THE DOUBLE-CRESTED CORMORANT Kate Sheehan Clemson University, [email protected]

THE DOUBLE-CRESTED CORMORANT Kate Sheehan Clemson University, Kates@G.Clemson.Edu

Clemson University TigerPrints

All Dissertations Dissertations

12-2013 DISTRIBUTION, ECOLOGY, AND TROPHIC RELATIONSHIPS OF A COLONIAL WATERBIRD: THE DOUBLE-CRESTED CORMORANT Kate Sheehan Clemson University, [email protected]

Follow this and additional works at: https://tigerprints.clemson.edu/all_dissertations Part of the Ecology and Evolutionary Biology Commons

Recommended Citation Sheehan, Kate, "DISTRIBUTION, ECOLOGY, AND TROPHIC RELATIONSHIPS OF A COLONIAL WATERBIRD: THE DOUBLE-CRESTED CORMORANT" (2013). All Dissertations. 1230. https://tigerprints.clemson.edu/all_dissertations/1230

This Dissertation is brought to you for free and open access by the Dissertations at TigerPrints. It has been accepted for inclusion in All Dissertations by an authorized administrator of TigerPrints. For more information, please contact [email protected]. DISTRIBUTION, ECOLOGY, AND TROPHIC RELATIONSHIPS OF A COLONIAL WATERBIRD: THE DOUBLE-CRESTED CORMORANT

______

A Dissertation Presented to the Graduate School of Clemson University

______

In Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy Wildlife and Fisheries Biology

______

by Kate Lyn Sheehan December 2013

______

Accepted by: Dr. Ron J. Johnson, Committee Co-Chair Dr. Greg K. Yarrow, Committee Co-Chair Dr. Alan Johnson Dr. David Tonkyn ABSTRACT

The realized distribution of an organism is dependent on the environmental characteristics of the landscape, biological interactions within the communities in which it lives, and geographic barriers to dispersal. Changes in habitat can influence landscape characteristics, and consequently, the distribution of organisms within the landscape. The

Double-crested Cormorant ( Phalacrocorax auritus ) is a piscivorous waterbird implicated

in human-wildlife conflicts at aquaculture facilities and natural aquatic systems where

they compete for resources (fish) with anglers and commercial fishing guides. Impacts of

P. auritus on aquatic systems result from their consumption of fish stocks and their

contamination of water and soil with guano near nesting and roosting locations. The

consequences of top-down and bottom-up forcing associated with P. auritus colonies

have been evaluated singularly within particular components of a food web, but they have

not been evaluated individually from a community-wide perspective. We observed food

chains and trophic networks of communities from lakes where P. auritus breed and

compared their composition, biomass, and topologies to those of a mesocosm system

where the effects of P. auritus were simulated with the addition of fertilizer and the

removal of fishes. When organisms in the lake systems were pooled into trophic levels

within food chains, the patterns of relative biomass showed evidence of top-down and

bottom-up forcing. In the mesocosms, we also were able to capture differences in top-

down, bottom-up, and combined forcing in the topological assessments of trophic

networks. The addition of nutrients (bottom-up forcing) was associated with smaller, yet

more plentiful fishes. The removal of fishes (top-down forcing during all experimental ii

phases) was associated with a high biomass of fish. The combination of top-down and bottom-up forcing had no impact on the aquatic community when applied at low levels but, at high-intensities, these factors led to a sharp reduction of fish biomass. Thus, the impact of P. auritus in freshwater communities is unlikely to be negative unless their numbers and duration of use is extreme, a rarely realized condition in natural systems.

P. auritus has experienced population declines and rebounds within the last century. An increase in the number of water bodies such as reservoirs, ponds, and aquaculture facilities has changed the cormorant carrying capacity of the North American landscape. Consequently, the distribution of P. auritus has expanded to new geographic areas where foraging and breeding success is high. Human-wildlife conflicts with P. auritus have led to culling programs in many states, with the exception that a resident subspecies of P. auritus that breeds in the southeastern United States is protected from culling and harassment. Because the distribution of multiple subspecies of P. auritus can overlap within southeastern states, it is important to measure differences in habitat use among the subspecies to minimize conflicts in management programs. We developed a species distribution model for two subspecies of P. auritus from their known breeding areas ( P. a. auritus from Minnesota and P. a. floridanus from Florida) and transferred those models to South Carolina, where there is question about which subspecies is breeding on reservoir lakes. The models indicate that the breeding habitats of the two subspecies differ. The Florida model correctly predicted nesting locations in South

Carolina. The Minnesota model was also able to predict some nesting sites in South

Carolina, but with low prediction values that suggested the habitat in South Carolina was

iii

not suitable for nesting P. a. auritus . Thus, our models support the presence of the

protected P. a. floridanus subspecies breeding in South Carolina.

Landscape characteristics influence the distribution and movements of P. auritus , which in turn reflect the composition and geographic distribution of organisms that they encounter, specifically their helminthic parasites. Trophically transmitted parasites require multiple host species to complete a single revolution of a life cycle. P. auritus obtain helminthic parasites directly from the organisms on which they feed. Thus, the suite of parasites that a host P. auritus contains can indicate the complexity of the aquatic communities from which it fed. We assessed the intestinal parasites of 218 P. auritus that had been collected by state and federal agencies during culling activities. We document

15 types of parasites in P. auritus , many of which had not been previously reported in this species, and others from geographic regions not previously reported. We assessed similarities and differences in the parasite assemblages of P. auritus at local and regional scales, and between migratory ( P. a. auritus ) and resident ( P. a. floridanu s) subspecies.

The parasite assemblages found within P. auritus were distinct among many sampling locations, among geographic regions, and between resident and migratory subspecies.

This appears to be a useful indicator of host grouping and movement and could be investigated further by including additional geographic regions and host species.

Moreover, the parasite differences between resident and migratory subspecies add credence to the habitat model finding that the resident subspecies breeding in South

Carolina is the protected P. a. floridanus .

Our assessments of local habitat characteristics in relation to the distribution of P. auritus are a new way of demonstrating the differences between sympatric subspecies.

Our methodology is able to confirm differences between subspecies that current molecular techniques have been unable to capture. Furthermore, we document how P. auritus predation and defecation can influence the aquatic and parasitic organisms that they encounter in aquatic communities. Because these interactions are restricted to areas where P. auritus occur, our contributions to understanding the distribution of cormorant subspecies and potential impacts on aquatic communities are critical for evaluating management needs and options.

ACKNOWLEDGEMENTS

I have the sincerest appreciation for the help, guidance, and opportunities while performing this research project. Dr. Ron Johnson was a generous source of wisdom and strength who was instrumental in the completion of my dissertation research. Dr. Greg

Yarrow has provided invaluable insight into the meaning and application of my research, and I am grateful for his assistance and encouragement. Dr. David Tonkyn helped me to critically assess the meaning and application of my work, and his enthusiasm for my research results were truly inspiring. I am grateful to have had the specialty knowledge of

Dr. Alan Johnson, who greatly contributed to the final assessment of my most complicated assessments. I thank the entire committee for their guidance and support throughout this process.

My field research was made possible by the Minnesota Department of Natural

Resources (MNDNR), the South Carolina Department of Natural Resources (SCDNR), and the National Parks Service (NPS) in Minnesota. These agencies provided me with support in kind, boats, personnel, freezer space, and lodging. Mary Catherine Martin and

Chad Holbrook were kind enough to provide boat time to me not only in my personal research endeavors, but to my classes of students interested in learning field sampling techniques. I would not have contacted many of these agencies without the suggestions of Dr. William Bowerman, and I thank him for his encouragement while in the project development phase.

I owe many thanks to Dr. Patrick Jodice, who encouraged my transition into waterbird research, specifically with that of the Double-crested Cormorant. His assistance in project development was critical to the success of this project. Finally, I thank Dr.

Michael Childress for getting me to Clemson, where these opportunities were made available to me.

The Clemson University Creative Inquiry (CI) program provided me an amazing opportunity to teach, perform research, and mentor undergraduate researchers. With the help of Dr. Ron Johnson, Dr. Greg Yarrow and Dr. Patrick Jodice I created two CI programs that were wildly successful in drawing in the brightest undergraduates who wanted to perform field studies in aquatic systems or work with waterbirds. There were over 50 students who participated in these courses, and I would like to thank each one for helping to make me a more competent and patient instructor. In particular, I would like to thank Madeline Kral, David Saari, and Justin Holiday for their assistance in the development of the mesocosm system. In addition to these very special students, I thank

Scott Davis at the Clemson Bottoms for his assistance with pond maintenance and the use of supplies, equipment, and vehicles of the facility. Brandon Newton and Andrew

Begrowicz were student interns that performed numerous samplings in the mesocosms, and Charles Palmer, Meghan Philips, Madeline McMillan, Courtney Adams, Shauna

Gearhart, and Braden Stocks, who helped to assess the bulk of organism samples in the laboratory. Without the help of such stellar students, I would not have finished this program in 5 years. Furthermore, the CI program provided financial support for this

vii

research and for the presentation of its findings at scientific meetings by me and my CI students. None of this would have been possible without that support.

Most importantly, I would like to thank my family and friends for their support during the tenure of my stay at Clemson. My parents were supportive and encouraging throughout the duration of my dissertation, even under the most stressful situations. My other family members have been quite understanding of my noticeable absence from family events while studying one thousand miles away. Their unwavering support of my decisions to continue with schooling helped me to maintain my stamina. My Coast Guard family in Mobile, AL and Savannah, GA were understanding with my extended commute and the need for batched drilling while working in the U.S. Coast Guard Selected

Reserves. They never ceased to remind me of the importance of a strong education and the constant pursuit for knowledge and training. My new friends and lab mates at

Clemson were welcoming, supportive, and inspirational to me and I hope the relationships that we have forged will grow stronger and last the rest of my life. Finally,

I thank Dr. Samuel Esswein for his support, patience, and understanding while I was knee-deep in my dissertation research. As a previous Ph. D. student in the program, he knew better than anyone what was required of a high-performing graduate student.

viii

TABLE OF CONTENTS

Page

TITLE PAGE ...... i

ABSTRACT ...... ii

ACKNOWLEDGEMENTS ...... vi

LIST OF TABLES ...... xii

LIST OF FIGURES ...... xv

CHAPTER

I. PROJECT DESCRIPTION ...... 1

The Costs and Benefits of Colonial Living ...... 4 Avian Colonies and Trophic Web Structure and Function ...... 5 Model Colonial Species: The Double-crested Cormorant ...... 7 Cormorant Impacts on Aquatic Trophic Webs ...... 10 Consequences of Movements of Double-crested Cormorants at Different Scales ...... 12 Figures...... 19 Literature Cited ...... 21

II. COMPARISON OF AQUATIC COMMUNITIES NEAR COLONIAL WATERBIRD COLONIES AND MESOCOSM EXPERIMENTS OF SIMULATED COLONIES ...... 31

Abstract ...... 31 Introduction ...... 32 Methods...... 35 Results ...... 45 Discussion ...... 50 Tables ...... 55 Figures...... 60 ix

Table of Contents (Continued) Page

Acknowledgements ...... 73 Literature Cited ...... 74

III. NESTING HABITAT SUITABILITY OF TWO DOUBLE- CRESTED CORMORANT SUBSPECIES ( PHALACROCORAX AURITUS AURITUS AND P. A. FLORIDANUS ) ...... 79

Abstract ...... 79 Introduction ...... 80 Methods...... 83 Results ...... 92 Discussion ...... 97 Tables ...... 106 Figures...... 108 Acknowledgements ...... 121 Literature Cited ...... 122

IV. INTESTINAL PARASITES OF CULLED DOUBLE-CRESTED CORMORANTS ( PHALACROCORAX AURITUS) ...... 131

Abstract ...... 131 Introduction ...... 132 Methods...... 133 Results ...... 134 Discussion ...... 145 Tables ...... 148 Acknowledgements ...... 153 Literature Cited ...... 154

V. PARASITE ASSEMBLAGES OF DOUBLE-CRESTED CORMORANTS AS INDICATORS OF GEOGRAPHICALLY SIMILAR SUBSPECIES ...... 160

Abstract ...... 160 Introduction ...... 161 Methods...... 164

Table of Contents (Continued) Page

Results ...... 168 Discussion ...... 173 Tables ...... 179 Figures...... 186 Literature Cited ...... 190

VI. CLOSING REMARKS ...... 193

APPENDICES ...... 196

Appendix A: Supplemental Materials for Chapter II ...... 197 Appendix B: Supplemental Material for Chapter III ...... 218 Appendix C: Supplemental Material for Chapter V ...... 222 Appendix D: Historical Excerpt from Literature ...... 234

LIST OF TABLES Page

Table

2.1 Sample sites on lakes where nesting colonies of P. auritus occur...... 55

2.2 Univariate analyses of trophic variables measured for aquatic food chains of two lakes where P. auritus nest...... 56

2.3 Comparison of variables combined among sample types in aquatic communities...... 57

2.4 Contingency table of output values from the Canonical Discriminant Analysis corresponding with the classifications of food webs based on network topological characteristics...... 58

2.5 Correlations of the eigenvectors for the three significant axes describing the separation of mesocosm food webs based on treatments...... 59

3.1 Results of Student’s T-tests used to compare model predictions based on prediction value and values truncated at the threshold for maximum training sensitivity plus specificity (MTSS) ...... 106

3.2 Variable contribution for parameters included in the final nesting habitat models developed with Maxent in Minnesota and Florida. Factors that appear in both models are highlighted ...... 107

4.1 Trematodes reported in cormorants from the United States and Canada...... 148

4.2 Nematodes reported in cormorants from the United States and Canada...... 150

4.3 Acanthocephalans and cestodes reported in cormorants from the United States and Canada ...... 151

4.4 Location, collection information, and parasite diversity estimates (within site [alpha] and regional [gamma]) for all culled birds assessed in this study...... 152

xii

List of Tables (Continued)

Page

Table

5.1 Sampling sites and characterizations used in univariate and multivariate analyses...... 179

5.2 Number (sum), intensity (±SE; avg int), rank occurrence (avg rank), and prevalence (prev) of each documented parasite...... 180

5.3 P-values of linear regression analysis of parasite presence/absence (prevalence) and abundance (intensity) over geographic gradients (latitude and longitude)...... 182

5.4 P-values and R-square values for t-test analyses of parasite prevalence and intensity (if applicable) of P. auritus parasites. Significant differences indicated in bold...... 183

5.5 Model parameters and output used to compare the performance of Discriminant Analyses performed for three categorical groupings (sampling site, geographic region, migration behavior) on three iterations of the same parasite data (infection intensity, prevalence, and ranked intensity) ...... 184

5.6 Comparison of model performance in predicting the correct classification of P. auritus by sample site, geographic region, or migration status (migratory or resident)...... 185

A.1 Averaged variable values for sample units for each treatment (Treatm) during each Phase...... 203

A.2 Trophic Relationship matrix for organisms collected from mesocosms...... 204

A.3 Proportion of trophic group positions held for organisms within food webs (Webs) collected from lakes where P. auritus breed...... 205

A.4 Proportion of trophic group positions held for organisms within food webs (Webs) collected in the mesocosm system...... 206

xiii

List of Tables (Continued)

Page

Table

B.1 Variable description and sources of data used in Maxent models...... 218

B.2 Description of steps taken to derive each variable for the states of Minnesota, Florida, and South Carolina ...... 220

C.1 Binary table of significant findings from pairwise Adonis analyses...... 222

C.2 Classification table for Discriminant Analysis grouping based on host collection site...... 223

C.3 Classification table for Discriminant Analysis grouping based on host collection region...... 224

C.4 Classification table for Discriminant Analysis grouping based on host migration status without winter collection sites included in the initial model development...... 225

C.5 Classification table for Discriminant Analysis grouping based on parasite assemblages as identified based on the Adonis Analysis ...... 226

xiv

LIST OF FIGURES

Page

Figure

1.1 Diagram of organism abundance based on Elton’s “Pyramid of Numbers” (Elton 1927). The base of the food web provides resources that upper trophic levels consume...... 19

1.2 Diagram of a trophic cascade (left 4 columns) along a left-to- right gradient with increasing numbers of trophic levels where resources limit the abundance of Producers. The standing biomass of Producers and Primary Consumers is influenced...... 20

2.1 Trophic cascade as presented by Holtz and others (2000)...... 60

2.2 Eltonian pyramid modified to represent biomass estimates from Sukhdeo (2010) based on data from Lindeman (1942)...... 61

2.3 Aerial photograph of the Bottoms region of Clemson University...... 62

2.4 Relative biomass estimates of organisms collected from field sites in Minnesota...... 63

2.5 Relative biomass estimates of organisms collected from field sites in South Carolina...... 64

2.6 Biplots of significant ANCOVA models for microconsumers and zooplankton over time...... 65

2.7 Biplots of significant ANCOVA models for benthic organisms over time...... 66

2.8 Relative estimates of biomass for the three trophic levels with phytoplankton estimates for the base of the web (basal)...... 67

2.9 Relative abundance of trophic levels using only phytoplankton estimates as the base of the trophic chain (basal)...... 68

2.10 Food webs depicting the relationships among trophic groups from Kabetogama Lake...... 69 xv

List of Figures (Continued)

Page

Figure

2.11 Food webs depicting the relationships among trophic groups from the Stumphole area of Lake Marion...... 70

2.12 Food web topologies from all treatments for four major sampling sessions ...... 71

2.13 Canonical plot of food webs in ordinate space...... 72

3.1 Aerial imagery of nesting sites reported in Florida where sandy and shell-hash substrates connect P. auritus nesting areas (inset image) in trees. Example polygons drawn around P. auritus colonies...... 108

3.2 Prediction of P. auritus nesting habitat in the state of Minnesota based on characteristics from known cormorant breeding colonies...... 109

3.3 Receiver operating characteristic (ROC) curve for the training presence records used to develop the Maxent model for P. auritus nesting habitat in Minnesota...... 110

3.4 Prediction of P. auritus nesting habitat in Florida based on known cormorant breeding colonies...... 111

3.5 Receiver operating characteristic (ROC) curve for the training presence records used to develop the Maxent model for P. auritus nesting habitat in Florida...... 112

3.6 Prediction values of P. auritus nesting habitat in South Carolina based on parameters that describe the ecological niche of cormorants nesting in Minnesota...... 113

3.7 Prediction values of P. auritus nesting habitat in South Carolina based on parameters that describe the ecological niche of cormorants nesting in Florida...... 114

xvi

List of Figures (Continued)

Page

Figure

3.8 Receiver operating characteristic (ROC) curve for the training presence records used to develop the Maxent model for random points generated in Minnesota...... 115

3.9 Receiver operating characteristic (ROC) curve for the training presence records used to develop the Maxent model for random points generated in Florida...... 116

3.10 Variable contribution plot for parameters included in the final Maxent model of P. auritus nesting habitat in Minnesota...... 117

3.11 Predictor profile plots for the 5 most explanatory variables for the Minnesota model...... 118

3.12 Variable contribution plot for variables included in the final Maxent model of cormorant nesting habitat in Florida...... 119

3.13 Predictor profile plots for the 5 most explanatory variables for the Florida model...... 120

5.1 Biplot of ranked intensity data illustrating the distribution of Double-crested Cormorant parasite assemblages from different sites in ordinate space...... 186

5.2 Biplot of ranked intensity data illustrating the distribution of Double-crested Cormorant parasite assemblages from different regions in ordinate space...... 187

5.3 Biplot illustrating the distribution of migratory and resident Double-crested Cormorant parasite assemblages in estimated ordinate space...... 188

5.4 Biplot of parasite assemblages based on output of the Adonis Analysis illustrating the distribution of Double-crested Cormorant parasite assemblages in ordinate space ...... 189

xvii

List of Figures (Continued)

Page

Figure

A.1 Map of pond system...... 207

A.2 Relative biomass estimates for the three trophic levels with emergent vegetation as the base of the food web (basal)...... 208

C.1 Biplots of parasite assemblages based on intensity count data including variable direction rays, which are informative for the migration of points in ordinate space, for sample site...... 227

C.2 Biplots of parasite assemblages based on prevalence data including variable direction rays, which are informative for the migration of points in ordinate space, for sample site...... 228

C.3 Mosaic plot based on the contingency table compiled from results of ranked site data ...... 230

C.4 Mosaic plot based on the contingency table compiled from results of ranked region data...... 231

C.5 Mosaic plot based on the contingency table created with the results from the ranked intensity data used to predict migration status without data from birds collected in winter...... 232

C.6 Mosaic plot based on the contingency table created with the results from parasite assemblages identified by the Adonis Analysis...... 233

xviii

1. CHAPTER I: PROJECT DESCRIPTION

My graduate career for the tenure of my dissertation has been rather unconventional in the sense that I had the great fortune to develop, fund, and carry out my research without prior obligation. My previous graduate training was in Marine

Science with an emphasis on the host-parasite interactions of an estuarine shrimp and a trematode worm (Sheehan et al. 2011). Working with the intermediate (invertebrate) hosts in that system was extremely informative and helped me to develop a useful skillset, but I longed to perform research on definitive (vertebrate) hosts of trematode parasites. Working with vertebrate hosts adds challenges associated with animal welfare, and moral obligations to be scientifically thorough while being prudent and ethically defensible with sample sizes.

Parasites are not organisms that most early researchers are interested in working with, but I found early on that working with less-desirable organisms allowed me the freedom of scientific creativity. As I investigated potential definitive host species to work with for my dissertation, I focused on ‘nuisance’ species that represented opportunities for study with sufficient sampling size and distribution. Ultimately, this led me to work with a waterbird, the Double-crested Cormorant ( Phalacrocorax auritus ).

P. auritus populations are reduced through culling activities in many states and

are considered by some to be ‘death/murder on wings,’ ‘black death,’ or a ‘feathered

pariah’ because of historic and current conflicts with people (Jackson and Jackson 1995).

I consulted the working groups of managers that manage cormorant populations in the

United States, read the vast literature on the subject, and spoke with anglers and local citizens where cormorants live, and other members of the scientific community that work with them. The overwhelming impression from my initial literature research (Colonial

Waterbirds Special Publication 1, 1995) was that this particular species was being lethally managed to control for population growth primarily in areas where cormorants compete for resources in human-made environments (e.g., aquaculture facilities, lakes and ponds stocked with fish; Brugger 1995, Krohn et al. 1995, Price and Nickum 1995).

The ability of P. auritus to extract a resource that humans are also interested in exploiting (fisheries species; Pauly et al. 2002, Vitousek et al. 1997), has led to the development of management plans throughout North America (Cowx and Gerdeaux

2004). To me, P. auritus was a perfect subject for my parasite research: an undesirable subject (parasitology) in an equally undesirable host. Beyond the parasitology, I wanted to assess the assertions that were being made about P. auritus surrounding two main factors that contribute to their current management protocols: 1) that their predation of fishes was detrimental to aquatic communities of freshwater systems, and 2) that their gregarious/colony-forming habits that concentrate fecal matter in high-use areas could impair water quality and aquatic food webs.

The timing of my research was quite fortunate in that there was growing concern regarding cormorant depredation and the field of cormorant research was expanding

(Taylor and Dorr 2003, Coleman and Richmond 2007, Boutin et al. 2011), but no one

appeared to be working with their parasites. I was able to acquire gastrointestinal tracts from cormorant culling operations in Minnesota, Mississippi, Alabama, and Vermont. I was also invited to perform my trophic web studies at Voyageurs National Park in

Minnesota with housing, a boat, and field assistance provided free-of-charge. The

Minnesota Department of Natural Resources (MNDNR) agreed to send me fishes from their seasonal samplings and added sites to their schedule that corresponded with my sampling regime on Kabetogama Lake. They were quite interested in cormorant ecology.

The interest in South Carolina was similarly strong, and I was able to arrange identical support here on Lakes Marion and Moultrie working with the South Carolina Department of Natural Resources (SCDNR).

Our current era in geological history is now recognized by many people as the

Anthropocene, dominated by the activities and consequences of humans including loss of natural habitats, species extinctions, and effects on water, carbon and nitrogen cycles

(Lande 1998, Travis 2003, Sayre et al. 2013). Since the industrial revolution of the mid-

1800s, human activities have had increasing and varied impacts on natural habitats and populations of wild animals (White 1967, Vitousek et al. 1997). Habitat losses occurred as forests and other areas were cleared for agricultural expansion (Brooks et al. 1999,

Brook et al. 2003). Agricultural intensification and widespread use of pesticides in the

20 th Century had additional negative effects on wild birds, such as cormorants, and other

animals (Carson 1962, Johnson et al. 2011). While many species declined with land use

change, others expanded their ranges as the carrying capacity of new landscapes allowed

(Jokimaki and Suhonen 1998, Williams et al. 2006).

This dissertation explores unknown ecological consequences of expanding populations of nuisance cormorant colonies by using field surveys and mesocosms to understand trophic structure and ecological relationships in freshwater ecosystems.

Additionally, parasite analysis and species distribution models are employed to understand the local and widespread consequences of landscape alterations that contribute to changes in the distribution of P. auritus and the diseases they could spread

to fisheries resources.

The Costs and Benefits of Colonial Living

Colonial living (in groups of conspecifics) allows for cooperative foraging, vigilance, and territory defense, which, in many cases, increases survival of individuals while minimizing energy expended for resource acquisition (Horn 1968, Brown and

Brown 1987, Ekman and Hake 1988). Flocking Green Finches ( Carduelis chloris ) exhibit reduced starvation because individuals use flock mates to locate food sources (Ekman and Hake 1988). Cliff Swallows in Nebraska were more likely to avoid predators when nesting in colonies where vigilance effort was increased (Brown and Brown 1987).

Brewer’s Blackbird colony arrangements have been shown to relate directly to food resource distribution (Horn 1968).

Although it can be energetically advantageous to persist in colonies, living in high-density groups can lead to increased intraspecific competition for food, cover, and sexual resources (Brown and Brown 2002, Forrero et al. 2002, Cronin et al. 2012), and increased predation attention (Nisbet 1975, Brown et al. 1990, Clode 1993). Colonial

living has the potential to overexploit local resources. Ashmole (1963) documented a halo-effect of decreased food availability around avian colonies. Penguins living in large colonies must forage far from nesting sites because of local resource depletion, resulting in longer hunting trips with no net increase in food stores (Elliott et al. 2009).

Conflicts of interest occur when animals living in colonies are perceived as sharply conflicting with human interests, yet are provided protection from harassment and persecution. Colonial organisms are often more conspicuous than solitary ones, thus, the majority of human/wildlife conflicts involve animals that live in large groups (Peer et al. 2003, Sijtsma et al. 2013). Flocks of blackbirds in the United States and Dickcissels in

Venezuela cause millions of dollars of crop damage (Basili and Temple 1999, Peer et al.

2003) and inspired the creation of numerous management methods (Winter et al. 2009).

While economic endpoints are often more easily measured, the ecological significance of wildlife conflicts (or benefits) and their underlying ecological mechanisms are unknown in many cases. As such, removal of nuisance species might not ultimately benefit ecosystem functioning even if it appears to reduce an immediate conflict.

Avian Colonies and Trophic Web Structure and Function

Blackall et al. (2008) suggested that avian colonies contaminate terrestrial run-off waters with phosphorus and the atmosphere with ammonia. Choy et al. (2010) documented decreased productivity in lower trophic levels of aquatic systems where persistent avian droppings were high in organic pollutants. Furthermore, ecosystem services such as water quality and suitability of an area for recreational use can decline

near large avian colonies. Bottom-up effects on water quality and community productivity have the potential to cause shifts in aquatic community assemblages (Figure

1.1).

Increased primary productivity in aquatic communities can lead to an overabundance of microalgae and subsequent blooms of zooplankters, in addition to increases in submerged and emergent aquatic vegetation (Carbiener et al. 1990,

Livingston 2001, Heisler et al. 2008). While macrophytes provide refuge and food resources for some grazers (Schriver et al. 1995), emergent vegetation transfers fewer nutrients to upper trophic levels compared to microalgae (Mann 1988) and, by shading the water column, can diminish dissolved oxygen concentrations required by most consumers for respiration (Gee et al. 1997, Wilcock and Nagels 2001). While the storage of carbon in standing biomass can be a positive consequence of increased primary productivity (Smith 1981), decreased water quality, restriction of water circulation, and reduced visibility from thick stands of vegetation can ultimately reduce the biomass of organisms at upper trophic levels in these areas (Frodge et al. 1990, Jeppesen et al. 1997,

Diaz 2001). Thus, increased macrophyte production could force larger fishes into open waters (Petry et al. 2003, Troutman et al. 2007) where avian colonies have a better chance of predating them (Kersten et al. 1991).

By consuming large predatory fish, pursuit-diving birds also have the potential to alter aquatic communities through trophic cascades where the abundance of upper trophic levels (top predators) directly influence the abundance of prey populations and indirectly

influence the abundance of food sources on which the prey species rely (Figure 1.2).

Many studies have shown the powerful influence of top predators on the abundance and structure of lower organisms within a food web (Letourneau and Dyer 1998, Tessier and

Woodruff 2002), but few studies have experimentally separated the effects of top-down and bottom-up influences on aquatic communities (Moon and Stilling 2002), which is critical for determining the true impacts that avian colonies have on aquatic communities.

Model Colonial Species: The Double-crested Cormorant

P. auritus is a piscivorous pursuit-diving waterbird that can consume ¾ kilograms

of fish per bird each day (Kelly 2008, Goktepe 2012). These birds nest and forage

colonially, often in flocks of dozens to thousands, mixed with other waterbirds. Multiple

population crashes of P. auritus have led to reduced and patchy distribution of this

species for decades (Kirsch 1995, Jackson and Jackson 1995). Legislative protection of

P. auritus from persecution and the ban of widespread persistent organic pesticide use paved the way for their current recovery and apparent range expansion (Chapdelaine and

Bedard 1995, Krohn et al. 1995).

Foraging and nesting activities of current populations of P. auritus have resulted in conflicts with humans. Historically, human removal of eggs from breeding colonies provided a food source to Native American tribes and early settlers (Cott 1953).

Amendments to the Migratory Bird Act (16 USC Chapter 7, Subchapter II) protected this species from harvest in 1972 and, because the meat of P. auritus is not considered to be

palatable (Fowler 1965, Appendix D), there is no hunting season for this species. Recent

conflicts are characterized primarily in three ways: 1) competition with fishing interests,

2) depredation at aquaculture facilities, and 3) impacts on trees and other vegetation where nesting and roosting colonies persist (Thompson et al. 1995, Hatch and Weseloh

1999). Other specific conflicts such as impacts on endangered salmon restoration efforts are recognized (Hawkes 2013).

P. auritus has been shown to be an opportunistic forager (Kirsch 1995, Fenech et al. 2004, Withers and Brooks 2004). Anglers argue, in many locations, that the most abundant fishes consumed by cormorants could be young sport fish (Trapp et al. 1997); however, P. auritus diet studies have failed to confirm this assertion (Kirsch 1995, Trapp et al. 1997, Withers and Brooks 2004). Although sport fish are often rare in P. auritus gut contents, correlative studies of cormorant colonies and sport fish recruitment have suggested some interaction between P. auritus and yellow perch (Goktepe et al. 2012) and small mouth bass (DeVault et al. 2012, Farquhar et al. 2012). A more ecologically- relevant argument suggesting P. auritus reduces sport fish stocks through competition for forage fish was refuted by Wollkind (1976) and Abrams (1982), who demonstrated that the inclusion of multiple predators into trophic models can stabilize the populations of competitors and prey species (Armstrong and McGehee 1980 and Grover 1997 provide mechanistic descriptions). Although resource competition might limit sport fish populations in some areas, many lakes of the southeastern United States are suggested to have abundant forage fish populations that are limited by intraspecific competition rather than by predation (personal communication: Scott Lamprecht, SCDNR).

Other impacts to economic fisheries have been attributed to expanding populations of waterbirds (Duffy 1995). For example, White Pelicans ( Pelecanus erythrorhynchos ) have been shown to transmit parasitic worms to farmed catfish, effectively ruining millions of kilograms of commercial meat (Overstreet et al. 2002,

Overstreet and Curran 2004). P. auritus are also carriers of various diseases and parasites that can be transmitted to aquatic species such as molluscs, fishes, and other waterbirds

(Kuiken et al. 1999, Friend and Franson 1999, Allison et al. 2005, Gilchrist 2005). In aquaculture ponds filled with commercially important food items, cormorants are capable predators that can depredate hundreds of thousands of farmed fishes in a season (Brugger

1995, Glahn and Stickley 1995, Glahn et al. 1995). Although the aquaculture industry illustrates a clear case for cormorant management, ecological consequences of P. auritus on natural systems have yet to be established (Erwin 1995).

The apparent negative effects that P. auritus have on water quality (Tamisier and

Boudouresque 1994, Klimaszyk et al. 2008), farmed and natural fish populations (Dorr et al. 2004, Adkins et al. 2010), human structures (e.g., docks, lighthouses), and shoreline forest communities (Hobara et al. 2005, Breuning-Madsen et al. 2008) have led to lethal colony controls of breeding and wintering cormorants in many states. Although some researchers have deemed lethal management strategies successful anecdotally (Bedard et al. 1995, Dorr et al. 2010, Farquhar et al. 2012), the impacts of cormorant colonies and the potential influences of cormorant control programs on aquatic ecosystems have yet to be investigated.

Cormorant Impacts on Aquatic Trophic Webs

It is possible that cormorant guano can actually provide a pulse of nutrients to aquatic systems and rejuvenate or support local food webs, thereby enhancing ecosystem services through a bottom-up effect (right column, Figure 1.2). Simultaneously, as a top predator of aquatic food webs, P. auritus has the potential to alter the community

composition through top-down forcing.

Although many studies have shown that top-down forcing overshadows the effects of bottom-up processes (Jeppesen et al. 1997), others have demonstrated that manipulation of upper trophic levels alone is not sufficient to cause meaningful changes in ecosystem services (Kasprzak et al. 2003). The extent to which these impacts are realized will vary between lentic (lake and pond) and lotic (river and tidally influenced) systems, freshwater and marine ecosystems, and large and small avian colonies (Jeppesen et al. 1997). Thus, to describe the influence of piscivorous colonial waterbirds in general terms is challenging.

Here, I attempt to identify the impacts of P. auritus on colonies of lentic (glacial lakes) and lotic (southern reservoirs) freshwater environments. Furthermore, I experimentally separate top-down and bottom-up factors that cormorant colonies could bring about in freshwater systems. By coupling these studies of environmental (water quality characteristics) and food web dynamics, I will demonstrate to what degree cormorant colonies of different sizes influence food webs of aquatic communities and how they might induce changes in ecosystem functioning through productivity. Below, I

describe the premise and questions behind each chapter of this dissertation that pertain to avian colonies and aquatic community structure.

Chapter II: Surveys of Cormorant Colonies in the Field and Simulated Colonies

In this chapter, I described the aquatic communities near cormorant colonies in the wild. Field samples of aquatic communities were collected from a glacial lake in

Minnesota and a reservoir lake in South Carolina. I compared the aquatic community composition of sites impacted by cormorants to reference sites that have little to no influence from cormorant foraging and defecation.

To complement the field surveys, I used experimental mesocosm ponds and a replicated 2x2 factorial design to test the major factors associated with piscivorous avian colonies that could influence community structure and functioning: predation of fishes and deposition of waste/guano. Using a pond system located on the campus of Clemson

University, I simulated the effects of cormorants on the food web by removing fish

(predation treatments, top-down forcing), applying guano fertilizer (nutrient treatments, bottom-up), a combination of these factors (predation + nutrients), and controls where neither treatment is applied. A better understanding of how colonies of waterbirds such as

P. auritus affect aquatic communities can yield more informed decisions regarding the management of aquatic resources.

The research questions I addressed in Chapter II are:

• What is the composition of food webs for the following scenarios, and how do they compare to one another? o Top-down forcing where only predation is applied? o Bottom-up forcing where only fertilization is applied? o Cormorant impacts where predation and fertilization are applied? o Control treatments where no manipulation is applied?

• What is the composition of food webs where cormorant colonies persist in the wild?

• How do cormorant-influenced food webs compare to nearby reference sites where cormorants are less common?

• How do food webs from lakes where cormorants persist compare to our mesocosm treatments?

Consequences of Movements of Double-crested Cormorants at Different Scales

Double-crested Cormorants are historically migratory (Dolbeer 1991), and the entirety of the species range, with the exceptions of subspecies in Florida, Alaska, and the

Pacific Northwest, are considered to be a single population (Waits et al. 2003). This suggests a lack of site fidelity for breeding and/or wintering sites where, for example, a cormorant might breed in New England and winter in Florida one year and then breed in

Michigan and winter in Mississippi the next year. Similarly, molecular evidence suggests that some birds may switch over a lifetime between being migratory versus year-round residents (Green et al. 2006). Over the last few decades, the landscape has been altered 12

considerably with the creation of aquatic habitat such as ponds and reservoirs, particularly in the southeastern U.S. (Smith et al. 2002, Havel et al. 2005), that can serve as foraging and breeding grounds for cormorants (Campo et al. 1993). Consequently, small colonies of birds have started to utilize these areas for breeding, without the more traditional migration to northern latitudes in summer (Post and Seals 1991).

Resident colonies of P. auritus can now be found in the interior of most southeastern states where, historically, there was little standing water available (Ellers et al. 1988, Post and Seals 1991). Cuthbert et al. (2010 unpublished data MNDNR) assert that resident colonies in the south are a re-expanding population of the non-migratory

Florida subspecies ( Phalacrocorax auritus floridanus ). Resident colonies are typically small in size and density (dozens to hundreds of individuals; Post and Seals 1991) in comparison to migratory breeding colonies (hundreds to tens of thousands; Duffy 1995) and are often protected from harassment by state wildlife managers (Personal communication, Derrell Shipes, SCDNR). This protection stems from the understanding that small numbers of cormorants may be desirable and are not likely to cause harm (e.g., to fish stocks). However, many southern resident colonies are inundated with migratory birds during non-breeding seasons (King et al. 2012), which increases damage potential

(King et al. 2010). Because we do not know how interspersed resident and migratory birds are, lethal management is not currently used in these areas because of the possibility of resident birds being removed from the system (personal communication, Derrell

Shipes, SCDNR).

The migratory ( Phalacrocorax auritus auritus ) and Florida ( P. a. floridanus ) subspecies may cue in on very different landscape features when selecting suitable nesting sites. Southern breeding birds have been documented nesting in trees of swamps

(Post and Seals 1991), whereas northern colonies typically initiate nesting on forested islands where eventually the vegetation is removed entirely through defoliation and nutrient toxicity associated with cormorant guano (Breuning-Madsen et al. 2008). To address the potential differences in nesting site preferences between P. a. auritus and P. a. floridanus , I built predictive models from historical nesting data and biogeographic

parameters that might be important for the success of a breeding colony of P. auritus such as foraging, nesting, and anthropogenic impact.

With these models, I predicted the likelihood of colonization by each subspecies in specific areas (30m x 30m) where new cormorant colonies have been reported in the last three decades (Minnesota and Florida). This non-invasive technique may be useful for southeastern wildlife managers evaluating the likelihood of the more rare Florida subspecies to occur in their area. If there are differences in habitat use by each subspecies, there may also be differences in their impacts on fish stocks and other ecosystem components. This information can be useful for initiating management programs. I have also documented the parasites of cormorants from migratory and resident colonies from multiple locations throughout the eastern and central U.S. in an effort to establish potential differences in foraging habits between the subspecies of P. auritus .

Through migratory inundation of resident colonies, one would expect mixing of parasites at most cormorant foraging sites. This prediction assumes P. auritus acquire, carry, and distribute parasites to each location where they forage and defecate. Provided there are suitable intermediate hosts (the host a cormorant must consume to acquire a parasite) available to sustain viable parasite populations, one would also expect the suite of parasites carried by any given cormorant to be similar to all other cormorants. The level of parasite similarity would add understanding to the feeding behavior and distribution of resident and migratory hosts. Differentiation of parasite assemblages could also help assess the notion that P. auritus is effectively a single population/subspecies in eastern North America and that the distinction between P. a. auritus and P. a. floridanus is strictly behavioral. I address these issues in the next three chapters of my dissertation.

Chapter III: Habitat Suitability of Two Subspecies of Double-crested Cormorant

Using the program Maxent (Phillips and Dudik 2008, Elith et al. 2011), I developed habitat species distribution models for the states of Minnesota and Florida. I then used each model to predict the distribution of cormorant colonies in South Carolina and compared the prediction output maps to known colonies and areas where cormorants do not currently nest. This is a particularly useful tool for managers interested in assessing potential areas susceptible to cormorant establishment in the future. The research questions that I address in chapter III are:

• Where are P. auritus auritus most likely to nest? o In MN – Do the predictions agree with historical nesting data? o In SC – Do the predictions agree with current colony locations?

• Where are P. a. floridanus most likely to nest? o In FL – Do the predictions agree with historical nesting data? o In SC – Do the predictions agree with current colony locations?

• What factors are associated with nesting site suitability for the two subspecies?

• Are there locations where the two subspecies are likely to overlap successfully?

Chapter IV: Parasites of Double-crested Cormorants

Two-hundred eighteen P. auritus collected from 11 sites throughout the central and eastern U.S. were assessed for intestinal parasites. Prior to identification of each parasite species recovered from host intestines, I performed a literature search to determine what parasites could infect the birds in this survey. Based on the distribution of breeding and wintering cormorants in my assessments, I assembled a list of over 30 parasites that might be recovered from intestine samples. Ultimately, I describe 15 parasites from my survey, many of which are common, but others that had not previously been reported in the United States. This information is useful for those interested in identifying potentially important locations for conservation, because larger diversities of parasites often coincide with greater diversities of intermediate host species (fishes).

Furthermore it is powerful information for those interested in the conservation of parasites (Byers 2009). The research questions that I address in Chapter IV are:

• What intestinal parasites have been reported in P. auritus ?

• What intestinal parasites can be found from frozen cormorants collected during culling activities? 16

Chapter V: Distribution of Parasite Assemblages of Double-crested Cormorants

Parasite prevalence and count data from the parasite survey described in Chapter

IV were used to determine parasite assemblages within each host intestine processed.

These data were assessed for similarities based on presence, abundance, collection year, and geographic location to determine whether similar parasite assemblages exist across geographic gradients, or between birds exhibiting differences in migration behavior. I developed predictive models which reliably outperformed null models in determining the origin or migratory status of a host at three geographic scales. The research questions I addressed in Chapter V are:

• Do the distributions of specific parasites change along spatial gradients?

• Are parasite communities similar among different geographic collection locations?

• Are parasite communities similar among different geographic regions?

• Do parasite assemblages of P. auritus differ between resident and migratory host collections?

I complete this dissertation with a brief reflection organized into a ‘Closing

Remarks’ chapter that includes perspectives on research and management needs that follow the dissertation findings. Results from this research enhance our understanding of the role that colonial waterbirds might play as agents of change in aquatic systems of

North America. This is particularly important information that conservationists should consider when developing management plans for colonial waterbird species like the

Double-crested Cormorant, their prey species of concern, and their parasites. Findings from this dissertation will provide valuable insights to local residents, state and federal agencies, and Native American nations interested in effective conservation and management strategies for cormorants and their allies.

Figures

Figure 1.1. Diagram of organism abundance based on Elton’s “Pyramid of Numbers” (Elton 1927). The base of the food web provides resources that upper trophic levels consume.

Figure 1.2. Diagram of a trophic cascade (left 4 columns) along a left-to-right gradient with increasing numbers of trophic levels where nutrient resources and Consumers influence the abundance of Producers.

Literature Cited

Abrams, P. A. 1982. Functional responses of optimal foragers. The American Naturalist 120:382-390.

Adkins, J.Y., D.D. Roby, D. Battaglia, R. Bradley, P. Capitolo, H. Carter, T. Chatwin, K. Collis, K.N. Courtot, M. Elliott, A. Evans, F. Gress, S. M. Haig, L. Harvey, J. Jahncke, P. J. Loschl, R. Lowe, D. E. Lyons, G. McChesney, D. Mercer, K. Molina, C. Moulton, M. Naughton, M. Rauzon, C. Robinson-Nilsen, W. D. Shuford, S. Stephensen, Y. Suzuki, and W. B. Tyler. 2010. A status assessment of the Double-crested Cormorant ( Phalacrocorax auritus ) in western North America: 1998-2009. 69pp. USGS – Oregon Cooperative Fish and Wildlife Research Unit, Oregon State University, Corvallis, Oregon. Also available at: http://www.birdresearchnw.org/Feature-Story/385085.aspx.

Allison, A. B., N. L. Gorrdenker, D. E. Stallknecht. 2005. Wintering of neurotropic velogenic Newcastle Disease virus and West Nile virus in double-crested cormorants ( Phalacrocorax auritus ) from the Florida Keys. Avian Diseases 49:292-297.

Armstrong, R. A. and R. McGehee. 1980. Competitive exclusion. The American Naturalist 115:151-170.

Ashmole, N. P., 1963. The regulation of numbers of tropical oceanic seabirds. Ibis 103:458–473.

Basili, G. D. and S. A. Temple. 1999. Dickcissels and crop damage in Venezuela: defining the problem with ecological models. Ecological Applications 9:732-739.

Bedard, J., A. Nadeau, and M. Lepage. 1995. Double-crested Cormorant culling in the St. Lawrence Estuary. Colonial Waterbirds 18:78-85.

Blackall, T. D., L. J. Wilson, J. Bull, M. R. Theobald, P. J. Bacon, K. C. Hamer, S. Wanless, and M. A. Sutton. 2008. Temporal variation in atmospheric ammonia concentrations above seabird colonies. Atmospheric Environment 42:6942-6950.

Boutin, C., T. Dobbie, D. Carpenter, and C. E. Hebert. 2011. Effects of Double-crested Cormorants ( Phalacrocorax auritus Less.) on island vegetation, seedbank, and soil chemistry: evaluating island restoration potential. — Restoration Ecology 19: 720-727.

Breuning-Madsen, H., C. B. Ehlers, and O. K. Borggaard. 2008. The impact of perennial cormorant colonies on soil phosphorus status. Geoderma 148:51-54.

Brook, B. W., N. S. Sodhi, and P. L. Ng. 2003. Catastrophic extinctions follow deforestation in Singapore. Nature 424:420-423.

Brooks, T., J. Tobias, and A. Balmford. 1999. Deforestation and bird extinctions in the Atlantic forest. Animal Conservation 2:211-222.

Brown, C. R. and M. B. Brown. 1987. Group-living in Cliff Swallows as an advantage in avoiding predators. Behavioral Ecology and Sociobiology 21:97-107.

Brown, C. R. and M. B. Brown. 2002. Does intercolony competition for food affect colony choice in Cliff Swallows? The Condor 104: 117-128.

Brown, C. R., B. J. Stuchbury, and P. D. Walsh. 1990. Choices of colony size in birds. Trends in Ecology and Evolution 5:398-403.

Brugger, K. E. 1995. Double-crested Cormorants and fisheries in Florida. Colonial Waterbirds 18: 110-117.

Byers, J. E. 2009. Including parasites in food webs. Trends in Parasitology 25:55–57.

Campo, J. J., B. C. Thompson, C. Barron, R. C. Telfair II, P. Durocher, and S. Gutreuter. 1993. Diet of Double-crested Cormorants wintering in Texas. Journal of Field Ornithology 64: 135-144.

Carbiener, R., M. Tremolieres, J. L. Mercier and A. Ortcheit. 1990. Aquatic macrophyte communities as bioindicators of eutrophication in calcareous oligosaprobe stream waters (Upper Rhine plain, Alsace). Vegetatio 86: 71-88.

Carson, R. 1962. Silent Spring. Mariner, Boston (2002).

Chapdelaine, G. and J. Bedard. 1995. Recent changes in the abundance and distribution of the Double-crested Cormorant in the St. Lawrence River, estuary and gulf, Quebec, 1878-1990. Colonial Waterbirds 18:70-77.

Choy, E. S., L. E. Kimpe, M.L. Mallory, J. P. Smol, and J. M. Blais. 2010. Contamination of an arctic terrestrial food web with marine-derived persistent organic pollutants transported by breeding seabirds. Environmental Pollution 158:3431-3438.

Clode, D. 1993. Colonially breeding seabirds: predators or prey. Trends in Ecology and Evolution 8:336-338.

Coleman, J. T. H., M. E. Richmond, L. G. Rudstam, and P. M. Mattison. 2005. Foraging location and site fidelity of the Double-crested Cormorant on Oneida Lake, New York. — Waterbirds 28: 498-510.

Conover, D. O., J. J. Brown and A. Ehtisham. 1997. Countergradient variation in growth of young striped bass ( Morone saxatilis ) from different latitudes. Canadian Journal of Aquatic Science 54:2401-2409.

Cott, H. B. 1953. The exploitation of wild birds for their eggs. Ibis 95:409-449.

Cowx, I. G. and D. Gerdeaux. 2004. The effects of fisheries management practices on freshwater ecosystems. Fisheries Management and Ecology 11:145-151.

Cronin, A. L., P. Federici, C. Doums, and T. Monnin. 2012. The influence of intraspecific competition on resource allocation during dependent colony foundation in a social insect. Oecologia 168: 361-369.

DeVault, T. L., R. B. Chipman, S. C. Barras, J. D. Taylor, C. P. Cranker III, E. M. Cranker and J. F. Farquhar. 2012. Reducing impacts of Double-crested Cormorants to natural resources in central New York: a review of a collaborative research, management, and monitoring program. Waterbirds 35:50-55.

Diaz, R. J. 2001. Overview of hypoxia around the world. Journal of Environmental Quality 30:276-281.

Dolbeer, R. A. 1991. Migration patterns of Double-crested Cormorants east of the Rock Mountains. Journal of Field Ornithology 62: 83-93.

Dorr, B. S., T. Aderman, P. H. Butchko, and S. C. Barras. 2010. Management effects on breeding and foraging movements of Double-crested Cormorants in the Les Cheneaux Islands, Lake Huron, Michigan. Journal of Great Lakes Research 36: 224-231.

Dorr, B. S., T. King, M.E. Tobin, J. B. Harrel, and P. L. Smith. 2004. Double-crested Cormorant movements in relation to aquaculture in eastern Mississippi and western Alabama. The International Journal of Waterbird Biology 27: 147-154.

Duffy, D. C. 1995. Why is the Double-crested Cormorant a problem? Insights from cormorant ecology and human sociology. Colonial Waterbirds 18: 25-32.

Ekman, J. and M. Hake. 1988. Avian flocking reduces starvation risk: an experimental demonstration. Behavioral Ecology and Sociobiology 22: 91-94.

Elith, J., S. J. Phillips, T. Hastie, M. Dudik, Y. E. Chee, and C. J. Yates. 2011. A statistical explanation of MaxEnt for ecologists. Diversity and Distributions 17: 43-57.

Ellers, J. M., D. H. Landers, and D. F. Brakke. 1988. Chemical and physical characteristics of lakes in the southeastern United States. Environmental Science and Technology 22: 172-177.

Elliott, K.H., K.J. Woo, A.J. Gaston, S. Benvenuti, L. dall’Antonia, G.K. Davoren. 2009. Central-place foraging in an Arctic seabird provides evidence for Storer- Ashmole’s Halo. The Auk 12:613-625.

Elton, C. 1927. Animal Ecology: Chapter V, The animal community. Sidgewick and Jackson, London.

Erwin, R. M. 1995. The ecology of cormorants: some research needs and recommendations. Colonial Waterbirds 18: 240-246.

Farquhar, J. F., I. M. Mazzocchi, R. D. McCullough, R. B. Chipman, and T. L. DeVault. 2012. Mitigation of Double-crested Cormorant impacts of Lake Ontario: from planning and practice to product delivery. Waterbirds 35:56-65.

Fenech, A. S., S. E. Lochmann, and A. A. Radomski. 2004. Seasonal diets of male and female double-crested cormorants from an oxbow lake in Arkansas, USA. Waterbirds 27: 170-176.

Forrero, M. G., J. L. Tella, K. A. Hobson, M. Bertellotti, and G. Blanco. 2002. Conspecific food competition explains variability in colony size: a test in Magellanic Penguins. Ecology 83: 3466-3475.

Fowler, W. M. W. (1965, republished 2006). Countryman's Cooking. Excellent Press, Ludlow. ISBN 1-900318-29-6.

Friend, M. and J. C. Franson. 1999. Field Manual of Wildlife Diseases. General Field Procedures and Diseases of Birds Geological Survey, Madison Wisconsin Wildlife Biology Biological Resources Division. (No. ITR-1999-001).

Frodge, J. D., G. L. Thomas, and G. B. Pauley. 1990. Effects of canopy formation by floating and submergent aquatic macrophytes on the water quality of two shallow Pacific Northwest lakes. Aquatic Biology 38: 231-248.

Gablehouse, Jr., D. W. 1984. A length-categorization system to assess fish stocks. North American Journal of Fisheries Management 4:273-285.

Gee, J. H. R., B. D. Smith, K. M. Lee, and S. W. Griffiths. 1997. The ecological basis of freshwater pond management for biodiversity. Aquatic Conservation: Marine and Freshwater Ecosystems 7: 91-104.

Gilchrist, P. 2005. Involvement of free-flying wild birds in the spread of the viruses of avian influenza, Newcastle disease and infectious bursal disease from poultry products to commercial poultry. World’s Poultry Science Journal 61: 198-214.

Glahn, J.F. and A.R. Stickley, Jr. 1995. Wintering double-crested cormorants in the delta region of Mississippi: population levels and their impact on the catfish industry. Colonial Waterbirds 18: 137-142.

Glahn, J.F., P.J. Dixson, G.A. Littauer, and R.B. McCoy. 1995. Food habits of double- crested cormorants wintering in the delta region of Mississippi. Colonial Waterbirds 18: 158-167.

Goktepe, O., P. Hundt, W. Porter, and D. Pereira. 2012. Comparing bioenergetics models of Double-crested Cormorant ( Phalacrocorax auritus ) fish consumption. Waterbirds 35: 91-102.

Grover, J. P. 1997. Resource Competition. Chapman and Hall, London.

Hatch, J. J. and D. V. Weseloh. 1999. Double-crested Cormorant ( Phalacrocorax auritus ), The Birds of North America Online (A. Poole, Ed.). Ithaca: Cornell Lab of Ornithology; Retrieved from the Birds of North America Online: http://bna.birds.cornell.edu/bna/species/441.doi:10.2173/bna.441

Havel, J. E., C. E. Lee, and M. J. vander Zanden. 2005. Do reservoirs facilitate invasions into landscapes? BioScience 55:518-525.

Hawkes, J. P. 2013. Assessing efficacy of non-lethal harassment of Double-crested Cormorants to improve Atlantic salmon smolt survival. Northeastern Naturalist 20:1-18.

Heisler, J., P. M. Glibert, J. M. Burkholder, D. M. Anderson, W. Cochlan, W. C. Dennison, Q. Dortch, C. J. Gobler, C. A. Heil, E. Humphries, A. Lewitus, R. Magnien, H. G. Marshall, K. Sellner, D. A. Stockwell, D. K. Stoecker, and M. Suddleson. 2008. Eutrophication and harmful algal blooms: a scientific consensus. Harmful Algae. 8:3-13. Also available at: http://digitalcommons.unl.edu/usepapapers/16.

Hobara, S., K. Koba, T. Osono, N. Tokuchi, A. Ishida, and K. Kameda. 2005. Nitrogen and phosphorus enrichment and balance in forests colonized by cormorants: Implications of the influence of soil absorption. Plant and Soil 268: 89-101.

Horn, H. S. 1968. The adaptive significance of colonial nesting in the Brewer’s Blackbird (Euphagus cyanocephalus ). Ecology 49:682-694.

Jackson, J. A and B. J. S. Jackson. 1995. The Double-crested Cormorant in the south- central United States: habitat and population changes of a feathered pariah. Colonial Waterbirds 18: 118-130.

Jeppesen, E., J. P. Jensen, M. Sondergaard, T. Lauridsen, L. J. Pedersen and L. Jensen. 1997. Top-down control in freshwater lakes: the role of nutrient state, submerge macrophytes and water depth. Hydrobiologia 342:151-164.

Johnson, R. J., J. A. Jedlicka, J. E. Quinn and J. R. Brandle. 2011. Global perspectives on birds in agricultural landscapes. Pages 55-140 in: Issues in agroecology – present status and future prospectus, Volume 1, Integrating Agriculture, Conservation and Ecotourism: Examples from the Field. Springer. NY.

Jokimaki, J. and J. Suhonen. 1998. Distribution and habitat selection of wintering birds in urban environments. Landscape and Urban Planning 39:253-263.

Kasprzak, P., R. Koschel, L. Krienitz, T. Gonsiorczyk, K. Anwand, U. Laude, K. Wysujack, H. Brach and T. Mehner. 2003. Reduction of nutrient loading, planktivore removal and piscivore stocking as tools in water quality management: the Feldberger Haussee biomanipulation project. Limnologica 33:190-204.

Kelly, A. 2008. A population survey and foraging analysis of the Double-crested Cormorant ( Phalacrocorax auritus ) on the Santee Lakes, South Carolina. Thesis, Clemson University, 55 pages.

Kersten, M., R. H. Britton, P. J. Dugan, and H. Hafner. 1991. Flock feeding and food intake in Little Egrets: the effects of prey distribution and behavior. Journal of Animal Ecology 60: 241-252. 26

King, D. T., B. Blackwell, and B. Dorr. 2010. Effects of aquaculture and movement patterns of Double-crested Cormorants. USDA National Wildlife Research Center – Staff Publications. Paper 925. http://digitalcommons.unl.edu/icwdm_usdanwrc/925.

King, D. T., B. K. Strickland and A. A. Radomski. 2012. Winter and summer home ranges and core use areas of Double-crested Cormorants captured near aquaculture facilities in the southeastern United States. Waterbirds 35:124-131.

Kirsch, E. M. 1995. Double-crested Cormorants along the upper Mississippi River. Colonial Waterbirds 18:131-136

Klimaszyk, P., T. Joniak, T. Sobczynski, and W. Andrzejewski. 2008. Impact of a cormorant ( Phalacrocorax carbo L.) colony on surface water quality. Overland flow a factor of nutrient transfer from colony to the lake. The Functioning and Protection of Water Ecosystems. Poznan: Department of Water Protection, Faculty of Biology, Adam Mickiewicz University.

Krohn, W.B., R.B. Allen, J.R. Moring, and A.E. Hutchinson. 1995. Double-crested cormorants in New England: population and management histories. Colonial Waterbirds 18: 99-109.

Kuiken, T., F. A. Leighton, G. Wobeser, and B. Wagner. 1999. Causes of morbidity and mortality and their effect on reproductive success in Double-crested Cormorants from Saskatchewan. Journal of Wildlife Diseases 35:331-346.

Lande, R. 1998. Anthropogenic, ecological and genetic factors in extinction and conservation. Researches on Population Ecology 40:259-269.

Letourneau, D. K. and L. A. Dyer. 1998. Experimental test in lowland tropical forest shows top-down effects through four trophic levels. Ecology 79: 1678-1687.

Livingston, R. J. 2001. Eutrophication Processes in Coastal Systems: Origin and Succession of Plankton Blooms and Effects on Secondary Production in Gulf Coast Estuaries. CRC press, Boca Raton, FL.

Mann, K. H. 1988. Production and use of detritus in various freshwater, estuarine, and coastal marine ecosystems. Limnology and Oceanography 33:910-930.

Moon, D. C. and P. Stilling. 2002. Top-down, bottom-up, or side to side? Within-trophic- level interactions modify trophic dynamics of a salt marsh herbivore. Oikos 98:480-490. 27

Nisbet, I. C. T. 1975. Selective effects of predation in a tern colony. The Condor 77:221- 226.

Overstreet, R. M., S. S. Curran, L. M. Pote, D. T. King, C. K. Blend, and W. D. Grater. 2002. Bolbophorus damnificus n. sp. (Digenea: Bolbophoridae) from the channel catfish Ictalurus punctatus and American white Pelican Pelecanus erythrorhynchos in the USA based on life-cycle and molecular data. Systematic Parasitology 52: 81-96.

Overstreet, R. M. and S. S. Curran. 2004. Defeating diplostomoid dangers in USA catfish aquaculture. Folia Parasitologica 51:153-135.

Pauly, D., V. Christensen, S. Guenette, T. J. Pitcher, U. R. Sumalia, C. J. Walters, R. Watson and D. Zeller. 2002. Towards sustainability in world fisheries. Nature 418:689-695.

Peer, B. D., J. Homan, G. M. Linz, and W. J. Bleier. 2003. Impact of Blackbird damage to sunflower: bioenergetics and economic models. Ecological Applications 13: 248-256.

Petry, P., P. B. Bayley and D. F. Markle. 2003. Relationship between fish assemblages, macrophytes and environmental gradients in the Amazon River floodplain. Journal of Fish Biology 63:547-579.

Phillips, S. J. and M. Dudik. 2008. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31:161-175.

Post, W. and C. A. Seals. 1991. Breeding biology of a newly-established Double-crested Cormorant population in South Carolina, USA. Colonial Waterbirds 14: 34-38.

Price, I. M. and J. G. Nickum. 1995. Aquaculture and birds: the context for controversy. Colonial Waterbirds 18:33-45.

Russell, I., D. Parrott, M. Ives, D. Goldsmith, S. Fox, D. Clifton-Dey, A. Prickett and T. Drew. 2008. Reducing fish losses to cormorants using artificial refuges: an experimental study. Fisheries Management and Ecology 15:189-198.

Sayre, N. F., R. R. J. McAllister, B. T. Bestelmeyer, M. Moritz and M. D. Turner. 2013. Earth stewardship of rangelands: coping with ecological, economic, and political marginality. Frontiers in Ecology and the Environment 11:348-354.

Schriver, P., J. Bogestrand, E. Jeppesen and M. Sondergaard. 1995. Impact of submerged macrophytes in fish-zooplankton-phytoplankton interactions: large-scale enclosure experiments in shallow eutrophic lake. Freshwater Biology 33:255-270.

Sheehan, K. L., K. D. Lafferty, J. O’Brien, and J. Cebrian. 2011. Parasite distribution, prevalence, and assemblages of the Grass shrimp, Palaemonetes pugio , in southwestern Alabama, U.S.A. Comparative Parasitology 78:245-256.

Sijtsma, M. T. J., J. J. Vaske, and M. H. Jacobs. 2013. Acceptability of lethal control of wildlife that damage agriculture in the Netherlands. Society and Natural Resources: an International Journal 25: 1308-1323.

Smith, S. V, W. H. Renwick, J. D. Bartley and R. W. Buddemeier. 2002. Distribution and significance of small, artificial water bodies across the United States Landscape. The Science of the Total Environment 299:21-36.

Smith, S. V. 1981. Marine macrophytes as global carbon sink. Science 211:838-840.

Tamisier, A. and C. Boudouresque. 1994. Aquatic bird populations as possible indicators of seasonal nutrient flow at Ichkeul Lake, Tunisia. Hydrobiologia 279:149-156.

Taylor II, J. D. and B. S. Dorr. 2003. Double-crested Cormorant impacts to commercial and natural resources. Proceedings, Wildl. Damage Manag. Conference. 10:43-51.

Tessier, A. J. and P. Woodruff. 2002. Cryptic trophic cascade along a gradient of lake size. Ecology 83: 1263-1270.

Thompson, B. C., J. J. Campo and R. C. Telfair, II. 1995. Origin, population attributes, and management conflict resolution for Double-crested Cormorants wintering in Texas. Colonial Waterbirds 18: 181-188.

Trapp, J. L., S. J. Lewis and D. M. Pence. 1997. Double-crested Cormorant impacts on sport fish: literature review, agency survey, and strategies. Symposium on Double-crested Cormorants: Population Status and Management Issues in the Midwest. Paper 9:87-96. Also available at: http://digitalcommons.unl.edu/nwrccormorants/

Travis, J. M. J. 2003. Climate change and habitat destruction: a deadly anthropogenic cocktail. Proceedings of the Royal Society of London. Series B: Biological Sciences 270:467-473.

Troutman, J. P., D. A. Rutherford, and W. E. Kelso. 2007. Patterns of habitat use among vegetation-dwelling littoral fishes in the Atchafalaya River Basin, Louisiana. Transactions of the American Fisheries Society 136: 1063-1075.

Vitousek, P. M., H. A. Mooney, J. Lubchenco, and J. M. Melillo. 1997. Human domination of Earth’s ecosystems. Science 277:494-499.

Waits, J. L., M. L. Avery, M. E. Tobin, and P. L. Leberg. 2003. Low mitochondrial DNA variation in Double-crested Cormorants in eastern North America. Waterbirds 26: 196-200.

White, Jr., L. 1967. The historical roots of our ecologic crisis. Science 155:1203-1207.

Wilcock, R. J. and J. W. Nagels. 2001. Effects of aquatic macrophytes on physio- chemical conditions of three contrasting lowland streams: a consequence of diffuse pollution from agriculture? Water Science and Technology 43: 163-168.

Williams, N. S. G., M. J. McDonnel, G. K. Phelan, L. D. Keim and R. van der Ree. 2006. Range expansion due to urbanization: increased food resources attract Grey- headed Flying-foxes ( Pteropus poliocephalus ) to Melbourne. Austral Ecology 31:190-198.

Winter, J., G. M. Linz, and W. Bleier. 2009. Avian use of rice-baited trays attached to cages with live decoy blackbirds in central North Dakota: research update. USDA National Wildlife Research Center - Staff Publications. Paper 933. http://digitalcommons.unl.edu/icwdm_usdanwrc/933

Withers, K. and T.S. Brooks. 2004. Diet of double-crested cormorants ( Phalacrocorax auritus ) wintering on the central Texas coast. The Southwestern Naturalist 49: 48- 53.

Wollkind, D. J. 1976. Exploitation in three trophic levels: an extension allowing intraspecies carnivore interaction. The American Naturalist 110:431-442.

CHAPTER II: COMPARISON OF AQUATIC COMMUNITIES NEAR COLONIAL WATERBIRD COLONIES AND MESOCOSM EXPERIMENTS OF SIMULATED COLONIES

Abstract: Colony forming birds have the potential to alter aquatic communities through both top-down and bottom-up forcing. Human-wildlife conflicts with Double-crested

Cormorants, Phalacrocorax auritus , focus on reduced fishery resources, suggesting predation by P. auritus is capable of damaging aquatic communities. Large colonies of P. auritus also have the potential to concentrate and redistribute nutrient resources, and in

turn, influence local productivity in aquatic systems. Here, we investigate the

communities where top-down, bottom-up and a combination of top-down and bottom-up

forcing is applied on freshwater systems. We assessed aquatic communities in lake

systems where P. auritus forage and defecate and experimentally simulated their effects

in a mesocosm system where top-down and bottom-up processes could be studied

independently and together. Nutrient subsidies associated with bird defecation had

negligible impacts on all basal and intermediate trophic groups of aquatic communities,

with the exception of an alteration of seasonal emergence for specific organism groups.

Interestingly, in South Carolina we observed a higher abundance of sport fish in areas

where P. auritus defecate but do not forage. Predation alone was tested only in our

mesocosm system and had negligible effects of aquatic community structuring. In lake

systems with P. auritus , we observed conflicting food chain patterns near nesting

colonies that suggested the impact of P. auritus are system-specific. Trophic networks of

areas where colonies forage and defecate contained fewer sport fish, but retained a high

abundance of forage fish. In our mesocosm system, the combination of nutrient addition

and predation substantially reduced fish abundance, but only once treatment intensity was increased to represent P. auritus densities that far exceed the current population size of the species.

Introduction

Population management of predatory animals is employed in situations where resource preservation is needed and the resources are shared by wildlife and humans. In some cases, management is required to maintain the ecological integrity of natural systems, including species diversity and resilience of organismal communities

(Mawdsley et al. 2009). Foraging and nesting activities of the Double-crested Cormorant,

Phalacrocorax auritus, are the basis for conflicts with humans. P. auritus is a piscivorous pursuit-diving waterbird that can consume ¾ kilogram of fish per bird each day (Kelly 2008, Goktepe 2012). Defecation of P. auritus is shown to degrade vegetative communities where they nest and roost (Blackall et al. 2008) and to damage private property and human-made structures (USFWS 2003). To reduce these conflicts and suspected negative influences of P. auritus on aquatic communities, culling programs have been established in many states (Jackson and Jackson 1995).

By consuming predatory fish, pursuit-diving birds have the potential to alter aquatic communities through trophic cascades (top-down forcing) where the abundance of upper trophic levels (top predators) directly influences the abundance of prey populations and indirectly influences the abundance of food sources that prey species rely on (Figure 2.1). Alternatively, community composition can be influenced by bottom-up

forcing (Figure 2.2), where the resources available to basal trophic levels determine the total biomass and energy available to support the entire food web (Venterink 2003). More realistically, resources such as nutrients dictate the total potential biomass that an ecosystem can sustain. Many studies have shown top-down forcing to overwhelm the effects of bottom-up forcing (Hunter and Price 1992, Letourneau and Dyer 1998), but others have demonstrated that manipulation of the upper trophic levels alone is not sufficient for meaningful changes in ecosystem services (Kasprzak et al. 2003, Eriksson et al. 2012). Consequently, top-down forcing is likely to structure food webs/chains (i.e., a 3-level food chain [Figure 2.1C] where only fishes exhibit forcing on the aquatic community versus a 4-level food chain [Figure 2.1D] where P. auritus exhibit forcing on the aquatic community), while bottom-up forcing contributes to total productivity or the movement of energy through a web.

Previous studies of P. auritus have demonstrated the influence of colonies on vegetation and terrestrial organisms on islands where breeding colonies persist (Craig et al. 2012). Other approaches have used energetics modeling to determine the potential impacts that P. auritus could have on fisheries, based on their daily caloric requirements

(Kelly 2008). While these studies are informative at local scales, the impacts to entire lake communities or community constituents beyond those measured in their analyses are difficult to interpret at a larger scale. Understanding the consequences of P. auritus colonies requires a comprehensive approach to assess aquatic communities because these birds are not just consuming fish, they are potentially influencing other organisms through top-down and bottom-up forcing. Thus, we suggest an assessment based on the 33

evaluation of the entire aquatic community. This could be accomplished by assessing complex communities in food webs (trophic networks), or more simply in food chains, which are developed from aggregations of organisms into similar trophic feeding groups.

We assessed the organismal communities of lakes where two subspecies of P.

auritus forage and nest, each in distinct habitats (Chapter V). We use these field-based

assessments in lakes to determine the patterns of community structure where P. auritus

might impact aquatic systems. To account for the natural stochasticity associated with

replicated lake systems, we also assessed the aquatic communities in an experimental

pond system, where we manipulated predation pressure and nutrient deposition to

simulate P. auritus foraging and defecation. This experimental approach also allowed us

to assess top-down and bottom-up forcing independently, which in natural systems can be

difficult to isolate. Our aquatic community assessments using data from both natural

lakes and experimental ponds were designed to help separate the impacts that this highly-

mobile predatory species can have on aquatic communities where their foraging and

roosting takes place. Clarification of the community-wide implications of P. auritus in

aquatic systems is important for development of suitable management plans.

Trophic Assessments

Variables that could illustrate the influence of top-down and bottom-up factors in

food chains include relative biomass and abundance of organisms or trophic levels. If a

trophic cascade was shaping a food chain, the combined biomass of organisms within

each respective trophic level should form a distinct pattern (Figure 2.1). We expect to

detect pattern switching of relative biomass within a food chain if a trophic cascade was influencing community composition. For example, with sufficient P. auritus foraging, we expect the relative biomass pattern of a three-level trophic community (Figure 2.1C) changing to that of a community with four trophic levels (Figure 2.1D). Alternatively, if bottom-up forcing is driving community composition, the productivity of the aquatic community is expected to increase. As such, we would expect to detect changes in the abundance and size of organisms in food chains, particularly those at the base of the food web. Because food chains simplify the true structure of aquatic communities, they may not capture changes in important components of aquatic systems. Thus, we also assess community dynamics by evaluating the topologies of trophic networks for changes in food web properties that could be associated with P. auritus colonies. Network topology characteristics are used to evaluate the efficiency of energy transfer among nodes (Kuhn et al. 2010) and the stability (Sangiovanni-Vincentelli et al. 1977) and resilience

(Ulanowicz et al. 2009) of a system. If the composition of species, species interaction, transfer of energy, or stability and resilience of a community changes as a consequence of

P. auritus predation, defecation, or both, we would expect to detect these changes in the topologies of food webs.

Methods

To assess changes in aquatic community dynamics associated with P. auritus , we used field assessments of lake communities where P. auritus breed and an experimental mesocosm system where top-down and bottom-up forcing could be simulated. In the

experimental mesocosm system, removal of organisms (fishes) represented top-down forcing (simulating P. auritus predation on fishes, Figure 2.1D), fertilization represented bottom-up forcing (simulating defecation of P. auritus into the water), and a combination of the two represented conditions where P. auritus forage and defecate.

Field Surveys

We sampled the aquatic communities of two lakes that contain nesting sites of P. auritus : Kabetogama Lake, MN, and Lake Marion, SC (Table 2.1). These nesting sites

are well-established and have been used by breeding P. auritus for many years (Steve

Windels, National Parks Service, Mary-Catherine Martin, SCDNR, personal communication). We collected field samples early (2010) and mid (2012) breeding season for Minnesota and South Carolina colonies, respectively. The Kabetogama Lake colony (Little Pine Island) was on a remote island with no standing vegetation and contained approximately 100 nests built from sticks and twigs placed on the bare ground.

The waters surrounding Little Pine Island were >20m deep. The Lake Marion colony of

P. auritus contained about 30 nests, which were constructed in a dense stand of live bald

cypress or tupelo trees where the surrounding water was approximately 6m deep.

Foraging under the nesting site in Lake Marion was unlikely because of obstruction of

open waters from standing and fallen/floating trees; however, foraging was possible at

nearby roosting and reference (control) sites where tree density and debris were less of an

obstacle for foraging and flying P. auritus . In each lake system, we sampled as close to

nesting colonies as benthic sampling would allow (Knox Island adjacent to the nesting

island on Kabetogama Lake, MN and under the nests in the Stumphole area of Lake

Marion, SC) and at one or more reference sites where P. auritus did not commonly roost or forage (Steve Windels, National Parks Service, Mary-Catherine Martin, SCDNR, personal communication). We also sampled from sites (Echo Island and Wood Duck

Island in Minnesota and a roosting site in the Stumphole area of South Carolina) where we observed foraging or roosting P. auritus on at least two occasions within the breeding season (Table 2.1).

Mesocosm System

Our experimental system was located at the Calhoun Outdoor Research

Laboratory of Clemson University in a two-acre pond previously used for catfish aquaculture research (Figure 2.3). Concrete walls separated the system into six water bodies (hereafter ponds). All other sides of each pond consisted of clay and soil. The bottom substrate of each pond was a clay, silt, and sand mixture. Each pond was partitioned lengthwise into six ‘raceways’ using heavy polyvinyl chloride sheeting. To reduce variation associated with differing experimental unit attributes (associated with the number of concrete and earthen walls), only the center four raceways within the center four ponds were used in out experiments (16 total experimental units). Each raceway measured approximately 4m across and 64m long. To avoid resampling from the same locations, we marked the partitions of each raceway so that 15 4m x 4m “sampling areas” could be referenced from a pond map (Appendix A Figure A.1). We chose areas to sample for each sampling period prior to the start of our study using random number generation; however, if an area had been sampled within the last month, it was reassigned. The average water depth in our ponds was 0.43m, making sample collection

and walking within the water easy; however, we caution that these conditions differ from the lake systems sampled, where waters were deeper.

Treatment Applications

We applied three experimental treatments to our pond system in addition to a reference treatment where no experimental manipulation was applied. To simulate bottom-up forcing that could be caused by defecating P. auritus , we developed a Nutrient

Treatment, where we distributed 293g of fertilizer (a combination of 182g Peruvian

Seabird Guano® [10%K: 10%N: 10%P] Sunleaves Garden Products, Bloomington, IN and 111g Triple Super Phosphate® [0%K: 0%N: 45%P] North Fremantle, WA) into every raceway for each application period. Fertilizer pellets were combined with water and broadcast by hand throughout each raceway. The N:P of fertilizer added represented the guano that 4,000 P. auritus /hectare were expected to excrete in a 24-hour period

(32grams of guano/bird at 3.28% nitrogen and 14.32% phosphorus, Marion et al. 1994,

Hahn et al. 2007). In the treatment representing foraging waterbirds (Predation) we removed fish from the corresponding raceways. Fish removal was not consistently successful (as is the case with foraging birds, Gremillet et al. 2006); but we used a consistent sampling effort for each application period and removed an average of 50 fish from each raceway. To apply a treatment that would represent the effects of waterbird colonies that both defecate and forage in a water body, we used a Nutrient plus Predation

Treatment (N+P) that included both added fertilizer and removal of fish.

Each treatment application occurred during one of three temporal phases: Phase I represented a low-impact, low-density waterbird colony that might occur at the beginning of the nesting season, similar to what we experienced in the lake systems; Phase II represented a high-impact, high-density waterbird colony that might occur at the end of a nesting season when adult and fledglings are hunting and defecating around nesting colonies; and the Recovery Phase represented post-colony conditions where no additional treatment was applied. These phases were timed to coincide with nesting seasons of P.

auritus (Guillaumet et al. 2011), where nesting initiation occurs in late spring (April),

fledglings forage with adults in late summer (August), and seasonal colony abandonment

occurs in the fall (September). During Phase I, we applied treatments every three weeks

from April 2 to July 30, 2012, collecting two sets of environmental samples for each

Phase I treatment: once 1-4 days before treatment application and once 1-4 days after

treatment application for a total of 11 Phase I samples for each organism type. During

Phase II, we applied two treatments each week and sampled weekly from July 31 to

September 16, 2012 for a total of 7 Phase II samples. Phase II represents a 6-fold increase

in P. auritus use of our system. We did not apply experimental treatments during the

Recovery Phase and sampled every 10 days from September 19 to November 19, 2012,

for a total of 6 Recovery samples. We began our mesocosm experiments March 28, 2012

and completed them November 19, 2012. Each set of environmental samples (25 grand

total) was considered as a “sample session,” which we used as a proxy for time in

statistical analyses.

Lake and Mesocosm Sampling

We measured the following environmental variables: water depth (cm), water temperature (°C), and dissolved oxygen (in mg/L and %) of the water with a YSI 45 oxygen meter (Yellow Springs, OH) in the mesocosm system. From each location where water parameters were measured, we collected an 800mL water sample in an acid- cleaned Nalgene ® bottle. Water samples were stored on ice until they could be filtered in the laboratory for particulate organic matter, Chlorophyll α concentrations, and

dissolved inorganic nutrients. We collected 10cm-deep sediment samples with a 17.5cm

diameter PVC benthic corer or a 17.5cmx17.5cm Ekman benthic grab (depending on

water depth; Flannagan 1970) and sieved sediments at 0.5mm. Benthic samples were

placed in Ziploc® bags and frozen for laboratory assessment where organisms were

identified, measured for length, and weighed. A 64um mini-plankton net with a 17.5cm

aperture was used to collect small organisms from the top of the water column. We

performed plankton tows over 4 to 28m transects along raceways and 100m transects in

lakes. We preserved planktonic organisms in Lugol’s iodine solution (Choi and Stoecker

1989) until they could be processed for organism identification, length, and (pooled)

weight. Vegetation was approximated on a 4m x 4m basis in mesocosms. We estimated

total percent cover of vegetation and the percent cover and average length (cm) of 5

leaves for the three most abundant vegetation species within each 4m x 4m sampling unit.

In lake systems, Minnesota and South Carolina DNR provided fish samples from

ongoing fish assessments using electrofishing boats and gill net surveys that were

performed within 4 weeks of other sample collections. Fishes within size classes suitable

for P. auritus foraging (2 – 25cm; Craven and Lev 1987, Campo et al. 1993) were frozen and sent to Clemson University for identification and measurement. In the mesocosms, we collected fishes in minnow traps fitted with 1mm window screen. Because baited traps could attract fishes from a distance, we restricted fish movement in and out of sampling areas by erecting walls of 1cm mesh around 4m x 4m sampling units. Traps were allowed to soak for 4 to 5 hours and any organism recovered from traps was measured for length and released back into the water.

Trophic Assessments

To test for evidence of bottom-up and top-down forcing in lake and mesocosm

aquatic communities, we compared average size, abundance, and biomass of each

organismal sample type (emergent vegetation, phytoplankton, zooplankton, benthic

organisms, and nekton-primarily fish) against variables that represented either treatment-

related or seasonal factors. Pairwise assessments were used to test for differences in

treatment types at an alpha level of 0.05. Any significant interactions identified in

multiple treatments or phases of study were also tested for covariance with ANCOVA.

We used JMP Pro 10.0.0 ® to perform statistical assessments. Any parameters that did

not meet parametric assumptions of normal distribution and equal variance were

transformed to meet one or both criteria. We assessed patterns of biomass for the trophic

levels documented in our mesocosm experiments using phytoplankton estimates as the

base of the food chain. We performed similar assessments using emergent vegetation as

the food chain base; however, we did not consider this to be representative of our systems

as the only organism that consumes live emergent vegetation that we were able to collect

were crayfish (Gutierrez-Yurrita et al. 1998). Thus, we used only phytoplankton for our estimates of relative biomass and abundance of organisms within the mesocosms.

Patterns of relative abundance for each lake sampling site and mesocosm treatment for each phase were compared to those expected based on the presence or absence of a second-order predator (Figure 2.1).

We used the n_w program 1 for analysis of trophic networks to determine the

trophic position of each organism collected in our field surveys. Each interaction was

assigned a binary value (0 or 1) within matrices where columns represent

consumers/predators and rows represent resources/prey. An entry of 1 in a food web

matrix indicates that the predator within the column of interest consumes the prey item of

the corresponding row (Table A.2 and reference list in Appendix A). The n_w program

computes a directed network with a number of topological characteristics for a given

food web. These characteristics include those associated with the composition of species

(the number of species [S] and the number of basal [B] intermediate [I] and top [T]

species) species interactions (the number of links between species [L], the link density

[L/S], connectivity [L/S 2], occurrence of omnivory [O], cannibalism [Loops], and organismal cycles where the starting node is the same as the end node [Cycles]), the transfer of energy (the average height of paths between basal and top species [Height], the maximum path length between basal and top species [Hmax], the average length of all directed paths within a web [Path Length], longest path among most direct

1 http://www.biologie.ens.fr/~legendre/n_w/n_w.html 42

connections between [Radius], the shortest link length between any two nodes

[Characteristic Length]), or stability and resilience (robustness of a network to perturbation [Entropy], and the node-independent network complexity [Scaled Entropy]).

Because any of the topological characteristics of a trophic network can be informative on the differences among food webs, we considered all for variable selection when building descriptive models of P. auritus -impacted food webs.

We aggregated species collected from each site and treatment by trophic similarity (criteria for aggregation in n_w) in order to reduce the number of redundant feeding groups (Sugihara et al. 1997). The two most common metrics used to describe food webs are species richness (S) and connectance (C). Martinez (1992) demonstrated that aggregation of species based on trophic similarity increased connectance as the number of species declined (when the level of aggregation exceeds ½ of the original number of species); however, directed connectance is the only characteristic of food webs that is robust to aggregation and allows for comparisons of community food webs

(Martinez 1991). Because we wanted to detect differences in species richness, it was appropriate to use connectance (the number of links per species squared) as the variable that determined the threshold for trophic web evaluation.

Some comparisons of web characteristics are based on connectance values

(Dunne et al. 2002, Ruiz-Moreno et al. 2006), while others have successfully compared communities by holding the value of connectance constant (Scotti et al. 2009). We chose to use a fixed value for connectance in order to compare the topological characteristics of

food webs within our field and mesocosm aquatic communities. Because we expected connectance to increase with aggregation, we chose a static value of connectance (0.25 ±

0.03) that was higher than our largest value of unaggregated data (0.23). The connectance threshold value we used was somewhat higher than the connectance reported in other studies (Dunne et al. 2002) and is reflective of a food web that is dominated by generalist consumers (Warren 1994). Aggregation to the point where consumers are considered as generalists allows for more coarse-grained comparisons of food webs, which could be useful for addressing fundamental differences among many food webs.

To test for underlying differences among site types and treatments, we performed a linear canonical discriminant analysis (CDA) to identify groupings of trophic webs in ordinate space. CDA is used to define the relationship of one set of variables to another by maximizing differences among qualitatively similar data sets (Moore et al. 1991,

Moore and de Ruiter 1991, Gil-Agudelo et al. 2006). Aggregation to similar functional feeding groups was a critical requirement prior to this analysis. Because we were interested in identifying similarities between our mesocosm communities and those collected from lakes where P. auritus breed, we used the CDA models built from

mesocosm communities to predict the treatment of each lake site. CDA model success

was assessed with Chi-square analysis for actual vs. predicted treatment groups within

the mesocosms. Trophic network characteristics were compared to canonical

eigenvectors for significant (p<0.05) linear interactions. We used Analysis of Variance

(ANOVA) to test for differences in treatment for all characteristic parameters included in

the final CDA model (Moore and de Ruiter 1991, Demopoulos et al. 2007).

Results

Food Chain Assessment

We found no differences in species richness, biomass, abundance, size, or trophic position for any of the trophic groups collected from our field assessments in

Kabetogama Lake or Lake Marion (Table 2.2). Furthermore, within-lake comparisons of treatments (assigned using the predictions of the CDA), yielded no differences among sites. Between-lake comparisons found that benthic organisms were larger (p=0.004), had a higher biomass (p=0.008) and trophic position (p=0.015) in Kabetogama Lake than in

Lake Marion. We also found that fishes and predatory invertebrates had higher average trophic levels in Kabetogama Lake (p=0.0031, Table 2.2). The patterns of relative biomass for the trophic groups collected from Kabetogama Lake (Figure 2.4), with the exception of Knox Island, showed higher relative biomass for predators and basal trophic levels compared to low-level consumers. These are similar to those expected in a system with a three-level food chain (Figure 2.1C), where fishes are the top predators exhibiting trophic forcing on lower trophic levels. Knox Island, in contrast, exhibited a pattern where predator (fish) biomass was suppressed (similar to Figure 2.1D) and consumer biomass was released from predation pressure. Interestingly, the pattern of trophic groups at Knox Island, exhibited a high relative biomass of primary producers even when low- level consumer biomass increased, also indicating that subsidies to primary producers

(guano) allowed for resistance to consumer pressures.

The sites where P. auritus nest and roost in South Carolina (Figure 2.5) exhibited

food chain patterns consistent with fishes exerting top-down forcing on the aquatic

community (Figure 2.1C), with no evidence to suggest P. auritus predation or guano

impacted the relative biomass of aquatic organisms. The site used as a reference had

relatively little biomass of predators or consumers with high primary producer biomass.

Using mesocosm data, univariate tests among treatments revealed significant differences of organism size, abundance, and biomass for emergent plants, phytoplankton, zooplankton, benthic organisms, and nekton (fishes, anurans, reptiles, and crayfish; Table 2.3); however, the majority of these significant changes were also seasonal. Analysis of Covariance with season included as a covariate revealed few differences among treatments when seasonality was considered (Figures 2.6 and 2.7).

Thus, we were unable to distinguish treatment effects from seasonal effects for most organism variables. Similarly, when emergent vegetation was included as a covariate in

ANCOVA, the remaining variables found to be significantly different among treatments were found to be related to the density of vegetation. The exception to this was a significant difference in the length of zooplankton, where organism length was greatest in the Nutrient Treatment (p<0.05).

The patterns of biomass for the trophic levels documented in our mesocosm experiments were similar among all treatments. The relative biomass of fish and basal organisms was lower than that documented for benthic organisms and zooplankton

(Figures 2.8 and 2.9), similar to that expected in a four-level food chain (Figure 2.1D)

Trophic Network Assessments

Trophic networks developed for the field and mesocosm assessments showed variable network structure prior to aggregation. Analysis of Variance comparing treatments of original webs revealed no significant differences among sites or treatments, and assessment of treatments over time could not be performed due to low replication numbers. Aggregation of species within food webs allowed us to increase the number of webs compared, because we could include all webs with a connectance value of 0.25 ±

0.03. This increased the number of webs from 24 to 46. Because we use varying degrees of aggregation to create webs, the same organism could fall within different trophic groups for different aggregation levels. The dominant group each type of organism was assigned to is documented for lake (Table A.3) and mesocosm (Table A.4) webs.

Field assessments of aquatic communities in Minnesota exhibited complex food webs that, even when simplified by aggregation, contained multiple feeding groups at the same trophic levels (Figure 2.10). Aquatic communities in South Carolina were simple

(Figure 2.11) and when aggregated to a connectance level of 0.25, consisted of a maximum of three trophic levels. The relatively simple food webs documented in Lake

Marion are similar in structure to those from our mesocosm assessments (Figure 2.12).

Food webs from our mesocosm study were combined for four extensive sampling events: the initial system prior to treatment applications (Initial), following 18 weeks of low-intensity treatments (Phase I), following 6 weeks of high-intensity treatments, (Phase

II), and following 9 weeks where no additional treatment was applied (Recovery). Many

food web characteristics changed in a parabolic fashion over time, but when assessed with ANCOVA with treatment and time as covariates, time was not found to be significant unless part of an interaction with treatment. Where significant treatment effects were identified, we consistently found values for each characteristic to be higher for the Predation Treatment when compared to other treatments, specifically when compared to the Nutrient and Reference Treatments.

Our CDA of mesocosm communities experiencing top-down, bottom-up, and the combination of top-down and bottom up forcing correctly placed 80.43% of food webs within their predefined treatments (p<0.0001, R-square=0.602, -2LogLikelihood = 34.06,

Table 2.4). This high success rate suggests that the characteristics of trophic networks experiencing various levels of top-down and bottom-up forcing are relatively robust.

Only three of the nine misclassified food webs were from a survey taken during experimental manipulation of the aquatic community (e.g., a Reference Treatment predicted by the model to be a Nutrient Treatment). All other misclassified food webs were sampled during the Recovery period (Table 2.4). All misclassified webs were predicted to be Nutrient webs. The final model included 14 food web characteristics

(Table 2.5).

The first canonical axis separates treatments where N+P Treatments have low species richness, a low number of links, low average and maximum species height within a web, lower average and maximum trophic level within a web, fewer cannibals and high efficiency of energy transfer within a web. Nutrient and Predation Treatments had

moderate values in terms of the aforementioned characteristics, and Reference

Treatments had relatively high web topology characteristics. The second canonical axis demonstrates trends in lower average and maximum number of trophic levels, lower frequency of cannibalism, and higher occurrence of cycles. The opposite was true for

Reference and N+P Treatments, with Nutrient Treatments exhibiting moderate values.

Separation of treatments along the third canonical axis was negatively associated with species richness, the number of links in a web, average and maximum higher of species in a web, the omnivory index, the number of cycles, and web complexity (Entropy), and was positively associated with the efficiency of energy transfer within a web. Treatments separated along axis three where Nutrient Treatments had significantly higher values than all other treatments. Similar associations for network characteristics and their correlation directions were observed for Axis 1 and 3 as well as Axis 1 and 2 (Table 2.5), so we used only Axis 2 and Axis 3 to visualize separation of points in ordinate space (Figure 2.13).

The patterns of relative biomass in the trophic networks of the mesocosms

changed seasonally. An increase in web complexity occurred at different times when

fertilized and unfertilized treatments were compared. In the Reference and Predation

Treatments, an increase in the number of feeding groups at the consumer trophic level

was captured 18-weeks into the study (Phase I). The same increase in web complexity

was observed in the N+P and Nutrient Treatments, but only after the high-intensity

treatments had been applied (6 weeks later).

Discussion

We designed this study to compare bottom-up and top-down forcing that might occur as a result of piscivorous waterbird colony use on a freshwater resource. We compared the communities of lake systems to those in our experimental system in an attempt to elucidate the trophic consequences that P. auritus causes in freshwater

systems. Our interpretations of the results incorporate the limitation that the top predators

in our surveys were fishes that, in reality, are unlikely to be top-predators within

freshwater communities. Thus, our results of the “top” trophic level are more indicative

of P. auritus prey items (2.5-12.5cm length fish; Campo et al.1993) than they are of

targets for anglers and commercial fisheries (typically >13 cm; Gablehouse, Jr. 1984).

Food Chains

Our assessments of trophic level biomass, abundance, and organism size yielded

few differences among treatments. In our mesocosm systems, treatment effects were not

distinguishable in many cases once seasonal trends were considered. The consistent

treatment effects that could be documented, despite seasonal factors, were those

associated with nutrient addition through fertilization. We saw an increase in the average

length of zooplankton, but a decrease in fish biomass in Nutrient Treatments. However,

in the field, the relative biomass patterns exhibited near the nesting colony in Minnesota

are consistent with top-down and bottom-up forcing. Top-down influence is indicated by

the lower relative biomass of fishes compared to benthic organisms and zooplankton

combined. This release from predation pressure should induce top-down forcing on basal

species by benthic organisms and zooplankton, but instead, we observe consistently high relative biomass of particulate matter and primary producers, consistent with eutrophication (Pennock et al.1995, Friedrichs et al. 2011).

In South Carolina, we observed a lack of predation pressure on fishes at the nesting and roosting sites. We expected predation of fishes by P. auritus to be negligible under the nests because tree density would impede take-off for flight (Figure 2.1C); however, we expected to see an effect of predation at the roosting site (Figure 2.1D).

Instead, the patterns of relative abundance were higher for fishes in the roost site than the reference sites, suggesting alternative variables control the abundance or movement of fishes within Lake Marion. Lake communities exhibited significantly different trophic parameters from one another (Table 2.2), suggesting that lakes, rather than within-lake community dynamics, are likely to differ because of top-down and bottom-up forcing.

Thus, using food chains, we were able to demonstrate bottom-up forcing consistently throughout our mesocosm, and to a lesser extent, the lake systems. Patterns consistent with three-level and four-level food chains were observed in nearly all communities documented in our surveys and top-down forcing (whether initiated by P. auritus or fishes) was evident in the patterns of relative biomass in the Knox Island, Stump Nest, and Stump Roost field sites.

Trophic Networks

Our trophic network assessments were used to confirm many patterns observed in

the food chain relative biomass assessments while identifying underlying differences

among the communities in our mesocosm treatments (Figure 2.10-2.12). This analysis demonstrated the effects of predation on communities through increases in community composition, species interactions, and energy transfer within a food web. The CDA developed from our mesocosm data consistently predicted the treatment of experimental webs, but because we pooled web data over time in order to achieve optimal replication, seasonal changes among treatments could not be captured with CDA.

The patterns of relative biomass of the trophic networks in the mesocosms indicated that seasonal changes in organism abundance were occurring at different rates within fertilized and unfertilized treatments. Nutrients could promote the continuation of an early-seasonal bloom of a single dominant feeding group of consumers that persisted until nutrient concentrations exceeded the ability of that group to outcompete other groups. Nutrient additions clearly influenced the timing of changes in trophic web complexity, as this seasonal shift occurred later in mesocosm treatments where fertilizers were applied (Figure 2.12). Thus, the trophic network assessments were able to capture bottom-up forcing of aquatic communities in conditions designed to simulate P. auritus colony impacts.

In locations where P. auritus colonies roost in winter or nest during summer, fecal deposition is unlikely to occur in the same habitat as foraging (Birt et al. 1987). This dissimilarity of resting and foraging habitat is suggested to be associated with the eventual depletion of food resources near colonies (a.k.a. Ashmole’s Halo, Gaston et al.

2007), and/or could be a preventative measure to reduce disease transmission through

fecal contamination of food resources (Haynes 1987). Small forage fish were abundant under roost sites (Stump Roost, Lake Marion, SC) where P. auritus forage and defecate, but larger sport fishes were decidedly absent. Although it is possible that P. auritus reduced sport fish populations through predation and this evidence supports the hypothesis of Ashmole’s halo, the diet of P. auritus is well-documented as consisting

primarily of small, forage fish (Craven and Lev 1987). Smaller fish could be less

desirable to anglers; however, smaller mesopredator fishes (such as those assessed here)

could provide a better food source for economically important sport fish species (Persson

et al. 1996, Persson et al. 2007). In turn, sport fish populations could be enhanced by

defecation of P. auritus , and provide a better resource for recreational and commercial fisheries. This supposition agrees with some intraguild predation hypotheses that suggest the addition of a top predator, such as P. auritus , which compete with existing predators such as Striped bass, will ultimately enhance the quality of top predator populations

(Wissinger and McGrady 1993, Holt and Polis 1997, Mylius et al. 2001).

Locations where moderate fish predation occurs without substantial defecation of

P. auritus are unlikely to reduce the diversity, complexity, or stability of aquatic communities. We were unable to locate foraging grounds of cormorants that did not also receive guano, but our mesocosm Predation Treatments maintained similar communities to the Reference Treatments (Figure 2.12). Although predation alone was uncommon in lake systems where P. auritus nest, we could compare communities from foraging grounds to those that P. auritus could not access. Changes in fish biomass coincided with predation in Minnesota, but not in South Carolina. Thus, the effects of P. auritus

predation should be assessed independently among systems, as these effects appear to be lake-specific. This could be because the fish assemblages differ among lakes and the appropriate size classes for consumption of P. auritus are more abundant in Minnesota.

Higher growth rates of fishes have been documented in higher latitudes (Conover et al.

1997), thus, in addition to losses of small fish to predation, growth of fishes out of the size class suitable for cormorant consumption are also possible in Minnesota.

Alternatively, the difference in predation effect between lakes could indicate a more productive fish assemblage in the warmer waters of South Carolina (Carpenter et al.

1992). We observed high relative biomass of fishes at both lake sites, indicating that even in conditions where P. auritus reduce fish biomass, refuges remain that can maintain relative high fish abundance. Wetlands provide suitable refuge for small fishes and refuges impede the foraging success of P. auritus (Russell et al. 2008).

The combination of predation and nutrient addition resulted in the smallest body sizes and lowest abundance of fishes in our mesocosms, but only after 6 weeks of high- intensity treatments, representing 4,000 birds/hectare (Phase II), which if scaled to the size of Kabetogama Lake, would be over 11 million cormorants, 4 times the current P. auritus population (Wires et al. 2001). We expect moderate levels of predation and guano deposition would be similar to those of fertilizer-only conditions or N+P Treatments during Phase I. Thus, we only see negative impacts of P. auritus when their densities are extremely high. It appears unlikely for natural concentrations of P. auritus to have negative impacts on aquatic communities in freshwater systems.

Tables Table 2.1. Sample sites on lakes used to examine trophic structure. Nesting colonies of P. auritus occur on Knox Island and at Stump Nests and a roosting colony at Stump Roost.

Other sites served as reference where P. auritus did not commonly roost or forage.

Lake Site Treatment Latitude Longitude Kabetogama, MN Knox Island N + P 48.4486° 92.9305° Blind Ash Bay East Reference 48.4304° 92.8688° Blind Ash bay West Reference 48.4346° 92.8756° Echo Island Reference 48.4727° 93.0614° Wood Duck Island Reference 48.4916° 93.0528° Marion, SC Stump Nests Nutrient 33.5900° 80.5182° Stump Roost N + P 33.5867° 80.5246° Stump Control Reference 33.5957° 80.5328°

Table 2.2. Univariate analyses of trophic variables measured for organisms (Org. Type) in aquatic food chains of two lakes where P. auritus nest 2.

Kabetogama, MN Lake Marion, SC Lakes Org. Type Variable p-value R-sq. p-value R-sq. p-value R-sq.

Biomass 0.1344 0.580 N/A N/A 0.2420 0.219 Phyto

Length 0.8224 0.001 0.6680 0.126 0.1272 0.051 Trophic Position 0.8511 0.001 N/A N/A 0.2102 0.035 Richness 0.6358 0.054 0.7911 0.133 0.5328 0.051 Biomass 0.9592 <0.001 0.8802 0.042 0.8083 0.001

Zooplankton Abundance 0.3407 0.025 0.8276 0.061 0.1247 0.052

Length 0.1257 0.052 0.6264 0.110 0.0040 0.139 K Trophic Position 0.8571 <0.001 0.7903 0.057 0.0147 0.103 K Richness 0.5910 0.049 0.8717 0.167 0.0664 0.085 Biomass 0.1509 0.046 0.7007 0.085 0.0082 0.118 K Benthic Abundance 0.9367 <0.001 0.5120 0.154 0.6600 0.003

Length 0.4831 0.010 0.1216 0.410 0.5187 0.007 Trophic Position 0.8791 <0.001 0.1628 0.365 0.0031 0.137 K Richness 0.7722 0.051 0.3695 0.037 0.0161 0.337 K

Nekton Biomass 0.9162 <0.001 0.3353 0.024 0.6747 0.003 Abundance 0.3802 0.016 0.4314 0.190 0.1992 0.027

2 Within-lake assessments were compared using analysis of variance. Between-lake differences were identified using Student’s t-test and the lake with the higher value of a parameter is indicated (K=Kabetogama Lake). 56

Table 2.3. Comparison of variables combined among sample types in aquatic communities 3.

Treatment Nutrients Predation Date Test p R-sq. Output Test p R-sq. Output Test p R-sq. Test p R-sq. Output

Biomass A 0.2937 0.028 T 0.0587 0.026 T 0.7697 0.001 LR 0.0012 0.000 − Phyto

Length A <0.0001 0.154 N+P > R > P T <0.0001 0.129 + T 0.7666 0.001 PR 0.0013 0.086 − Richness A 0.0768 0.013 T 0.0243 0.023 − T 0.6462 0.001 LR 0.0003 0.018 − Biomass A <0.0001 0.110 N > P > R T <0.0001 0.153 + T 0.4235 0.003 LR 0.1221 0.012 Em. Plants Plants Em. Abundance A <0.0001 0.101 N > P > R T 0.0003 0.058 − T 0.3883 0.003 LR 0.5194 0.002

Length nT 0.0455 0.055 N > N+P T 0.7218 0.001 T 0.2242 0.010 PR 0.0085 0.064 − Trophic Position A 0.4664 0.018 nT 0.0493 0.008 − T 0.9364 0.000 PR 0.0195 0.053 − Richness A 0.8646 0.005 T 0.5357 0.003 T 0.6004 0.002 LR 0.0050 0.051 − Biomass A 0.5849 0.013 nT 0.0021 0.096 − T 0.9459 0.000 LR 0.0002 0.093 − Zooplankton Zooplankton Abundance A 0.9246 0.003 nT 0.0007 0.113 − T 0.6536 0.001 LR <0.0001 0.132 −

Length nT 0.1565 0.026 N+P > R T 0.0614 0.018 − T 0.2530 0.007 LR 0.0019 0.048 − Trophic Position A 0.7253 0.007 T 0.8921 0.001 T 0.9445 0.001 LR 0.1718 0.000 Richness A 0.1576 0.026 T 0.6351 0.001 T 0.3632 0.001 LR 0.1057 0.013 Benthic Benthic Biomass A 0.1263 0.029 T 0.0298 0.024 + T 0.8795 0.000 LR <0.0001 0.161 + Abundance A 0.1111 0.030 T 0.0168 0.029 + T 0.8619 0.000 LR <0.0001 0.179 +

Length A 0.3785 0.019 T 0.2427 0.009 T 0.1992 0.010 LR 0.0014 0.062 − Trophic Position A 0.1108 0.034 T 0.1314 0.013 T 0.0615 0.020 PR <0.0001 0.254 + Richness A 0.4335 0.016 T 0.6417 0.001 T 0.5459 0.002 LR 0.0055 0.044 − Nekton Nekton Biomass A 0.0497 0.049 RP > N T 0.0085 0.001 − T 0.7438 0.001 PR <0.0001 0.120 − Abundance A 0.0566 0.043 T 0.0109 0.037 − T 0.2678 0.007 PR <0.0001 0.223 −

3 Analysis of Variance (A) was used to compare treatments (R=reference, N=nutrient-only, P=predation-only, N+P=nutrient and predation). Student’s T-test (T) compared samples among fertilized and unfertilized treatments, and predation to non-predation treatments. If needed, nested t-tests were used at an alpha value of 0.05 (nT). Sample session was used as a proxy for date, and linear or quadratic regression was used to test for changes over time. 57

Table 2.4. Contingency table of output values from the Canonical Discriminant Analysis corresponding with the classifications of food webs based on network topological characteristics.

Predicted Treatment Sample Webs R N + P N P R 16 0.69 1 0 0.31 0 N+P 10 0 0.8 2 0.2 0 N 7 0 0 13 0 P 13 0 0 0.15 0.85 4 Blind Ash E 1 1 0 0 0 Blind Ash W 5 0.80 0.20 0 0 Echo Island 4 1 0 0 0 Knox Island 7 0.71 0 0 0.29 Wood Duck I 3 1 0 0 0 Stump C 1 0 0 1 0 Stump N 5 0.20 0.80 0 0 Stump R 1 0 1 0 0

1 The proportion of reference webs that were correctly classified by the CDA model. 2 The proportion of webs from treatments with nutrients added and fish removed that were correctly classified by the CDA model. 3 All webs from treatments with nutrients added were correctly classified by the CDA model. 4 The proportions of webs from predation treatments that were correctly classified by the CDA model. 58

Table 2.5. Correlations of the eigenvectors for the three significant axes describing the separation of mesocosm food webs based on treatments.

Canonical Variable Description Axis Sig. Diff. 1 1 2 3 S Number of species – – P > N* L Number of links – – P > CN* Conn Number of connected components Height 2 Average height of non-basal species – – P > N+P NC† Hmax Maximum height of all species within a web – – P > CN† TroLev Average trophic level of all species within a web – – TLmax Maximum trophic level of all species within a web – – OI Omnivory index – P > C* Loops Number of cannibalistic species – – Cycles Number of feedback cycles (start=finish node) + – P > N+P NC* CycLen Average length of cycles within a web P > NC† Radius 3 Minimum distance between connected species (directed) Minimum distance between connected species CharLen 4 (undirected) – + Entropy Resilience of a web to withstand perturbation †5 –

1 Significant positive (+) and negative (–) relationships are indicated with axes and the direction/magnitude (>) of significant differences among treatments. 2 Represents the quantity of energy that is transferred to an organism and food web and the height of a web is the average trophic height of all consumer (non-basal) species in a trophic network. 3 The average difference in height among species, effectively capturing the most efficient transfer of energy considering the species interactions within a trophic web. 4 The maximum possible efficiency for the transfer of energy through a network. 5 * Indicates significant associations for treatment, time, and a treatment x time interaction; †indicates significant interaction between treatment and time.

Figures

Figure 2.1. Trophic cascade as presented by Holtz et al. (2000).

Figure 2.2. Eltonian pyramid revised to represent biomass estimates like Sukhdeo (2010) based on data from Lindeman (1942).

Figure 2.3. Aerial photograph of the Bottoms region of Clemson University 1.

1 Aerial photograph provided by Dr. Christopher Post, Clemson University. 62

1600 basal 1400 ben/zoop

1200 fish

1000

800

600

400 Biomass g/1000L Biomass

200

0 Blind Ash E Bline Ash W Echo Island Knox Island Wood Duck I Field Site (MN)

Figure 2.4. Relative biomass estimates of organisms collected from field sites in Minnesota 2.

2 Knox Island is the closest site to the nesting colony of P. auritus on Kabetogama Lake, MN. 63

1000 basal 900 ben/zoop 800 fish 700 600 500 400 300

Biomass g/1000L Biomass 200 100 0 Stump Nest Stump Roost Stump Control Field Site (SC)

Figure 2.5. Relative biomass estimates of organisms collected from field sites in South Carolina 3.

3 The reference site (Stump Control) was upstream of nesting (Stump Nest) and roosting sites (Stump Roost) sites from Lake Marion, SC. 64

Figure 2.6. Biplots of significant ANCOVA models for microconsumers and zooplankton over time 1 from mesocosms.

Sample sessions represent samples taken through time from March 29 to November 19, 2012.

1 Separate slopes are drawn for each treatment where R=reference [orange], N=nutrients added [red], P=fish removed [blue], and N+P=nutrients added and fish removed [green]. Significant linear regressions for treatments are displayed under descriptive statistics of the full model.

Figure 2.7. Biplots of significant ANCOVA models for benthic organisms over time 1.

1 Separate slopes are drawn for each treatment where R=reference, N=nutrients added, P=fish removed, and N+P=nutrients added and fish removed. Significant linear regressions for treatments are displayed under descriptive statistics of the full model. 66

25 basal ben/zoop 20 fish

5 Biomass (g/1000L) Biomass

0 ImpactN + P Nutrient Predation Reference Mesocosm Treatment

Figure 2.8. Relative estimates of biomass for the three trophic levels within mesocosm assessments with phytoplankton estimates for the base of the food chain (basal) 1.

1 Benthic organisms and zooplankton are considered as first-order consumers (ben/zoop), and nekton collected in minnow traps as second-order consumers (primarily fish). 67

200 40,000+ 30,000+ 55,000+ 110,000+ 180 basal 160 ben/zoop 140 fish 120 100 80 60 40 20 Abundance (Number/1000L) Abundance 0 ImpactN + P Nutrient Predation Reference Mesocosm Treatment

Figure 2.9. Relative abundance of trophic levels using only phytoplankton estimates as the base of the trophic chain (basal) 1.

1 Benthic organisms and zooplankton are considered as first-order consumers, and nekton collected in minnow traps as second- order consumers (fish). 68

Figure 2.10. Food webs depicting the relationships among trophic groups from Kabetogama Lake 1.

1 Knox Island is the closest site to the P. auritus nesting colony, with all other sampling sites considered as a reference. Green nodes represent basal species, blue nodes represent consumers, and yellow nodes represent predators. 69

Figure 2.11. Food webs depicting the relationships among trophic groups from the Stumphole area of Lake Marion.

Figure 2.12. Food web topologies from all treatments in mesocosms for four major sampling sessions1

1 Initial = no prior treatment applications, Phase I = after 18 weeks of low-intensity predation and nutrient applications, Phase II = following 6 weeks of high-impact predation and nutrient applications, Recovery = following 9 weeks with no additional treatments applied. Green nodes represent basal species, blue nodes represent consumers, yellow nodes represent predators. 71

Figure 2.13. Canonical plot of mesocosm food webs in ordinate space 2.

2 Canonical axis 3 is positively associated with network connectivity, maximum path height, average trophic level, average directed path length, network radius and average undirected path length. Canonical axis 2 is negatively associated with the number of basal species, average directed path length, network entropy and scaled entropy and negatively associated with average trophic level. 72

Acknowledgements

This study was supported by the Clemson Creative Inquiry program. The funding, student participants, and interns that assisted with the completion of sample collection and processing were critical to the project. We thank Scott Davis at the Calhoun Field

Research facility for maintaining the mesocosm system water levels and access to its waters. Numerous undergraduate researchers enrolled in the Aquatic Communities

Assessment and Analysis Creative Inquiry course were instrumental in the set-up and completion of this research, and to all those that contributed we communicate our deepest appreciation. We express thanks to Dr. John Rodgers and his graduate students who allowed us to use their fluorometric equipment and chlorophyll standards for primary producer assessments of the water column. Water filtration efficiency increased substantially with the use of Clemson University’s microbiological facilities. We would also like to thank the friendly reviewers of this manuscript who improved delivery of its message and readability.

Literature Cited

Birt. V., T. P. Birt, D. Goulet, D. K. Cairns, and W. A. Montevecchi. 1987. Ashmole’s halo: direct evidence for prey depletion by a seabird. Marine Ecology Progress Series 40:205-208.

Blackall, T.D., L.J. Wilson, J. Bull, M.R. Theobald, P.J. Bacon, K.C. Hamer, S. Wanless, and M.A. Sutton. 2008. Temporal variation in atmospheric ammonia concentrations above seabird colonies. Atmospheric Environment 42:6942-6950.

Campo, J.J., B.C. Thompson, C. Barron, R.C. Telfair II, P. Durocher, and S. Gutreuter. 1993. Diet of Double-crested Cormorants wintering in Texas. Journal of Field Ornithology 64: 135-144.

Carpenter, S. R., S. G. Fisher, N. B. Grimm and J. F. Kitchell. 1992. Global change and freshwater ecosystems. Annual Review of Ecology and Systematics 23:119-139.

Choi, J. W. and D. K. Stoecker. 1989. Effects of fixation on cell volume of marine planktonic protozoa. Applied and Enviroonmental Microbiology 55:1761-1765.

Craig, E. C., S. B. Elbin, J. A. Danoff-Burg and M. I. Palmer. 2012. Impacts of Double- crested Cormorants ( Phalacrocorax auritus ) and other colonial waterbirds on plant and arthropod communities o islands in an urban estuary. Waterbirds 35:4- 12.

Demopoulos, A. W. J., B. Fry and C. R. Smith. 2007. Food web structure in exotic and native mangroves: a Hawaii-Puerto Rico comparison. Oecologia 153:675-686.

Dunne, J. A., R. J. Williams, and N. D. Martinez. 2002. Food-web structure and network theory: the role of connectance and size. Proceedings of the National Academy of Sciences of the United States of America 99:12917 – 12922.

Elton, C. 1927. Animal Ecology: Chapter V, The animal community. Sidgewick and Jackson, London.

Eriksson, B. K., A. Rubach, J. Batsleer and H. Hillebrant. 2012. Cascading predator control interacts with productivity to determine the trophic level of biomass accumulation in a benthic food web. Ecol. Res. 27: 203-210.

Flannagan, J. F. 1970. Efficiencies of various grabs and corers in sampling freshwater benthos. Journal of the Fisheries Board of Canada 27:1691-1700.

Friedrichs, S. J., K. D. Zimmer, B. R. Herwig, M. A. Hanson, and J. R. Fieberg. 2011. Total phosphorus and piscivore mass as drivers of food web characteristics in shallow lakes. Oikos 120:756-765.

Gablehouse, Jr., D. W. 1984. A length-categorization system to assess fish stocks. North American Journal of Fisheries Management 4:273-285.

Gaston, A. J., R. C. Ydenberg, and G. E. J. Smith. 2007. Ashmole’s halo and population regulation in seabirds. Marine Ornithology 35:119 – 126.

Gil-Agudelo, C. Myers, G. W. Smith and K. Kim. 2006. Changes in the microbial communities associated with Gorgonia ventalina during aspergillosis infection. Diseases of Aquatic Organisms 69:89-94.

Goktepe, O., P. Hundt, W. Porter, and D. Pereira. 2012. Comparing bioenergetics models of Double-crested Cormorant ( Phalacrocorax auritus ) fish consumption. Waterbirds 35: 91-102.

Gremillet, D., M. R. Enstipp, M. Boudiffa, and H. Liu. 2006. Do cormorants injure fish without eating them? An underwater video study. Marine Biology 148:1081 - 1087.

Guillaumet, A., B. Dorr, G. Wang, J. D. Taylor, II, R. B. Chipman, H. Scherr, J. Bowman, K. F. Abraham, T. J. Doyle, and E. Cranker. 2011. Determinants of local and migratory movements of Great Lakes Double-crested Cormorants. Behavioral Ecology 3:1096 - 1103.

Gutierrez-Yurrita, P. J., G. Sancho, M. A. Bravo, A. Baltanas, and C. Montes. 1998. Diet of the red swamp crayfish Procambarus clarkia in natural ecosystems of the Donana National Park temporary fresh-water marsh (Spain). Journal of Crustacean Biology 18:120-127.

Hahn, S., S. Bauer, and M. Klaassen. 2007. Estimating the contribution of carnivorous waterbirds to nutrient loading in freshwater habitats. Freshwater Biology 52:2421 - 2433. 75

Haynes, A. M. 1987. Human exploitation of seabirds in Jamaica. Biological Conservation 41:99 - 124.

Holt, R. D. and G. A. Polis. 1997. A theoretical framework for intraguild predation. The American Naturalist 149:745 - 764.

Holz, J. C., K. D. Hoagland, and A. Joern. 2000. Aquatic food web interactions: Microcosms as Lake Models. Pages 305-323, in Tested studies for laboratory teaching, Volume 21 (S. J. Karcher, Editor). Proceedings of the 21st Workshop/Conference of the Association for Biology Laboratory Education (ABLE), 509 pages. http://www.ableweb.org/volumes/vol-21/15-holz.pdf.

Hunter, M. D. and P. W. Price. 1992. Playing chutes and ladders: heterogeneity and the relative roles of bottom-up and top-down forces in natural communities. Ecology 73: 724 – 732.

Jackson, J. A and B. J. S. Jackson. 1995. The Double-crested Cormorant in the south- central United States: habitat and population changes of a feathered pariah. Colonial Waterbirds 18: 118-130.

Kasprzak, P., R. Koschel, L. Krienitz, T. Gonsiorczyk, K. Anwand, U. Laude, K. Wysujack, H. Brach, and T. Mehner. 2003. Reduction of nutrient loading, planktivore removal and piscivore stocking as tools in water quality management: The Feldberger Haussee biomanipulation project. Limnologica 33: 190-204.

Kelly, A. 2008. A population survey and foraging analysis of the Double-crested Cormorant ( Phalacrocorax auritus ) on the Santee Lakes, South Carolina. Thesis, Clemson University, 55 pages.

Kuhn, F., N. Lynch and R. Oshman. 2010. Distributed composition in dynamic networks. In Proceedings of the 42nd ACM symposium on Theory of computing. ACM, New York, NY, USA, 513-522.

Lindeman, R. L. 1942. The trophic dynamic aspect of ecology. Ecology 23: 399-418.

Letourneau, D. K. and L. A. Dyer. 1998. Experimental test in lowland tropical forest shows top-down effects through four trophic levels. Ecology 79:1678-1687.

Marion, L., P. Clergeau, L. Brient, and G. Bertu. 1994. The importance of avian- contributed nitrogen (N) and phosphorus (P) to Lake Grand-Lieu, France. Hydrobiologia 279:133 – 147.

Martinez, N. D. 1991. Artifacts or attributes? Effects of resolution on the Little Rock Lake food web. Ecological Monographs 61:367-392.

Martinez, N. D. 1992. Constant connectance in community food webs. The American Naturalist 139:1208-1218.

Mawdsley, J. R., R. O’Malley and D. S. Ojima. 2009. A review of climate-change adaptation strategies for wildlife management and biodiversity conservation. Conservation Biology 23:1080-1089.

Moore, J. C. and P. D. de Ruiter. 1991. Temporal and spatial heterogeneity of trophic interactions within below-ground food webs. Agriculture, Ecosystems and Environment 34:371-397.

Mylius, S. D., K. Klumpers, A. M. de Roos and L. Persson. 2001. Impact of intraguild predation and stage structure on simple communities along a productivity gradient. The American Naturalist 158:259-276.

Pascual, M., and Dunne, J. A. (Eds.). 2005. Ecological Networks: Linking Structure to Dynamics in Food Webs. Oxford University Press.

Pennock, J. R., J. H. Sharp, and W. W. Schroeder. 1995. What controls the expression of estuarine eutrophication? In: Dyer K.R., R.J. Orth (Eds). Changes in Fluxes in Estuaries. Olsen & Olsen, Fredensborg. pp. 139-146.

Persson, L., J. Andersson, E. Wahlstrom and P. Eklov. 1996. Size-specific interactions in lake systems: predator gape limitation and prey growth rate and mortality. Ecology 77:900-911.

Persson, L., P. A. Amundsen, A. M. De Roos, A. Klemetsen, R. Knudsen, and R. Primicerio. 2007. Culling prey promotes predator recovery-alternative states in a whole-lake experiment. Science 316:1743-1746.

Ruiz-Moreno, D., M. Pascual, and R. Riolo. 2006. Exploring network space with genetic algorithms: modularity, resilience and reactivity. Ecological Networks: Linking Structure to Dynamics in Food Webs. Oxford University Press. 187-208.

Sangiovanni-Vincentelli, A., L. Chen and L. O. Chua. 1977. An efficient heuristic cluster algorithm for tearing large-scale networks. IEEE Transactions on Circuits and Systems 12:709-717.

Scotti, M., C. Bondavalli, A. Bodini and S. Allesina. 2009. Using trophic hierarchy to understand food web structure. Oikos 118:1695-1702.

Sugihara, G., L. F. Bersier and K. Schoenly. 1997. Effects of taxonomic and trophic aggregation on food web properties. Oecologia 112:272-284.

Sukhdeo, M.V.K. 2010. Food webs for parasitologists: a review. Journal of Parasitology 96: 273-284.

Ulanowicz, R. E., S. J. Goerner, B. Lietaer and R. Gomez. 2009. Quantifying sustainability: resilience, efficiency and the return of information theory. Ecological Complexity 6:27-36.

USFWS. 2003. Final Environmental Impact Statement: Double-Crested Cormorant Management. US Department of Interior Fish and Wildlife Service in cooperation with WS Department of Agriculture APHIS Wildlife Services. Div. of Migratory Bird Management, Arlington, VA.

Venterink, H. O., M. J. Wassen, A. W. M. Verkroost, and P. C. de Ruiter. 2003. Species richness-productivity patterns differ between N-, P-, and K-limited wetlands. Ecology 84:2191-2199.

Warren, P. H. 1994. Making connections in food webs. Trends in Ecology and Evolution 9:136-141.

Wissinger, S. and J. McGrady. 1993. Intraguild predation and competition between larval dragonflies: direct and indirect effects of shared prey. Ecology 74:207 – 218.

2. CHAPTER III: NESTING HABITAT SUITABILITY OF TWO DOUBLE- CRESTED CORMORANT SUBSPECIES (PHALACROCORAX AURITUS AURITUS AND P. A. FLORIDANUS )

Abstract: Developing effective management plans and conservation initiatives for

similar subspecies requires an understanding of differences in their ecology and

geographic distributions. Two subspecies of the Double-crested Cormorant

(Phalacrocorax auritus ) occur in South Carolina, but molecular evidence for separation

of the subspecies is lacking. Instead, migration to northern nesting sites differentiates

migratory P. a. auritus from resident P. a. floridanus . Recent population declines and

recoveries have altered the subspecies distributions and there is uncertainty whether birds

breeding in South Carolina are P. a. auritus or P. a. floridanus . We use Maxent to

develop species distribution models and to compare the habitats used in South Carolina to

nesting habitat characteristics from the historical breeding ranges of P. a. floridanus

(Florida) and P. a. auritus (Minnesota). The nesting habitat in South Carolina more

closely resembles the habitat characteristics associated with P. a. floridanus. Our findings

for habitat differences between these two subspecies could be used by managers in

refining management strategies for human conflicts with overabundant P. a. auritus and,

at the same time, conservation initiatives needed for P. a. floridanus .

Introduction

Subspecies are commonly defined by the isolation of geographic distributions of organisms within the same species (Rand and Traylor 1950, Mallet 1995, Haig et al.

2006). Other factors, such as behavior, can also lead to differentiation of subspecies

(Mayr 1982, Stanford and Nkurunungi 2003). Behavioral traits such as habitat use and foraging preference, in turn, may influence management imperatives (Carranza and Winn

1954, Fonteneau et al. 2009). In South Carolina, resource managers are tasked to develop management strategies for different subspecies of the Double-crested Cormorant,

Phalacrocorax auritus (personal communication, Derrell Shipes, SCDNR). Following population bottlenecks when cormorant abundance in North America declined from millions to only a few thousand birds (Wires and Cuthbert 2006, Wild 2012), geographic distributions of the migratory ( P. a. auritus ) and resident ( P. a. floridanus ) subspecies during the breeding season were well defined and non-overlapping (Brugger 1995, Hatch

1995). Although P. auritus populations are now considered to be recovered (Hatch 1995,

Wires et al. 2001), it is unclear whether the contemporary breeding colonies in South

Carolina belong to P. a. floridanus or to P. a. auritus because the breeding ranges of both subspecies have expanded around human-created water bodies such as those developed in the 1950s in South Carolina. Migratory birds that winter in South Carolina are the target of nuisance wildlife control, whereas resident birds that nest in the state in summer are of conservation concern (personal communication, Derrell Shipes, SCDNR, DNR News

11/20/2013). Using information on the known geographic distributions of the two subspecies, we seek to identify the origin of the colonies now nesting in South Carolina.

There are currently four suggested subspecies of Phalacrocorax auritus (Hatch

1995, Waits et al. 2003); however, there is no molecular evidence to suggest separation of subspecies of central and eastern North America (Waits et al. 2003, Green et al. 2006, and Mercer 2008). Mercer (2008) provides molecular evidence to support the separation of a Pacific subspecies ( P. a. albociliatus ) and an Alaskan subspecies ( P. a. cincinatus ), but their results agree with Green and others (2006) who argue that migratory and resident birds in the central and eastern U.S. are, in fact, a single subspecies. The differentiation between the migratory birds in the interior and eastern regions of North

America ( P. a. auritus ) and the non-migratory birds that occur in the southeastern United

States ( P. a. floridanus ) is behavioral (Green et al. 2006). Migration behavior is influenced by climatic, biological, and anthropogenic factors (Hutto, 1985, Walther et al.

2002) and different subspecies are likely to respond to these variables in distinct ways.

Using the environmental characteristics of the known nesting sites of cormorants, we can develop ecological niche models to describe the habitat of P. a. auritus and P. a. floridanus during the breeding season.

Species distribution models (SDM) are useful tools that allow users to predict current distributions of rare or cryptic species (Raxworthy et al. 2003, Engler et al. 2004), potential distributions of invasive species (Peterson et al. 2003, Young et al. 2013), and future distributions of organisms in relation to climate change (Thomas et al. 2004). SDM can be used to develop ecological-niche models where response data can be presence- only, presence-absence, or count data. The type of input data can restrict the statistical assessment used to develop predictive models (Elith et al. 2011, Aarts et al. 2012, Hastie

and Fithian 2013). Using presence-absence and count data limits the number of observations included in models to those where observational data are available (Phillips and Dudik 2008, Van Couwenberghe et al. 2013). Alternatively, presence-only data can increase sampling robustness by assuming all locations not listed as presence points are absence points (Phillips and Dudik 2008, Elith et al. 2011). This can be problematic for species for which observations of occupancy are unreliable (Guisan and Thuiller 2005), but could be informative as to what conditions exist at presence sites that do not occur elsewhere. Maximum entropy (Maxent) is an increasingly popular method for predicting the geographical distributions of organisms based on presence-only data (Phillips and

Dudik 2008, Gormley et al. 2011, Evans et al. 2010). Applicability and methodology associated with Maxent are heavily documented (Peterson et al. 2007, Phillips and Dudik

2008, Oppel et al. 2012, Renner and Warton 2013, Merow et al. 2013).

Here, we build species distribution models to investigate three suites of environmental variables expected to be important for successful breeding of P. auritus .

Breeding waterbird colonies are relatively conspicuous and there is a low likelihood of missed detection (Ridgway 2010), thus, presence-only models are suitable for modeling

P. auritus nesting habitat distribution. We compare important habitat variables of contemporary breeding colonies of P. auritus within the states of Minnesota and Florida using presence and absence data and predict nesting habitat in South Carolina. Although absence data were not used to create our models, we used known absence points to validate model predictions. Count data for nesting sites was also used to assess the prediction values for correlations with colony size.

Methods

Nesting Colony Data

Long-term nesting surveys of P. auritus have been conducted throughout the geographic distribution of the species (Craven and Lev 1987, Nisbet et al. 2002,

Anderson et al. 2004). We developed nesting habitat models using data from Minnesota for the migratory subspecies ( P. a. auritus ) and data from Florida for the resident subspecies ( P. a. floridanus ). Nest site count data were based on waterbird surveys from

1977 through 2010 for Minnesota from the USDA/APHIS Wildlife Services and

MNDNR/University of Minnesota (Brian Dorr, unpublished data, Cuthbert et al. 2005,

Wires et al. 2010). The 2004 and 2010 nest surveys were the most comprehensive, although all years were used to calculate the mean colony density for each nesting site in

Minnesota. We acquired nesting data for Florida from the Florida Fish and Wildlife

Conservation Commission (Nisbet et al. 2002). These data contained three statewide waterbird surveys performed within three decades (1970’s, 1980’s and 1990’s). Data for nesting sites in South Carolina were based on colonies documented by SCDNR in 2011 and 2012 (unpublished data, Christy Hand, SCDNR) and by publications reporting contemporary nesting locations (Post and Seals 1991). All count data were converted to presence only for Maxent model creation and presence/absence for model validation.

Layer Development for Individual Parameters

The attributes of P. auritus nesting locations were developed under three different criteria: foraging habitat, nesting habitat, and anthropocentric parameters. These variables

were derived from data layers obtained through publicly available web downloads including the National Atlas 3, National Land Cover Database (NLCD) 4, National

Wetlands Inventory 5, and the National Hydrography Dataset (NHD) 6 (Appendix B Table

B.1). Fish consumption advisories were obtained from the Environmental Protection

Agency 7. Fish stocking activity data were obtained from state fisheries agencies. Climate

variable data were downloaded from the PRISM Climate Group 8, Oregon State

University.

Nesting sites were determined by dissolving NHD water layers into a single

object and identifying non-water features (land) smaller than 10,000 km 2. We overlaid this layer on satellite imagery to confirm island locations, and created island polygons for nesting sites not captured based on water locations, as was the case when rookeries occupied islands smaller than the spatial resolution of the source dataset (30m x 30m,

Figure 3.1). In areas where nesting occurred in swamps and on mainland peninsulas, polygons were created to estimate colony location. Nesting count data were then joined with this ‘islands’ layer in order to estimate colony densities. Colony polygons were then converted to points, corresponding with cell centers, to ensure distinct habitat differences within each nesting colony could be captured within the Maxent models.

3 http://nationalatlas.gov/ 4 http://www.mrlc.gov 5 http://www.fws.gov/wetlands/ 6 http://nhd.usgs.gov/ 7 http://water.epa.gov/ 8 http://www.prism.oregonstate.edu/ 84

Foraging Habitat

P. auritus is a piscivorous pursuit diving waterbird that propels itself underwater

while foraging for fish (Strod et al. 2004). We assumed that ideal foraging habitat would

be lentic or lotic water bodies large enough to contain suitable prey items (fish) of the

proper size classes (2-25mm total length; Craven and Lev 1987) that are not restricted by

emergent or submerged vegetation. We expected that water bodies large enough to

appear in the NHD (30 meter resolution, Eadie et al. 1986, Scheffer et al. 2006) would

provide suitable habitat for fish. Although aquatic vegetation impedes foraging of P.

auritus (Esler 1992, Traut and Hostetler 2004), wetlands can be important habitat for

fishes (Rozas and Odum 1988). Thus, we expected wetland habitat to be an important

variable for P. auritus foraging (Campo et al. 1993, Coleman et al. 2012, Goktepe et al.

2012). Supplemental fish stocking activities tend to increase the density of fishes of

adequate foraging size for P. auritus (Lorenzen 2000, Halverson 2008). Thus, fish

stocking programs were expected to enhance foraging habitat of P. auritus . Fish stocking

information is available from most state fisheries agencies; however, a compiled useable

national database is not publicly available. To develop a layer for fish stocking, we joined

data from known stocking locations with water body layers.

Nesting Habitat

Most cormorant nesting sites in Minnesota occur on islands (K. L. Sheehan,

personal observation). Many of these islands initially contained forested habitat. Over

time, trees and understory vegetation were defoliated because of fecal contamination

(Breuning-Madsen et al. 2008, Boutin et al. 2011), and nesting sites now occur on bare

ground. Vegetation damage might not occur as often in Florida and South Carolina, where P. auritus nest in forested wetlands and swamps, because guano is likely to dissolve in the surrounding waters (Wayne 1910, Post and Seals 1991). Differentiating between land cover types, such as forested lands and wetlands, could be important for identifying potential nesting sites of P. auritus . Forest cover may be important for both nest placement and acquisition of nesting material such as woody sticks (Baicich and

Harrison 1997). We reclassified data from the NLCD into three categories, one of which was a ‘forested’ classification. The other two categories identified areas of bare ground, which we considered important for nesting habitat, and landscape altered for human use, an anthropogenic variable. Upon inspection of islands where colonies of P. auritus occur, we found nesting sites were often classified in the NLCD as bare ground. Climate variables such as maximum temperature and precipitation also influence nesting and fledging success, because they affect the time adults spend foraging away from a nest

(Anderson et al. 2004, Coleman and Richmond 2007). Ten years of climate data

(minimum temperature, maximum temperature, average precipitation) for three months of the breeding season (March, June, September) were obtained from the PRISM Group at Oregon State University. Each climate variable was the average for each respective month for all years between 2000 and 2010.

Human Influence

The decline and eventual recovery of P. auritus was shaped profoundly by human activities associated with this bird and its environment (Wires et al. 2001). The number of people available to interact (directly or indirectly) with cormorants and their habitats can

strongly influence nesting success (Rodgers and Schwikert 2002). Direct interactions between humans and P. auritus include those associated with harassment or island disturbance by recreational activities (Ellison and Cleary 1978, Carney and Sydeman

1999, Rodgers and Schwikert 2002). Indirect influences can include decreased reproductive success because of low quality forage (contaminated fish), lack of forested areas nearby to support nesting birds or birds collecting nesting material (Dirksen et al.

1995, Larson et al. 1996). As a measure of human influence near nesting areas, human population estimates were derived from 2001 county-level Census data.

The proximity of a nesting area to developed lands (urban, suburban, and other residential classes) and agricultural lands (pasture, cultivated croplands, etc.) may also decrease the success of cormorant colonies (Carney and Sydeman 1999). We derived a layer from NLCD consisting of data for residential/developed land classifications and a layer based on the proportion of farmed lands by county was used to estimate agricultural lands. Impervious surfaces and intensity of night lights often increase with higher human populations and land development for anthropogenic activities. We included these layers as potential covariates to human population.

Human actions associated with industry and agriculture have led to a decline in the cleanliness and safety of aquatic habitats worldwide (Foley et al. 2005).

Consequently, fishing advisories have been enacted in many water bodies throughout the

United States. These advisories suggest limitations for consumption of fish and, in many cases, are quantitative and species specific (e.g., 4lb of brown trout/month because of

mercury contamination; Cunningham et al. 1994). Here, we identified water bodies that had fish advisories associated with Mercury, PCBs, and ‘other contaminants’, as well as rescinded advisories of any kind.

Derivation of Parameters

The aforementioned layers were converted (if necessary) to raster and snapped to a common registration point with a cell size of 30x30m. The Albers Equal Area Conic projection was chosen because of its preservation of area and minimal distortion of shape and distance within the conterminous United States. Prior to analysis, ‘NoData’ pixels were converted to a number which allowed for focal statistical analysis using the ArcGIS for Desktop software. The likelihood of any given location to be impacted positively or negatively by the values of other nearby cells was either summed, averaged, or maximized at a radius of either 3.5km or 10km (Appendix B Table B.2). We based focal statistic radii on foraging distances reported during the breeding season (Dorr et al. 2012,

Coleman et al. 2005). Guisan and Thuillier (2005) recommend focal statistics for highly mobile organisms because observations are likely to vary between potential and realized distributions when organisms move large distances to avoid disturbance. Layers developed from focal statistical analysis were clipped to the shape of the corresponding state so that raster values outside the political boundaries of Minnesota, Florida, or South

Carolina became ‘NoData’. This ensured that background data would be selected only for localities where all variables contained data. The final state-based raster layers were converted to the Tagged Image File Format (tif). Data were imported into the R statistical computing environment (R Core Team 2013) using the Dismo package. The Dismo

package is also used to wrap the functionality of the Maxent application, which executes inside of a Java Virtual Machine (Appendix Table B.1).

In some cases, we developed multiple focal statistic outputs using different spatial scales or alternate statistical metrics. For example, in the initial model, we included the maximum value of pixels containing water within a 10km and 3.5km radius of all points and the sum of pixel values for water bodies within a 10 km and 3.5km radius (Appendix

B Table B.2). We considered a conservative foraging radius of P. auritus to be 3.5km and a broader foraging radius to be 10km based on the variety of distances travelled by foraging P. auritus during the nesting season (Custer and Bunck 1992, Anderson et al.

2004, Coleman et al. 2004). We expected many of these layers to covary. During the model development phase, we identified groups of correlated parameters. Within each group, the variable that explained the most variance in nesting site distribution was retained and the remaining group members were removed from the model (York et al.

2011, Young et al. 2013).

Conspecific Parameters

In addition to environmental parameters, layers based on cormorant data were created to account for potentially important conspecific attractants that might influence nesting site selection. Craven and Lev (1987) document consistent reuse of colonies in

Wisconsin and Anderson et al. (2004) report similar findings in the Columbia River estuary. Thus, a layer reporting previous use of a given colony site could be informative for future nesting activity. We documented previously used sites by counting the number

of decades before 2000 that they had been used by P. auritus for nesting. Similarly, proximity to other nesting sites, either current or historical, might influence the use of a given location by nesting P. auritus . We measured the Euclidean distance to the nearest nesting colony for any pixel within each state. Additionally, we determined the density of the nearest nesting colony to any pixel. We expected these three layers would be highly influential within our models, effectively obscuring any influence of other environmental variables. This was important to consider because our intention was to develop methods for states without long-term cormorant/waterbird surveys. Thus, we sequentially removed these conspecific variables in order of importance in subsequent models. The final models presented here do not include these variables.

Species Distribution Models

We assessed nesting sites of P. auritus for influence of environmental parameters

(derived parameters) using the Dismo package in R to communicate with the Maxent program. For Minnesota and Florida, we stacked derived variables (Phillips et al. 2006) and a Maxent algorithm was run on presence-only data on the parameter stack (Hijmans and Elith 2013). We used variable contribution outputs to determine the most influential parameters on nesting location predictions. We built models through a series of iterations

(5 models for each step), removing environmental variables in the following order: variables that contributed no explanatory power to the model (providing 0% contribution); variables that provided 0.5% or less explanatory contribution; variables that covaried significantly with highly explanatory parameters. The final predicted distribution was plotted within the study area of each state and an additional predicted

distribution was plotted based on the models of Florida and Minnesota for South

Carolina.

Testing the Model Output

The Maxent program uses presence data and 10,000 additional ‘background’ points, which are considered to be locations of absence, in order to develop predictive models (Merow et al. 2013). A common metric used to test the predictive success of

SDM is the receiver operating characteristic (ROC) area under the curve (AUC; Merow et al. 2013). This value communicates the likelihood of a model to assign a higher prediction value for any randomly chosen presence point when compared to any randomly chosen background point (Merow et al. 2013). Additionally, we had real absence data points in Minnesota and Florida (Nisbet et al. 2002, Cuthbert et al. 2005,

Wires et al. 2010), which we used to test the predictive ability of each model. We sampled point data from the predictive map outputs based on known presence or absence data for each state using ArcGIS. We compared raw prediction values with known presence-absence data with Analysis of Variance (ANOVA). Chi-square analyses were used to compare presence-absence data and binary data derived from thresholds of the

Maximum Training Sensitivity plus Specificity (MTSS) and Balance training omission

(Balanced) values (Jimenez-Valverde and Lobo 2007 show derivations and descriptions).

MTSS and Balanced values are derived by Maxent and we used them as points of data truncation in order to categorize each point as good (1) or poor (0) nesting habitat of P. auritus (Jimenez-Valverde and Lobo 2007, York et al. 20011, Cao et al. 2013, Merow et al. 2013). Chi-square analyses of contingency tables based on actual vs. predicted data

were used to assess the ability of each model to correctly predict the actual status of a site observed for nesting P. auritus . Additionally, we wanted to determine whether model predictions for presence of P. auritus nesting sites corresponded with colony size. We used linear regression to compare nest density (nests/km 2) to prediction values derived by the Minnesota and Florida models.

To test for the possibility that the variables that explain cormorant nesting

locations are merely a representation of the landscape, we created a random set of 5,000

points within the state boundaries of Minnesota and Florida. We developed models for

random point locations in the same manner as the non-random models: removing

individual factors that contributed nothing or little to the explanatory power (small

changes to AUC) of the model, then parameters that might covary with variables that had

high explanatory power. These predictive maps of the random point models were tested

against the real nesting census data for Minnesota, Florida, and South Carolina.

Results

State Models

The Maxent Model for migratory cormorant habitat in Minnesota produced distribution prediction values ranging from 0.00004 to 0.956 (Figure 3.2). The receiver operating characteristic (ROC) gave an area under the curve (AUC) value of 0.911

(Figure 3.3). The ANOVA performed on observed nesting accounts confirmed the prediction success of this model for migratory birds (p<0.0001, Table 3.1) with a mean prediction of absence sites of 0.1205 ± 0.0062 (95% = CI 0.1084-0.1325) and a mean

prediction of presence sites of 0.5396 ± (95% CI 0.5349-0.5444). The threshold for the

MTSS as determined by Maxent was 0.277 and the Balanced threshold was set at 0.077.

A Chi-square analysis of the binary data agreed with known nesting site data for both the

MTSS (p<0.0001, R-square = 0.814) and the Balanced threshold (p<0.0001, R-square

0.394). We assessed the ability of Maxent to estimate colony size based on prediction values using linear regression. In Minnesota, higher prediction values (those most likely to be P. auritus nesting sites) corresponded with small colony sizes rather than large colony sizes (p = 0.0278, R-square = 0.0023).

The Florida model (Figure 3.4) successfully predicted presence and absence of P.

auritus nesting colonies (p<0.0001, Table 3.1). The mean prediction value for absence

sites was 0.0405 ± 0.0023 (95% CI = 0.0360-0.0450) and for presence values was 0.4955

± 0.0023 (95% CI = 0.4910-0.5000). The ROC of the Florida model indicated an AUC of

0.887, Figure 3.5). The threshold values used for presence/absence designation of a site

were 0.298 (MTSS) and 0.087 (Balanced). Chi-square indicated significant predictive

ability of the Florida model to identify nesting locations for both the MTSS threshold

(p<0.0001, R-square square 0.827) and the Balanced threshold (p<0.0001, R – square =

0.642). Linear regression analysis for Florida colony densities agreed with prediction

values derived from Maxent (p<0.0001, R-square = 0.203), with large colonies exhibiting

higher predicted values.

Model Predictions for South Carolina

In South Carolina, there are few current nesting sites of P. auritus (personal

communication, Christy Hand, SCDNR). Historical nesting sites of P. auritus in South

Carolina prior to the 1950’s are available (Wayne 1910). Contemporary colonies nesting

in the state now persist in reservoir lakes created in the 1950’s (Post and Post1988, Post

and Seals 1991, K. L. Sheehan, personal observation). Thus, we did not consider

historical nesting sites for this assessment. The models for Minnesota and Florida (Figure

3.6) performed well when prediction values were compared with ANOVA (Minnesota

model: p<0.0001 and R-square = 0.256; Florida model: p<0.0001 and R-square = 0.218;

Table 3.1). The same threshold values were used for the predictions of presence/absence

of nesting habitat in South Carolina, yielding successful nesting habitat prediction based

on MTSS or Balanced thresholds for the Minnesota model. Nesting sites of P. auritus

based on the Florida model identified two colonies with the MTSS threshold values

(Figure 3.7; p<0.0001, R – square = 0.3401).

Model Validation Results

When tested for prediction success in Florida, the Minnesota model performed

well (p<0.0001), but explained little of the variance in nest presence (R-square = 0.082)

using prediction values. When truncated to presence/absence of nesting habitat based on

threshold values, the Minnesota model identified no P. auritus nesting locations in

Florida. The Florida model did not successfully predict nesting locations in Minnesota

with prediction values (p=0.5069), but was able to correctly identify nesting locations

when truncated (p=0.0043, R-square = 0.0038).

In addition to cross-validation of data among states, we tested whether variables in the final Minnesota and Florida models predicted P. auritus nesting sites better than

models built from random data points. The random Minnesota model did not successfully

predict nesting sites of P. auritus (p = 0.4132, AUC = 0.527; Figure 3.8). The Florida

model created with random points was able to predict nesting habitat of P. auritus (p =

0.033, R-square = 0.0009, AUC = 0.536; Figure 3.9); however, the prediction values for

absence points (mean = 0.941748 ± 0.0049, 95% CI = 0.9322-0.9513) were higher than

those for presence points (0.9271 ± 0.0049, 95% CI = 0.9176-0.9366). When used to

predict nesting habitat in South Carolina, the random Minnesota model performed well

when using prediction values (p<0.0001), but identified the non-nesting sites as more

suitable for breeding P. auritus. We tested the MTSS and Balanced threshold values for

nesting assignment in South Carolina and found no suitable nesting habitat based on the

random Minnesota model. Similarly, the random Florida model predicted that the most

suitable nesting habitats for P. auritus were locations where colonies were absent, and the

truncated models identified no nesting sites in South Carolina.

Model Parameters

Seventeen parameters were included in the final model predicting P. auritus

nesting habitat in Minnesota (Table 3.2, Figure 3.10). The variable that explained the

most variance in P. auritus distribution was the cumulative water area available for fish

habitat (23.3% variable contribution). Interestingly, in smaller water sources, prediction

values increased when all other variables are held constant (Figure 3.11A). The presence

of water within a 3.5 km radius was also important for model performance (16.5%

variable contribution). Higher prediction values for nesting habitat occurred where there was a higher incidence of water (Figure 3.11B). The quantity of impervious surfaces was also a strong contributing factor (13.4% variable contribution), where moderate imperviousness values correspond with higher predictions of P. auritus nesting habitat

(Figure 3.11 C). The presence of forested habitat within a 3.5km radius was negatively associated with P. auritus nesting probability (9.3% variable contribution; Figure 3.11D).

Fish advisories associated with mercury contamination were also important, contributing

7.3% to model performance. Higher P. auritus distribution predictions occurred when mercury advisories were either very high or very low, when holding all other contributing variables constant (Figure 3.11E).

As in Minnesota, seventeen parameters were included in the final model predicting P. auritus nesting habitat in Florida (Table 3.2, Figure 3.12). The presence of water within a 10km radius was negatively related to prediction value when all other variables were held constant (19.7% variable contribution, Figure 3.13A). Undeveloped lands, consisting of unforested areas that were not cleared for urbanization or agricultural use, exhibited a trend for a positive association with predicted value (16.1% variable contribution, Figure 3.13B). A negative relationship is apparent between land area used for agriculture and P. auritus nest site predictions (12% variable contribution, Figure

3.13C). Minimum temperature in March contributed 12% to the predictive ability of the final model for Florida. When all other variables were held constant, sites with high prediction values for nest sites of P. auritus were cooler than those without nests (Figure

3.13D). Like in the Minnesota model, forested lands were also a strong contributor for

the prediction of P. auritus nesting sites. High densities of forested lands correlated with lower prediction values for both the Minnesota and Florida models (11% variable contribution, Figure 3.13E).

When we compared the model parameters included in the Minnesota and Florida models, there were important parameters that appeared in both models. Variables occurring in both models included water features, forested lands, fish advisories, availability of wetland habitat, land use change, undeveloped lands, avian mortalities associated with orthophosphate poisoning, and the minimum temperature for the onset of the breeding season (March). Factors that differed between the models include proximity to lands developed for agriculture, climate variables, presence of Native American lands, and avian mortalities associated with Botulism. Although more than half of the variables included in the Minnesota and Florida models were the same, the importance of each varied.

Discussion

One option to consider for effective management of avian subspecies with differing conservation imperatives is to identify habitats that could promote or discourage the establishment of colonies. The Maxent algorithm was used to predict suitable habitat for nesting cormorants in Minnesota and Florida based on local environmental and anthropocentric parameters that we determined could be important for the foraging and nesting success of the subspecies. We detected similar important variables for the

Minnesota and Florida models, but the importance of each parameter varied. We avoided

over-fitting of the models due to multicolinearity by removing data layers found to be correlated (Evangelista et al. 2005, York et al. 20011, Young et al. 2013), and in doing so, reduced the model parameters from 42 to 17.

In Minnesota, cormorant nesting sites occur near large water bodies that were relatively isolated from other waters (Cuthbert et al. 2005, Wires et al. 2010). This is likely a descriptor of the water features of the state, as formation of many of the lakes and ponds in Minnesota are associated with glacial deposits (Herwig et al. 2010, Sepulveda-

Villet and Stepin 2012). Nesting habitat is predicted to occur more commonly in areas with moderate coverage of impervious surfaces. We interpret this finding as an indication that landscapes containing few impervious surfaces (heavily forested areas, agricultural lands with unpaved roads, and the surfaces of water bodies) do not contain habitat suitable for nest formation. Isolated water bodies with no impervious surfaces nearby are less likely to be stocked with fish. Similarly, densely populated areas where impervious surface coverage is very high are unlikely candidates for P. auritus nesting because of

disturbance, although exceptions to this have been observed in California and Canada

(Stenzel et al. 1995, Magnuson et al. 1998, Chatwin et al. 2002). We observe moderate

impervious surface coverage near the perimeters of many lakes and rivers, where island

nesting sites occur most frequently (Arnold and Gibbons 1996, Langen et al. 2005). Low

densities of forested lands were associated with high prediction values of P. auritus

nesting habitat. Nesting sites on islands in Minnesota are common (Wires et al. 2001),

where the average area consisting of forested lands is low, considering small island area

and large water area around island nesting sites.

The potential impacts that fish advisories can have on piscivores is complicated, but the presence of highly contaminated fish can be detrimental to top predators (Ludwig et al. 1995, Scheuhammer et al. 2007). P auritus is relatively insensitive to heavy mercury contamination (Henny et al. 2002, Heinz et al. 2009); however morbidity associated with mercury poisoning has occurred (Sepulveda et al. 1998). Thus, low cormorant colony densities may be maintained through contaminated food. Alternatively,

P. auritus largely consume forage fish (Campo et al. 1993), but mercury testing of fishes concentrates on sport fishes from the highest trophic positions within an aquatic community (McClain et al. 2006). Thus, fish advisories can indicate increased complexity of trophic webs that contain higher-order top predators (McClain et al. 2006,

McIntyre and Beauchamp 2007). It is possible that cormorants nest and forage near aquatic systems where there is little competition with higher-order predatory fishes for forage fish resources. Alternatively, dilution of mercury through many types of predators such as cormorants and predatory sport fish, may decrease the accumulation of contaminants in cormorant and fish tissues. Because sampling for fish contamination is typically concentrated around populated places (Burger 2013), it is likely that these data do not represent the actual likelihood of a cormorant eating a contaminated fish.

The occurrence of water within a 10km radius of a given point was an important

variable with a negative correlation in predicting the presence of nesting habitat in

Florida. Water classification describes open water bodies and flowing water systems and

was classified separately from wetland habitat. The separation of waters and wetlands

resulted in an increase in the variation in water body size and connectivity. Thus, a

negative association with how frequently water occurs within a 10km radius outperformed other water variables in predicting P. auritus nesting habitat. Predicted values of cormorant nesting habitat were higher at very low and moderate quantities of undeveloped land. We classified undeveloped lands based on the NLCD where barren land (rocks/sand/clay, NLCD Legend), shrub/scrub habitat, and grassland/herbaceous habitats grouped together. The majority of nest sites in Florida occurred on small islands dominated by sandy habitat; however, many arboreal nesting colonies occurred in wetlands (Nisbet et al. 2002). Forested wetlands were classified as forested habitat, which also was a strong contributor to prediction value in the Florida model. Similar to the trends with forest cover in Minnesota, P. auritus nest habitat was predicted more frequently where forest was present, but covered little of the area in a 3.5km radius. A negative association between the proportion of area covered by agricultural land and prediction value was also demonstrated for P. auritus nesting sites in Florida. The conversion of wetland area for agricultural use and the diversion of flowing waters to supplement farming practices in Florida limit the aquatic habitat available to piscivorous waterbirds for foraging (Guardo et al. 1995, Kautz et al. 2007). Thus, although heavily farmed areas are unsuitable for cormorant nesting, agricultural lands to some degree can be informative of P. auritus nesting habitat by indicating areas where aquatic habitat occurs. Beyond landscape components, we found that the minimum temperature in June was an important variable for developing a prediction value for nesting success of P. auritus in Florida. The minimum temperature ranged from 19°C to 26°C and higher P. auritus nesting habitat prediction values were associated with lower minimum

100

temperatures. Lower temperatures in the morning could allow for longer, farther foraging bouts of adult P. auritus . This could increase the provisioning ability and nest success of

P. auritus in climates where foraging time is reduced by high temperatures that require parents to shield eggs and chicks from heat and desiccation (Coleman et al. 2005).

We transferred the predictions of the Minnesota and Florida models to the extent of South Carolina. Transferability of a model is dependent on the similarities between the region used to develop the model and the area the model is transferred to (Thuiller et al.

2004, Randin et al. 2006, Peterson et al. 2007, Warren and Seifert 2011). In our models, data that might have differed significantly in range was ranked prior to focal statistic transformation. This allowed for focal statistic numbers to comprise the same range of values when being calculated, preventing transferability problems of interpretation associated with clamping (Phillips et al. 2006). Because focal statistics were used on ranked data, the final layers that we included in our models were continuous rather than nominal. The default settings of Maxent assume data are continuous (Phillips and Dudik

2008), and we did not need to alter the settings when developing our models. Climate variables were the only environmental parameters not used with focal statistics.

Temperature averages in March and September are higher for the two southern states

(South Carolina and Florida) than for Minnesota (Easterling et al. 1997); however, temperature variables in the Minnesota model contributed a cumulative 10% to the contribution of prediction value assignments. Thus, we do not expect this model to be incompatible with the variable values for South Carolina or Florida. We suggest that the

101

results of the threshold tests in South Carolina are correct and conclude that the nesting colonies of South Carolina are P. a. floridanus and not P. a. auritus .

Additional Considerations for the Model

The large, national extent and relatively fine resolution (30m) at which we developed our models prevented us from using some biological and environmental variables that might have been informative on P. auritus nesting habitat prediction at local scales. Nonetheless, our models successfully predicted presence and absence points in the states for which they were designed and were transferred to the state of South

Carolina with some success. Many of the variables included in our models, although relatively general, are likely to capture the essence of important nesting and foraging habitat of P. auritus . Here, our intention was to determine whether nesting conditions of

P. auritus in South Carolina resemble characteristic traits of nesting P. a. auritus in

Minnesota or P. a. floridanus in Florida. Different goals for understanding the mechanistic biology and ecology of nesting waterbirds might require additional local variables of importance to the species of interest. Specific parameters that might be informative include: fine-scale submerged and emergent vegetation data, climate conditions that would contribute to exposure severity such as lake fetch accumulation and forest cover density, as well as recreation variables that might be useful for estimating the human use of each water body (water depth, boat launches, beaches, etc.).

102

Data Points Excluded from Model Development

Some of the survey points in the Minnesota and Florida datasets were excluded from model development because they were known zero values. Maxent does not consider absence, population size, or density, but rather assumes that any point entered is a presence value (Elith et al. 2011); thus, many useful points where cormorants were decidedly absent were not used to develop habitat prediction values. We feel that our use of these data in model validation was a good compromise, because we were interested in identifying differences between nesting habitat and all other types of habitat. If instead, we were interested in recognizing differences in presence and absence points, we would have limited model robustness by decreasing the number of background/absence points and excluding descriptive landscape information that would allow for transferal to succeed (Thuiller et al. 2004, Randin et al. 2006). P. auritus has a contentious history in

North America where harassment and exploitation of nesting colonies by humans is common (Wires et al. 2006). Today, although these birds are protected by the Migratory

Bird Treaty Act (16 USC Chapter 7, Subchapter II - MIGRATORY BIRD TREATY), they are still harassed and forced to abandon suitable nesting habitat (Tobin et al. 2002,

Farquhar III et al. 2003, Wires et al. 2006). Thus, the realized ecological niche where P. auritus nests does not necessarily represent the potential ecological niche. Thus, human disturbance represents a portion of the disparity between the fundamental and realized ecological niche of P. auritus . Our models include anthropocentric variables that could help increase the accuracy of predictions of nest site suitability. Many of these anthropocentric variables can be controlled to some degree by urban planning and natural

103

resource management, points to consider for management goals of altering the potential distribution of P. auritus .

Conclusion and Future Directions

The nesting habitat of P. a. floridanus is dissimilar from the nesting habitat of P. a. auritus . Our models suggest that the habitat available in South Carolina for nesting colonies of P. auritus resemble nesting habitat of P. a. floridanus and not that of P. a. auritus . The models described suggest geographic parameters such as water density and forested land are critical predictors for the distribution of P. auritus . Anthropogenic

parameters such as the quantity and distribution of impervious surfaces were also

important. These and other parameters can be manipulated through changes in land

management practices. For example, connecting and converting undeveloped lands to

forested habitat near potential nesting sites might reduce the attractiveness of a site for P. a. auritus colonies. Additionally, the removal or alteration of roosting habitat (standing dead cypress trees) in areas where P. a. auritus are undesirable could prevent colony establishment. Furthermore, mechanistic characteristics of parameters such as fish stocking activities could be explored in greater detail by managers to elucidate whether timing, stocking numbers, richness of species stocked, or size of stocked fish in specific lakes could be altered to deter or attract the occurrence and density of P. auritus nesting colonies. We encourage managers to consider using similar methods to identify potential factors that could be manipulated to decrease the attractiveness of managed lands to undesirable cormorant colonies while still preserving ecosystem services. Such models

104

could be useful for conservationists interested in differentiating between migratory and resident populations in the absence of reliable molecular evidence.

Our models demonstrate how readily available environmental variables can be used to develop Maxent models that accurately describe the distribution of colonial waterbirds. Our Florida model successfully identified current nesting sites of P. auritus in

South Carolina; whereas, our Minnesota model was unable to successfully classify current nesting habitat in South Carolina. Molecular assessments of populations and subspecies of P. auritus failed to separate the migratory P. a. auritus and resident P. a. floridanus subspecies (Waits et al. 2003, Green et al. 2006, Mercer 2013). Thus, the current differentiation of these groups is based solely on breeding behavior. Our assessments support a separation of subspecies based on distinct nesting habitat use and suggest that the breeding colonies of P. auritus in South Carolina are the Florida subspecies, P. a. floridanus . This information could be used to refine management plans for both subspecies in states where the two types of cormorants overlap in geographic distribution.

105

Tables Table 3.1. Results of Student’s T-tests used to compare model predictions based on prediction value and values truncated at the threshold for maximum training sensitivity plus specificity (MTSS).

Model State Test T-test P R-square Truncated T -test P R-square Minnesota Minnesota <0.0001 0.653 <0.0001 0.750 Florida Florida <0.0001 0.791 <0.0001 0.827

Minnesota Florida <0.0001 0.082 N/A N/A Florida Minnesota 0.507 0.0004 <0.0001 0.004 Minnesota S.Carolina <0.0001 0.256 N/A N/A Florida S.Carolina <0.0001 0.218 <0.0001 0.036

106

Table 3.2. Variable contribution for parameters included in the final nesting habitat models developed with Maxent in Minnesota and Florida. Factors that appear in both models are highlighted.

Minnesota Florida Conspecific Avian Botulism Death 0.1 Undeveloped Land 2.7 16.1 Foraging Avg. Wetland Area 4 6.9 Lbs. Fish Stocked 10k Lbs. Fish Stocked 3.5k 0.7 Min Temp September Num Fish Stocked 10k Water Availability 3.5k 23.3 Water Presence 10k Water Quantity 3.5k 16.5 Water Quantity 10k 19.7 Nesting Forested Land 9.3 11 Max Temp June 1.8 2.6 Max Temp March 3.7 Max Temp September Min Temp March 1.3 12 Min Temp September 4.4 Min Tempt June 2.9 Precipitation March 3.4 Precipitation September 5.3 Anthropocentric Anthropogenic Land Agriculture Quantity 12 Avian Lead Poisoning Avian Pesticide Poison 1.1 0.7 Human Pop. Density 2.1 Impervious Surf. Quant 13.4 4.8 Indian Land 1 Land Use Change 1.9 2.7 Mercury Fish Advisory 7.3 0.6 Rescinded Fish Adv. 2.7 2

107

Figures

Figure 3.1. Aerial imagery of island nesting sites reported in Florida where sandy and

shell-hash substrates (light areas in photo) connect P. auritus nesting areas (inset image)

in trees. Example polygons drawn around P. auritus colonies. 108

Figure 3.2. Prediction of P. auritus nesting habitat in the state of Minnesota based on characteristics from known cormorant breeding colonies.

109

Figure 3.3. Receiver operating characteristic (ROC)9 curve for the training presence

records used to develop the Maxent model for P. auritus nesting habitat in Minnesota.

The black line represents random predictions with 0.5 probability of correctly identifying nest locations whereas the area under the curve (AUC) represented by the red line represents a much higher 0.91 probability.

9 Area under the curve (AUC) demonstrates the probability of positive predictions to be ranked higher than negative prediction values. 110

Figure 3.4. Prediction of P. auritus nesting habitat in Florida based on known cormorant breeding colonies.

111

Figure 3.5. Receiver operating characteristic (ROC 10 ) curve for the training presence

records used to develop the Maxent model for P. auritus nesting habitat in Florida.

10 Area under the curve (AUC) demonstrates the probability of positive predictions to be ranked higher than negative prediction values.

112

Figure 3.6. Prediction values of P. auritus nesting habitat in South Carolina based on parameters that describe the ecological niche of cormorants nesting in (A) Minnesota and (B) Florida.

113

Figure 3.7. P. auritus nesting habitat in South Carolina based on parameters that describe the ecological niche of cormorants nesting in Florida. Continuous prediction values were converted to “good” and “poor” habitat based on an MTSS threshold value derived from the MAXENT_FL model.

114

Figure 3.8. Receiver operating characteristic (ROC)11 curve for the training presence records used to develop the Maxent model for random points generated in Minnesota.

11 Area under the curve (AUC) demonstrates the probability of positive predictions to be ranked higher than negative prediction values. 115

Figure 3.9. Receiver operating characteristic (ROC)12 curve for the training presence records used to develop the Maxent model for random points generated in Florida.

12 Area under the curve (AUC) demonstrates the probability of positive predictions to be ranked higher than negative prediction values. 116

Water Availability 3.5k Water Quantity 3.5k Imervious Surf. Quant Forested Land Mercury Fish Advisory Precipitation Septmber Min Temp September Avg Wetland Area Precipitation March

Variable Undeveloped Land Rescinded Fish Adv Land Use Change Max Temp June Min Temp March Avian Pesticide Poison Indian Land Lbs Fish Stocked 3.5k

0 5 10 15 20 25 Percent Contribution

Figure 3.10. Variable contribution plot for parameters included in the final Maxent model of P. auritus nesting habitat in

Minnesota. 117

Figure 3.11. Predictor profile plots for the 5 most explanatory variables 13 for the

Minnesota model.

13 Behavior of a P. auritus nesting habitat prediction value related to the magnitude of each factor is displayed while holding all other variable values constant. 118

Water Quantity 10k Undeveloped Land Agriculture Quantity Min Temp March Forested Land Avg Wetland Area Imervious Surf. Quant Max Temp March Min Tempt June Land Use Change Variable Max Temp June Human Pop. Density Rescinded Fish Adv Avian Pesticide Poison Mercury Fish Advisory Avian Botulism Death Lbs Fish Stocked 3.5k

0 5 10 15 20 25 Percent Contribution

Figure 3.12. Variable contribution plot for variables included in the final Maxent model of cormorant nesting habitat in

Florida. 119

Figure 3.13. Predictor profile plots for the 5 most explanatory variables 14 for the Florida

model.

14 Behavior of a P. auritus nesting habitat prediction value related to the magnitude of each factor is displayed while holding all other variable values constant.

120

Acknowledgements The development of this project would not have happened without the inspiration of Dr. Robert Baldwin, who encouraged the initial ideas in his Landscape Ecology and

Conservation course. We appreciate the use of the GIS computing facilities on the campus of Clemson University, and the support provided by ESRI staff for overcoming obstacles in the project implementation. We thank the Clemson University faculty who reviewed this publication prior to submission, as their recommendations for improvement were highly valuable.

121

Literature Cited Aarts, G. et al. 2012. Comparative interpretation of count, presence–absence and point methods for species distribution models. — Methods in Ecology and Evolution 3: 177-187.

Anderson, C. D. et al. 2004. Foraging patterns of male and female Double-crested Cormorants nesting in the Columbia River estuary. — Canadian J. of Zoology 82: 541-554.

Arnold, Jr., C. L. and Gibbons, C. J. 1996. Impervious surface coverage: the emergence of a key environmental indicator. — J. Amer. Plan. Assoc. 62:243-258.

Baicich, P. J. and Harrison, C. J. O. 1997. A guide to the nests, eggs, and nestlings of North American birds (2nd ed.). — Academic Press, San Diego.

Boutin, C. et al. 2011. Effects of Double-crested Cormorants ( Phalacrocorax auritus Less.) on island vegetation, seedbank, and soil chemistry: evaluating island restoration potential. — Restoration Ecology 19: 720-727.

Breuning-Madsen, H. et al. 2008. The impact of perennial cormorant colonies on soil phosphorus status. — Geoderma 148: 51-54.

Brugger, K. E. 1995. Double-crested Cormorants and fisheries in Florida. — Colonial Waterbirds 18:110-117.

Burger, J. 2004. Fish consumption advisories: knowledge, compliance and why people fish in an urban estuary. — J. Risk Res. 7:463-479.

Campo, J. J. et al. 1993. Diet of Double-crested Cormorants wintering in Texas. — J. Field Ornithology 64: 135-144.

Cao, Y. et al. 2013. Using Maxent to model the historic distributions of stonefly species in Illinois streams: the effects of regularization and threshold selections. — Ecol. Modelling 259:30-39.

Carney, K. M. and Sydeman, W. J. 1999. A review of human disturbance effects on nesting colonial waterbirds. — Waterbirds 22: 68-79.

Carranza, J. and Winn, H. E. 1954. Reproductive behavior of the Blackstripe Topminnow, Fundulus notatus. — Copeia 1954:273-278.

122

Chatwin, T. A. et al. 2002. Changes in pelagic and Double-crested Cormorant nesting populations in the Strait of Georgia, British Columbia. — Northwestern Naturalist 83:109-117.

Coleman, J. T. H. and Richmond, M. E. 2007. Daily foraging patterns of adult Double- crested Cormorants during the breeding season. — Waterbirds 30: 189-198.

Coleman, J. T. H. et al. 2005. Foraging location and site fidelity of the Double-crested Cormorant on Oneida Lake, New York. — Waterbirds 28: 498-510.

Coleman, J. T. H. et al. 2012. Eating the invaders: the prevalence of Round Goby (Apollonia melanostomus ) in the diet of Double-crested Cormorants on the Niagara River. — Waterbirds 35: 103-113.

Craven, S. R. and Lev, R. E. 1985. Double-crested Cormorant damage to a commercial fishery in the Apostle islands, Wisconsin. — Eastern Wildlife Damage Control Conference. 2: page 5.

Craven, S. R. and Lev, R. E. 1987. Double-crested Cormorants in the Apostle Islands, Wisconsin, USA: population trends, food habits, and fishery depredations. — Colonial Waterbirds 10: 64-71.

Cunningham, P. A. et al. 1994. A national fish consumption advisory data base: a step toward consistency. — Fisheries 19:14-23.

Custer, T. W. and Bunck, C. 1992. Feeding flights of breeding Double-crested Cormorants at two Wisconsin colonies. J. Field Ornithol. 63:203-211.

Dirksen, S. et al. 1995. Reduced breeding success of cormorants ( Phalacrocorax carbo sinensis ) in relation to persistent organochlorine pollution of aquatic habitats in the Netherlands. — Env. Pollution 88: 119-132.

Dorr, B. S. et al. 2012. Summer and migrational movements of satellite-marked Double- crested Cormorants from a breeding colony managed by egg-oiling in Lake Ontario, USA. Waterbirds 35: 114-123.

Eadie, J. M. et al. 1986. Lakes and rivers as islands: species-area relationships in the fish faunas of Ontario. — Env. Bio. Of Fishes 15: 81-89.

Easterling, D. R. et al. 1997. Maximum and minimum temperature trends for the globe. — Science 277:364-367.

123

Elith, J., S. et al. 2011. A statistical explanation of MaxEnt for ecologists. — Diversity and Distributions 17: 43-57.

Engler, R. et al. 2004. An improved approach for predicting the distribution of rare and endangered species from occurrence and pseudo-absence data. — J. Appl. Ecology 41:263-274.

Esler, D. 1992. Habitat use by piscivorous birds on a power plant cooling reservoir. — J. Field Ornith. 63:241-249.

Evangelista, P. H. et al. 2009. Mapping invasive Tamarisk ( Tamarix ): a comparison of single-scene and time-series analyses of remotely sensed data. — Remote Sens. 1:519-533.

Evans, R. J. et al. 2010. Comparative nest habitat characteristics of sympatric White- tailed Haliaeetus albicilla and Golden Eagles Aquila chrysaetos in western Scotland. — Bird Study.57:473-482.

Farquhar, III, J. F. et al. 2003. Human harassment and the Double-crested Cormorant Phalacrocorax auritus nesting at three colonies in eastern Lake Ontario, New York, USA: observations from a management program. — Vogelwelt 124:319- 324.

Foley, J. A. et al. 2005. Global consequences of land use. — Science 309: 570-574.

Fonteneau, F. et al. 2009. Relationships between bird morphology and prey selection in two sympatric Great Cormorant Phalacrocorax carbo subspecies during winter. — Ibis 1-13.

Goktepe, O. et al. 2012. Comparing bioenergetics models of Double-crested Cormorant (Phalacrocorax auritus ) fish consumption. — Waterbirds 35: 91-102.

Gormley, A. M. et al. 2011. Using presence-only and presence-absence data to estimate the current and potential distributions of established invasive species. — J. Appl. Ecology 48:25-34.

Green, M. C. et al. 2006. Microsatellite variation of Double-crested Cormorant populations in eastern North America. — J. Wildlife Manag. 70: 579-583.

Guardo, M. et al. 1995. Large-scale constructed wetlands for nutrient removal from stormwater runoff: an Everglades restoration project. — Env. Manag. 19:879-889.

124

Guisan, A. and Thuiller, W. 2005. Predicting species distribution: offering more than simple habitat models. — Ecology Letters 8:993-1009.

Haig, S. M. et al. 2006. Taxonomic considerations in listing subspecies under the U.S. Endangered Species Act. USGS Staff - Published Research. Paper 671. http://digitalcommons.unl.edu/usgsstaffpub/671

Halverson, M. A. 2008. Stocking trends: a quantitative review of governmental fish stocking in the United States, 1931 to 2004. — Fisheries 33:69-75.

Hastie, T. and Fithian, W. 2013. Inference from presence-only data; the ongoing controversy. — Ecography 36:864-867.

Hatch, J. J. 1995. Changing populations of Double-crested Cormorants. — Colonial Waterbirds 18: 8-24

Heinz, G. H. et al. 2009. Species differences in the sensitivity of avian embryos to methylmercury. Arch. Environ. Contam. Toxicol. — 56:129-138.

Henny, C. J. et al. 2002. Nineteenth Century mercury: hazard to wading birds and cormorants of the Carson River, Nevada. Ecotoxicol. — 11:213-231.

Herwig, B. R. et al. 2010. Factors influencing fish distributions in shallow lakes in prairie and prairie-parkland regions of Minnesota, USA. — Wetlands 30:609-619.

Hijmans R. J. and Elith, J. 2011. Species distribution modeling with R. 72 pp. Available online at http://cran.r-project.org/web/packages/dismo/vignettes/sdm.pdf. Hutto, R. L. 1985. Habitat selection by nonbreeding, migratory land birds. Habitat Selection in Birds — Academic Press, Orlando, FL. 455-476.

Jimenez-Valverde, A. and Lobo, J. M. 2007. Threshold criteria for conversion of probability of species presence to either-or presence-absence. — Acta Oecologia 31:361-369.

Kautz, R. et al. 2007. Florida vegetation 2003 and land use change between 1985-89 and 2003. — Florida Scient. 70:12-23.

Langen, T. A. 2005. Pelagic bird surveys on Lake Ontario following Hurricane Isabel, September 2003: observations and remarks on methodology. — J. Great Lakes Res. 31:219-226.

125

Larson, J. M. et al. 1996. Reproductive success, developmental anomalies, and environmental contaminants in Double-crested Cormorants ( Phalacrocorax auritus ). — Env. Tox. Chem. 15: 553-559.

Lorenzen, K. 2000. Allometry of natural mortality as a basis for assessing optimal release size in fish-stocking programmes. — Can. J. Fish. Aquat. Sci. 57:2374-2381.

Ludwig, J. P. et al. 1995. Evaluation of the effects of toxic chemicals in Great Lakes cormorants: has causality been established? — Colonial Waterbirds 18:60-69.

Magnuson, J. J. et al. 1998. Isolation vs. extinction in the assembly of fishes in small northern lakes. — Ecology 79:2941-2956.

Mallet, J. 1995. A species definition for the modern synthesis. — TREE 10: 294-299.

Mayr, E. 1982. Of what use are subspecies? — The Auk 99:593-595.

McClain, W. C. et al. 2006. Mercury concentrations in fish from Lake Meredith, Texas: implications for the issuance of fish consumption advisories. — Env. Monitor. Assess. 123:249-258.

McIntyre, J. K. and Beauchamp, D. A. 2007. Age and trophic position dominate bioaccumulation of mercury and organichlorines in the food web of Lake Washington. — Sci. Total Env. 372:571-584.

Mercer, D. M. et al. 2013. Phylogeography and population genetic structure of Double- crested Cormorants ( Phalacrocorax auritus ). — Conserv. Genet. 14:823-836.

Merow, C. 2013. A practical guide to Maxent for modeling species’ distributions: what it does, and why inputs and settings matter. — Ecography 36:1-12.

Nisbet, I. C. T. 1975. Selective effects of predation in a tern colony. — The Condor 77:221-226.

Nisbet, S. A. et al. 2002. Atlas of waterbird colonies in Florida during 1999, with an analysis of the accuracy of aerial estimates. Final Study Report. Bureau of Wildlife Diversity Conservation, Gainesville, FL 32601 (Study 7613). FY1998B2002.

Oppel, S. et al. 2012. Comparison of five modelling techniques to predict the spatial distribution and abundance of seabirds. — Biol. Conserv. 156:94-104.

126

Peterson, A. T. et al. 2003. Predicting the potential invasive distributions for four alien plant species in North America. Weed Sci. 51:863-868.

Peterson, A. T. et al. 2007. Transferability and model evaluation in ecological niche modeling: a comparison of GARP and Maxent. — Ecography 30:550-560.

Phillips, S. J. 2008. Transferability, sample selection bias and background data in presence-only modelling: a response to Peterson et al. (2007). — Ecography 31: 272-278.

Phillips, S. J. and Dudik, M. 2008. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. — Ecography 31:161-175.

Phillips, S. J. et al. 2006. Maximum entropy modeling of species geographic distributions. — Ecol. Modeling 190: 231-259.

Post, W. and Post, C. A. 1988. Double-crested Cormorant nesting in South Carolina. The Chat 52:34-35.

Post, W. and Seals, C. A. 1991. Breeding biology of a newly-established Double-crested Cormorant population in South Carolina, USA. — Colonial Waterbirds 14: 34-38.

R Core Team. 2013. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org/.

Rand, A. L. and Traylor, M. A. 1950. The amount of overlap allowable for subspecies. The Auk 67:169-183.

Randin, C. F. et al. 2006. Are niche-based species distribution models transferable in space? J. Biogeogr. 33:1689-1703.

Raxworthy, C. J. et al. 2003. Predicting distributions of known and unknown reptile species in Madagascar. — Nature 426:837-841.

Renner, I. W. and Warton, D. I. 2013. Equivalence of Maxent and Poisson Point Process Models for Species Distribution Modelling in Ecology. — Biometrics 69:274- 281.

Ridgway, M. S. 2010. Line transect distance sampling in aerial surveys for Double- crested Cormorants in coastal regions of Lake Huron. — J. of Great Lakes Res. 36:403-410.

127

Rodgers, Jr., J. A and Schwikert, S. T. 2002. Buffer-zone distances to protect foraging and loafing waterbirds from disturbance by personal watercraft and outboard- powered boats. — Con. Biol. 16: 216-224.

Rozas, L. P and W. E. Odum. 1988. Occupation of submerged aquatic vegetation by fishes: testing the roles of food and refuge. — Oecologia 77:101-106.

Scheffer, M. et al. 2006. Small habitat size and isolation can promote species richness second-order effects on biodiversity in shallow lakes and ponds. — Oikos 112:227-231.

Scheuhammer, A. M. et al. 2007. Effects of environmental methylmercury on the health of wild birds, mammals, and fish. — Ambio 36:12-18.

Sepulveda, M. S. et al. 1998. Concentrations of mercury and selenium in tissues of Double-crested Cormorants ( Phalacrocorax auritus ) from southern Florida — Colonial Waterbirds 21:35-42.

Sepulveda-Villet, O. J. and Stepien, C. A. 2012. Waterscape genetics of the yellow perch (Perca flavescens ): patterns across large connected ecosystems and isolated relict populations — Mol. Ecol. 21:5795-5826.

Stanford, C. B. and Nkurunungi, J. B. 2003. Behavioral ecology of sympatric chimpanzees and gorillas in Bwindi Impenetrable National Park, Uganda: diet. — Int. J. Primatology 24:901-918.

Stenzel, L. E. et al. 1995. Breeding success of Double-crested Cormorants in the San Francisco Bay area, California. — Colonial Waterbirds 18:216-224.

Strod, T. et al. 2004. Cormorants keep their power: visual resolution in a pursuit-diving bird under amphibious and turbid conditions. — Current Biology, 14: R376- R377.

Thomas, C. D. et al. 2004. Extinction risk from climate change. Letters to Nature 427:145-148.

Thuiller, W. et al. 2004. Effects of restricting environmental range of data to project current and future species distributions. — Ecography 27:165-172.

Tobin, M. E. et al. 2002. Effect of roost harassment on cormorant movements and roosting in the delta region of Mississippi. — USDA NWRC Staff Publications. Paper 501.

128

Traut, A. H. and Hostetler, M. E. 2004. Urban lakes and waterbirds: effects of shoreline development on avian distribution. — Landscape and Urban Planning 69:69-85.

Van Couwenberghe, R. et al. 2013. Can species distribution be used to describe plant abundance patterns? — Ecography 36:665-674.

Waits, J. L. et al. 2003. Low mitochondrial DNA variation in Double-crested Cormorants in eastern North America. — Waterbirds 26: 196-200.

Walther, G. R. et al. 2002.Ecological responses to recent climate change. — Nature 416:389-395.

Warren, D. L. and Seifert, S. N. 2011. Ecological niche modeling in Maxent: the importance of model complexity and the performance of model selection criteria. — Eco Apps. 21:335-342.

Wayne, A. T. 1910. Birds of South Carolina (Vol. 1). — Daggett Printing Company, Charleston, SC.

Wild, D. 2012. The Double-crested Cormorant: Symbol of Ecological Conflict. — University of Michigan Press, Ann Arbor, MI.

Wires, L. R. and Cuthbert, F. J. 2006. Historic populations of the Double-crested Cormorant ( Phalacrocorax auritus ): implications for conservation and management in the 21 st Century. — Waterbirds 29:9-37.

Wires, L. R. et al. 2001. Status of the Double-crested Cormorant ( Phalacrocorax auritus ) in North America. — Final Report to USFWS. http://digitalcommons.unl.edu/usfwspubs/400/.

Wires, L. R. et al. 2005. The Double-crested Cormorant and American White Pelican in Minnesota: a statewide status assessment. Final Report to Minnesota Department of Natural Resources’ Statewide Wildlife Grants Program. Bemidji, MN.

Wires, L. R. et al. 2011. The American White Pelican and Double-crested Cormorant in Minnesota in 2010: Distribution, abundance and population change. Final Report to MN Dept. Nat. Res. State Wildlife Grants Program. Bemidji, MN.

York, P. et al. 2011. A habitat overlap analysis derived from Maxent for Tamarisk and the South-western willow Flycatcher. — Front. Earth. Sci. 5:120-129.

129

Young, K. E. et al. 2013. Estimating suitable environments for invasive plant species across large landscapes: a remote sensing strategy using Landsat 7 ETM+. — Int. J. Biodiv. Cons. 5:122-134.

130

3. CHAPTER IV: INTESTINAL PARASITES OF CULLED DOUBLE- CRESTED CORMORANTS ( PHALACROCORAX AURITUS)

ABSTRACT : Two hundred eighteen Double-crested Cormorants ( Phalacrocorax

auritus ) culled from 11 sites in Alabama, Minnesota, Mississippi, and Vermont were

examined for intestinal helminthes. Every bird we assessed was infected with at least

1 species of trematode with the exception of 3 birds infected with only cestodes, and

2 birds infected only with acanthocephalans. Over 96% of birds carried more than 1

species of intestinal parasites with an average richness of 3 species per bird. The

average intestinal parasite load was 63 worms per bird, with the maximum infection

rate occurring in an individual with 1,488 parasites. We document Echinochasmus sp.

and Ribeiroia sp. infections of Double-crested Cormorants for the first time in the

United States.

Key Words: Double-crested Cormorant, Phalacrocorax auritus , trematode,

Drepanocephalus spathans, Austrodiplostomum ostrowskiae, Hysteromorpha triloba,

Neodiplostomum, Echinochasmus, Ribeiroia, Strigeidae, Amphimerus , Nematoda,

Capillaria , Contracaecum

131

INTRODUCTION

The Double-crested Cormorant, ( Phalacrocorax auritus ) is a large-bodied piscivorous waterbird ubiquitous in North America. Historically, cormorants breeding in the interior of the United States and Canada wintered along the Gulf of Mexico, and coastal colonies moved south during non-breeding seasons (Hatch, 1995; Wires and

Cuthbert, 2006). Many studies have documented parasites of P. auritus ; however, the definitive text listing all groups of parasites in cormorants (Forrester and Spalding, 2003) documents the parasites of primarily the Florida subspecies ( Phalacrocorax auritus floridanus ), a non-migratory bird. Double-crested Cormorants are opportunistic pursuit- divers that forage on the most abundant fishes between 2 and 25 cm in length (Campo et al., 1993; Kirsch, 1995; Fenech et al., 2004). This can be problematic in stocked ponds and aquaculture facilities where cormorants and white pelicans readily consume fish grown for human uses (Jackson and Jackson, 1995; Overstreet and Curran, 2004; King et al., 2012). In natural systems, the most frequently consumed fish species can vary from season to season (Gido and Matthews, 2000; Anderson et al., 2004; Coleman and

Richmond, 2007) and, because fish assemblages (along with their parasites) can vary among water bodies, cormorants eat an assortment of fishes and parasites as they move from one foraging ground to another. P. auritus feed on a high diversity of prey items; thus, the selection of parasites with which they interact is also likely to be vast. Here, we focus on parasites recovered from the intestines of P. auritus from central and eastern

United States.

132

Cormorant species that overlap in range with P. auritus during breeding or wintering seasons (Tables 4.1-4.3) have the potential to share parasites with our focal species (Holmes and Price, 1980; Freeland, 1983; Fallon et al., 2005). P. auritus has 20 known intestinal helminthes in the eastern U.S. and could serve as host to 14 intestinal parasites of other cormorant species found in North America (Tables 4.1- 4.3). We had the opportunity to assess the intestinal parasites of P. auritus collected for diet research or culling activities. Here, we document parasites collected from 218 intestines of P. auritus from various locations of the eastern and central U.S. Birds were culled from 3 locations in Minnesota, 5 locations in Mississippi, 2 locations in Alabama, and 1 location in

Vermont, U.S. (Table 4.4). Based on the distribution of parasites that could infect P. auritus in these regions, we expected to find parasite assemblages similar to those of cormorants in Florida (Hutton and Sogandares-Bernal, 1960; Hutton, 1964; Threlfall,

1982), Texas (Dronen, 2009), Mississippi (Overstreet and Curran, 2004; Doffitt et al.,

2009), central Canada (Chandler and Rausch, 1984; Kuiken et al., 1999; Robinson et al.,

2008; Robinson et al., 2009; Wagner et al., 2012), and Mexico (Violante-Gonzalez et al.,

2011; Garcia-Varela et al., 2012; Tables 4.1-4.3).

METHODS

The U.S. Department of Agriculture Animal Plant Health Inspection Service

(USDA/APHIS) Wildlife Services of Minnesota, the USDA/APHIS National Wildlife

Research Center, and the Leech Lake Division of Resource Management shipped intestines or entire carcasses to Clemson University School of Agricultural, Forest, and

Environmental Sciences. The agencies collecting culled birds froze the carcasses

133

immediately after harvest and in some cases, 70% ethanol was poured down the esophagus to preserve stomach and intestinal contents prior to freezing. Chemical preservation resulted in desiccation and deterioration of the tissues of some intestines, making parasite assessment unreliable. We do not present data from those specimens here.

We processed intestines in wildlife laboratories at Clemson University.

Gastrointestinal tracts were defatted prior to emptying of the contents by stripping the lining of the intestine by hand (Wildlife Necropsy Videos, Southeastern Cooperative

Wildlife Disease Study). We then washed the contents in a 64 µm sieve, and fixed them in buffered formalin or 80% ethanol prior to parasite assessment. Contents were viewed under 3-70x magnification and any whole or partial parasites were removed for identification. Parasites were stored in 80% ethanol prior to identification. We identified recovered parasites to the lowest taxonomic level and a subsample of each species from each site was Carmine stained and mounted for deposition (Gower 1939) at the U. S.

National Parasite Collection.

RESULTS

All 218 intestines of P. auritus assessed for this study contained helminthic parasites with a mean intensity of 63 worms. One bird had a single infection (one worm), and the most highly infected bird carried nearly 1,500 parasites. We document a mean parasite richness of 3.7 species per host with the highest species richness found in birds from Bee Lake and Swamp Roost, Mississippi (Table 4.4). Regional diversity was similar

134

for hosts from Mississippi/Alabama and Minnesota, but cormorants in Vermont have a lower gamma diversity (7; Table 4.4). Here, we document 15 distinct groups of helminthes from P. auritus ; however, positive identification to species was not possible for many groups. Thus, our results underestimate the species richness of P. auritus parasites. Ten parasites were trematodes, only six previously reported in P. auritus .

Trematoda Diplostomidae Austrodiplostomum ostroweskiae Szidat and Nani, 1951

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Swamp Roost, MS: nine of 10 hosts sampled (90%, 174 ± 153, 1-1397); Cat Island, AL: nineteen of 22 hosts sampled (86%, 4 ± 1, 1-12); Port of Columbus, MS: eight of 10 hosts sampled (80%, 10

± 3, 1-24); Whittington Channel, MS: six of 9 hosts sampled (67%, 20 ± 8, 1-48); Bee

Lake, MS: two of 5 hosts sampled (40%, 4 ± 0, 4); Mossy Lake, MS: four of 11 hosts sampled (36%, 4 ± 3, 1-12); Lake Waconia, MN: three of 30 hosts sampled (10%, 1 ± 0,

1-2); Lake Champlain, VT: two of 25 hosts sampled (8%, 3 ± 2, 1-4); Wells Lake, MN: two of 30 hosts sampled (7%, 2 ± 1, 1-2).

Previous records : In Texas (Dronen, 2009).

135

Hysteromorpha triloba Lutz, 1931

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Leech Lake, MN: seven

of 29 hosts sampled (97%, 15 ± 12, 1-88); Bee Lake, MN: three of 5 hosts sampled

(60%, 7 ± 6, 1-18); Whittington Channel, MS: five of 9 hosts sampled (56%, 16 ± 8, 1-

41); Swamp Roost, MS: four of 10 hosts sampled (40%, 13 ± 10, 2-42); Lake

Guntersville, AL: thirteen of 35 hosts sampled (35%, 4 ± 3, 1-34); Wells Lake, MN: eight

of 30 hosts sampled (27%, 3 ± 2, 1-15); Mossy Lake, MS: three of 11 hosts sampled

(27%, 2 ± 1, 1-4); Port of Columbus, MS: two of 10 hosts sampled (20%, 2 ± 1, 1-2);

Lake Champlain, VT: two of 25 hosts sampled (8%, 6 ± 3, 3-8); Lake Waconia, MN: one

of 30 hosts sampled (3%, 1, 1).

Previous records : Reported widely throughout the range of P. auritus (Table 4.1).

Neodiplostomum sp., Lutz 1928

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Bee Lake, MS: four of 5 hosts sampled (80%, 7 ± 4, 1-17); Whittington Channel, MS: one of 9 (11%, 10, 10);

Lake Guntersville, AL: seventeen of 37 hosts sampled (46%, 8 ± 2, 1-32); Swamp Roost,

MS: three of 10 hosts sampled (30%, 83 ± 42, 1-137); Wells Lake, MN: seven of 30 hosts sampled (23%, 3 ± 1, 1-8); Leech Lake, MN: five of 29 hosts sampled (17%, 4 ± 3, 1-

136

16); Lake Waconia, MN: four of 30 hosts sampled (13%, 1 ± 0, 1-2); Mossy Lake, MN: one of 11 (9%, 1, 1).

Previous records : Metacercariae observed in amphibians, adults in Falconiformes and

Strigiformes previously reported in Spain (Sanmartin et al., 2004).

Remarks : Neascus-type adult diplostomids uncommon in waterbird hosts. Bipartite body

form, absence of pseudosuckers, and thick holdfast organ with median slit and

conspicuous genital opening. It is possible that this is the Neascus stage of another

diplostomid. We often documented concurrent infection Neodiplostomum sp. with H. triloba ; however, that species is well documented having a Diplostomulum larval form with pseudosuckers (Hugghins, 1954).

Echinostomatidae Drepanocephalu spathans Dietz, 1909

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Swamp Roost, MS: ten of

10 (100%, 31 ± 11, 4-120), Port of Columbus, MS: ten of 10 hosts sampled (100%, 64 ±

13, 11-124); Bee Lake, MS: five of 5 hosts sampled (100%, 16 ± 7, 1-44); Leech Lake,

MN: twenty-eight of 29 hosts sampled (97%, 100 ± 21, 1-406); Lake Guntersville, AL: thirty-five of 37 hosts sampled (95%, 33 ± 4, 1-100); Lake Waconia, MN: twenty-eight of 30 hosts sampled (93%, 15 ± 3, 1-60); Mossy Lake, MS: ten of 11 hosts sampled

(91%, 39 ± 17, 2-157); Wells Lake, MN: twenty-seven of 30 hosts sampled (90%, 71 ±

13, 1-263); Whittington Channel, MS: eight of 9 (89%, 38 ± 18, 2-128); Lake Champlain,

137

VT: sixteen of 25 hosts sampled (64%, 14 ± 3, 1-42); Cat Island, AL: thirteen of 22 hosts sampled (59%, 5 ± 2, 1-24).

Previous records : Reported in P. auritus subspecies (Threlfall, 1982; Flowers et al.,

2004; Robinson et al., 2008) and P. brasilianus (Montiero et al., 2011)

Remarks : We recovered multiple size classes of ranging from 0.5mm to 12mm in length.

Echinochasmus sp. Dietz, 1909

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Bee Lake, MS: four of 5

hosts sampled (80%, 37 ± 35, 1-143); Mossy Lake, MS: two of 11 hosts sampled (18%, 1

± 0, 1); Lake Guntersville, AL: six of 37 hosts sampled (16%, 11 ± 9, 1-57); Cat Island,

AL: three of 22 hosts sampled (14%, 2 ± 0, 1-2); Lake Champlain, VT: one of 25 hosts

sampled (4%, 30, 30).

Previous records : This genus has been documented in other species of Phalacrocorax

in Mexico and Europe (Table 4.1).

Remarks : Protrusion of the oral sucker gives the anterior region a similar appearance

to Ascocoytle (Phagicola ) sp.; however, we did not find Ascocoytle sp. in our survey.

138

Psilostomidae Ribeiroia ondatrae Travassos, 1939

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Lake Guntersville, AL: two of 37 hosts sampled (5%, 1 ± 0, 1); Lake Waconia, MN: 1 out of 30 (3%), 1, 1).

Previous records : Previous records of R. ondatrae exist for P. auritus auritus in

Canada and P. auritus brasilianus in South America (Drago et al., 2011; Montiero et al.,

2011; Violante-Gonzalez et al., 2011). Other researchers have considered similar

specimens to be Pseudopstilostoma varium (O’Hear et al. 2012, unpublished data), but we do not consider the worms recovered here to belong to that taxon considering Jones et al. (2005) treat Pseudopstilostoma as a genus inquirendum after Lumsden and Zischke

(1963) placed the genus in synonymy with Ribeiroia .

Opisthorchiidae Amphimerus sp. Barker, 1911

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Wells Lake, MN: four of

30 hosts sampled (13%, 1 ± 0, 1); and Lake Waconia, MN: two of 30 hosts sampled (7%,

2 ± 1, 1-2).

Previous Records : Reported in the liver and bile duct of P. auritus in Louisiana by

Pense and Childs (1972).

139

Unknown/Unidentified 1

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance: Leech Lake, MN: one of

29 hosts sampled (3%, 1, 1).

Remarks : This organism came from the gallbladder as it was stained green. It appeared to be similar in structure to the Amphimerus , but was distinct in terms of body length and the size and number of eggs present.

Strigeidae Unknown/Unidentified 2

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Leech Lake, MN: one of

29 hosts sampled (3%, 1, 1).

Remarks : Three strigeids that were unidentifiable due to degradation were collected from an individual cormorant from Minnesota U.S.A. Although these animals strongly resemble the previously reported Strigea falconis from P. auritus (Violante-Gonzalez et al., 2011), we were unable to obtain a positive identification for this particular organism.

140

Sclerodistomoididae Unknown/Unidentified

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Cat Island, AL: one of 22 hosts sampled (5%, 1, 1).

Remarks : A single parasite collected from the intestine of one cormorant.

Nematoda Capillaridae Capillaria carbonis Rudolphi, 1819

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Swamp Roost, MS: eight

of 10 hosts sampled (80%, 5 ± 2, 1-15); Lake Guntersville, AL: twenty-six of 37 hosts

sampled (70%, 2 ± 0, 1-6); Port of Columbus, MS: seven of 10 hosts sampled (70%, 2 ±

1, 1-6); Bee Lake, MS: three of 5 hosts sampled (60%, 2 ± 1, 1-4); Leech Lake, MN:

seventeen of 29 hosts sampled (59%, 2 ± 0, 1-4): Cat Island, AL: twelve of 22 hosts

sampled (55%, 2 ± 0, 1-6); Lake Waconia, MN: sixteen of 30 hosts sampled (53%, 3 ± 1,

1-10); Wells Lake, MN: fifteen of 30 hosts sampled (50%, 2 ± 0, 1-5); Whittington

Channel, MS: four of 9 hosts sampled (44%, 4 ± 2, 1-7); Lake Champlain, VT: seven of

25 hosts sampled (28%, 2 ± 0, 1-4); Mossy Lake, MS: two of 11 hosts sampled (18%, 7 ±

5, 2-12).

Previous records : Many have reported this genus in P. auritus (Table 4.2).

141

Anisakidae Contracaecum rudolphii Rudolphi, 1819

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Lake Waconia, MN: sixteen of 30 hosts sampled (53%, 3 ± 1, 17); Leech Lake, MN: twelve of 29 hosts sampled (41%, 3 ± 1, 1-13); Wells Lake, MN: eleven of 30 hosts sampled (37%, 3 ± 1, 1-

9); Whittington Channel, MS: three of 9 hosts sampled (33%, 1, 1); Lake Champlain, VT: seven of 25 hosts sampled (28%, 1 ± 0, 1-3); Lake Guntersville, AL: nine of 37 hosts sampled (24%, 1 ± 0, 1-3); Swamp Roost, MS: two of 10 hosts sampled (20%, 2 ± 1, 1-

2); Cat Island, AL: two of 22 hosts sampled (9%, 6 ± 5, 1-11); Mossy Lake, MS: one of

11 hosts sampled (9%, 1, 1) .

Previous records : This is a commonly reported species in P. auritus (Table 4.2).

Remarks : We did not attempt to identify Contracaecum specimens to species considering the current contention of the phylogeny within the genus (Szostakowska and

Fagerholm, 2012; Garbin et al., 2011). Whole specimens in ethanol are available for molecular identification in the museum collections.

142

Acanthocephala Polymorphidae Unknown/Unidentified

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Lake Champlain, VT: four of 25 hosts sampled (16%, 3 ± 1, 1-6); Cat Island, AL: three of 22 hosts sampled

(14%, 2 ± 1, 1-3); Whittington Channel, MS: one of 11 hosts sampled (11%, 1, 1).

Remarks : Frozen specimens unsuitable for specific identification because rostellum of

each animal was not everted.

Andracantha sp. Schmidt, 1975

Stage : Adult

Localities, prevalence, mean intensity± SE and abundance : Leech Lake, MN: one of

29 hosts sampled (3%, 1, 1).

Previous records : Reported in P. auritus from Texas (Threlfall, 1982) and Florida

(Fedynich et al., 1997).

Remarks : The rostellum of this specimen was not fully extended upon death, thus, although its appearance is consistent with reports of Andracantha sp., we question its identity. This organism is dissimilar to the more common unknown/unidentified

Acanthocephalan listed previously as the size and body shapes were quite dissimilar.

143

Cestoda Dilipidae

Stage : Adult

Localities, prevalence, mean intensity± SD and abundance : Bee Lake, MS: five of 5 hosts sampled (100%, 3 ± 2, 1-9); Lake Waconia, MN: twenty-seven of 30 hosts sampled

(90%, 35 ± 6, 1-106); Leech Lake, MN: twenty-five of 29 hosts sampled (86%, 8 ± 1, 1-

32); Lake Champlain, VT: twenty-one of 25 hosts sampled (84%, 9 ± 2, 1-36); Mossy

Lake, MS: nine of 11 hosts sampled (82%, 10 ± 6, 1-57);Whittington Channel, MS: seven of 9 hosts sampled (78%, 3 ± 0, 1-5); Swamp Roost, MS: seven of 10 hosts sampled

(70%, 15 ± 12, 1-84); Wells Lake, MN: twenty-one of 30 hosts sampled (70%, 7 ± 2, 1-

25); Port of Columbus, MS: six of 10 hosts sampled (60%, 3 ± 1, 1-6); Cat Island, AL: eight of 22 hosts sampled (36%, 3 ± 0, 1-5); Lake Guntersville, AL: six of 37 hosts sampled (16%, 3 ± 1, 1-4).

Previous records : Reported in P. a. auritus and P. a. floridanus (Table 4.3)

Remarks : Rostellar hooks and reproductive structures were not available as they remained in host tissue or were contracted. Thus, we could not perform positive identification to species. Specimens for molecular identification are available in the museum collection.

In addition to traditional endoparasites, we observed evidence of ectoparasite infection in four cormorants. Exoskeletons of lice were collected from intestinal contents of P. auritus from Alabama, Mississippi, and Minnesota. Whole bird necropsies 144

documented a low incidence of lice infestation at Leech Lake (1 bird out of 59 exhibited feather loss associated with lice); however, no lice were documented in those intestines.

DISCUSSION

Many researchers have documented the intestinal parasites of P. auritus at single localities or within a narrow region of its distribution, often focusing on one particular class of parasite (Hutton, 1964; Threlfall, 1982; Flowers et al., 2004; Dronen, 2009;

Wagner et al., 2012). This is the first study to document parasites from multiple regions within the range of P. auritus . We document infections in P. auritus previously reported in other cormorant species or subspecies. Further, we document the distribution of many parasites as widespread within the host range. In particular, D. spathans was found in all locations sampled in this study.

Furthermore, we reveal the frequency of tapeworm infections of P. auritus .

Although we were unable to positively identify the cestode species from frozen P. auritus , we assert that this is a previously underrepresented parasite group in cormorants and additional studies on unfrozen hosts will elucidate the identity and distribution of P. auritus cestode species.

The richness of parasite species common throughout the range of P. auritus

(Table 4.4) suggests that similar intermediate hosts (suitable of supporting these species) exist in northern breeding, southern wintering, and southern breeding locations. This may speak more to a lack in host specificity for these common parasites, as the likelihood of snail and fish host assemblages being similar among regions is low. Nonetheless, it is

145

possible that that some widely distributed intermediate hosts, such as the bullhead

(Ameiurus sp.), gizzard shad ( Dorosoma cepedianum ), and yellow perch ( Perca flavescens ), function as intermediate hosts to many trematodes and promote infection over a large geographic area (Holl, 1932; Arnold, Jr., 1934; Thomas, 1937; Krueger,

1954; Carney and Dick, 2000; Poulin and Dick, 2007). Alternatively, parasites persisting in the intestine of cormorant hosts between seasons could explain the similarities between winter and summer parasite assemblages. This possibility is particularly interesting because it suggests that interspecific interactions between parasites could exclude

(competition) or promote (facilitation) parasites that are expanding in range (Lello et al.,

2004; Johnson and Buller, 2011) through changes in host distribution.

Distinct diversities of parasites among sites may be useful indicators for targeted sampling in the future. Considering the relatively large geographic area assessed here, a combination of alpha (within-site) and gamma diversity (within-region) evaluations are quite informative. We document higher parasite diversities in birds of the southeast and

Minnesota, and relatively low parasite diversity in birds culled in Vermont. Further exploration of spatial changes of parasite assemblages could clarify competitive dynamics among parasitic fauna.

The high variety of parasites found in P. auritus confirms the generalist feeding habits of this waterbird. The parasite diversity of individual sites may not represent the intermediate host diversity of any particular location, because these birds will often forage in multiple locations around a colony when available. Thus, the suite of parasites

146

within any host represents the local and historical intermediate host populations that they have consumed. Nonetheless, collections of intermediate hosts from localities where particularly diverse or interesting parasite assemblages have been collected could be of interest to conservationists.

147

Tables Table 4.1. Trematodes reported in cormorants from the United States and Canada 15 . TREMATODES P. auritus P.a. floridanus P. carbo P. brasi lianus Ascocoytle sp. * TX 1 TX 1 Austrodiplostomum compactum BRA 2, MEX 3 A. mordax * TX 1 TX 1, BRA 2, ARG 4 A. ostrowskiae * TX 5 ARG 4 Clinostomum attenuatus * FL 7 C. complanatum ‡ MEX 3 Diplostomum compactum * NC 8 Drepanocephalus olivaceus ‡ BRA 2, MEX 3 D. spathans * † NC 8, ON 9, MS 10 , FL 12 BRA 2, ARG 4, TX 1 SK 11 , FL 12 , TX 1 Echinoschasmus coaxatus LTU 13 E. leopoldinae ‡ MEX 3 Euhaplorchis californiensis ‡ MEX 3 Holostephanus dubinini LTU 13 Hysteromorpha triloba * † NC 8, WI & FL 6,7 LTU 13 MS 15 , BRA 2, ARG 4, MB 14 , MS 15 , TX 1 Mehrastomum minutum ‡ FL 12 MEX 3 Mesoophordodisplstomum pricei * TX 1

15 Two-letter abbreviations: Europe, Central, and South America (three-letter country codes: ARG= Argentina, BRA=Brazil, CZE=Czech Republic, ITA=Italy, LTU=Lithuania, MEX=Mexico, POL=Poland). * Parasite species expected based on distribution in southern U.S. † Parasite species expected based on occurrence in north central U.S. and Canada. ‡ Parasite species expected based on distribution in northeastern Mexico. 1(Fedynich et al. 1997), 2(Montiero et al. 2011), 3(Violante-Gonzalez et al. 2011), 4(Drago et al. 2011), 5(Dronen 2009), 6(Hutton and Sogandares-Bernal 1960), 7(Hutton 1964), 8(Flowers et al. 2004), 9(Robinson et al. 2008), 10 (Doffitt et al. 2009), 11 (Wagner et al. 2012), 12 (Threlfall 1982), 13 (Svazas et al. 2011), 14 (Chandler and Rausch 1984), 15 (Overstreet and Curran 2004). 148

Table 4.1. (cont.) Trematodes reported in cormorants from the United States and Canada 16 . TREMATODES P. auritus P.a. floridanus P. carbo P. brasilianus Mesostephanus appendiculatoides *FL 12 FL 7,12 M. splendiculatoides * FL 12 Mesostephanus sp. * FL 7 Metorchis xanthosomus LTU 13 Odhneria raminellae ‡ MEX 3 Parorchis acanthus * FL 12 P. diminuta * FL 6,7 Paryphostomum parvicephalum ARG 4 P. radiatum LTU 13 POL 16 P. segregatum BRA 2, ARG 4 Petasiger phalacrocoracis Europe 13,17,18 Phagicola longa * ‡ FL 12 FL 6,7,12 MEX 3 Phocitremoides butionis TX 1 TX 1 Prosthogonimus ovatus BRA 2, ARG 4 Ribeiroia ondatrae †‡ SK 11 BRA 2, MEX 3, ARG 4 Strigea falconis brasiliana ARG 4 Tylodelphys adulta ARG 4 T. clavata LTU 13

16 Two-letter abbreviations: Europe, Central, and South America (three-letter country codes: ARG= Argentina, BRA=Brazil, CZE=Czech Republic, ITA=Italy, LTU=Lithuania, MEX=Mexico, POL=Poland). * Parasite species expected based on distribution in southern U.S. † Parasite species expected based on occurrence in north central U.S. and Canada. ‡ Parasite species expected based on distribution in northeastern Mexico. 1(Fedynich et al. 1997), 2(Montiero et al. 2011), 3(Violante-Gonzalez et al. 2011), 4(Drago et al. 2011), 6(Hutton and Sogandares-Bernal 1960), 7(Hutton 1964), 11 (Wagner et al. 2012), 12 (Threlfall 1982), 13 (Svazas et al. 2011), 16 (Biedunkiewicz et al. 2012), 17 (Faltynkova et al. 2008), 18 (Nasincova et al. 1994). 149

Table 4.2. Nematodes reported in cormorants from the United States and Canada 17 . NEMATODES P. auritus P.a. floridanus P. carbo P. brasilianus Capillaria carbonis * FL 1 FL 1 ContracaecumC. spiculata * microcephalum TX 2 LTU 3 TX 2 C. multipapillatum ‡ MEX 5 C. rudolphii * † reported C. SK 4,6 , TX 2 FL 7 ITA 8, CZE 9, BRA 10,13 , TX 2 spiculigerum in FL LTU 3, POL 11 C. yamagutii † ON 12,14 Cosmocephalus obvelatus LTU 14 Euccoleus contortus BRA 10 Eustrongyloides sp. LTU 3 Ornithocapillaria appendiculata BRA 10 Skrjabinocara squamatum * FL 1 Syncuaria squamata * † FL 1, SK 4, TX 2 FL 1 ITA 8, LTU 3, BRA 10 , TX 2 Tetrameres sp. * FL 1, TX 2 FL 1 POL 11 BRA 10 , TX 2

17 United States and Canada given with two-letter abbreviations, Europe, Central and South America given with three-letter country codes: BRA=Brazil, CZE=Czech Republic, ITA=Italy, LTU=Lithuania, MEX=Mexico, POL=Poland. * Parasite species expected based on distribution in southern U.S. † Parasite species expected based on occurrence in central Canada. ‡ Parasite species expected based on distribution in northeastern Mexico. 1(Threlfall 1982), 2(Fedynich et al. 1997), 3(Svazas et al. 2011), 4(Wagner et al. 2012), 5(Violante-Gonzalez et al. 2011), 6(Kuiken et al. 1999), 7(Hutton 1964), 8(Dezfuli et al. 2002), 9(Moravec 2009), 10 (Montiero et al. 2011), 11 (Biedunkiewicz et al. 2012), 12 (Robinson et al. 2008), 13 (Torres et al. 2005), 14 (Robinson et al. 2009) 150

Table 4.3. Acanthocephalans and cestodes reported in cormorants from the United States and Canada 18 . ACANTHOCEPHALANS P. auritus P.a. floridanus P. carbo P. brasilianus Andracantha gravida * TX 2 TX 2 A. tandemtesticulata BRA 10 Andracantha sp. * FL 1 FL 1 Corynosoma sp. * FL 7 Polymorhynchus bulbocolli † SK 4 Polymorphous obtusus * FL 1 Southwellina hispida ‡ MEX 3 ITA 8 MEX 5,15 CESTODES Gryporhynchus sp. LTU 3 Ligula colymbi † SK 4 Paradilepis caballenoi *†‡ ON 1,2 , SK 4, BRA 10 , MEX 5, TX 2 P. scolecina † TX 2 LTU 3, POL 11,16 Parvitaenia * FL 1 FL 1 Schistocephalus soldus † SK 4

18 United States and Canada given with two-letter abbreviations, Europe, Central and South America given with three-letter country codes: BRA=Brazil, CZE=Czech Republic, ITA=Italy, LTU=Lithuania, MEX=Mexico, POL=Poland. * Parasite species expected based on distribution in southern U.S. † Parasite species expected based on occurrence in central Canada. ‡ Parasite species expected based on distribution in northeastern Mexico. 1(Threlfall 1982), 2(Fedynich et al. 1997), 3(Svazas et al. 2011), 4(Wagner et al. 2012), 5(Violante-Gonzalez et al. 2011), 7(Hutton 1964), 8(Dezfuli et al. 2002), 10 (Montiero et al. 2011), 11 (Biedunkiewicz et al. 2012), 15 (Garcia-Varela et al. 2012), 16 (Dziekonska-Rynko and Dzika 2011) 151

Table 4.4. Location, collection information, host sample size, and parasite species richness and diversity (within site [alpha] and regional [gamma]) estimates for all culled birds assessed in this study.

Collection Diversity Cull Site State Latitude Longitude Sample Species Season Year Size Richness Alpha Gamma Bee Lake MS 33.0476 -90.3470 Fall 2010 5 10 5.8 13 Cat Island AL 30.3191 -88.2100 Winter 2012 22 7 2.7 13 Lake Guntersville AL 34.3194 -86.3160 Summer 2009 37 10 3.2 13 Lake Champlain VT 44.5866 -73.3800 Spring 2010 25 8 2.4 7 Lake Waconia MN 44.8610 -93.7846 Spring 2010 15 7 2.9 13 Lake Waconia MN 44.8610 -93.7846 Spring 2011 26 8 3.7 13 Leech Lake MN 47.1063 -94.3720 Fall 2010 15 8 3.7 13 Leech Lake MN 47.1063 -94.3720 Spring 2010 16 6 2.9 13 Mossy Lake MS 33.3474 -90.3980 Fall 2010 10 8 2.9 13 Port of Columbus MS 33.4798 -88.4430 Winter 2011 10 5 3.3 13 Swamp Roost MS 33.0320 -91.0800 Spring 2011 10 7 4.3 13 Wells Lake MN 44.2881 -93.3485 Spring 2010 16 7 2.7 13 Wells Lake MN 44.2881 -93.3485 Spring 2011 29 9 3.8 13 Whittington MS 32.9353 -90.5430 Winter 2011 9 8 3.9 13 Channel

152

Acknowledgements

This research would not have been possible without the support of the faculty at

Clemson University and the University of Maryland, particularly Dr. Patrick Jodice and

Dr. William Bowerman. Gary Nohrenberg (MN USDA/Aphis) was kind enough to provide us with whole birds and intestines from birds in Minnesota, and Steve Mortensen

(Leech Lake DCR) allowed us to obtain intestines from birds collected for his diet assessments of cormorants. We appreciate the use of facilities and equipment at Clemson

University, appropriated in part through funds provided by the Clemson Creative Inquiry program.

153

Literature Cited

Anderson, C. D., D. D. Roby and K. Collis. 2004. Foraging patterns of male and female Double-crested Cormorants nesting in the Columbia River estuary. Waterbirds 27:155-160.

Arnold, Jr., J. G. 1934. Some trematodes of the common bullhead Ameiurus nebulosus (Le Sueur). Transactions of the American Microscopical Society 53:267-276.

Biedunkiewicz, A., J., Dziekonska-Rynko, and J. Rokicki. 2012. Black Cormorant Phalacrocorax carbo (L., 1758) as a vector of fungi and parasites occurring in the gastrointestinal tract. Biologia 67:417-424.

Campo, J. J., B. C. Thompson, J. C. Barron, R. C. Telfair, III, P. Durocher and S. Gutreuter. 1993. Diet of Double-crested Cormorants wintering in Texas. Journal of Field Ornithology 64:135-144.

Carney, J. P. and T. A. Dick. 2000. Helminth communities of yellow perch ( Perca flavescens (Mitchill)): determinants of pattern. Canadian Journal of Zoology 78:538-555.

Chandler, A. C. and R. Rausch. 1984. A contribution to the study of certain avian strigeids (Trematoda). The Journal of Parasitology 34:207-210.

Coleman, J. T. H. and M. E. Richmond. 2007. Daily foraging patterns of adult Double- crested Cormorants during the breeding season. Waterbirds 30:189-198.

Dezfuli, S., S. Volponi, I. Beltrami and R. Poulin. 2002. Intra- and interspecific density- dependent effects on growth in helminth parasites of the cormorant, Phalacrocorax carbo sinensis . Parasitology 124:537-544.

Doffitt, C. M., L. M. Pote and T. King. 2009. Experimental Bolbophorous damnificus (Digenea: Bolbophoridae) infections in piscivorous birds. Journal of Wildlife Diseases 45:684-691.

Dolbeer. R. A. 1991. Migration patterns of Double-crested Cormorants east of the Rocky Mountains. Journal of Field Ornithology 62:83-93.

Drago, F. B., L. I. Lunaschi and M. Schenone. 2011. Digenean parasites of the Neotropic Cormorant Phalacrocorax brasilianus (Gmelin, 1789) (Aves: Phalacrocoracidae)

154

from Argentina: Distribution extension and new host records. Check List 7:871- 878.

Dronen, N. O. 2009. Austrodiplostomum ostrowskiae n. sp. (Digenea: Diplostomidae: Diplostominae) from the Double-crested Cormorant, Phalacrocorax auritus (Phalacrocoridae) from the Galveston, Texas area of the Gulf of Mexico, U.S.A. Comparative Parasitology 76:34-39.

Dziekonska-Rynko and J., E. Dzika. 2011. The tapeworm Paradilepis scolecina (rudolphi , 1819) (Cestoda: Cyclophyllidea) invasion in Great Cormorant [Phalacrocorax carbo sinensis (Blumenbach)] from the breeding colony in Lake Selment Wielki (northern Poland). Helminthologia 48:23-28.

Fallon, S. M., E. Bermingham and R. E. Ricklefs. 2005. Host specialization and geographic localization of Avian Malaria parasites: a regional analysis in the Lesser Antilles. The American Naturalist. 165:466-480.

Faltynkova, A., D. I. Gibson and A. Kostadinova. 2008. A revision of Petasiger Diets, 1909 (Digenea: Echinostomatidae) and a key to its species. Systematic Parasitology 71:1-40.

Fedynich, A. M., D. B. Pence and J. F. Bergan. 1997. Helminth community structure and pattern in sympatric populations of Double-crested and Neotropic Cormorants. Journal of the Helminthological Society of Washington 64:176-182.

Fenech, A. S., S. E. Lochman, and A. A. Radomski. 2004. Seasonal diets of male and female Double-crested Cormorants from an oxbow lake in Arkansas, USA. Waterbirds 27:170-176.

Flowers, J. R., M. F. Poore, J. E. Mullen and M. G. Levy. 2004. Digeneans collected from piscivorous birds in North Carolina, U.S.A. Comparative Parasitology 71:243-244.

Forrester, D. J. and M. G. Spalding. 2003. Parasites and diseases of wild birds in Florida . University Press of Florida, Gainsville.

Freeland, W. J. 1983. Parasites and the coexistence of animal host species. The American Naturalist 121:223-236.

155

Garbin, L., S. Mattiucci, M. Paoletti, D. Gonzalez-Acuna and G. Nascetti. 2011. Genetic and morphological evidences for the existence of a new species of Contracaecum (Nematoda: Anisakidae) parasite of Phalacrocorax basilianus (Gmelin) from Chile and its genetic relationships with congeners from fish-eating birds. Journal of Parasitology 97:476-492.

Garcia-Varela, M., F. J. Aznar, R. P. Rodriguez and G. Perez-Ponce de Leon. 2012. Genetic and morphological characterization of Southwellina hispida Van Cleave, 1925 (Acanthocephala: Polymorphidae), a parasite of fish eating birds. Comparative Parasitology 79:192-201.

Gido, K. B. and W. J. Matthews. 2000. Dynamics of the offshore fish assemblage in a southwestern reservoir (Lake Texoma, Oklahoma-Texas). Copeia 2000:917-930.

Gower, W. C. 1939. A modified stain and procedure for trematodes. Biotechnic and Histochemistry 14:31-32.

Hatch, J. J. 1995. Changing populations of Double-crested Cormorants. Colonial Waterbirds 18:8-24.

Holl, F. J. 1932. The ecology of certain fishes and amphibians with special reference to their helminth linguatulid parasites. Ecological Monographs 2:83-107.

Holmes, J. C. and P. W. Price. 1980. Parasite communities: the roles of phylogeny and ecology. Systematic Zoology 29:203-213.

Hugghins, E. J. 1954. Life history of a strigeid trematode, Hysteromorpha triloba (Rudolphi, 1819) Lutz, 1931. II Sporocyst through adult. Transactions of the American Microscopical Society 73:221-236.

Hutton, R. F. 1964. A second list of parasites from marine and coastal animals of Florida. Transactions of the American Microscopical Society 83:439-447.

Hutton, R. F. and F. Sogandares-Bernal. 1960. Studies on helminth parasites from the coast of Florida. II. Digenic trematodes from shore birds of the west coast of Florida. Bulletin of Marine Science of the Gulf and Caribbean 10:40-54.

Jackson, J. A. and B. J. S. Jackson. 1995. The Double-crested Cormorant in the south- central United States: habitat and population changes of a feathered pariah. Colonial Waterbirds 18:118-130.

156

Johnson, P. T. J. and I. D. Buller. 2011. Parasite competition hidden by correlated coinfection: using surveys and experiments to understand parasite interactions. Ecology 92:535-541.

Jones, A., Bray, R. A., and Gibson, D. I. 2005. Keys to the Trematoda: Volume 2. CABI Publishing, London.

Kirsch, E. M. 1995. Double-crested Cormorants along the upper Mississippi River. Colonial Waterbirds 18:131-136

Krueger, R. F. 1954. A survey of the helminth parasites of fishes from Van Buren Lake and Rocky Ford Creek. Ohio Journal of Science 54:277-297.

Kuiken, T., F. A. Leighton, G. Wobeser and B. Wagner. 1999. Causes of morbidity and mortality and their effect on reproductive success in Double-crested Cormorants from Saskatchewan. Journal of Wildlife Disease 35:331-346.

Lello, J., B. Boag, A. Fenton, I. R. Stevenson and P. J. Hudson. 2004. Competition and mutualism among the gut helminths of a mammalian host. Nature 428:840-844.

Lumsden, R. D. and Zischke, J. A. 1963. Studies on the trematodes of Louisiana birds. Zeitschrift für Parasitenkunde 22:316-366.

Montiero, C. M., J. F. R. Amato and S. B. Amato. 2011. Helminth parasitism in the Neotropical Cormorant, Phalacrocorax brasilianus , in southern Brazil: effect of host size, weight, sex, and maturity state. Parasitology Research 109:849-855.

Moravec, F. 2009. Experimental studies on the development of Contracaecum rudolphii (Nematoda: Anisakidae) in copepod and fish paratenic hosts. Folia Parasitologica 56:185-193.

Nasincova, V., T. Scholz and F. Moravec. 1994. Redescription of Petasiger exaeretus Diets, 1909 and P. phalacrocoracis (Yamaguti, 1939) (Trematoda: Echinostomatidae), parasites of cormorants. Systematic Parasitology 27:139-147.

157

Overstreet, R. M. and S. S. Curran. 2004. Defeating diplostomoid dangers in USA catfish aquaculture. Folia Parasitologica 51:153-135.

Pense, D. B. and G. E. Childs. 1972. Pathology of Amphimerus elongus (Digenea: Opisthorchiidae) in the liver of the Double-crested Cormorant. Journal of Wildlife Disease 8: 221-224.

Poulin, R. and T. A. Dick. 2007. Spatial variation in population density across the geographical range in helminth parasites of yellow perch Perca flavescens . Ecography 30:629-636.

Robinson, S. A., M. R. Forbes, C. E. Hebert and J. D. McLaughlin. 2008. Male-biased parasitism by common helminthes is not explained by differences in body size or spleen mass of breeding cormorants Phalacrocorax auritus. Journal of Avian Biology 39:272-276.

Robinson, S. A., M. R. Forbes and C. E. Hebert. 2009. Parasitism, mercury contamination, and stable isotopes ion fish-eating Double-crested Cormorants: no support for the co-ingestion hypothesis. Canadian Journal of Zoology 87:740-747.

Sanmartín, M. L., F. Álvarez, G. Barreiro and J. Leiro. 2004. Helminth fauna of Falconiform and Strigiform birds of prey in Galicia, northwest Spain. Parasitology Research 92:255-263.

Svazas, S., N. Chukalova, G. Grishanov, Z. Putys, A. Sruoga, D. Butkauskas, L. Raudonikis and P. Prakas. 2011. The role of Great Cormorant ( Phalacrocorax carbo sinensis ) for fish stock and dispersal of helminthes parasites in the Curonian Lagoon area. Veterinarian Medical Zoology 55:79-87.

Szostakowska, B. and H. P. Fagerholm. 2012. Coexistence and genetic variability of Contracaecum rudolphi A and Contracaecum rudolphii B (Nematoda: Anisakidae) in cormorants, Phalacrocorax carbo sinensis, in the Baltic Region. Journal of Parasitology 98:472-478.

Thomas, L. J. 1937. On the life cycle of Contracaecum spiculigerum (Rud.). The Journal of Parasitology 23:429-431.

158

Threlfall, W. 1982. Endoparasites of the Double-crested Cormorant ( Phalacrocorax auritus ) in Florida. Proceedings of the Helminthological Society of Washington 49:103-108.

Torres, P., J. Ortega and R. Schlatter. 2005. Nematode parasites of the digestive tract in Neotropic Cormorant checks ( Phalacrocorax brasilianus ) from the River Cruces Ramsar site in southern Chile. Parasitology Research 97:103-107.

Violante-Gonzalez, J., S. Monks, S. Gil-Guerrero, A. Rojas-Herrera, R. Flores-Garza and E. Larumbe-Moran. 2011. Parasitic communities of the Neotropical Cormorant Phalacrocorax brasilianus (Gmelin) (Aves, Phalacrocoracidae) from two coastal lagoons in Guerrero state, Mexico. Parasitology Research 109:1303-1309.

Wagner, B. A., E. P. Hoberg, C. M. Somers, C. Soos, H. Fenton and E. J. Jenkins. 2012. Gastrointestinal helminth parasites of Double-crested Cormorants ( Phalacrocorax auritus ) at four sites in Saskatchewan, Canada. Comparative Parasitology 79:275- 282.

Wires, L. R. and F. J. Cuthbert. 2006. Historic populations of the Double-crested Cormorant ( Phalacrocorax auritus ): implications for conservation and management in the 21 st century. Waterbirds 29:9-37.

159

4. CHAPTER V: PARASITE ASSEMBLAGES OF DOUBLE-CRESTED CORMORANTS AS INDICATORS OF GEOGRAPHICALLY SIMILAR SUBSPECIES

ABSTRACT: The Double-crested Cormorant (DCCO, Phalacrocorax auritus ) is culled

in many states because of the real and presumed damage it inflicts on farmed and

recreational fisheries, as well as other ecosystem services. P. auritus can decrease fish

abundance, and their contemporary dispersal to artificial aquatic features has the potential

to introduce trophically transmitted parasites into food webs that previously had been free

of cormorant influence. Migratory P. auritus encounter a variety of habitats and

intermediate host communities during migration and are likely to contain parasite

communities that differ from those of non-migratory, resident birds. Resident colonies of

cormorants may be re-expanding groups of the Florida subspecies ( P. a. floridanus ),

which are protected in many states from lethal management. Here, we document five

distinct assemblages of helminth parasites collected from 218 P. auritus culled from 11

sites in Alabama, Minnesota, Mississippi, and Vermont. We provide evidence for mixing

of cormorants at a regional scale using Discriminant Analysis, which suggests a single

population of the migratory subspecies ( P. a. auritus ). Furthermore, our models provide

strong support for two subspecies of P. auritus in eastern North America. The

assemblages of P. auritus parasites are distinct among many sampling locations and can

be used to correctly predict where a host cormorant has been feeding based on its

intestinal parasite community. Our models could serve as effective tools for managers

interested in population control of P. a. auritus and conservation of P. a. floridanus ; 160

however, we acknowledge that the use of such models can be challenging because parasite assemblages change with host distribution change across a human-altered landscape. Latitudinal and longitudinal gradients of parasite abundance on a species-by- species basis could be a useful way to determine parasitization risk of fishes and cormorants for a given site based on the current distribution of P. auritus .

Introduction

To reduce human-wildlife conflicts in developed landscapes, legislative acts have given authority to state managers to carry out population control activities on mammals and birds that pose a threat to human health, safety, and apparent wellbeing (50 CFR

21.47 eCFR, FGC §4181, and T14 §401). Consequently, control programs for nuisance wildlife have been initiated, including programs to limit the colony sizes of the Double- crested Cormorant, Phalacrocorax auritus through culling and egg-oiling programs

(Bedard and Lepage 1995, Taylor and Dorr 2003, DeVault et al. 2012, SCDNR News

Release 2013). Management programs of P. auritus have now been developed in South

Carolina, where both migratory and breeding resident birds occur 9SCDNR News

Release 2013). The examination of culled birds for parasites allowed us to assess the spatial distribution of parasite communities in relation to movement patterns of resident and migratory subspecies, which overlap in range in the southeastern U.S.

P. auritus is a colonial, pursuit diving waterbird that primarily consumes forage fish (Brugger 1995, Fenech et al. 2004). Large colonies of P. auritus are highly conspicuous and have been suggested to deplete fish stocks from lakes, rivers, reservoirs,

161

estuaries, and coasts in addition to farmed (aquaculture) aquatic ecosystems (Brugger

1995, Taylor and Dorr, 2003). Colony numbers and densities of P. auritus are higher today than in the last 50 years (Jackson and Jackson 1995), and growing colonies are commonly culled to reduce presumed effects to ecosystem services such as decreased sport fisheries, diminished water quality, and fouling of nesting islands that could serve as sites for recreation. Damage management programs for other wildlife species have been economically and ecologically justified (Basili and Temple 1999, Peer et al. 2003,

Sijtsma et al. 2013); however, scientific evidence of reduced ecosystem services associated with cormorant colonies is lacking (Erwin 1995).

Like many other avian top-predators, P. auritus experienced population

bottlenecks in the middle of the 20 th century, declining in numbers from millions to a few thousand (Kirsch 1995, Krohn 1995, Wires and Cuthbert 2006). The legal protection of migratory birds ( P. auritus was added to the Migratory Bird Act in 1972 (eCFR 2013))

and the prohibition of persistent organic pesticides such as DDT contributed to the

successful rebound of P. auritus to over a million birds in North America (Hatch 1995,

Wires et al. 2001). Four distinct subspecies of P. auritus have been described: P. a. auritus widespread throughout the interior and east coast of North America south to the

Gulf of Mexico, Caribbean, and Mexico; P. a. cincinatus in Alaska; P. a. albociliatus of

coastal California; and P. a. floridanus in the southeastern United States (Dolbeer 1991,

Wires and Cuthbert 2006). Based on winter sightings of banded birds, Dolbeer (1991)

suggested that two distinct populations of P. a. auritus breed in the interior (migrating

along the Mississippi flyway) and Atlantic coast of North America. However, molecular

162

studies do not support multiple populations of any of the subspecies and indicate that only three subspecies exist (Alaskan, Pacific coast, and interior/eastern North American); confirming regular genetic exchange between each region and subspecies therein.

Furthermore, the lack of molecular differentiation suggests nesting behavior is the factor that separates migratory and resident P. auritus subspecies of eastern North America

(Mercer et al. 2013). P. a. auritus is observed to overlap in distribution with P. a. floridanus during the winter months as migratory birds travel to the southeastern U.S.

Double-crested Cormorants of the interior and Atlantic coast continue to migrate between breeding seasons with notable exceptions associated with the creation of freshwater reservoirs and aquaculture facilities. These facilities have provided opportunities for wintering birds to reduce migration distances, and in some cases, resident colonies now persist year-round in the southeastern U.S. Many of these locations are within the historical range of P. a. floridanus , and Wires and Cuthbert (2006) suggest that these ‘new’ resident colonies are the recovering Florida subspecies rather than P. a. auritus . Genetic assessments have not confirmed this supposition (Waits et al. 2003,

Green et al. 2006), and we ask whether parasites can be used as biomarkers to distinguish among P. auritus from different foraging sites and regions in addition to breeding locations.

In this study, we examine whether internal helminthic parasites differ among foraging groups of P. auritus and, if so, whether they can be used to identify the regional breeding locations and migratory status of individual hosts. If parasite assemblages differ

163

among breeding regions, it is possible that distinct populations of coastal and interior P. a. auritus occur. Similarly, if parasite assemblages of resident P. auritus are distinct from those observed in migratory birds, it is likely that parasites could be used to differentiate between P. a. auritus and P. a. floridanus . We first describe the sampling methodology and assessment of parasite communities, then define distinct assemblages of parasites, and use predictive models to assign hosts to a site, region, migration status, and parasite assemblage. Finally, we test our models based on predicted versus actual characteristics of each host.

Methods

Sample Processing

Cormorant intestines (from base of proventriculus to cloaca) were assessed for the presence of helminthic parasites. The USDA/APHIS Wildlife Service of Minnesota, the

Leech Lake Natural Resource Division, and the USDA/APHIS National Wildlife

Research Center collected and assisted in preparation of frozen intestines of P. auritus , which were collected from 11 sites from 2010 to 2012 (Table 5.1). Two sites included in our surveys were previously reported resident breeding sites of P. auritus (Swamp Roost,

Mississippi and Lake Guntersville, Alabama, U.S.A.; Hanson et al. 2010), and five sites were established northern breeding colonies (Minnesota and Vermont, U.S.A.). The remaining five locations had not been documented as P. auritus breeding sites

(Mississippi and Alabama, U.S.A.). A total of 218 intestines were assessed. Intestines were frozen prior to parasite assessment and defatted before the thawed contents were removed by cutting down the length of the intestine and stripping the contents and lining

164

of the intestine from the tissue. We washed intestinal contents in a 64 µm sieve and fixed

parasites either in 70% ethanol or buffered formalin. Parasites were identified to the

lowest possible taxonomic level based on previously reported parasites of the

Phalacrocorax genus in North America (Chapter V has a list and description of parasites

identified).

We analyzed parasite data at three levels of resolution, raw counts (intensities) of

parasites in individual birds, ranks of relative abundance (1 most common to 7 least

common) and presence or absence of each parasite species (prevalence). Raw counts

include maximum sample information, but this includes many 0s and some very large

numbers, making parametric analyses challenging because of unequal variance.

Prevalence omits all information except presence or absence, but can be simplest to

analyze. Rank data are more appropriate here and allow for non-parametric tests of

concordance. It was important that we establish the appropriate level of resolution in

parasite data to assess whether parasites could be used as biomarkers of host movement.

Geographic Trends of Parasite Abundance

To assess for changes in parasite distribution over the geographic range of P.

auritus sampled in this study, we used regression analysis to test for linear and quadratic

associations between infection intensity and prevalence with latitude and longitude. Note

that these are the highest and lowest resolution measures that we considered.

165

Parasite Community Assessments

We used a 1,000-permutation iterated multivariate Analysis of Variance (Adonis in the vegan package of R© version 2.0-9; McArdle and Anderson 2001, Anderson 2001,

Oksanen et al. 2013) on untransformed intensity data to test for significant differences in parasite assemblages among 11 sites. Using intensity data ensured that the assemblage analysis accounted not only for differences in the parasite composition of a site, but also the magnitude of infection. To determine how parasite assemblages differed, we subsequently ran pairwise tests for all sites using the same methodology as the whole- system test (1,000-permutation Adonis; McArdle and Anderson 2001, Anderson 2001).

This technique combines correlation matrices of response variables (parasite data) and sampling information (latitude and longitude, sample year etc.; Table 5.1), thus, it attempts to correct for sampling autocorrelation among sites. Results of the Adonis analysis were confirmed with a Mantel test (Oksanen et al. 2013). Significant differences among pairwise tests were compiled into a binary matrix using the most conservative

(highest) p-value of the Mantel and Adonis analyses. The binary matrix was used in a

Cluster Analysis (Average Hierarchical method using JMP® Pro 10.0.0) to combine sites into distinct parasite assemblages. Assemblages identified as clusters were assessed qualitatively for confirmation of similarities or differences. Sites that did not differ (at an alpha level of 0.05) from at least 4 other sites were considered to be similar to other parasite assemblages and were not assigned to specific assemblages.

166

Predictive Models Using Parasite Data

We performed Regularized Discriminant Analyses (DA; lambda = 1, gamma = 1) with stepwise selection using JMP® Pro 10.0.0 to build predictive models to identify P. auritus groups (sample site, region, migratory status, parasite assemblage) based on the parasite assemblages of individual hosts. We performed this assessment on each resolution of data (intensity, ranked abundance, and prevalence). Separation of host points in multivariate space was poor when using intensity data (Appendix C Figure C.1), but ranked intensity data (Figure 5.1) and prevalence (Appendix C Figure C.2) revealed improved point separation.

Ranked data is more likely to meet or successfully transform to meet assumptions of parametric distribution (Kanno and Vokoun 2008). Nonetheless, having many rare species in computations will ultimately skew the distribution of a dataset by introducing an abundance of zeros. To overcome this issue, we removed uncommon species from our raking system. To qualify for removal, a species must have met at least two of three criteria: 1) a parasite was never the most abundant species in any individual bird, 2) the prevalence among all hosts was ≤ 1%, or 3) it occurred in a single host. These criteria

allowed us to exclude five uncommon parasites, leaving eleven types of helminthes for

our assessments using ranked data.

We assessed DA model outputs by comparing the proportions of correctly identified hosts. A relatively high degree of misidentification does not necessarily indicate that a model is not useful in future evaluations. In the absence of prior

167

information, we might expect to correctly assign a bird to migratory vs. non-migratory status 50% of the time if the parasites were equally common among these groups, and if our model is more accurate than that, it could prove useful to managers interested in conservation or population control of cormorants. Furthermore, since we sampled 11 distinct sites, we only have to be more accurate than 9.1% of the time to beat the null expectation. To test model success vs. random predictions, we used Fisher’s exact and

Chi-square analyses on actual vs. predicted contingency table outputs from each model.

Results

The majority of parasites we found were digenetic trematodes, which were in

87% of the intestinal tracts examined (Table 5.2). We also observed Cestodes (65% prevalence), nematodes (53% prevalence), and acanthocephalans (4% prevalence) in the intestines of P. auritus (Chapter V). Additionally, ectoparasites (feather lice) were recovered from the intestinal contents of some hosts (4% prevalence).

Geographic Trends of Parasite Abundance

Four species of trematode changed significantly in intensity or prevalence with latitude and/or longitude (Table 5.3). Drepanocephalus spathans intensity increased with latitude and decreased with longitude and prevalence was highest in the west.

Hysteromorpha triloba decreased in prevalence with latitude. Neodiplostomum sp. infections decreased in occurrence with both latitude and longitude. Prevalence of

Echinochasmus sp. and Austrodiplostomum ostrowskiae increased from north to south and east to west. Amphimerus sp. prevalence increased with latitude. Cestodes prevalence

168

and intensity increased with latitude and prevalence increased with longitude, with more parasites found in the east. The nematode Contracaecum rudolphi was more common in the east and north (Table 5.3). Acanthocephalan prevalence was highest in the east.

Parasite Assemblage Analysis

When all sites were considered in the same model, the model developed with

Adonis program in R indicated that multiple parasite assemblages of P. auritus exist among the 11 sites we sampled (p < 0.0001). Cluster analysis of the binary matrix that we created from the output of pairwise Adonis assessments (0 for significant differences, 1 for non-significant differences; Appendix B Table C.1) revealed five distinct parasite community assemblages across 11 sites. Two sites, Wells Lake and Lake Waconia, MN, shared assemblages with the other five groups and were not assigned to a distinct parasite assemblage. We compared the prevalence and average infection intensity (when infections were found) of each parasite for hosts within each community assemblage and all other hosts combined (Table 5.4).

The first distinct parasite assemblage occurred at Lake Guntersville, AL, where P. auritus are year-round residents. Distinctive characteristics of this parasite community include low occurrence of cestodes and A. ostrowskiae, higher prevalence of H. triloba,

Neodiplostomum sp., C. carbonis, Echinochasmus sp. and arthropods, and the presence of a rare parasite, Ribeiroia sp. Cestodes were significantly less common (16% prevalence) at Lake Guntersville when compared to all other sites combined (96%, p<0.0001).

Average intensity of cestode infections was also lower at Lake Guntersville (1 worm /

169

host) when compared to intensities at other sites (13 worms / host, p = 0.009). A. ostrowskiae was absent from Lake Guntersville. H. triloba was more prevalent at Lake

Guntersville (35% of hosts infected) when compared to birds from all other sites (17% prevalence, p = .0347), as was Neodiplostomum sp. (46% at Guntersville vs. 17% elsewhere, p < 0.0001), C. carbonis (70% prevalence at Lake Guntersville vs. 50% prevalence elsewhere, p = 0.0263), and Echinochasmus sp. (16% of hosts infected at

Lake Guntersville vs. 6% elsewhere p = 0.0230). Feather lice were recovered from intestines significantly more frequently at Lake Guntersville (11% vs. 3% elsewhere, p =

0.0249). Ribeiroia sp. was also collected from Lake Waconia; however, that parasites assemblage of Lake Waconia was statistically indistinguishable from all other sampling locations and was not assigned to a parasite community assemblage.

Leech Lake, MN also had a statistically unique parasite assemblage.

Characteristics of this parasite community include high intensities of D. spathans (97 worms / host vs. 29 elsewhere, p < 0.0001), the absence of Echinochasmus sp. and A. ostrowskiae , and the presence of two unique strigeids that we found at no other sites

(Table 5.4). At Leech Lake, 97% of birds carried D. spathans . At all other sites

combined, 86% of birds contained D. spathans (p < 0.0001). There were only three sites that lacked Echinochasmus sp.: Leech Lake, Wells Lake, and Lake Waconia. All three of these sites are located in Minnesota and Wells Lake and Lake Waconia had parasite assemblages that were similar to multiple parasite assemblages. We were unable to positively identify the rare strigeid parasites to species.

170

Two sites in Mississippi grouped together in a unique parasite assemblage, Mossy

Lake and Bee Lake. These sites were dissimilar in terms of parasite prevalence and intensity, but the communities of parasites present were nearly identical (Table 5.4). One notable difference between the parasite communities of these two sites was the presence of two unique unidentified species that were recovered from two birds from Bee Lake.

Characteristics of the Mossy Lake/Bee Lake parasite assemblage were higher prevalences of cestodes (93% infection rate compared to 63% elsewhere, p = 0.0175) and

Echinochasmus sp. (40% prevalence in this assemblages versus 5% for all other sites combined, p < 0.0001). Although occurrence of Capillaria sp. infection was similar in the Mossy Lake/Bee Lake assemblage (33% prevalence) when compared to other assemblages (55% of hosts infected, p = 0.1026), the intensities of Capillaria sp. were significantly higher at Mossy Lake/Bee Lake (4 worms/host) than those documented at other sites (2 worms/host, p = 0.0355).

Three of the seven sites from the Southeast region (Cat Island, AL, Swamp Roost,

MS, and Port of Columbus, MS) grouped into a single parasite assemblage. This suite of parasites is characterized by high intensities yet lower prevalence of the trematode

Neodiplostomum sp., high prevalence of the diplostomid trematode A. ostrowskiae, and low occurrences of the nematode C. rudolphi, and Cestode parasites (Table 5.4).

Neodiplostomum sp. infected only 7% of hosts from these sites but infected 22% of hosts elsewhere (p = 0.0266). However, when birds at these sites were infected with

Neodiplostomum sp., they harbored significantly more intense infections (83 worms/host) than birds at other sites (6 worms/host, p <0.0001). A. ostrowskiae infected a larger

171

proportion of hosts from these sites (86%) when compared to P. auritus from all other sites combined (7% prevalence, p < 0.0001). We also documented lower prevalence of C. rudolphi (10%) in the Cat Island/Swamp Roost/Port of Columbus assemblage when compared to other sites (34% of hosts infected, p = 0.0019). Similarly, lower prevalence of Cestode parasites occurred at these sites (50% in this assemblage versus 69% elsewhere, p = 0.0219).

The final parasite assemblage grouped P. auritus from Lake Champlain, VT and

Whittington Channel, MS together. The most striking difference between this assemblage

and all others was the high incidence of acanthocephalan parasites (15% prevalence vs.

2% elsewhere, p = 0.0003). We found fewer infections of D. spathans in hosts at Lake

Champlain and Whittington Channel (71% infection vs. 90% elsewhere, p = 0.0016) and

lower intensities (15 worms / host compared to 42 worms / host elsewhere, p = 0.0163).

We also documented lower prevalence of Neodiplostomum sp. (3% infected vs. 22%

elsewhere, p = 0.0085) and Capillaria sp. (32% prevalence vs. 58% elsewhere, p =

0.0065) in this assemblage. Cestodes were found in more hosts from Lake Champlain and

Whittington Channel (62% prevalence) than P. auritus examined from other sties (62%

prevalence, p = 0.0218).

Predictive Models Using Parasite Data

Models using intensity data consistently underperformed (correctly classifying

only 37% of hosts; Table 5.5) when compared to prevalence and ranked intensity data for

sample site (prevalence data successfully predicted 49% of hosts and ranked data

172

correctly classified hosts 46% of the time). Ranked data used to classify hosts grouped by region was correct 73% of the time. When P. auritus were classified based on migration

status, rank data yielded correct identification of host status 78% if the time (Table 5.5).

Biplots depict canonical separation of ranked data for individual hosts for site (Figure

5.1), region (Figure 5.2), migration status (Figure 5.3), and parasite assemblages (as

identified by the Adonis analyses; Figure 5.4) in ordinate space. Results of classifications

are summarized in tables based on the proportion correct classification for each

assessment (see Appendix C Table C.2 = site prediction, Table C.3 = region prediction,

Table C.4 = migration, Table C.5 = parasite assemblage data). We assessed model

performance by arranging actual and predicted group assignments into contingency tables

(represented in mosaic plots: see Appendix C Figure C.3 = site, Figure C.4 = region,

Figure C.5 = migration status, Figure C.6 = parasite assemblage assignments) with

Fisher’s Exact and Chi-Square tests. These analyses indicated that all of our models

significantly outperformed random classification (p < 0.0001; Table 5.6).

Discussion

Our methodology is effective for detecting host characteristics such as specific site of collection, regional site of collection, and migration status using parasite assemblages. Important to the success of our models was the proper treatment of data: we used ranked intensity data for parasites that were common to at least two sites, two birds, or were ranked as the most common parasite in at least one bird. Using unranked intensity data or presence/absence data was not effective in our predictive assessments because intensity data could not meet assumptions of normal distribution and equal

173

variance required for the Adonis analysis we used, and presence/absence data omitted information regarding relative abundance. Converting count data into ranked data has many advantages although, in doing so, some data regarding the magnitude of infection intensity is lost. Ranked intensity data is still preferred over prevalence data because converting infection information to presence/absence removes even more information as all intensity data is lost. We were able to confirm that intensity was important in determining the assemblages of parasites through univariate assessments among parasite assemblages (Table 5.4). This type of data treatment is less common in community assessments (Lennon et al. 2004, Kanno and Vokoun 2008, Mazaris et al. 2013), but we suggest it be considered as an alternative to count data.

We were able to detect unique parasite assemblages for two resident colonies of

P. auritus: breeding birds from Lake Guntersville, AL and wintering birds from Swamp

Roost, MS (Hanson et al. 2010). Cat Island, AL was included in the assemblage for Port of Columbus and Swamp Roost, MS, but is ecologically distinct from the Mississippi sites as it is an island located in the marine waters of the Mississippi Sound. Cat Island provides nesting habitat to other (non-P. auritus ) waterbirds in summer and is a roosting site for many species in winter (K. L. Sheehan, personal observation). Based on the parasite assemblage that includes Cat Island and two inland freshwater sites in

Mississippi, it is possible that southern breeding cormorants forage along the Gulf of

Mexico in winter or that these sites attract previously migratory birds to southern breeding colonies.

174

Our models correctly classified the largest proportion of hosts (78%) when predicting migratory status (migratory vs. resident), suggesting differences in the fish assemblages or feeding habits of migratory and resident P. auritus . This is compelling evidence for separation of subspecies based on nesting behavior that molecular studies have been unable to document. Many researchers have defined parasite communities based on host phylogeny (Price 1989, Arneberg et al. 1998); here we use parasite communities to suggest host sub-speciation. Differences in parasite communities are indicative of distinct feeding ecologies of migratory and resident host subspecies.

We based our migratory status assignments of sites in Mississippi on communications with managers who were involved in culling programs and familiar with breeding and wintering activities at those sites. Because some birds were culled during non-breeding seasons (winter months), when both resident and migrant birds could be present, they were not used to build the model for migration status. Because Discriminate

Analysis allows for classification of individuals not included in model development, it is a useful tool to identify the breeding status of birds collected outside of the breeding season. Interestingly, Swamp Roost, where birds were collected in winter, is a summer nesting site for resident cormorants, but was classified as a foraging group of migratory

P. auritus . Bee Lake was the only site classified as having resident cormorants in winter.

We received only 5 hosts from Bee Lake, and the unique classification might be attributed to low sampling size, but it is possible that this site serves as a foraging ground for resident P. auritus in winter. Predictions for all of the Mississippi sites were poor regarding migration status, suggesting that although these individual sites carried distinct

175

assemblages of parasites, they were likely to be comprised of migratory and resident birds, which could contribute to unique combinations of parasite communities. Larger datasets of parasite assemblages from resident and migratory P. auritus will help to

determine the identity of wintering birds with more confidence. This assessment could be

particularly useful for managers interested in conserving groups of P. a. floridanus, while controlling foraging groups if P. a. auritus in the southeastern U.S.

Regional grouping performed similarly to migration status, correctly classifying

73% of P. auritus , which is significantly better than random assignment would predict

(null = 67% misclassification, p<0.0001). The majority of misclassifications occurred for

Lake Champlain. This site was the only collection in the northeast region and shared parasite assemblage characteristics with Whittington Channel in Mississippi, located in the Southeast region. Acanthocephalan prevalence was high at these sites (>10% infected hosts), although these parasites were also documented in the two Alabama sites. Other misclassifications for region occurred between sites in the Southeast and Northcentral

(Minnesota) regions. Similarities in parasite assemblages between Cat Island (Southeast) and Lake Waconia (Northcentral) contributed to misclassifications (Appendix C Table

C.3), as did similarities in parasite assemblage between Lake Guntersville (Southeast),

Bee Lake (Southeast), and Wells Lake (Northcentral). Interestingly, Swamp Roost birds

(Southeast) were incorrectly classified to the northeast nearly as often as they were correctly classified (Table C.3). The inability of our models to distinguish birds of the

Northeast from those of the Southeast and Northcentral regions suggests that there is a single population of P. a. auritus . This is contrary to findings from banding recovery

176

records (Dolbeer 1991), and could be investigated further by assessing additional sites from New England and the Northcentral region of North America for confirmation.

P. auritus groupings from Discriminant Analyses misclassified host site of collection half of the time. While this may be a less than desirable outcome, it performs significantly better than would be expected if classification were based on random assignment (correct prediction for 1 of 11 attempts; p < 0.0001). Classifying sites based on parasite assemblages identified by Adonis Analysis revealed better model performance in terms of predictive power (misclassification 30% of the time, p < 0.0001) than grouping by site alone. This suggests that groups of sites can provide comparable fish assemblages with similar parasite assemblages. Pooling sampling sites in this way could be very informative for researchers interested in documenting the successful introduction of parasites into aquatic systems as newly formed wintering and breeding colonies are established. Consequently, because parasite communities may change with fluctuations in the geographic distribution of P. auritus and other waterbirds that might serve as competent hosts of cormorant parasites, the use of historical parasite assemblages might not be useful for comparison with contemporary samples.

General trends in parasite presence and abundance along latitudinal and longitudinal gradients can also prove informative for predicting the likelihood of parasitization for intermediate (invertebrate and fish) and avian definitive hosts. Larger sampling efforts of P. auritus and other waterbirds that might serve as hosts to cormorant parasites will further clarify these trends in host and parasite distribution. Our findings

177

have established clear linkages between cormorant parasite assemblages and their associated migratory patterns. As future research efforts further clarify how birds group in space and time, parasite assemblages could become a useful tool in characterizing cormorant colonies and understanding associated control or conservation options. Using parasite assemblages to understand how P. auritus use fish resources could inform managers and conservationists interested in properly managing cormorant colonies, while protecting ecologically important resources.

178

Tables Table 5.1. Sampling sites and characterizations used in univariate and multivariate analyses.

Site State Latitude Longitude Year Bee Lake MS 33.048 -90.347 2010 Cat Island AL 30.319 -88.210 2012 Lake Champlain VT 44.587 -73.380 2010 Lake Guntersville AL 34.319 -86.316 2009 Leech Lake MN 47.106 -94.372 2010 Mossy Lake MS 33.347 -90.398 2010 Port of Columbus MS 33.480 -88.443 2011 Swamp Roost MS 33.032 -91.080 2011 Lake Waconia MN 44.861 -93.785 2010 & 2011 Wells Lake MN 44.288 -93.349 2010 & 2011 Whittington Channel MS 32.935 -90.543 2011

179

Table 5.2. Number (sum), intensity (±SE; avg. int.), rank occurrence (avg. rank), and prevalence (prev.) of each parasite.

Lake Leech Wells Mossy Bee Cat Swamp Lake Whittington Port of Lake Parasite Guntersville Lake Lake Lake Lake Island Roost Champlain Channel Columbus Waconia sum 1168 2803 1926 391 78 63 305 219 304 640 429 avg int 33 ± 4 100 ± 21 71 ± 3 39 ± 17 16 ± 7 5 ± 1 31 ± 11 14 ± 3 38 ± 17 64 ± 13 15 ± 3 D. spathans avg rank 1 1 2 2 2 4 2 3 2 1 2 prev 95% 97% 90% 91% 100% 59% 100% 64% 89% 100% 93% sum 56 108 25 6 20 0 50 11 78 3 1 avg int 4 ± 2 15 ± 6 3 ± 1 2 ± 1 7 ± 5 13 ± 7 6 ± 1 16 ± 7 2 ± 0 1 ± 0 H. triloba avg rank 6 6 6 6 4 7 6 7 5 6 7 prev 35% 24% 27% 27% 60% 0% 40% 8% 56% 20% 3% sum 140 21 21 1 28 0 250 0 10 0 5 avg int 8 ± 2 4 ± 1 3 ± 1 1 ± 0 7 ± 3 83 ± 30 10 ± 3 1 ± 0 Neodiplostomum sp. avg rank 5 6 6 7 3 7 6 7 7 7 7 prev 46% 17% 23% 9% 80% 0% 30% 0% 11% 0% 13% sum 12 188 140 93 13 21 102 199 24 15 954 avg int 2 ± 0 8 ± 1 7 ± 1 10 ± 6 3 ± 2 3 ± 1 15 ± 10 9 ± 2 3 ± 1 3 ± 1 35 ± 6 Cestoda avg rank 6 3 3 3 3 5 4 3 4 5 2 prev 16% 86% 70% 82% 100% 36% 70% 84% 78% 60% 90% sum 51 29 26 14 7 21 42 14 17 15 41 avg int 2 ± 0 2 ± 0 2 ± 0 7 ± 3 2 ± 1 2 ± 0 5 ± 2 2 ± 0 4 ± 1 2 ± 1 3 ± 1 Capillaria carbonis avg rank 4 5 5 6 5 4 4 6 5 4 5 prev 70% 59% 50% 18% 60% 55% 80% 28% 44% 70% 53% sum 12 32 33 1 0 12 3 9 3 0 44 Contracaecum avg int 1 ± 0 3 ± 1 3 ± 1 1 ± 0 6 ± 2 2 ± 0 1 ± 0 1 ± 0 3 ± 1 rudolphi avg rank 6 5 5 7 7 7 7 6 6 7 5 prev 24% 41% 37% 9% 0% 9% 20% 28% 33% 0% 53% sum 2 0 0 0 0 0 0 0 0 0 1 Ribeiroia sp. avg int 1 ± 0 1 ± 0 prev 5% 0% 0% 0% 0% 0% 0% 0% 0% 0% 3% sum 13 0 9 0 3 0 0 0 0 0 1 avg int 3 ± 0 3 ± 1 3 ± 1 1 ± 0 Arthropoda avg rank 7 7 7 7 6 7 7 7 7 7 7 prev 11% 0% 10% 0% 20% 0% 0% 0% 0% 0% 3%

180

Table 5.2. (cont). Number (sum), intensity (±SE; avg int), rank occurrence (avg rank), and prevalence (prev) of each parasite.

Lake Leech Wells Mossy Bee Cat Swamp Lake Whittington Port of Lake Parasite Guntersville Lake Lake Lake Lake Island Roost Champlain Channel Columbus Waconia Echinochasmus sp. sum 68 0 0 2 149 5 0 30 0 0 0 avg int 11 ± 4 1 ± 0 37 ± 2 ± 0 30 ± 6 avg rank 6 7 7 6 4 6 7 7 7 7 7 prev 16% 0% 0% 18% 80% 14% 0% 4% 0% 0% 0% A. ostroweskiae sum 0 0 3 16 8 77 1567 5 120 80 4 avg int 2 ± 0 4 ± 2 4 ± 2 4 ± 1 174 ± 3 ± 1 20 ± 8 10 ± 3 1 ± 0 avg rank 7 7 7 5 5 2 2 7 4 3 7 prev 0% 0% 7% 36% 40% 86% 90% 8% 67% 80% 10% Unknown sum 0 0 0 0 1 0 0 0 0 0 0 Opisthorchiidae avg int 1 ± 0 prev 0% 0% 0% 0% 20% 0% 0% 0% 0% 0% 0% Unknown Strigeidae sum 0 0 0 0 1 0 0 0 0 0 0 avg int 1 ± 0 prev 0% 0% 0% 0% 20% 0% 0% 0% 0% 0% 0% Acanthocephala sum 2 0 0 0 0 5 0 12 1 0 0 avg int 2 ± 0 2 ± 0 3 ± 1 1 ± 0 avg rank 7 7 7 7 7 6 7 6 7 7 7 prev 3% 0% 0% 0% 0% 14% 0% 16% 11% 0% 0% Unknown Fluke sum 0 0 0 0 0 0 0 0 0 0 1 avg int 1 ± 0 prev 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 3% Unknown Strigeidae sum 0 3 0 0 0 0 0 0 0 0 0 avg int 3 ± 1 prev 0% 3% 0% 0% 0% 0% 0% 0% 0% 0% 0% Unknown sum 0 3 0 0 0 0 0 0 0 0 0 Sclerodistomoididae avg int 3 ± 1 prev 0% 3% 0% 0% 0% 0% 0% 0% 0% 0% 0% Amphimerus sp. sum 0 0 4 0 0 0 0 0 0 0 3 avg int 1 ± 0 2 ± 0 avg rank 7 7 7 7 7 7 7 7 7 7 7 prev 0% 0% 13% 0% 0% 0% 0% 0% 0% 0% 7%

181

Table 5.3. P-values of regression analysis of parasite presence/absence (prevalence) and abundance (intensity) over geographic gradients (latitude and longitude) 19 .

Latitude Longitude Prevalence Intensity Prevalence Intensity D. spathans 0.4396 0.0015(+) <0.0001 (–) 0.0001(–) 0.0457 0.0675 0.0666 H. triloba 0.0277* (–) 0.7448 0.2996 0.4456 0.0328 Neodiplostomum sp. 0.0065* (–) 0.0799 0.0108* (–) 0.8819 0.0458 0.0412 Cestoda <0.0001 (+) 0.0135 (+) <0.0001* (+) 0.0069* (+) 0.1359 0.0427 0.2328 0.0692 Capillaria carbonis 0.1897 0.1179 0.0163* (–) 0.6431 0.0376 Contracaecum rudolphi <0.0001(+) 0.4624 0.0006* (+) 0.1305 0.0748 0.0670 Ribeiroia sp. 0.6669 NA 0.9836 NA

Arthropoda 0.7287 0.6718 0.6070 0.6604

Echinochasmus sp. 0.0002 (–) 0.6287 0.0009* (–) 0.9489 0.0639 0.0629 A. ostroweskiae <0.0001 (–) 0.8467 <0.0001* (–) 0.5379 0.2843 0.1186 Unknown Opisthorchiidae 0.3117 NA 0.8156 NA

Unknown Strigeidae 0.3117 NA 0.8156 NA

Acanthocephala 0.3319 0.2341 0.0004 (+) 0.2059 0.0570 Unknown Fluke 0.3820 NA 0.4352 NA Unknown Strigeidae 0.2173 NA 0.3822 NA Unknown Sclerodistomoididae 0.2173 NA 0.3822 NA

Amphimerus sp. 0.0430 (+) 0.1778 0.0680 0.1778 0.0188

19 R – square of significant relationships are listed below p-values, and results derived from quadratic regression given with asterisks. Positive (+) and negative (–) associations given in parentheses. 182

Table 5.4. P-values and R-square values for t-test analyses of parasite prevalence and intensity (if applicable) of P. auritus parasite assemblages compared to all other sites.

Cat Island Lake Mossy Lake Swamp Roost Champlain Parasite Guntersville Leech Lake Bee Lake Port Columbus Whittington D. spathans 0.139 0.1053 0.461 0.0647 0.0016 0.4658 <0.0001 0.8964 0.0905 0.0163

H. triloba 0.0347 0.7678 0.0823 0.1799 0.8276 0.3972 0.1434 0.6034 0.8194 0.3378 Neodiplostomum sp. <0.0001 0.7678 0.1537 0.0266 0.0085 0.5429 0.5322 0.6284 <0.0001 0.9605

Cestoda <0.0001 0.0104 0.0175 0.0219 0.0218 0.2118 0.1971 0.333 0.1648 0.2088

Capillaria carbonis 0.0263 0.5679 0.1026 0.1258 0.0065 0.3217 0.2123 0.0355 0.1915 0.5079 Contracaecum 0.5027 0.1123 0.0493 0.0019 0.9431 rudolphi 0.2050 0.6617 0.6039 0.2788 0.1276

Ribeiroia sp. 0.0209 0.4967 0.6373 0.3965 0.4557

Arthropoda 0.0249 0.2320 0.6106 0.1357 0.1895 0.7040 0.9633 Echinochasmus sp. 0.0230 0.1045 <0.0001 0.9569 0.2866 0.7181 0.4566 0.4807 0.7082

A. ostroweskiae <0.0001 0.0007 0.1738 <0.0001 0.8049 0.6802 0.4608 0.7652 Unknown 0.6522 0.6962 0.0002 0.6263 0.6683 Opisthorchiidae Unknown Strigeidae 0.6522 0.6962 0.0002 0.6263 0.6683

Acanthocephala 0.6342 0.232 0.4073 0.2766 0.0007 0.8968 0.5099 0.4774

Unknown Fluke 0.6522 0.6962 0.7865 0.6263 0.6683 Unknown Strigeidae 0.6522 0.0104 0.7865 0.6263 0.6683 Unknown Sclerodistomoididae 0.6522 0.0104 0.7865 0.6263 0.6683

Amphimerus sp. 0.2365 0.3328 0.5018 0.2269 0.2878

183

Table 5.5. Model parameters and output used to compare the performance of

Discriminant Analyses performed for three categorical groupings (sampling site, geographic region, migration behavior) on three iterations of the same parasite data

(infection intensity, prevalence, and ranked intensity) 20 .

Correct Parasites Classification Sig. Cumulative Wilk's p Included Number Percent Axes Explanation 11 98 37 3 81 <0.0001 Intensity Site all 94 35 3 77 <0.0001

Prevalence 15 107 49 3 82 <0.0001 all 104 48 4 90 <0.0001

Rank 10 100 46 4 93 <0.0001 all 100 46 4 93 <0.0001

Region Intensity 6 142 60 1 89 <0.0001 all 133 55 1 88 0.0008

Prevalence 8 154 71 2 100 <0.0001 all 147 67 2 100 <0.0001

Rank 6 160 73 2 100 <0.0001 all 158 72 2 100 <0.0001

Migration Intensity 12 134 56 4 93 <0.0001 all 128 53 1 79 0.0032

Prevalence all 170 78 1 69 <0.0001

Rank all 169 78 1 69 <0.0001

20 Model output for the optimum model (yielding the fewest misclassified hosts) is given above that of the full model , which includes all parasites unless the full model resulted in the fewest misclassifications. The number of parasites included in each model, the number and percent of misclassified hosts are based on the Discriminant Analysis. Canonical estimation of point separation provides the number of significant axes, the cumulative variance explained by the significant axes, and a Wilk’s Lambda p-value for model performance. 184

Table 5.6. Comparison of model performance in predicting the correct classification of P.

auritus by sample site, geographic region, or migration status (migratory or resident) 21 when compared to null model. Significant increases in model performance given as difference (Diff.) measured with Chi-square analysis.

Grouping # Groups Null DA Diff. Sig.

Site (all) 11 9% 43% 4.7 P<0.0001 Site (assemblages) 5 20% 70% 3.5 P<0.0001

Region 3 33% 74% 2.2 P<0.0001

Migration (all) 2 50% 83% 1.7 P<0.0001

Migration (known) 2 50% 84% 1.7 P<0.0001

21 Models built with all hosts (all) and subsets of hosts based on incomplete information (assemblages, known) are indicated with grouping descriptions. The proportion of correct classifications based on random selection (null) is based on the number of groups within a particular grouping category. Performance of the Discriminant Analysis (DA) compared to the null (diff. & sig.) was derived with Chi-Square analysis. 185

Figures

Figure 5.1. Biplot of ranked intensity data illustrating the distribution of Double-crested

Cormorant parasite assemblages from different sites in ordinate space 22 .

22 Mean confidence limit contour lines are drawn around each group mean. Canonical axis 1 explains 53.5% of the separation of the means with an eigenvalue of 2.008, correlation of 0.817, DF=110, p<0.0001. Canonical axis 2 explains 25.6% of the separation of the means with an eigenvalue of 0.959, correlation of 0.700, DF = 90, P<0.0001. Three additional canonical axes were found to be significant (C3 p < 0.0001, C4 p < 0.0001, C5 p = 0.0004) for a cumulative explanatory value of 93%. 186

Figure 5.2. Biplot of ranked intensity data illustrating the distribution of Double-crested

Cormorant parasite assemblages from different regions in ordinate space 23 .

23 50% normal contour lines are drawn around each group mean. Canonical axis 1 explains 86.6% of the separation of the means with an eigenvalue of 0.917, correlation of 0.692, DF=14, p<0.0001. Canonical axis 2 explains 13.4% of the separation of the means with an eigenvalue of 0.142, correlation of 0.353, DF=6, p<0.0001. 187

Figure 5.3. Biplot illustrating the distribution of migratory and resident Double-crested

Cormorant parasite assemblages in estimated ordinate space 24 where red points represent hosts that are resident, blue points migratory hosts, and black points represent hosts that could not be classified because of collection during winter.

24 50% normal contour lines are drawn around each group mean. Colored points represent hosts where migration status was known, black points represent hosts collected in the southeastern region during winter (unassigned migration status). Canonical axis 1 explains 100% of the separation of the means with an eigenvalue of 0.478, correlation of 0.569, DF=8, p<0.0001. Note, because axis 1 explains all variation, points represented in two-dimensional space are estimated based on inflated noise within the model. 188

Figure 5.4. Biplot of parasite assemblages based on output of the Adonis Analysis illustrating the distribution of Double-crested Cormorant parasite assemblages in ordinate space 25 .

25 Normal contour lines are drawn around each group mean. Group 1 represents Lake Guntersville 2 = Leech Lake, 3 = Mossy Lake/Bee Lake, 4 = Cat Island/Swamp Roost/Port of Columbus, 5 = Lake Champlain/Whittington Channel. Canonical axis 1 explains 64.4% of the separation of the means with an eigenvalue of 1.803, correlation of 0.802, DF=32, p<0.0001. Canonical axis 2 explains 24.8% of the separation of the means with an eigenvalue of 0.695, correlation of 0.447, DF=21, p<0.0001. Two additional Canonical axes were found to be significant (axis 3 p<0.0001, axis 4 p = 0.0257), contributing to a 100% cumulative explanation of point separation. 189

Literature Cited

Anderson, M. J. 2001. A new method for non-parametric multivariate analysis of variance. Austral Ecology 26, 32–46.

Arneberg, P., A. Skorping, B. Grenfell, and A. F. Read. 1998. Host densities as determinants of abundance in parasite communities. Proc. Royal Soc. London B 265, 1283-1289.

Basili, G. D. and S. A. Temple. 1999. Dickcissels and crop damage in Venezuela, defining the problem with ecological models. Ecol. Appl. 9,732-739.

Bedard, J, A. Nadeau, and M. Lepage. 1995. Double-crested Cormorant culling in the St. Lawrence River estuary. Colonial Waterbirds 18, 78-85.

Brugger, K. E. 1995. Double-crested Cormorants and fisheries in Florida. Colonial Waterbirds 18, 110-117.

DeVault, T. L., R. B. Chipman, S. C. Barras, J. D. Taylor, C. P. Cranker III, E. M. Cranker, and J. F. Farquhar. 2012. Reducing impacts of Double-crested Cormorants to natural resources in central New York: a review of a collaborative research, management, and monitoring program. Waterbirds 35, 50-55.

Dolbeer, R. A. 1991. Migration patterns of Double-crested Cormorants east of the Rocky Mountains. J. Field Ornithol. 62, 83-93. e-CFR (Electronic Code of Federal Regulations). 2013. Title 50-Wildlife and Fisheries, Chapter 1, Subchapter B, Part 21-Migratory Bird Permits, Subpart D--Control Of Depredating And Otherwise Injurious Birds. Available at: http://www.ecfr.gov/cgi-bin/text- idx?c=ecfr&tpl=/ecfrbrowse/Title50/50cfr21_main_02.tpl.

Erwin, R. M. 1995. The ecology of cormorants: some research needs and recommendations. Colonial Waterbirds 18, 240-246.

Fenech, A. S., S. E. Lochmann, and A. A. Radomski. 2004. Seasonal diets of male and female Double-crested Cormorants from an oxbow lake in Arkansas, USA. Waterbirds 27, 170-176.

Green, M. C. J. L. Waits, M. L. Avery, M. E. Tobin, and P. L. Leberg. 2006. Microsatellite variation of Double-crested Cormorant populations in eastern North America. J. Wildl. Manag. 70, 579-583.

190

Hanson, K. C., T. L. DeVault, and S. J. Dinsmore. 2010. Increased abundance and first breeding record of the Neotropic Cormorant ( Phalacrocorax brasilianus ) on the alluvial plain of Mississippi. SDA National Wildlife Research Center – Staff Publications. Paper 991. http://digitalcommons.unl.edu/icwdm_usdanwrc/991.

Hatch, J. J. 1995. Changing populations of Double-crested Cormorants. Colonial Waterbirds 18, 8-24.

Jackson, J. A and B. J. S. Jackson. 1995. The Double-crested Cormorant in the south- central United States: habitat and population changes of a feathered pariah. Colonial Waterbirds 18, 118-130.

Kanno, Y. and J. C. Vokoun. 2008. Biogeography of stream fishes in Connecticut: defining faunal regions and assemblage types. Northeast. Naturalist 15, 557-576.

Kirsch, E. M. 1995. Double-crested Cormorants along the upper Mississippi River. Colonial Waterbirds 18, 131-136.

Krohn, W. B., R. B. Allen, J. R. Moring, and A. E. Hutchinson. 1995. Double-crested Cormorants in New England: population and management histories. Colonial Waterbirds 18, 99-109.

Lennon, J. J., P. Koleff, J. J. D. Greenwood and K. J. Gaston. 2004. Contribution of rarity and commonness to patterns of species richness. Ecol. Letters. 7, 81-87.

Mazaris, A. D., M. A. Tsianou, A. Sigkounas, P. Dimopoulos, J. D. Pantis, S. P. Sgardelis and A. S. Kallimanis. 2013. Accounting for the capacity of common and rare species to contribute to diversity spatial patterns: is it a sampling issue or a biological effect? Ecol. Indic. 32, 9-13.

McArdle, B. H. and M. J. Anderson. 2001. Fitting multivariate models to community data: A comment on distance-based redundancy analysis. Ecol. 2, 290–297.

Mercer, D. M., S. M. Haig, D. D. Roby. 2013. Phylogeography and population genetic structure of Double-crested Cormorants ( Phalacrocorax auritus ). Conserv. Gen. 14, 823-836.

Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, P. R. Minchin, R. B. O’Hara, G. L. Simpson, P. Solymos, M. H. H. Stevens, and H. Wagner. 2013. Package ‘vegan’. Community Ecol. Package. Version 2.0-9.

191

Peer, B. D., J. Homan, G. M. Linz, and W. J. Bleier. 2003. Impact of Blackbird damage to sunflower: bioenergetics and economic models. Ecol. Appl. 13, 248-256.

Price, P. W. 1989. Host populations as resources defining parasite community organization. In Parasite Communities: Patterns and Processes. Springer Netherlands. pp. 21-40

Sijtsma, M. T. J., J. J. Vaske, and M. H. Jacobs. 2013. Acceptability of lethal control of wildlife that damage agriculture in the Netherlands. Soc. and Nat. Resour. 25, 1308-1323.

Taylor II, J. D. and B. S. Dorr. 2003. Double-crested Cormorant impacts to commercial and natural resources. Proceedings of the 10 th Wildl. Damage Manag. Conference Pages 43-51.

Waits, J. L., M. L. Avery, M. E. Tobin, and P. L. Leberg. 2003. Low mitochondrial DNA variation in Double-crested Cormorants in eastern North America. Waterbirds 26, 196-200.

Wires, L. R., F. J. Cuthbert, D. R. Trexel, and A, R. Joshi. 2001. Status of the Double- crested Cormorant ( Phalacrocorax auritus ) in North America. Final Report to USFWS. http://digitalcommons.unl.edu/usfwspubs/400/.

Wires, L. R., F. J. Cuthbert. 2006. Historic populations of the Double-crested Cormorant (Phalacrocorax auritus ): implications for conservation and management in the 21 st Century. Waterbirds 29, 9-37.

192

5. CLOSING REMARKS

The interactions among wildlife, their environment, and the other organisms around them can be complex and variable in time and space. While management plans to enhance or restrict population growth of wildlife and fisheries are often enacted to reduce the strain on shared resources, they can also assist in the conservation of species. Just as human-wildlife interactions should be considered when developing wildlife management plans, so too should the wildlife- wildlife interactions that can result as a consequence of management applications. Furthermore, these ecological relationships should be documented on a case-by case basis wherever management practices are employed.

In this dissertation, I have documented ways in which different organisms that interact with the Double-crested Cormorants ( Phalacrocorax auritus ) are distributed. This includes two subspecies of P. auritus that overlap in their non-breeding distribution, but apparently nest in very different habitat types in summer. Understanding the environmental variables that contribute to the suitability of various habitats for nesting P. auritus could help managers develop non-lethal management plans that influence the summer habitat for P. auritus rather than the species itself. I demonstrated the use of simple landscape-scale characteristics that defined the breeding habitat of two subspecies of P. auritus and show that these models can be used to predict potential breeding areas in states where nesting data are not readily available. This could be a useful tool for conservation and management alike, as I was able to provide evidence to suggest that the P. auritus nesting in South Carolina are a subspecies that is of conservation concern for state wildlife agencies in the southeastern U.S.

193

I also documented the patterns of distribution of the helminthic parasites of P. auritus in the eastern U.S. My observations documented species that had not been reported in many of the states or regions for which the birds in my study were collected and could be indicative of changes in the distribution of both hosts and parasites. The suites of parasites carried within a host P. auritus appear to follow predictable patterns that suggest similarities in the feeding preferences (or availabilities) within and among regions. Furthermore, we have further evidence of subspeciation of P. auritus through assessments of their parasites, a new type of biological indicator for subspeciation. This could also be a useful tool for managers interested in documenting the feeding habits of waterbirds that are changing in response to alteration of environmental conditions in which they breed, forage, or migrate.

Finally, I attempted to clarify the magnitude and mechanisms behind aquatic community changes where P. auritus feed and breed. I used a combination of field and mesocosm assessments of aquatic communities to test for patterns that would indicate top-down and bottom-up forcing. There were apparent influences of both top-down and bottom up forcing in real and simulated P. auritus colonies; however, the enrichment of nutrients with avian droppings appears to be a more powerful factor influencing the dynamics in aquatic communities. When top-down and bottom-up forcing is applied at low levels simultaneously (as is predicted to occur in many natural systems) their impacts are negligible, but when combined forcing is strong, there is a potential for damage to high-level consumers (i.e., predators) within aquatic systems. This could be particularly informative for managers dealing with colonies of different sizes and colonies that may or may not forage and roost/nest near resources that humans use for sport and recreational purposes.

194

This project was the product of the kindness and cooperative involvement of managers from four states, SC, MN, MS, and FL. It was also an opportunity for me, as a mentor, to teach undergraduates the dynamic nature of field and lab work. The collaborative nature of this research program allowed for the scope and focus of each component of this research project to fluctuate and transform to fit the abilities, and needs of the teams associated with each part.

Consequently, the chapters presented here may not form a single clear picture, but all have a clear focus and demonstrate potential methods that can be used to answer questions that are important to managers and conservationists before management plans are set in motion. This research drives toward an understanding of the factors that affect perspectives and decisions about P. auritus , such as impacts on ecosystems, distribution patterns and subspecies as indicated by habitat and by parasite loads. The tools employed here could also be used to test how effective current management plans are assisting in the management not only of P. auritus , but the resources and communities in which they interact.

195

APPENDICES

196

Appendix A: Supplemental Materials for Chapter II

Background Information on Trophic Webs and their Measurements Defining Trophic Levels: A trophic level is the functional feeding habit of an organism. Common examples of trophic levels include: primary producers (photosynthetic organisms) that derive their resources from the surrounding environment; herbivores that obtain sustenance by eating primary producers, and predators that eat herbivores. Plant defense against herbivory is well- developed in aquatic systems (Coley et al. 1985) and many organisms consume dead, decaying plant matter (Newman 1991) which is less-heavily defended. Thus, detritivory is common in aquatic food webs (Martinson et al. 1991). Omnivory is the consumption of food sources from multiple types of trophic groups, for example, an organism that eats plant matter and herbivores.

Additionally, bacteria-based components of many food webs can be important food sources for microconsumers, which occupy small size classes (1-65 µm).

Problems Associated with Biomass/Energy Modeling : Static measurements of community composition such as biomass, energy transfer, trophic position, and interaction strengths of a food web are commonly used individually (rather than simultaneously) for community assessments (Holz 2000). Collection of these data is time and resource consuming and can be difficult to quantify accurately. Populations fluctuate asynchronously with neighboring trophic levels (Otto and Day 2007). Models developed from population estimates rely heavily on assumptions that reduce their realistic nature. Exhaustive field measurements can be particularly daunting and often require assumptions on variables that cannot easily or reliably be measured with great accuracy, such as population size measurements and productivity rates (Holz 2000).

197

Problems Associated with Organism Size : Size is a variable commonly used in trophic analyses.

Walters and Post (2008) evaluated trophic consequences of disturbance (drought) in an aquatic

system and found that measurements of size were more informative endpoints for evaluation of

disturbance than relative biomass or food chain length. While these estimates may be implied in

the Eltonian pyramid (size-abundance relationships within trophic levels, Sukhdeo 2010) and

Hariston-Smith-Slobodkin (1960; predator control of herbivores allows for vegetation to be

abundant) concepts, size was never explicitly incorporated into the initial theories (Hariston et al.

1960). Elton (1927) suggested the Pyramid of Numbers to represent the abundance of organism

residing at specific feeding levels of a trophic web (with casual observation that most species

consume organisms smaller than themselves). Lindeman (1947) expanded on this and used this

abundance estimate along with biomass to represent the Eltonian pyramid in terms of total

biomass of each trophic level. Size was never explicitly taken into consideration, although one

could easily argue that size and biomass are directly linked (Hechinger et al. 2011). Similarly,

size is not taken into consideration in the trophic cascade model, as each level is comprised of

consumers that eat from the same portion of the food web. For example, deer and rabbits are

considered to have similar trophic positions within a community, even though their sizes differ

substantially (Romanuk et al. 2010).

Problems Associated with Trophic Position : The concept of trophic position is abstract in many

ways. Lindeman (1947) described communities in general terms of diet, which was defined as

consuming inorganic materials, detritus, plant matter, or animal matter. These dietary groupings

help to establish general feeding strategies and a hierarchy of organisms in relation to their

position within a food web, or more simply a food chain . It was the food chain concept that

198

allowed for the simplification and introduction of trophic cascade theories (Carpenter et al.

1985). Unfortunately, many organisms within a community will not fall neatly within a given trophic level. A food chain does not take into consideration multiple food sources for a given species, and thus, does not account for omnivory (where an organism will feed on individuals from multiple trophic levels). Problems associated with omnivory in trophic webs is one of the major criticisms with trophic ecology (Cousins 1987).

199

Additional Results

Here we describe in detail the patterns of organisms grouped by sample type. The five sample types collected include chlorophyll a, which is used as a proxy for phytoplankton abundance, emergent vegetation, organisms collected in plankton tows (zooplankton), and organisms collected in minnow traps (nekton). Tests were performed at four levels, where organismal measurements were compared by treatment (Reference, Nutrient-only, Predation- only, and Nutrient and Predation combined), whether fertilizers were added (Nutrient-only and

Nutrient+Predation) or not (Reference and Predation-only), whether predation was applied

(Predation and Nutrient+Predation) or not (Reference and Nutrient-only), based on the quantity of fertilizer that had been added to the water at the time of sampling, and by sampling session (a representation of time).

Phytoplankton Phytoplankton abundance, was not significantly different among treatments (Reference,

Nutrient-only, Predation-only, and Nutrient and Predation combined; p=0.6676), when fertilized and non-fertilized treatments were compared (p=0.9789), when predation and non-predation treatments were compared (p=0.8748), or when the quantity of nutrients added was considered

(p=0.6136). We did detect a significant decline in phytoplankton biomass over time (p=0.0011).

Emergent Vegetation Emergent vegetation increased in height when nutrients were added (p<0.0001) and showed an increasing then decreasing trend in height over time when all treatments were pooled

(p= 0.0013). This captured the growing season and senescence of vegetation at the end of the growing season where plants begin to die-back. Plant richness was higher in fertilized treatments

200

(p=0.0243), with the highest diversity occurring in Nutrient Treatments. Plant biomass was highest in ponds that received fertilizer (p<0.0001), with the highest biomass occurring in raceways with Nutrient Treatments and Nutrient raceways having significantly higher plant cover than Predation Treatments, which, in turn, had higher plant cover than Reference

Treatments (p<0.05).

Zooplankton Zooplankton size changed in a parabolic fashion over time, where organism length increased then decreased throughout the study period. Although ANOVA and ANCOVA did not capture a significant difference among treatments, a pairwise assessment of nutrient and predation combined and nutrient raceways revealed larger zooplankton were found in the

Nutrient Treatments (P<0.05). Trophic position of the average zooplankter increased then decreased over the course of our assessment with all treatments pooled (p=0.0195). Trophic position of zooplankton (p=0.0493) as well as the biomass (p=0.0021) and abundance

(p=0.0007) of zooplankton in ponds that received Nutrient Treatments increased then decreased over time. We detected significant reductions in zooplankton biomass (p=0.0002) and abundance

(p<0.0001) over the course of our field assessment.

Benthic Organisms Benthic organisms decreased in average length over time (p=0.0019). Although ANOVA assessments did not reveal significant differences among treatments, pairwise assessments for benthic organism length (N+P > C), species richness (C > P), biomass (C > N), and abundance

(C > N) were used to establish characterizations of each treatment using an alpha value of 0.05.

Benthic organism biomass (p = 0.0298) and abundance (p<0.0001) were lower in Nutrient

Treatments compared to reference ponds; however, within fertilized ponds, biomass and 201

abundance of benthic organisms increased as more fertilizer was applied (p<0.0001). Because fertilizer was applied over time, we cannot distinguish this variable from date, which was also a significant predictor of benthic organism biomass (p<0.0001) and abundance (p<0.0001).

Nekton Organisms collected in minnow traps in Nutrient Treatments were more abundant as more fertilizer was applied (p=0.00278) and in all treatments over time (p=0.0014). The average trophic position of nekton increased then decreased with time (p<0.0001) and a trend for higher trophic levels in Reference Treatments vs. N+P Treatments was observed using pairwise assessments of trophic position. Species richness declined over time (p=0.0055). Nekton biomass

(p=0.0085) and abundance (p=0.0109) were lower in fertilized treatments than all other treatments and these variables showed a negative parabolic trend over time (biomass p=0<0.0001, abundance p<0.0001).

202

Tables Table A.1. Averaged variable values for sample units for each treatment (Treat) during each Phase 26 .

Phase Treat Fert Temp Depth Diss. Oxygen ChlA Zooplankton Benthic Nekton Emergent POM SOM mg DO Abu Ric Bio Abu Ric Bio Abu Ric Bio Abu Ric Bio I R 0 23.0 44.0 3.45 41.3 7.61 127 4.0 27.6 515 1.3 120 4.16 0.2 170 2.32 1.2 534 7.0 0.60 I N+P 992 23.3 43.5 4.35 51.7 8.37 178 4.4 43.9 458 1.5 137 3.21 0.1 152 6.57 1.4 202 6.3 0.64 I N 944 23.1 44.0 5.50 51.2 9.23 111 4.3 39.5 397 1.3 102 2.67 0.3 119 6.71 1.6 191 6.4 0.53 I P 0 22.7 41.7 3.70 43.2 8.27 244 3.8 49.8 362 1.1 99 4.11 0.2 178 2.07 1.0 462 7.1 0.68 II R 0 28.8 42.4 7.84 104. 8.56 162 3.7 37.5 700 1.4 174 12.2 0.1 367 1.74 0.7 322 4.3 0.62 II N+P 398 26.0 43.2 3.11 38.7 7.49 489 4.1 15.3 333 1.3 78 5.30 0.0 220 6.45 1.1 177 N/ 0.83 II N 390 26.2 39.1 3.52 43.9 7.63 732 2.0 38.4 347 1.1 67 5.28 0.2 202 6.08 1.1 141 N/ 0.69 II P 0 29.0 38.3 7.46 98.7 9.38 361 3.0 15.6 597 1.2 135 11.1 0.1 387 4.84 0.8 106 4.9 0.55 R R 0 18.9 46.5 9.58 101. 4.98 645 3.6 18.4 356 1.3 743 3.45 0.0 232 2.29 0.7 425 7.3 0.73 R N+P 540 18.3 41.1 9.20 101. 5.72 355 3.0 12.4 167 1.1 352 0.56 0.0 136 3.87 0.6 114 7.9 0.67 R N 540 18.2 42.6 10.0 112. 6.55 124 4.3 31.5 105 1.1 266 2.36 0.0 127 8.43 0.9 195 7.7 0.48 R P 0 19.7 42.2 8.52 93.4 7.64 118 5.5 45.1 245 1.1 565 3.03 0.0 174 4.96 0.8 791 8.1 0.45

26 Phases included a low intensity (I), high intensity (II) and recovery (R). Treatments had a reference ( R) with no manipulation of nutrient concentration or fish abundance, N=addition of nutrients, P=removal of fish, N+P=addition of nutrients and removal of fish, Fert= fertilizer added (grams), Temp= water temperature (°C), Depth=water depth (cm), dissolved oxygen mg=mg/L. ChlA= Chlorophyll A (grams). Abundance (abu), species richness (ric), and biomass (bio). Particulate organic matter of the water column (POM) and organic matter of the sediment (SOM). 203

Table A.2. Trophic Relationship matrix for organisms collected from mesocosms. An entry of 1 in a food web matrix indicates that the predator within the column (numbers for each indicated next to organism name) of interest consumes the prey item of the corresponding row. Trophic level that each organism was grouped by indicated (Tro Lvl).

Tro 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Lvl 1. Alligator Weed 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 2. Annelid 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 3. Bosmina 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 2 4. Calanoid 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 5. Cattail 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 6. Chlorophyll A 0 0 1 1 0 0 1 1 0 1 1 0 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 0 1 7. Chydorid 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 2 8. Cladocera 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 2 9. Crayfish 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 3 10. Cyclops 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 2 11. Daphnia 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 2 12. Detritus 0 1 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 1 0 0 1 0 0 2 13. Dragonfly 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 3 14. Frog 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 2 15. Greenear Sunfish 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 16. Holopedium 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 2 17. Juncus 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 18. Mosquito Larva 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 2 19. Mosquitofish 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 20. Ostracod 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 2 21. POM 0 0 1 1 0 0 1 1 0 1 1 0 0 0 0 1 0 0 1 1 0 0 0 1 0 0 1 0 0 0 1 22. Predatory Midge 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 3 23. Shredding Midge 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 2 24. Sida 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 25. Snail 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 26. Spider 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 3 27. Tadpole 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 2 28. Water Beetle 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 29. W. Beetle Larva 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 30. Willow 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 204

Table A.3. Proportion of trophic group positions held for organisms within food webs (Webs) collected from lakes where P. auritus breed 27 .

Organism Webs Filter Con Pred Mi Pred Basal Graz Mi Con Phyto Zoop Amphipod 5 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Black bass 3 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 Black Crappie 1 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 Bluegill 1 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 Bosmina 7 0.14 0.71 0.00 0.00 0.00 0.00 0.00 0.00 0.14 Brook silverside 1 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Calanoid 5 0.00 0.80 0.00 0.00 0.00 0.00 0.00 0.00 0.20 Chlorophyll A 7 0.00 0.00 0.00 0.00 0.43 0.00 0.00 0.57 0.00 Clam 7 0.29 0.71 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Cyclops 5 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Daphnia 4 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Dragonfly 3 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Gizzard shad 2 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Golden shiner 1 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Hydrobiid 5 0.00 0.60 0.00 0.00 0.00 0.40 0.00 0.00 0.00 Leech 4 0.00 0.00 0.25 0.75 0.00 0.00 0.00 0.00 0.00 Nematode 2 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 Plenorbid 3 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 POM 7 0.00 0.00 0.00 0.00 0.43 0.00 0.57 0.00 0.00 Pred. Midge 4 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Pugnose minnow 2 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Rock bass 3 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 Sauger 3 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 Sh. Midge 6 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Shadow Bass 2 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 Sida 3 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Smallmouth bass 2 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 Tadpole Madtom 3 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 Troutperch 3 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 Walleye 3 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 White crappie 2 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 White sucker 5 0.00 0.40 0.60 0.00 0.00 0.00 0.00 0.00 0.00 Yellow perch 6 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00

27 Filter=filter/suspension feeders, Con=consumers, Pred=predators, Mi Pred=micropredators/parasites, Basal=organisms with no prey, Graz=herbivores/ grazers, Mi Con=microzooplankton/POM, Phyto=phytoplankton, Zoop=zooplankton. Dominant trophic groups are bolded for each organism. 205

Table A.4. Proportion of trophic group positions held for organisms within food webs (Webs) collected in the mesocosm system 28 .

Organism Webs Pred Con EmVeg Phyto Basal Zoop Det Alligator Weed 6 0.00 0.17 0.83 0.00 0.00 0.00 0.00 Annelid 9 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Bosmina 9 0.00 0.89 0.00 0.00 0.00 0.11 0.00 Calanoid 13 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Cattail 16 0.00 0.06 0.94 0.00 0.00 0.00 0.00 Chlorophyll A 16 0.00 0.13 0.00 0.13 0.75 0.00 0.00 Chydorid 6 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Cladocera 3 0.00 0.67 0.00 0.00 0.00 0.33 0.00 Crayfish 4 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Cyclops 15 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Daphnia 11 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Detritus 16 0.00 0.00 0.00 0.00 0.81 0.00 0.19 Dragonfly 6 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Frog 1 1.00 0.00 0.00 0.00 0.00 0.00 0.00 Greenear Sunfish 16 0.69 0.31 0.00 0.00 0.00 0.00 0.00 Holopedium 5 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Juncus 11 0.00 0.09 0.91 0.00 0.00 0.00 0.00 Mosquitofish 7 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Mosquito Larvae 2 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Ostracod 8 0.00 1.00 0.00 0.00 0.00 0.00 0.00 POM 16 0.00 0.13 0.00 0.00 0.88 0.00 0.00 Predatory Midge 5 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Shredding Midge 14 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Sida 6 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Snail 13 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Spider 3 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Tadpole 1 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Water Beetle 2 0.50 0.50 0.00 0.00 0.00 0.00 0.00 Water Beetle Larva 2 0.00 1.00 0.00 0.00 0.00 0.00 0.00 Willow 11 0.00 0.09 0.91 0.00 0.00 0.00 0.00

28 (Pred=predators, Con=consumers, EmVeg=emergent vegetation, Phyto=phytoplankton, Basal=organisms with no prey, Zoop=zooplankton, Det=detritus). Dominant trophic groups are bolded for each organism. 206

Figure A.1. Map of pond system 29 .

29 Individual ponds (labeled 1-4) are separated by concrete walls, which are water tight for nutrient treatments. Vinyl sheeting separates raceways within ponds designed to prevent movement of fishes. Treatment raceways C=reference, N+P=nutrient and predation, N=nutrient-only treatment, P=predation-only treatment. 207

1200 basal ben/zoop 1000 fish 800

600

400 Biomass (g/1000L) Biomass 200

0 Impact Nutrient Predation Reference Mesocosm Treatment

Figure A.2. Relative biomass estimates for the three trophic levels with emergent vegetation as the base of the food web (basal) 30 .

30 Benthic organisms, zooplankton, and nekton collected in minnow traps (fish) as consumers. 208

Reference Literature Used to Develop the Trophic Matrices.

Acharya, K., P. A. Bukaveckas, J. D. Jack, M. Kyle, and J. J. Elser. 2006. Consumer growth linked to diet and RNA-P stoichiometry: response of Bosmina to variation in riverine food resources. Limnology and Oceanography 51:1859-1869.

Anderson, D. H., S. Darring, and A. C. Benke. 1998. Growth of crustacean meiofauna in a forested floodplain swamps: implications for biomass turnover. Journal of the North American Benthological Society 17:21 – 36.

Arslan, N. and Oktener, A. 2012. A general review of parasitic Annelida (Hirudinea) recorded from different habitats and hosts in Turkey. Turkish Journal of Zoology 36:141-145.

Bailey, M. M. 1972. Age, growth, reproduction, and food of the burbot, Lota lota (Linnaeus), in southwestern Lake Superior. Transactions of the American Fisheries Society 101:667-674.

Balaya, D. J. and B. Moss. 2004. Relative importance of grazing on algae by plant- associated and open-water microcrustacea (Cladocera). Arch. Hydrobiol. 161:199-224.

Bellgraph, B. J., C. S. Guy, W. M. Gardner, and S. A. Leathe. 2008. Competition potential between saugers and walleyes in nonnative sympatry. Transactions of the American Fisheries Society 137:790-800.

Blanco, S., S. Romo, and M. J. Villena. 2004. Experimental study on the diet of mosquitofish ( Gambusia holbrooki ) in different ecological conditions in a shallow lake. International Review of Hydrobiology 89:250-262.

Blinn, D. W., R. W. Davies, and B. Dehdashti. 1987. Specialized pelagic feeding by Erpobdella montezuma (Hirudinea). Holarctic Ecology 10:235-240.

Branstrator, D. K. and J. T. Lehman. 1991. Invertebrate predation in Lake Michigan: regulation of Bosmina longirostris by Leptodora kindtii . Limnology and Oceanography 38:482-493.

Brett, M. T., D. C. Muller-Navarra, A. P. Ballantyne, J. L. Ravet, and C. R. Goldman. 2006. Daphnia fatty acid composition reflects that of their diet. Limnology and Oceanography 51:2428-2437.

209

Browne, R. A. and D. Lutz. 2010. Lake ecosystem effects associated with top-predator removal due to selenium toxicity. Hydrobiologia 655:137-148.

Burcher, C. L. and L. A. Smock. 2002. Habitat distribution, dietary composition and the life history characteristics of odonate nymphs in a blackwater coastal plain stream. American Midland Naturalist 148:75-89.

Calow, P. 1975. The feeding strategies of two freshwater gastropods, Ancylus fluviatilis Mull. And Planorbis contortus Linn. (Pulmonata) in terms of ingestion rates and absorption efficiencies. Oecologia 20:33-49.

Carey, M. P., K. O. Maloney, S. R. Chipps, and D. H. Wahl. 2010. Effects of littoral habitat complexity and sunfish composition on fish production. Ecology of Freshwater Fish 19:466-476.

De Carvalho, E. M. and V. S. Uleda. 2009. Diet of invertebrates sampled in leaf bags incubated in a tropical headwater stream. Zoologia 26:694-704.

Denlinger, J. C. S., R. S. Hale, and R. A. Stein. 2006. Seasonal consumptive demand and prey use by stocked saugeyes in Ohio reservoirs. Transactions of the American Fisheries Society 135:12-27.

Dixon, C. J. and J. C. Vokoun. 2010. Population structure and diet of burbot ( Lota lota ) in small streams near the southern extent of the species’ range. Journal of Freshwater Ecology 25:49-58.

Edlund, M. B. and D. R. Francis. 1999. Diet and habitat characteristics of Pagastiella ostansa (Diptera: Chironomidae). Journal of Freshwater Ecology 14:293-300.

Elliott, J. M. 2005. Contrasting diel activity and feeding patterns of four instars of Rhyacophila dorsalis (Trichoptera). Freshwater Biology 50:1022-1033.

Elser, J. J., J. H. Schampel, M. Kyle, J. Watts, E. W. Carson, T. E. Dowling,C. Tang, and P. D. Roopnarine. 2005. Response of grazing snails to phosphorus enrichment of modern stromatolitic microbial communities. Freshwater Biology 50:1826-1835.

Folsom, T. C. and N. C. Collins. 1984. The diet and foraging behavior of the larval dragonfly Anax junius (Aeshnidae), with an assessment of the role of refuges and prey activity. Oikos 42:105-113.

210

George, E. L. and W. F. Hadley. 1979. Food and habitat partitioning between rock bass (Ambloplites rupestris ) and smallmouth bass ( Micropterus dolomieui ) young of the year. Transactions of the American Fisheries Society 108:253-261.

Gonzalez, J. M. and M. A. S. Graca. 2003. Conversion of leaf litter to secondary production by a shredding caddis-fly. Freshwater Biology 48:1578-1592.

Gutowski, M. J. and J. R. Stauffer, Jr. 1993. Selective predation by Noturus insignis (Richardson) (Teleostei: Icraluridae) in the Delaware River. American Midland Naturalist 129:309-318.

Hamm, B. J. and L. Zrum. 1997. Littoral microcrustaceans (Cladocera, Copepoda) in a prairie coastal wetland: seasonal abundance and community structure. Hydrobiologia 357:37-52.

Horth, L. 2004. Predation and the persistence of melanic male mosquitofish ( Gambusia holbrooki ). Journal of Evolutionary Biology 17:672-679.

Johnson, R. L., S. D. Henry, and S. W. Barkley. 2010. Distribution and population characteristics of shadow bass in two Arkansas Ozark streams. North American Journal of Fisheries Management 30:1522-1528.

Kennedy, C. H. 1922. The ecological relationships of the dragonflies of the bass islands of Lake Erie. Ecology 3:325-336.

Kerfoot, W. C. 1978. Combat between predatory copepods and their prey: Cyclops , Epischura , and Bosmina . Limnology and Oceanography 23:1089-1102.

Klecka, J. and D. S. Boukal. 2012. Who eats whom in a pool? A comparative study of prey selectivity by predatory aquatic insects. PloS ONE 7:e37741.

Kraatz, W. C. 1928. Study of the food of the blunt-nosed minnow, Pimephales notatus. The Ohio Journal of Science 28:86-98.

Lachner, E. A. 1950. Food, growth and habits of fingerling northern smallmouth bass, Micropterus dolomieu dolonieu Lacepede, in trout waters of western New York. The Journal of Wildlife Management 14:50-56.

211

MacNeil, C., J. A. T. Dick, and R. W. Elwood. 1997. The trophic ecology of freshwater Gammarus spp. (Crustacea: Amphipoda): problems and perspectives concerning the functional feeding group concept. Biol. Rev. 72:349-364.

Maezono, Y., R. Kobayashi, M. Kusahara, and T. Miyashita. 2005. Direct and indirect effects of exotic bass and bluegill on exotic and native organisms in farm ponds. Ecological Applications 15:638-650.

Mansfield, S. and B. H. McArdle. 1998. Dietary composition of Gambusia affinis (Family Poeciliidae) populations in the northern Waikato region of New Zealand. New Zealand Journal of Marine and Freshwater Research 32:375-383.

Matthews, W. J., F. P. Gelwick, and J. J. Hoover. 1992. Food of and habitat use by juveniles of species of Micopterus and Morone in a southwestern reservoir. Transactions of the American Fisheries Society 121:54-66.

Miyake, M. and T. Miyashita. 2011. Identification of alien predators that should not be removed for controlling invasive crayfish threatening endangered odonates. Aquatic Conservation 21:292-298.

Moens, T. and M. Vincx. 1997. Observations on the feeding ecology of estuarine nematodes. Journal of the Marine Biological Association of the U. K. 77:211-227.

Mullan, J. W., R. L. Applegate, and W. C. Rainwater. 1968. Food of logperch ( Percina caprodes ), and brook silverside ( Labidesthes sicculus ) in a new and old Ozark reservoir. Transactions of the American Fisheries Society 97:300-305.

Noatch, M. R. and G. W. Whitledge. 2011. An evaluation of hydrated lime and predator sunfish as a combined chemical-biological approach for controlling snails in aquaculture ponds. North American Journal of Aquaculture 73:53-59.

Oliver, J. D. 1991. Consumption rates, evacuation rates and diets of Pygmy Killifish, Leptolucania ommata , and Mosquitofish, Gambusia affinis in the Okefenokee Swamp. Faculty Publications and Presentations. Paper 92. http://digitalcommons.liberty.edu/bio_chem_fac_pubs/92.

Paine, M. D., J. J. Dobson, and G. Power. 1982. Habitat and food resource partitioning among four species of darters (Percidae: Etheostoma ) in a southern Ontario stream. Canadian Journal of Zoology 60:1635-1641.

212

Pelham, M. E., C. L. Pierce, and J. G. Larscheid. 2001. Diet dynamics of the juvenile piscivorous fish community in Spirit Lake, Iowa, USA, 1997-1998. Ecology of Freshwater Fish 10:198-211.

Persaud, A. D. and P. J. Dillon. 2011. Differences in zooplankton feeding rates and isotopic signatures from three temperate lakes. Aquatic Science 73:261 – 273.

Persaud, A. D., P. J. Dillon, L. A. Molot, and K. E. Hargan. 2012. Relationships between body size and trophic position of consumers in temperate freshwater lakes. Aquatic Science 74:203-212.

Phillips, E. and R. V. Kilambi. 1996. Food habits of four benthic fish species (Etheostoma spectabile, Percina caprodes, Noturus exilis, Cottus carolinae ) from northwest Arkansas streams. The Southwestern Naturalist 41:69-73.

Rader, R. B. and J. V. Ward. 1989. Influence of impoundments on mayfly diets, life histories, and production. Journal of the North American Benthological Society 8:64-73.

Rick, A. R., J. R. Hodgson, and D. A. Seekell. 2011. Foraging specialization by the opportunistic largemouth bass ( Micropterus salmoides ). Journal of Freshwater Ecology 26:435-439.

Roberts, M. E., B. M. Burr, M. R. Whiles, and V. J. Santucci, Jr. 2006. Reproductive ecology and food habits of the blacknose shiner, Notropis heterolepis , in northern Illinois. American Midland Naturalist 155:70-83.

Roell, M. and D. J. Orth. 1993. Trophic basis of production of stream-dwelling smallmouth bass, rock bass, and flathead catfish in relation to invertebrate bait harvest. Transactions of the American Fisheries Society 122:46-62.

Sammons, S. M. 2012. Diets of juvenile and sub-adult size classes of three Micropterus spp. In the Flint River, Georgia: potential for trophic competition. Southeastern Naturalist 11:387-404.

Schmid, P. E. and J. M. Schmid-Araya. 1997. Predation on meiobenthic assemblages: resource use of a tanypod guild (Chironomidae, Diptera) in a gravel stream. Freshwater Biology 38:67-91.

Schulz, C. A., M. V. Thomas, S. Fitzgerald, and M. Faisal. 2011. Leeches (Annelida: Hirudinida) parasitizing fish of Lake St. Clair, Michigan, U. S. A. Comparative Parasitology 78:73-83. 213

Stasiak, R. 2006. Lake chub ( Couesinus plumbeus ): a technical conservation assessment. Technical Report: USDA Forest Service, Rocky Mountain Region.

Strayer, D. L., N. F. Caraco, J. J. Cole, S. Findlay, and M. L. Pace. 1999. Transformation of freshwater ecosystems by bivalves. Bioscience 49:19-27.

Tahir, H. M and A. Butt. 2009. Predatory potential of three hunting spiders inhabiting the rice ecosystems. Journal of Pest Science 82:217-225.

Tarkowska-Kukuryk, M. and T. Mieczan. 2008. Diet composition of epiphytic chironomids of the Cricotopus sylvesteris group (Diptera: Chironomidae) in a shallow hypertrophic lake. International Journal of Freshwater Entomology 30:285-294.

Tifnouti, A., O. Cherifi, and A. Chifaa. 1994. A study of the diet of five species of Cladocera in the reservoir Lalla Takerkoust (Morocco). Annls. Limnol. 30:285- 296.

Tolonen, A., J. Kjellman, and J. Lappalainen. 1999. Diet overlap between burbot ( Lota lota (L.)) and whitefish ( Coregonus lavaretus (L.)) in a subarctic lake. Ann. Zool. Fennici 36:205-214.

Turner, A. M. 2008. Predator diet and prey behavior: freshwater snails discriminate among closely related prey in a predator’s diet. Animal Behaviour 76:1211-1217.

Tuten, T., M. Allen, and C. Cichra. 2008. Effects of benthic prey composition and abundance on diet and growth of black crappies in three Florida lakes. Transactions of the American Fisheries Society 137:1778-1790.

Van Buskirk, J. 1992. Competition, cannibalism, and size class dominance in a dragonfly. Oikos 65:455-464.

Viosca, Jr., P. 1936. A new rock bass from Louisiana and Mississippi. Copeia 1936:37- 45.

Walker, E. D., E. J. Olds, and R. W. Merritt. 1988. Gut content analysis of mosquito larvae (Diptera: Culcidae) using DAPI stain and epifluorescence microscopy. Journal of Medical Entomology 25:551-554.

Wang, N. and A. Appenzeller. 1998. Abundance, depth distribution, diet composition and growth of perch ( Perca fluviatilis ) and burbot ( Lota lota ) larvae and juveniles in the pelagic zone of Lake Constance. Ecology of Freshwater Fish 7:176-183.

214

Weidel, B. C., D. C. Josephson, and C. C. Krueger. 2000. Diet and prey selection of naturalized smallmouth bass in an oligotrophic Adirondack lake. Journal of Freshwater Ecology 15:411-420.

Yoshioka, T., E. Wada and H. Hayashi. 1994. A stable isotope study in seasonal food web dynamics in a eutrophic lake. Ecology 75:835-846.

215

Literature Cited for Supplemental Material

Carpenter, S. R., J. F. Kitchell, and J. R. Hodgson. 1985. Trophic interactions and lake productivity. BioScience 35:634 – 639.

Coley, P. D., J. P. Bryant, and F. S. Chapin. 1985. Availability and plant antiherbivore defense. Science 230:895 – 899.

Cousins, S. 1987. The decline of the trophic level concept. Trends in Ecology and Evolution 70: 312-316.

Elton, C. 1927. Animal Ecology: Chapter V, The animal community. Sidgewick and Jackson, London.

Hariston, N. G., F. E. Smith, L. B. Slobodkin. 1960. Community structure, population control, and competition. The American Naturalist 94: 421-425.

Hechinger, R. F., K. D. Lafferty, A. D. Dobson, and A. M. Kuris. 2011. A common scaling rule for abundance, energetics, and production of parasitic and free-living species. Science 333: 445-448.

Lindeman, R.L. 1942. The trophic dynamic aspect of ecology. Ecology 23: 399-418.

Martinson, H. M., K. Schneider, J. Gilbert, J. E. Hines, P. A. Hambäck, and W. F. Fagan. 2008. Detritivory: stoichiometry of a neglected trophic level. Ecological Research 23:487–491.

Newman, R. M. 1991. Herbivory and detritivory on freshwater macrophytes by invertebrates : a review. Journal of North American Benthological Society10:89– 114.

Otto S. P. and T. Y. Day. 2007. A Biologists Guide to Mathematical Modeling. Princeton University Press, Princeton, NJ. pp 64-67.

216

Romanuk, T. N., A. Hayward and J. A. Hutchings. 2011. Trophic level scales positively with body size in fishes. Global Ecology and Biogeography 20:231-240.

Sukhdeo, M. V. K. 2010. Food webs for parasitologists: a review. Journal of Parasitology 96: 273-284.

Walters, A. W. and D. M. Post. 2008. An experimental disturbance alters fish size structure but not food chain length in streams. Ecology 89:3261–3267.

217

Appendix B: Supplemental Material for Chapter III Table B.1. Variable description and sources of data used in Maxent models. Variable Type Data Description Data Source Data Type Layer Name Water Presence/Absence NHD Ras WaterY3.5kSUM Water Presence/Absence NHD Ras WaterY3.5kMAX Water Presence/Absence NHD Ras WaterY10kSUM Water Presence/Absence NHD Ras WaterY10kMAX Water Area(sqkm) NHD Ras WaterA3.5kSUM Water Area(sqkm) NHD Ras WaterA3.5.MAX Water Area(sqkm) NHD Ras WaterA10kSUM Water Area(sqkm) NHD Ras WaterA10kMAX Wetland Area(sqkm) USFWS Ras WetlandSUM Wetland Area(sqkm) USFWS Ras WetlandAVG Wetland Area(sqkm) USFWS Ras WetlandMax NLCD Presence/Absence USGS Ras Anthropogenic NLCD Presence/Absence USGS Ras Forest NLCD Presence/Absence USGS Ras Undeveloped NiteLights Intensity USGS Ras NightLights Impervious % imperviousness USGS Ras Impervious Indian Lands Presence/Absence National Atlas Poly NativeAmerican Agricultural % land in agriculture USDA Poly Ag Lands LandsLand Change DCCO type USGS Ras LandUseChange Climate Hundredths of mm Prism Climate Group Ras Precip June Climate Hundredths of mm Prism Climate Group Ras Precip March Climate Hundredths of mm Prism Climate Group Ras Precip September Climate Degrees Celsius Prism Climate Group Ras Min Temp June

218

Table B.1. (cont.) Variable description and sources of data used in Maxent models. Climate Degrees Celsius Prism Climate Group Ras Min Temp March Climate Degrees Celsius Prism Climate Group Ras Min Temp Sept Climate Degrees Celsius Prism Climate Group Ras Max Temp June Climate Degrees Celsius Prism Climate Group Ras Max Temp March Climate Degrees Celsius Prism Climate Group Ras Max Temp Sept Conspecific Presence nest points USDA/FFWCC Ras BirdDensity Sample Sites Nests per colony USDA/FFWCC Ras AllNestSites Nest Presence All sites sampled USDA/FFWCC Ras NestPresentPoints Conspecific bird density USDA/FFWCC Ras Allocate Conspecific bird density USDA/FFWCC Ras Euclidean Conspecific Presence/absence of nests USDA/FFWCC Pnt PriorDecade Mortality Avian botulism deaths USGS NWHC Poly Botulism Mortality Avian cholera deaths USGS NWHC Poly Cholera Mortality Avian lead poisoning deaths USGS NWHC Poly Lead Mortality Avian orthophosphate deaths USGS NWHC Poly Pesticide Fish Advisories Current mercury advisories USEPA Poly, Line, Mercury Fish Advisories Current advisories of other USEPA PtPoly, Line, OtherPollutant Fish Advisories Currentpollutants PCB advisories USEPA Poly,Pt Line, PCBs Fish Advisories Advisories no longer in effect USEPA Poly,Pt Line, Rescind Human Number of people per county U.S. Census Bureau PtPoly CountyPop PopulationHuman Number of people per sqkm U.S. Census Bureau Poly PopDensity PopulationFish Stocking Pounds of fish stocked MNDNR, FFWCC Pnt LbsFishStocked3.5k Fish Stocking Pounds of fish stocked MNDNR, FFWCC Pnt LbsFishStocked10k Fish Stocking Number of Fish Stocked MNDNR, FFWCC Pnt FishStockSqkm3.5k Fish Stocking Number of Fish Stocked MNDNR, FFWCC Pnt FishStockSqkm10k

219

Table B.2. Description of steps taken to derive each variable for the states of Minnesota, Florida, and South Carolina.

Layer Name Data Treatment Steps Populate Poly to Ras Reclass Range Focal Stats AllNestSites Create Polygon Join point data Area Y Y 0/1 Presence

NestPresentPoints N Y ND/1 Presence

Conspecific N Y

Botulism Y N 3.5k MAX Cholera Y N 3.5k MAX

Foraging N N

WaterY3.5kSUM Merged Waterbodies Created Waterbody Col. 1 Y Y 0/1 3.5k SUM WaterY3.5kMAX Merged Waterbodies Created Waterbody Col. 1 Y Y 0/1 3.5k MAX

WaterY10kSUM Merged Waterbodies Created Waterbody Col. 1 Y Y 0/1 10k SUM

WaterY10kMAX Merged Waterbodies Created Waterbody Col. 1 Y Y 0/1 10k MAX

WaterA3.5kSUM Merged Waterbodies Calc. polygon area Y Y 0 thru 10 Geometric Interval 3.5k SUM

WaterA3.5.MAX Merged Waterbodies Calc. polygon area Y Y 0 thru 10 Geometric Interval 3.5k MAX

WaterA10kSUM Merged Waterbodies Calc. polygon area Y Y 0 thru 10 Geometric Interval 10k SUM

WaterA10kMAX Merged Waterbodies Calc. polygon area Y Y 0 thru 10 Geometric Interval 10k MAX

WetlandSUM Merged State Wetlands Created Wetland Col. 1 Y Y 0 thru 10 Geometric Interval 3.5k SUM

WetlandAVG Merged State Wetlands Created Wetland Col. 1 Y Y 0 thru 10 Geometric Interval 3.5k MEAN

WetlandMax Merged State Wetlands Created Wetland Col. 1 Y Y 0 thru 10 Geometric Interval 3.5k MAX

LbsFishStocked3.5k Spatial Join Y Y 1 thru 10 Geometric Interval 3.5k SUM

LbsFishStocked10k Spatial Join Y Y 2 thru 10 Geometric Interval 10k SUM

FishStockSqkm3.5k Spatial Join Calc. waterbody area Area Y Y 3 thru 10 Geometric Interval 3.5k SUM

FishStockSqkm10k Spatial Join Calc. waterbody area Area Y Y 4 thru 10 Geometric Interval 10k SUM

220

Table B.2. (cont.). Description of steps taken to derive each variable for the states of Minnesota, Florida, and South Carolina. Layer Name Data Treatment Steps Convert to Raster Populate Poly to Ras Reclass Focal Stats Nesting Undeveloped N N Y 0/3 Undeveloped land 3.5k SUM Forest N N Y 0/2 Forested 3.5k SUM Precip June N N Precip March N N Precip September N N Min Temp June N N Min Temp March N N Min Temp Sept N N Max Temp June N N Max Temp March N N Max Temp Sept N N Anthropocentric Anthropogenic N N 1 thru -1 DCCO change NightLights N N Y 0/1 Anthropogenic 3.5k SUM Impervious N N Y 1 thru 5 Geometric Interval 3.5k SUM NativeAmerican N N Y 1 thru 5 Geometric Interval 3.5k SUM Ag Lands N Y Y 0/1 Reservation present 3.5k SUM LandUseChange N Y Y 0 thru 10 Geometric Interval 3.5k SUM Lead N Y N 3.5k MAX Pesticide N Y N 3.5k MAX Mercury Y Cell stats Max N N 3.5k SUM OtherPollutant Y Cell stats Max N N 3.5k SUM PCBs Y Cell stats Max N N 3.5k SUM Rescind Y Cell stats Max N N 3.5k SUM CountyPop N Y Y 1 thru 10 Geometric Interval 3.5k MAX PopDensity N Calc Area Pop/area Y Y 2 thru 10 Geometric Interval 3.5k MAX 221

Appendix C: Supplemental Material for Chapter V

Tables Table C.1. Binary table of significant findings from pairwise Adonis analyses. 0 = significant difference observed and 1 = no significant difference between parasite assemblages detected.

Bee Cat Lake Lake Leech Mossy Port of Swamp Lake Wells Whittington Site Lake Island Champlain Guntersville Lake Lake Columbus Roost Waconia Lake Channel

Bee Lake 1 0 1 1 0 1 0 0 1 1 1

Cat Island 0 1 0 0 0 0 1 1 1 0 0

Lake Champlain 1 0 1 0 0 1 1 1 1 0 1

Lake Guntersville 1 0 0 1 0 0 0 0 0 1 0

Leech Lake 0 0 0 0 1 0 0 0 1 1 0

Mossy Lake 1 0 1 0 0 1 0 0 1 1 1

Port of Columbus 0 1 1 0 0 0 1 1 1 0 1

Swamp Roost 0 1 1 0 0 0 1 1 1 0 1

Lake Waconia 1 1 1 0 1 1 1 1 1 1 1

Wells Lake 1 0 0 1 1 1 0 0 1 1 0 Whittington 1 0 1 0 0 1 1 1 1 0 1 Channel

222

Table C.2. Classification table for Discriminant Analysis grouping based on host collection site 31 .

Cat Lake Lake Leech Mossy Port of Swamp Lake Wells Whittington Site Bee Lake Island Champlain Guntersville Lake Lake Columbus Roost Waconia Lake Channel

Bee Lake 0.80 0.05 0.04 0.11 0.00 0.18 0.00 0.00 0.00 0.00 0.00

Cat Island 0.00 0.68 0.08 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.11

Lake Champlain 0.00 0.05 0.44 0.00 0.00 0.00 0.00 0.00 0.23 0.07 0.11

Lake Guntersville 0.00 0.00 0.00 0.76 0.17 0.00 0.00 0.00 0.00 0.20 0.11

Leech Lake 0.20 0.05 0.24 0.08 0.55 0.27 0.20 0.10 0.17 0.33 0.00

Mossy Lake 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

Port of Columbus 0.00 0.18 0.00 0.00 0.00 0.09 0.50 0.30 0.03 0.00 0.11

Swamp Roost 0.00 0.00 0.00 0.00 0.00 0.09 0.20 0.60 0.00 0.00 0.11

Lake Waconia 0.00 0.00 0.12 0.00 0.28 0.09 0.00 0.00 0.40 0.07 0.11

Wells Lake 0.00 0.00 0.08 0.05 0.00 0.09 0.00 0.00 0.17 0.27 0.00 Whittington Channel 0.00 0.00 0.00 0.00 0.00 0.18 0.10 0.00 0.00 0.07 0.33

N 5 22 25 37 29 11 10 10 30 30 9

31 Proportion of correct classifications in bold. Total number of hosts collected from each site listed in last row (N). 223

Table C.3. Classification table for Discriminant Analysis grouping based on host collection region 32 .

Whittington Site Bee Cat Lake Lake Leech Mossy Port of Swamp Lake Wells Lake Island Champlain Guntersville Lake Lake Columbus Roost Waconia Lake Channel

Region SE SE NE SE NC SE SE SE NC NC SE

Northcentral 0.20 0.05 0.48 0.24 0.86 0.27 0.10 0.05 0.93 0.70 0.22

Northeast 0.00 0.18 0.36 0.03 0.03 0.18 0.40 0.45 0.07 0.10 0.11

Southeast 0.80 0.77 0.16 0.73 0.10 0.55 0.50 0.50 0.00 0.20 0.67

N 5 22 25 37 29 11 20 20 30 30 9

32 Actual region is given below site names (NC = northcentral; NE = northeast; SE = southeast). Proportion of correct classifications in bold. Total number of hosts collected from each site listed in last row (N). 224

Table C.4. Classification table for Discriminant Analysis grouping based on host migration status without winter collection sites included in the initial model development 33 .

Whittington Site Bee Cat Lake Lake Leech Mossy Port of Swamp Lake Wells Lake Island Champlain Guntersville Lake Lake Columbus Roost Waconia Lake Channel

Status NA NA M R M NA NA NA M M NA

Migratory 0.20 0.68 0.96 0.22 0.83 0.91 0.70 0.60 0.97 0.80 0.53

Resident 0.80 0.32 0.04 0.78 0.17 0.09 0.30 0.40 0.03 0.20 0.47

N 5 22 25 37 29 11 10 10 30 30 17

33 Actual migration status (status) is given below site names (M = migratory; R = resident, NA = no prior migration status assigned). Proportion of correct classifications in bold. Total number of hosts collected from each site listed in last row (N). 225

Table C.5. Classification table for Discriminant Analysis grouping based on parasite assemblages as identified based on the

Adonis Analysis 34

Whittington Site Bee Cat Lake Lake Leech Mossy Port of Swamp Lake Wells Lake Island Champlain Guntersville Lake Lake Columbus Roost Waconia Lake Channel

Assemblage 3 4 5 1 2 3 4 4 NA NA 5

1 0.20 0.00 0.08 0.86 0.17 0.09 0.00 0.00 0.03 0.27 0.11

2 0.00 0.05 0.40 0.11 0.79 0.36 0.20 0.10 0.60 0.53 0.11

3 0.80 0.05 0.04 0.03 0.00 0.36 0.00 0.10 0.00 0.03 0.00

4 0.00 0.86 0.04 0.00 0.00 0.18 0.80 0.80 0.03 0.03 0.56

5 0.00 0.05 0.44 0.00 0.03 0.00 0.00 0.00 0.33 0.13 0.22

N 5 22 25 37 29 11 10 10 30 30 9

34 Actual assemblage number is given below site names, NA indicated for sites without assignment to a unique parasite assemblage. Proportion of correct classifications in bold. Total number of hosts collected from each site listed in last row (N). 226

Figures

Figure C.1. Biplots of parasite assemblages based on intensity count data including variable direction rays, which are informative for the migration of points in ordinate space, for sample site.

227

Figure C.2. Biplots of parasite assemblages based on prevalence data including variable direction rays, which are informative for the migration of points in ordinate space, for sample site.

228

Mosaic Plot Description We developed mosaic plots to visually represent the contingency tables associated with prediction output of Canonical Discriminant Analyses. Each mosaic plot is comprised of columns, which represent the origin of a host, and rows, which represent the predicted output of the model. The width and height of columns and rows depict the proportion of P. auritus contained within a cell. Cell colors correspond with specific groupings of parasites (legend at the right side of each mosaic plot), and the height of a row within each column should be interpreted as the number of hosts classified therein. If a predictive model is effective, there should be a trend for cells to be largest along a diagonal moving from bottom left to upper right.

229

Figures

Figure C.3. Mosaic plot based on the contingency table compiled from results of ranked site data 35 .

35 N=218, DF=90, -Log Likelihood=201.42, R – square (U) =0.429, Likelihood Ratio Chi- Square= 402.85 p<0.0001, Pearson Chi-Square=553.01 p<0.0001, Fisher’s Exact one-sided = 6.43x10 -27 , Fisher’s two-sided <0.0001. 230

Figure C.4. Mosaic plot based on the contingency table compiled from results of ranked region data 36 .

36 N=218, DF=4, -Log Likelihood=56.10, R – square (U) = 0.315, Likelihood Ratio Chi-Square= 112.20 p<0.0001, Pearson Chi-Square=105.94 p<0.0001, Fisher’s Exact one-sided = 6.43x10 -27 , Fisher’s two-sided <0.0001.

231

Figure C.5. Mosaic plot based on the contingency table created with the results from the ranked intensity data used to predict migration status without data from birds collected in winter 37 .

37 N=151, DF=1, -Log Likelihood=23.19, R – square (U) = 0.260, Likelihood Ratio Chi-Square= 26.39 p<0.0001, Pearson Chi-Square = 50.53 p < 0.0001, Fisher’s Exact one-sided = 2.24x10 - 11 where the probability (Predicted = Y is greater than Actual = Y) <0.0001, Fisher’s two-sided <0.0001. 232

Figure C.6. Mosaic plot based on the contingency table created with the results from parasite assemblages identified by the Adonis Analysis 38 .

38 N=218, DF=1, -Log Likelihood=28.88, R – square (U) = 0.222, Likelihood Ratio Chi-Square = 57.75 p < 0.0001, Pearson Chi-Square = 62.37 p < 0.0001, Fisher’s Exact one-sided = 5.23x10 -14 where the probability (Predicted = Y is greater than Actual = Y) <0.0001, Fisher’s two-sided <0.0001. 233

Appendix D: Historical Excerpt from Literature

Countryman's Cooking, by W.M.W Fowler circa 1965. Cormorant Recipe “Having shot your cormorant, hold it well away from you as you carry it home; these birds are exceedingly verminous and the lice are said to be not entirely host- specific. Hang up by the feet with a piece of wire, soak in petrol and set on fire. This treatment both removes most of the feathers and kills the lice.

When the smoke has cleared away, take the cormorant down and cut off the beak.

Send this to the local Conservancy Board who, if you are in the right area, will give you

3/6d or sometimes 5/- for it. Bury the carcase, preferably in a light sandy soil, and leave it there for a fortnight. This is said to improve the flavour by removing, in part at least, the taste of rotting fish.

Dig up and skin and draw the bird. Place in a strong salt and water solution and soak for 48 hours. Remove, dry, stuff with whole, unpeeled onions: the onion skins are supposed to bleach the meat to a small extent, so that it is very dark brown instead of being entirely black.

Simmer gently in seawater, to which two tablespoons of chloride of lime have been added, for six hours. This has a further tenderising effect. Take out of the water and allow to dry, meanwhile mixing up a stiff paste of methylated spirit and curry powder.

Spread this mixture liberally over the breast of the bird.

Finally roast in a very hot oven for three hours. The result is unbelievable. Throw it away. Not even a starving vulture would eat it.”

234