recession rates, but this was only observed in ibises when DSD availability was high. This behavior in ibis could reflect their tendency to feed on crayfish at relatively higher landscape water depths, but then switch to fish when they become highly concentrated under dryer conditions (Dorn et al. 2011). Egrets have a broader depth tolerance, but will opportunistically feed in shallow water to access concentrated fish that high recession rates provide (Gawlik 2002, Beerens et al. 2011). In contrast to egret and ibises, the storks’ use of shallow water was not associated with recession rates, but instead with longer hydrological inundation. This behavior would 44

allow storks to target larger sizes of marsh fishes that develop in deeper pools that may not dry annually (Ogden et al. 1976).

Recession

All species used higher recession rates when available, demonstrating its important function as a prey concentrating process. Selection of higher rates was most evident when recession rate heterogeneity was high, driven by local rainfall events.

These temporary interruptions to the drying process can lead to prey dispersal and reduced rates of capture (Gawlik 2002, Russell et al. 2002). However, this study demonstrates that wading birds can identify locations where disruptions have been minimal. Higher recession rates were also selected in a dryer landscape when these sites could be accessed in shallow water. Similar to depth, selection occurred when a particular resource was potentially limiting and when it could be exploited.

Days Since Drydown

Few studies have directly modeled wading bird selection of multi-annual hydrological patterns; however, the population dynamics of fishes at this scale are well documented ( Loftus and Eklund 1994, Trexler et al. 2002, Chick et al. 2004). A higher frequency of severe drydowns (i.e., lower DSD) limits the population of large fishes, whereas small fishes rapidly recolonize and crayfish emerge from burrows in newly flooded areas. This likely causes small variations in wading bird species’ responses to

DSD availability based on preferences of prey species. Thus, the DSD of a site is only relevant to birds when it enters and remains in the species’ depth range and was modeled accordingly. Similar to recession rate use, all species used higher DSD when available, demonstrating its utility in describing long-term prey dynamics. 45

In a previous study, egrets and ibises selected for DSD in a year with high prey availability, but not in years when prey was limiting (Beerens et al. 2011). Egrets and ibises in the current study selected for longer DSD when mean available landscape depths were greater than 15- and 10-cm, respectively; however, this selection decreased as their selection of shallower water increased. This demonstrates the importance of DSD during wet season conditions and as a prerequisite for efficient foraging (when combined with steady recession) later in the dry season. In contrast, DSD selection by storks did not vary with landscape conditions because they still used high DSD in shallow water and did not select higher DSD in wet conditions. All species also used longer DSD by selecting locations with greater depths, a response that likely benefits the species with a broader depth tolerance allowing them to forage in sites with a more abundant and diverse prey base (Loftus and Eklund 1994, Trexler et al. 2002) with larger body sizes (Chick et al.

2004). For egrets, DSD selection was slightly higher as recession rates slowed, which could indicate their ability to locate higher density sites when prey is not being further concentrated by recession.

Temporal Abundance

Similar to resource selection patterns, temporal models of both flock and individual abundance indicate that wading birds are responding to and may be limited by processes interacting over the multiple spatio-temporal scales that influence prey availability. For example, for all species, the largest numbers of foraging birds were observed when the mean DSD used was ~400-600 days, consistent with the idea that long-term wetland inundation (across a gradient of hydroperiods) allows for prey production. Even under these conditions, however, the timing of greatest numbers of 46

foraging birds differed among species depending on the availability and selection of suitable depths to access the increased prey production. Egrets were most abundant when locations with a long DSD (~500 days) were used and individuals fed in deeper water (15

– 25 cm), but egret abundance declined under the driest landscape conditions observed (<

15 cm availability). Ibises were also more abundant when they used longer DSD (~450 days), but the increase in abundance was instead strengthened when ibises used shallower water (10 – 20 cm). Indeed, in a year with high prey production and high nesting success, collected boluses of ibises predominately contained small fish that were concentrated in shallow water (Cook and Herring 2007, Dorn et al. 2011). Abundance of storks did not directly respond to changing selection for DSD, but rather their abundance was highest in a dry landscape when individuals fed in shallow water (0 – 10 cm). The increased use of shallower water in a high production (i.e., DSD) environment demonstrates the fundamental importance of an established multi-annual prey base for all species. These conditions would occur when the majority of foraging habitat did not dry in the prior year.

Peak flock and individual abundance were observed in all species when landscape depth heterogeneity was high and individuals had a variety of depth choices. Under the wettest conditions, high heterogeneity was available to egrets, which can exploit deeper water. For ibis and storks, high depth heterogeneity occurred at the middle one-third of the hydrological gradient and limited ibis and stork abundance under wetter (and dryer) conditions. This reaffirms the importance of a landscape high in spatial heterogeneity from previous reports (Fleming et al. 1994, Beerens et al. 2011) and demonstrates that the significant loss of short hydroperiod wetlands may have caused a delay in nesting 47

initiations of searcher species by decreasing the depth heterogeneity at the beginning of particularly wet years (Fleming et al. 1994, Gawlik 2002). This spatial limitation also increases wading bird dependence on high drying rates to provide shallow depths early in the dry season that may have changed the species response to recession rate. While high recession rates were not necessarily a historical requirement of successful breeding, they provide new high quality patches that are exposed in suitable depths and may ease the loss of foraging habitat that would have been suitable under wetter conditions (Frederick and Spalding 1994, Beerens et al. 2011).

In this study, egret abundance was higher with the use of intermediate recession rates in a form similar to a model predicting their daily nest survival (Herring et al. 2010), with a peak in abundance at ~0.5 cm/day. Stork abundance was similarly higher with the use of intermediate recession rates. For these species, other studies have demonstrated a positive linear response between abundance or use and increasing recession rate (Russell et al. 2002, Beerens et al. 2011, Borkhataria et al. 2013), although there are several examples when recession rate does not contribute to increased probability of use (Beerens et al. 2011, Herring and Gawlik 2011), demonstrating that the species response to recession rate varies given available landscape conditions. In a year with high food availability, both egrets and storks did not demonstrate selection for recession rate illustrating that selection is more likely to occur in less than optimal conditions (Beerens et al. 2011, Herring and Gawlik 2011) and may be a misleading model for habitat quality without the context of processes occurring at longer temporal scales.

For all species, distinct landscape conditions caused individual abundance to rise more rapidly than flock abundance. For example, egrets clustered as the landscape dried, 48

indicated by a negative response of flocks but not individuals. Individuals cluster into larger flocks when the benefits of conspecific presence outweigh the costs of interference

(Folmer et al. 2012). This dichotomy provides a good indicator of when species with higher costs of interference use high quality habitat. As visual foragers, egrets incur higher interference costs when foraging in groups; however they clustered when food was plentiful as species converged on the most profitable prey resources (Gawlik 2002,

Beerens et al. 2011). Ibis dispersed following dry season reversals, and storks clustered at intermediate depths, demonstrating that social attraction can co-vary with foraging strategy, water-level changes, and habitat quality. As searchers and tactile foragers, ibises and storks are less sensitive to interference and maintained a higher level of conspecific attraction regardless of habitat conditions, likely resulting from specialized foraging strategies where conspecific presence helps reduce search and travel time

(Kushlan 1976, Gawlik and Crozier 2007). Therefore, flock abundance in ibises and storks is more likely to indicate habitat quality and represent a larger area of the landscape suitable for foraging (Frederick et al. 2009).

Spatial Abundance

Typical of hydrologically-integrated systems, the values of variables sampled at nearby locations were not independent of each other, resulting in spatial autocorrelation

(Tobler 1970). Accordingly, high residual spatial correlation in wading bird foraging distributions was substantially reduced by applying a non-parametric radial smoother, indicating spatial covariates not captured by the environmental variables sampled

(Dormann et al. 2007). While the proxies for temporally-specific resources in this study are strongly linked with wading bird habitat use, other spatially correlated variables, not 49

available in this study, can influence prey availability such as nutrients, topography and vegetation structure (Gawlik 2002, Beerens et al. 2011). Topography could play a particularly important role in accounting for spatial autocorrelation at the landscape scale, because large areas of wetland could drain into particular cells and highly concentrate fish leading to frequent occurrences of birds over a decadal scale.

While it has been demonstrated that environmental relationships can invert when accounting for spatial autocorrelation (Kühn 2006), the results from this study confirm spatio-temporal interactions identified by the TFC model. Again, the co-occurrence of cells with long DSD (> 400 days) and shallow water was associated with more frequent use by all species. Conversely, frequency of use of a cell declined in shallow water when

DSD (and therefore time for prey reproduction) was low, indicating that birds may use locations less even with optimal depths. If applied on an annual scale, this could indicate longer site fidelity in more resource dense cells, with birds potentially relying less on recession to expose new patches and less habitat overall over the length of the breeding season. In addition, these species used cells with a longer hydroperiod more frequently, demonstrating their energetic demand for larger and more abundant prey that could be exploited over repeated foraging bouts to the same location. This study also supports the idea that the egret’s selection of higher recession rates when prey density was low provides an enhancement in foraging conditions that increases their occurrence. While a similar selection response was noted in ibises (Beerens et al. 2011), this behavior did not increase spatial occurrence over the gradient of habitat conditions in the current study.

The behavior was only demonstrated (and may only be effective) in a particularly dry year, when ibis switch to concentrated fish as a prey resource (Dorn et al. 2011). 50

Similar to the TFC, a sharper decline was evident in spatial occurrence for ibis and storks when DSD exceeded 600 days; however, use of these cells tended to co-occur with the use of deeper water. This could demonstrate the searchers’ higher giving up density (GUD) in these depths (Gawlik 2002). Alternatively, the species’ response could have decreased because of changes in the prey community as a result of long periods of inundation. Under these habitat conditions, cells become dominated by larger predatory fishes and limit crayfish populations (Kellogg and Dorn 2012). Periodic droughts may benefit ibis indirectly by increasing crayfish populations, potentially driving the large pulses in ibis nesting effort that follow drought years.

Application

My findings provide context to species-specific resource selection and distributions and are consistent with other patterns observed, but offer a better model for understanding how recent trends in water levels and spatio-temporal limitations have driven species-specific population trends. Results can also be applied to evaluate the multi-annual resource pulses of real-time conditions, climate change scenarios, or restorative hydrological regimes by tracking changing seasonal and annual locations of high quality foraging patches.

The TFC predicts the daily abundance of flocks and individuals on the landscape within a changing area of suitable depths throughout the species-specific nesting season.

By developing a daily temporal model focused on the average response to resource levels within suitable habitat (i.e., depths) while simultaneously modeling how the resource heterogeneity impacts the response, the changing preference for resource-levels is captured and evaluated by the landscape response. Because only resources within 51

suitable habitat were modeled, the response can represent patch quality, summed over the extent of the Everglades. The SFC predicts spatial occurrence for each cell by integrating spatial dynamics unaccounted for by the chosen set of predictors (i.e., spatial correlation).

When this model is averaged spatially over daily time steps, it can serve as a surrogate measure of patch abundance. Finally, combining daily output from the SFC and TFC can jointly account for the seasonal increase and eventual decrease in patch abundance as the landscape dries, and the increase in patch quality as habitat with longer periods of inundation becomes available in suitable water depths.

In the Everglades, wading birds are intimately connected with rapidly changing hydrology and with each other, thus serving as indicators of ecosystem health and restoration efforts mandated by the Water Resources Development Act. Ecological models that incorporate species’ ecology such as dynamic habitat selection, conspecific attraction, and responses over environmental gradients provide more realistic predictions to facilitate restoration success.

52

LITERATURE CITED

Arthur, S. M., B. F. J. Manly, L. L. McDonald, and G. W. Garner. 1996. Assessing

habitat selection when availability changes. Ecology 77:215–227.

Austin, M. 2007. Species distribution models and ecological theory: A critical assessment

and some possible new approaches. Ecological Modelling 200:1–19.

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 615–658 in Everglades: The Ecosystem and

Its Restoration (S. M. Davis and J. C. Ogden, Eds.). St. Lucie Press, Delray

Beach, Florida.

Beerens, J. M., D. E. Gawlik, G. Herring, and M. I. Cook. 2011. Dynamic habitat

selection by two wading bird species with divergent foraging strategies in a

seasonally fluctuating wetland. The Auk 128:651–662.

Beerens, J. M., and M. I. Cook. 2010. Using Wood Stork distribution data to develop

water management guidelines. Pages B1–B7 in USFWS multi-species transition

strategy for Water Conservation Area 3A.

Beyer, H. L., D. T. Haydon, J. M. Morales, J. L. Frair, M. Hebblewhite, M. Mitchell, and

J. Matthiopoulos. 2010. The interpretation of habitat preference metrics under

use-availability designs. Philosophical transactions of the Royal Society of

London. Series B, Biological Sciences 365:2245–54.

53

Borkhataria, R. R., P. C. Frederick, J. M. Beerens, and A. L. Bryan. 2013. Predicting

Wood Stork (Mycteria americana) nesting dynamics in response to hydrological

change in ENP using simulation modeling, nesting records, and satellite telemetry

data. Final Report Submitted to Everglades National Park.

Botson, B., and D. E. Gawlik. 2010. Aquatic fauna seasonal concentrations. Final Report

Submitted to the South Florida Water Management District. Florida Atlantic

University.

Burnham, K. P., and D. R. Anderson. 2002. Model Selection and Multimodel Inference:

A Practical Information-Theoretic Approach, 2nd ed. Springer-Verlag, New York.

Campomizzi, A. J., J. a. Butcher, S. L. Farrell, A. G. Snelgrove, B. a. Collier, K. J.

Gutzwiller, M. L. Morrison, and R. N. Wilkins. 2008. Conspecific attraction is a

missing component in wildlife habitat modeling. Journal of Wildlife Management

72:331–336.

Chick, J. H., C. R. Ruetz, and J. C. Trexler. 2004. Spatial scale and abundance patterns of

large fish communities in freshwater marshes of the Florida Everglades. Wetlands

24:652–664.

Cook, M. I., and H. K. Herring. 2007. System-wide summary. Pages 1-2 in South Florida

Wading Bird Report, vol. 13 (M. I. Cook and H. K. Herring, Eds.). South Florida

Water Management District, West Palm Beach, Florida.

Dormann, C. F. 2007. Effects of incorporating spatial autocorrelation into the analysis of

species distribution data. Global Ecology and Biogeography 16:129–138.

54

Dorn, N. J., M. I. Cook, G. Herring, R. A. Boyle, J. Nelson, and D. E. Gawlik. 2011.

Aquatic prey switching and urban foraging by the White Ibis Eudocimus albus are

determined by wetland hydrological conditions. Ibis 153:323–335.

Elith, J., and J. R. Leathwick. 2009. Species distribution models: Ecological explanation

and prediction across space and time. Annual Review of Ecology, Evolution, and

Systematics 40:677–697.

Fleming, D., W. Wolff, and D. DeAngelis. 1994. Importance of landscape heterogeneity

to wood storks in Florida Everglades. Environmental Management 18:743–757.

Folmer, E. O., H. Olff, and T. Piersma. 2012. The spatial distribution of flocking

foragers: Disentangling the effects of food availability, interference and

conspecific attraction by means of spatial autoregressive modeling. Oikos

121:551–561.

Frederick, P. C., D. E. Gawlik, J. C. Ogden, M. I. Cook, and M. Lusk. 2009. The White

Ibis and Wood Stork as indicators for restoration of the Everglades ecosystem.

Ecological Indicators 9:S83–S95.

Frederick, P. C., and J. C. Ogden. 1997. Philopatry and nomadism : Contrasting long-

term movement behavior and population dynamics of White Ibises and Wood

Storks. Colonial Waterbirds 20:316–323.

Frederick, P. C., and M. G. Spalding. 1994. Factors affecting reproductive success of

wading birds (Ciconiiformes) in the Everglades ecosystem. Pages 659–691 in

Everglades, The Ecosystem and its Restoration (S. M. Davis and J. C. Ogden,

Eds.). St. Lucie Press, Delray Beach, FL.

55

Frid, A., and L. Dill. 2002. Human-caused disturbance stimuli as a form of predation risk.

Conservation Ecology 6:11.

Garshelis, D. L. 2000. Delusions in habitat evaluation: measuring use, selection, and

importance. Pages 111-164 in Research techniques in animal ecology:

Controversies and consequences (L. Boitani, and T. K. Fuller, Eds.). Columbia

University Press, New York.

Gawlik, D. E., and G. E. Crozier. 2007. A test of cues affecting habitat selection by

wading birds. The Auk 124:1075.

Gawlik, D. E. 2002. The effects of prey availability on the numerical response of wading

birds. Ecological Monographs 72:329–346.

Gillies, C. S., M. Hebblewhite, S. E. Nielsen, M. a. Krawchuk, C. L. Aldridge, J. L. Frair,

D. J. Saher, C. E. Stevens, and C. L. Jerde. 2006. Application of random effects to

the study of resource selection by animals. Journal of Animal Ecology 75:887–

898.

Godvik, I. M. R., L. E. Loe, J. O. Vik, V. Veiberg, R. Langvatn, and A. Mysterud. 2009.

Temporal scales, trade-offs, and functional responses in red deer habitat selection.

Ecology 90:699–710.

Hebblewhite, M., and E. Merrill. 2008. Modelling wildlife-human relationships for social

species with mixed-effects resource selection models. Journal of Applied Ecology

45:834–844.

Herring, G., D. E. Gawlik, M. I. Cook, and J. M. Beerens. 2010. Sensitivity of nesting

Great Egrets (Ardea alba) and White Ibises (Eudocimus albus) to reduced prey

availability. The Auk 127:660–670. 56

Herring, H. K., and D. E. Gawlik. 2011. Resource selection functions for Wood Stork

foraging habitat in the southern Everglades. Waterbirds 34:133–142.

Johnson, D. H. 1980. The comparison of usage and availability measurements for

evaluating resource preference. Ecology 61:65–71.

Kahl, M. P. 1964. Food ecology of the wood stork (Mycteria americana) in Florida.

Ecological Monographs 34:97–117.

Kellogg, C. M., and N. J. Dorn. 2012. Consumptive effects of fish reduce wetland

crayfish recruitment and drive species turnover. Oecologia 168:1111–21.

Kühn, I. 2006. Incorporating spatial autocorrelation may invert observed patterns.

Diversity & Distributions 13:66–69.

Kushlan, J. 1976. Wading bird predation in a seasonally fluctuating pond. The Auk

93:464–476.

Lantz, S. M., D. E. Gawlik, and M. I. Cook. 2011. The effects of water depth and

emergent vegetation on foraging success and habitat selection of wading birds in

the Everglades. Waterbirds 34:439–447.

Liu, Z., J. C. Volin, D. Owen, L. G. Pearlstine, J. R. Allen, F. J. Mazzotti, and A. L.

Higer. 2009. Validation and ecosystem applications of the EDEN water-surface

model for the Florida Everglades. Ecohydrology 2:182–194.

Loftus, W. F., and A. M. Eklund. 1994. Long-term dynamics of an Everglades small-fish

assemblage. Pages 461–483 in Everglades, The Ecosystem and its Restoration (S.

M. Davis and J. C. Ogden, Eds.). St. Lucie Press, Delray Beach, FL.

57

Mazzotti, F. J., G. R. Best, L. A. Brandt, M. S. Cherkiss, B. M. Jeffery, and K. G. Rice.

2009. Alligators and crocodiles as indicators for restoration of Everglades

ecosystems. Ecological Indicators 9:S137–S149.

McCarter, K., and E. Burris. 2010. Accounting for spatial correlation using radial

smoothers in statistical models used for developing variable-rate treatment

prescriptions. Proceedings of the 10th International Conference on Precision

Agriculture. Denver, CO.

McLoughlin, P. D., D. W. Morris, D. Fortin, E. Vander Wal, and A. L. Contasti. 2010.

Considering ecological dynamics in resource selection functions. The Journal of

Animal Ecology 79:4–12.

Mysterud, A., and R. Ims. 1998. Functional responses in habitat use: availability

influences relative use in trade-off situations. Ecology 79:1435–1441.

Ogden, J. C. 2005. Everglades ridge and slough conceptual ecological model. Wetlands

25:810–820.

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 Everglades, The Ecosystem and its Restoration (S.

M. Davis and J. C. Ogden, Eds.). St. Lucie Press, Delray Beach, FL.

Ogden, J. C., J. A. Kushlan, and J. T. Tilmant. 1976. Prey selectivity by the Wood Stork.

The Condor 78:324–330.

Orians, G., and J. Wittenberger. 1991. Spatial and temporal scales in habitat selection.

American Naturalist 137:S29–S49.

58

RECOVER. 2004. CERP monitoring and assessment plan. The purpose and scope of

CERP monitoring and assessment plan.

http://www.evergladesplan.org/pm/recover/recover_docs/map/MAP_1.0_Purpose.

pdf.

Rittenhouse, C. D., F. R. Thompson, W. D. Dijak, J. J. Millspaugh, and R. L. Clawson.

2010. Evaluation of habitat suitability models for forest passerines using

demographic data. Journal of Wildlife Management 74:411–422.

Russell, G. J., O. L. Bass, and S. L. Pimm. 2002. The effect of hydrological patterns and

breeding-season flooding on the numbers and distribution of wading birds in

Everglades National Park. Animal Conservation 5:185–199.

SAS Institute. 2010a. SAS 9.2 Help and Documentation. MIXED Procedure. SAS

Institute, Cary, North Carolina.

SAS Institute. 2010a. SAS 9.2 Help and Documentation. GLIMMIX Procedure. SAS

Institute, Cary, North Carolina.

Tobler, W. 1970. A computer movie simulating urban growth in the Detroit region.

Economic Geography 46:234–240.

Trexler, J. C. 2010. Greater Everglades aquatic trophic levels DRAFT small-sized

freshwater fish density performance measure. Draft DECOMP Performance

Measure Documentation Sheet Submitted to the Army Corp of Engineers.

Trexler, J. C., and C. W. Goss. 2009. Aquatic fauna as indicators for Everglades

restoration: Applying dynamic targets in assessments. Ecological Indicators

9:S108–S119.

59

Trexler, J. C., W. F. Loftus, F. Jordan, and J. H. Chick. 2002. Ecological scale and its

implications for freshwater fishes in the Florida Everglades. Pages 153–181 in

The Everglades, Florida Bay, and Coral Reefs of the Florida Keys: An Ecosystem

Sourcebook (J. W. Porter and K. G. Porter, Eds.). CRC Press, New York.

60

Figure 3.1. South Florida study system displaying Everglades hydrological basins

(regions) and Systematic Reconnaissance Flight (SRF) survey extent. The regions of coverage include Water Conservation Areas (WCA) 1, 2A, 2B, 3A, 3B, Big Cypress

National Park (BCNP), and Everglades National Park (ENP). 61

Figure 3.2. Daily mean landscape flocks and individuals (fourth-root transformed) predicted by the model-averaged terms for the Great Egret. Predicted presence and abundance are highest with an average landscape depth (within available depths) of 15-22 cm and when Great Egrets are using an average DSD of ~500 days.

62

Figure 3.3. Daily mean landscape flocks and individuals (fourth-root transformed) predicted by the model-averaged terms for the White Ibis. Predicted presence and abundance are highest with an average landscape depth (within available depths) of 10-17 cm and when White Ibis are using an average DSD of ~450 days.

63

Figure 3.4. Daily mean landscape flocks and individuals (fourth-root transformed) predicted by the model-averaged terms for the Wood Stork. Predicted presence and abundance are highest with an average landscape depth (within available depths) less than 15 cm and when Wood Storks are using an average recession rate of ~0.5 cm/day.

64

Figure 3.5. Map displaying XY parameter estimates accounting for residual spatial correlation of Great Egret frequency of use. Locations in red are frequented more often across time after accounting for hydrological predictors.

65

Table 3.1. Ranking of candidate models describing variables influencing daily mean depth use of Great Egrets, White Ibises, and Wood Storks in the Florida Everglades (Proc

Mixed). Models are ranked by differences in Akaike’s information criterion and only

d candidate models within ΔAICc ≤ 4.0 are presented. Model selection results are followed by model averaging results for each species. The R2 represents the model fit for the estimated mean daily depth use vs. model averaged predicted values.

b c d 2 GREAT EGRET MODEL K AICc modelid ΔAICc weight R Depth, Depth SD, Depth2, Recess 7 1588.7 15 0.00 0.22 0.81 Depth, Depth SD, Depth2, Recess, DSD, DSD2 9 1589.0 20 0.23 0.19 Depth, Depth2, Recess, Depth*Recess 7 1589.9 23 1.15 0.12

Variable N Avg PE SE Importance Intercept 27 -18.875 2.83 1.00 Depth 24 2.300 0.42 1.00 Recess 16 -1.652 1.16 0.94 Depth SD 14 0.614 0.33 0.91 Depth2 10 -0.017 0.01 0.84

b c d 2 WHITE IBIS MODEL K AICc modelid ΔAIC c weight R Depth, Depth SD, Recess*DSD 6 1472.6 3 0.00 0.30 0.79 Depth, Depth SD, Recess, Recess*DSD 7 1473.3 21 0.71 0.21 Depth, Depth2, Recess, Recess*DSD 7 1475.0 14 2.43 0.09 Depth, Depth SD, Recess, Recess*DSD, Cells 8 1475.3 26 2.78 0.07

Variable N Avg PE SE Importance Intercept 27 -12.304 2.61 1.00 Depth 24 1.726 0.21 1.00 Recess*DSD 14 -0.004 0.00 0.87 Depth SD 12 0.558 0.24 0.84

b c d 2 WOOD STORK MODEL K AICc modelid ΔAICc weight R Depth, Depth2, DSD SD, Depth*DSD 6 1229.2 26 0.00 0.18 0.78 Depth, Depth2, Recess, DSD SD, Depth*DSD 7 1229.6 21 0.43 0.15 Depth, Depth2, Recess2, DSD SD, Depth*DSD 7 1229.8 22 0.67 0.13 Depth, Depth2, DSD SD, Depth*DSD, Recess*DSD 7 1230.3 16 1.16 0.10

Variable N Avg PE SE Importance Intercept 27 -19.170 3.24 1.00 Depth 22 3.199 0.28 1.00 Depth2 22 -0.037 0.01 1.00 Depth*DSD 14 -0.001 0.00 0.95 DSD SD 14 0.012 0.01 0.80

66

Table 3.2. Ranking of candidate models describing variables influencing daily mean 2- week recession rate use of Great Egrets, White Ibises, and Wood Storks in the Florida

Everglades (Proc Mixed). Models are ranked by differences in Akaike’s information

d criterion and only candidate models within ΔAICc ≤ 4.0 are presented. Model selection results are followed by model averaging results for each species. The R2 represents the model fit for the estimated mean daily recession rate use vs. model averaged predicted values.

b c d e 2 GREAT EGRET MODEL K AICc modelid ΔAICc wi R Recess, Recess SD, Depth SD, Depth*Recess 6 -337.7 23 0.00 0.44 0.95 Recess, Recess SD, Recess2, Depth SD, Depth*Recess, Recess*DSD 8 -335.5 9 2.23 0.14 Recess, Recess SD, Recess2, Depth, Depth SD, Depth*Recess, Cells 9 -334.9 8 2.85 0.11

Variable N Avg PE SE Importance Intercept 27 0.123 0.04 1.00 Recess 16 1.074 0.04 1.00 Recess SD 10 0.247 0.05 1.00 Depth SD 10 -0.012 0.00 0.98 Depth*Recess 10 -0.004 0.00 0.83

b c d e 2 WHITE IBIS MODEL K AIC c modelid ΔAIC c w i R Recess, Recess SD, Depth*Recess 5 -281.5 23 0.00 0.34 0.95 Recess, Recess SD, Recess2, Depth*Recess 6 -281.1 8 0.46 0.27 Recess, Recess SD, Recess2, Depth SD, DSD, Depth*DSD, Depth*Recess, Depth*DSD 9 -279.5 26 2.01 0.13

Variable N Avg PE SE Importance Intercept 27 0.017 0.02 1.00 Recess 16 1.156 0.04 1.00 Recess SD 10 0.150 0.11 1.00 Depth*Recess 10 -0.013 0.05 1.00

b c d e 2 WOOD STORK MODEL K AICc modelid ΔAICc wi R Recess, Recess SD, Depth SD, Depth*Recess 6 -46.9 24 0.00 0.33 0.85 Recess, Recess SD, Recess2, Depth SD, Depth*Recess, Recess*DSD 8 -46.5 26 0.41 0.27 Recess, Recess SD, Depth SD, Depth*Recess, Recess*DSD 7 -45.0 9 1.83 0.13 Recess, Recess SD, Depth*Recess 5 -45.0 23 1.91 0.13

Variable N Avg PE SE Importance Intercept 27 0.055 0.08 1.00 Recess 18 1.128 0.07 1.00 Recess SD 12 0.560 0.11 1.00 Depth*Recess 10 -0.010 0.00 1.00 Depth SD 12 -0.011 0.00 0.97

67

Table 3.3. Ranking of candidate models describing variables influencing daily mean days since drydown (DSD) use of Great Egrets, White Ibises, and Wood Storks in the Florida

Everglades (Proc Mixed). Models are ranked by differences in Akaike’s information

d criterion and only candidate models within ΔAICc ≤ 4.0 are presented. Model selection results are followed by model averaging results for each species. The R2 represents the model fit for the estimated mean daily DSD use vs. model averaged predicted values.

b c d e 2 GREAT EGRET MODEL K AICc modelid ΔAICc wi R Depth, Depth SD, Depth2, DSD, DSD SD, Depth Use*Depth, Depth*DSD, Recess*DSD 10 2676.7 9 0.00 0.88 0.92 Depth, Depth SD, Depth2, DSD, DSD SD, DSD, 2Depth Use*Depth, Depth*DSD, Depth*Recess, Recess*DSD 12 2680.7 12 4.08 0.11

Variable N Avg PE SE Importance Intercept 27 -134.269 25.47 1.00 Depth 16 33.356 3.57 1.00 Depth SD 10 -6.146 2.98 1.00 Depth2 10 -1.687 0.12 1.00 DSD 14 0.483 0.08 1.00 DSD SD 12 0.171 0.07 1.00 Depth Use* Depth 10 0.554 0.04 1.00 Depth*DSD 14 0.021 0.00 1.00 Recess*DSD 14 -0.069 0.04 1.00

b c d e 2 WHITE IBIS MODEL K AIC c modelid ΔAIC c w i R Depth, Depth2, DSD, DSD SD, Depth Use*Depth, Depth*DSD 8 2609.1 9 0.00 0.51 0.92 Depth, Depth2, DSD, DSD SD, Depth Use*Depth, Depth*DSD, Cells 9 2609.6 15 0.51 0.40

Variable N Avg PE SE Importance Intercept 27 -96.078 23.19 1.00 Depth 18 26.009 2.89 1.00 Depth2 10 -1.802 0.15 1.00 DSD 14 0.548 0.08 1.00 DSD SD 10 0.122 0.06 1.00 Depth Use*Depth 12 0.638 0.05 1.00 Depth*DSD 14 0.024 0.00 1.00

b c d e 2 WOOD STORK MODEL K AICc modelid ΔAICc wi R Depth, Depth2, DSD, DSD SD, Depth Use*Depth 7 2149.0 19 0.00 0.43 0.76 Depth, Depth2, DSD, DSD SD, Depth Use*Depth, Depth*Recess 8 2150.5 9 1.45 0.21 Depth, Depth2, DSD, DSD SD, Depth Use*Depth, Depth*DSD 8 2150.7 15 1.64 0.19

Variable N Avg PE SE Importance Intercept 27 -79.957 38.01 1.00 DSD 14 0.809 0.08 1.00 Depth 18 17.188 3.47 1.00 DSD SD 10 0.246 0.11 1.00 Depth Use*Depth 12 0.418 0.06 1.00 Depth2 10 -0.930 0.13 1.00 68

Table 3.4. Ranking of candidate models describing variables influencing daily mean flock abundance of Great Egrets, White Ibises, and Wood Storks in the Florida

Everglades (Proc Mixed). Models are ranked by differences in Akaike’s information

d criterion and only candidate models within ΔAICc ≤ 4.0 are presented. Model selection results are followed by model averaging results for each species. The R2 represents the model fit for the estimated daily flock abundance vs. model averaged predicted values.

b c d e 2 GREAT EGRET MODEL K AICc modelid ΔAICc wi R Depth SD, Depth2, DSD SD, DSD2, Reversal, Depth Use*Depth 8 626.3 5 0.00 0.41 0.39 Depth SD, Depth2, DSD SD, Reversal, Depth*DSD, Depth Use*Depth 8 628.45 9 0.16 0.38

Variable N Avg PE SE Importance Intercept 27 1.132 0.33 1.00 Depth SD 10 0.194 0.03 1.00 Depth2 10 -0.002 0.00 1.00 DSD SD 10 0.004 0.00 1.00 Reversal 14 -1.675 0.30 1.00 Depth Use*Depth 18 0.003 0.00 1.00 DSD2 16 -0.000 0.00 0.62

b c d e 2 WHITE IBIS MODEL K AICc modelid ΔAICc wi R Depth SD, DSD, DSD SD, DSD2, Depth*DSD 7 558.6 15 0.00 0.70 0.32 Depth, Depth SD, DSD, DSD SD, DSD2, Depth*DSD 8 560.7 17 2.01 0.26

Variable N Avg PE SE Importance Intercept 27 0.453 0.32 1.00 Depth SD 10 0.128 0.04 1.00 DSD 10 0.005 0.00 1.00 DSD2 10 0.000 0.00 1.00 DSD SD 12 0.003 0.00 1.00 Depth*DSD 14 -0.000 0.00 0.95

b c d e 2 WOOD STORK MODEL K AICc modelid ΔAICc wi R Depth A, Depth SD, Depth2, Recess2, Reversal, Depth Use*Depth 8 -211.8 21 0.00 0.77 0.44 Depth A, Depth, Depth SD, Depth2, Recess2, Reversal, DSD2, Depth Use*Depth 10 -209.4 14 2.45 0.23

Variable N Avg PE SE Importance Intercept 27 0.578 0.05 1.00 Depth A 10 0.024 0.00 1.00 Depth SD 12 -0.032 0.01 1.00 Depth2 10 0.000 0.00 1.00 Recess2 14 0.053 0.02 1.00 Reversal 10 0.127 0.08 1.00 Depth Use*Depth 12 -0.001 0.00 1.00

69

Table 3.5. Ranking of candidate models describing variables influencing daily individual abundance of Great Egrets, White Ibises, and Wood Storks in the Florida Everglades

(Proc Mixed). Models are ranked by differences in Akaike’s information criterion and

d only candidate models within ΔAICc ≤ 4.0 are presented. Model selection results are followed by model averaging results for each species. The R2 represents the model fit for the estimated daily individual abundance vs. model averaged predicted values.

b c d e 2 GREAT EGRET MODEL K AICc modelid ΔAICc wi R Depth A, Depth SD, Depth2, DSD SD, Reversal, Depth Use*Depth 9 797.6 12 0.00 0.44 0.35 Depth A, Depth SD, Depth2, Recess2, DSD SD, Reversal, Depth Use*Depth 9 797.6 2 0.08 0.42

Variable N Avg PE SE Importance Intercept 27 3.176 0.54 1.00 Reversal 10 -2.482 0.44 1.00 Depth Use*Depth 11 0.005 0.00 1.00 Depth SD 9 0.268 0.05 1.00 DSD SD 9 0.005 0.00 1.00 Depth2 11 -0.003 0.00 1.00 Depth A 9 -0.109 0.04 1.00

b c d e 2 WHITE IBIS MODEL K AIC c modelid ΔAIC c w i R Depth A, Depth SD, Depth2, Recess SD , DSD, DSD2, Reversal, Depth Use*Depth 10 888.3 4 0.00 0.75 0.31 Depth SD, Depth2, Recess SD, DSD, Reversal, DSD Use*DSD 8 891.3 11 2.98 0.17

Variable N Avg PE SE Importance Intercept 27 2.764 0.72 1.00 DSD 13 0.008 0.00 1.00 Reversal 9 -2.946 0.66 1.00 Depth SD 9 0.336 0.10 1.00 Recess SD 11 1.590 0.63 1.00 Depth2 9 -0.003 0.00 1.00 Depth A 9 -0.167 0.08 0.87 DSD2 10 0.000 0.00 0.76 Depth Use*Depth 9 0.007 0.00 0.76

b c d e 2 WOOD STORK MODEL K AICc modelid ΔAICc wi R Depth A, Depth SD, Depth2, Recess SD, Recess2, Reversal, Depth Use*Depth 9 481.1 5 0.00 0.79 0.30 Depth A, Depth, Depth SD, Depth2, Recess, Recess SD, Recess2, DSD SD, Reversal, Depth Use*Depth 12 484.9 6 3.72 0.12

Variable N Avg PE SE Importance Intercept 27 2.794 0.39 1.00 Depth SD 11 0.141 0.38 1.00 Depth A 10 -0.114 0.03 1.00 Reversal 9 -1.471 0.58 0.99 Recess SD 9 0.907 0.48 0.98 Recess2 13 -0.338 0.13 0.98 Depth Use*Depth 9 0.003 0.00 0.94 Depth2 9 -0.001 0.00 0.94 70

Table 3.6. Ranking of candidate models describing variables influencing frequency of cell use (i.e., spatial occurrence) over the study period for the Great Egret, White Ibis, and Wood Stork (Proc Glimmix). Models are ranked by differences in Akaike’s

d information criterion and only candidate models within ΔAICc ≤ 4.0 are presented.

Model selection results are followed by model averaging results for each species. The R2 represents the model fit for the estimated spatial occurrence vs. model averaged predicted values.

GREAT EGRET MODEL N AICC modelid d_aic weight R2 Depth, Depth2, DSD, DSD2, HP, Depth*DSD, Recess*DSD 12 880.59 5 0.00 0.74 0.88 Depth, Depth2, DSD, DSD2, Reversal, HP, Depth*DSD, Depth*Recess, Recess*DSD 14 883.83 18 3.24 0.15 Variable N Avg PE SE Importance Intercept 27 -0.584 7.48 1.00 Depth 14 -0.006 0.00 1.00 Depth2 14 -0.000 0.00 1.00 Depth*DSD 15 0.000 0.00 1.00 DSD2 15 -0.000 0.00 1.00 HP 14 0.004 0.00 1.00 DSD 13 0.001 0.00 0.99 Recess*DSD 11 -0.000 0.00 0.99

WHITE IBIS MODEL N AICC modelid d_aic weight R 2 Depth2, Recess, Recess2, Depth*DSD, Depth*Recess 8 750.2 4 0.00 0.55 0.83 Depth2, Recess, Recess2, DSD2, Depth*DSD 8 750.9 6 0.67 0.39

Variable N Avg PE SE Importance Intercept 27 1.126 7.79 1.00 Depth2 17 -0.000 0.00 1.00 Recess2 15 -0.012 0.01 1.00 Recess 12 0.061 0.02 0.99 Depth*DSD 16 0.000 0.00 0.94 Depth*Recess 15 0.000 0.00 0.58

WOOD STORK MODEL N AICC modelid d_aic weight R2 Depth, Depth2, Recess2, DSD, DSD2, HP, HP2, Depth*DSD 13 -485.9 12 0.00 0.49 0.55 Depth, Depth2, Recess2, DSD, DSD2, Reversal, HP, HP2, Depth*DSD 14 -485.1 18 0.72 0.34

Variable N Avg PE SE Importance Intercept 27 0.987 0.63 1.00 Depth 15 0.004 0.00 1.00 Depth2 15 0.000 0.00 1.00 DSD 13 -0.001 0.00 1.00 DSD2 14 0.000 0.00 1.00 Depth*DSD 13 -0.000 0.00 1.00 HP2 9 -0.000 0.00 1.00 HP 15 0.001 0.00 1.00 71

CHAPTER 4: HABITAT EVALUATION OVER MULTIPLE PHASES OF BREEDING

USING DYNAMIC SPATIO-TEMPORAL HABITAT SUITABILITY INDICES

ABSTRACT

Habitat evaluation relates a species’ habitat to their survival and reproduction and is a fundamental step in determining habitat quality for wildlife populations. Species occurrence and density tend to be disconnected from habitat quality when resources are highly seasonal, unpredictable over time, and patchy. In these environments, scale- specific spatial and temporal heterogeneity can often drive the habitat selection response.

The Florida Everglades is a highly seasonal, dynamic wetland that provides intermittent pulses of productivity to top predators such as wading birds, which are used as indicators of ecosystem health. I used a spatio-temporal, multi-scale distribution model (SDM) framework that estimated patch quality and abundance for Great Egrets (Ardea alba),

White Ibises (Eudocimus albus), and Wood Storks (Mycteria americana) and linked these indices to responses at the nest initiation and nest provisioning phases of the breeding cycle from 1993-2013. Predictor variables at a variety of temporal resolution

(daily-multiannual) captured spatio-temporal dynamics of ephemeral resources to demonstrate time-sensitive breeding responses to changes in habitat quality. The timing of increases and overall magnitude of resource pulses predicted by the selection based

72

model (i.e., patch quality) in March and April were strongly linked to breeding responses by all species. Great Egret nesting effort and success were higher when birds also showed greater conspecific aggregation. Wood Stork nesting effort was closely related to the timing of concurrently high levels of patch quality (regional scale) and abundance (400-m scale), indicating the importance of a multi-scaled approach. Responses to changes in timing of habitat quality at different phases of breeding can provide inferences to population-level changes that may result from restoration and/or climate change.

INTRODUCTION

Understanding how species are linked with their habitat, such as determining what resources and conditions are necessary for occupancy, is integral to managing wildlife populations. It is also important to outline the underlying mechanisms that relate habitat to survival and reproduction, termed habitat evaluation (Van Horne 1983). The complex process of habitat evaluation involves determining how patterns of habitat use indicate selection, how selection reflects preference, and how preference is shaped by differential fitness among habitat resources (Garshelis 2000). Therefore, habitat suitability (or quality) generally conveys the ability to sustain life and support population growth.

While controlled studies can reveal preferences by observing use by species being offered equal amounts of a differing resource, preference in natural settings must be inferred from patterns of observed use in environments with dynamic resource levels.

Static habitat selection models have been recently improved to describe changes in habitat preference with altered habitat availability (functional response in habitat selection; Mysterud & Ims, 1998). This development has allowed plasticity in foraging

73

behavior to be incorporated into models of systems with variation in the resource base

(Godvik et al. 2009).

In most habitat selection studies, the use of selected habitat is assumed to increase reproductive success and be reinforced by demographics. In practice, distributions (or densities) of animals are statistically correlated with environmental variables thought to be surrogates for features used by animals to select habitat. These features may indicate habitat attributes such as food availability, shelter from predators, or distance from anthropogenic structures. However, habitat selection is often unrelated to habitat quality in the case of an ecological (Gilroy et al. 2011, Hollander et al. 2011) or perceptual trap

(Patten and Kelly 2010). In these examples animals are unable to directly assess habitat quality and instead use environmental cues such as social attraction to reduce assessment time, consequently decoupling the habitat selection and fitness response. In these cases, habitat selection models can be quite misleading (Gaillard et al. 2010) or decline in their predictive power (Folmer and Piersma 2012).

Species occurrence and density tend to be disconnected from habitat quality when resources are highly seasonal, unpredictable over time, and patchy (Van Horne 1983).

When spatial and temporal variation in resources occurs, scale-specific heterogeneity of the environment can often drive the selection response (Gaillard et al. 2010).

Furthermore, habitat selection is hierarchical and individuals respond to limiting factors at multiple spatial (or temporal) scales (Johnson 1980). This might produce a gradient of selection responses reflecting trade-offs across multiple scales from year to year, with animals maximizing their fitness given the set of resources available. For example, within-year selection for a short-term process related to survival could compensate for a 74

limitation at a longer-term scale that drives reproductive success. The models resulting from such a short-term study would imply that the system be managed to increase the short-term process, potentially resulting in lower fecundity. Thus, long-term studies of habitat selection linked with measures of fitness are necessary to understand the trade- offs across resource gradients.

In the Everglades of Florida, a resource trade-off across temporal scales is exemplified by two species of wading birds (White Ibis [Eudocimus albus], Great Egret

[Ardea Alba]; Beerens et al. 2011) whose breeding populations are limited by food availability (Frederick and Spalding 1994, Ogden 2005, Herring et al. 2011) and availability of prey is driven by water depth and density of prey (Gawlik 2002). In this shallow, subtropical wetland system, length of inundation over periods of months and years increases the density of the prey base (Trexler 2010), whereas dry season drying and ponding of water over periods of weeks concentrates prey and is linked to wading bird foraging density (Russell et al. 2002). The Great Egret and White Ibis (hereafter egret and ibis) demonstrated spatial selection for the short-term drying process, but not the long-term hydroperiod process (i.e., prey production) in a year with lower prey availability and nesting success. However, in a year with higher prey availability and nesting success, both species selected foraging areas with higher prey production, but not the drying process (Beerens et al. 2011). Clearly, long and short term processes both limit prey availability to the birds, creating a hierarchy of temporally nested processes at the landscape scale. Understanding the interaction between these processes and the response of species distributions requires a dataset that includes varying combinations of resource levels over multiple years. 75

Many studies report selection and preference from observations of use, but few have linked selection of specific resources to fitness (Whitham 1980, Hollander et al.

2011), especially in dynamic environments where resources change daily. Thus, there is much impetus to establish a direct relationship between shifting patterns of observed use and measures of fitness (Mosser et al. 2009, Nielsen et al. 2010). This is significant in the Everglades because management recommendations from habitat selection studies guide long-term (30 yr) and large-scale (cf 4,000 km2) restoration projects.

The Wading Bird Distribution Evaluation Models (WADEM) were developed as a framework for isolating and modeling the spatial and temporal responses of egret, ibis, and Wood Stork (Mycteria americana; hereafter stork) distributions over nearly a decade of environmental variation (see Chapter 3). WADEM utilizes long-term observational records combined with environmental predictors of high temporal resolution specifically to capture changing preference and trade-offs across temporal scales. In WADEM, a spatial foraging conditions model (SFC) predicts wading bird abundance over time at a fixed spatial scale and a temporal foraging conditions model (TFC) predicts abundance across space at a fixed temporal scale. These two approaches are necessary because wading bird occurrence patterns are highly variable in time and space as individuals track changing seasonal and annual locations of high quality foraging patches. In addition, the indices resulting from the two models represent proxies for different components of patch dynamics. Patch quality within suitable depths is reflected by TFC and landscape patch abundance by SFC. In a seasonal wetland, patch abundance is expected to increase to a maximum when the greatest area is within a species’ suitable depth range, and decrease as the landscape dries. In contrast, patch quality is expected to continually increase as 76

patches with longer hydroperiods, and thus higher prey density, become available within suitable depths (Chapter 3). The product of these two indices (area × quality; or foraging index [FI]) provides a metric to account for both processes.

In the current study, I use the TFC, SFC and FI to predict annual nesting effort by all species over a period of 21 years and nesting success for egrets and ibises over a period of 13 years. I predicted that the strong interactions among temporal scales, such as the requirement for high prey density in shallow water, would drive variation in both responses. Further, I expected that the timing of peaks in patch quality and abundance during key periods of the breeding cycle would be linked to responses at different phases of breeding. I also predicted that dispersion of foraging individuals would be an important consideration, since high patch quality might result in strong clustering of individuals (individuals/flock) in preferred habitat, whereas large numbers of patches might result in a low ratio. I expected the individual/flock ratio to vary between species as foraging conditions improve during the dry season and for this clustering to be related to peak nesting effort and success at a certain optima.

METHODS

Study Area

The Florida Everglades is a dynamic subtropical wetland subject to rapid seasonal spatial and temporal resource pulses (Frederick et al. 2009). Wading birds time their breeding to coincide with these resource pulses; however, their populations have declined by an estimated 70% since the 1930s (Crozier and Gawlik 2003) because of habitat reduction and hydrological alteration. In contrast to historically high volumes of sheet

77

flow traversing the wetland, current water flows are now considerably reduced because of peripheral drainage and regulation by water managers through a set of management units.

These practices have altered the location, seasonal timing, and magnitude of prey concentrations relative to traditional colony locations (Ogden 2005) and placed spatio- temporal limitations on the range of climatic conditions that increase populations

(Fleming et al. 1994). As recent as the 1990s, there were significant reductions in pulses of productivity because of sustained high water conditions, more detrimental to populations of searcher species (i.e., ibises and storks) that rely on higher prey concentrations (Gawlik et al. 2002). Thus, a goal of on-going management and restoration of the Everglades is to provide ecological functions more similar to the historical system, signified by a growing wading bird population (Frederick et al. 2009).

Wading Bird Distribution Evaluation Models (WADEM)

From 1985-2012, Systematic Reconnaissance Flights (SRF) have been consistently used to document the abundance, flock composition, and spatiotemporal distribution of foraging wading birds across the Greater Everglades system (Water

Conservation Areas, Big Cypress National Park and Everglades National Park). From

January–June, during low altitude (61m) flights, observers estimated numbers and species of birds in belt transects spaced at 2-km intervals (Bancroft et al. 1994). To develop the

WADEM, SRF occurrence data from 2000–2009 were paired with daily hydrological variables calculated from water depths generated by the Everglades Depth Estimation

Network (EDEN). The EDEN is an integrated 400-m grid network of real-time water- level monitoring, ground elevation data, and surface water modeling that provides

78

estimated water depth information for the entire freshwater portion of the greater

Everglades (Telis 2006).

To predict daily landscape-wide wading bird flocks and abundance, the TFC model used daily mean hydrological characteristics calculated regionally throughout the

Everglades. Three explanatory variables representing hydrological conditions across a gradient of temporal scales were used as proxies for prey dynamics: days since drydown

(DSD) was used as an indicator for long-term prey production (Trexler 2010), recession rate was used as an indicator for prey concentration dynamics (Russell et al. 2002,

Beerens et al. 2011), and daily water depth was used as an indicator of short-term prey availability (Gawlik 2002, Beerens et al. 2011). These temporally-specific “resources” were considered available only when cell depths were in the foraging depth range of each species (see Chapter 3). The spatially explicit mean levels and heterogeneity (SE) of each resource were used to predict resource selection, which subsequently was used to predict the abundance of flocks and individuals across the landscape. Flock presence was defined as one or more birds of the target species (e.g., egrets, ibises, and storks) detected in a cell, whereas individual abundance counted the total number of birds present. Both individual and flock responses were modeled because wading birds are highly social and select foraging habitat based in part on the presence of conspecifics, a process that may increase or decrease individual fitness (Campomizzi et al. 2008). For the TFC, daily output summed over each region (Water Conservation Area (WCA) 1, WCA-2, WCA-

3N, WCA-3S, Big Cypress National Park (BCNP), and Everglades National Park (ENP);

Fig. 4.1) represents the mean patch quality of the landscape, within the suitable water depths of each species. 79

The SFC used a different approach by grouping foraging observations over time at a fixed cell and integrating spatial dynamics (e.g., topography) unaccounted for by the chosen set of hydrological predictors (i.e., spatial correlation; Chapter 3). By accounting for patterns in the spatial variation of the landscape, the noise independent of the hydrologic parameters can be reduced to better capture the species-specific behavioral response to rapidly changing habitat conditions (Dormann 2007). In addition to depth, recession, and DSD, the SFC used dry-to-wet reversal, hydroperiod, and Cartesian position, averaged over each instance of cell use. The dry-to-wet reversal variable estimated when a cell had gone dry and rewet, that results in highly depleted fish populations (Trexler et al. 2002). Hydroperiod defined the 10-year mean in the annual length of cell inundation, which influences wading bird distributions through long-term changes in microtopography and vegetation communities (Gunderson 1994). Frequency of cell use was defined as the number of times over the study period that a species used a given cell. Output from the SFC averaged over the landscape can serve as a surrogate measure of the abundance of high-quality patches.

In both the TFC and SFC, interaction terms among depth, recession rate, and DSD quantified a common trade-off in aspects of prey availability to birds: the tendency of the wetland system to produce prey through spatial immigration and reproduction over long periods of inundation (>6 months; DeAngelis et al. 2005) versus the shorter term (1-2 week) tendency of prey to become concentrated into pools and shallow areas through drying trends.

80

Nesting Effort

Egret, ibis, and stork maximum nesting effort (numbers of nest starts – nests having at least one egg) from 1993 – 2013 was obtained from annual South Florida

Wading Bird Reports (South Florida Water Management District, West Palm Beach, FL) and Crozier and Gawlik (2003). Nesting effort was monitored annually throughout the study area by systematic aerial colony search and survey efforts (February–June, monthly) performed at 240m altitude by two observers in overlapping east-west transects, through targeted visits to colonies, and through systematic ground surveys by airboat (see

Frederick and Ogden 2003 for further details of the monitoring program).

Nesting Success

Nest success data was provided by P. Frederick (Univ. of Florida, unpublished data). Nest success was monitored by checking individually identifiable nests every 5 – 7 d during 1993 – 2013. Nest success was defined as the probability of any nest producing one or more young to age 14 or 21 days for ibises and egrets, respectively. Sample size for stork nesting success was not robust enough for similar analysis. Colonies monitored in each year were selected based on large size (largest 3-5 colonies), species composition, and geographic representation. Many colonies are occupied by several species, and not all colony locations are active in each year (Frederick and Spalding 1994). Nests for study were marked within colonies along 4-m wide belt transects oriented from the colony edge to areas of greatest nest density, marking all nests within the belts with numbered surveyors flagging. Colonies were monitored from the time most nests had progressed to incubation until all nests on the transects had either failed or produced young. On each visit, all nests were checked for contents. Nests were identified to 81

species based on construction materials, size, and egg and chick characteristics

(McVaugh 1972). Nest start date was defined as the date of laying of the first egg, determined based on either laying or hatching schedule. Nests were assumed to have failed when all eggs or chicks disappeared or were found dead prior to the fledging age.

Barring more detailed evidence at the nest, timing of nest failure was assigned to the midpoint between nest checks. Nest success was expressed as a probability of the nest surviving to produce young of a predetermined age summarized over all nests of each species from all colonies within any breeding year (Mayfield 1961, 1975, Hensler 1985).

Variables & Statistical Methods

I applied the egret, ibis, and stork TFC and SFC to hydrological data calculated from the EDEN during the dry seasons (Jan-May) and species-specific nesting effort and nest success estimates during 1993-2013. Stork nesting effort was fourth-root transformed to meet assumptions of normality, whereas egret and ibis nesting effort did not require transformations to meet these assumptions. Daily regional outputs from the

TFC were summed (see Fig. 4.1), whereas cell outputs from the SFC were averaged to represent a daily landscape score for each model. The daily FI was calculated by multiplying the daily individual TFC by the daily SFC. A flocking index was obtained by dividing the individual TFC by the flock TFC to determine the mean number of individuals per flock on a given day. The flocking index was then averaged over the dry season to indicate the annual degree of foraging aggregation (Ratio) and its variation

(Ratio SE). The daily change in the individual TFC, calculated to determine whether the

TFC was increasing or decreasing, was averaged by month to indicate monthly changes in foraging conditions (Jan Δ, Feb Δ, Mar Δ, and Apr Δ). These changes were included 82

in the model set to focus on the time frame when adults and nestlings were more sensitive to fluctuations in habitat quality. In addition, the daily individual TFC and FI were averaged by month (Jan , Feb , Mar , and Apr ) to determine whether the timing of the absolute level in the TFC or FI were driving nesting effort and success. The mean dry season flock TFC, individual TFC, FI, and their variations (SE) were also tested to determine if overall annual patterns explained additional variation in nest effort and success.

Models predicting species-specific nesting effort included variables describing dry season means, SEs, and monthly changes and means (Jan, Feb, and Mar), and were analyzed using a generalized linear mixed model (GLMM) in SAS 9.2 (SAS Institute

2010a). The random effect Decade was included in the models to account for the much wetter hydrological regime that occurred in the study period from 1993-1999 (Fig. 4.2).

Models predicting species-specific nesting success included variables describing dry season means, SEs, and monthly changes and means (Feb, Mar, and Apr), and were analyzed using a generalized linear model (GLM) in SAS 9.2 (SAS Institute 2010c). A

Priori models were constructed and evaluated using Akaike’s Information Criterion for small sample sizes (AICc) to determine which models were most parsimonious (Burnham and Anderson and 2002). Delta AIC (Δi, Akaike’s Information Criterion) and AIC weights (ωi) were calculated from AICc values. Models with the lowest AICc value were considered the best explanatory models, although additional competing models with

ΔAICc < 2 were considered equally plausible given the data (Burnham and Anderson

2002). Models with ΔAICc > 4 were considered to have little to no support (Burnham and Anderson 2002). Model-averaged coefficients and standard errors (SE) were 83

calculated for each parameter by averaging all models containing the variable in proportion to the AIC weight. The importance of a specific variable was determined by summing the weights of all models containing that term.

RESULTS

Great Egret

Egret nesting effort weakly correlated with nesting success (R2 = 0.22, N = 14) because there were years with low nest effort, but high success (e.g., 1993-1995) and years with high nest effort, but low success (e.g., 2012-2013; Fig. 4.3). The lowest egret nesting effort of 2,308 nesting pairs was observed in 2008 and the highest nesting effort of 13,211 nesting pairs occurred in 2009. These two years also corresponded to the highest (0.72 ± 0.04) and lowest (0.00 ± 0.00) years of nesting success (Fig. 4.3). Year

2008 was characterized by low initial water levels and extreme reversals during the breeding period; whereas initial water levels in 2009 were high and a steady recession was maintained throughout the breeding period.

Nesting effort of egrets was best explained by the model that included the variables Ratio SE, Mar Δ TFC, and the FI SE. The second best model included the term

Apr TFC rather than Mar Δ TFC, but received 4× less support. A third model with 4× less support did not include the term for FI SE (Table 4.1). The random effect Decade caused the G-matrix to no longer be positive definite and thus was unrelated to variation in nesting effort after accounting for the fixed effects and was removed. Egret nesting effort was higher with increases in the variability of the flocking index, caused by steady clustering of individuals into a smaller number of flocks. An increased Mar Δ TFC, but

84

steady FI was also related to higher nesting effort. The Mar Δ TFC received less support

(0.52) than the FI SE (0.79) and Ratio SE (0.95), and this model explained 61% of the variation in egret nesting effort.

The most parsimonious model to explain the nesting success of egrets included the terms Ratio SE, Apr TFC, and the FI SE. This model is 1.3× more likely to explain the response than when removing the FI SE and 1.4× more likely than when substituting the TFC mean for the Apr TFC (Table 4.2). Similar to nest effort, egret nesting success increased with increases in the variability of the flocking index through steady clustering of individuals. Nesting success also increased with a high Apr TFC, but steady FI SE.

The sums of the variable weights indicate high variable importance for the Ratio SE

(0.97) and moderate importance for the Apr TFC (0.59) and FI SE (0.50), and this model explained 75% of the variation in egret nesting success (Table 4.2).

White Ibis

Ibis nesting success was more closely linked with nesting effort (R2 = 0.47, N =

13), but similar to the egret, ibises had the highest nesting effort (43,415 nesting pairs) and success (0.85 ± 0.4) in 2009 (Fig. 4.4). Nesting effort was lowest in 1993 (818 nesting pairs) and nesting success was lowest in 2010 (0.01 ± 0.01), with both years having minimal water level recession.

The model that best described ibis nesting effort contained the terms Mean TFC,

Mar Δ TFC, and Decade. This model is 5.5× more likely to explain nesting effort than when substituting the Mean TFC for the Mar TFC (Table 4.1). An overall negative effect on nesting effort of ibis was evident in the period from 1993-1999 in comparison to

2000-2013. Ibis nests increased with an increasing Mar Δ and high mean TFC, variable 85

importance was high for both of these (>0.7), and the top model explained 76% of the variance in nesting effort (Table 4.1).

The top model for ibis nesting success included the term Apr TFC and a second model with 7× less support included the term Mar Δ TFC (Table 4.2). Ibis nest success increased with a higher Apr TFC, which had a high variable importance (1.0), and this one term explained 80% of the variance in ibis nesting effort (Table 4.2).

Wood Stork

Similar to the other species, stork nesting effort was highest in 2009 with 4,063 nesting pairs, but the low of 25 nesting pairs occurred in 1998 (Fig. 4.5). The most parsimonious model to describe stork nest effort consisted of the terms Mar FI and

Decade (Table 4.1). This model is 6× more likely than the second best model with the additional term February FI. Similar to ibises, an overall negative effect of high water conditions was evident from 1993-1999 on stork nesting effort, in comparison to the more recent hydrological regime. Stork nests increased with a high Mar FI, when a peak in patch abundance (SFC) co-occurred with high patch quality (TFC).

DISCUSSION

This study provides the first empirical evidence that dynamic selection processes are linked with wading bird nesting effort and success, which collectively drive fecundity. Predictor variables at a variety of temporal resolutions (daily-multiannual) captured spatio-temporal dynamics of ephemeral resources to demonstrate time-sensitive breeding responses to changes in habitat quality (McPherson and Jetz 2007). Because wading birds have flexible breeding schedules (Heath et al. 2003) and are primarily

86

limited by food availability (Ogden 1994, Gawlik 2002, Herring et al. 2011), it is not surprising that nesting responses were linked to the timing of increases in patch quality and abundance. However, the fact that nesting effort is highly flexible suggests either that large sections of the population are not breeding in many years, or that annual breeding strategies by these highly mobile animals are formed by choices made at larger range-wide scales. Further, nesting responses within the Everglades corresponded to multiple spatial scales at the 400-m and landscape resolution indicating the importance of a multi-scaled SDM approach (Johnson et al. 2004, Flesch and Steidl 2010).

The timing of increases and overall magnitude of resource pulses predicted by the selection based model (TFC) were strongly linked to both ibis nesting effort and nesting success. Increases in patch quality in March, followed by sustained patch quality in April strongly predicted ibis nesting effort and success, respectively. Overall high patch quality from January-May provided an additional lift in nest effort. Because most chicks are fledged in May, it is expected that high patch quality in May is also critical to fledging success and survival (Frederick and Spalding 1994). Therefore, increasing patch quality in March and sustained patch quality in April-May would allow ibises to reproduce successfully in their relatively short nesting cycle of 60-80 days (Frederick et al. 2009). Ibises also provided a functional contrast between very wet conditions (middle

1990s) when ibis nesting effort and success was markedly low, and later years when nesting effort and success improved markedly through increases in the availability of prey. Indeed, the historical benchmark of a 1.6 year interval between exceptional ibis nesting events (>16,977 nesting pairs) has been achieved over the last 5 years (Frederick

2013). Results from this study suggest that a hydrological pattern of alternating wet and 87

dry years results in optimal wetland inundation of 450 days throughout the landscape (see

Chapter 3) that in the dry year is slowly exposed in suitable depths from drying. This fluctuating regime would increase crayfish populations through release of top down limitation by large predatory fishes (Kellogg and Dorn 2012) and also provide dense concentrations of small fishes as longer hydroperiod sites are exposed (Trexler 2010).

Similar to ibis, nesting effort and success was increased with increasing patch quality in March and high patch quality in April, respectively. However, variable importance was lower for these time-sensitive estimates than for metrics describing foraging conditions over the whole breeding season. In particular, egret nesting effort and success were higher in years when a high abundance of foraging egrets clustered into fewer flocks. This link demonstrates that an increase in conspecific attraction for species with higher interference costs can be used as a measure to describe increasing habitat quality (Folmer and Piersma 2012). Denser foraging aggregations under improving conditions could account for a decline in the predictive power of environmental variables

(area under the curve; AUC) that has been demonstrated in resource selection functions for egrets (Beerens et al. 2011).

After accounting for the above fixed effects, egret nest effort was not affected by the wetter conditions of the 1990s and nest success was high from 1993-1995. While patch quality remained relatively higher for egrets than ibises in wetter conditions, a lower flocking index was associated with the limited nesting attempts of the 1990s.

Conditions subsequently improved to support high nest success, but without high nesting effort, fewer chicks fledged (obtained by calculating nest effort × nest success).

Regardless, these wetter conditions still favored egrets because of their broader depth 88

tolerance and preference to exploit prey communities that develop over a longer period of inundation (up to 600 days; Chapter 3). Thus, the documented increase of egrets (a visual forager) relative to tactile foragers (ibises and storks) suggests habitat quality has declined for species with more specific habitat requirements that rely on higher prey concentrations (Gawlik 2002, Beerens et al. 2011).

For both egrets and ibises, breeding responses were more closely linked with patch quality than patch abundance. However, patch quality also captured resource heterogeneity (SD), a metric associated with increased choice of resources at the regional scale. For egrets, a stable foraging index (incorporating patch abundance) was related to increases in egret nesting effort and success. In an average year, patch abundance is expected to be low in January because of deeper landscape depths, rise to an optimum when a large portion of the landscape is within suitable depths, and decline as the landscape dries and exploited patches are no longer suitable. The egret foraging index remained stable when patch quality stayed high while patch abundance was declining.

Egret patch quality remained high in dryer conditions only when foraging sites had remained wet over multi-annual cycles, suggesting egrets are at a disadvantage when feeding in shallow depths following a drought year.

Stork nesting effort was closely related to the timing of concurrently high levels of patch quality and abundance. In contrast to egret and ibis nesting effort, which responded to increasing patch quality in March, patch quality (and patch abundance) in

March must both be high for storks to initiate nesting in large numbers. In addition, stork nesting effort was markedly lower in the 1990s because of wetter conditions and lack of early season foraging habitat within suitable depths. While stork nest success could not 89

be modeled, I suspect that a sustained high foraging index is crucial to reproductive success, especially in this species because storks have a very long reproductive cycle

(e.g., >110 days, Kahl 1964). Indeed, many years with high nesting success in storks

(Frederick and Fontaine 1999, Frederick and Hilburn 2002, Simon et al. 2006) also had a high foraging index in April and May, whereas low scores in April and May paralleled increased abandonment of chicks (Frederick and Battaglia 1997, Frederick et al. 2003,

Simon et al. 2005, Frederick and Simon 2008, 2010, 2012).

The results of this study indicate that the timing of patch abundance and quality may also have a profound effect on stork populations in the study area. Stork nest initiation date has shifted from November-December (1930-1960s) to January-March

(1980s-present; Frederick et al. 2009) because of the loss of early dry season habitat (i.e., short hydroperiod marsh; Fleming et al. 1994). As a result of late initiation, stork nesting now routinely continues into the beginning of the wet season, when food resources become very scarce due to rising water levels, and stork colonies typically abandon en masse, and fledging chicks have very low survival (Ogden 1994, Borkhataria et al. 2012).

Also, the maximum reproductive effort pre-1969 was 4000-5000 birds, whereas maximum reproductive effort following 1969 never exceeded 2500 birds (Ogden 1994), with the exception of 2009 (Fig. 4.5), when conditions were highly favorable through

May 15 (Cook and Kobza 2009).

If the overall timing of peaks in patch abundance and high patch quality can be shifted earlier in the breeding season via restoration of historical hydrological patterns, it is expected that storks will nest earlier and be more likely to fledge successful offspring

(Fleming et al. 1994, Ogden 1994). Under current water management, early peaking of 90

patch abundance only occurs at the start of extremely dry years (e.g., 2001, 2007 and

2011) resulting in a concomitantly reduced patch abundance during the critical later phases of breeding when the surface of the marsh is dry. Increasing the early season abundance and quality of short hydroperiod marsh in typically wetter years would slightly lengthen the breeding season and increase the time for stork reproduction before the rainy season. However, it is not expected to shift the nest initiation date to what it was historically because of the disproportionate reduction in short hydroperiod habitats that are not being restored.

Application

This study provides a modeling corollary for three out of the four wading bird performance measures that are used as indicators of Everglades restoration success: 1) the ratio of nesting ibises + storks to egrets, 2) the interval between years with exceptionally large ibis nesting events, and 3) the timing of nesting by storks (Frederick et al. 2009).

An analysis to predict the fourth performance measures (the proportion of all nests located in the estuarine/freshwater ecotone) is under development. Thus, hydrological input from restoration alternatives can be used to produce WADEM output to determine scenarios that best meet established nesting effort objectives and targets. There is also much impetus to determine restoration scenarios that produce high estimates of nest success in years with predictions of high nest effort (i.e., nest effort × nest success).

Additionally, WADEM output can be utilized to inform water management operations in real-time and determine the effects of climate change on wading bird habitat quality

(Catano et al. in review).

91

Patterns of species distributions in dynamic environments are often noisy, but incorporating components of species ecology such as flexible habitat selection of resources within and among temporal scales, responses to environmental gradients, conspecific attraction, and spatial autocorrelation can yield results that better approximate habitat quality. Furthermore, responses to changes in timing of habitat quality at different phases of breeding can provide inferences to population-level changes that may result from restoration and/or climate change.

92

LITERATURE CITED

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 615–658 in Everglades: The Ecosystem and

Its Restoration (S. M. Davis and J. C. Ogden, Eds.). St. Lucie Press, Delray

Beach, Florida.

Beerens, J. M., D. E. Gawlik, G. Herring, and M. I. Cook. 2011. Dynamic habitat

selection by two wading bird species with divergent foraging strategies in a

seasonally fluctuating wetland. The Auk 128:651-662.

Borkhataria, R. R., P. C. Frederick, R.A. Keller and J.A. Collazo. 2012. Temporal

variation in local wetland hydrology influences postdispersal survival of juvenile

wood storks (Mycteria americana). The Auk 129:517−528.

Burnham, K. P., and D. R. Anderson. 2002. Model Selection and Multimodel Inference:

A Practical Information-Theoretic Approach, 2nd ed. Springer-Verlag, New York.

Campomizzi, A. J., J. A. Butcher, S. L. Farrell, A. G. Snelgrove, B. A. Collier, K. J.

Gutzwiller, M. L. Morrison, and R. N. Wilkins. 2008. Conspecific attraction is a

missing component in wildlife habitat modeling. Journal of Wildlife Management

72:331–336.

Catano, C. P., S. Romanach, J. M. Beerens, L. Brandt, K. M. Hart, F. J. Mazzotti, L.

Pearlstine, and J. C. Trexler. In Review. Using scenario planning to evaluate the

93

impacts of climate change on wildlife populations and communities in the Florida

Everglades. Environmental Management.

Cook, M. I., and M. Kobza. 2009. System-wide Summary. Pages 1-2 in South Florida

Wading Bird Report, vol. 15 (M. I. Cook and M. Kobza, Eds.). South Florida

Water Management District, West Palm Beach, Florida.

Crozier, G. E., and D. E. Gawlik. 2003. Wading bird nesting effort as an index to wetland

system integrity. Waterbirds 26:303–324.

DeAngelis, D. L., J. C. Trexler, and W. F. Loftus. 2005. Life history trade-offs and

community dynamics of small fishes in a seasonally pulsed wetland. Canadian

Journal of Fisheries and Aquatic Sciences 62:781–90.

Dormann, C. F. 2007. Effects of incorporating spatial autocorrelation into the analysis of

species distribution data. Global Ecology and Biogeography 16:129–138.

Fleming, D. M., W. F. Wolff, and D. L. DeAngelis. 1994. Importance of landscape

heterogeneity of Wood Storks in Florida Everglades. Environmental Management

18:743–757.

Flesch, A. D., and R. J. Steidl. 2010. Importance of environmental and spatial gradients

on patterns and consequences of resource selection. Ecological Applications

20:1021–39.

Folmer, E. O., and T. Piersma. 2012. The contributions of resource availability and social

forces to foraging distributions: A spatial lag modelling approach. Animal

Behaviour 84:1371–1380.

94

Frederick, P. C. 2013. Status of wading bird recovery. Pages 35-38 in South Florida

Wading Bird Report, vol. 19 (M. I. Cook, Ed.). South Florida Water Management

District, West Palm Beach, Florida.

Frederick, P. C., and N. Vitale. 2012. Regional nesting reports: Water Conservation

Areas 2 and 3, and A.R.M. Loxahatchee National Wildlife Refuge. Pages 7-8 in

South Florida Wading Bird Report, vol. 18 (M. I. Cook and M. Kobza, Eds.).

South Florida Water Management District, West Palm Beach, Florida.

Frederick, P. C., and J. Simon. 2010. Regional nesting reports: Water Conservation Areas

2 and 3, and A.R.M. Loxahatchee National Wildlife Refuge. Pages 7-8 in South

Florida Wading Bird Report, vol. 16 (M. I. Cook and M. Kobza, Eds.). South

Florida Water Management District, West Palm Beach, Florida.

Frederick, P. C., D. E. Gawlik, J. C. Ogden, M. I. Cook, and M. Lusk. 2009. The White

Ibis and Wood Stork as indicators for restoration of the everglades ecosystem.

Ecological Indicators 9:S83–S95.

Frederick, P. C., and J. Simon. 2008. Regional nesting reports: Water Conservation Areas

2 and 3, and A.R.M. Loxahatchee National Wildlife Refuge. Pages 7-9 in South

Florida Wading Bird Report, vol. 14 (M. I. Cook and M. Kobza, Eds.). South

Florida Water Management District, West Palm Beach, Florida.

Frederick, P. C., and J. C. Ogden. 2003. Monitoring wetland ecosystems using avian

populations: Seventy years of surveys in the Everglades. Pages 321-350 in

Monitoring Ecosystems: Interdisciplinary Approaches for Evaluating Ecoregional

Initiatives (D. Bush and J. C. Trexler, Eds.). Island Press, Washington D.C.

95

Frederick, P. C., M. Schickel, J. Bartz, and B. Hylton. 2003. Regional nesting reports:

Water Conservation Areas 2 and 3, and A.R.M. Loxahatchee National Wildlife

Refuge. Pages 4-7 in South Florida Wading Bird Report, vol. 9 (G. E. Crozier and

D. E. Gawlik, Eds.). South Florida Water Management District, West Palm

Beach, Florida.

Frederick, P. C., and J. Hilburn. 2002. Regional nesting reports: Water Conservation

Areas and A.R.M. Loxahatchee National Wildlife Refuge. Pages 5-8 in South

Florida Wading Bird Report, vol. 8 (D. E. Gawlik, Ed.). South Florida Water

Management District, West Palm Beach, Florida.

Frederick, P. C., and P. Fontaine. 1999. Regional nesting reports: Water Conservation

Areas 2 and 3. Page 3 in South Florida Wading Bird Report, vol. 5 (D. E. Gawlik,

Ed.). South Florida Water Management District, West Palm Beach, Florida.

Frederick, P. C., and D. Battaglia. 1997. Regional nesting reports: Water Conservation

Areas 2 and 3. Page 2-3 in South Florida Wading Bird Report, vol. 3 (D. E

Gawlik, Ed.). South Florida Water Management District, West Palm Beach,

Florida.

Frederick, P., and M. Spalding. 1994. Factors affecting reproductive success of wading

birds (Ciconiiformes) in the Everglades ecosystem. Pages 659-692 in Everglades,

the Ecosystem and its Restoration (S. M. Davis and J. C. Ogden, Eds.). St. Lucie

Press, Delray Beach, FL.

Gaillard, J. M., M. Hebblewhite, A. Loison, M. Fuller, R. Powell, M. Basille, and B. Van

Moorter. 2010. Habitat-performance relationships: finding the right metric at a

96

given spatial scale. Philosophical transactions of the Royal Society of London.

Series B, Biological Sciences 365:2255–65.

Garshelis, D. L. 2000. Delusions in habitat evaluation: measuring use, selection, and

importance. Pages 111-164 in Research techniques in animal ecology:

Controversies and consequences (L. Boitani, and T. K. Fuller, Eds.). Columbia

University Press, New York.

Gawlik, D. E. 2002. The effects of prey availability on the numerical response of wading

birds. Ecological Monographs 72:329–346.

Gilroy, J. J., G. Q. A. Anderson, J. A. Vickery, P. V. Grice, and W. J. Sutherland. 2011.

Identifying mismatches between habitat selection and habitat quality in a ground-

nesting farmland bird. Animal Conservation 14:620–629.

Godvik, I. M. R., L. E. Loe, J. O. Vik, V. R. Veiberg, R. Langvatn, and A. Mysterud.

2009. Temporal scales, trade-offs, and functional responses in red deer habitat

selection. Ecology 90:699–710.

Heath, J. A., P. C. Frederick, T. M. Edwards, and L. J. Guillette. 2003. Reproductive

physiology of free-living White Ibises (Eudocimus albus) in the Florida

Everglades. General and Comparative Endocrinology 133:118–131.

Hensler, G. L. 1985. Estimation and comparison of functions of daily nest survival

probabilities using the Mayfield method. Statistics in Ornithology. Springer-

Verlag, New York.

Hollander, F. A., H. Van Dyck, G. San Martin, and N. Titeux. 2011. Maladaptive habitat

selection of a migratory passerine bird in a human-modified landscape. PloS one

6:e25703. 97

Johnson, C. J., D. R. Seip, and M. S. Boyce. 2004. A quantitative approach to

conservation planning: using resource selection functions to map the distribution

of mountain caribou at multiple spatial scales. Journal of Applied Ecology

41:238–251.

Johnson, D. H., and N. Prairie. 1980. The Comparison of Usage and Availability

Measurements for Evaluating Resource Preference. Ecology 61:65–71

Kellogg, C. M., and N. J. Dorn. 2012. Consumptive effects of fish reduce wetland

crayfish recruitment and drive species turnover. Oecologia 168:1111–21.

Mayfield, H. F. 1961. Nesting success calculated from exposure. Wilson Bulletin 73:255-

261.

Mayfield, H. F. 1975. Suggestions for calculating nest success. Wilson Bulletin 87:456-

466.

McPherson, J. M., and W. Jetz. 2007. Effects of species ecology on the accuracy of

distribution models. Ecography 30:135–151.

McVaugh, W., Jr. 1975. The development of four North American herons. Living Bird

Quarterly 14:163-183.

Mosser, A., J. M. Fryxell, L. Eberly, and C. Packer. 2009. Serengeti real estate: density

vs. fitness-based indicators of lion habitat quality. Ecology Letters 12:1050–60.

Mysterud, A., and R. Ims. 1998. Functional responses in habitat use: availability

influences relative use in trade-off situations. Ecology 79:1435–1441.

Nielsen, S. E., G. McDermid, G. B. Stenhouse, and M. S. Boyce. 2010. Dynamic wildlife

habitat models: Seasonal foods and mortality risk predict occupancy-abundance

and habitat selection in grizzly bears. Biological Conservation 143:1623–1634. 98

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 Everglades, the Ecosystem and its Restoration. (S.

M. Davis and J. C. Ogden, Eds.) St. Lucie Press, Delray Beach, FL

Ogden, J. C. 2005. Everglades ridge and slough conceptual ecological model. Wetlands

25:810-820

Patten, M. A, and J. F. Kelly. 2010. Habitat selection and the perceptual trap. Ecological

Applications 20:2148–56.

Russell, G. J., O. L. Bass, and S. L. Pimm. 2002. The effect of hydrological patterns and

breeding-season flooding on the numbers and distribution of wading birds in

Everglades National Park. Animal Conservation 5:185–199.

SAS Institute. 2010a. SAS 9.2 Help and Documentation. MIXED Procedure. SAS

Institute, Cary, North Carolina.

SAS Institute. 2010b. SAS 9.2 Help and Documentation. GLM Procedure. SAS Institute,

Cary, North Carolina.

Simon, J., P. C. Frederick, and R. Borkhataria. 2006. Regional nesting reports: Water

Conservation Areas 2 and 3, and A.R.M. Loxahatchee National Wildlife Refuge.

Pages 5-7 in South Florida Wading Bird Report, vol. 12 (M. I. Cook and E.M.

Call, Eds.). South Florida Water Management District, West Palm Beach, Florida.

Simon, J., P. C. Frederick, and R. Borkhataria. 2005. Regional nesting reports: Water

Conservation Areas 2 and 3, and A.R.M. Loxahatchee National Wildlife Refuge.

Pages 5-7 in South Florida Wading Bird Report, vol. 11 (M. I. Cook and E.M.

Call, Eds.). South Florida Water Management District, West Palm Beach, Florida. 99

Telis, P. A. 2006. The Everglades Depth Estimation Network (EDEN) for support of

ecological and biological assessments. U.S. Geological Survey Fact Sheet 2006-

3087.

Trexler, J. 2010. Prey-based freshwater fish density performance measure. RECOVER

greater Everglades performance measure documentation sheet.

http://www.evergladesplan.org/pm/recover/recover_docs/perf_measures/ge_pp_fi

sh_pm_051 611.pdf

Van Horne, B. 1983. Density as a misleading indicator of habitat quality. The Journal of

Wildlife Management 47:893–901.

Whitham, T. 1980. The theory of habitat selection: examined and extended using

Pemphigus aphids. American Naturalist 115:449–466.

100

Figure 4.1. South Florida study system displaying Everglades hydrological basins

(regions). The regions of coverage include Water Conservation Areas (WCA) 1, 2, 3-

North, 3-South, Big Cypress National Park (BCNP), and Everglades National Park

(ENP). 101

Figure 4.2. Hydrograph depicting Everglades Depth Estimation Network (EDEN) mean water depths (cm) during 1993–1999 and 2000–2013.

102

Figure 4.3. Great Egret nest effort (i.e., max nesting pairs) and success estimates (± SD) from 1993–2013.

103

Figure 4.4. White Ibis nest effort (i.e., max nesting pairs) and success (± SD) estimates from 1993–2013.

104

Figure 4.5. Wood Stork nest effort (i.e., max nesting pairs) and success (± SD) estimates from 1993–2013.

105

Table 4.1. Ranking of candidate models describing variables influencing nesting effort

(i.e., max nesting pairs) of Great Egrets, White Ibises, and Wood Storks in the Florida

Everglades (Proc Mixed). Models are ranked by differences in Akaike’s information criterion and only candidate models within ΔAICcd ≤ 4.0 are presented. Model selection results are followed by model averaging results for each species. The R2 represents the model fit for the estimated annual nest effort vs. model averaged predicted values.

b c d e 2 GREAT EGRET NESTING EFFORT K AICc modelid ΔAICc wi R Ratio SE, Mar Δ, FI SE 5 388.2 24 0.00 0.41 0.61 Ratio SE, Apr TFC, FI SE 5 391.0 9 2.76 0.10 Ratio SE, Mar Δ 4 391.0 11 2.78 0.10

Variable N Avg PE SE Importance

Intercept 27 2703.120 2195.373 1.00 Ratio SE 12 130485.162 28392.37 0.95 FI SE 8 -0.561 0.266 0.79 Mar Δ 9 0.277 0.195 0.52

b c d e 2 WHITE IBIS NESTING EFFORT K AICc modelid ΔAICc wi R Mean TFC, Mar Δ 5 445.2 7 0.00 0.61 0.76 Mar TFC, Mar Δ 6 448.5 10 3.36 0.11

Variable N Avg PE SE Importance

Intercept 27 -9736.618 13127.94 1.00 Mar Δ 8 29.225 12.62 0.82 Mean TFC 9 2.425 1.01 0.78

b c d e 2 WOOD STORK NESTING EFFORT K AICc modelid ΔAICc wi R Mar FI 4 65.2 22 0.00 0.74 0.73 Feb FI, Mar FI 5 68.7 8 3.41 0.13

Variable N Avg PE SE Importance

Intercept 27 1.263 1.21 1.00 Mean FI 9 0.003 0.00 0.98

106

Table 4.2. Ranking of candidate models describing variables influencing nesting success

(Mayfield method) of Great Egrets and White Ibises in the Florida Everglades (Proc

GLM). Models are ranked by differences in Akaike’s information criterion and only candidate models within ΔAICcd ≤ 4.0 are presented. Model selection results are followed by model averaging results for each species. The R2 represents the model fit for estimated annual nest effort vs. model averaged predicted values.

b c d e 2 GREAT EGRET NESTING SUCCESS K AICc modelid ΔAICc wi R Ratio SE, Apr TFC, FI SE 5 3.8 27 0.00 0.28 0.75 Ratio SE, Apr TFC 4 4.3 9 0.41 0.22 Ratio SE, Mean TFC, FI SE 5 4.4 4 0.46 0.22

Variable N Avg PE SE Importance

Intercept 27 -0.716 0.35 1.00 Ratio SE 12 10.145 2.66 0.97 Apr TFC 8 0.000 0.00 0.59 FI SE 8 -0.000 0.00 0.50

b c d e 2 WHITE IBIS NESTING SUCCESS K AICc modelid ΔAICc wi R Apr TFC 3 -12.9 27 0.00 0.78 0.80 Apr TFC, Mar Δ 4 -9.0 23 3.84 0.11

Variable N Avg PE SE Importance

Intercept 27 -0.409 0.12 1.00 Apr TFC 8 0.000 0.00 1.00

107

CHAPTER 5: CENTRAL EVERGLADES PLANNING PROJECT APPLICATION

INTRODUCTION

The goal of the Central Everglades Planning Project (CEPP) is to deliver a finalized plan for a suite of restoration projects in the central Everglades as part of the

Comprehensive Everglades Restoration Plan (CERP; USACE 2014). The overall intent in formulating CEPP alternative plans is restoration to more natural water flow, depth, and durations into and within the central Everglades by 1) increasing storage, treatment and conveyance of water south of Lake Okeechobee, 2) removing canals and levees within the central Everglades, and 3) retaining water within Everglades National Park and protect urban and agricultural areas to the east from flooding (USACE 2014). It is expected that these changes will result in a more natural mosaic of habitat conditions in

Water Conservation Area 3 (WCA-3), Everglades National Park (ENP), and Florida Bay.

A tentatively selected plan (TSP) was determined by evaluating scenarios that maximized the above objectives while controlling for cost (IMC 2014).

Using the wading bird distribution evaluation models (WADEM), my objectives were to 1) determine predicted changes in the timing of high quality foraging for wading birds by evaluating the TSP relative to the baseline (future without restoration; FWO) and

2) discuss how these changes may impact wading bird nesting effort and success.

108

METHODS

Alternative Restoration Scenarios

Hydrological restoration scenarios were developed using the Regional System

Model (RSM) version 2.3.5 (SFWMD 2005) using climate inputs from 1965-2005. For a detailed description of this model, alternative scenarios, and how the TSP was selected see USACE (2013) and IMC (2014).

Wading Bird Distribution Evaluation Models (WADEM)

I used a spatio-temporal species distribution model (SDM) framework to evaluate the foraging responses of Great Egrets, White Ibises, and Wood Storks to the TSP

(Alternative 4r2) relative to the FWO scenario. Using a multi-model approach, a spatial foraging conditions model (SFC) predicted wading bird abundance over time at a fixed spatial scale (400-m) and a temporal foraging conditions model (TFC) predicted daily abundance across space (Chapter 3). The resulting indices represent proxies for different components of patch dynamics: patch quality within suitable depths is reflected by TFC and landscape patch abundance by SFC. The product of these two indices (area × quality; or foraging index [FI]) provides a metric to account for both processes.

To evaluate the effects of the selected suite of restoration projects (i.e., the TSP) on wading bird patch quality and patch abundance, I calculated the cumulative percent change in the daily TFC, SFC, and FI during the breeding months of Jan-May, for the

TSP relative to the FWO. I also calculated the mean values of each foraging metric by month for each scenario to determine changes in the timing of high quality foraging conditions and mapped the cumulative percent change over each cell (SFC) to determine spatial changes over the length of the simulation. These estimates were then interpreted 109

using the factors that predict wading bird nesting effort and nesting success (Chapter 4) to determine how restoration may impact measures of reproduction.

RESULTS

Great Egret

For the Great Egrets, there was an increase in all foraging metrics for the TSP relative to the baseline, with the exception of the flocking index. Great Egret flock TFC increased ~850%, whereas individual TFC increased ~650% across the 37 year simulation (Fig. 5.1). This suggests that with restored hydrology, there will be more

Great Egrets, but their distributions will not be as clumped. Currently, increases in aggregation are linked with Great Egret nesting effort and success, but with higher patch quality this relationship may not persist. Great Egrets aggregate under shallow water conditions, which are reduced in the TSP, partly explaining the decrease in the flocking ratio. Similar to TFC, SFC increased in the TSP over the simulation (Fig. 5.2), primarily in northern Water Conservation Area (WCA) 3A, WCA-3B, and north-east Everglades

National Park (ENP; Fig. 5.3). Increases in Great Egret nesting effort and success are linked to increases in TFC in March and high TFC in April (Chapter 4). Under the TSP,

TFC are higher in every month, most noticeably in the earlier months of January,

February, and March (Fig. 5.4), which could drive increases in reproductive responses.

White Ibis

White Ibis individual TFC increased ~600% for the TSP relative to the baseline, but flock TFC only increased slightly over the 37 year simulation (Fig. 5.5). Similar to the Great Egret, the individual TFC was higher in all months, with the most noticeable

110

increases in January, February, and March (Fig. 5.6). These increases in TFC during critical periods of the breeding cycle are linked with both nesting effort and success and suggest that the White Ibis breeding populations may benefit from restored hydrology.

The White Ibis SFC increased the least of the three species under the TSP (Fig. 5.2), demonstrating that the new spatial configuration of patches may provide a greater benefit to species that forage more frequently in longer hydroperiods (e.g., Great Egrets and

Wood Storks; Chapter 3). However, this metric was not linked with breeding and SFC increases were still predicted in north-west WCA-3A and eastern ENP (Fig. 5.7).

Wood Stork

For Wood Storks, there was a slight decrease in both individual TFC (~50%) and flock TFC (~150%) over the 37 year simulation (Fig. 5.8); however, these decreases were offset by the increases in SFC (Fig. 5.2) in northern WCA-3A, WCA-3B, and large portions of ENP (Fig. 5.9). As a result, the predicted FI for Wood Storks was slightly higher for the TSP in January, February, and March (Fig. 5.10). While increased nesting effort is linking to an increased FI in March, the predicted benefit from the TSP could promote earlier nesting initiation, and shift the relationship between peak nesting effort and the FI to an earlier month.

Summary

Changes in foraging responses by all species suggest that wading bird reproduction over multiple phases of breeding may benefit from the selected restoration scenario. Models predicting nesting responses were not directly run with output from the foraging indices because it is expected that current relationships will change as the timing of patch quality and abundance change. Therefore, I encourage continued monitoring of 111

nesting effort and success throughout restoration projects to monitor the dynamic linkage between foraging metrics and nesting responses. Regardless, the WADEM framework provides a novel method of restoration planning to restore more natural hydropatterns by optimizing the location, timing, and duration of appropriate water levels for indicator species of wading birds.

112

LITERATURE CITED

Interagency Modeling Center. 2014. Central Everglades Planning Project (CEPP)

tentatively selected plan model documentation report (DRAFT). US Army Corps

of Engineers, Jacksonville, FL and the South Florida Water Management District,

West Palm Beach, FL.

South Florida Water Management District. 2005. Regional simulation model -

Hydrologic Simulation Engine (HSE) user manual. South Florida Water

Management District, West Palm Beach, FL.

US Army Corps of Engineers. 2014. Central Everglades Planning Project fact sheet,

http://www.evergladesplan.org/docs/fs_cepp_jan_2014.pdf. US Army Corps of

Engineers, Jacksonville, FL.

US Army Corps of Engineers. 2013. Central Everglades Planning Project (CEPP) draft

integrated Project Implementation Report (PIR) and Environmental Impact

Statement (EIS). US Army Corps of Engineers, Jacksonville, FL.

113

Figure 5.1. Cumulative mean percent change in Great Egret temporal foraging conditions

(TFC) for the tentatively selected plan (TSP), relative to the baseline (future without restoration; FWO) during the breeding months of Jan-May, 1967-2004.

114

Figure 5.2. Cumulative mean percent change in Great Egret, White Ibis, and Wood Stork spatial foraging conditions (SFC) for the tentatively selected plan (TSP), relative to the baseline (future without restoration; FWO) during the breeding months of Jan-May,

1967-2004.

115

Figure 5.3. The coloration in the map represents the mean percent change in Great Egret spatial foraging conditions (Jan – May, 1967-2004) for the tentatively selected plan (Alt

4r2) relative to future without restoration (FWO). Red demonstrates benefits from restoration, whereas blue represents losses.

116

Figure 5.4. Mean Great Egret temporal foraging conditions (TFC) for the tentatively selected plan (A4r2) and baseline (future without restoration; FWO) during the breeding months of Jan-May, 1967-2004.

117

Figure 5.5. Cumulative mean percent change in White Ibis temporal foraging conditions

(TFC) for the tentatively selected plan (TSP), relative to the baseline (future without restoration; FWO) during the breeding months of Jan-May, 1967-2004.

118

Figure 5.6. Mean White Ibis temporal foraging conditions (TFC) for the tentatively selected plan (A4r2) and baseline (future without restoration; FWO) during the breeding months of Jan-May, 1967-2004.

119

Figure 5.7. The coloration in the maps represents the mean percent change in White Ibis spatial foraging conditions (Jan – May, 1967-2004) for the tentatively selected plan (Alt

4r2) relative to future without restoration (FWO). Red demonstrates benefit from restoration, whereas blue represents losses.

120

Figure 5.8. Cumulative mean percent change in Wood Stork temporal foraging conditions

(TFC) for the tentatively selected plan (TSP), relative to the baseline (future without restoration; FWO) during the breeding months of Jan-May, 1967-2004.

121

Figure 5.9. The coloration in the maps represents the mean percent change in Wood Stork spatial foraging conditions (Jan – May, 1967-2004) for the tentatively selected plan

(A4r2) relative to future without restoration (FWO). Red demonstrates benefit from restoration, whereas blue represents losses.

122

Figure 5.10. Mean Wood Stork temporal foraging conditions (TFC) for the tentatively selected plan (A4r2) and baseline (future without restoration; FWO) during the breeding months of Jan-May, 1967-2004.

123

CHAPTER 6: CLIMATE CHANGE APPLICATION

INTRODUCTION

Significant alteration of the Everglades landscape has resulted in wading bird populations being more affected by 1) climatic variation that causes fluctuations in the production and availability of prey (Gawlik 2002) and 2) management strategies to protect human and ecological interests from flooding and droughts (Light and Dineen

1994). Restoration of historical hydrologic conditions is expected to improve wading bird prey availability (Chapter 5), but climate change increases the uncertainty of these benefits. Using the wading bird distribution evaluation models (WADEM), my goal was to evaluate wading bird responses across the Everglades under alternative climate scenarios with varying rainfall (RF) and evapotranspiration (ET) associated with increased air temperature.

METHODS

Alternative Climate Scenarios

Four environmental scenarios were created using the South Florida Water

Management Model (SFWMM; Obeysekera et al., in review) by adjusting observed hydrologic conditions from 1965 - 2005 across the Everglades based on relationships between hydrologic parameters and climatic variables. The SFWMM is a regional scale 124

model used for Everglades restoration planning that produces spatially explicit hydrologic data at a grid cell size of 3.2 x 3.2 km. The first scenario (BASE) is a baseline established on current demands and land use in 2010. The second scenario (+ET) simulated a 1.5˚ C temperature increase with a 30.5 cm sea level rise. The third (-RF+ET) and fourth (+RF+ET) scenarios simulated the same temperature and sea level rise with a

10% decrease and increase in precipitation, respectively. These scenarios were chosen to represent likely bounds of possible climate conditions affecting the Everglades. See

Obeysekera et al. (in review) for complete details of the SFWMM development and climate scenarios.

Preparation of the SFWMM scenarios

Water depth is a critical variable in many Everglades ecological models. The

SFWMM is often used to model Everglades hydrology; however, the spatial resolution of

3.2 x 3.2 km is too coarse to capture landscape heterogeneity that may be important to evaluations of species’ habitat. The Everglades Depth Estimation Network Digital

Elevation Model (EDEN-DEM, Jones and Price 2007) provides finer resolution topography that is used to calculate water depths at 500 x 500 m resolution used for all wildlife models other than fish. This was achieved by Delaunay triangulation (de Berg et al. 2000) of the SFWMM water stages and then subtracting the interpolated water stage surface from the EDEN-DEM ground elevation values.

Wading Bird Distribution Evaluation Models (WADEM)

I used a species-specific, spatially-explicit foraging conditions model (SFC) developed for evaluating hydrologic scenarios for Everglades restoration (see Chapter 3) to examine potential changes in Great Egret, White Ibis, and Wood Stork abundance 125

under the SFWMM climate scenarios. Interaction terms among depth, recession rate, and

DSD quantified a common trade-off in foraging response; the spatial immigration and reproductive development of the prey base over long periods of inundation (i.e., prey production) versus the spatial and temporal concentration of prey animals into intermediate depths over the shorter period of drying (i.e., prey concentration). These modifiers are important model inputs because wading birds show increasing selection for the shorter-term process of concentration to mitigate the loss of productive foraging habitat from a shorter period of inundation (Beerens et al. 2011). Final models predicting frequency of cell use from hydrologic and spatial characteristics were developed using

Proc Glimmix (SAS Institute) and evaluated for parsimony using Akaike’s Information

Criterion for small sample sizes (AICc; Burnham and Anderson 2002). When output from this model is averaged over the landscape, it can serve as a surrogate measure of the abundance of high-quality patches.

To evaluate the effects of potential climate change on wading bird distributions, I calculated the cumulative percent change in the daily SFC, during the breeding months of

Jan-May, between the baseline (BASE) and each of three SFWMM climate scenarios. I also produced habitat suitability maps for each of the four climate scenarios to determine the changes in patch abundance over the 38 year simulation.

RESULTS

Across all wading bird species, there was a slight negative response to the +ET scenario and a slight positive response to the +RF+ET scenario (relative to BASE; Fig.

6.1). Under –RF+ET scenario, drier conditions have a negative impact on the foraging

126

response of all bird species (Figures 6.2-4); particularly the Great Egret and Wood Stork which typically use deep water habitats. Additionally, any water loss through ET

(evapotranspiration) or reduced rainfall lowers landscape DSD, hydroperiods, and resulting prey production, such that prey density is not as high when depths are shallow.

127

LITERATURE CITED

Beerens, J. M., D. E. Gawlik, G. Herring, and M. I. Cook. 2011. Dynamic habitat

selection by two wading bird species with divergent foraging strategies in a

seasonally fluctuating wetland. The Auk 128:651–662.

De Berg, M., M. van Kreveld, M. Overmars, and O. Schwarzkopf. 2000. Computational

Geometry: Algorithms and Applications. Springer-Verlag, New York.

Jones J. W., S. D. Price. 2007. Initial Everglades Depth Estimation Network (EDEN)

digital elevation model research and development: U.S. Geological Survey Open-

File Report 2007-1034.

Obeysekera, J. In review. Predicting responses of the greater Florida Everglades to

climate change and future hydrologic regimes: climate sensitivity runs and

regional hydrologic modeling. Environmental Management.

128

Figure 6.1. Cumulative mean percent change in Great Egret (GE), White Ibis (WI), and

Wood Stork (WS) cell use for the simulations +RF+ET, +ET, and -RF+ET, relative to the baseline during the breeding months of Jan-May, 1967-2005.

129

Figure 6.2. Predicted mean Great Egret habitat suitability maps (1967-2005) for 4 climate scenarios (clockwise: BASE, +RF+ET, -RF+ET, +ET). Dark green represents the highest frequency of use, whereas dark blue represents the lowest. The area of high-quality habitat is reduced with decreasing modeled rainfall, with the largest loss occurring in the

-RF+ET scenario. 130

Figure 6.3. Predicted mean White Ibis habitat suitability maps (1967-2005) for 4 climate scenarios (clockwise: BASE, +RF+ET, -RF+ET, +ET). Dark green represents the highest frequency of use, whereas dark blue represents the lowest. The area of high-quality habitat is reduced with decreasing modeled rainfall, with the largest loss occurring in the

-RF+ET scenario. 131

Figure 6.4. Predicted mean Wood Stork habitat suitability maps (1967-2005) for 4 climate scenarios (clockwise: BASE, +RF+ET, -RF+ET, +ET). Dark green represents the highest frequency of use, whereas dark blue represents the lowest. The area of high- quality habitat is reduced with decreasing modeled rainfall, with the largest loss occurring in the -RF+ET scenario. 132