Duke University Dissertation Template

Home , Bay mud

TOP PREDATOR DISTRIBUTION AND FORAGING ECOLOGY IN FLORIDA BAY,

FLORIDA

Leigh Gabriela Torres

Department of Environment Duke University

Date:______Approved:

______Andrew J. Read, Supervisor

______Patrick N. Halpin

______Larry B. Crowder

______Michael R. Heithaus

______Hans Paerl

Dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Environment in the Graduate School of Duke University

2007

ABSTRACT

TOP PREDATOR DISTRIBUTION AND FORAGING ECOLOGY IN FLORIDA BAY,

FLORIDA

Leigh Gabriela Torres

Department of Environment Duke University

Date:______Approved:

______Andrew J. Read, Supervisor

______Patrick N. Halpin

______Larry B. Crowder

______Michael R. Heithaus

______Hans Paerl

An abstract of a dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Environment in the Graduate School of Duke University

2007

Abstract

The heterogeneous landscape of Florida Bay provides habitats for a variety of predators and prey. This dissertation examined the bottom‐up transfer of affects from environmental variability through prey composition up to competition and predation affects on top predator distribution and foraging ecology in Florida Bay. Line transect surveys for bottlenose dolphins and seabirds were conducted in Florida Bay during the summer months of 2002 ‐ 2005. Photo‐identification techniques were implemented to identify individual dolphins. Synoptic with this survey effort, habitat characteristics

(temperature, salinity, dissolved oxygen, turbidity, chlorophyll a, depth and bottom type) and prey composition (bottom trawl or gillnet) were sampled. Comparison of envelope maps from generalized additive models determined that predictive capacity of dolphin habitat did not improve by incorporating fish distribution data. However, models of dolphin distribution based solely on environmental proxies of fish distribution resulted in high predictive capacity. During the 2005 summer, shark distribution was sampled using a longline. The abundance of sharks was only correlated to fish catch from trawls on a regional scale. Larger sharks, of species that may threaten dolphins, were only caught in the Gulf zone of the Bay. Analysis of dolphin distribution revealed high individual site and foraging tactic fidelity. Dolphins were spatially coincident with habitat characteristics that encouraged the use of each individual’s

preferred foraging tactic. Depth was identified as the primary variable determining dolphin foraging tactic choice. Depth plays a significant role in the benthic composition of Florida Bay, which subsequently impacts prey communities and affects dolphin distribution, foraging and social ecology. Ordinations determined that fish distribution was also principally affected by depth and bottom type. Shallow environments frequently corresponded with mudbank habitat (depth < 1m) where the sighting rates of seabirds (cormorants, osprey, pelicans, terns) and foraging dolphins peaked. In conclusion, subtle relief in South Florida’s bedrock topography dramatically affect benthic composition within Florida Bay, providing patchy habitats for prey and predators. The Florida Bay ecosystem will change with expected sea level rise, including spatial shifts of mudbank habitats. Top predator populations in Florida Bay will be forced to modify their distribution and foraging ecology accordingly.

Dedication

Para La Gran Familia Torres

cerca y lejos

mayor y menor

antepasados y descendientes

Contents

Abstract ...... iv

Dedication ...... vi

List of Tables...... xi

List of Figures ...... xiii

Acknowledgements ...... xvi

Introduction ...... 1

Chapter 1: Influence of teleost abundance on the distribution and abundance of sharks in Florida Bay, USA...... 10

Introduction...... 11

Methods ...... 14

Study Site...... 14

Field Methods...... 16

Analysis...... 18

Results ...... 19

Discussion...... 24

Conclusions ...... 28

Chapter 2. Fine‐scale habitat modeling of a top marine predator: Does prey data improve predictive capacity? ...... 29

Introduction...... 30

Methods ...... 36

Study Site...... 36

vii

Field Methods...... 38

Benthic Habitat and Water Quality Sampling ...... 40

Fish Sampling ...... 43

Pseudo‐Absence Generation...... 44

Grid Sampling...... 47

Analysis...... 49

Generalized Additive Models (GAMs)...... 49

Mantel’s Tests...... 50

Predictive Maps...... 51

Model Evaluation ...... 51

Hypothesis Testing...... 54

Exploratory Exercise...... 54

Ho 1 ...... 57

Ho 2 ...... 62

Results ...... 63

Exploratory Exercise ...... 63

Ho 1...... 70

Ho 2...... 75

Discussion...... 80

Conclusions ...... 86

Chapter 3. How to catch a fish? The ecology of bottlenose dolphin (Tursiops truncatus) foraging tactic fidelity in Florida Bay, Florida...... 88

viii

Introduction...... 89

Methods ...... 92

Study Site...... 92

Field Methods...... 96

Foraging Behavior Observations ...... 97

Water Quality Sampling ...... 100

Analyses ...... 102

Results ...... 106

Ripley’s K ...... 107

Classification and Regression Tree ...... 110

ANOVA Tests...... 112

Individual Site & Foraging Tactic Fidelity...... 116

Discussion...... 122

Chapter 4: A kaleidoscope of mammal, bird and fish: Habitat use patterns of top predators in Florida Bay, Florida...... 131

Introduction...... 132

Methods ...... 135

Study Site...... 135

Field Methods...... 138

Benthic Habitat and Water Quality Sampling ...... 140

Species Diets...... 143

Fish Sampling ...... 146

Analysis...... 147

Results ...... 154

Discussion...... 170

Chapter 5: A ‘Trophin Fountain’: The bottom‐up synthesis of top predator distribution and foraging ecology in Florida Bay, Florida...... 178

References ...... 205

Biography...... 224

List of Tables

Table 1.1: Number of individual shark species caught in Florida Bay...... 20

Table 1.2: Abundance of fishes caught in trawls within each region in Florida Bay...... 21

Table 1.3: Influence of biotic & abiotic factors on catch rates of sharks in Florida Bay ....23

Table 2.1: A confusion matrix used to tabulate the predictive capacity...... 53

Table 2.2: Description of four types of GAMs created for each zone & tested in Ho 1...... 59

Table 2.3: GAM specifications of models used to test Ho 2...... 63

Table 2.4: Binomial Mantel tests p‐values on 2004 & 2005 dolphin presence/absence .....65

Table 2.5: Predictive capacity of dolphin distribution from 2005 survey data based on predictive maps created by the four tested model types in Ho 1 based on 2004 data for each zone...... 71

Table 2.6: Overall model performance by the four tested model types in Ho 1...... 75

Table 2.7: Ho 2 prediction results of 2003 dolphin presence/absence by 2002 zonal CPUE ~ ENV ‘GAMvelope’ models...... 76

Table 2.8: Overall predictive capacity results of dolphin distribution by zone for Ho 2...79

Table 3.1: Description of six environmentally distinct zones in Florida Bay ...... 96

Table 3.2: ANOVA results between environmental conditions at foraging tactics...... 113

Table 4.1: Environmental description and percent bottom type cover for six zones ...... 137

Table 4.2: Diets of predator species in Florida Bay based on literature review ...... 144

Table 4.3: Description of fish groups used in trawl NMS...... 152

Table 4.4: Description of fish groups used in gillnet NMS...... 153

Table 4.5: Three‐way ANOVA results for the sighting rate of each species group as associated with the factors zone, year and bottom type...... 156

Table 4.6: Pearson’s correlation coefficients between ordination axes and explanatory variables from 2D NMS ordination of the six predator species groups...... 160

Table 4.7: Pearson’s correlation coefficients between ordination axes and explanatory variables from 2D NMS ordination on trawl catch data...... 163

Table 4.8: Pearson’s correlation coefficients between ordination axes and explanatory variables from 2D NMS ordination of gillnet catch data...... 166

Table 5.1: Number of individual shark species caught in Florida Bay by zone...... 195

xii

List of Figures

Figure A: Location of Florida Bay at the southern tip of the Florida peninsula ...... 2

Figure 1.1: Location of shark and teleost sampling sites (*) within five regions of Florida Bay, USA...... 16

Figure 1.2: Influence of regional teleost abundance on shark catch rates...... 24

Figure 2.1: Schematic of a three level ecosystem between the environment, prey and predators...... 32

Figure 2.2: Bottom types and zones of Florida Bay...... 37

Figure 2.3: Survey areas and tracklines during 2004 and 2005 field seasons...... 40

Figure 2.4: Example ‘GAMvelope’ predictive map for the Atlantic zone 23‐July‐05...... 61

Figure 2.5: GAM plot comparison between significant explanatory variables for dolphin distribution and CPUE in the Atlantic, Central, and Gulf zones...... 68

Figure 3.1: Benthic habitat types and zones in Florida Bay ...... 94

Figure 3.2: Photograph of dolphins mud ring feeding in Florida Bay...... 100

Figure 3.3: Spatial distribution of 84 dolphin foraging events observed in Florida Bay...... 107

Figure 3.4: Ripley’s K plots describing the spatial distribution of foraging tactic groups...... 109

Figure 3.5: Classification and Regression Tree of three groups of foraging tactics employed by bottlenose dolphins in Florida Bay...... 111

Figure 3.6: Box plots depicting the distance from shore, depth, and group size range of each foraging tactic group...... 115

xiii

Figure 3.7: Zone and foraging tactic fidelity of 61 dolphins observed 5 times or more in Florida Bay ...... 118

Figure 3.8: Comparison of zone and foraging tactic fidelity by 64 dolphins observed at mud ring feeding events, and 138 dolphins observed at deep diving events...... 120

Figure 3.9: Group composition at all dolphin sightings in Florida Bay...... 122

Figure 4.1: Bottom types and zones of Florida Bay...... 138

Figure 4.2: Regions of concentrated survey effort during 2004 and 2005 field seasons and locations of fish sampling...... 140

Figure 4.3: Comparison of the distribution and density of seabird and dolphin sightings from 2003, 2004 and 2005 surveys in Florida Bay...... 155

Figure 4.4: Sighting rates of each species group by Bottom Type and Zone...... 157

Figure 4.5: Path diagram of Mantels tests on 2003, 2004 and 2005 species group presence/absence compared to environmental characteristics...... 159

Figure 4.6: Nonmetric multidimensional scaling (NMS) plots, showing the habitat preferences of the six predator groups examined ...... 161

Figure 4.7: Nonmetric multidimensional scaling (NMS) plot of trawl data, describing the distribution of 9 fish groups in habitat space...... 164

Figure 4.8: Distribution of trawl samples within the habitat space created by the 2D NMS on 9 fish groups, symbolized by Bottom Type and Zone ...... 165

Figure 4.9: Nonmetric multidimensional scaling (NMS) plot from 2D NMS of gillnet catch, showing the habitat preferences of catfish and jacks...... 167

Figure 4.10: GAM plots of mullet catch versus depth and bottom type ...... 169

Figure 5.1: Benthic habitat types and zones in Florida Bay...... 180

Figure 5.2: Schematic of dissertation research on the determinants of top predator distribution and foraging ecology...... 181

xiv

Figure 5.3: Sighting rates of each predator group by bottom type...... 187

Figure 5.4: Spatial distribution of 84 dolphin foraging events observed in Florida Bay...... 189

Figure 5.5: Photograph of dolphins mud ring feeding in Florida Bay...... 190

Figure 5.6: Box plots depicting the (a) distance from shore, (b) depth, and (c) group size range of each foraging tactic group...... 191

Figure 5.7: Zone and foraging tactic fidelity of 61 dolphins observed 5 times or more..193

Figure 5.8: Illustration of a ‘trophic fountain’ with bottom‐up affects on predator distribution in Florida Bay...... 201

Acknowledgements

There are so many people to thank for their help and support. I will start by thanking my advisor, Andrew Read, who challenged and inspired me to ask and answer thoughtful questions, not just interesting questions. You taught me what the word

“ecology” really means and demonstrated through example how to successfully merge science and conservation. I hope to follow your path and make you proud. I also want to acknowledge my entire committee (Pat Halpin, Mike Heithaus, Larry Crowder and

Hans Paerl) for their wise feedback and support of my research efforts. I especially want to thank Pat, for guiding me from when I could not make an arc file to when you named my analytical technique ‘GAMvelope’. And, Mike, your enthusiasm is infectious and you always responded with positive, critical and sage advice. You both managed to squeeze me into your busy schedules, contributing to my research and helping me find the stories in my data. I also must thank David Hyrenbach for encouraging me to include seabirds in my research and for superb encouragement and advice. Finally, I thank Dean Urban for answering a plethora of emails about spatial analyses.

Funding during my first two summers in Florida Bay (2002 & 2003) was provided by The Dolphin Ecology Project and The Florida Keys National Marine

Sanctuary. I thank Laura Engleby for her financial and logistical support, as well as the

xvi

lessons I learned. Funding and support for the last two years of my field research (2004

& 2005) was provided by the NOAA/Coastal Ocean Program. Data was collected under

Everglades National Park permits EVER‐2002‐SCI‐0049, EVER‐2003‐SCI‐0042, EVER‐

2004‐SCI‐0062, and EVER‐2005‐SCI‐0062, General Authorizations 911‐1466 and 572‐1639, and approved by the Duke University Institutional Animal Care & Use Committee

(IACUC). Data were provided by the SERC‐FIU Water Quality Monitoring Network which is supported by SFWMD/SERC Cooperative Agreement #C‐15397 as well as EPA

Agreement #X994621‐94‐0. Mullet data was supplied by the marine fisheries‐ independent monitoring program of the Florida Fish and Wildlife Conservation

Commission. I sincerely thank the Everglades National Park for supporting my research.

Additionally, I thank The Keys Marine Lab and Steve Lohmayer for their field assistance.

The days were long, the sun burned, the mosquitoes were thick, thunderstorms abounded, the green heads bit, the YSI was heavy, the trawl got tangled, and the gillnet tore. But, the water was clear, the sights beautiful and remote, the fish colorful and plentiful, and the dolphins & birds came and fed. I thank the following people for assisting me with my field work and sharing the adventures: Michelle Barbierie, Todd

Chandler, Kate Freeman, Larry Fulford, Gretchen Lovewell, Erica Morehouse, Liz Tuoy‐

Sheen, Anne Starling, Danielle Waples, Margaret Worthington, and Mango.

xvii

When I was not in Florida Bay, I was in Beaufort, North Carolina where another network of people supported me. First, I want to thank the faculty, staff and students of the Duke Marine Laboratory. During my eight years here I have grown in so many ways due to the unique, collegial environment of the lab that fosters creativity, conservation and intellectual growth. In particular, I thank the Readlab, those who inspired my path

(Heather Koopman, Andrew Westgate, Caterina D’Agrosa, Tara Cox, Damon Gannon,

Dave Johnston), my contemporaries in the trenches (Vicki Thayer, Ari Friedlaender,

Caroline Good, Catherine McClellen), those that will follow the tradition (Leslie Thorn,

Lynne Williams), and the leaders of the pack (Andy Read, Kim Urian, Danielle Waples).

I also thank all the PhD students of the marine lab during my tenure who provided comradery and daily support on all matters. Special thanks to the friendships of Danielle

Waples, Erin LaBrecque, Kelly Stewart, Caroline Good, Vicki Thayer, Zoë Meletis, James

Abbott, Josh Osterberg, Elliott Hazen, Matt Ogburn, Maria Wise, Leslie Thorn, Varun

Swamy, and Eric Treml. Lastly, I doubly thank Zoë and Elliott for fighting with me to make The Green Wave a reality. It’s working. I can only hope that our swell continues to build.

My off‐island times in Beaufort are what kept me sane and kept me focused. I thank the entire Beaufort community for giving me friendships and memories that are unforgettable. I send the biggest hug I can to Gretchen Lovewell for her undying love,

xviii

laughter, support and faith. Our friendship is a rock that we will always lean on. Next, I thank Lookout Ultimate (Who’s House?!). You all redefined the word “team” for me.

I’ve played on a lot of teams, but never before have my teammates also been my best friends. On and off the field, you all helped me relax, laugh, forget the stress and layout!

Thank you for all your friendships! And, I can’t forget to thank the Backstreet Pub: my meeting place, watering hole, music scene and dance hall, and place of escape when I needed it.

Lastly, it is my family that got me here; for you all I am eternally grateful. Mom and Dad, no other parents could have been more supportive of my journey. You all provided shelter during hurricanes, financial support during droughts, and waves of encouragement and love when I felt flat. Without your belief in me I would be lost.

Phillip, whether you knew it or not, you have always pushed me, physically and professionally. Now, your beautiful family inspires me. We will continue to grow together. Grandma, Grandpa and Grandma Torres, your ethic of education brought me here. I will carry your values of consciousness and conservation forever, passing it along to others. I cannot forget Ilana Simons who has given me strength and love from age four to age 104 (I hope).

xix

Todd Chandler, you are my even keel. You make me feel brave when I am weak, love me when I feel unloved, provide a reality check when I go too far, pick me up when

I feel down, and hold my hand throughout all the pendulum swings. I know our love will carry us through our many adventures and voyages, in Russamee and in life. And

Mango, my most loyal friend. Who greets me with a smile no matter what, who patiently waits for me to finish working, who gives me love whenever I need it, and who is always by my side with a fluffy, orange, sweet nuzzle.

And, one more thank you: To Florida Bay, your endless beauty and character are breathtaking and inspired so much curiosity in me. I hope my research aids in the conservation of the Florida Bay ecosystem and the sustainable growth of South Florida.

Introduction

Florida Bay lies at the tip of the Florida peninsula and is a dynamic ecosystem: ecologically, politically and economically. Since its creation as a result of sea level rise

4500‐3000 years before present (Wanless and Tagett 1989), the Bay has been a unique habitat because of its shallow depth and position at the confluence of three water sources: The Gulf of Mexico to the west, The Atlantic Florida Keys reef track to the east and south, and freshwater flow through the Florida Everglades from the north. The environmental characteristics of Florida Bay are distinctive and diverse, providing a variety of available niche space for flora and fauna (Figure A).

Figure A: Location of Florida Bay at the southern tip of the Florida peninsula. Inset: Florida Bay is the area in the box. Note heterogeneous mosaic of bottom types.

Florida Bay is currently the terminus of the nation’s large ecosystem restoration project, a direct repercussion of the fact that, for over 100 years, Florida Bay has experienced the consequences of the nation’s largest ecosystem engineering project.

From the time Europeans entered South Florida in the late 1800’s, they were determined to ‘tame the Everglades’ in an enduring effort to convert the swampland of the

Everglades into fertile soil for agriculture and profitable real estate (Grunwald 2006).

This ambitious mission meant re‐engineering all natural freshwater flow through South

Florida. Historically, freshwater flowed slowly south from Lake Okeechobee along a low gradient (slope of 3 cm/km) in a wide shallow sheet that covered all of South Florida

(10,000 km2) from the Atlantic Coastal Ridge to the Gulf of Mexico margin (Obeysekera et al. 1999). This gentle freshwater flow was the sustenance to life and vitality in the

Florida Everglades, a wild ecosystem in a delicate and soggy balance of land and water; its organisms well adapted to the periodic droughts and floods that characterized this region. Freshwater from Lake Okeechobee flowed south through the Everglades, then into the Shark River and Taylor sloughs, and finally into Florida Bay. It was at this time that Florida Bay could have truly been described as an ‘estuary’, where freshwater mixed with saltwater.

However, European settlers did not like the unpredictable and uncontrollable nature of this freshwater flow, nor the inability to farm, live on, or make money from this land. Therefore, an endless battle with the robust Everglades ecosystem began as humans tried their best to build dikes and levees and dig canals in an effort to control the water flow and force it east and west of Lake Okeechobee. ‘Drain the Everglades’ was the phrase of choice in the early and mid 1900’s in South Florida (Grunwald 2006),

and since 1900 one half of the 1.2 million ha once covered by the Everglades has been converted to agriculture or development (Davis et al. 1997). The ecological ramifications of this effort were drastic; South Florida and Florida Bay will never be the same. To make matters worse for Florida Bay, the construction of Henry Flagler’s Overseas

Railroad also caused major habitat alterations. Between 1902 and 1912, in a commercially driven effort to link Jacksonville, Florida with Key West, dredges and construction crews went to work building many railroad bridges to connect the Florida

Keys. This construction effort filled passes and blocked natural flow between the

Atlantic side of the Florida Keys and Florida Bay, impacting the ecosystem’s circulation patterns and salinity regimes (Smith et al. 1989; Brewster‐Wingard and Ishman 1999).

Freshwater, originally discharged into Florida Bay, now meanders through many opportunities to be rerouted before it reaches the Bay. With 59% less freshwater reaching

Florida Bay than in the pre‐drainage period (Smith et al. 1989), these diversions have turned an estuary into an environment that is often hyper‐saline. These unnatural conditions in Florida Bay are frequently exacerbated by conditions in South Florida: If a drought occurs in South Florida, water is retained in conservation areas and canals, causing hyper‐saline conditions in Florida Bay. If rains are heavy in South Florida, canal gates are open and freshwater flows generously into Florida Bay causing lower salinity

levels. As a consequence of such water management practices in South Florida and the construction of the Flagler Railroad, the habitat of many species has been lost or altered in Florida Bay.

As the adverse environmental affects of these destructive efforts to control water flow through the Everglades were realized, the Comprehensive Everglades Restoration

Project (CERP) was ratified by Congress in 2000. A total of $7.8 billion was appropriated to improve the quality, quantity and timing of freshwater flow through the Everglades and into Florida Bay. Sadly, the Florida Everglades will never be restored to its original state; South Florida’s population has exploded and continues to expand with an estimated population of 4 million in 2000 and a population growth rate of 17% between

1990‐2000 (U.S. Census 2000). This population growth puts a heavy burden on the environment. The most positive outcome from CERP is a balance of adequate freshwater flow through the Everglades to prevent further ecosystem degradation (and possibly revive currently marginal habitat), with a sufficient freshwater supply to South Florida’s demanding urban areas and agricultural activities.

Florida Bay is composed of a heterogeneous mosaic of habitat types and water quality gradients. The ecosystem is dominated by macroalgae (seagrasses) with a large

planktonic primary production base, both consumed by secondary producers which subsequently provide a food source to higher trophic‐level predators. The spatial variability of these resources in Florida Bay are the foundation of diverse prey and predator communities. It is the distribution of these communities relative to each other and their environment in Florida Bay that are the focus of my dissertation. Considerable research has been conducted in Florida Bay on geology, hydrology, vegetative communities, fish composition and wading bird communities (See: special issue of

Estuaries, 1999 v22, Porter and Porter 2002). However, prior to this study, no research had been conducted on Florida Bay’s top predators: bottlenose dolphins, sharks

(numerous species) and piscivorous birds (cormorants, osprey, pelicans and terns). My work provides baseline data for these populations and their distribution, but also aims to understand the ecology of these predators in Florida Bay within a framework of spatial scale, predator‐prey dynamics, foraging behavior, intra and interspecific competition, habitat use and habitat partitioning. In an effort to understand the chain and pathways of influences on predator distribution and habitat selection, I incorporated data from a hierarchy of levels within the Florida Bay ecosystem: environmental variability, prey communities and predator distributions. I collected original data during four summer field seasons in Florida Bay and integrated findings

from past research conducted in the Bay, such as the definition of environmentally homogeneous zones, into field methods and analyses.

Ecology is based on the foundation that species distribution is non‐random.

Individuals select habitats based on multiple, sometimes conflicting, parameters. For instance, fish do not select habitat based solely on the distribution prey, but must make tradeoffs between resource availability and predator avoidance. Likewise, when top predators select habitats they must account for prey availability, the effects of intra and interspecific competition, and density‐dependent relationships. In this dissertation, I considered the influence of predator‐prey interactions and environmental variability on the distribution of predators. Additionally, I paid close attention to the foraging behavior and tactics of predators to elucidate the functional mechanisms of habitat and prey resources.

Spatial scale directly affects species distribution, as well as how researchers study, perceive and ultimately understand species distribution. The term ‘habitat’ has no defined scale or size associated with it. Therefore, the same geographic location (X,Y coordinate) can have dramatically different habitat characteristics depending on the scale of the sampling unit: the exact X,Y location, a 5 m grid cell, or a 5 km grid cell.

Likewise, an individual may recognize different habitat qualities at different scales and choose its habitat dependently. For example, a fish may select a 10 km2 area with a preferred water temperature range, but also a 1 km2 seagrass bed with good refuge from predators, and then finally a 10m2 area of increased detritus availability for feeding.

Species can also interact and distribute themselves relative to each other at various scales. Therefore, spatial scale is a critical issue to consider. A study can identify or miss relationships between species distribution and predictor variables as a direct result of the scale associated with measurement and analysis. Additionally, spatial autocorrelation between explanatory variables can affect results by masking real determinants of distribution and/or identifying false or corollary relationships. In this dissertation, I used analytical methods that accounted for the effects of spatial scale and autocorrelation, or when this was not possible, I considered any relevant assumptions or biases.

Continued habitat alterations are expected for the Florida Bay ecosystem, not only as a result of the restoration project and population growth, but also due to sea level rise stemming from global climate change (Wanless, Parkinson, and Tedesco 1997).

By documenting the relationships between species distribution and their environment, the impacts of such habitat changes can be anticipated. More importantly, through

applying such knowledge with proper planning, loss of habitat and biodiversity can be avoided. And finally, I hope that future ecologists use this study as a baseline to examine how the predator species of Florida Bay adapted and evolved in the face of habitat alterations.

Chapter 1: Influence of teleost abundance on the distribution and abundance of sharks in Florida Bay, USA

Introduction

Understanding the factors influencing the distribution of animals is important for predicting the likely impacts of anthropogenic changes to the environment. Such studies are particularly important in Florida Bay, a large semi‐enclosed body of water between south Florida and the Florida Keys. Once considered an estuary, many parts of

Florida Bay now experience hypersalinity because up to 70% of the natural freshwater flow through the Florida Everglades is diverted by upstream management to support agriculture, control floods, and provide water to the growing population of South

Florida (McPherson and Halley 1996; Light and Dineen 1994). These large‐scale anthropogenic changes to the natural freshwater flow throughout south Florida have disturbed the greater Everglades ecosystem, including Florida Bay. The Comprehensive

Everglades Restoration Project (CERP), which aims to restore more natural water flow to the Everglades ecosystem, will modify the quality, quantity, and timing of freshwater inputs into this system (Ogden et al. 1999). As anthropogenic alterations to this ecosystem continue, understanding factors influencing habitat use and abundance of species at high trophic levels will allow managers to apply effective management schemes and monitor the results.

Previous studies have shown that alterations of habitats within Florida Bay, such as increased salinity and elevated nutrient levels, have stimulated algal blooms, resulting in large‐scale dieoffs of seagrasses, sponges and mangroves (Robblee et al.

1991; Butler et al. 1995; Rudnick et al. 1999). This habitat degradation has resulted in reductions in the abundance and diversity of teleosts and changes in the composition of their communities (Matheson et al. 1999). However, there is currently no information on the effects of these changes on large predators like sharks.

The distribution and abundance of predators may be driven primarily by the abundance of prey resources, with predators distributed across habitats proportional to prey availability (e.g. ideal free distribution; Fretwell and Lucas 1970). While the distribution of large marine predators often coincides with that of their prey at large spatial scales, there is often a mismatch between the distribution of predators and prey at small spatial scales in marine environments (e.g. Sih 1984; Fauchald, Erikstad, and

Skarsfjord 2000; Guinet et al. 2001). A lack of covariation in predator and prey distributions may be caused by a number of factors including anti‐predator behavior of prey (e.g. Sih 1984), physical attributes of the habitat that influence prey capture probability (Hugie and Dill 1994), and a lack of predictability of prey resources.

Florida Bay supports a diverse community of sharks that are relatively large‐ bodied predators and may have important influences on the distribution of fishes and other species in Florida Bay through predator‐prey interactions (e.g. Heithaus 2004).

However, no studies have specifically investigated the possible links between shark distributions and the distribution of their prey. Due to large variation in the physical habitats of Florida Bay and spatial variation in potential prey abundance (Thayer and

Chester 1989; Sogard, Powell, and Holmquist 1989; Matheson et al. 1999; Thayer, Powell, and Hoss 1999), shark abundances likely are spatially variable as well.

Previous studies have shown variable support for the hypothesis that sharks match the distribution of their prey. In open ocean habitats, basking sharks forage in frontal regions with high zooplankton densities (Sims and Quayle 1998), while in coastal seagrass habitats of Australia tiger sharks show a preference for habitats with the highest prey density (Heithaus et al. 2002). In contrast, juvenile blacktip sharks within a coastal nursery do not spend the majority of their time in areas where fish trap catches of teleosts are highest (Heupel and Hueter 2002). Thus, it is unclear to what extent the distribution of sharks is impacted by prey resources, or vice versa, at both regional and habitat spatial scales.

Methods

Study Site

Florida Bay is a large (1800 km2) and complex estuarine system that lies between the southern tip of the Florida peninsula and the Florida Keys. For this study, five environmentally distinct regions were selected based on salinity, water clarity, productivity, water depth, and bottom sediment types (Figure 1.1). Within regions, benthic habitats are spatially heterogeneous and may include seagrass, mud, hardbottom, and sand in varying levels and combinations. However, regions were selected to be as homogeneous as possible. The Eastern region is most influenced by freshwater discharge through the canals and freshwater sheet flow through the

Everglades. The Eastern region is characterized by low planktonic productivity, low to moderate salinity, moderate depth (1.5 – 2.5 m), high water clarity, and the bottom substrate is dominated by seagrass (Thalassia testudinum). The Central region is very shallow with an average depth of 1 m and experiences limited water circulation due to extensive mudbanks. The Central region has relatively high salinity (up to 50ppt), benthic habitats dominated by mud and mudbanks, low water clarity, and high productivity. The Atlantic region is heavily influenced by water flow from the Atlantic

Ocean through various passes between the middle Florida Keys. This region has oceanic salinity (~35 ppt), high water clarity, moderate to deep waters (1.5 – 3 m), is composed of 14

hardbottom and sparse seagrass benthic habitats, and has moderate productivity levels.

The Gulf region is open to the Gulf of Mexico on its western and southwestern sides. It is characterized by oceanic salinities (~35 ppt) with moderate to high productivity rates, relatively low water clarity, and is the deepest region of this study (1.5 to 3.5 m). The substrates are dominated by mixed seagrass (Thalassia testudinum and Syringodium filform), mud or sand. The Flamingo region lies between the Central and Gulf regions of

Florida Bay and, therefore, is influenced to some degree by oceanic waters of the Gulf of

Mexico. The Flamingo region is a shallow area dominated by mudbanks and seagrass

(Thalassia testudinum and Halodule wrightii) with numerous channels that connect basins.

The Flamingo region is typically very turbid, has moderate productivity, and normal to high salinity levels.

Figure 1.1: Location of shark and teleost sampling sites (*) within five regions of Florida Bay, USA.

Field Methods

From 29 June – 23 July, 2005, teleosts, sharks, and environmental conditions were sampled at 43 sites (Figure 1.1). Sample sites were chosen to represent the diversity of habitats present within each region so a range of habitats and fish communities would be sampled while also obtaining adequate spatial coverage of each region. At each site, environmental data (temperature, salinity, turbidity, percent dissolved oxygen, and

chlorophyll a) was collected using a YSI 6600 Sonde. At a minimum of two sites in each region, water samples were obtained and filtered to calibrate the chlorophyll a readings produced by the YSI. These filters were later extracted in the lab and a linear regression model was developed to convert the YSI chlorophyll a readings into accurate chlorophyll a values. The depth and bottom habitat type were also recorded at each sampling site. A trawl was then conducted to sample the teleost community and a longline was set approximately 100m from the trawled area to sample the shark community. Trawls were conducted at each sampling location immediately before the longline was set. The relatively short duration of trawls relative to longline soak times and the distance between trawls and longlines minimized potential impacts of trawls on catch rates of sharks. A total of 43 trawls and 43 longline sets were conducted.

The trawl sample used a 3‐m research otter trawl towed at approximately 4 km/hr for 3 minutes. All captured fish were identified and their total length (TL) measured and recorded, before being released alive. GPS position was recorded at the start and end of each trawl in order to calculate the exact distance trawled using GIS

(ArcGIS; Version 8.2). Catch per unit effort (CPUE) was calculated as fish captured per meter of trawling.

Sharks were captured on a 600m longline baited with 35‐50 hooks spaced approximately 10‐15m apart. Each hook (size 13/0‐15/0 Mustad tuna circles) was baited with mullet or squid (bait types were distributed randomly across regions) and attached to an approximately 3m long individual clip line made of 900lb monofilament. The longline was allowed to soak for approximately one hour. Upon retrieval, the presence or absence of bait on each hook was noted. All captured sharks were identified, measured, tagged, and released alive. Shark CPUE is expressed as the number of sharks captured per hour of bait soaking. To account for variation in the number of hooks deployed on the longline, soak time for a set was expressed as the sum of soak times for each hook where hooks that captured a shark or lost bait were considered to have lost the bait half way through the soak time (see Heithaus 2001).

Analysis

The influence of teleost abundance and physical features of the environment on shark CPUE was determined using ANOVA on log (x+1) transformed data. To determine whether sharks responded to teleost abundance at large (regional) and small

(the sampling site) spatial scales, teleost CPUE at the site and mean CPUE for the region was included in the analysis. Because not all fishes that were captured are likely to be consumed by sharks, analysis was restricted to include only fishes over 2.0 cm TL. Also, six groups of fish that are unlikely to be prey items of sharks were eliminated from the

analysis: rainwater killifish (Lucania parva), blenny sp. (Blenniidae Spp.), seahorse sp.

(Hippocamous Spp.), goby sp. (Gobiidae Spp.), sheepshead minnow (Cyprinodon variegatus variegatus), and pipefish sp. (Syngnathidae Spp.). Results were similar for analyses that included all species of all sizes, all species of fishes over 2.0 cm TL, and all fishes over 5.0 cm TL. I did not attempt to analyze only fish species that are known prey of sharks because of a lack of data on shark diets in the region and the likelihood of considerable geographic variation in shark diets (see Simpfendorfer, Gotreid, and McAuley 2001;

Weatherbee and Cortes 2004). Furthermore, teleost data from trawls serve as an index of secondary production within habitats and regions rather than precise measures of food available to sharks.

Results

Seven shark species and 45 species of teleosts (Table 1.1 & 1.2) were captured. The Gulf region had the greatest diversity of shark species and the highest shark CPUE, followed closely by the Flamingo region (Table 1.1). The most commonly captured shark species were bonnethead (Sphyrna tiburo), lemon (Negaprion brevirostris) and nurse sharks

(Ginglyostoma cirratum). The most frequently captured teleosts were gulf killifish

(Fundulus grandis grandis), mojarra sp. (Eucinostomus spp.), and pinfish (Lagodon rhomboides). The Flamingo and Gulf regions also had the highest teleost CPUE. 19

Table 1.1: Number of individual shark species caught in Florida Bay, broken down by region. A = Atlantic, C = Central, E = Eastern, F = Flamingo, G = Gulf, T = Total

Common Name Scientific Name A C E F G T Bonnethead shark Sphyrna tiburo 3 2 5 8 2 20 Nurse shark Giglyostoma cirratum 3 1 2 2 8 Atlantic sharpnose shark Rhizoprionodon terraenovae 1 4 5 Lemon shark Negaprion brevirostris 4 7 11 Blacktip shark Carcharhinus limbatus 5 5 Bull shark Carcharhinus leucas 2 2 Great hammerhead shark Sphyrna mokarran 1 1 Total 7 6 6 17 16 52

Table 1.2: Abundance of fishes caught in trawls within each region sampled in Florida Bay. A = Atlantic, C = Central, E = Eastern, F = Flamingo, G = Gulf, T = Total.

Common Name Scientific Name A C E F G T Gulf killifish Fundulus grandis grandis 7 15 716 738 Mojarra Diapterus Spp. 12 48 54 62 181 357 Pinfish Lagodon rhomboides 28 4 14 65 117 228 Planehead filefish Stephanolepis hispidus 37 3 1 50 91 Pipefish Syngnathidae Spp. 5 4 29 41 11 90 Silver perch Bairdiella chrysoura 32 31 63 Juvenile grunt Haemulon Spp. 29 1 31 61 Sheepshead minnow Cyprinodon variegatus variegatus 46 46 Pigfish Orthopristis chrysoptera 1 34 7 42 Fringed filefish Monacanthus ciliatus 34 3 37 Gulf toadfish Opsanus beta 4 1 10 11 7 33 White grunt Haemulon plumierii 13 1 19 33 Lane snapper Lutjanus synagris 14 7 8 29 Blue striped grunt Haemulon sciurus 23 1 3 27 Scrawled cowfish Acanthostracion quadricornis 11 1 1 6 19 Striped burrfish Chilomycterus schoepfi 7 2 2 4 15 Gray snapper Lutjanus griseus 5 6 3 14 Parrotfish spp. Scaridae Spp. 6 6 12 Yellow tailed snapper Ocyurus chrysurus 10 10 Grass porgy Calamus arctifrons 2 2 5 9 Sardines Sardinella Spp. 7 7 Bandtail puffer Sphoeroides spengleri 6 6 Wrasse spp. Labridae Spp. 6 6 Barracuda Sphyraena barracuda 4 1 5 Barred hamlet Hypoplectrus puella 5 5 Goby spp. Gobiidae Spp. 1 3 1 5 Hogfish Lachnolaimus maximus 3 2 5

Table 1.2 (continued)

Common Name Scientific Name A C E F G T Bonnetmouth Emmelichthyops atlanticus 4 4 Blenny spp. Blenniidae Spp. 1 2 3 Rainwater killifish Lucania parva 3 3 Sea horse Hippocampus Spp. 1 2 3 Gulf flounder Paralichthys albigutta 1 1 2 Sand perch Diplectrum formosum 2 2 Schoolmaster snapper Lutjanus apodus 1 1 2 Scorpian fish Scorpaenidae Spp. 1 1 2 Yellow stingray Urolophus jamaicensis 2 2 Balloonfish Diodon holocanthus 1 1 Damselfish Spp. Pomacentridae Spp. 1 1 French angelfish Pomacanthus paru 1 1 Inshore lizardfish Synodus foetens 1 1 Inland silverside Menidia beryllina 1 1 Mutton snapper Lutjanus analis 1 1 Seaweed blenny Parablennius marmoreus 1 1 Trunkfish Lactophrys trigonus 1 1 Unidentified fish 1 1 Total 272 68 142 1040 503 2025

Shark CPUE was not influenced by any of the physical features of the environment that were sampled, nor was it correlated with chlorophyll a levels (Table

1.3). Similarly, there was no relationship between shark catch rates and the CPUE of teleosts at a particular sampling site. However, shark catch rates were significantly higher in regions where teleost CPUE was also higher (Table 1.3, Figure 1.2).

Table 1.3: Influence of biotic and abiotic factors on catch rates of sharks in Florida Bay. Only fish over 2.0 cm TL were included in the analysis. Species mentioned in text were eliminated from analyses.

Factor F1,42 P

Regional fish abundance 4.4 0.04

Sample site fish abundance 1.0 0.33

Depth 0.09 0.77

Temperature 0.6 0.43

Salinity 0.06 0.81

Turbidity 0.6 0.45

Chlorophyll a 0.03 0.86

Figure 1.2: Influence of regional teleost abundance on shark catch rates. Error bars represent ± SE.

Discussion

Shark catch rates in Florida Bay were spatially variable, but were not significantly correlated with physical factors, chlorophyll a levels or the abundance of teleosts at small spatial scales. Shark abundance was, however, greater in regions where teleost abundance was highest. A lack of covariation in marine predator abundance and their prey at small spatial scales with linked distributions at larger spatial scales is consistent with findings in other, primarily open, marine systems (e.g. Mehlum et al.

1999; Fauchald, Erikstad, and Skarsfjord 2000; Guinet et al. 2001). However, previous studies of marine predators have found that predator and prey distributions tend to coincide at small spatial scales when prey resources are predictable. For example, pied cormorants (Phalacrocorax varius) match the distribution of their prey at the scale of microhabitats and habitat patches in a seagrass ecosystem with predictable teleost distributions (Heithaus 2005). Similarly, tiger sharks prefer shallow seagrass habitats where prey is most abundant (Heithaus et al. 2002) and basking sharks actively select energetically profitable patches of zooplankton (Sims and Quayle 1998). Because the spatial heterogeneity present in Florida Bay should lead to relatively predictable teleost distributions, a significant effect of teleost abundance on shark CPUE at the level of sampling sites might be predicted. However, sharks have relatively low feeding rates

(Weatherbee and Cortes 2004), and therefore, may not concentrate their movements in microhabitats of high prey abundance, especially when they are at risk of predation themselves and safer areas have lower prey abundance (Heithaus 2004). Many of the sharks captured were juveniles, which often exhibit relatively restricted movements (e.g. lemon sharks, Morrissey and Gruber 1993; blacktip sharks, Heupel and Hueter 2002).

For example, one blacktip shark that was tagged during this study was subsequently recaptured on two occasions within several kilometers of its release location. Therefore,

it is possible that sharks remain within regions of high prey density even if they do not match prey abundance at small spatial scales.

The relationship between relative teleost abundance and shark abundance may have been obscured due to sampling limitations. Specifically, the rapid removal of bait from the longline in some locations, especially the Atlantic and Gulf regions, where relatively high teleost catch rates occurred, may have prevented an accurate estimation of shark abundance. Bait loss in these instances is likely due to the abundance of untargeted scavenger species, especially crabs. This problem was addressed in the calculation of shark CPUE by assuming bait loss occurred half way through the soak time. While it is likely that this technique only partially accounts for the high volume of bait loss in some regions, it is important that habitat‐ and region‐specific rates of bait loss be considered in studies of elasmobranches that use hook‐capture methods.

The finding that shark abundance at a regional scale is related to the average abundance of teleosts within the region suggests that proposed increases in freshwater flow to Florida Bay may have profound consequences for the abundance and distribution of these top predators. Currently, the CERP project has not defined the restoration plans or goals for Florida Bay, but has initiated the Florida Bay & Florida

Keys Feasibility Study. The goal of this study is to evaluate the Florida Bay ecosystem and determine the modifications that are needed to successfully restore water quality and ecological conditions of the Bay. As part of this feasibility study, an upper trophic level modeling component will be developed to consider the response of upper trophic level species, including fish and sharks, in an effort to provide recommendations for

Florida Bay’s restoration. Although the precise habitat alterations caused by CERP are yet to be determined, it is likely that the freshwater increase will alter the habitat quality throughout Florida Bay. Shark habitat selection at large spatial scales will likely be mediated by the response of their teleost prey to these habitat alterations. This in turn could lead to cascading effects in the ecosystem caused by top‐down effects of sharks

(reviewed in Heithaus 2004). Further studies of links between patterns of shark abundance, including species‐specific analyses, with biotic and abiotic factors will be of great value to predicting the effects of proposed modifications on the distribution and abundance of sharks. Furthermore, studies that investigate the effects of sharks on other species within Florida Bay will provide important information on how changes to shark populations and habitat use may cascade through the community.

Conclusions

No relationship between shark distribution and physical features of the environment or phytoplankton primary productivity were identified. Although shark

CPUE was not affected by teleost abundance at a sampling site, shark abundance was positively correlated with mean teleost CPUE for a region. Due to sampling limitations in this study, further research across a longer temporal period and that examines the links between individual shark species and their prey will greatly enhance our understanding of these top predators in Florida Bay. This study suggests that shark abundance is likely to be impacted by changes to teleost communities that are predicted to occur during future anthropogenic and natural (i.e.: hurricanes, sea level rise) changes to the Florida Bay ecosystem.

Chapter 2:

Fine-scale habitat modeling of a top marine predator: Does prey data improve predictive capacity?

Introduction

Modeling species distribution is a valuable tool of biological conservation efforts, especially predictive models of marine predators due to the logistical difficulties of monitoring their distributions at sea. For instance, managers of whale and dolphin populations can benefit from accurate model‐derived predictions of cetacean habitat to mitigate anthropogenic effects such as fisheries by‐catch (Torres et al. 2003), sonar (Cox et al. 2006), and the impacts of habitat alterations on ecosystem function (Baumgartner et al. 2000; DʹAmico et al. 2003), in order to protect critical habitat (Hooker, Whitehead, and Gowans 1999; Gregr and Trites 2001) and to understand the ecology of these animals (Hamazaki 2002). By assuming that the distribution of cetaceans is non‐random relative to environmental variability, predictive models of cetacean distribution typically identify the ecological relationships between the environment and species habitat selection. With the goal to improve conservation applications of modeling efforts, my study examines the potential for increased predictive capacity by models of dolphin distribution that include direct prey data as an explanatory variable.

Abiotic variables may be correlated with the distribution of dolphins (i.e. temperature, salinity, depth, dissolved oxygen, distance from shore). However, these metrics often have little direct influence on the actual selection of habitats by dolphins.

In reality, these abiotic variables are frequently used as proxies for prey distribution.

Predictive models of dolphin distribution rarely include direct data on prey distribution because prey sampling is more difficult than sampling of abiotic variables. Therefore, as top marine predators, dolphins are removed from the direct influence of the environmental variability that is commonly used to characterize their habitat.

The ability to predict dolphin distribution requires an understanding of the ecological factors relevant to dolphin habitat selection (Fig. 2.1). The schematic in Figure

2.1 is similar to the three trophic‐level system applied to predator‐prey interactions based on game theory (Hugie and Dill; Lima 2002; Sih 1998). Water and habitat quality compose the first level of my simplified ecosystem schematic (termed ENV), which represents the environmental characteristics that influence phytoplankton and zooplankton distribution, as well as the physiological limits of fish distribution. Primary and secondary predators (FISH in this study) respond to the distribution of these environmental factors because they are poikilotherms and therefore are tied to changes in their environment. Predators are at the other end of the schematic; The distribution of homeothermic DOLPHINS is contingent upon the spatial structuring of their prey

(FISH). The relationships between ENV & FISH and FISH & DOLPHINS are well understood, but the extent to which ENV can be used to predict dolphin distribution

without any information on the intermediary step (FISH) has not been evaluated.

Nevertheless, the presumed relationship between ENV & DOLPHINS is supported by well‐developed predator‐prey theory; Lima (2002) states that, due to predator avoidance by prey, predators will distribute themselves relative to the distribution of the resources of their prey (i.e.: high primary productivity or zooplankton aggregations), rather than the distribution of the prey itself.

Figure 2.1: Schematic of a three level ecosystem between the environment, prey and predators. Solid lines depict direct relationships between levels and the dashed line indicates an unconfirmed indirect relationship.

In a recent paper reviewing the techniques for cetacean habitat modeling

(Redfern et al. 2006), the authors note that 1) environmental variables are used as proxies of prey distribution and, 2) cetacean‐habitat models should ideally incorporate data on prey populations and understand the ecological relationships between predators, prey and their environment. It is generally believed that predictive models of cetacean

distribution would improve with the inclusion of direct data on prey distribution and densities (Ferguson, Barlow, Fiedler et al. 2006; Redfern et al. 2006; Davis et al. 1998), but this assertion has never been tested. Few studies have actually sampled prey and monitored dolphin distributions synoptically to described correlations with dolphin habitat use (Allen et al. 2001; Benoit‐Bird and Au 2003; Heithaus and Dill 2002). In this chapter, I ask whether including prey distribution data in fine‐scale predictive models of bottlenose dolphin (Tursiops truncatus) habitat selection in Florida Bay, FL, USA improves predictive capacity.

I begin with an exploratory exercise to compare the response of dolphin distribution and fish catch to a suite of environmental variables. I assess the pragmatism of predictive models to rely on proxy relationships between environment and dolphin distribution under the assumption that dolphin predators mimic the spatial distribution of their prey. If this assumption is true, dolphins and fish should display similar responses to the same environmental variables. This phase of analysis also evaluates the optimal descriptive metric of fish sampling data (i.e. catch per unit effort, diversity, richness) for use in models of dolphin habitat selection. Next, in hypothesis 1 (Ho 1) I assert that dolphin habitat selection can be predicted without recourse to describing the distribution of their prey. I assess four types of models, with and without fish sampling

data as explanatory variables, according to their capacity to predict dolphin habitat selection. Hypothesis two (Ho 2) builds on the results of Ho 1 to determine if accurate predictive maps of dolphin distribution can be produced by modeling areas of high fish catch based on significant environmental characteristics. The model tested in Ho 2

predicts habitat with high prey availability and assumes that dolphins will distribute themselves accordingly.

I collected data in Florida Bay during four summer field seasons (2002 ‐2005) to test these hypotheses. These data include standardized dolphin surveys and prey sampling by bottom trawls, both with associated habitat quality measurements. Florida

Bay is a remarkably heterogeneous ecosystem (Durako, Hall, and Merello 2002;

Fourqurean and Robblee 1999; Zieman, Fourqurean, and Iverson 1989; Sogard, Powell, and Holmquist 1989) which lends itself to a fine‐scale study of dolphin habitat selection due to large habitat variation over small spatial distances. Ample data are available from these four years of fieldwork to use independent datasets to train and test models. Shark sampling conducted in Florida Bay found few sharks of the species and sizes that would pose a threat to dolphins or cause behavior modifications (Torres, Heithaus, and Delius

2006 (Chapter 1); Wiley and Simpfendorfer 2007; Heithaus 2001). Thus, I assume a minimal predation effect on dolphin habitat selection in Florida Bay. Commercial fishing

is prohibited and recreational boating is minimal in Florida Bay, so I also assume that fine‐scale dolphin distribution is not heavily influenced by boat traffic, fishing activities or other proximate anthropogenic activities.

I use Generalized Additive Models (GAMs) to model dolphin and fish catch distribution because they are non‐parametric and describe non‐linear relationships.

GAMs have been used in previous work to detect significant non‐linear relationships between cetacean distribution and environmental variables (Forney 2000, 1999; Hedley,

Buckland, and Borchers 1999; Ferguson, Barlow, Reilly et al. 2006). Moreover, Segurado and Araujo (2004) found that GAMs are an appropriate technique to model species with complex distribution and behavior patterns relative to environmental variables.

My spatial analysis is performed on presence/absence data, incorporating each sighting equally regardless of group size to minimize the effect of population density of habitat selection patterns. A fine grain resolution of 50 m is applied to all analyses because of the extreme habitat heterogeneity in Florida Bay. The scale of study is an important and, often confounding, variable related to distribution matching between predator and prey in marine environments (Heithaus and Dill 2006; Logwerwell and

Hargreaves 1996; Schneider and Piatt 1986). Working at a small scale fosters the

incorporation of small, yet important, refuges for prey and the ability of predators to use patchy habitats with high prey availability. Moreover, fine‐scale studies of species biogeography coupled with behavioral data can reveal considerable insight about the biological mechanisms underlying associations between distribution, environment and space. Therefore, I recorded the behavior state of observed dolphins at each sighting to denote information on the functional mechanisms of habitat selection (Hastie et al. 2004;

Heithaus and Dill 2002). Dolphins may socialize or travel through marginal habitat, but forage in areas where critical resources exist; such areas may deserve increased management attention and protection. Therefore, due to the importance of identifying feeding areas in conservation modeling applications, I assess each model’s ability to predict dolphin foraging habitat, as well as the model’s overall predictive capacity.

Methods

Study Site

Florida Bay lies at the southern tip of the Florida Peninsula and is approximately

1800 km2. Florida Bay is composed of a heterogeneous mosaic of benthic habitat types including seagrass, mud, sand and hardbottom areas composed of sponge and coral structures (Fig. 2.2). Previous research in Florida Bay sub‐divided the bay into

environmentally distinct zones based on fish composition (Thayer, Powell, and Hoss

1999; Sogard, Powell, and Holmquist 1989; Thayer and Chester 1989), water quality

(Boyer, Fourqurean, and Jones 1999; McIvor, Ley, and Bjork 1994) and seagrass distribution (Hall et al. 1999; Zieman, Fourqurean, and Iverson 1989). I used these definitions in my study to divide the bay into 6 relatively homogeneous zones: Atlantic,

Central, Eastern, Gulf, Flamingo and Western.

Figure 2.2: Bottom types and zones of Florida Bay. Inset: Florida Bay lies at the southern end of the Florida peninsula (area in box enlarged). Black lines designate boundaries between the six environmentally homogeneous zones. Bottom types map based on USGS OFR 97‐526.

Field Methods

I conducted a fine‐scale study of dolphin habitat selection in Florida Bay during the summers of 2002 through 2005. This field work consisted of trackline surveys for dolphin presence and absence. I drove a 17ft Boston Whaler at approximately 16 knots with 3 trained observers searching for dolphins. When a dolphin sighting occurred, I stopped the research vessel so that the radial distance to the sighting and angle from trackline could be estimated. I used these data to determine the perpendicular distance from trackline (Buckland et al. 2001) and to generate pseudo absences (See Pseudo‐

Absence Generation below). The field crew then slowly approached the dolphin(s) to record data including GPS location, depth, benthic habitat type, water quality metrics

(see below), group size and behavior state (forage, travel, socialize, rest, or unknown).

For these analyses, I classified dolphin behavior at each sighting as foraging or non‐ foraging. I determined foraging behavior if dolphins were observed chasing fish, catching fish, or surfacing erratically and quickly within one area using fluke‐out dives to promote deep diving profiles. Additionally, I recorded water quality and bottom type at the start and end of each survey, as well as every 30 minutes if no dolphins were sighted. These methods allowed me to assess environmental conditions at locations of both dolphin presence (sightings) and absence.

Due to limited knowledge on dolphin habitat use in Florida Bay prior to fieldwork in 2002, I used a random approach to dolphin surveys during the 2002 and

2003 field seasons. During each of these field seasons I surveyed the entirety of Florida

Bay twice. This allowed me to identify areas of high use by dolphins. From these surveys in 2002 and 2003, I established tracklines in areas of high dolphin use within 3 of the environmentally distinct zones in 2004 and 2005: Atlantic, Central, and Gulf (Fig.

2.3). I completed replicate surveys along each trackline four times in 2004 and twice in

2005.

Figure 2.3: Survey areas and tracklines during 2004 and 2005 field seasons. Grayed areas represent surveyed regions within larger zones. Black lines designate the trackline followed within each zone for replicate surveys.

Benthic Habitat and Water Quality Sampling

The USGS produced a bottom types map of Florida Bay in 1997 (USGS, OFR 97‐

526). I used this map as the foundation for benthic habitat classification throughout

Florida Bay. To refine this map, I assessed and recorded benthic habitat type at all water and habitat quality sampling locations. These 1092 sampling locations were used to 40

update and improve the resolution of the original USGS map (see Fig. 2.2). I classified habitat types by visual inspection (through the water column) or, when turbidity did not allow the former, a small bottom grab sampler (2” diameter; 3” deep) was used. I categorized bottom types with the same classification system used by the USGS map with the addition of two classes: 1) Patchy seagrass: areas composed of dense patches of seagrass interspersed in barren landscapes, and 2) Hardbottom with seagrass: areas composed on equal amounts of sponge and coral habitat as seagrass habitat. Thus, I used nine bottom type classifications in this study: sparse seagrass, intermediate seagrass, dense seagrass, patchy seagrass, hardbottom, hardbottom with seagrass, mud, sand, and mudbank.

In 2002 and 2003, I used a YSI 30 to measure salinity and temperature at the mid‐ point of the water column. To estimate turbidity, percent dissolved oxygen and chlorophyll a values during these years, I acquired data through the Southeast

Environmental Research Center’s Water Quality Monitoring Network. This program has

24 water quality stations with monthly sampling periods placed throughout Florida Bay.

Using ArcGIS® (v. 9.1, ESRI), I interpolated these data points with an inverse distance weighting (IDW) technique which preserves local variation between sample points. I chose an IDW interpolation technique as the most appropriate method to create raster

grids for preservation of the water quality variation that exists in Florida Bay across spatial scales less then or equal to the distances between the 24 water quality stations. In total, I created 18 separate rasters: one grid for each variable (turbidity, percent dissolved oxygen, chlorophyll a) for each month of the study period (June, July, August of 2002 and of 2003). Finally, I used each location (sightings, absences, trawls, and 30‐ minute survey interval water quality sites) to sample the appropriate temporal set of three grids. I used these values of turbidity, percent dissolved oxygen and chlorophyll a in all further analyses regarding 2002 and 2003 data points.

During the 2004 and 2005 field seasons, I used a YSI 6600 Sonde to sample the following water quality data in the upper and lower portions of the water column: temperature (C°), salinity (ppt), turbidity (NTU), percent dissolved oxygen, and chlorophyll a. I applied the average value of the upper and lower water column measurements for each location in all analyses because no discernable difference was detected between water samples from the well mixed, shallow waters of Florida Bay. I obtained and filtered water samples to calibrate the YSI‐based chlorophyll a fluoresence readings. I used Whatman GF/F 25mm diameter filters with nominal porosity of 0.7 μm.

I later extracted these filters in the lab and used a Turner Designs Fluorometer to make absolute chlorophyll a readings. From these readings, I developed a linear regression

model to convert the YSI chlorophyll a readings into accurate chlorophyll a (μg/l) values.

I collected water samples at 12 spatially and temporally distributed sites in 2004 and 7 sites in 2005, with 3 replicate filters extracted from each sample.

Fish Sampling

I sampled the fish community throughout Florida Bay using a 3‐m demersal research otter trawl towed at approximately 4 km/hr for 3 minutes. I randomly generated the locations of trawl sampling sites, and stratified them by benthic habitat type to sample the different bottom types within each zone. I conducted a minimum of three trawls on each survey day. All captured fish were identified, measured, and then released alive. I recorded all water and habitat quality metrics prior to each trawl, and the GPS positions at the start and end of each trawl to calculate the exact distance trawled.

I converted the catch from each trawl into four descriptive metrics: Catch per unit effort (CPUE), dolphin prey per unit effort (DPPUE), Simpson’s diversity index per unit effort (SPUE), and Margalef’s species richness index per unit effort (MPUE). CPUE is the total number of fish captured per meter of trawling. DPPUE is a subset of CPUE consisting of catch which may be potential dolphin prey items. I determined this subset by fish species and size. Due to the expansive, uninhabited nature of Florida Bay,

stranded dolphins are rarely recovered and to date no data exist from stomach contents in this region. Therefore, I assumed the diet of dolphins in Florida Bay to be similar to the diets of bottlenose dolphins documented in adjacent areas (Barros 1987, 1993; Barros and Odell 1990; Barros and Wells 1998). SPUE is Simpson’s diversity index (∑pi2) for

each trawl divided by the length of the trawl in meters. I calculated Margalef’s richness index ((S‐1) / log N) for each trawl and converted it into the per meter of trawling metric

MPUE. I implemented these four metrics in the exploratory exercise to determine the fish catch metric most correlated with dolphin habitat selection.

Pseudo-Absence Generation

While surveying for dolphins, I constantly collected absence data. However, this data is in the form of strip transect areas, as opposed to dolphin sightings which are point locations. Therefore, to test for dolphin habitat selection with binomial presence/absence data, I generated pseudo‐absence points in the areas of absence by accounting for detection probabilities (MacKenzie et al. 2002; Brotons et al. 2004).

The quality of a model can be significantly influenced by the location of absence points (Engler, Guisan, and Rechsteiner 2004). Therefore, I proportionally distributed pseudo‐absence locations while accounting for survey effort to avoid labeling areas of missed sightings as absence locations. This approach allowed me to be confident that the

pseudo‐absences are real absences, and not locations where dolphins could have been present but were not detected. I generated ten times the number of sightings as pseudo‐ absences according to probability of detection functions; I randomly distributed these pseudo‐absences in proportion to survey effort, distance from trackline, and sighting conditions.

The ability of an observer to detect dolphins is a function of the dolphin’s distance from the trackline and the sighting conditions. Beaufort Sea State (BSS) scale was used as a measure of sighting conditions in Florida Bay. BSS is an integer scale used to describe the sea conditions (wind speed and wave height) which affect the sightability of dolphins. For this study in Florida Bay, survey conditions ranged from a BSS 0 (flat calm) to BSS 4 (constant white caps). As conditions deteriorated from a BSS 0 to a BSS 4, the ability to detect a dolphin decreases and an observer is less likely to spot a dolphin further from the trackline. Using the program DISTANCE 5.0 (Thomas et al. 2005) probability of detection functions were created using sightings data from 2004 and 2005 relative to BSS and perpendicular distance from trackline. The analysis was run using

Multiple Covariates Distance Sampling with BSS as the only covariate. The data set consisted of 99 sightings, 2228.6km of trackline, and 543 samples. The optimal model, based on the lowest Akaike information criterion (AIC), was a hazard‐rate model with a

cosine expansion function, 5% truncation, and manually set intervals of 50 m. The resulting detection functions for BSS 0‐4 were then used to generate pseudo‐absences.

For each zone/year combination, 10X the number of sightings in each BSS class were created as pseudo‐absences. The choice to generate 10X as many absences as presences was arbitrary, but due to the rarity of true sightings relative to the large amount of survey effort, the contrast between presence and absence locations was increased by generating more pseudo‐absences according to stringent dispersal patterns.

Moreover, through multiple trials it was determined that the ability of both the binomial

Mantel Tests and GAMs to differentiate habitat characteristics between presence and absence points dramatically increased with a larger sample size while maintaining robustness despite an unbalanced sample (more absences than presences).

Each trackline was buffered out to 450 m because this was the maximum perpendicular sighting distance according to DISTANCE (five sightings were made at perpendicular distances between 450‐820m, but these were considered outliers by the program). Each 450 m buffered polygon was segmented into consecutive 50 m distance from trackline intervals and then classified by BSS (0, 1, 2, 3 or 4). Using these polygons, the area of survey effort, on each survey day, in each BSS was calculated. Additionally,

to prevent absence points from occupying the same habitat as actual sightings, a 450 m buffer was created around each sighting, within which no pseudo‐absence point could be created.

Each probability of detection function relative to BSS and distance from trackline was used to calculate the percent area under the curve at each distance interval of 50 m.

The percentage for each distance class was multiplied by 10X the number of sightings in the same BSS/zone/year class. This number of absences was then distributed between survey days proportional to the amount of area surveyed within each distance bin.

Finally, Hawth’s Tools in ArcGIS (Beyer 2004) was used to randomly distribute the correctly apportioned pseudo‐absence points within each BSS class and distance from trackline bin for every survey day. In summation, 10X the number of sightings in each BSS/zone/year group were generated according to the probability of detection functions and randomly distributed in proportion to the amount of survey effort within each distance class bin.

Grid Sampling

I created interpolated raster grids, with 50 m grid cell size, to obtain water quality values for all absence points. I recorded water quality and habitat metrics at all

dolphin sightings so it was unnecessary to sample presence locations (except to acquire turbidity, percent dissolved oxygen, and chlorophyll a values for 2002 and 2003 sightings, as described above). I created daily water quality grids for each survey day through interpolations of all water quality sampling locations on each survey day

(sightings, trawls, 30‐minute survey intervals, and start and end of surveys samples). I employed a kriging interpolation method to generate daily water quality grids because numerous data points occupied relatively small spatial extents, allowing the kriging method to accurately interpolate spatial trends between data points.

In contrast to daily water quality grids, I created seasonal interpolated surfaces of trawl data of the four fish metrics. With only three trawls conducted on each survey day

I did not have enough spatial coverage to adequately describe the daily spatial variability of fish distribution. Therefore, I used all sampling locations conducted in each zone and summer field season to interpolate fish grids. With this method I make the assumption that fish community composition within each zone of Florida Bay does not change throughout a summer (Gaertner 2000; Thayer, Powell, and Hoss 1999; Matheson et al. 1999). I created CPUE, DPPUE, SPUE and MPUE grids for each zone during each summer between 2002 and 2005. I interpolated these grids from trawl data points using

either a spline or kriging method, determined by the accuracy of real data point representation.

Analysis

Generalized Additive Models (GAMs)

Using S‐plus 7.0 (Insightful Corporation, Seattle), I created models of dolphin habitat selection using Generalized Additive Models (Hastie and Tibshirani 1990) with a smoothing function and backwards variable selection. I chose the optimal model, composed of the combination of variables that best fit the observed data, based on the lowest Akaike information criterion (AIC) (Burnham and Anderson 1998). GAMs generate smoothed curves representing the relationship between the response and each predictor variable in the model. GAMs are particularly good at identifying and describing non‐linear relationships which are more typical than linear relationships in ecology (Oksanen and Minchin 2002). I conducted GAMs with two types of data: 1) binomial dolphin presence/absences data with a logit link function, and 2) continuous fish catch data using a gaussian family model with an identity link function.

Mantel’s Tests

Mantel’s tests (Mantel 1967) are able to overcome many problems associated with examining species‐environment relationships. They are multivariate, explicitly test

for the effect of space on the response variable, account for multicolinearity between predictor variables, and identify and account for spatial autocorrelation of explanatory variables (Schick and Urban 2000). Mantel’s test are a more robust analysis then GAMs which have been criticized for their tendency to overfit data and give artificially high p‐ values (Guisan, Edwards, and Hastie 2002; Guisan and Zimmermann 2000; Vaughan and Ormerod 2005). Unlike GAMs, greater correlation between predictor and response variables is required to obtain a significant p‐value from a Mantel’s test. Therefore, for the exploratory exercise, I used Mantel’s tests (run in S‐Plus 7.0) to determine significant predictor variables correlated with dolphin presence/absence and limit the number of variables included in the GAMs.

A Mantel’s test is a linear regression between dissimilarity matrices that describe the ecological distance between sample values. Mantel’s tests are non‐parametric with tests for significance done through randomization of the data by permutation of the distance matrices. To create the distance matrices, all explanatory variables were first rescaled and normalized to Z‐scores so that all variables had common units. The dissimilarity matrices for explanatory variables and space (X, Y location) were computed using the Euclidean distances between values.

Predictive Maps

I used the zero line on each GAM plot to divide the range of the explanatory variable that had a positive effect from the range that had a negative effect, on the response variable. I applied simple threshold cutoffs from these GAM plots in a

Geographic Information System (GIS) framework to define dolphin habitat. Where the response curve was above the zero line, the explanatory variable was used to select habitat. This threshold approach is an extension of envelope models which generate reproducible results based on minimum and maximum values of explanatory variables

(Redfern et al. 2006). Using GIS, I selected grid cells from interpolated surfaces representing various variables (i.e. salinity, bottom type) based on the thresholds determined by the GAM models. I refer to this technique as ‘GAMvelopes’.

Model Evaluation

Model evaluation is a critical step in producing predictive species distribution models (Vaughan and Ormerod 2005; Rushton, Ormerod, and Kerby 2004). Although evaluation standards can vary depending on the goal of the model, most evaluation techniques must reflect the models ability to correctly select habitat, as well as exclude non‐habitat. I assessed the predictive capacity of each GAM model used to map dolphin habitat using the True Skill Statistic (TSS) (Allouche, Tsoar, and Kadmon 2006) calculated from traditional model accuracy measures of sensitivity and specificity. 51

Sensitivity and specificity are products of a confusion matrix used to evaluate binomial distribution models (See Table 2.1). A confusion matrix compares the ability of a habitat model to accurately predict observed presences and absences by tabulating true positives (TP), false positives (FP), false negatives (FN) and true negatives (TN) predictions. Sensitivity is a measure of commission error (TP/(TP+FN)) and specificity is a measure of omission error (TN/(TN+FP)). Sensitivity and specificity are calculated independently of each other and also independently of prevalence (the proportion of presence locations). Sensitivity and specificity values range from 0, indicating a high error rate, to 1 which describes perfect agreement between observed and predicted values. TSS is defined as:

TSS = Sensitivity + Specificity ‐ 1

Table 2.1: A confusion matrix used to tabulate the predictive capacity of presence/absence models. TP = presence observed and predicted by model; FP = absence observed but predicted as a presence location; FN = presence observed but location predicted as absence; TN = absence observed and predicted by model.

OBSERVED Presence Absence True Positive False Positive Presence (TP) (FP) False Negative True Negative Absence (FN) (TN) PREDICTED PREDICTED

Sensitivity and specificity have long been used to generate other measures of model performance (Fielding and Bell 1997; Segurado and Araujo 2004; Manel,

Williams, and Ormerod 2001; Pearce and Ferrier 2000) such as receiver operator curves

(ROC). The ROC technique generates probabilities which are inappropriate for models based on binomial data. The Cohen’s kappa statistic (Cohen 1960) is a popular accuracy measure of binomial predictions, but many recent studies criticize the kappa statistic for its inherent dependence on prevalence and subsequent generation of bias (Manel,

Williams, and Ormerod 2001; Fielding and Bell 1997; Vaughan and Ormerod 2005;

McPherson, Jetz, and Rogers 2004). TSS is a variant of a ROC that is applicable as an accuracy measure of binomial models. TSS is insensitive to both prevalence and the size of the validation set, combines both errors of commission and omission, and is simple to calculate and interpret. TSS ranges from ‐1 to +1, where +1 indicates perfect model performance. A zero value means the model performed no better than random and a negative value indicates that the model performed worse than random guessing would have predicted. TSS assigns equal weight to sensitivity and specificity which makes false positive as unwanted as false negatives.

Hypothesis Testing

As explained above, different survey methods were used in 2002 and 2003 than during the field seasons of 2004 and 2005. These two sets of years also experienced

different environmental conditions: 2002 and 2003 were wet years yielding similar conditions, as compared to the more dry years of 2004 and 2005. I performed the following analyses by creating dolphin habitat selection models with training datasets and evaluating each models’ predictive capacity with independent testing datasets. I matched testing and training datasets based on consistent field methods: I used data collected in 2004 and 2005 in my exploratory exercise; For Ho 1, I used 2004 data to train models and 2005 data to test each model; I explored Ho 2 with models trained based on

2002 data and tested by predicting data collected in 2003.

Exploratory Exercise

To examine the effect of the same environmental variables on predator and prey,

I used data from the Atlantic, Central and Gulf zones of Florida Bay in 2004 and 2005 to,

1) determine the predictors of dolphin presence/absence with binomial Mantel’s tests, and 2) compare GAM response curves of dolphin presence/absence and trawl catch to the same environmental variables.

The data files used in the binomial Mantel’s tests and GAMs were composed of all sighting locations and generated pseudo‐absences within each zone with associated explanatory environmental data for each point. I tested the following explanatory variables against dolphin presence/absence in the binomial Mantel’s tests: temperature,

salinity, turbidity, chlorophyll a, percent dissolved oxygen, distance from mudbanks,

CPUE, DPPUE, MPUE, and SPUE. I transformed the last five variables to a log +1 scale.

I performed three types of Mantel’s tests. First, global Mantel’s tests evaluated the significance of the overall effect of explanatory variables on dolphin distribution, space on dolphin distribution, and space on the variability of the explanatory variables.

Additionally, the effect of space can be accounted for when testing for the effect of explanatory variables on dolphin distribution. Likewise, the effect of the explanatory variables can be removed from a test of the effect of space on dolphin distribution. The results of these two tests can identify the driving force on dolphin distribution, space or the explanatory variables. If space is found to be significant, this is an indication that none of the tested explanatory variables are significantly correlated with the response variable.

The second type of Mantel’s test performed identified the spatial autocorrelation of each individual explanatory variable. The third type of test, termed pure partial

Mantel’s tests, identified which explanatory variable(s) are correlated to the dependent variable, dolphin presence/absence while accounting for multicolinearity and autocorrelation. The partial effect of each explanatory variable on the dependent

variable is given while accounting for correlations with all other explanatory variables and the effect of spatial structure. A significant result for the predictor variable space, indicates that there is inherent spatial structure in the distribution of dolphins which cannot be attributed to a correlation with any of the other tested explanatory variables.

The last step of my exploratory exercise modeled fish catch relative to environmental variability; I used the fish metric identified as most correlated with dolphin presence/absence by the Mantel’s tests. Datasets were composed of trawl catch data from all trawls conducted in the Atlantic, Central and Gulf zones of Florida Bay in

2004 and 2005. Those explanatory variables found to be correlated with the fish catch metric were then tested in a binomial GAM of dolphin presence/absence for that same zone. To conclude this exploratory exercise, I created and compared GAM plots of the response fish catch metric and binomial dolphin presence/absence to the same explanatory variables.

Ho 1

I used trawl catch data and independent surveys conducted in each zone

(Atlantic, Central and Gulf) in 2004 to create four types of models that used different sets of explanatory variables. The first model predicted dolphin presence/absence based on environmental factors (notation: DOLPHIN ~ ENV). The next model predicted dolphin

distribution based only on fish catch data (notation: DOLPHIN ~ FISH). The third model predicted dolphin habitat by modeling both environmental factors and fish catch data

(notation: DOLPHIN ~ ENV + FISH). The final model predicted areas of high fish catch

(CPUE) based on environmental factors (notation: FISH ~ ENV). This model essentially predicted habitat with high prey availability and assumed that hungry dolphins would distribute themselves appropriately. I created the first three types of models using presence and absence points from four surveys conducted in each zone in 2004. I developed models of the fourth type based on trawl catch data from all trawls conducted in each of the three zones during the 2004 summer.

I produced GAM plots for each model between the response variable and any explanatory variable with a significant parameter estimate (p < 0.05). I determined thresholds to be applied in GIS as predicted dolphin habitat from the range of each significant explanatory variable where the response curve fell above the zero line. I tested the predictive capacity of the four model types with presence and absence points from two independent surveys conducted in 2005 in each zone. Thresholds determined by the GAM plots were applied to daily water quality grids for each 2005 survey day, seasonal CPUE grids for each zone in 2005, and a static bottom type grid of each zone. I tabulated a confusion matrix for each predictive map applied to a 2005 survey day (two

in each zone). I subsequently calculated the sensitivity, specificity and true skill statistic

(TSS) for each predictive map. Additionally, I tallied the ratio of foraging sighting locations predicted to foraging sightings observed. Using six 2005 surveys as test datasets, I compared the TSS and percent foraging sightings correctly predicted between the four model types. Figure 2.4 depicts an example ‘GAMvelope’ predictive map based on GAM plots resulting from a DOLPHIN ~ ENV model in the Atlantic zone of 2004 and tested by a dolphin survey in the Atlantic zone on 23‐July‐05. The GAMs and threshold ranges applied to create the predictive maps of dolphin distribution for each model type tested in Ho 1 are described in Table 2.2.

Table 2.2: Description of the four types of GAMs created for each zone and tested in Ho 1: DOLPHIN ~ ENV + FISH, DOLPHIN ~ ENV, DOLPHIN ~ FISH, and FISH ~ ENV. Column one describes the number of presence and absence locations used to fit the first three GAMs for each zone. The number of trawls used to fit the last GAM (FISH ~ ENV) is described in the bottom row for each zone. The selected model, based on the minimum AIC produced by the backwards selection process, is described under GAM Model with the corresponding R2 value. Thresholds to be applied in GIS to produce predictive maps were only generated for those explanatory variables that were significant at the 0.05 level. Abbreviations: DOLPHIN = Dolphin presence or absence; Bottom = Bottom Type; SAL = Salinity; TEMP = Temperature; CHLA = Chlorophyll a; TURB = Turbidity; DO = Percent dissolved oxygen; CPUE = catch per unit effort; DPPUE = dolphin prey per unit effort; MPUE = Margalef’s richness index per unit effort; SPUE = Simpson’s diversity index per unit effort; Dist.MB = Distance from Mudbanks (log+1); Dense = Dense seagrass; INT = Intermediate seagrass; Sparse = Sparse seagrass; HBSG = Hardbottom with seagrass; Mudbank = Mudbank; MUD = mud; SAND = sand.

Models based on ‘GAMvelope’ thresholds applied in GIS data from Zone/Year Model Type GAM Model R2 (p‐value from GAM) Dolphins ~ Bottom + 59 DOLPHIN ~ s(DPPUE) + s(TEMP) + Bottom= HBSG, INT, Sparse; DPPUE < 0.005 or ENV + FISH s(MPUE) 0.9762 > 0.022 (0.05) Atlantic 2004: Dolphins ~ Bottom + 9 Presences DOLPHIN ~ s(SAL) + s(Dist.MB) + Bottom= HBSG, INT, Sparse; SAL < 38.4 (0.01); 137 Absences ENV s(TURB) 0.7188 TURB < 2.2 or > 2.5 & <3.0 (0.02) DOLPHIN ~ Dolphins ~ s(SPUE) + FISH s(DPPUE) 0.3761 SPUE > 0.0185 & < 0.0225 (0.05) 12 Trawls FISH ~ ENV CPUE ~ Bottom 0.3610 Bottom = INT

Table 2.2: (Continued)

Models based on ‘GAMvelope’ thresholds applied in GIS data from Zone/Year Model Type GAM Model R2 (p‐value from GAM) P.A ~ Bottom + s(DPPUE) + Bottom = Dense, Mudbank; SAL < 42 or >50 s(SAL) + s(TEMP) + (0.002); DPPUE > 0.011 and < 0.0225 (0.004); DOLPHIN ~ s(CHLA) + s(Dist.MB) + TEMP < 28.7 or > 31.3 (0.015); TURB < 4.1 Central 2004: ENV + FISH s(TURB) 0.7846 (0.016) 21 Presences P.A ~ Bottom + s(TEMP) + 206 Absences DOLPHIN ~ s(CHLA) + s(Dist.MB) + Bottom = Dense, Mudbank; TEMP <28.5 or >31 ENV s(TURB) 0.6676 (0.0009); TURB < 4.5 (0.002) DOLPHIN ~ FISH P.A ~ s(CPUE) + s(DPPUE) 0.2274 CPUE > 0.05 (<0.0001) 13 Trawls FISH ~ ENV CPUE ~ Bottom + s(SAL) 0.9807 Bottom = Mudbank P.A ~ Bottom + s(CPUE) + 60 DOLPHIN ~ s(TEMP) + s(CHLA) + Bottom = INT, MUD, SAND; DO > 102 (0.001); ENV + FISH s(TURB) + s(DO) 0.5827 TURB < 9 (0.04) Gulf 2004: P.A ~ Bottom + s(Dist.MB) + 21 Presences DOLPHIN ~ s(SAL) + s(TEMP) + Bottom = INT, MUD, SAND; Dist.MB < 4.1 or > 145 Absences ENV s(CHLA) + s(TURB) + s(DO) 0.5134 7.2 (0.004); SAL < 34.7 or > 38.1 (0.008) DOLPHIN ~ SPUE > 0.013 and < 0.0275 (0.003); CPUE < 0.08 FISH P.A ~ s(SPUE) + s(CPUE) 0.1750 or > 0.42 (0.005) 12 Trawls FISH ~ ENV CPUE ~ Bottom + s(TEMP) 0.9234 Bottom = INT, MUDBANK

Figure 2.4: Example ‘GAMvelope’ predictive map for the Atlantic region on 23‐July‐ 05. Habitat selected by thresholds generated from GAM plots shown modeling 2004 dolphin presence/absence based on environmental variables (DOLPHIN ~ ENV). The dashed horizontal black lines represents the zero line in each plot. The vertical dotted lines delineate the range of the explanatory variable above the zero line used as thresholds. The dashed lines bracketing the response curves are twice the standard error and function as confidence limits of the model. Tick marks on the x‐axis represent sampling intensity. Note that the scale on the y‐axis is different between plots. The habitat of six out of six sightings (open boxes) was selected including three out of three foraging sightings (demarcated with an F). The location of 45 out of 61 absences (open circles) was not selected as habitat. This predictive map yielded a sensitivity of 1, specificity of 0.7377, and a TSS of 0.7377. Notation: Dense = Dense seagrass; HBSG = Hardbottom with seagrass; Hard = Hardbottom; INT = Intermediate seagrass; Mudbank = Mudbank; Sparse = Sparse seagrass.

Ho 2

This hypothesis explicitly tests the predictive capacity of dolphin distribution by the FISH ~ ENV model type using large training (2002) and testing (2003) datasets. I created a five GAMs of CPUE for five different zones of Florida Bay based on trawls conducted in 2002. These fives zones were Atlantic, Central, Gulf, Eastern, and the combined region of the Flamingo and Western zones. The Flamingo and Western zones are geographically adjacent and have similar environments, which allowed me to combine the zones to increase the sample size of trawls used to create the 2002 fish catch

GAM. I applied derived ‘GAMvelope’ thresholds of CPUE versus significant explanatory variables in GIS to predict dolphin sightings from 2003 surveys in each zone

(See Table 2.3 for model specifications). The predictor variables tested for a significant relationship with CPUE were bottom type, salinity, temperature, chlorophyll a, turbidity, percent dissolved oxygen, and distance from mudbanks (log+1). Habitat was selected based on daily grids of the water quality metrics and a static bottom type grid for each zone. I overlaid dolphin sightings and generated pseudo‐absence points from each 2003 survey on the corresponding predictive map to assess sensitivity, specificity, the true skill statistic, and the percent foraging sightings correctly predicted. In total, I created five zonal dolphin habitat selection models from 99 trawls conducted in 2002. I validated these CPUE ~ ENV models with 36 independent dolphin surveys from 2003.

Table 2.3: GAM specifications of models used to test Ho 2: Catch Per Unit Effort as a function of environmental characteristics (CPUE ~ ENV) from trawls conducted in each zone in 2002. Abbreviations: Bottom = Bottom Type; CHLA = Chlorophyll a; Dist.MB = Distance from Mudbanks (log+1); Dense = Dense Seagrass; INT = Intermediate Seagrass; Mudbank = Mudbank.

CPUE ~ ENV # of Trawls GAM based on used to ‘GAMvelope’ thresholds data from create applied in GIS (p‐value from Zone/Year Model R2 GAM GAM) CPUE ~ Atlantic 2002 Bottom 0.4805 22 Bottom= Dense, Int CPUE ~ Bottom + Bottom = Dense, Int, s(CHLA) + Mudbank; Dist.MB = 0 or > Central 2002 s(Dist.MB) 0.9662 23 5.6 (0.002) CPUE ~ Eastern 2002 Bottom 0.2073 27 Bottom= Dense, Int, Mud CPUE ~ Gulf 2002 s(Dist.MB) 0.5827 14 Dist.MB > 5.75 & < 7.7 (0.04) Western & CPUE ~ Flamingo 2002 Bottom 0.4483 13 Bottom = Dense, Mudbank

Results

Exploratory Exercise

The global Mantel’s tests for each zone found significant correlation between the explanatory variables and dolphin presence/absence, even when the effect of space was

accounted for (Table 2.4). Many of the explanatory variables had significant spatial autocorrelation but the pure partial Mantel’s tests were able to compensate for these spatial trends to identify those explanatory variables with direct correlation to dolphin distribution. In the Atlantic zone during 2004 and 2005, CPUE and DPPUE were the only variables significantly related to dolphin presence/absence once the effect of spatial location was removed. Salinity, chlorophyll a, percent dissolved oxygen, and CPUE were significantly correlated with dolphin distribution in the Central zone. The predictor variables correlated to dolphin presence/absence in the Gulf zone in 2004 and

2005 were salinity, chlorophyll a, and CPUE.

Table 2.4: Mantel p‐value from binomial Mantel’s tests on 2004 and 2005 dolphin presence/absence, grouped by zone. Bold values significant at 0.09 cutoff. NA values for Percent dissolved oxygen result of broken sensor. NA** due to strong correlation between Turbidity and Chlorophyll a; necessary to remove turbidity from the analysis (Turbidity not significant without Chlorophyll a in dataset). Notation: Explanatory = explanatory variables; Dolphins = dolphin presence/absence; Space= the geographic distance between data points. See text for explanation of tests.

Atlantic Central Gulf Global Mantel Tests Explanatory on Dolphins 0.023 0.007 0.088 Space on Dolphins 0.511 0.967 0.509 Space on Explanatory 0.001 0.001 0.001 Explanatory on Dolphins accounting for Space 0.016 0.001 0.071 Space on Dolphins accounting for Explanatory 0.714 1 0.847 Spatial Autocorrelation of Explanatory Variables Temperature 0.002 0.001 0.001 Salinity 1 0.001 0.001 Turbidity 0.144 0.001 NA** Chlorophyll a 1 0.001 0.575 Percent dissolved oxygen NA 0.001 NA Distance from Mudbanks (log +1) 0.001 0.95 1 CPUE (log +1) 0.001 0.001 0.001 DPPUE (log +1) 0.001 0.001 0.001 MPUE (log +1) 0.001 0.001 0.001 SPUE (log +1) 0.001 0.001 0.001 Pure Partial Mantel’s Tests Temperature 0.56 0.993 0.497 Salinity 0.83 0.069 0.041 Turbidity 0.326 0.515 NA** Chlorophyll a 0.842 0.002 0.01 Percent dissolved oxygen NA 0.001 NA Distance from Mudbanks (log +1) 0.959 0.179 0.4 CPUE (log +1) 0.001 0.003 0.041 DPPUE (log +1) 0.064 0.998 0.959 MPUE (log +1) 0.473 0.573 0.447 SPUE (log +1) 0.506 0.786 0.842 Space 0.817 0.962 0.188 65

With the effects of autocorrelation and multicollinearity removed in the pure partial Mantel’s tests, CPUE was identified as the fish metric most correlated with dolphin presence/absence in all three zones. No other fish metric was found significant, with the exception of DPPUE in the Atlantic zone. Based on this result, CPUE was used as the response variable tested in the second step of this exploratory exercise (GAMs of trawl catch) and in the FISH ~ ENV models of Ho 1 and Ho 2.

Despite using different datasets to create GAM models of CPUE and dolphin distribution, results from this exploratory exercise showed strong evidence that dolphins and fish catch responded to the same environmental variables. Figure 2.5 compares

GAM plots between significant variables related to CPUE with GAM plots of binomial dolphin distribution versus the same variables in each zone. Each zonal GAM of CPUE ~

ENV returned bottom type and one water quality variable as significant explanatory variables. In Figure 2.5, the datasets used to fit each GAM and the resulting models are described in the right column adjacent to the plots from that model. Bottom type (plots in central column) was a significant explanatory variable in all GAMs. The left column for each zone (panel a, b, or c) compares plots of the same explanatory variable related to dolphin presence/absence and CPUE. The dashed horizontal black lines represents the zero line in each plot. The vertical dotted lines delineate the points at which the response

curves moves above or below the zero line. The dashed lines bracketing the response

curves are twice the standard error and function as confidence limits of the model. Tick marks on the x‐axis represent sampling intensity. Note that the scale on the y‐axis is different between plots. Notation: DOLPHIN signifies dolphin presence/absence; D =

Dense seagrass; H = Hardbottom; HS = Hardbottom with seagrass; I = Intermediate seagrass; M = Mud; MB = Mudbank; P = Patchy seagrass; S = Sand; SP = Sparse seagrass.

Figure 2.5: GAM plot comparison between significant explanatory variables for dolphin distribution (upper plots in each panel) and CPUE (lower plots in each panel) in the a) Atlantic, b) Central, and c) Gulf zones.

Salinity and bottom type were the significant predictor variables of trawl catch in the Atlantic zone. The GAM plots comparing the effects of salinity on fish catch and dolphin distribution showed that a similar range of salinity had a positive effect on both response variables (Fig. 2.5a). More fish were caught at salinity levels less then 37ppt, and in the GAM plot of dolphin presence/absence a similar range of salinity, less then

38ppt, had a positive effect on dolphin presence. Moreover, the same bottom types had similar effects on CPUE and dolphin presence in the Atlantic zone: Intermediate seagrass was positively associated with both dolphin presence and CPUE, while hardbottom habitats were negatively correlated with both dolphin presence and CPUE.

In the Central zone, more fish were caught in habitats with salinity levels greater than 47ppt and in mud or mudbank bottom types (Fig. 2.5b). The GAM plots of dolphin presence/absence depict that the same habitats had a positive impact on dolphin presence: salinity > 47ppt and in mud and mudbank bottom types. Sparse seagrass habitats had negative effects on both CPUE and dolphin presence. These GAM plots indicate that both dolphins and fish responded positively to the hyper‐saline conditions of the Central zone.

Chlorophyll a was significantly correlated with both CPUE and dolphin distribution in the Gulf zone. Plots from the Gulf zone GAM models show that chlorophyll a values between ~1.5 and 4 μg/l had a positive influence on dolphin presence and CPUE (Fig. 2.5c). The effects of bottom type were less congruent in the

Gulf zone; only intermediate seagrass and mudbank habitats had positive effects on both dolphin presence and CPUE.

Ho 1

The predictive capacity of dolphin habitat by the four model types created from

2004 data and tested by each 2005 dolphin surveys is described in Table 2.5. The TSS for all prediction maps ranged from a low of ‐0.3087, poor predictive capacity, to a high of

0.7377, very good predictive capacity. For every 2005 survey day used to validate the four model types, the resulting TSS values were used to rank each set of four models.

The highest TSS value received a rank of one and the lowest TSS value received a rank of four.

Table 2.5: Predictive capacity of dolphin distribution from 2005 survey data based on predictive maps created by the four tested model types in Ho 1 based on 2004 data for each zone. “Ratio of Foraging Sightings Selected” is the number of foraging sightings correctly predicted / number of foraging sightings observed. True Positive (TP), True Negatives (TN), False Negatives (FN), and False Positives (FP) were used to calculate Sensitivity (Se) and Specificity (Sp), which were then used to calculate the True Skill Statistic (TSS) for each prediction map. (See Methods for details.) True Observed False Survey TSS True # False # Not Not Commission TN/ TP/(TP+FN) (# (# (# (# 2004

Sensitivity Specificity of of from Omission TSS Sightings Foraging Selected) Selected) Selected

Ratio Absences Absences Sightings Sightings

Errors Errors

(Se+Sp Absences Zone Sightings Type

Negatives Selected) Selected) (

Positives Negative

TN+FP

Positive

Model Rank

Date 2005

Data

‐

: :

)

Dolphin ~ ENV +

71 22 July FISH 6 48 0/1 0 48 6 0 0 1 0 3 2005 Dolphin ~ ENV 6 48 1/1 6 29 0 19 1.0000 0.6042 0.6042 1 Atlantic Dolphin ~ FISH 6 48 0/1 1 32 5 16 0.1667 0.6667 ‐0.1667 4 FISH ~ ENV 6 48 0/1 4 35 2 13 0.6667 0.7292 0.3958 2 Dolphin ~ ENV + 23 July FISH 6 61 0/3 0 61 6 0 0.0000 1.0000 0.0000 3 2005 Dolphin ~ ENV 6 61 3/3 6 45 0 16 1.0000 0.7377 0.7377 1 Atlantic Dolphin ~ FISH 6 61 0/3 1 32 5 29 0.1667 0.5246 ‐0.3087 4 FISH ~ ENV 6 61 3/3 3 50 3 11 0.5000 0.8197 0.3197 2

Table 2.5: (Continued) True Observed False Survey True TSS False # # Not Not Commission TN/ TP/(TP+FN) (# (# (# (# 2004 Sensitivity Specificity of of from Omission TSS Sightings Foraging Selected) Selected) Se

Ratio Absences Absences Sightings Sightings

Errors Errors

(Se+Sp Absences Zone Sightings Type

Negatives Selected) Selected) (

l Positives Negative

TN+FP e Positive

Model Rank

Date 2005 t e of

d Data

‐

: :

)

Dolphin ~ ENV + 29 June FISH 8 34 0/2 0 34 8 0 0.0000 1.0000 0.0000 2 2005 Dolphin ~ ENV 8 34 0/2 0 34 8 0 0.0000 1.0000 0.0000 2 Central Dolphin ~ FISH 8 34 0/2 0 34 8 0 0.0000 1.0000 0.0000 2 FISH ~ ENV 8 34 2/2 4 30 4 4 0.5000 0.8824 0.3824 1 Dolphin ~ ENV + 1 July FISH 3 45 0/1 0 45 3 0 0.0000 1.0000 0.0000 2 72 2005 Dolphin ~ ENV 3 45 1/1 1 41 2 4 0.3333 0.9111 0.2444 1 Central Dolphin ~ FISH 3 45 0/1 0 45 3 0 0.0000 1.0000 0.0000 2 FISH ~ ENV 3 45 1/1 1 41 2 4 0.3333 0.9111 0.2444 1

Table 2.5: (Continued) True Observed False Survey TSS True # False # Not Not Commission TN TP/ (# (# (# (# 2004 Sensitivity Specificity of of from Omission TSS Sightings Foraging Selected) Selected) S

Ratio Absences Absences Sightings Sightings

Errors Errors /( e

(Se+Sp Absences Zone Sightings ( Type

Negatives Selected) Selected)

l Positives TP+FN Negative

TN+FP e Positive

Model Rank

Date 2005 e of

d Data

‐

: :

1) )

)

Dolphin ~ ENV + FISH 5 33 0/0 3 15 2 18 0.6000 0.4545 0.0545 1 2 July Dolphin ~ ENV 5 33 0/0 0 33 5 0 0.0000 1.0000 0.0000 3 2005 Gulf Dolphin ~ FISH 5 33 0/0 1 28 4 5 0.2000 0.8485 0.0485 2 FISH ~ ENV 5 33 0/0 0 30 5 3 0.0000 0.9091 ‐0.0909 4 Dolphin ~ ENV + FISH 4 35 1/1 3 8 1 27 0.7500 0.2286 ‐0.0214 3 3 July 73 Dolphin ~ ENV 4 35 0/1 0 35 4 0 0.0000 1.0000 0.0000 2 2005 Gulf Dolphin ~ FISH 4 35 0/1 0 33 4 2 0.0000 0.9429 ‐0.0571 4 FISH ~ ENV 4 35 1/1 1 33 3 2 0.2500 0.9429 0.1929 1

The overall evaluation of predictive performance by the four model types was determined using three ranking methods: Sum of TSS Ranks, Average TSS Value, and the Percent Foraging Sightings Correctly Predicted (Table 2.6). The order of predictive performance by the four model types was identical between the Sum of TSS ranks and the Average TSS value. These two evaluation methods determined that the DOLPHIN ~

ENV models performed best, but only slightly better than the FISH ~ ENV models.

However, the ability of the DOLPHIN ~ ENV models to predict foraging habitat was

25% worse than the ability of the FISH ~ ENV models. The FISH ~ ENV models were able to correctly predict dolphin foraging locations 87.5% of the time. The predictive capacity of the DOLPHIN ~ ENV + FISH models placed third with an average TSS value only marginally better than random. Those models which attempted to predict dolphin distribution based on fish distribution alone, DOLPHIN ~ FISH, performed the worst and were able to predict 0% of dolphin foraging sightings.

Table 2.6: Overall model performance by the four tested model types in Ho 1. GAMs trained with 2004 data and tested by dolphin presence and absence from 2005 surveys. Predictive capacity is assessed in three ways: the sum of TSS ranks (lower rank indicates a higher TSS, which means better performance), the average TSS for all survey dates, and the percent observed foraging sightings on each survey date correctly predicted by model.

Dolphin ~ ENV Dolphin ~ Dolphin ~ + FISH ENV FISH FISH ~ ENV Sum of TSS Ranks 14 10 18 11 Average TSS Value 0.0055 0.2644 ‐0.0807 0.2407 Percent Foraging Sightings Predicted 12.50 62.50 0 87.50

Ho 2

Predictive maps of 2003 dolphin distribution produced by GAM models of 2002

CPUE ~ ENV were evaluated by the TSS and percent foraging sightings correctly predicted (Table 2.7). Only 7 of 36 predictive maps produced a TSS score of zero or less.

Two maps had perfect predictive capacity and 3 other maps produced a TSS of 0.80 or higher.

Table 2.7: Ho2 prediction results of 2003 dolphin presence/absence by 2002 zonal CPUE ~ ENV ‘GAMvelope’ models. “# Foraging Sightings Selected” = # foraging sightings correctly predicted / # foraging sightings observed. (See Methods for details.)

Zone ATLANTIC CENTRAL 28‐ 27‐ 11‐ 13‐ 16‐ 24‐ 29‐ 28‐ 25‐ 18‐ 15‐ 18‐ 22‐ 29‐ 9 ‐ Jun Jun Jun Jun Jun Jun Jun Jul Jul Jul Jul Jul Jul Jul Jul

‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ 03 03 03 03 03 03 03 03 03 03 03 03 03 03 03

2003 Survey Date # of Sightings 3 3 3 1 1 2 2 2 5 1 3 1 1 4 1 # of Absences 11 32 4 5 35 12 19 38 27 3 138 2 10 76 21 # Foraging Sightings 1/1 0/1 2/2 0/0 0/1 0/0 0/0 0/1 1/1 1/1 1/1 1/1 0/0 1/1 1/1 Selected True Positives 3 2 2 1 0 2 2 1 4 1 2 1 1 3 1 (# Sightings Selected) True Negatives (# of 9 28 4 3 12 11 15 30 23 2 120 2 8 60 16 76 Absences Not Selected) False Negative (# 0 1 1 0 1 0 0 1 1 0 1 0 0 1 0 Sightings Not Selected) False Positive (# 2 4 0 2 23 1 4 8 4 1 18 0 2 16 5 Absences Selected) Sensitivity: Commission 1.00 0.67 0.67 1.00 0.00 1.00 1.00 0.50 0.80 1.00 0.67 1.00 1.00 0.75 1.00 Errors TP/(TP+FN) Specificity: Omission 0.82 0.88 1.00 0.60 0.34 0.92 0.79 0.79 0.85 0.67 0.87 1.00 0.80 0.79 0.76 Errors TN/(TN+FP) TSS (Se+Sp ‐1) 0.82 0.54 0.67 0.60 ‐0.66 0.92 0.79 0.29 0.65 0.67 0.54 1.00 0.80 0.54 0.76

Table2.7: (Continued)

Zone EASTERN FLAMINGO & WESTERN 23 15 11 5 6 7 8 9 14 16 10 12 8 8 9 ‐ ‐ ‐ ‐ ‐ ‐ Aug Aug Aug Aug Aug ‐ ‐ ‐ ‐ ‐ Jun ‐ ‐ ‐ ‐ Jul Jul Jun Jun Jun Jul Jul Jul Jul

‐ ‐ ‐ ‐ ‐ ‐ ‐ 03 03 ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ 03 03 03 03 03 03 03 03 03 03 03 03 03 2003 Survey Date # of Sightings 1 1 1 1 1 4 1 2 1 2 3 1 1 5 2 # of Absences 18 16 20 34 15 3 25 8 5 22 13 15 4 36 11 # Foraging Sightings 0/0 0/1 0/0 1/1 1/1 2/2 0/0 1/1 0/0 2/2 1/1 0/0 0/0 3/3 1/1 Selected True Positives 1 0 0 1 1 3 1 2 1 2 3 0 0 5 2 (# Sightings Selected) True Negatives (# of 16 14 17 19 6 1 14 8 0 11 1 15 3 3 3 Absences Not Selected)

77 False Negative (# 0 1 1 0 0 1 0 0 0 0 0 1 1 0 0 Sightings Not Selected) False Positive 2 2 3 15 9 2 11 0 5 11 12 0 1 33 8 (# Absences Selected) Sensitivity: Commission 1.00 0.00 0.00 1.00 1.00 0.75 1.00 1.00 1.00 1.00 1.00 0.00 0.00 1.00 1.00 Errors TP/(TP+FN) Specificity: Omission 0.89 0.88 0.85 0.56 0.40 0.33 0.56 1.00 0.00 0.50 0.08 1.00 0.75 0.08 0.27 Errors TN/(TN+FP) TSS (Se+Sp ‐1) 0.89 ‐0.13 ‐0.15 0.56 0.40 0.08 0.56 1.00 0.00 0.50 0.08 0.00 ‐0.25 0.08 0.27

Table 2.7: (Continued)

Zone GULF 10‐ 7 9 11‐ 10‐ 2 ‐ ‐ Aug Aug ‐ Aug Jul Jul Jul ‐ ‐ ‐ 03 ‐ ‐ 03 03 ‐ 03 03 03 2003 Survey Date # of Sightings 1 1 1 1 2 2 # of Absences 16 52 34 24 36 20 # Foraging Sightings 1/1 1/1 0/1 1/1 1/1 0/0 Selected True Positives 1 1 0 1 2 2 (# Sightings Selected) True Negatives (# of 5 21 11 13 15 7 Absences Not Selected) 78 False Negative 0 0 1 0 0 0 (# Sightings Not Selected) False Positive 11 31 23 11 21 13 (# Absences Selected) Sensitivity: Commission 1.00 1.00 0.00 1.00 1.00 1.00 Errors TP/(TP+FN) Specificity: Omission 0.31 0.40 0.32 0.54 0.42 0.35 Errors TN/(TN+FP) TSS (Se+Sp ‐1) 0.31 0.40 ‐0.68 0.54 0.42 0.35

The five 2002 models of CPUE ~ ENV produced positive average TSS values ranging from 0.2247 to 0.6557 when validated by 2003 observed presence and absence locations (Table 2.8). The average TSS value for 36 predictive maps was 0.3905, denoting that the maps performed on average 40% better at predicting dolphin habitat than random guessing would have produced. Additionally, each zone model was able to predict 60% or more of all dolphin foraging locations, including 100% accuracy in the combined region of the Western and Flamingo zones. In total, models of high fish catch habitat generated by 2002 trawl data accurately predicted 83% of all dolphin foraging sightings observed in 2003 in Florida Bay.

Table 2.8: Overall predictive capacity results of dolphin distribution by zone for Ho 2. Models of dolphin presence/absence trained by data from 2002 trawls (CPUE ~ ENV) and tested by 2003 dolphin surveys. Ratio for “Percent Foraging Sightings Correctly Predicted” is the number of foraging sightings correctly predicted / number of foraging sightings observed.

Number of Average Surveys TSS for all Percent Foraging Sightings ZONE Tested models Correctly Predicted (Ratio) Atlantic 7 0.5251 60 (3/5) Central 8 0.6557 86 (6/7) Eastern 5 0.3145 76 (2/3) Gulf 6 0.2247 80 (4/5) Western & Flamingo 10 0.2326 100 (10/10) TOTAL, ALL ZONES 36 0.3905 83 (25/30)

Discussion

My results indicate that incorporating prey availability data into fine‐scale models of dolphin habitat selection does not improve predictive capacity. However, predictive modeling of prey distribution based on environmental variability did produce high predictive performance of dolphin habitat selection, particularly foraging habitat. The FISH ~ ENV model assumes that dolphins will track their prey to maximize their encounter rate with potential prey items. This assumption is corroborated by the exploratory exercise that depicted an overall congruency of dolphin distribution and fish catch response to the same environmental characteristics. It is difficult to distinguish whether dolphin predators are tracking their prey, or the resources of their prey, but with the FISH ~ ENV approach to dolphin habitat modeling I successfully used environmental characteristics as proxies of prey distribution to predict dolphin distribution.

Model testing from Ho 1 concluded that the DOLPHIN ~ ENV or FISH ~ENV models were the most appropriate to predict fine‐scale dolphin habitat selection. Why did those models of dolphin distribution which included fish catch data (DOLPHIN ~

ENV + FISH and DOLPHIN ~ FISH) not perform as well? I conclude that the scale of fish sampling was inappropriate for such a fine‐scale study. Prey items are able move

between habitats within a grain size of 50 m on a weekly or possibly daily basis, and my fish sampling scheme was unable to capture such variability. The assumption that the spatial distribution of the fish community does not change during a three month summer field season was likely incorrect, leading to imprecise seasonal fish catch grids.

Conversely, the more frequent water quality sampling enabled greater spatial and temporal resolution, allowing me to capture this fine scale variability. Therefore, those models of dolphin distribution which used only environmental factors as explanatory variables (DOLPHIN ~ ENV) were more successful because of the increased spatial and temporal resolution at which these parameters were sampled. The scale at which data are collected and analyzed has important consequences to model output (Redfern et al.

2006; Levin 1992; Guisan, Thuiller, and Gotelli 2005) and previous studies have also encountered scale‐dependent relationships between marine predators and their prey

(Piatt and Methven 1992; Schneider and Piatt 1986; Guinet et al. 2001; Fauchald,

Erikstad, and Skarsfjord 2000).

The distribution of predators and prey revolves around complex trade‐offs between resource distribution, refuge availability, and predation risk, all of which have varying levels of importance relative to the temporal and spatial scale of the interaction.

Due to the mutual responses of predator and prey to each other, a mismatch in their

distributions at some spatial and temporal scales may be expected. For instance, the divergence of bottom type influence in the Gulf zone between dolphin presence and areas of high fish catch may be due to an increased encounter rate in this zone of Florida

Bay with sharks of the size and species that may modify the distribution of dolphins

(Torres, Heithaus, and Delius 2006 (Chapter 1); Wiley and Simpfendorfer 2007; Heithaus

2001). Therefore, predictive models of marine predator habitat use can either attempt to incorporate all variables relevant to predator‐prey interactions (i.e. anti‐predator tactics, resource availability, predation and competition effects, density‐dependent relationships), which most models cannot do (Redfern et al. 2006), or focus on those variables with weaker or no response to predator and prey distribution (ENV). The later option provides a simpler and effective approach to modeling dolphin habitat selection based on the distribution of the resources of their prey, as demonstrated in this study.

The link between dolphins and environment is easier to model than the relationship between dolphins and their prey because both predator and prey are mobile and biotic.

The distribution of a predator has less complex and more spatially consistent relationships with abiotic explanatory variables which determine the distribution of their prey than with the prey itself. At small spatial scales, in variable coastal ecosystems, it appears ineffective to predict the distribution of one complex biotic

predator based on the complex distribution patterns of multiple biotic prey items, especially because these two interact at multiple spatial and temporal scales.

Despite conducting almost 400 trawls in Florida Bay, my fish sampling methods were too broad to accurately describe temporal and spatial overlap between dolphins and their prey at the small scale these data were applied. To successfully incorporate prey distribution data into my predictive models in Florida Bay, I would need to trawl at a frequency equal to that of my environmental quality sampling scheme. This protocol would achieve an increased resolution of fish distribution but require increased effort that would significantly detract from dolphin survey time. Alternatively a different prey sampling method could be used. Benoit‐Bird and Au (2003) accredit their ability to identify congruent distribution patterns between spinner dolphins and their prey at multiple spatial and temporal scales to their employed sampling methodology. An echosounder was used to simultaneously measure the abundance of dolphins and prey in the water column. Such a sampling method has increased sampling resolution and overcomes problems of scale associated with modeling predator distribution based on prey distribution. In lieu of such a synchronous sampling protocol, my results suggest that, due to high habitat heterogeneity and the unpredictability of anti‐predator behavior, fine‐scale models of marine predator habitat selection in coastal habitats will

be more successful if environmental variables are used as proxies of both prey and predator distributions rather than relying on direct prey distribution data.

I conclude that methods and scales of prey data collection and incorporation into models must be appropriate and considered with forethought. For instance, depending on the spatial coupling between predators and their prey, direct prey data may be more effective in models of pelagic predator distribution. Models for pelagic ecosystems are typically applied to larger spatial extents but provide decreased spatial resolution. The resulting larger grid cells allow greater predictability of prey, which, in conjunction with minimal prey refuges, may allow increased predictability of predators based on direct prey data. Again, due to the patchy nature of fish distribution and abundance, fish sampling intensity must be enough so that the applied grain size is able to realistically represent the spatial and temporal variability of the ecosystem.

The FISH ~ ENV models performed well because there is no issue of spatial or temporal scale associated with these models. At every trawl location, environmental quality was sampled in situ. No spatial interpolations or temporal assumptions were necessary to associate fish catch rates with environmental variability. Therefore, when these models of fish catch based on environmental characteristics were produced, there

were no residual effects of scale to misrepresent the factors influencing fish catch distribution. The FISH ~ENV models in Ho 1 and Ho 2 predicted fish catch well and the dolphin predators distributed themselves appropriately, with 83% of 2003 and almost

90% of 2005 foraging sightings correctly predicted. These results reveal a strong spatial coupling between foraging habitat and areas of increased fish catch. Therefore, due to the importance of identifying feeding areas in conservation modeling applications and the lack of scaling issues associated with the FISH ~ ENV models, I believe that fine scale predictions of top marine predator distribution can be reliably produced by understanding and modeling the environmental factors that determine prey distribution.

As apex predators, bottlenose dolphins are frequently considered an indicator of healthy habitats (Torres and Urban 2005; Caro and OʹDoherty 1999). However, the mere presence of dolphins does not necessarily imply a healthier habitat, but a sighting of foraging dolphins does indicate a habitat with a prey resource. Therefore, because marine predators such as dolphins preferentially select different habitat characteristics when foraging as compared to socializing, traveling, or resting, behavioral data should be incorporated when using to predators as an indicator of relative habitat health because such data can identify the functional importance of habitats.

My method of pseudo‐absence generation worked well to contrast dolphin habitat (actual dolphin sightings) with non‐dolphin habitat (pseudo‐absences) and this method performed well with the use of the TSS to quantify the uncertainty of model‐ derived predictions. Moreover, the ‘GAMvelope’ technique was used successfully to define critical thresholds in the data and was easy to implement in a GIS to produce predictive maps. I characterized a complex marine ecosystem with a simplified approach using GAMs to explore trends in the datasets and generate habitat envelop models. I encourage others to apply these modeling methods in different ecosystems.

Conclusions

On average, measuring environmental conditions is easier, demands less time, and costs less than fish sampling. Moreover, at fine spatial scales, fish aggregations can be ephemeral leading to poor performance as a spatial proxy for predator distribution, while the environmental dynamics that aggregate prey are easier to sample and more consistent. Therefore, managers are left with two options to model and predict dolphin distribution at fine‐scales. High predictive capacity of dolphin habitat selection can be achieved by modeling either 1) environmental variability correlated with dolphin presence (DOLPHIN ~ ENV), or 2) the significant environmental factors that determine the distribution of dolphin prey (FISH ~ ENV). Option 1 assumes that dolphins and their 86

prey respond to environmental variability similarly (verified by the exploratory exercise) and that dolphin predators track their prey. Option 2 also assumes that dolphin predators track their prey. Predictive maps of dolphin distribution based on FISH ~ ENV models performed with similar success as DOLPHIN ~ ENV models, and were appreciably better at predicting dolphin foraging habitat, proving tight spatial links between foraging dolphins, their prey community and their environment.

Limits on funding, time and expertise force managers of marine predator

populations to make decisions about which predictor variables are the most appropriate to devote resources to measuring. Based on the sampling intensity necessary to successfully incorporate measures of prey into predictive models, my study concludes that it is not worth the effort or cost to include prey distribution data in fine‐scale predictive models of top marine predator habitat selection. The models including fish catch data (DOLPHIN ~ ENV + FISH and DOLPHIN ~ FISH) were unable to account for the spatial variation of fish distribution in Florida Bay, and subsequently the spatial relationship between dolphin presence and high prey densities decoupled. These results suggest that, without greatly increasing fish sampling effort, predictive capacity of fine‐ scale marine predator habitat selection does not improve by including data on prey distribution as a predictor variable in the model.

Chapter 3:

How to catch a fish? The ecology of bottlenose dolphin (Tursiops truncatus) foraging tactic fidelity in Florida Bay, Florida

Introduction

Competition for prey can facilitate evolution of novel foraging behaviors, as individuals should forage in a manner that maximizes their net energy intake rate

(Schoener 1974; Emlen 1966; Krebs 1978). Therefore, within an intraspecific community of competitors, individual predators may develop unique methods to search for, handle, and capture prey. Successful foraging tactics may be closely tied to specific environmental conditions, prey items, and/or particular individuals.

Foraging specializations by marine mammals have been reported for individual sea otters (Estes et al. 2003), between sympatric killer whale ecotypes (Baird, Abrams, and Dill 1992), and populations of humpback whales (Weinrich, Schilling, and Belt

1992), minke whales (Hoelzel, Dorsey, and Stern 1989), and bottlenose dolphins.

Bottlenose dolphins (Tursiops truncatus) exhibit considerable variation in habitats and prey items and display innovative foraging techniques: Mud plume feeding (Lewis and

Schroeder 2003), fish herding (Gazda et al. 2005), kerplunking (Connor, Heithaus et al.

2000), crater feeding (Rossbach and Herzing 1997), strand feeding (Hoese 1971), sponge‐ feeding (Smolker et al. 1997), and fishing gear depredation (Zollett and Read 2006; Read et al. 2003; Chilvers, Corkeron, and Puotinen 2003; Corkeron, Bryden, and Hedstrom

1990). This remarkable variation of foraging behavior raises questions about the factors

that promote such diversification and the associated behavioral, social and ecological ramifications.

In the Florida Bay ecosystem, bottlenose dolphins forage in a remarkably heterogeneous environment. Dolphins have developed numerous foraging tactics that may vary in their use spatially and across individuals. The goal of this study is to identify the major determinants of tactic choice: individual preference, spatial location, or habitat characteristics. Possible impacts of these foraging specializations on the ranging, social and reproductive patterns of this dolphin population are considered.

Baird, Abrams and Dill (1992) postulated that two populations of killer whales

(Orca orcinus) in the Northeast Pacific, known as ‘residents’ and ‘transients’, evolved and are currently speciating due to differences in target prey items and associated foraging tactics. These authors suggest that increased intraspecific competition led to density‐ dependent prey depletion, which encouraged some individuals to practice an alternative foraging strategy. Similarly, Estes et al. (2003) explored the mechanisms that led to the development of individual sea otter (Enhydra lutris) foraging specializations along the central California coast. These authors also propose that density‐dependent prey depletion increased intraspecific competition among sea otters and led to frequency‐

dependent diversification of individual foraging behaviors. Their conclusions are based on the framework presented by Partridge and Green (1985) that delineates three conditions which promote the diversification of a generalist population into alternative foraging strategies: environmental and prey variation, phenotypic variation among individuals, and frequency‐dependent prey choices relative to foraging habits of conspecifics. Application of the ideal free distribution theory (Fretwell and Lucas 1970) to the evolution of foraging specializations outlines a scenario in which individuals employ a foraging tactic that maximizes their intake rate even if the prey item is inferior energetically or less abundant. This approach dilutes competition for resources between predators because different prey items are consumed by individual predators. Like killer whales and sea otters, bottlenose dolphins are generalist feeders. Based on this foundation and the results from this analysis, the evolutionary mechanisms behind dolphin foraging specializations in Florida Bay and subsequent population level affects are explored.

I conducted field work in Florida Bay, which is composed of varied habitats and prey communities. I hypothesized that, 1) dolphins preferentially employ specific foraging tactics in certain environments, and 2) individual dolphins specialize in particular foraging tactics in accordance with their distribution patterns. Bottlenose

dolphins have plastic foraging behaviors and Florida Bay contains diverse environmental conditions and fish communities, but individual dolphins may specialize in one or two tactics and subsequently limit their ranging patterns to coincide with habitats that facilitate success using that foraging tactic.

Sargeant et al. (2007) explored the role of environmental variation in structuring foraging tactics within a the population of Indian Ocean bottlenose dolphins (Tursiops aduncus) in Shark Bay, Australia. In this population, although social learning may play a role in the development and maintenance of foraging tactics, habitat variation was also an important correlate of many tactics. In Florida Bay, I used a fine spatial grain of 50 m to explore foraging tactic choice relative to nine environmental variables and geographic location. Additionally, individual foraging tactic and site fidelity were examined to determine the consequences of habitat‐related foraging tactics on dolphin distribution patterns.

Methods

Study Site

Florida Bay lies at the southern tip of the Florida peninsula (Fig. 3.1). Florida Bay is open to the Gulf of Mexico on the western edge. Along the northeast boundary there are isolated passes to the southern end of Biscayne Bay. A few large (~ 3 km wide) passes on the southeastern boundary connect Florida Bay to the Florida Keys reef track.

Figure 3.1: Benthic habitat types and zones in Florida Bay. Updated version of “Florida Bay Bottom Types map” created by Robert Halley and Ellen Prager (1997), USGS Open‐file reports OFR 97‐526, USGS, Reston, VA. Thick black lines denote zone boundaries. Inset: Florida Bay lies at the southern end of the Florida peninsula (area in box enlarged).

Florida Bay is approximately 1800 km2 and is composed of a heterogeneous mosaic of benthic habitat types including dense seagrass, mud, sand and hardbottom areas composed of sponge and coral structures (Fig. 3.1). This heterogeneity persists at small spatial scales with benthic substrates and biological communities varying among

areas less than 10 m apart. An extensive network of mudbanks runs throughout the Bay.

These mudbanks and chains of mangrove islands delineate basins which are connected through isolated passes and channels. Previous research in Florida Bay sub‐divided the bay into environmentally distinct zones based on fish composition (Thayer, Powell, and

Hoss 1999; Sogard, Powell, and Holmquist 1989; Thayer and Chester 1989), water quality (Boyer, Fourqurean, and Jones 1999; McIvor, Ley, and Bjork 1994) and seagrass distribution (Hall et al. 1999; Zieman, Fourqurean, and Iverson 1989). These zone definitions were utilized to divide the bay into six relatively homogeneous regions:

Atlantic, Central, Eastern, Flamingo, Gulf, and Western (Table 3.1; Fig. 3.1).

Table 3.1: Description of six environmentally distinct zones in Florida Bay

Zone Typical water quality characteristics Dominant benthic habitat Atlantic Large passes to Atlantic reef track, normal Hardbottom salinity, high water clarity, low productivity, moderate to deep depth (1.5-3 m) Central Locked between mudbanks, moderate to Mud and mudbanks hyper-saline, low water clarity, high productivity, shallow depth (0.5-1.5 m) Eastern Greatest freshwater inflow, low to moderate Mixture of variable levels salinity, high water clarity, low productivity, of seagrass, mud, moderate depth (1.5-2.5 m) hardbottom, and barren Flamingo Generally shallow (1 m) with one wide deep Mudbanks and dense channel (2.5m), moderate to high salinity, seagrass beds high productivity, low water clarity Gulf Open to Gulf of Mexico, moderate salinity, Mixture of dense seagrass moderate to low water clarity, moderate beds, mud, sand productivity, deep depth (1.5-3 m) Western Moderate to high salinity, high water clarity, Dense seagrass beds low to moderate productivity, moderate depth (1-2 m)

Florida Bay is extremely shallow with depths ranging from less than 1 m to 4 m and can be described as where the gently sloping (3 cm / km) South Florida peninsula descends below sea level (National Research Council 2002; Obeysekera et al. 1999). This statement is particularly true in the interior region (Central and Flamingo zones) of

Florida Bay where the average depth is just 1 m. Additionally, mudbanks are most prevalent in this area which restrict water circulation (Fourqurean and Robblee 1999). 96

The shallow depth and minimal water flow in this area often lead to hyper‐saline conditions in the Central zone, peaking during my field work at 52 ppt.

Field Methods

A fine‐scale study of dolphin habitat selection was conducted in Florida Bay during the summers of 2002 through 2005. This field work consisted of trackline surveys for dolphin presence and absence. A 17ft Boston Whaler was driven at approximately 16 knots with three trained observers searching for dolphins within any distance range. The field crew stopped at all dolphin sightings to record data including GPS location, depth, benthic habitat type, water quality metrics (see below), group size and behavioral state.

Photographs were taken of the dorsal fin of all dolphins in each sighting to identify individuals through photo‐identification analysis (Würsig and Würsig 1977).

Additionally, focal follows were conducted on individual dolphins with distinct dorsal fins. Location, group size, behavior state, and habitat and water quality characteristics were recorded at 1:30 min intervals during focal follows. This field work yielded 248 dolphin sightings with foraging observed at 84 sightings.

Foraging Behavior Observations

A unique GPS location was recorded at each foraging event. Only one foraging event per sighting or focal follow was used in the dataset in order to achieve

independence of data. Each foraging event was classified into one of three distinct foraging tactic groups based on field notes: (1) deep diving with erratic surfacings, (2) herd & chase, and (3) mud ring feeding. Two other foraging tactics were observed in

Florida Bay but were not included in the analysis. “Bottom associated feeding with fluke‐out dives” could only be documented when water clarity allowed. To avoid this sampling bias, these observations were labeled as deep diving events. Additionally, a foraging tactic termed “Poke under seagrass rafts” was observed only five times, which did not allow an adequate sample size for analysis.

Deep diving with erratic surfacings (referred to as “deep diving”) is typified by dolphins performing fluke‐out dives (tail flukes come out of the water to allow a more vertical dive profile) with very quick surface times. Surfacing locations are irregular but the dolphin(s) stay in one general area and show no consistent direction of travel.

Herd & chase behavior involved frequent speed burst by the dolphin(s) in which it was possible to observe the fish, or group of fish, chased by the dolphin(s). This foraging behavior always involved the used of a barrier to herd the fish up against such as a mudbank, mangrove island, or seagrass bed.

Mud ring feeding is a dramatic foraging behavior only documented in Florida

Bay (Engleby and Waples 2001). One dolphin creates a mud plume in the shape of a ring. As the ring is about to close, the dolphin and any other conspecifics present (up to about six) wait at the surface for fish to jump out of the water, from inside the ring to outside the ring. The dolphin(s) then lunge for the fish in the air and catch them in their mouths (Fig. 3.2). These prey have been identified as mullet (Mugil sp.), and some specifically identified as white mullet (Mugil curema), a species known to be adapted to highly saline waters (Hoese and Moore 1992). This foraging behavior is unique from mud plume feeding described by Lewis and Schroeder (2003) because it requires cooperation among multiple individuals, occurs primarily in mud sediment habitats, the target prey is mullet, and the mud plume is circular.

Figure 3.2: Photograph of dolphins mud ring feeding in Florida Bay. See text for description of foraging behavior.

Water Quality Sampling

At each foraging location during the 2004 and 2005 field seasons, the following water quality metrics were recorded in upper and lower 0.5 m of the water column using a YSI 6600 Sonde (YSI Incorporated, Yellow Springs, OH): temperature (C°), salinity (ppt), turbidity (NTU), percent dissolved oxygen, and chlorophyll a. The average value of the upper and lower water column measurements was used for each location in

100

the following analyses because no discernable difference was detected between water samples from the well mixed, shallow waters of Florida Bay. At a minimum of two sites in each zone, water samples were obtained and filtered to calibrate the YSI‐based chlorophyll a fluoresence readings. Water samples were filtered with Whatman GF/F

25mm diameter filters with nominal porosity of 0.7 μm. These filters were later extracted in the lab and a Turner Designs Fluorometer was used to make absolute chlorophyll a readings. From these readings, a linear regression model was developed to convert the

YSI chlorophyll a readings into accurate chlorophyll a (μg/l) values. Water samples for chlorophyll a were collected at 12 spatially and temporally distributed sites in 2004 and seven sites in 2005, with filters extracted from three sub‐samples collected at each site.

Each sub‐sample of 500 ml was extracted for 24 hours and assessed using the fluorometer.

In 2002 and 2003, a YSI 30 was used to measure salinity and temperature at the mid‐point of the water column. To estimate turbidity, percent dissolved oxygen and chlorophyll a values during these years, data were acquired through the Southeast

Environmental Research Center’s Water Quality Monitoring Network. This program has

24 water quality stations with monthly sampling periods placed throughout Florida Bay.

Using ArcGIS® (v. 9.1, ESRI) these data points were interpolated with an inverse

101

distance weighting (IDW) technique which preserves local variation between sample points. In total, 18 separate rasters were created: one grid for each variable (turbidity, percent dissolved oxygen, chlorophyll a) for each month of the study period (June, July,

August of 2002 and of 2003). Finally, each foraging event location was used to sample the appropriate temporal set of three grids and these values of turbidity, percent dissolved oxygen and chlorophyll a were used in all further analyses.

Analyses

Four complimentary analytical techniques were used to understand the distribution patterns of the foraging tactics employed by bottlenose dolphins in Florida

Bay: Ripley’s K, Classification and Regression Trees, one‐way Analysis of Variance

(ANOVA), and analysis of the dolphin photo‐id dataset.

A Ripley’s K (Ripley 1981, 1979) point pattern analysis was conducted on the three foraging tactics to examine within group, as well as between group, distribution patterns. Ripley’s K is the cumulative frequency distribution of observations within a given distance class; the function tallies the proportion of points that fall within each distance class. A Ripley’s K analysis of the spatial distribution of points indicates the presence or absence of a pattern, as well as the scale and intensity of a pattern because it preserves distances at multiple scales. Ripley’s K analyses were run in the SpatialStats

102

module of S‐Plus (Venables and Ripley 1997, Insightful Corporation) which provides

95% confidence limits based on 100 trials. Edge effects, created by the boundary of a study area, can skew point pattern results because there are zero points where surveys were not conducted to one side of each boundary. To remove edge effects, the analyses were restricted to distances of half the smallest dimension of Florida Bay (25.7 km).

Two extensions of Ripley’s K were used in this study: 1) univariate tests that assess the clustering, over‐dispersion or random pattern of points within a group, and 2) bivariate tests that assess the patterns between points from two different groups.

Bivariate tests indicate whether the groups are attracted, repulsed or randomly distributed relative to each other. Results indicating repulsion or attraction between groups were interpreted as foraging tactics that occurred in spatially disparate areas or in spatially coincident areas, respectively. Traditionally, Ripley’s K analyses have been used in ecological studies of tree and seedling dispersal over variations in landscapes

(Barot, Gignoux, and Menaut 1999; Urban et al. 2000). I applied Ripley’s K to the distribution of point locations of foraging tactics used by dolphins in Florida Bay to assess the influence of space on tactic choice.

103

Significance of a Ripley’s K test is determined by generating confidence limits based on multiple randomizations of the test data. Because Florida Bay is a complex landscape which infers an overall spatial pattern of dolphin foraging locations (i.e.: dolphins cannot forage on mangrove islands or on impassable mudflats) I used a label permutation randomization technique. Confidence limits generated by label permutation identify spatial patterns of points which already exist within a predominant pattern by leaving all X and Y coordinates stationary but changing the label of each point within the test group(s).

In the second analytical method, a Classification and Regression Tree (CART)

(Venables and Ripley 1997) was used to describe the primary environmental variables influencing foraging tactic used by dolphins. A CART recursively partitions data using an algorithm that splits observations into groups based on a single best predictor variable chosen for each branch until all points are classified. This method of partitioning gives the maximum deviance in the response variable and the resulting sub‐ groups are partitioned until the final groups are relatively homogeneous. A CART analysis is nonparametric and does not assume linearity, homogeneous variables or independence. Although CART was originally created for use in the medical field

(Breiman et al. 1984), this approach to cluster analysis is increasingly used in ecological

104

studies that describe geographic distributions of points as they relate to environmental factors (Verbyla 1987; Iverson and Prasad 1998; Michaelsen et al. 1994; Torres et al. 2003).

The dataset used in the CART analysis was composed of all 84 observed foraging events with supplemental data for each event. This supplemental data included depth, temperature, salinity, turbidity, percent dissolved oxygen, chlorophyll a, distance from mudbank (log+1), bottom type and zone. The last two variables are categorical while the others are continuous. The resulting tree was pruned using a cross‐validation function to select the tree size with both minimum classification error and minimum tree complexity (number of branches).

The final analysis used an extensive photo‐id dataset to examine the individual site fidelity and foraging tactic fidelity of dolphins observed in Florida Bay during my study period. Two hundred and forty sightings occurred throughout the six zones of

Florida Bay, and included 437 photographically identifiable dolphins. To examine individual site and foraging tactic fidelity, the preliminary analysis focused on animals with multiple sightings. Using a Geographic Information System (GIS) all sightings of 61 individual dolphin seen five or more times each were plotted to examine site fidelity.

The use of five sightings was arbitrary, but offered a representative sample with which

105

to examine individual fidelity. To describe site fidelity, the frequency each individual dolphin was observed in each of the six zones was tallied. The number of foraging events each individual dolphin was present at was also summed and the foraging tactic used at each of those foraging events was recorded.

To elucidate the ranging patterns of all 437 identified dolphins throughout

Florida Bay, the group composition at 240 sightings was examined. The sighting history of each individual dolphin’s zone use was compiled (i.e.: dolphin X was observed 4 times in the Gulf zone and 1 time in the Flamingo zone). From this information, a second dataset was created to describe the group composition at each sighting based on the cumulative sighting history of each dolphin identified at that sighting. This dataset was then visually depicted in a spatial context.

Results

Foraging observations (n = 84) had an uneven distribution (Fig. 3.3). Mud ring feeding was observed 42 times, deep diving with erratic surfacings occurred 29 times, and herd & chase events were recorded 13 times.

106

Figure 3.3: Spatial distribution of 84 dolphin foraging events observed in Florida Bay. Triangles denote deep diving events. Stars represent Herd & Chase events. Circles are locations of mud ring feeding events.

Ripley’s K

A univariate Ripley’s K test of mud ring feeding events found this foraging behavior to be highly clustered at all spatial scales (Fig. 3.4a). This result is in contrast to the Ripley’s K univariate tests on herd & chase events (Fig. 3.4b) and deep diving events

107

(Fig. 3.4c). The overall spatial distribution of herd & chase foraging events showed no pattern except slight over dispersion at spatial scales between 12 and 18 km, but this is likely a spurious result due to a relatively small sample size. The univariate Ripley’s K test on deep diving events showed no point pattern until spatial scales of 10 km or greater, at which point this foraging tactic was clustered in space. However, the significant clustering of deep diving events was not as strong as observed for mud ring feeding events.

Both bivariate tests involving mud ring feeding events, contrasted with the distribution of herd & chase events (Fig. 3.4d) and deep diving events (Fig. 3.4e), showed strong spatial repulsion. This result indicates that mud ring feeding occurred in a spatially disparate area relative to all other foraging tactics. Few other foraging tactics were observed in areas where mud ring feeding events occurred. The bivariate Ripley’s

K tests between herd & chase and deep diving foraging events detected no significant point patterns at any spatial scale (Fig. 3.4f).

108

Figure 3.4: Ripley’s K plots describing the spatial distribution of foraging tactic groups. Left column (a‐c) are univariate tests of each foraging tactic group. Right column (d‐f) are bivariate tests between foraging tactic groups. Note different scales on the Y‐axis. Diamond shapes represent the observed values. Triangles are the upper confidence limit (UCL) and squares are the lower confidence limit (LCL), for each distance class. Each test used 50 distance classes with intervals of 513.56 m. Interpretation of univariate plots: observed line is above the UCL, points are clustered; observed line is below the LCL, points are over‐dispersed; observed line between UCL and LCL, no significant point pattern. Interpretation of bivariate plots: observed line is above the UCL, groups are spatially attracted; observed line is below the LCL, groups are spatially repulsed; observed line between UCL and LCL, no significant point pattern between groups. 109

Classification and Regression Tree

Depth was the primary variable selected by the CART to divide tactics into groups (Fig. 3.5). In areas of Florida Bay where the water depth is less than 1.5 m (the left side of the CART in Fig. 3.5), the vast majority of foraging events (42 of 51) were mud ring feeding. The variable zone was selected next in the model to subdivide this group further: 41 out of 43 foraging events in less than 1.5 m of water in the Central,

Eastern and Flamingo zones were mud ring feeding, while 7 of 8 foraging events in these shallow waters in the Atlantic or Gulf zones used the herd & chase foraging tactic.

110

Figure 3.5: Classification and Regression Tree of three groups of foraging tactics employed by bottlenose dolphins in Florida Bay. Terminal nodes are represented by squares, all other nodes are circles. Variables selected by CART at each branch are in italics. The dominant foraging tactic within each node is written, with the ratio describing the number of misclasses/total number of events. The misclassification error rate for the entire CART is 0.07 = 6/84.

The right side of the tree describes foraging tactic distribution in water depths greater then 1.5 m, where all but 4 out of 33 foraging events used the deep diving with erratic surfacings foraging tactic. Bottom type was the variable selected next in the

111

model to group tactics more homogeneously. All twenty foraging events in deeper waters in mud, sand, sparse seagrass or hardbottom with seagrass benthic habitat types were deep diving, creating a homogeneous terminal node. The remaining branch of the tree, in water greater than 1.5 m depth with bottom types of hardbottom, dense seagrass and intermediate seagrass, used the predictor variable turbidity to separate another deep diving group from another herd & chase group.

ANOVA Tests

Non‐parametric Kruskal‐Wallis ANOVA tests determined highly statistically significant differences between foraging tactics relative to depth and distance from mudbank (log+1) (Table 3.2; Fig. 3.6). Minor correlations between foraging tactics and other environmental variables (salinity, turbidity, and chlorophyll a) are likely due to environmental differences between zones of Florida Bay where foraging tactics tended to occur rather than dolphin habitat selection patterns.

112

Table 3.2: ANOVA results between environmental conditions at foraging tactic locations. Kruskal‐Wallis one‐way ANOVA results comparing environmental conditions at event locations of the three foraging tactic groups. Post‐hoc pairwise comparisons between foraging tactic groups conducted with Dunnʹs Method. Stars indicate a significant difference between groups at the P < 0.05 level. Dashes indicated a non‐significant result.

Distance Pairwise from Comparison Temperature Salinity Turbidity Chlorophyll a mudbanks Depth

Not H2=6.8 H2=8.9 H2=10.2 H2=28.3 H2=61.1 Significant (P=0.034) (P=0.012) (P =0.006) (P<0.001) (P<0.001)

113 Deep diving vs. --- *** *** *** *** *** Mud ring feeding Deep diving vs. ------*** *** Herd & Chase Herd & Chase vs. ------*** --- *** Mud ring feeding

The box plots confirm results of ANOVAs and describe the extent and direction of post‐hoc differences between foraging tactic groups (Fig. 3.6). While mud ring feeding and herd & chase events were observed closer to mudbanks than deep diving (Fig. 3.6a), mud ring feeding foraging behavior consistently occurred in shallower waters than herd

& chase and deep diving (Fig. 3.6b). No mud ring feeding events were observed in water depths greater than 1.5 m, exactly the same divide chosen as the primary splitting variable in the CART analysis.

114

Figure 3.6: Box plots depicting the (a) distance from shore, (b) depth, and (c) group size range of each foraging tactic group. Boxes range from the 25th to 75th percentiles. Black dots represent 5th and 95th percentile outliers. Solid line is the median value and the dotted line is the mean value.

115

A one‐way ANOVA was also conducted on dolphin group size variation between foraging tactic groups. A significant difference (F2,13 = 4.966, P = 0.0009) between dolphin group sizes at each type of foraging event was observed with a significant difference between the mean group sizes of deep diving (X ± SE = 11 ± 13.9, N = 29) and mud ring feeding (X ± SE = 4.5 ± 3.5, N = 42) dolphins (Tukey post‐hoc test; P = 0.007)

(Fig. 3.6c). This result is in contrast to a Kruskal‐Wallis one way ANOVA run on all 248 sightings which found a significant difference in group sizes among zones (H 5 = 17.787,

P = 0.003), but no significant pairwise comparison between zones.

Individual Site & Foraging Tactic Fidelity

The sighting histories of the 61 dolphins observed five or more times revealed dramatic site and foraging tactic fidelity. Individuals were rarely observed in more than one zone, either within or between years of the study. Additionally, no dolphin observed at a mud ring feeding event was ever seen at a sighting where the deep diving foraging tactic was employed, and vice versa.

Dolphins were classified based on their predominant zone of sighting locations:

Atlantic dolphins, Central dolphins, Eastern dolphins, Flamingo dolphins, or Gulf dolphins. Atlantic dolphins displayed extreme zone fidelity with all 38 sightings of seven dolphins in the Atlantic zone (Fig. 3.7a). The deep diving foraging tactic was

116

employed by these dolphins at 7 out of 8 observed foraging events (Fig. 3.7b). Similarly, very few (3% of 157) sightings of the 26 Gulf dolphins were outside the Gulf zone. Of the

60 foraging sightings Gulf dolphins were observed at, 72% employed deep diving.

Central and Eastern dolphins also showed high zone fidelity, with 93% and 97% respectively, of sightings within one zone. These 14 Central and 7 Eastern dolphins also displayed strong foraging tactic fidelity to mud ring feeding, used at 92% of their 52 foraging sightings. The seven Flamingo dolphins showed the most variability in zone use but were observed most frequently in the Flamingo zone (69%), and their foraging tactic of choice appeared to be mud ring feeding.

117

Figure 3.7: Zone and foraging tactic fidelity of 61 dolphins observed 5 times or more. a) The percent of sightings individual dolphins were observed in by zone. Numbers above each column denote the number of dolphins, and in parentheses are the number of sightings those dolphins were observed at. Grey = Atlantic; White = Central; Speckled = Eastern; Vertical stripes = Flamingo; Black = Gulf; Hatched = Western. b) The percent of foraging sightings broken down by foraging tactic used. Numbers above each column denote the total number of foraging sightings observed. Grey = Deep diving; Black = Mud ring feeding; White = Herd & Chase; Diagonal stripes = Other or unknown foraging tactic. 118

Site fidelity and foraging behavior of dolphins in Florida Bay was also explored using the entire photo‐id dataset of 240 sightings (437 individually identified dolphins).

First, to examine the flexibility of foraging tactic choice by individual dolphins, the sighting and foraging tactic histories of the 64 dolphins observed mud ring feeding were compared with those of the 138 dolphins observed foraging with the deep diving tactic

(Fig. 3.8). These two groups of dolphins demonstrated high, yet contrasting, zone and foraging tactic fidelity. Ninety‐three percent of sightings of mud ring feeding dolphins were observed in the Central, Flamingo or Eastern zones and 88% of their 133 foraging events employed the mud ring feeding tactic. These dolphins rarely used Herd & Chase

(8 events) or the deep diving tactic (2 events). Of those dolphins observed deep diving,

93% of their sightings were in the Gulf or Atlantic zones and 88% of their foraging events employed deep diving. Only two of these dolphins were observed at mud ring feeding events (on four occasions), and four other dolphins participated in four herd & chase events. This analysis underscores the overall tendencies of dolphins to specialize on a particular foraging tactic within a given zone.

119

Figure 3.8: Comparison of zone and foraging tactic fidelity by a) 64 dolphins observed at mud ring feeding events and, b) 138 dolphins observed at deep diving events. Numbers adjacent to pie slices indicate the number of sightings or events. Sightings by zone: Grey = Atlantic; White = Central; Speckled = Eastern; Vertical stripes = Flamingo; Black = Gulf; Hatched = Western. Foraging tactic use: Grey = Deep diving; Black = Mud ring feeding; White = Herd & Chase; Diagonal stripes = Other or unknown foraging tactic.

120

The cumulative sighting history of all photographically identified dolphins at each sighting is spatially depicted in Fig. 3.9. There are five distinct areas of relatively homogeneous zone use by dolphins. Sightings in the Atlantic zone were composed of animals rarely seen beyond the Atlantic region of Florida Bay. Likewise, the group composition of sightings in the Eastern, Gulf and Central zones were dominated by dolphins only previously observed in these zones, respectively. Areas of mixed group composition within sightings were also evident (depicted by dashed circles in Fig. 3.9).

The Flamingo region had the most diverse composition with frequent sightings of dolphins also observed in the Flamingo, Gulf, Central and Western zones. It is important to note that the Flamingo zone is substantially smaller (155 km2) than the Central (494 km2), Eastern (393 km2), Atlantic (369 km2), and Gulf (258 km2) zones. These zones definitions were constructed a priori, using variation in environmental factors, without incorporation of dolphin distribution patterns and area. It is interesting to notice that although dolphins observed in the Flamingo zone were occasionally observed in the western side of the Central zone, they were never observed in the eastern portion of the

Central zone. Additionally, dolphins observed in the Flamingo zone were also seen in the eastern reaches of the Gulf zone, but rarely was a dolphin observed in the Central or

Atlantic zones also seen in the Gulf zone.

121

Figure 3.9: Group composition at all dolphin sightings in Florida Bay. The cumulative sighting history, by zone, of each dolphin identified at all sightings is included in each pie chart. To avoid pie chart overlap, the black lines lead to actual locations of sightings. Pie charts are size scaled to reflect group size of each sighting. Dashed black circles indicate areas of increased heterogeneity of group composition within sightings.

Discussion

Mud ring feeding events were spatially clustered in a distinct area relative to all other dolphin foraging tactics in Florida Bay. Habitat characteristics associated with

122

mud ring feeding were shallow waters (<1.5 m) close to mudbanks. This foraging tactic only occurred in the Central, Flamingo and Eastern zones. In contrast, deep diving events occurred in deep water habitats (>1.5 m), further from mudbanks and typically in the Gulf or Atlantic zones. The herd & chase tactic was also used in shallow habitats near mudbanks, but it was primarily employed in different zones than mud ring feeding where prey other than mullet are available (unpublished data provided by Florida Fish and Wildlife Conservation Commission; Leigh Torres, unpublished data).

Those dolphins that used mud ring feeding primarily occupied the Central and

Flamingo zones (occasionally in the Eastern zone), and the deep diving foraging tactic was used exclusively by dolphins in the Gulf and Atlantic zones. Given the high spatial fidelity of dolphins and the strong correlation between environmental features and specific foraging tactics, it is perhaps not surprising that mud ring feeding was never observed in the Gulf or Atlantic zones. It appears that dolphins in the Central and

Flamingo zones have culturally evolved a novel foraging behavior, mud ring feeding, which only occurs in certain environments and is only practiced by individuals that occupy those environments. This foraging adaptation allows these dolphins to occupy a distinct habitat relative to other dolphins in Florida Bay. Central dolphins exploit prey in an unusually hyper‐saline and shallow environment with limited interspecific

123

competition for mullet from osprey (Pandion haliaetus), tarpon (Megalops atlanticus), and possibly nurse sharks (Ginglymostoma cirratum) (Leigh Torres, unpublished data). My results show that environmental characteristics, primarily depth and habitat type, have strong ecological effects on foraging tactic use, which promote the co‐existence of multiple foraging specialists within one population of bottlenose dolphins.

Foraging specializations are expected to occur in predator populations limited by resource availability where interspecific competition is weak and intraspecific competition is strong (Estes et al. 2003). Moreover, Smith and Skulason (1996) suggest that resource polymorphisms, such as foraging specialties, evolve in “open” niches or areas of underutilized resources. Based on these prerequisites, the environmental heterogeneity of the Florida Bay ecosystem may have allowed dolphin foraging specializations to evolve through opportunities to utilize a relatively unexploited habitat and resource.

I postulate that the heterogeneity of Florida Bay, with its corresponding prey variation, has led to the evolution of at least two discrete groups of dolphins that occupy distinct habitats and geographic locations, and utilize different foraging tactics: 1) The

Gulf and Atlantic dolphins who exploit the deeper exterior habitats of Florida Bay and,

124

2) the Central and Flamingo dolphins, adapted to living in the shallow tongue of the

Everglades that extends into the interior region of Florida Bay. These feeding adaptations are likely inherited through matrilineal lines during calf development much like the foraging specializations of California sea otters (Estes et al. 2003), bottlenose dolphins in Sarasota Bay, FL (Weiss 2006) and killer whales in the Crozet Archipelago

(Guinet 1991), but long‐term investigation is needed to document this process. It is not clear whether or not these two dolphin groups in Florida Bay are demographically or reproductively isolated, which could give rise to divergence, similar to that observed between resident and transient killer whale populations in the northeast Pacific (Baird,

Abrams and Dill 1992; Baird and Dill 1995). Like Florida Bay dolphins, the foraging tactics and prey items of these killer whale populations are different.

Although individual dolphins do not roam through the entire Bay, I do not assume that the boundaries of Florida Bay are the limits of their range. Rather, it is likely that the dolphins on the margins of the Florida Bay (in the Gulf, Atlantic, and Eastern zones) disperse beyond the Bay’s boundaries. However, the Central and Flamingo dolphins have limited physical access to other regions through tight and shallow passes.

Similar to the role of depth variation as a significant determent of dolphin foraging tactic, the bathymetry in Florida Bay may be an important factor of dolphin distribution

125

patterns. The mudbanks and shallow depths of the interior region of Florida Bay may prevent dolphins unfamiliar with the habitat, environment and prey from easily navigating across the Florida Bay landscape.

Foraging specializations, like mud ring feeding, can provide increased foraging success but could also limit foraging opportunities if (1) the behavior is obligate, and (2) the targeted resource or environment alter. Dolphins in Florida Bay exhibit variation, although slight, in their foraging tactics so it appears that these behaviors are not obligate. Yet, biomass of mullet in Florida Bay is lowest during the summer months

(Scott et al. 1989), when this research was conducted. Even with this relatively low abundance level of the targeted resource (approximately 64% below the yearly average biomass), dolphins continued to employ mud ring feeding to capture mullet in the

Central and Flamingo zones, instead of switching habitats or target prey items. The sighting histories of dolphins who do not employ mud ring feeding indicate that these dolphins do not occupy the shallow, interior areas of Florida Bay, but rather stay in habitats where deep diving, herd & chase or another foraging tactic can be employed

(Fig 3.8b). Strong intraspecific competition may have achieved a frequency‐equilibrium of foraging tactics where the relative amount of energy gained versus the amount of energy expended to catch each prey item is equalized between predators. Similar to

126

killer whale populations in the NW Pacific, bottlenose dolphin groups in Florida Bay

(those dolphins that mud ring feed and those dolphins that deep dive) may minimize intraspecific competition for prey through foraging tactics specialized to target certain prey items found in specific habitat types.

These divergent foraging strategies may manifest in the social dynamics of dolphins in Florida Bay. Group size variation in marine mammals can be a function of predation risk and/or foraging behavior (Packer and Ruttan 1988; Baird and Dill 1996;

Heithaus 2001; Heithaus and Dill 2002). With low predation risk group size is expected to be a function of foraging success, relative to the size of the prey school. Previous research on killer whales (Baird and Dill 1996) documented optimal group sizes to maximize net energy intake during foraging bouts. Conversely, if a greater predation threat exists, there are trade‐offs in group size between protection and foraging success.

Larger group sizes at deep diving events may be associated with a defense strategy against increased predation risk from sharks in the Gulf and Atlantic zones. Sampling in

Florida Bay has revealed few predatory sharks of size and species that could pose a threat to dolphins, with the exception of the Gulf and Atlantic zones (Torres, Heithaus, and Delius 2006 (Chapter 1); Wiley and Simpfendorfer 2007). In these latter areas, large sharks were caught, including a known predator of bottlenose dolphins, the bull shark

127

(Carcharhinus leucas) (Heithaus 2001). The small observed group sizes of dolphins at mud ring feeding events may be a reflection of relatively lower predation risk in the Central and Flamingo zones, as well as the optimal group size for successful mud ring feeding.

Mud ring feeding is a coordinated feeding strategy which requires communication and cooperation. Prey are shared among individuals participating in each foraging event.

Therefore, in the absence of predation risk to dolphins, the optimal group size should maximize the energetic reward for participating dolphins. This social dynamic is in contrast to deep diving foraging events that are typically performed by individual dolphins where success and failure depend on each dolphins’ pursuit. Thus, it is possible that there is greater selection for close social bonds among dolphins that employ mud ring feeding than dolphins who rely on a deep diving strategy. Here we see how depth, this time mediated through foraging behavior, distribution patterns and predation risk, could be an ecological determinant of dolphin social relationships in

Florida Bay.

The social implications of foraging specializations can be extended to gender specific roles. In the polygynous social system of bottlenose dolphins (Connor, Wells et al. 2000), female dolphins may use a habitat with reliable prey and high foraging success to maximize energy intake and support the demands of reproduction (Whitehead and

128

Mann 2000). In this social system, males will maximize reproductive opportunities by attempting to mate with numerous females (Connor, Read, and Wrangham 2000). It is possible that males from the Central and Flamingo zones travel greater distances than females to increase the number of reproductive opportunities. If this is true, males may be more plastic in their foraging behavior since they will encounter increased habitat and prey variation. At the present time I do not have sufficient information to address this hypothesis.

Despite South Florida’s minimal topographic relief (maximum of 6.22 m), small variations in elevation are defining physical characteristics that control surfacewater flow, water retention, and groundwater seepage (Obeysekera et al. 1999). Topographic changes of only a few cm strongly influence floral and faunal communities within the

Everglades ecosystem (Gleason and Stone 1997). Marine communities are similarly affected by this gradual slope with small bathymetric changes in Florida Bay shaping sediment characteristics and seagrass distribution which subsequently affect fish communities (Thayer, Powell, and Hoss 1999; Thayer and Chester 1989; Zieman,

Fourqurean, and Iverson 1989). The impacts of this slight, but influential, topographic relief within South Florida can be extended to include variation in dolphin foraging behavior, distribution and social ecology. The foraging success of wading birds, another

129

top predator in Florida Bay, is also affected by bathymetric structure because these birds are limited to feeding on shallow mudbanks where the length of their legs must exceed the water depth (Holmquist, Powell, and Sogard 1989). As South Florida inundates from continued sea level rise due to global warming (Wanless, Parkinson, and Tedesco 1997), the ecological implications may span from geology and hydrology, through vegetative community structure, and up to top predator distribution and foraging ecology.

My work demonstrates how foraging strategies, distribution patterns and social systems can evolve within a diverse habitat mosaic. Depth variation combined with spatially structured habitat types and prey communities appear to drive behavioral and ecological diversification within the bottlenose dolphin community in Florida Bay.

Dolphins exploit particular environments and employ distinct foraging strategies to maximize foraging success. These dolphins exhibit spatially coincident foraging tactics and spatial distribution patterns. The bathymetric and spatial structure of Florida Bay facilitate the existence of multiple dolphin foraging strategies, with subsequent constraints on the distribution patterns and social dynamics of dolphins that employ each particular strategy.

130

Chapter 4:

A kaleidoscope of mammal, bird and fish: Habitat use patterns of top predators in Florida Bay, Florida

131

Introduction

Habitat selection by predators is a function of both prey availability and resource competition (Fretwell and Lucas 1970; Schoener 1971), and, in marine ecosystems, predation is a significant determinant of species distribution (Heck and Orth 1980).

Species with similar requirements can reduce interspecific competition through disparate distribution patterns (Connell 1961; MacArthur 1958), to carve out distinct habitat niches. Such interspecific competition for resources can vary in intensity and the number of species involved. MacArthur (1972) defined the term ‘diffuse competition’ to mean the cumulative competitive effects of a number of interspecific competitors within an ecosystem. Competition between predators can be reduced by targeting different prey items, different prey sizes, prey in different habitats, or using different foraging tactics. These methods of prey partitioning act as ‘niche dimensions’ to reduce competition and facilitate coexistence among interspecific competitors: the greater the number of niche dimensions, the greater potential for diffuse competition (Pianka 1974).

May and MacArthur (1972) examined niche overlap between species as a function of environmental variability in an ecosystem with strong, constant competition.

Pianka (1974) built upon this work to describe his ‘niche overlap hypothesis’, applicable when the supply of resources is variable. Pianka illustrates that interspecific niche

132

overlap does not infer competition, but rather when resources are plentiful organisms can share them without conflict (high niche overlap = reduced competition). Conversely, disparate niches between competitors indicates competition avoidance when resources are scarce (low niche overlap = increased competition).

Comparative habitat use studies between seabird species or assemblages in relation to environmental conditions and foraging tactics have primarily been conducted in pelagic environments (Vilchis, Ballance, and Fiedler 2006; Jaquemet et al. 2005;

Hyrenbach et al. 2007; Hyrenbach et al. 2006; Pocklington 1979). Comparable studies in coastal ecosystems are rare (Baltz and Morejohn 1977; Skov and Prins 2001). Of all these studies only one included mammal competitors (Skov et al. 1995), or direct data on prey distribution or diets (Baltz and Morejohn 1977) in the analyses. In this study, I attempt to distinguish habitat use patterns of five seabird species groups and one dolphin species relative to each other, their prey fields and their environment within a dynamic coastal ecosystem.

Within the patchy landscape of Florida Bay, data were collected from standardized surveys for bottlenose dolphins (Tursiops truncatus), double‐crested cormorants (Phalacrocorax auritus), osprey (Pandion haliaetus), brown pelicans (Pelecanus

133

occidentalis), and terns (genus Sterna) and combined with synoptic fish and environmental sampling to address two questions: 1) Do these top predators in Florida

Bay exhibit habitat separation or overlap?, and 2) Do these habitat characteristics correspond with the habitat use patterns of each predator’s major prey items? These predators utilize different foraging tactics to locate and capture their prey and, while the prey items and sizes of these predators are diverse, there is overlap. Although it is unknown if any prey items are a limiting resource for these predators in Florida Bay, the potential for resource competition and the means of prey partitioning between predators are considered.

Due to the complexities of scale‐dependent predator‐prey interactions of both seabirds and marine mammals with their prey fields (Fauchald, Erikstad, and Skarsfjord

2000; Schneider and Piatt 1986; Vlietstra 2005; Chapter 2; Redfern et al. 2006), spatial scale is essentially removed from this analysis. Rather, habitat use patterns are established by comparing and contrasting the habitat qualities at the time and location each predator species was observed and each prey item was captured. Habitat characteristics and complexity play a significant role in predator and prey distributions in marine ecosystems (Heck and Orth 1980). The Florida Bay landscape is a diverse mosaic of habitat types that fosters a heterogeneous community of prey. This diversity

134

of habitats provides a variety of refuges for prey and foraging opportunities for predators. Therefore, the influence of habitat variability and structure on predator and prey distribution in Florida Bay is closely examined.

This study makes a few important assumptions: 1) prey items do not significantly shift habitat use patterns relative to predator distribution over the temporal scale examined, 2) there is minimal predation on the top predators examined in this study that would affect their habitat use patterns (including a low abundance of shark species that pose a predation risk to dolphins (Torres, Heithaus, and Delius 2006

(Chapter 1); Wiley and Simpfendorfer 2007)), and 3) competition from other top predators in Florida Bay (sharks, tarpon, large piscivorous fish) for similar prey items is minimal and does not influence the habitat use patterns of predators or prey examined in this study.

Methods

Study Site

Florida Bay lies at the southern tip of the Florida peninsula, is approximately

1800 km2, and is composed of a heterogeneous mosaic of benthic habitat types, water

135

quality gradients, and floral and faunal communities (Fig. 4.1). Previous research in

Florida Bay sub‐divided the bay into environmentally distinct zones based on fish composition (Thayer, Powell, and Hoss 1999; Sogard, Powell, and Holmquist 1989;

Thayer and Chester 1989), water quality (Boyer, Fourqurean, and Jones 1999; McIvor,

Ley, and Bjork 1994) and seagrass distribution (Hall et al. 1999; Zieman, Fourqurean, and Iverson 1989). These zone definitions were utilized to divide the bay into six relatively homogeneous regions: Atlantic, Central, Eastern, Flamingo, Gulf, and Western

(Fig. 4.1, Table 4.1). Florida Bay is extremely shallow with depths ranging from less than

1 m to 4 m. An extensive network of mudbanks and mangrove islands run throughout the Bay which delineate basins connected through isolated passes and channels. These mudbanks are typically covered with seagrass vegetation.

136

Table 4.1: Environmental description and percent bottom type cover for six zones. HBSG = Hardbottom with seagrass.

Zone Typical water quality characteristics Hardbottom HBSG Mud Mudbank Sand Seagrass Atlantic Large passes to Florida Keys reef track, normal salinity, high water clarity, low productivity, 53 1.5 0 8.5 0 37 moderate to deep depth (1.5‐3 m) Central Locked between mudbanks, moderate to hyper‐ saline, low water clarity, high productivity, shallow 9 0 8 39 0 44 depth (0.5‐1.5 m) Eastern Greatest freshwater inflow, low to moderate salinity, high water clarity, low productivity, 44 1 3 2 0 50 moderate depth (1.5‐2.5 m)

137 Flamingo Generally shallow (1 m) with one wide deep channel (2.5m), moderate to high salinity, high 0 0 17 55 9 19 productivity, low water clarity Gulf Open to Gulf of Mexico, moderate salinity, moderate to low water clarity, moderate 0 8 34 2 20 36 productivity, deep depth (1.5‐3 m) Western Moderate to high salinity, high water clarity, low to moderate productivity, moderate depth (1‐2 m) 0 0 1 50 0 49

All of 23 2 9 21 4 41 Florida Bay

Figure 4.1: Bottom types and zones of Florida Bay. Inset: Florida Bay lies at the southern end of the Florida peninsula (area in box enlarged). Thick black lines designate boundaries between the six environmentally homogeneous zones. Bottom types map based on USGS OFR 97‐526. Bottom types: gray = mudbanks, speckled = mud, horizontal lines = seagrass, hatched = hardbottom, diagonal lines = hardbottom with seagrass, vertical lines = sand, black = land.

Field Methods

Tracklines were surveyed throughout Florida Bay during the summer months

(June, July and August) of 2003, 2004 and 2005. A 17ft Boston Whaler was driven at approximately 16 knots with three trained observers searching for bottlenose dolphins 138

within any distance range and double‐crested cormorants, osprey, brown pelicans, royal terns (Sterna maxima), least terns (Sterna antillarum), and sandwich terns (Sterna sandvicensis) within a 100 m radius of the vessel. Water quality and bottom type (see below) were recorded at the start and end of each survey, at each dolphin sighting, and every 30 minutes if no dolphins were sighted. At all seabird and dolphin sightings, the field crew recorded the GPS location and the behavior state of the animal. Seabirds were classified as either sitting on stoop, sitting on water, flying or feeding. Dolphins were classified as foraging or non‐foraging. The species of any seabird present at a dolphin sighting was also recorded.

During the 2003 field season, the entirety of Florida Bay was surveyed 2 times to gather baseline information on the distribution of these piscivorous species. Based on these results, survey tracklines were established in areas of high species density within five different zones and run six times each during 2004 and 2005 (Fig. 4.2). This approach was taken to concentrate survey effort in areas of high habitat use by multiple species, within various environments, with possible increased competition and habitat partitioning among the predators observed.

139

Figure 4.2: Regions of concentrated survey effort during 2004 and 2005 field seasons in gray within larger zones outlined in black. Locations of fish sampling shown: Circle = trawl, Cross = gillnet set, Triangle = mullet seine.

Benthic Habitat and Water Quality Sampling

The United States Geological Society (USGS) produced a bottom types map of

Florida Bay in 1997 (USGS, OFR 97‐526). This map was used as the foundation for benthic habitat classification throughout Florida Bay. To refine this map, benthic habitat type was assessed and recorded at all water and habitat quality sampling locations.

140

These 1092 sampling locations were used to update and improve the resolution of the original USGS map (see Fig. 4.1). Habitat types were classified by visual inspection

(through the water column) or, when turbidity did not allow the former, use of a small bottom grab sampler (2” diameter; 3” deep). The same classification system used by the

USGS map was implemented with the addition of “hardbottom with seagrass” which were areas composed on equal amounts of sponge and coral habitat as seagrass habitat.

For this analysis six bottom type classifications were established: seagrass (all sparse, intermediate and dense seagrass habitats), mud, mudbank, sand, hardbottom, and hardbottom with seagrass.

During the 2004 and 2005 field seasons, the following water quality metrics were recorded in upper and lower 0.5 m of the water column using a YSI 6600 Sonde (YSI

Incorporated, Yellow Springs, OH): temperature (C°), salinity (ppt), turbidity (NTU) and chlorophyll a. The average value of the upper and lower water column measurements was used for each location in the following analyses because no discernable difference was detected between water samples from the well mixed, shallow waters of Florida

Bay. At a minimum of two sites in each zone, water samples were obtained and filtered to calibrate the YSI‐based chlorophyll a fluoresence readings. Water samples were filtered with Whatman GF/F 25mm diameter filters with nominal porosity of 0.7 μm.

141

These filters were later extracted in the lab and a Turner Designs Fluorometer was used to make absolute chlorophyll a readings. From these readings, a linear regression model was developed to convert the YSI chlorophyll a readings into accurate chlorophyll a

(μg/l) values. Water samples for chlorophyll a were collected at 12 spatially and temporally distributed sites in 2004 and seven sites in 2005, with filters extracted from three sub‐samples collected at each site. Each sub‐sample of 500 ml was extracted for 24 hours and assessed using a fluorometer.

In 2003, a YSI 30 was used to measure salinity and temperature at the mid‐point of the water column. To estimate turbidity and chlorophyll a values during this year, data were acquired through the Southeast Environmental Research Center’s (SERC)

Water Quality Monitoring Network. This program has 24 water quality stations with monthly sampling periods placed throughout Florida Bay. Using ArcGIS® (v. 9.1, ESRI) these data points were interpolated with an inverse distance weighting (IDW) technique which preserves local variation between sample points. In total, 6 separate rasters were created: one grid for each variable (turbidity and chlorophyll a) for each month of the study period (June, July, August of 2003).

142

Water quality sampling points from each survey day were interpolated into rasters of daily temperature, salinity, chlorophyll a, and turbidity variation. (With the exception of chlorophyll a and turbidity in 2003 which were monthly interpolations, as described above.) Rasters of static variables depth, bottom type and distance from mudbank (Euclidean) were also composed.

Species Diets

An extensive literature search on the diets of the studied predator species was conducted in order to facilitate realistic comparisons between the habitat use patterns of predators and potential prey items. Unfortunately, information on the diets of these predators in Florida Bay was only available for the osprey. Diet information on the other predators was gathered from studies in geographically adjacent areas in Florida, except for terns, for which diet information was very limited (Table 4.2).

143

Table 4.2: Diets of predator species in Florida Bay based on literature review. C = Common; M = Moderate; R = Rare. *Atlantic threadfin herring (Poludactylus octonemus) is also a major prey item of brown pelicans in Florida but were not caught in Florida Bay.

Diet

Major prey items found in Catch

Predator Florida Bay, Florida Frequency Frequency Region References Mojarra (Family Gerreidae) C C South Barros, Bottlenose Pinfish, grass & sheepshead porgy Floirda, 1987, 1993; Dolphin (Family Sparidae) C C SE Barros and Grunts, pigfish (Family United Odell, Haemulidae) C C States 1990; Striped mullet, white mullet Barros and (Family Mugilidae) C ‐‐‐‐‐‐ Wells, 1998 Gulf Toadfish (Opsanus beta) C C Gafftopsail catfish, hardhead catfish (Family Ariidae) C C Silver perch, spotted seatrout C M (Family Sciaenidae) Jacks (Family Carangidae) M M Spanish Hogfish (Family Labridae) R R Gulf Toadfish (Opsanus beta) C C Biscayne Cummings Double‐ Grunts, pigfish (Family Bay, 1987 C C crested Haemulidae) Florida Cormorant Parrotfish (Family Scaridae) C M Pinfish (Lagodon rhomboides) M C Filefish (Family Monachanthidae) M C Gray Snapper (Lutjanus griseus) M M Tern Silversides (Mendidia bermellina; General Bent, 1921; C R (Royal, Family Atherinidae) diet; McGinnis Least, & North & Emslie Herring/Sardines/Anchovy Sandwich) C R Carolina 2001 (Order Clupeiformes)

144

Table 4.2: (Continued)

Diet

Major prey items found in Catch

Predator Florida Bay, Florida Frequency Frequency Region References Striped mullet, white mullet Florida Odgen C ‐‐‐‐‐ Osprey (Family Mugilidae) Bay, 1968; Gafftopsail catfish, hardhead Florida; Ogden C C catfish (Family Ariidae) Florida 1973; Jacks (Family Carangidae) C M Ogden Spotted seatrout (Cynoscion 1978; M M nebulosus) Poole, 1989 Pinfish (Lagodon rhomboides) R C Filefish (Family Monachanthidae) R C Sheepshead porgy (Calamus penna; R M Family Sparidae) Striped mullet, white mullet Florida Fogarty et (Family Mugilidae) C ‐‐‐‐‐ al., 1981; * Brown Herring/Sardines/Anchovey Audubon, Pelican (Order Clupeiforms) C R 1972 Pinfish (Lagodon rhomboides) M C Silver perch, spotted seatrout (Family Sciaenidae) M M Silversides (Mendidia bermellina) M R Gulf Toadfish (Opsanus beta) R C Mojarra (Family Gerreidae) R C Grunts, pigfish (Family Haemulidae) R C Killifish/sheepshead minnow (Family Cyprinodontidae) R C Jacks (Family Carangidae) R M

145

Fish Sampling

Because of the diversity of fish distributed throughout Florida Bay and catch biases associated with each fish sampling method, three different methods were used to sample the fish community. The primary method used a 3‐m research otter trawl towed at approximately 4 km/hr for 3 minutes. The locations of 268 trawl sampling sites were randomly generated, and stratified by benthic habitat type, to sample the different bottom types within each zone (Fig. 4.2). A minimum of three trawls were conducted on each survey day. Secondly, because many faster, larger fish can out‐swim the bottom trawl, a 100 m gillnet composed of two 50 m panels of 3 in and 3 1/2 in mesh were soaked for 30 minutes. These gillnet sets were constantly monitored during soak times and were also stratified by bottom type and zone. These data provided a description of the distribution of larger, faster fish throughout Florida Bay. A total of 50 gillnet sets were conducted in 2003, and 32 sets in 2004. No gillnets sets were performed in 2005. All captured fish from trawls and gillnet sets were identified, measured, and then released.

GPS position, and all water and habitat quality metrics were recorded prior to each trawl and gillnet set.

Lastly, data on the distribution of mullet (Mugil sp.) was collected by the

Fisheries Independent Monitoring Program of the Florida Fish and Wildlife

146

Conservation Commission in Florida Bay during June, July and August of 1997 and

August of 2006 using a 183‐m haul seine (FWC‐FWRI 2007). Due to the speed and patchiness of mullet, only one mullet was caught in the 82 gillnet sets, which is highly unrepresentative of the large abundance of mullet in Florida Bay (Scott et al. 1989). Due to the importance of mullet in the diets of dolphins, osprey and pelicans in Florida Bay, data from these mullet seines were incorporated into this analysis to provide a description of their habitat association preferences. Depth, temperature, salinity, GPS location and number of mullet caught were provided by the FWC for each sampling location. Chlorophyll a and turbidity values were obtained through interpolated rasters of data provided by the SERC Water Quality Monitoring Network during same monthly time period in 1997 and 2006. A backwards‐step generalized additive model (GAM) was run on these data to describe the non‐linear relationship between mullet catch and the environmental characteristics at each sampling site.

Analysis

Royal terns, least terns and sandwich terns target similar prey items (Bent 1921) and therefore were grouped into a single ‘tern’ group to increase sample size. Only 4% of all seabird sightings consisted of feeding activity and this percentage decreased further when categorized by species. Therefore, birds were not grouped by behavior.

Foraging was observed at 28% of dolphin sightings, allowing a big enough sample size

147

to have 2 groups of dolphins: foraging and non‐foraging. The following analyses were conducted on six predator groups: Double‐crested cormorants (1985 sightings), brown pelicans (234 sightings), osprey (96 sightings), terns (229 sightings), foraging bottlenose dolphins (375 sightings), and non‐foraging bottlenose dolphins (765 sightings).

A three‐way Analysis of Variance (ANOVA) was performed on the sighting rates of the six predator groups. The three factors tested were year, zone and bottom type.

Survey routes, buffered 100 m on each side (survey range for seabirds), were used to sum the amount of survey effort (km2) in each year‐zone‐bottom type combination. The number of sightings of each predator group in each of these year‐zone‐bottom type combinations was tabulated and divided by the amount of survey effort. These date were log transformed to conduct the three‐way ANOVA.

Although data on the distribution of seabirds and dolphins in Florida Bay were collected as points, ordinations must be run on plot data. Therefore, predator point data were compiled into plot data using ArcGIS®. Each survey route from 2003, 2004, and

2005 was divided into 1 km segments and buffered on each side to 100 m (the survey distance for seabirds). These 1 km X 200 m segments became plots. The number and species of each sighting that fell in each plot was tabulated. Any plot with no sightings

148

was eliminated from the dataset. The average value of each static and water quality raster within each plot was determined, as was the percentage of bottom type cover.

Finally, the centroid of each plot was calculated and used at its spatial location. A total of 1275 plots were analyzed.

Mantels tests (Mantel 1967) were used to identify those explanatory variables significantly correlated with the distribution of the six predator groups while accounting for spatial autocorrelation among and between variables. Mantel’s tests are able to overcome many problems associated with examining species‐environment relationships because they are multivariate, explicitly test for the effect of space on the response variable, account for the effects of multicolinearity between predictor variables, and identify and account for spatial autocorrelation of explanatory variables (Schick and

Urban 2000). Mantel’s tests were run in S‐plus (v. 7.0, Insightful Corporation, Seattle) on the above described plot data. To reduce the number of variables incorporated into the ordinations, only those with significant relationships to species presence/absence from the Mantels tests were incorporated. Ordinations do not account for spatial autocorrelation. Therefore, the use of Mantel’s tests as a preliminary step to ordinations allows more confidence in the resulting ordination plots of species in habitat space. The tested explanatory variables were, a) the plot average values of chlorophyll a, salinity,

149

temperature, turbidity, depth, distance from mudbank (log+1), b) the proportion of each plot covered by seagrass, mudbank, mud, sand, hardbottom, or hardbottom with seagrass, and c) zone (converted to dummy variables).

Nonmetric Multidimensional Scaling (NMS) (Kruskal 1964; Shepard 1962) was used to describe the association between the six predator groups and the dataset of environmental variables identified by the Mantel’s test as significant. A NMS is a non‐ parametric ordination that is free from assumptions about linearity, dimensionality of the data, or underlying species responses to gradients. Each axis describes a gradient of habitat characteristics so that samples are plotted in ordination space to preserve the relative pairwise ecological distance among samples. Predator groups and samples from similar habitats are plotted closer together in ordination space. The statistical software package PC‐ORD (McCune and Mefford 1999) was used to conduct all NMS analyses and to create plots of species distributions in habitat space. The number of axes was chosen based on the results of a step‐down procedure and Monte Carlo tests to achieve low stress and stable solutions. A Sorensen (Bray‐Curtis) distance measure was used to create the distance matrices.

150

A NMS was used to describe the environmental distribution of the six predator groups within 1275 plots. The continuous variables (depth, distance from mudbank

(log+1), salinity, temperature, chlorophyll a, turbidity) were transformed using a general relativization (p=1) and the proportional data (percent plot cover of each bottom type) was transformed with an arcsine squareroot transformation.

Ecological community datasets are typically plagued by a large number of zeros.

Beals (1984) referred to this as the “zero truncation problem” and developed a data transformation method for such sparse community matrices that replaces binomial data points with probabilities of occurrence based on the pattern of co‐occurrences in the matrix (McCune 1994). Beals smoothing, as this transformation is called, is effective for heterogeneous data as it reduces the number of zero occurrences and improves the detection of compositional gradients (McCune 1994; Unterseher and Tal 2006). Beals smoothing was used to reduce the percent of zero occurrences in the predator species dataset from 80% to 1.425%.

Although over 69 fish and 12 invertebrates species were caught in the fish sampling methods described above, catch data used in the ordination analyses were lumped by taxonomic family into 14 groups of frequently caught fish that are known

151

prey of dolphins, cormorants, pelicans, terns or osprey (Table 4.3 and 4.4). All invertebrates were excluded from the analysis. Trawls and gillnets sets with no catch were excluded from the analysis.

Table 4.3: Description of fish groups used in trawl NMS. *The last three groups were identified as outliers and removed from the analysis.

Group Number Number of Name Caught Occurrences Group Description Family Gerreidae (Mojarra including spotfin MOJARRA 1915 172 mojarra, yellowfin mojarra, silver jenny) Lagodon rhomboides, Calamus arctifrons, Calamus penna (pinfish, grass porgy, SPARIDAE 1655 122 sheepshead porgy) Haemulon spp. (grunts including bluestriped, HAEMUL 737 78 white & french grunts, pigfish, tomtate ) MONACH 338 70 Monocanthus spp. (filefish) Family Sciaenidae (drums including silver SCIAENID 288 39 perch, spotted seatrout, cubbyu, high‐hat) Family Lutjanidae (snappers including lane, LUTJAN 147 61 mutton, schoolmaster, gray, yellowtail) TOADFISH 81 51 Opsanus beta (gulf toadfish) SCARIDAE 77 33 Family Scaridae (parrotfish) Family Labridae (wrasses, hogfish, spanish hogfish, slippery dick, pearly razor, LABRIDAE 59 25 puddingwife) Family Cyprinodontidae (killfish, sheepshead CYRPI* 1422 233 minnow) ATHERIN* 67 7 Family Atherinidae (silversides) Order Clupeiforms (herrings, sardines, CLUPEIF* 14 6 thread herrinhs, shads, anchovies)

152

Table 4.4: Description of fish groups used in gillnet NMS.

Group Number Number of Name Caught Occurrences Group Description Family Ariidae (catfish including gafftopsail & CATFISH 72 23 hardhead) Family Carangidae (jacks including crevalle, blue JACKS 21 9 runner, leatherjack, lookdown)

A NMS was run on a dataset consisting of 12 fish groups and 216 trawls with the following associated environmental variables: depth, distance from mudbank (log+1), salinity, temperature, chlorophyll a, turbidity, year, zone, bottom type. The last three variables are categorical and not included in the matrix but used as grouping variables.

Two groups (families Atherinidae and Clupeidae) were initially included in the ordinations but due to infrequent occurrences were identified as outliers because the average distance of these groups was much larger then the average of all other groups.

Additionally, the group CYPRI (family Cyprinodontidae) was identified as an outlier because of its predominance in a small number of trawls. Therefore these three groups were removed from the trawl dataset. A second fish catch NMS was conducted on 2 groups of fish, catfish and jacks (families Ariidae and Carangidae), which were caught in

27 gillnets sets. These fish are important prey items of dolphins, osprey and occasionally brown pelicans, and were not caught in trawl samples. Unfortunately different sampling methods preclude a single ordination of all fish catch. However, catch from gillnets in

153

this NMS was compared to the same associated environmental variables as the trawl dataset. For both the trawl and gillnet ordinations, the fish group dataset was relativized by species maximum to equalize common and uncommon species.

Results

Figure 4.3 compares the distribution of the six predator groups sighted during surveys in 2003, 2004 and 2005 in Florida Bay. Uneven survey effort partially accounts for the patchy distribution of sightings throughout the bay, yet comparisons can be made between species distribution. Cormorants appear to be distributed throughout the bay, with very high concentrations in the Atlantic zone. The distribution of terns is fairly uniform throughout Florida Bay. Dolphins were observed throughout the bay with foraging dolphins concentrated within particular regions of each zone. Osprey were mainly observed in the northern reaches of the bay, particularly within the Central and

Flamingo zones and pelicans were most abundant in the northwestern areas of Florida

Bay.

154

Figure 4.3: Comparison of the distribution and density of seabird and dolphin sightings from 2003, 2004 and 2005 surveys in Florida Bay. Shaded areas are regions were survey effort was concentrated in 2004 and 2005 and therefore have higher sighting rates. Circle: Cormorant; Square: Tern; Solid star: Foraging dolphin; Open star: Non‐foraging dolphin; Diamond: Osprey; Triangle: Pelican.

Results from the three‐way ANOVA indicated that variation in sighting rates of each predator group was consistently associated with zone (exception: Dolphin ‐

155

Foraging) and bottom type, but never year (Table 4.5). This result supports grouping the data from all three years for the Mantels tests and ordinations. The sighting rates in mudbank habitat was significantly different for each predator group except non‐ foraging dolphins (Fig. 4.4a). Osprey and pelicans sighting rates were significantly greater in the Flamingo zone (Fig. 4.4b).

Table 4.5: Three‐way ANOVA results for the sighting rate of each species group as associated with the factors zone, year and bottom type. Pairwise comparisons conducted with a Tukey test (α < 0.05). Zone codes: A=Atlantic, C=Central, E=Eastern, F=Flamingo, W=Western. Bottom Type codes: H=Hardbottom, HBSG=Hardbottom with Seagrass, MB=Mudbank, M=Mud, S=Sand, SG=Seagrass.

F‐ Significant Pairwise Species Group Factor Statistic P‐value Comparisons Zone 2.757 0.028 A‐W Cormorant Year 1.046 0.359 Bottom Type 4.244 0.003 MB‐S, MB‐H, MB‐M Zone 1.515 0.202 Dolphin ‐ Foraging Year 1.138 0.329 Bottom Type 3.35 0.011 MB‐H, MB‐S, MB‐HBSG Zone 4.091 0.003 C‐W, A‐W Dolphin ‐ Non‐ Year 2.033 0.142 Foraging Bottom Type 2.406 0.05 ‐‐‐‐ Zone 8.398 <0.001 F‐All Zones but C Osprey Year 1.476 0.238 Bottom Type 7.512 <0.001 MB‐All Bottom Types Zone 3.461 0.009 F‐E Pelican Year 2.125 0.130 Bottom Type 6.9 <0.001 MB‐All Bottom Types Zone 4.2 0.003 C‐W, C‐A Tern Year 1.517 0.229 MB‐All Bottom Types but Bottom Type 6.255 <0.001 M

156

Figure 4.4: Sighting rates of each species group by a) Bottom Type and b) Zone. Speckled = Cormorant, Black = Dolphin ‐ Foraging, White = Dolphin ‐ Non‐Foraging, Diagonal stripes = Osprey, Vertical stripes = Pelican, Horizontal stripes = Tern.

157

The Mantels tests indicated that 8 out of 13 explanatory variables were significantly correlated with the presence/absence of the six predator groups once spatial autocorrelation was accounted for: zone, chlorophyll a, salinity, turbidity, depth, mudbank, mud and sand (Fig. 4.5). Space (the XY coordinates) was marginally non‐ significant (p = 0.063) indicating that the majority of the variation in species distribution was explained by the explanatory variables. All bottom type characteristics (including those not found to be significant in the Mantels tests) were included in the NMS ordination of predator group habitat preferences. This choice was made to facilitate contrast between the significant bottom types of mud, mudbank and sand. If predator groups were not found in these bottom types it was reasoned that they must be in seagrass, hardbottom or hardbottom with seagrass habitats and therefore important to include in the ordination.

158

Figure 4.5: Path diagram of Mantels tests on 2003, 2004 and 2005 species group presence/absence compared to environmental characteristics. Arrows on the left side indicate spatial autocorrelation of individual variables. Arrows on the right side indicate significance of pure partial Mantels test in which autocorrelation of variables and correlation among variables is accounted for. Arrows are weighted by significance: thick lines are highly significant, a dotted line is less significant, and no arrow indicates no correlation.

The NMS procedure selected two axis which accounted for 88% of the variability observed in the structure of predator group distribution in habitat space (Fig 4.6). Mud,

159

turbidity, depth and hardbottom were the explanatory variables most correlated with the both axes (Table 4.6), and the predator groups associated differently with each axis.

The variability in Chlorophyll a concentration was also captured in both axes, and the influence of mudbank and seagrass habitats was describe by the gradient along axis 1.

Some of the predator groups appear to occupy different habitat space: Cormorants were in deeper water with less mud, turbidity and chlorophyll a and greater hardbottom and seagrass habitat. This is in contrast to foraging dolphins which were observed in shallow, muddy, turbid, productive waters with low seagrass. Non‐foraging dolphins, however, were found in habitats characterized by deeper water with relatively lower levels of mud, turbidity, and chlorophyll a, and greater hardbottom and seagrass habitat. Osprey and pelicans occurred in similar habitat space as foraging dolphins.

Table 4.6: Pearson’s correlation coefficients between ordination axes and explanatory variables from 2D NMS ordination of the six predator species groups.

Axis 1 Axis 2 Chlorophyll a ‐0.153 ‐0.127 Salinity 0.044 ‐0.001 Turbidity ‐0.232 ‐0.173 Depth 0.219 ‐0.003 Seagrass 0.132 0.045 Mudbank ‐0.192 0.014 Mud ‐0.253 ‐0.197 Sand ‐0.075 ‐0.094 Hardbottom 0.219 0.159 Hardbottom with seagrass 0.089 0.027

160

161

Figure 4.6: Nonmetric multidimensional scaling (NMS) plots, showing the habitat preferences of the six predator groups examined. Left panel is the combined plot: the amount of variation explained by each axis is in parentheses. Plus signs indicate where the species falls in ordination space. The list of variables at the edge of each axis indicates the variables, and their direction of magnitude, that characterize each axis. These lists are in order from most significant to least significant based on Pearson’s correlation coefficients. Right panel of plots describe the magnitude of each species at each plot in ordination space: Larger triangles indicate greater group association with that habitat space.

Two axes were also selected by the NMS run on the trawl data which cumulatively explained 51% of the variation. Depth was the dominant explanatory variable of fish group distribution in the habitat space created by the NMS run on the trawl data (Fig. 4.7, Table 4.7). Chlorophyll a, temperature and turbidity variation explained a small portion of the gradient along axis 2. The fish groups associated variably with each axis. The explanatory capacity of the categorical variables bottom type and zone were not included in the ordination but were used as grouping variables to symbolize each trawl (Fig. 4.8). With each trawl symbolized by category, it is evident that the distribution of trawl samples within the ordination space fell along gradients of bottom type and zone. Trawls on the left side of the ordination plot were predominantly conducted in mud or mudbank habitats within the Central and Eastern zones. This is in contrast to trawls performed in deeper habitats on the right side of the plot which primarily sampled sand, hardbottom, or hardbottom with seagrass habitats in the Gulf and Atlantic zones.

162

Table 4.7: Pearson’s correlation coefficients between ordination axes and explanatory variables from 2D NMS ordination on trawl catch data.

Axis 1 Axis 2 Depth 0.475 0.140 Temperature 0.046 ‐0.126 Salinity 0.090 ‐0.016 Chlorophyll a ‐0.007 ‐0.132 Turbidity 0.050 ‐0.118 Distance from Mudbank ‐0.019 0.032

163

164

Figure 4.7: Nonmetric multidimensional scaling (NMS) plot of trawl data, describing the distribution of 9 fish groups in habitat space. The amount of variation explained by each axis is in parentheses. Plus signs indicate where the fish group fell in ordination space. The list of variables at the edge of each axis indicates the variables, and their direction of magnitude, that characterize each axis. These lists are in order from most significant to least significant based on Pearson’s correlation coefficients.

165

Figure 4.8: Distribution of trawl samples within the habitat space created by the 2D NMS on 9 fish groups, symbolized by a) Bottom Type, and b) Zone. Bottom type symbols: triangles = seagrass, squares = mud, upside‐down triangles = mudbank, diamonds = hardbottom, circles = hardbottom with seagrass, X = sand. Zone symbols: triangles = Atlantic, square = central, upside‐down triangles = eastern, diamonds = flamingo, circles = gulf, X = western. Arrows below plots describe the general distribution of trawls by categorical group(s).

Catfish and jacks occupied opposite areas of the two axes NMS ordination plot that cumulatively explained 85% of the variation in fish catch by gillnet sets (Fig. 4.9,

Table 4.8). Catfish were caught in shallower water, closer to mudbanks, with higher chlorophyll a and turbidity levels. Conversely, jacks preferred deeper habitat, further from mudbanks, with lower chlorophyll a and turbidity levels. Because the categorical variable bottom type could not be integrated into the ordination, ‘distance from mudbank’ can be considered a proxy for the relative amount of mudbank bottom type available at each sampling site.

Table 4.8: Pearson’s correlation coefficients between ordination axes and explanatory variables from 2D NMS ordination of gillnet catch data.

Axis 1 Axis 2 Depth 0.182 0.249 Temperature 0.076 ‐0.150 Salinity 0.158 ‐0.096 Chlorophyll a ‐0.291 ‐0.255 Turbidity ‐0.160 ‐0.229 Distance from Mudbank 0.320 0.102

166

Figure 4.9: Nonmetric multidimensional scaling (NMS) plot from 2D NMS of gillnet catch, showing the habitat preferences of catfish and jacks. Amount of variation explained by each axis is in parentheses. Plus signs indicate where the species fell in ordination space. The list of variables at beside each axis indicates the variables, and their direction of magnitude, that characterize each axis. These lists are in order from most significant to least significant based on Pearson’s correlation coefficients. 167

A backwards selection GAM of mullet catch returned bottom type and depth as the significant predictor variables. Although the r2 value for this model was only 0.38, the p‐value of depth was 0.06 and the GAM plot of mullet catch versus depth depicts decreasing catch rates with increasing water depth (Fig 4.10a). The plot of bottom type relative to mullet catch shows greater catch rates in mud and mudbank habitats (Fig

4.10b).

168

Figure 4.10: GAM plots of mullet catch versus a) depth and b) bottom type. Ticks on the x‐axis represent sampling intensity. The dashed lines bracketing the response curve are twice the standard error and function as confidence limits of the model. Hard = hardbottom, Int = intermediate seagrass, Sparse = sparse seagrass. 169

Discussion

The ordination of the six predator groups described foraging dolphins, osprey and pelicans occupying similar habitat space: shallow, turbid, productive waters with high proportions of mud and mudbank bottom types (Fig. 6). These three predators also share common dominant prey items in Florida Bay, mullet and catfish, which were also described as occupying habitats of similar characteristics (Figs. 9 and 10). It is likely that within the Central and Flamingo zones of Florida Bay, where mud and mudbank habitats dominate, dolphins, osprey and pelicans are competitors for mullet and catfish prey. This competition is likely diluted by variation in foraging tactics among the predators used to capture their prey: Dolphins in Florida Bay employ a unique, cooperative foraging tactic to corral mullet and force them to jump out of the water where they wait with open mouths to catch the mullet out of the air (Chapter 3); Osprey dive feet first after prey and penetrate a maximum on 1 m below the surface (Poole

1989); The dominant foraging tactic of pelicans is to plunge about 1‐2 m into the water column and ‘gulp’ up prey in their pouch (Gunter 1958; Johnsgard 1993). Throughout my fieldwork, it was not uncommon to witness dolphins, osprey and pelicans foraging together. In fact, 10% of all osprey sightings and 5% of all pelican sightings were observed at foraging dolphin sightings, as compared to 3% of tern and 1% of cormorant sightings. Poole (1989) described such a scene:

170

“On Florida Bay, I have watched [osprey], six or eight at a time, hovering together over schools of mullet that [dolphins] (Tursiops truncatus) had driven into shallow water. Brown Pelicans (Pelicanus occidentalis) and various gulls often joined the diving melee, a pleasing kaleidoscope of mammal, bird, and fish...”

Other significant prey items of dolphins in Florida Bay are gulf toadfish, mojarra, pinfish (family Sparidae), grunts and pigfish (family Haemulidae), and silver perch and spotted seatrout (family Sciaenidae). These prey items were commonly caught in trawls conducted in Florida Bay and, similar to their dolphin predators, the ordination determined their preferred habitat space to be shallow waters with high chlorophyll a and turbidity levels, composed of mud and mudbank habitats (Figs. 7 and 8). This is in contrast to three of the major prey items of cormorants: parrotfish (family Scaridae), filefish (family Monachanthidae), and gray snapper (Lutjanus griseus). These three groups of prey items were caught in deeper habitats with low chlorophyll a and turbidity levels, in seagrass, sand, hardbottom or hardbottom with seagrass bottom types of the Atlantic and Gulf zones. These habitat characteristics correspond with the habitat conditions described for cormorants by the predator group ordination: deeper water with low chlorophyll a and turbidity levels, less mud and mudbank habitat, and a greater proportion of hardbottom and seagrass bottoms.

While cormorants, like dolphins, are also known to consume gulf toadfish, pinfish, grunts and pigfish, cormorants in Florida Bay may concentrate their foraging

171

behavior on prey items with less interspecific competition. Although cormorants inhabit all regions of Florida Bay and had the highest sighting rates in each zone, cormorant sighting rates were especially dominant in areas of Florida Bay where mud and mudbank bottom types are minimal. In the muddy regions of the Central, Flamingo and

Western zones, sighting rates of dolphins, osprey, pelicans and terns rivaled that of cormorants. The habitat use patterns of cormorants described by the ordination is typified by deeper, clearer waters in hardbottom and seagrass habitats of the Gulf and

Atlantic zones. This habitat description is highly distinct from that of foraging dolphins.

Of the five predator species examined in this study, only dolphins and cormorants are able to exploit benthic prey items. Osprey, pelicans and terns are limited to surface prey.

Therefore, cormorants in Florida Bay may concentrate their foraging efforts on less competitive prey items and occur more frequently where these prey items dominate. It is possibly for this reason that cormorants and foraging dolphins occupy the most disparate habitats.

Prey size can also influence resource partitioning between predators. Dolphins generally pursue larger (35‐350 mm) prey than cormorants (<150 mm) with some overlap in size range (Barros 1993; Barros and Wells 1998; Johnsgard 1993). This mechanism of prey division may provide a less competitive foraging niche for

172

cormorants in the mud and mudbank habitats of Florida Bay and help to explain their continued use of these habitats. However, the prey items of osprey (250‐350 mm) and pelican (64‐257 mm) are within the same size range as dolphins, which increases the potential for competition between these predators for mullet and/or catfish (Poole 1989;

Pinson and Drummond 1993).

The behavior state of non‐foraging dolphins was either travel, socialize, rest or unknown. With the exception of resting behavior, these behaviors do not generally require any particular habitat characteristics and therefore contrast well with the habitat selection patterns of foraging dolphins that are constrained by relative prey availability.

While resting, dolphins may select for habitat that is more protected and with less risk of predation. These characteristics are also distinct from foraging dolphin habitat requirements. Indeed, the habitat characteristics of non‐foraging dolphins derived from the ordination were deeper waters, with lower turbidity and chlorophyll a levels and dominated by hardbottom and seagrass bottom types. This greatly contrasts the habitat space of foraging dolphins.

The habitat characteristics of terns fell in the middle of the ordination plot with no other species group in close proximity. This result denotes a more general habitat

173

preference pattern, unlike that of other predator species examined. Unfortunately, the prey items of terns could not be included in this analysis, mainly for lack of occurrence in fish sampling effort. Based on these two factors, it is difficult to draw conclusions about the habitat use of terns relative to other predators and their prey items in Florida

Bay. However, similar to the other predator groups examined, except non‐foraging dolphins, the sighting rate of terns increased in mud and mudbank bottom types.

Depth variation and bottom type consistently manifested as significant predictor variables. These static, benthic descriptors influenced the distribution of the six predator groups examined, fish catch from trawls, fish catch in gillnet sets and mullet catch from seines. Additionally, when survey effort was accounted for, all four seabird groups and foraging dolphins were found in mudbank habitats most frequently. Although mudbanks comprise about 25% of Florida Bay’s area (Zieman, Fourqurean, and Iverson

1989) previous research has documented that these mudbanks play a disproportionate role in the ecosystem. The mudbanks support half the bay’s standing crop of seagrass

(Powell, Holmquist, and Sogard 1989) and act as barriers to tides and currents which restrict circulation and cause hyper‐saline conditions (Wanless and Tagett 1989). These characteristics influence fish (Sogard, Powell, and Holmquist 1989; Thayer and Chester

1989) and crustaceans (Holmquist, Powell, and Sogard 1989) communities. The

174

mudbanks also provide essential foraging grounds for wading birds (Holmquist,

Powell, and Sogard 1989). This study documents that mudbanks in Florida Bay are also important habitat for the top predators within this ecosystem, including foraging bottlenose dolphins.

In Florida Bay, the roles of depth and bottom type are confounding: as depth varies so does bottom type. In the Florida Bay and Florida Keys region, depth has proven to influence seagrass (Hall et al. 1999; Landry 2005) and hardbottom (Shinn,

Lidz, and Harris 1994) communities which consequently effects prey availability

(Thayer, Powell, and Hoss 1999). Additionally, although the bedrock topography in

Florida Bay has no dramatic relief, subtle irregularities are the determinants of depth variation and, therefore, mudbank formation and location (Wanless and Tagett 1989).

These seagrass‐covered mudbanks concentrate diverse faunal communities in a narrow

(shallow) water column (Holmquist, Powell, and Sogard 1989; Sogard, Powell, and

Holmquist 1989) which attract a variety of predators.

In summary, it is the minimal variation of Florida Bay’s underlying limestone surface (1‐3 m below ground) that cause shifts in bathymetric structure and bottom type, thus influencing the floral and faunal communities, which subsequently determines the

175

distribution of top predators such as cormorants, osprey, pelicans, terns and foraging dolphins. Due to the concentration of prey in mudbank habitats, habitat use patterns of top predators overlap and may influence competition. Therefore, these predators must partition resources through other mechanisms such as prey size (i.e. dolphins and cormorants), foraging tactic (i.e. dolphins, osprey and pelicans), or prey item (i.e. dolphins and terns). Although this study identified habitat use by predators and prey, temporal and spatial scales were largely unaccounted for in this analysis. Therefore, it is also possible that predators minimize interspecific competition through spatial and/or temporal habitat segregation. These results highlight the importance of shallow, mudbank habitats in Florida Bay and raise questions about the effects of expected sea level rise throughout South Florida on the ecology of top predator populations.

Despite Florida Bay’s extreme habitat heterogeneity, there is high habitat overlap between predators within one particular habitat: the productive seagrass‐covered mudbanks. Application of Pianka’s ‘niche overlap hypothesis’ to this pattern of strong niche overlap by predators, implies that prey resources may not be limiting in Florida

Bay. Or, the diversity of foraging tactics and target prey items and sizes may provide ample ‘niche dimensions’ to adequately partition the resources and foster ‘diffuse competition’. An insightful companion study to these results would be to conduct this

176

research during a time of low resource supply in Florida Bay. It is expected that these predators would exhibit less habitat overlap and concentrate habitat use patterns in niches most suited to their behaviors and prey items. Despite currently adequate niche dimensions and resource availability, in a scenario of limited prey, a little bit of competitive inhabitation by many species (diffuse competition) can be equivalent to strong competition between a few species (Pianka 1974).

177

Chapter 5:

A ‘Trophic Fountain’: The bottom-up synthesis of top predator distribution and foraging ecology in Florida Bay, Florida

178

This chapter represents a bottom‐up synthesis of my doctoral dissertation at

Duke University on the distribution and foraging ecology of top predators in Florida

Bay, Florida, from environmental characteristics to predator behavior. Florida Bay is approximately 1800 km2 and lies at the southern tip of the Florida peninsula. It is a shallow bay positioned at the confluence of three water sources: the Gulf of Mexico to the west, the Atlantic Florida Keys reef track to the east and south, and freshwater flow through the Florida Everglades from the north (Fig. 5.1). I collected data in Florida Bay during four summers between 2002‐2005 on the distribution of bottlenose dolphins

(Tursiops truncatus), double‐crested cormorants (Phalacrocorax auritus), osprey (Pandion haliaetus), brown pelicans (Pelecanus occidentalis), royal terns (Sterna maxima), least terns

(Sterna antillarum), sandwich terns (Sterna sandvicensis), and sharks (numerous species in the subclass Elasmobranchii). My goal was to understand the ecological determinants of predator habitat selection and distribution in Florida Bay as a function of habitat variability, prey distribution, predator behavior, and competition and predation effects.

To achieve such an understanding, I studied aspects of the entire ecosystem from environmental quality (including productivity and physical properties such as temperature and salinity), through the structure of prey communities, up to predator distribution and behavior (Fig. 5.2).

179

Figure 5.1: Benthic habitat types and zones in Florida Bay. Updated version of “Florida Bay Bottom Types map” created by Robert Halley and Ellen Prager (1997), USGS Open‐file reports OFR 97‐526, USGS, Reston, VA. Thick black lines denote zone boundaries.

180

Figure 5.2: Schematic of dissertation research on the determinants of top predator distribution and foraging ecology. A simplified three level ecosystem including the environment, prey communities, and predator distribution and their behavior effects.

Florida Bay is composed of a heterogeneous mosaic of benthic habitat types, water quality gradients, and floral and faunal communities. In my field work and analytical methods, I defined six environmentally distinct zones based on findings from my own and previous studies in Florida Bay related to water quality (McIvor, Ley, and

Bjork 1994; Boyer, Fourqurean, and Jones 1999), vegetation (Zieman, Fourqurean, and

181

Iverson 1989; Hall et al. 1999), and fish distribution (Sogard, Powell, and Holmquist

1989; Thayer and Chester 1989; Thayer, Powell, and Hoss 1999) patterns: Atlantic,

Central, Eastern, Flamingo, Gulf, Western (Fig. 5.1). To document potential impacts of this environmental variability, I sampled habitat and fish community variation within the patchy landscape of Florida Bay in conjunction with distribution data collected on the above listed predators. I anticipated finding different environmental signals to be relevant to the distribution of the different predators. This was not the result of my research. Rather, the distribution of all predators (but sharks) was correlated with depth, and subsequently, habitat type. This surprising result lead me to investigate the bedrock topography of Florida Bay, expanding my dissertation to incorporate the role of geology in the distribution and foraging ecology of top predators in Florida Bay.

Despite Florida Bay’s minimal bathymetric relief (maximum range of only 4.0 m), small variations in depth alter benthic substrate composition (Hall et al. 1999; Landry

2005; Shinn, Lidz, and Harris 1994) with subsequent impacts on fish (Thayer, Powell, and Hoss 1999) and invertebrate (Shaw 1989) prey communities. This strong influence by subtle topographic relief is akin to the influence of the underlying topography in the terrestrial Florida Everglades where elevation is the foundation to the form and function of this unique ecosystem. Although the north/south absolute elevation grade through

182

the Everglades is only 2 cm / km, relative elevation is the primary determinant of plant communities (Gunderson 1997) by its influence on hydroperiods, fire, peat accumulation, freshwater flow patterns and the general structure of basins, terrestrial ridges and plateaus (White 1997). Because small changes in relative elevation provide habitat for specifically adapted vegetation, the distribution of tree islands and hammock peats is closely related to the bedrock topography (Gleason and Stone 1997). Such shifts in habitat structure and floral communities directly affect foraging and refuge habitats for the diversity of animals who live within the Everglades (Loftus and Eklund 1997;

Bancroft et al. 1997). To summarize, in both the terrestrial and marine (Florida Bay) portions of the Florida Everglades, the topographic expression of the bedrock surface has profound effects on the type of vegetation that grows and the corresponding faunal communities.

A defining feature of Florida Bay is the extensive network of carbonate mudbanks which runs throughout the bay, and with particular dominance in the interior region (Central and Flamingo zones). These mudbanks are typically < 1.0 m deep and covered by dense seagrass beds which provide refuge for important prey species such as mojarra (Family Gerreidae), pinfish (Lagodon rhomboides), gulf toadfish

(Opsanus beta) and mullet (Family Mugilidae). The median catch per unit effort (CPUE)

183

from trawls conducted in mudbank habitats was 0.113 fish per meter, second only to

CPUE from trawls in dense seagrass habitats (0.198 fish per meter). A Kruskal‐Wallis one way analysis of variance (ANOVA) determined a significant difference in CPUE between bottom types (H8 = 146.202, P < 0.001). Post‐hoc tests using the Dunn’s method demonstrated that CPUE from mudbanks was significantly different than catch rates in all other habitat types but dense seagrass.

The impact of mudbanks on the ecology of the interior region of Florida Bay is compounded by the fact that their shallow nature restricts water circulation and promotes evaporation, creating hyper‐saline conditions in this region of the Bay

(Wanless and Tagett 1989). The development and location of these mudbanks is a function of the Miami limestone bedrock that lies beneath Florida Bay and the physical processes of sedimentation (Wanless and Tagett 1989). Wanless and Tagett (1989) explain that coastal and freshwater peat and shore levee deposits, positioned by the limestone surface, were inundated and dissected by sea level rise to form the nuclei of

Florida Bay’s mudbanks. These authors divide the mudbanks of Florida Bay into four regions based on sediment deposition rates that cause mudbank migration, construction or destruction. These sedimentation rates are coincident with the bedrock depth gradient in Florida Bay which runs from northeast (< 1.5 m) to southwest (> 4.0 m). The

184

formation of Florida Bay has been dually influenced by sea level rise and bedrock elevation, with different parts of the Bay becoming inundated by sea water at different stages as a function of elevation (Davis 1980). The southwest region was inundated by sea level rise earliest (5500‐5000 YBP), is now the deepest area of Florida Bay, and has minimal sedimentation rates due to exposure to the Gulf of Mexico resulting in few mudbanks. Higher areas in interior region of Florida Bay were overstepped by sea level rise 5000‐4000 YBP, are very shallow, and have ample sediment supply to allow mudbank construction or migration. Finally, the northeastern region of Florida Bay, with the most elevated bedrock, was over taken by sea level rise 2500 YBP and is currently sediment starved due to restricted water circulation caused by extensive mudbanks to the southwest.

My research documented a bottom‐up transfer of effects due to Florida Bay’s variation in bathymetric structure from benthic substrate composition, through fish communities and finally to the distribution and behavior of top predators. ANOVA tests on data from systematic trackline surveys with synoptic environmental sampling revealed that cormorants (P = 0.003), osprey (P < 0.001), pelicans (P < 0.001), terns (P <

0.001), and foraging dolphins (P = 0.011) all preferentially selected mudbank habitat over other available habitat types (Fig. 5.3). While there is some diet overlap between these

185

predators, they do not all consume the same prey items. Rather, the overlying seagrass beds of these mudbanks provide refuge for prey of many difference species and sizes, and the shallow water column concurrently provides predators with a condensed habitat to more easily encounter and capture prey. Due to the concentration of prey in mudbank habitats, habitat use patterns of these top predators overlap and may influence interspecific competition. Therefore, if prey are limiting, these predators must partition resources through other non‐spatial mechanisms such as prey size, foraging tactic, or prey item. For instance, size partitioning is seen by cormorants who target smaller pinfish and mojarra than dolphins. Dolphins, osprey and pelicans partition mullet, an important prey item of all three predators, based on foraging tactic: dolphins capture mullet by mud ring feeding (see below), osprey dive feet first into the water column, and pelicans plunge head first through the water column. Furthermore, competition is completely avoided between dolphins and terns who target different prey items.

186

Figure 5.3: Sighting rates of each predator group by bottom type. Speckled = Cormorant, Black = Dolphin ‐ Foraging, White = Dolphin ‐ Non‐Foraging, Diagonal stripes = Osprey, Vertical stripes = Pelican, Horizontal stripes = Tern.

Depth and habitat type further influence competition patterns between predators in Florida Bay by promoting the co‐existence of multiple foraging tactics within one population of bottlenose dolphins. Behavioral observations of dolphins throughout

Florida Bay determined that dolphins employ different foraging tactics primarily as a function of depth, and the distribution of these tactics was highly spatially structured

(Fig. 5.4). Dolphins that occupied the mudbank rich Central and Flamingo zones of

Florida Bay were frequently observed performing a foraging tactic called ‘mud ring

187

feeding’ (see Chapter 3). This behavior is unique to Florida Bay; it is a cooperative foraging tactic to corral mullet and force them to jump out of the water where the dolphins wait with open mouths to catch the mullet out of the air (Fig. 5.5). This behavior was only witnessed in habitats less then 1.5 m and within close proximity to a mudbank (typically within 145 m) (Fig. 5.6). Mud ring feeding was never observed outside the Central and Flamingo zones. Foraging dolphins observed beyond of this region performed a ‘herd & chase’ foraging tactic in habitats that were also shallow

(typically less then 2 m), but further from mudbanks. In this behavior a dolphin used a barrier, such as a seagrass bed or mangrove tree roots, to assist in prey capture. Finally, those dolphins observed in the Gulf and Atlantic zones frequently employed a foraging tactic called ‘deep diving with erratic surfacings’. During this foraging tactic, dolphins performed fluke‐out dives (tail flukes come out of the water to allow a more vertical dive profile) with very quick breathing intervals. The dolphins surfaced in an irregular pattern but stayed in one general area and showed no consistent direction of travel.

These deep diving events were only observed in water deeper then 1.5 m and on average over 145 m from a mudbank.

188

Figure 5.4: Spatial distribution of 84 dolphin foraging events observed in Florida Bay. Triangles denote deep diving events. Stars represent herd & chase events. Circles are locations of mud ring feeding events.

189

Figure 5.5: Photograph of dolphins mud ring feeding in Florida Bay. See text for description of foraging behavior.

190

Figure 5.6: Box plots depicting the (a) distance from shore, (b) depth, and (c) group size range of each foraging tactic group. Boxes range from the 25th to 75th percentiles. Black dots represent 5th and 95th percentile outliers. Solid line is the median value and the dotted line is the mean value. 191

Analyses of the sighting history and foraging strategies used by 437 individually identified dolphins observed in Florida Bay demonstrated high individual site and foraging tactic fidelity. Although dolphins were encountered in all areas of the Bay, individual dolphins did not roam throughout the entire Bay but rather displayed high zone fidelity (Fig. 5.7). Dolphins remained in areas with habitat and prey characteristics which promoted the success of their preferred foraging tactic. This site and foraging tactic fidelity demonstrate how the bathymetry of Florida Bay influences dolphin behavioral ecology by limiting their distribution patterns; the mudbanks and shallow depths of the interior region of Florida Bay likely prevent dolphins unfamiliar with the habitat, environment and prey from easily navigating across the Florida Bay landscape.

The shallow habitats of the Central and Flamingo zones provide dolphins equipped with navigational knowledge and an effective foraging strategy the opportunity to exploit an underutilized resource (mullet). Thus, mudbanks in Florida Bay help to decrease competition among predators.

192

Figure 5.7: Zone and foraging tactic fidelity of 61 dolphins observed 5 times or more. a) The percent of sightings individual dolphins were observed in by zone. Numbers above each column denote the number of dolphins, and in parentheses are the number of sightings those dolphins were observed at. Grey = Atlantic; White = Central; Speckled = Eastern; Vertical stripes = Flamingo; Black = Gulf; Hatched = Western. b) The percent of foraging sightings broken down by foraging tactic used. Numbers above each column denote the total number of foraging sightings observed. Grey = Deep diving; Black = Mud ring feeding; White = Herd & Chase; Diagonal stripes = Other or unknown foraging tactic. 193

Discrete dolphin groups may exist in Florida Bay because of variation in their behavioral ecology, which is related to habitat selection patterns and foraging tactic specializations. Therefore, the effects of depth may extend to the social systems of these top predators. This hypothesis is founded on three lines of reasoning. First, social bonds between individual dolphins will vary based on the frequency of interaction between animals as a function of each dolphin’s habitat use and site fidelity patterns. Secondly, I documented a difference in group size at foraging tactic events (Fig. 5.6c). The mean group size at deep diving foraging events (X= 11, n=29) was significantly larger than at mud ring feeding events (X=4, n=42; F2,13 = 4.966, P = 0.0009), and this contrast can cause differences in the social dynamics of these dolphin groups. Group size variation in marine mammals can be a function of predation risk (protection in numbers) and/or foraging behavior (to maximize the net energy intake for each predator) (Packer and

Ruttan 1988; Baird and Dill 1996; Heithaus and Dill 2002). Because deep diving with erratic surfacings is an individualistic foraging tactic, larger group size at these events was likely a function of increased predation risk in the Gulf and Atlantic zones where this foraging tactic was performed. Shark sampling revealed that large sharks are distributed in these two zones, including the bull shark (Carcharhinus leucas), a known predator of bottlenose dolphins (Heithaus 2001). Few potentially predatory sharks to dolphins were found elsewhere in the Bay (Table 5.1) (Wiley and Simpfendorfer 2007;

194

Torres, Heithaus, and Delius 2006). In contrast, mud ring feeding is a coordinated group foraging strategy where prey are shared among individuals. With minimal predation effects on dolphins in the Central and Flamingo zones, the smaller group size at mud ring feeding events likely reflects an optimal group size to maximize the energetic reward for participating dolphins.

Table 5.1: Number of individual shark species caught in Florida Bay, broken down by zone. A = Atlantic, C = Central, E = Eastern, F = Flamingo, G = Gulf, T = Total

Lastly, the effects of depth variation may transfer to the social systems of dolphins in Florida Bay because mud ring feeding is a cooperative behavior, which requires communication and organization among individuals. Furthermore, it is likely that dolphins forage more frequently with the same individual(s) to improve their foraging efficiency. This repeated coordination among conspecifics may encourage tighter social bonds among individuals who mud ring feed in comparison to dolphins

195

who use a more individualistic foraging strategy. My research did not determine whether or not dolphin groups in Florida Bay are demographically or reproductively isolated, nor the rates of association between animals. However, based on the previous three points, it may be possible that dolphin groups in Florida Bay, delineated by foraging tactic and ranging patterns, are diverging into separate populations, similar to the divergence documented between resident and transient killer whale (Orca orcinus) populations in the northeast Pacific (Baird and Dill 1995) in association with their different foraging tactics and prey items. In summation, my research demonstrates how bathymetric variation in Florida Bay, as a function of slight, but influential, topographic relief, has strong ecological effects on the foraging, distribution, and possibly social and evolutionary patterns of this bottlenose dolphin population.

Predators and prey interact at various temporal and spatial scales and the behavioral state of a predator denotes important information about the interaction.

Observations of foraging behavior can facilitate fine‐scale links between predator, prey and habitat. Otherwise the occurrence of a predator does not necessarily infer that the availability of prey at that location is relevant. For instance, in my research on the distribution of sharks in Florida Bay as a function of habitat and prey variation, I found that only prey abundance on a regional scale (zone average) was correlated with

196

predator abundance; shark catch rates were significantly higher in zones where teleost catch per unit effort from trawls was also higher (F1,42=4.4, p=0.04). Sharks have relatively low feeding rates (Weatherbee and Cortes 2004), and therefore, may not concentrate their movements in microhabitats of high prey abundance unless they are foraging.

Teleost catch and shark catch at the same place and time in Florida Bay were not correlated, which suggests that sharks do not regularly occupy habitats with high prey availability until they are ready to feed. Thus, without behavioral data and fine‐scale knowledge of shark foraging locations, the functional use of specific habitats and their prey availability could not be assessed.

In contrast to this study on shark distribution, I was able to observe, record and incorporate behavior data into habitat modeling efforts of bottlenose dolphins throughout Florida Bay. The predictive capacity of four models, with and without prey distribution data or environmental quality data, were compared to determine the optimal set of explanatory variables to be used in predictive models of fine‐scale dolphin habitat selection. Due to extrapolation limitations associated with the temporal and spatial scale of sampling, models which utilized prey distribution data did not predict dolphin distribution well. However, models which predicted areas of high fish catch based on environmental characteristics correctly predicted over 80% of dolphin foraging

197

locations. This strong predictive capacity highlights the tight spatial links between a foraging predator, its prey and the environmental variables which influence the distribution of prey items.

Marine predators, such as dolphins, seabirds and sharks, select different habitat characteristics when foraging as compared to socializing or traveling when marginal habitat may suffice. The previous two examples of the factors influencing the distribution of dolphins and sharks in Florida Bay demonstrate how fine‐scale studies of species biogeography that are coupled with behavioral data can reveal considerable insight about the biological mechanisms underlying associations between species distribution, space and habitat and prey availability. Predators forage in areas where critical resources exist. Therefore, behavioral data should be incorporated into distribution studies, especially those with management applications, to reveal the functional importance of habitat selection patterns (Hastie et al. 2004; Heithaus and Dill

2002). In marine ecosystems, where the assessment of habitat quality is difficult, managers can rely on the distribution of an indicator or sentinel species to monitor relative habitat quality. Seabirds and cetaceans are ideal organisms to employ as indicators of ecosystem health because they are numerous and conspicuous predators, with large energetic requirements (Caro and OʹDoherty 1999; DʹAmico et al. 2003; Moore

198

and DeMaster 1998). Locations of observed foraging behavior can quickly be identified as productive fish communities, providing managers a useful tool to spatially and temporally quantify habitat quality and, through long‐term observational data collection, monitor spatial migration of productive habitats. In Florida Bay, such an approach could help managers mitigate anthropogenic effects on the ecosystem by identifying and protecting important habitat. The suitability of the bottlenose dolphin as an indicator species in Florida Bay was assessed by Torres and Urban (2005). Modeling results revealed high spatial overlap between dolphin foraging habitat and abundant fish populations, leading the authors to conclude that mangers can use foraging dolphins as an indictor of healthy fish habitats in Florida Bay.

In contrast to a ‘trophic cascade’ (Estes and Palmisano 1974) which occurs when top‐down effects permeate through multiple lower trophic levels (e.g.: impacts of sea otter predators on sea urchin densities on kelp forest health), the results of my research in Florida Bay describe a bottom‐up transfer of effects from the variation in bathymetric structure to top predator distribution and foraging ecology. My dissertation uncovered a

‘trophic fountain’ (Fig. 5.8) where subtle irregularities in the limestone bedrock beneath

Florida Bay influences depth variation, which is the primary determinant of mudbank formation and location (Wanless and Tagett 1989; Davis 1980; Gleason and Stone 1997).

199

Mudbanks provide appropriate conditions to be densely covered by seagrass beds due to their rich organic sediment and shallow water column that allows high light penetration (Zieman, Fourqurean, and Iverson 1989). Prey concentrate within these seagrass beds because of high refuge availability and nutrient availability (Holmquist,

Powell, and Sogard 1989; Sogard, Powell, and Holmquist 1989), which then attracts a variety of predators with a variety of foraging tactics. The spatially structured heterogeneous habitats of Florida Bay influence prey distribution, which indirectly determines the distribution patterns of dolphins, cormorants, osprey, pelicans, and terns, as well as the foraging tactics employed by the dolphin community. This ‘trophic fountain’ highlights the importance of mudbank habitat for predators and prey and raises questions about the affects of expected sea level rise on the ecology of top predator populations in Florida Bay.

200

Figure 5.8: Illustration of a ‘trophic fountain’ with bottom‐up affects of the bedrock geology in Florida Bay on the distribution of mudbank and seagrass habitats, prey availability and predator distribution. Foraging dolphins, cormorants, osprey, pelicans and terns all concentrate their habitat selection patterns in mudbank habitats where relatively high prey availability exists. Black = bedrock; gray = mudbank; striped = other bottom type; dark gray = terrestrial substrate.

Sea level is rising at a rate of 30‐40 cm / 100 years. This rate is significantly higher than estimates of sea level rise during pre‐modern Florida and is predicted to increase further due to melting of polar ice sheets as an effect of global warming (Gleason and

Stone 1997; Wanless, Parkinson, and Tedesco 1997). This rate of sea level rise is anticipated to trigger dramatic changes in coastal wetland communities of South Florida.

The southernmost terrestrial Everglades will be truncated as sea level overtops

201

freshwater marshes and promotes salt water and mangrove habitats (Gleason and Stone

1997). Shore margins will experience accelerated erosion rates and landward encroachment of marine wetlands (Wanless, Parkinson, and Tedesco 1997). The inverse perspective of this scenario is the expected northward expansion of Florida Bay. In fact, the northeastern coastline of Florida Bay is predicted to respond first and most dramatically to sea level rise due to this area’s demonstrated instability over the past

3000 years (Wanless, Parkinson, and Tedesco 1997). With sea water inundation, depths will increase across the landscape and cause the current habitat structure of Florida Bay to continue to evolve as a function of the interdependence between elevation and sea level. With increased water depth, low lying areas will become submerged, water circulation patterns will change and associated sedimentation rates will modify, causing

Florida Bay’s extensive network of mudbanks to construct, deconstruct and migrate into new areas.

Consequences of changes in Florida Bay’s benthic structure, including the spatial migration of old and formation of new mudbanks, will “trickle‐up” to the distribution of top predators. Wanless et al. (1997) predict that as habitats within coastal bays on South

Florida become deeper and mudbanks erode, waters will be less restricted and more turbid, thus fostering significant changes in benthic marine communities. Such changes

202

may included alterations in seagrass composition and distribution, with subsequent impacts on the spatial structure of fish communities. Results generated from my dissertation research indicate a strong habitat selection preference by dolphins, cormorants, osprey, pelicans and terns for shallow mudbank habitats, with minimal influence by the geographic location of the mudbanks. Therefore, I do not expect the foraging ecology of these predators to alter dramatically with sea level rise, but I do predict that the predators will spatially track the migration of mudbanks in Florida Bay.

In order for these top predators to maintain their behavioral association with the high prey availability and foraging success fostered by mudbank habitats, the predators will be forced to modify their distribution patterns accordingly. The exception may be the prevalence and distribution of dolphin foraging tactics. With increased water depth across Florida Bay, more habitat may become available for dolphins to employ the deep diving foraging tactic while mud ring feeding events may be more commonly observed in newly formed mudbank habitats in the northeastern region. In fact, on two occasions during my field work I was surprised to observe dolphins mud ring feeding in the most northeastern portion of Florida Bay right along the terrestrial margin, far away from all other observations of mud ring feeding (Fig. 5.4). Is it possible that habitat changes are already occurring, allowing mud ring feeding dolphins to exploit the expanding, shallow, muddy region of Florida Bay? As habitat alterations manifest, the interplay of

203

geology and sea level rise in Florida Bay will continue to modify all levels of the ecosystem, from the distribution and composition of benthic habitats and prey communities to the foraging and distribution ecology of top predators.

204

References

Allen, Mark C., Andrew J. Read, Jocelyn Gaudet, and Laela S. Sayigh. 2001. Fine‐scale habitat selection of foraging bottlenose dolphins Tursiops truncatus near Clearwater, Florida. Marine Ecology Progress Series 222:253‐264.

Allouche, Omri, Asaf Tsoar, and Ronen Kadmon. 2006. Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS). Journal of Applied Ecology 43 (6):1223‐1232.

Baird, Robin W., Peter A. Abrams, and Lawrence M. Dill. 1992. Possible indirect interactions between transient and resident killer whales: implications for the evolution of foraging specializations in the genus Orcinus. Oecologia 89:125‐132.

Baird, Robin W., and Lawrence M. Dill. 1995. Occurrence and behaviour of transient killer whales: seasonal and pod‐specific variability, foraging behaviour, and prey handling. Canadian Journal of Zoology 73:1300‐1311.

Baird, Robin W., and Lawrence M. Dill. 1996. Ecological and social determinants of group size in transient killer whales. Behavioral Ecology 7 (4):408‐416.

Baltz, D.M., and G. Victor Morejohn. 1977. Food habits and niche overlap of seabirds wintering on Monterey Bay, California. Auk 94:526‐543.

Bancroft, G. Thomas, Allan M. Strong, Richard J. Sawicki, Wayne Hoffman, and Susan D. Jewell. 1997. Relationship among wading bird foraging patterns, colony locations, and hydrology in the Everglades. In Everglades: The ecosystem and its restoration, edited by S. M. Davis and J. C. Ogden. Boca Raton: St. Lucie Press.

Barot, Sebastien, Jacques Gignoux, and Jean‐Claude Menaut. 1999. Demography of a savanna palm tree: Predictions from comprehensive spatial pattern analyses. Ecology 80 (6):1987‐2005.

Barros, Nelio B. 1987. Food habits of bottlenose dolphins (Tursiops truncatus) in the southeastern United States, with special reference to Florida waters. Master of Science, Marine Science, University of Miami, Coral Gables, FL.

205

Barros, Nelio B. 1993. Feeding ecology and foraging strategies of bottlenose dolphins on the central east coast of Florida. Doctor of Philosophy, Marine Science, University of Miami, Coral Gables, FL.

Barros, Nelio B., and Dan K. Odell. 1990. Food habits of bottlenose dolphins in the southeastern United States. In The bottlenose dolphin, edited by S. Leatherwood and R. R. Reeves. San Diego, California: Academic Press.

Barros, Nelio B., and Randall S. Wells. 1998. Prey and Feeding Patterns of Resident Bottlenose Dolphins (Tursiops truncatus) in Sarasota Bay, Florida. Journal of Mammalogy 79 (3):1045‐1059.

Baumgartner, Mark F., Keith Mullin, L. Nelson May, and T. D. Leming. 2000. Cetacean habitats in the northern Gulf of Mexico. Fisheries Bulletin 99 (2):219‐239.

Beals, E. W. 1984. Bray‐Curtis ordination: an effective strategy for analysis of multivariate ecological data. Advances in Ecological Research 14:1‐55.

Benoit‐Bird, Kelly J. , and Whitlow W. L. Au. 2003. Prey dynamics affect foraging by a pelagic predator (Stenella longirostris) over a range of spatial and temporal scales. Behavioral Ecology and Sociobiology 53 (6):364‐373.

Bent, A. C. 1921. Life histories of North American gulls and terns. Vol. Bulletin 113. Washington, D.C.: Smithsonian Institution.

Hawthʹs Analysis Tools for ArcGIS 3.23.

Boyer, Joseph N., James W. Fourqurean, and Ronald D. Jones. 1999. Seasonal and long‐ term trends in the water quality of Florida Bay (1989‐1997). Estuaries 22 (2B):417‐ 430.

Breiman, L., J. Freidman, R. Olshen, and C. Stone. 1984. Classification and regression trees. Belmont, California: Wadsworth.

Brewster‐Wingard, G. Lynn, and Scott E. Ishman. 1999. Historical trends in salinity and substrate in central Florida Bay: A paleoecological reconstruction using modern analogue data. Estuaries 22 (2B):369‐383.

206

Brotons, L., Wilfried Thuiller, M.B. Araujo, and Alexandre H. Hirzel. 2004. Presence‐ absence versus presence‐only modelling methods for predicting bird habitat suitability. Ecography 27:437‐448.

Buckland, S. T., D.R. Anderson, K.P. Burnham, J.L. Laake, D.L Borchers, and L. Thomas. 2001. Introduction to Distance Sampling: Estimating abundance of biological populations. New York: Oxford University Press Inc.

Burnham, K.P., and D.R. Anderson. 1998. Model selection and inference: a practical information‐theoretic approach. New York: Springer‐Verlag.

Butler, IV, M.J., J.H. Hunt, W.F. Herrnkind, M.J. Childress, R. Bertelsen, W. Sharp, T. Matthews, J.M. Field, and G. Marshall. 1995. Cascading disturbances in Florida Bay, USA: Cyanobacteria blooms, sponge mortality, and implications for juvenile spiny lobster Panulirus argus. Marine Ecology Progress Series 129:119‐125.

Caro, T. M., and Gillian OʹDoherty. 1999. On the use of surrogate species in conservation biology. Conservation Biology 13 (4):805‐814.

Chilvers, B. Louise, Peter J. Corkeron, and Marji L. Puotinen. 2003. Influence of trawling on the behaviour and spatial distribution of Indo‐Pacific bottlenose dolphins (Tursiops aduncus) in Moreton Bay, Australia. Canadian Journal of Zoology 81 (12):1947‐1955.

Cohen, J. 1960. A coefficient of agreement for nominal scales. Educ. Psychol. Meas. 20:37‐ 46.

Connell, Joseph H. 1961. The influence of interspecific competition and other factors on the distribution of the barnacle Chthamalus stellatus. Ecology 42 (4).

Connor, Richard C., Michael R. Heithaus, Per Berggren, and Jennifer L. Miksis. 2000. ʺKerplunkingʺ: Surface fluke‐splashes during shallow‐water bottom foraging by bottlenose dolphins. Marine Mammal Science 16 (3):646‐653.

Connor, Richard C., Andrew J. Read, and Richard Wrangham. 2000. Male reproductive strategies and social bonds. In Cetacean Societies: Field studies of dolphins and whales, edited by J. Mann, R. C. Connor, P. L. Tyack and H. Whitehead. Chicago and London: The University of Chicago Press.

207

Connor, Richard C., Randall S. Wells, Janet Mann, and Andrew J. Read. 2000. The bottlenose dolphin: Social relationships in a fission‐fusion society. In Cetacean Societies: Field studies of dolphins and whales, edited by J. Mann, R. C. Connor, P. L. Tyack and H. Whitehead. Chicago and London: University of Chicago Press.

Corkeron, Peter J., M.M. Bryden, and K.E. Hedstrom. 1990. Feeding by bottlenose dolphins in association with trawling operations in Moreton Bay, Australia. In The Bottlenose Dolphin, edited by S. Leatherwood and R. R. Reeves. San Diego: Academic Press.

Cox, TM, TJ Ragen, AJ Read, E Vos, RW Baird, K Balcomb, J Barlow, J Caldwell, T Cranford, L Crum, A DʹAmico, G DʹSpain, A Fernandez, J Finneran, R Gentry, W Gerth, F Gulland, J Hildebrand, D Houser, T Hullar, PD Jepson, D Ketten, CD MacLeod, P Miller, S Moore, DC Mountain, D Palka, S Rommel, T Rowles, B Taylor, P Tyack, D Wartzok, R Gisiner, J Mead, and L Benner. 2006. Understanding the impacts of anthropogenic sound on beaked whales. Journal of cetacean research and management 7 (3):177‐187.

DʹAmico, Angela, A. Bergamasco, P. Zanasca, S. Carniel, E. Nacini, N. Portunato, V. Teloni, C. Mori, and R. Barbanti. 2003. Qualitative correlation of marine mammals with physical and biological parameters in the Ligurian Sea. IEEE Journal of Oceanic Engineering 28 (1):29‐43.

Davis, R., G. S. Fargion, N. May, T. D. Leming, M. Brumgartner, W. E. Evans, Larry J. Hansen, and Keith Mullin. 1998. Physical Habitat of cetaceans along the continental slope in the north‐central and western Gulf of Mexico. Marine Mammal Science 14 (3):490‐507.

Davis, S.M., Lance H. Gunderson, Winifred A. Park, John R. Richardson, and Jennifer E. Mattson. 1997. Landscape dimension, composition, and function in a changing Everglades ecosystem. In Everglades: The ecosystem and its restoration, edited by S. M. Davis and J. C. Ogden. Boca Raton: St. Lucie Press.

Davis, T.D. 1980. Peat formation in Florida Bay and its significance in interpreting the recent vegetational and geological history of the Bay area. PhD dissertation, The Pennsylvania State University, University Park.

208

Durako, Michael, Margaret O. Hall, and Manuel Merello. 2002. Patterns of change in the seagrass dominated Florida Bay hydroscape. In The Everglades, Florida Bay, and coral reefs of the Florida keys: An ecosystem sourcebook, edited by J. W. Porter and K. G. Porter. Boca Raton: CRC Press.

Emlen, J. Merritt. 1966. The role of time and energy in food preference. The American Naturalist 100 (916):611‐617.

Engleby, L., and D. Waples. 2001. First observations of mud‐ring feeding by bottlenose dolphins (Tursiops truncatus) in Florida Bay. in Fourteenth Biennial Conference on the Biology of Marine Mammals, Vancouver, Canada.

Engler, Robin, Antoine Guisan, and Luca Rechsteiner. 2004. An improved approach for predicting the distribution of rare and endangered species from occurrence and pseudo‐absence data. Journal of Applied Ecology 41 (2):263‐274.

Estes, James. A., M. L. Riedman, M. M. Staedler, M. T. Tinker, and B. E. Lyon. 2003. Individual variation in prey selection by sea otters: patterns, causes and implications. Journal of Animal Ecology 72 (1):144‐155.

Estes, James A., and John F. Palmisano. 1974. Sea otters: Their role in structuring nearshore communities. Science 185:1058‐1060.

Fauchald, P.K., E. Erikstad, and H. Skarsfjord. 2000. Scale‐dependent predator‐prey interactions: the hierarchical spatial distribution of seabirds and prey. Ecology 81:773‐783.

Ferguson, Megan C., Jay Barlow, Paul Fiedler, Stephen B. Reilly, and Tim Gerrodette. 2006. Spatial models of delphinid (family Delphinidae) encounter rate and group size in the eastern tropical Pacific Ocean. Ecological Modelling 193 (3‐4):645‐662.

Ferguson, Megan C., Jay Barlow, Stephen B. Reilly, and Tim Gerrodette. 2006. Predictive Cuvierʹs (Ziphius cavirostris) and Mesoplodon beaked whale densities as functions of the environment in the eastern tropical Pacific Ocean. Journal of Cetacean Research Management 7 (3):287‐299.

209

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

Forney, Karin A. 1999. Trends in harbour porpoise abundance off central California, 1986‐1995: Evidence for interannual changes in distributions? Journal of Cetacean Research Management 1:73‐80.

Forney, Karin A. 2000. Environmental models of cetacean abundance: reducing uncertainty in populations trends. Conservation Biology 14 (5):1271‐1286.

Fourqurean, James W., and Michael B. Robblee. 1999. Florida Bay: A history of recent ecological changes. Estuaries 22 (2B):345‐357.

Fretwell, S. D., and H. L. Lucas, Jr. 1970. On territorial behavior and other factors influencing habitat distribution in birds. Acta Biotheoretica 19 (1):16‐36.

FWC‐FWRI. 2007. Fisheries‐Independent Monitoring Program Procedure Manual. St. Petersburg, Florida: Florida Fish and Wildlife Research Institute.

Gaertner, J.C. 2000. Seasonal organization patterns of demersal assemblages in the Gulf of Lions (north‐western Mediterranean Sea). Journal of Marine Biological Association of the United Kingdom 80 (5):777‐783.

Gazda, Stefanie K., Richard C. Connor, Robert K. Edgar, and Frank Cox. 2005. A division of labour with role specialization in group‐hunting bottlenose dolphins (Tursiops truncatus) off Cedar Key, Florida. Proceedings of the Royal Society of London 272:135‐140.

Gleason, Patrick J., and Peter Stone. 1997. Age, origin, and landscape evolution of the Everglades peatland. In Everglades: The ecosystem and its restoration, edited by S. M. Davis and J. C. Ogden. Boca Raton, Florida: St. Lucie Press.

Gregr, Edward J., and Andrew W. Trites. 2001. Predictions of critical habitat for five whale species in the waters of coastal British Columbia. Canadian Journal of Fish and Aquatic Science 58:1265‐1285.

210

Grunwald, Michael. 2006. The Swamp: The Everglades, Florida, and the politics of paradise. New York: Simon & Schuster.

Guinet, Christophe. 1991. Intentional stranding apprenticeship and social play in killer whales (Orcinus orca). Canadian Journal of Zoology 69:2712‐2716.

Guinet, Christophe, Laurent Dubroca, Mary‐Anne Lea, Simon Goldsworthy, Yves Cherel, Guy Duhamel, Francesco Bonadonna, and Jean‐Paul Donnay. 2001. Spatial distribution of foraging in female Antarctic fur seals Arctocephalus gazella in relation to oceanographic variables: a scale‐dependent approach using geographic information systems. Marine Ecology Progress Series 219:251‐264.

Guisan, Antoine, Thomas C. Edwards, and Trevor Hastie. 2002. Generalized linear and generalized additive models in studies of species distributions: setting the scene. Ecological Modelling 157 (2‐3):89‐100.

Guisan, Antoine, Wilfried Thuiller, and Nicholas Gotelli. 2005. Predicting species distribution: offering more than simple habitat models. Ecology Letters 8 (9):993‐ 1009.

Guisan, Antoine, and Niklaus E. Zimmermann. 2000. Predictive habitat distribution models in ecology. Ecological Modelling 135:147‐186.

Gunderson, Lance H. 1997. Vegetation of the Everglades: Determinants of community composition. In Everglades: The ecosystem and its restoration, edited by S. M. Davis and J. C. Ogden. Boca Raton: St. Lucie Press.

Gunter, Gordon. 1958. Feeding behavior of brown and white pelicans of the Gulf coast of the United States. Louisiana Academy of Sciences 21:34‐39.

Hall, Margaret O., Michael Durako, James W. Fourqurean, and Joseph C. Zieman. 1999. Decadal changes in seagrass distribution and abundance in Florida Bay. Estuaries 22 (2B):445‐459.

Hamazaki, Toshihide. 2002. Spatiotemporal prediction models of cetacean habitats in the mid‐western north Atlantic ocean (from Cape Hatteras, North Carolina, U.S.A. to Nova Scotia, Canada). Marine Mammal Science 18 (4):920‐939.

211

Hastie, Gordon D., Ben Wilson, L. J. Wilson, K. M. Parsons, and P. M. Thompson. 2004. Functional mechanisms underlying cetacean distribution patterns: hotspots for bottlenose dolphins are linked to foraging. Marine Biology 144:397‐403.

Hastie, T.J., and R.J. Tibshirani. 1990. Generalized additive models. Boca Raton, FL: Chapman & Hall/CRC.

Heck, Kenneth L., Jr., and Robert J. Orth. 1980. Seagrass habitats: The roles of habitat complexity, competition and predation in structuring associated fish and motile macroinvertebrate assemblages. Estuarine Perpectives:449‐464.

Hedley, S.L., S. T. Buckland, and D.L Borchers. 1999. Spatial modelling from line transect data. Journal of Cetacean Research Management 1:255‐264.

Heithaus, M.R. 2001. The biology of tiger sharks, Galeocerdo cuvier, in Shark Bay, Western Australia: sex ratio, size distribution, diet, and seasonal changes in catch rates. Environmental Biology of Fishes 61:25‐36.

Heithaus, Michael R. 2001. Predator‐prey and competitive interactions between sharks (order Selachii) and dolphins (suborder Odontoceti): a review. Journal of Zoology 253:53‐68.

Heithaus, Michael R. 2004. Predator‐prey interactions. In Biology of sharks and their relatives, edited by J. C. Carrier, J. A. Musick and M. R. Heithaus. Boca Raton: CRC Press.

Heithaus, Michael R. 2005. Habitat use and group size of pied cormorants (Phalacrocorax varius) in a seagrass ecosystem: possible effects of food abundance and predation risk. Marine Biology 147:27‐35.

Heithaus, Michael R., and Lawrence M. Dill. 2002. Feeding strategies and tactics. In Encyclopedia of Marine Mammals, edited by W. F. Perrin, B. Wursig and J. G. M. Thewissen. San Deigo: Academic Press.

Heithaus, Michael R., and Lawrence M. Dill. 2002. Food availability and tiger shark predation risk influence bottlenose dolphin habitat use. Ecology 83 (2):480‐491.

212

Heithaus, Michael R., and Lawrence M. Dill. 2006. Does tiger shark predation risk influence foraging habitat use by bottlenose dolphins at multiple spatial scales? Oikos 114 (2):257‐264.

Heithaus, Michael R., Lawrence M. Dill, G. J. Marshall, and B. Buhleier. 2002. Habitat use and foraging behavior of tiger sharks (Galeocerdo cuvier) in a seagrass ecosystem. Marine Biology 140:237‐248.

Heupel, M.R., and R.E. Hueter. 2002. Importance of prey density in relation to the movement patterns of juvenile blacktip sharks (Carcharhinus limbatus) within a coastal nursery area. Marine and Freshwater Research 53:543‐550.

Hoelzel, A. Rus, Eleanor M. Dorsey, and S. Jonathan Stern. 1989. The foraging specializations of individual minke whales. Animal Behaviour 38 (5):786‐794.

Hoese, H. D. 1971. Dolphin feeding out of water in a salt marsh. Journal of Mammalogy 52 (1):222‐223.

Hoese, H. D., and Richard H. Moore. 1992. Fishes of the Gulf of Mexico, Texas, Louisiana, and adjacent waters. College Station: Texas A&M University Press.

Holmquist, Jeff G., George V. N. Powell, and Susan M. Sogard. 1989. Decapod and stomatopod assemblages in an unusual system of seagrass‐covered mud banks in Florida Bay. Marine Biology 100 (4):473‐483.

Holmquist, Jeff G., George V. N. Powell, and Susan M. Sogard. 1989. Sediment, water level and water temperature characteristics of Florida Bayʹs grass‐covered mud banks. Bulletin of Marine Science 44 (1):348‐364.

Hooker, Sascha K., Hal Whitehead, and Shannon Gowans. 1999. Marine Protected Area Design and the Spatial and Temporal Distribution of Cetaceans in a Submarine Canyon. Conservation Biology 13 (3):592‐602.

Hugie, Don. M, and Lawrence M. Dill. 1994. Fish and game: a game theoretic approach to habitat selection by predators and prey. Journal of Fish Biology 45:151‐169.

213

Hyrenbach, K. D., Richard R. Veit, H. Weimerskirch, and George L. Jr. Hunt. 2006. Seabird associations with mesoscale eddies: the subtropical Indian Ocean. Marine Ecology Progress Series 324:271‐279.

Hyrenbach, K.D., Richard R. Veit, Henri Weimerskirch, Nicolas Metzl, and George L. Jr. Hunt. 2007. Community structure across a large‐scale ocean productivity gradient: Marine bird assemblages of the Southern Indian Ocean. Deep‐Sea Research I.

Iverson, Louis R., and Anantha M. Prasad. 1998. Predicting abundance for 80 tree species following climate change in the eastern United States. Ecological Monographs 68 (4):465‐485.

Jaquemet, Sebastien, Matthieu Le Corre, Francis Marsac, Michel Potier, and H. Weimerskirch. 2005. Foraging habitats of the seabird community of Europa Island. Marine Biology 147:573‐582.

Johnsgard, Paul A. 1993. Cormorants, darters, and pelicans of the world. Edited by M. Abbate. Washington: Smithsonian Institution Press.

Krebs, J.R. 1978. Optimal foraging: decision rules for predators. In Behavioural Ecology: an Evolutionary Approach, edited by J. R. Krebs and N. B. Davies. Oxford: Blackwell Scientific Publications.

Kruskal, J.B. 1964. Nonmetric multidimensional scaling: a numerical method. Psychometrika 29:115‐129.

Landry, Brooke. 2005. Changes in the distribution and density of Florida Bay macrophytes: 1995‐2004, Department of Marine Science, University of North Carolina Wilmington, Wilmington, North Carolina.

Levin, Simon A. 1992. The problem of pattern and scale in ecology. Ecology 73 (6):1943‐ 1967.

Lewis, Jennifer S., and William W. Schroeder. 2003. Mud plume feeding, a unique foraging behavior of the bottlenose dolphin in the Florida Keys. Gulf of Mexico Science 21 (1):92‐97.

214

Light, S.S., and J.W. Dineen. 1994. Water control in the Everglades: a historical perspective. In Everglades: The Ecosystem and its Restoration, edited by S. M. Davis and J. C. Ogden. Delray Beach, Florida: St. Lucie Press.

Lima, Steven L. 2002. Putting predators back into behavioral predator‐prey interactions. Trends in Ecology & Evolution 17 (2):70‐75.

Loftus, William F., and Anne‐Marie Eklund. 1997. Long‐term dynamics of an Everglades small‐fish assemblage. In Everglades: The ecosystem and its restoration, edited by S. M. Davis and J. C. Ogden. Boca Raton: St. Lucie Press.

Logwerwell, E. A., and N. B. Hargreaves. 1996. The distribution of sea birds relative to their fish prey off Vancouver Island: opposing results at large and small spatial scales. Fisheries Oceanography 5:163‐175.

MacArthur, Robert H. 1958. Population ecology of some warblers of Northeastern coniferous forests. Ecology 39 (4):599‐619.

MacArthur, Robert H. 1972. Geographical Ecology: Patterns in the distribution of species. New York: Harper and Row.

MacKenzie, Darryl I., James D. Nichols, Gideon B. Lachman, Sam Droege, J. Andrew Royal, and Catherine Langtimm. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83 (8):2248‐2255.

Manel, S., H.C. Williams, and S. J. Ormerod. 2001. Evaluating presence‐absence models in ecology: the need to account for prevalence. Journal of Applied Ecology 38:921‐ 931.

Mantel, N. 1967. The detection of disease clustering and a generalized regression approach. Cancer Research 27:209‐220.

Matheson, Richard E. Jr., David K. Camp, Susan M. Sogard, and Kimberly A. Bjorgo. 1999. Changes in seagrass‐associated fish and crustacean communities in Florida Bay mud banks: The effects of recent ecosystem changes? Estuaries 22 (2B):534‐ 551.

215

May, Robert M., and Robert H. MacArthur. 1972. Niche overlap as a function of environmental variability. Proceedings of the National Academy of Sciences of the United States of America 69 (5):1109‐1113.

McCune, Bruce. 1994. Improving community analysis with the Beals smoothing function. Ecoscience 1 (1):82‐86.

PC‐ORD. Multivariate Analysis of Ecological Data 4. MjM Software Design, Gleneden Beach, Oregon, USA.

McIvor, C.C., J.A. Ley, and R.D. Bjork. 1994. Changes in freshwater inflow from the Everglades to Florida Bay including effects on biota and biotic processes: a review. In Everglades: The ecosystem and its restoration, edited by S. M. Davis and J. C. Ogden. Delrey Beach, Florida: St. Lucie Press.

McPherson, B.E., and Robert B. Halley. 1996. The south Florida environment ‐ A region under stress. Reston, Virginia: United States Geological Survey.

McPherson, J.M., W. Jetz, and D.J. Rogers. 2004. The effects of speciesʹ range sizes on the accuracy of distribution models: ecological phenomenon or statistical artefact? Journal of Applied Ecology 41:811‐823.

Mehlum, F., G. L. Hunt Jr., Z. Klusek, and M.B. Decker. 1999. Scale‐dependent correlations between the abundance of Brunnich’s guillemots and their prey. Journal of Animal Ecology 68 (60‐72).

Michaelsen, J., D.S. Schimel, M.A. Friedl, F.W. Davis, and R.C. Dubayah. 1994. Regression tree analysis of satellite and terrain data to guide vegetation sampling and surveys. Journal of Vegetation Science 5:673‐686.

Moore, S. E., and Douglas P. DeMaster. 1998. Cetacean habitats in the Alaskan Arctic. Journal of Northwest Atlantic Fisheries Science 22:55‐69.

Morrissey, J. F., and S.H. Gruber. 1993. Home range of juvenile lemon sharks, Negaprion brevirostris. Copeia 2 (425‐434).

National Research Council. 2002. Florida Bay Research Programs and Their Relation to the Comprehensive Everglades Restoration Plan. Edited by N. R. C. Committee on

216

Restoration of the Greater Everglades Ecosystem. Washington, D.C.: The National Academies Press.

Obeysekera, Jayantha, Joan Browder, Lewis Hornung, and Mark A. Harwell. 1999. The natural South Florida system I: Climate, geology, and hydrology. Urban Ecosystems 3:223‐244.

Ogden, J.C., Joan Browder, John H. Gentile, Lance H. Gunderson, Robert Fennema, and John Wang. 1999. Environmental management scenarios: Ecological implications. Urban Ecosystems 3:279‐303.

Oksanen, J., and P.R. Minchin. 2002. Continuum theory revisited: What shape are species responses along ecological gradients? Ecological Modelling 157:119‐129.

Packer, Craig, and Lore Ruttan. 1988. The Evolution of Cooperative Hunting. The American Naturalist 132 (2):159‐198.

Partridge, L., and P. Green. 1985. Intraspecific feeding specializations and population dynamics. In Behavioral Ecology, edited by R. M. Sibley and R. H. Smith. Oxford, UK: Blackwell Press.

Pearce, Jennie, and Simon Ferrier. 2000. Evaluating the predictive performance of habitat models developed using logistic regression. Ecological Modelling 133 (3):225‐245.

Pianka, Eric R. 1974. Niche overlap and diffuse competition. Proceedings of the National Academy of Sciences of the United States of America 71 (5):2141‐2145.

Piatt, John F., and David A. Methven. 1992. Threshold foraging behavior of baleen whales. Marine Ecology Progress Series 84:205‐210.

Pinson, D., and H. Drummond. 1993. Brown pelican siblicide and the prey‐size hypothesis. Behavioral Ecology and Sociobiology 32:111‐118.

Pocklington, R. 1979. An oceanographic interpretation of seabird distribution in the Indian Ocean. Marine Biology 51:9‐21.

Poole, Alan F. 1989. Ospreys: A natural and unnatural history. Cambridge: Cambridge University Press.

217

Porter, James W., and Karen G. Porter. 2002. The Everglades, Florida Bay, and coral reefs of the Florida Keys: An ecosystem sourcebook. Edited by J. W. Porter and K. G. Porter. Boca Raton: CRC Press.

Powell, George V. N., Jeff G. Holmquist, and Susan M. Sogard. 1989. Physical and environmental characteristics of Florida Bay with emphasis on mud banks. Bulletin of Marine Science 44 (1):522.

Read, Andrew J., Danielle Waples, Kim W. Urian, and Dave Swanner. 2003. Fine‐scale behaviour of bottlenose dolphins around gillnets. Proceedings of the Royal Society Biological Sciences Series B 270 (Biology Letters Supplement 1).

Redfern, J.V., Megan C. Ferguson, E.A. Becker, K.D. Hyrenbach, Caroline Good, Jay Barlow, K. Kaschner, Mark F. Baumgartner, Karin A. Forney, Lisa T. Ballance, P.K. Fauchald, Pat Halpin, Toshihide Hamazaki, A.J. Pershing, S.S. Qian, Andrew J. Read, Stephen B. Reilly, Leigh G. Torres, and Fransico E. Werner. 2006. Techniques for cetacean‐habitat modeling. Marine Ecology Progress Series 310:271‐295.

Ripley, B. D. 1979. Tests of `randomnessʹ for spatial point patterns. Journal of the Royal Statistical Society. Series B (Methodological) 41 (3):368‐374.

Ripley, B. D. 1981. Spatial statistics. New York, New York, USA: Wiley.

Robblee, Michael B., T.R. Barber, P.R. Carlson Jr., Michael Durako, James W. Fourqurean, L.K. Muehlstein, D. Porter, L.A. Yarbro, R.T. Zieman, and Joseph C. Zieman. 1991. Mass Mortality of tropical seagrass Thalassia testudinum in Florida Bay (USA). Marine Ecology Progress Series 71:297‐299.

Rossbach, Kelly, and Denise L. Herzing. 1997. Underwater observations of benthic‐ feeding bottlenose dolphins (Tursiops truncatus) near Grand Bahama Island, Bahamas. Marine Mammal Science 13 (3):498‐504.

Rudnick, D. T., Z. Chen, D. L. Childers, J. N. Boyer, and T. D. Fontaine, III. 1999. Phosphorus and nitrogen inputs to Florida Bay: The importance of the Everglades watershed. Estuaries 22 (2B):398‐416.

218

Rushton, S. P., S. J. Ormerod, and G. Kerby. 2004. New paradigms for modelling species distributions? Journal of Applied Ecology 41 (2):193‐200.

Sargeant, Brook L., Aaron J. Wirsing, Michael R. Heithaus, and Janet Mann. 2007. Can environmental heterogeneity explain individual foraging variation in wild bottlenose dolphins (Turisops sp.)? Behavioral Ecology and Sociobiology 61:679‐688.

Schick, R. S., and Dean Urban. 2000. Spatial components of bowhead whale (Balaena mysticetus) distribution n the Alaskan Beaufort Sea. Canadian Journal of Fish and Aquatic Science 57:2193‐2200.

Schneider, D., and John F. Piatt. 1986. Scale‐dependent correlation of seabirds with schooling fish in a coastal ecosystem. Marine Ecology Progress Series 32:237‐246.

Schoener, Thomas W. 1971. Theory of feeding strategies. Annual Review of Ecology and Systematics 2:369‐404.

Schoener, Thomas W. 1974. Resource Partitioning in Ecological Communities. Science 185 (4145):27‐39.

Scott, Gerald P., Michael R. Dewey, Larry J. Hansen, Ralph E. Owen, and Edward S. Rutherford. 1989. How many mullet are there in Florida Bay? Bulletin of Marine Science 44 (1):89‐107.

Segurado, Pedro, and Miguel B. Araujo. 2004. An evaluation of methods for modelling species distributions. Journal of Biogeography 31 (10):1555‐1568.

Shaw, A.B. 1989. Distribution of mollusks in sediments of Florida Bay and reef tract. Bulletin of Marine Science 44 (1):523.

Shepard, R.N. 1962. The analysis of proximities: Multidimensional scaling with an unknown distance function. Psychometrika 27:125‐139.

Shinn, EA, BH Lidz, and MW Harris. 1994. Factors controlling distribution of Florida Keys Reefs. Bulletin of Marine Science 54 (3):1084.

Sih, Andrew. 1984. The behavioral response race of predators and prey. American Naturalist 123:143‐150.

219

Sih, Andrew. 1998. Game theory and predator‐prey response races. In Advances in Game Theory and the Study of Animal Behavior, edited by L. A. Dugatkin and H. K. Reeve: Oxford University Press.

Simpfendorfer, C. A., A.B. Gotreid, and R.B. McAuley. 2001. Size, sex, and geographic variation in the diet of the tiger shark, Galeocerdo cuvier, from Western Australian waters. Environmental Biology of Fishes 61 (37‐46).

Sims, David W., and Victoria A. Quayle. 1998. Selective foraging behaviour of basking sharks on zooplankton in a small‐scale front. Nature 393:460‐464.

Skov, Henrik, Jan Durinck, Finn Danielsen, and Dorete Bloch. 1995. Co‐Occurrence of Cetaceans and Seabirds in the Northeast Atlantic. Journal of Biogeography 22 (1):71‐88.

Skov, Henrik, and Erik Prins. 2001. Impact of estuarine fronts on the dispersal of piscivorous birds in the German Bight. Marine Ecology Progress Series 214:279‐287.

Smith, Thomas B., and Skuli Skulason. 1996. Evolutionary significance of resource polymorphisms in fishes, amphibians, and birds. Annual Review of Ecology and Systematics 27:111‐133.

Smith, Thomas J, III, Harold J. Hudson, Michael B. Robblee, George V. N. Powell, and Peter J. Isdale. 1989. Freshwater flow from the Everglades to Florida Bay: A historical reconstruction based on fluoresence banding in the coral solenastrea bournoni. Bulletin of Marine Science 44 (1):274‐282.

Smolker, Rachel, Andrew Richards, Richard C. Connor, Janet Mann, and Per Berggren. 1997. Sponge carrying by dolphins (Delphinidae, Tursiops sp.): A foraging specialization involving tool use? Ethology 103:454‐465.

Sogard, Susan M., George V. N. Powell, and Jeff G. Holmquist. 1989. Spatial distribution and trends in abundance of fishes residing in seagrass meadows on Florida Bay mudbanks. Bulletin of Marine Science 44 (1):179‐199.

220

Sogard, Susan M., George V. N. Powell, and Jeff G. Holmquist. 1989. Utilization by fishes of shallow, seagrass‐covered banks in Florida Bay: 1. Species composition and spatial heterogeneity. Environmental Biology of Fishes 24 (1):53‐65.

Thayer, Gordon W., and Alexander J. Chester. 1989. Distribution and abundance of fishes among basin and channel habitats in Florida Bay. Bulletin of Marine Science 44 (1):200‐219. Thayer, Gordon W., George V. N. Powell, and Donald E. Hoss. 1999. Composition of larval, juvenile, and small adult fishes relative to changes in environmental conditions in Florida Bay. Estuaries 22 (2B):518‐533.

Thomas, et al. 2005. Distance 5.0 Release 1. Research Unit for Wildlife Population Assessment, University of St. Andrews, UK.

Torres, Leigh G., Michael R. Heithaus, and Bryan Delius. 2006. Influence of teleost abundance on the distribution and abundance of sharks in Florida Bay, USA. Hydrobiologia 596:449‐455.

Torres, Leigh G., and Andrew J. Read. submitted. How to catch a fish? The ecology of bottlenose dolphins (Tursiops truncatus) foraging tactic fidelity of in Florida Bay, Florida. Behavioral Ecology.

Torres, Leigh G., Andrew J. Read, and Pat Halpin. submitted. Fine‐scale modeling of bottlenose dolphin habitat selection: Does prey data improve predictive capacity?

Torres, Leigh G., Patricia E. Rosel, Caterina DʹAgrosa, and Andrew J. Read. 2003. Improving management of overlapping bottlenose dolphin ecotypes through spatial analysis and genetics. Marine Mammal Science 19 (3):502‐514.

Torres, Leigh G., and Dean Urban. 2005. Using spatial analysis to assess bottlenose dolphins as an indicator of healthy fish habitat. In Estuarine Indicators, edited by S. A. Bortone. Boca Raton, Florida: CRC Press.

U.S. Census. 2000. South Florida Census 2000. Hollywood, FL 33021: South Florida Regional Planning Council.

221

Unterseher, Martin, and Ophir Tal. 2006. Influence of small scale conditions on the diversity of wood decay fungi in a temperate, mixed deciduous forest canopy. Mycological Research 110:169‐178.

Urban, Dean, C. Miller, N. L. Stephenson, and Pat Halpin. 2000. Forest pattern in Sierran landscapes: the physical template. Landscape Ecology 15:603‐620.

Vaughan, I. P., and S. J. Ormerod. 2005. The continuing challenges of testing species distribution models. Journal of Applied Ecology 42 (4):720‐730.

Venables, W.N., and B. D. Ripley. 1997. Modern applied statistics with S‐PLUS. 2nd ed. New York: Springer.

Verbyla, D. L. 1987. Classification trees: A new discrimination tool. Canadian Journal of Forest Research 17:1150‐1152.

Vilchis, L. Ignacio, Lisa T. Ballance, and Paul Fiedler. 2006. Pelagic habitat of seabirds in the eastern tropical Pacific: effects of foraging ecology on habitat selection. Marine Ecology Progress Series 315:279‐292.

Vlietstra, Lucy S. 2005. Spatial associations between seabirds and prey: effects of large‐ scale prey abundance on small‐scale seabird distribution. Marine Ecology Progress Series 291:275‐287.

Wanless, Harold R., Randall W. Parkinson, and Lenore P. Tedesco. 1997. Sea level control on stability of Everglades wetlands. In Everglades: The ecosystem and its restoration, edited by S. M. Davis and J. C. Ogden. Boca Raton, Florida: St. Lucie Press.

Wanless, Harold R., and Matthew G. Tagett. 1989. Origin, growth and evolution of carbonate mudbanks in Florida Bay. Bulletin of Marine Science 44 (1):454‐489.

Weatherbee, W. M., and E. Cortes. 2004. Food Consumption and Feeding Habits. In Biology of sharks and their relatives, edited by J. C. Carrier, J. A. Musick and M. R. Heithaus. Boca Raton: CRC Press.

222

Weinrich, Mason T., Mark R. Schilling, and Cynthia R. Belt. 1992. Evidence for acquisition of a novel feeding behaviour: lobtail feeding in humpback whales, Megaptera novaeangliae. Animal Behaviour 44 (6):1059‐1072.

Weiss, Jessica. 2006. Foraging habitats and associated preferential foraging specializations of bottlenose dolphin (Tursiops truncatus) mother‐calf pairs. Aquatic Mammals 32 (1):10‐19.

White, Peter S. 1997. Synthesis: Vegetation pattern and process in the Everglades ecosystem. In Everglades: The ecosystem and its restoration, edited by S. M. Davis and J. C. Ogden. Boca Raton: St. Lucie Press.

Whitehead, Hal, and Janet Mann. 2000. Female reproductive strategies of cetaceans: Life histories and calf care. In Cetacean Societies: Field studies of dolphins and whales, edited by J. Mann, R. C. Connor, P. L. Tyack and H. Whitehead. Chicago and London: University of Chicago Press.

Wiley, Tonya R., and Colin A. Simpfendorfer. 2007. The ecology of elasmobranchs occurring in the Everglades National Park, Florida: Implications for conservation and management. Bulletin of Marine Science 80 (1):171‐189.

Würsig, Bernd, and Melany Würsig. 1977. The photographic determination of group size, composition, and stability of coastal porpoises (Tursiops truncatus). Science 198:755‐756.

Zieman, Joseph C., James W. Fourqurean, and Richard L. Iverson. 1989. Distribution, abundance and productivity of seagrasses and macroalgae in Florida Bay. Bulletin of Marine Science 44 (1):292‐311.

Zollett, Erika, and Andrew J. Read. 2006. Depredation of catch by bottlenose dolphins (Tursiops truncatus) in the Florida king mackerel (Scomberomorus cavalla) troll fishery. Fishery Bulletin 104:343‐349.

223

Biography

LEIGH GABRIELA TORRES Birthplace: Mexico City, Mexico Date: October 15, 1975

EDUCATION DOCTOR OF PHILOSOPHY, October 2007 Department of the Environment Duke University Marine Lab Duke University, Durham, NC

MASTER OF ENVIRONMENTAL MANAGEMENT, May 2001 Nicholas School of the Environment and Earth Sciences Duke University, Durham, NC

BACHELOR OF ARTS, cum laude, May 1997 The American University, Washington, D.C.

GRANTS & AWARDS 2007 ‐Summer Support‐Oak Foundation, Duke Marine Lab, $3,000 2006 ‐Duke Marine Laboratory Dissertation Fellowship, $17,000 2004 & 2005 ‐ NOAA/Coastal Ocean Program, grant for research, $66,444 2000 ‐ Recipient of Sussman Fund Fellowship, conservation research, $4,000 2007 ‐ Best Student Presentation at SE and Mid Atlantic Marine Mammal Symposium 2001 ‐ Best Pre‐Doctoral Oral Presentation, 14th Biennial Marine Mammal Conference 2001 ‐ Best Student Presentation at SE and Mid Atlantic Marine Mammal Symposium 1993‐1997 ‐ Scholar ‐ Athlete for Colonial Athletic Association

SELECTED PUBLICATIONS Torres, L.G., M.R. Heithaus and B. Delius. 2006. Influence of teleost abundance on the distribution and abundance of sharks in Florida Bay, USA. Hydrobiologia 569:449‐455.

Redfern, J.V., M.C. Ferguson, E.A. Becker, K.D. Hyrenbach, C. Good, J. Barlow, K. Kaschner, M.F. Baumgartner, K.A. Forney, L. Balance, P. Fauchald, P. Halpin, T. Hamazaki, A. Pershing, S.S. Qian, A. Read, S.B. Reilly, L.G. Torres, R. Werner. 2006. Techniques for cetacean‐habitat modeling. Marine Ecology Progress Series 310:271‐295.

224

Torres, L.G., and D. Urban. 2005. Using spatial analysis to assess bottlenose dolphins as an indicator of healthy fish habitat. in: Estuarine Indicators. Ed: Stephen A. Bortone. CRC Press.

Torres, L.G., W.A. McLellan, E.M. Meagher, and D.A. Pabst. 2005. Seasonal distribution and relative abundance of bottlenose dolphins, Tursiops truncatus, along the US mid‐ Atlantic Coast. Journal of Cetacean Research and Management 7(2).

Johnston, D.W., A.S. Friedlaender, L.G. Torres, and D.M. Lavigne. 2005. Variation in ice cover on the east coast of Canada, 1969‐2002: climate variability and implications for harp and hooded seals. Journal of Climate Research 29:209‐222.

Torres, L.G., P. Rosel, A. Read, and C. D’Agrosa. 2003. Improving management of overlapping bottlenose dolphin ecotypes through spatial analysis and genetics. Marine Mammal Science 19(3):502‐514.

ADDITIONAL INFORMATION Member, The Society for Marine Mammalogy, 1997‐Present Member, The Ocean Conservancy, 2000‐Present Co‐Founder of The Green Wave, 2006

225