DEMOGRAPHICS AND MOVEMENT S OF MUTE SWANS IN MICHIGAN , USA By Randall Thomas Knapik A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Fisheries and Wildlife Doctor of Philosophy 2019 ABSTRACT DEMOGRAPHICS AND MOVEMENT S OF MUTE SWANS IN MICHIGAN , USA By Randall Thomas Knapik Mute swans ( Cygnus olor ) were introduced by humans to Michigan with translocation of one pair from Iowa into Michigan population peaked in 2013 with an estimated abundance of 17,520 individuals. This coincided with the Michigan Department of Natural Resources drafting a policy which sought fewer than 2,000 wild mute swans present in Michigan by 2030. However, uncertainty in life stage - specific demographic rates and movement s did no t allow for robust analys e s of levels and types of management needed to achieve the long - term goal. A pilot project was launched in 2014 to investigate inter - and intrastate movement s of mute swans within the Great Lakes region. This effort spawned a formal research partnership between Michigan State University, the Michigan Department of Natural Resources, and the Wildlife Services division of the U.S. Department of Agriculture Animal and Plant Health Inspection Service . The goals of this research were to refine mute swan management strategies in Michigan by incorporating region - specific parameters into demographic models and to understand se asonal movement s of mute swan s. We investigated nesting ecology and life stage - specific survival and movement s for mute swans located in the Lower Peninsula of Michigan. We used aerial surveys to locate nests and fledged young within site boundaries. We used boats to neck collar individuals and visit nests. Estimated nest survival ( = 0.701 ), mean egg volume ( 328.2 ± SD 26.6 cm 3 ), and mean incubation initiation date ( 8 Ap ril ) were comparable to estimates from other portions of mute swan range. Mean clutch size ( 7.0 ± SE 0.15 ) was slightly higher than in areas of their native range , but comparable to estimates from the introduced range in North America. Apparent cygnet surv ival ( i.e., hatch to estimated fledge; 0.27 ± 0.01 ), brood survival ( 0.58 ± 0.03 ), overall productivity (1.2 fledglings/pair) , and percentage of gray young in newly hatched broods ( 36.9 % gray plumage ) were slightly lower compared to portions of native range . Observed b reeding productivity related to saturation of characteristic nesting habitat ( ^ = - 0.9792 , p = 0.04 ) . Seven - month survival estimates for fledged young ( = 0.526, 95 % CI = 0.342 0 . 703 ) were slightly lower than areas of their native range and may be related to ratio of gray and leucistic morph individuals in our population ( ^ leucistic = - 0. 908 , 95 % CI = - 2.086 0 . 269) . Estimated annual survival for non - breeding ( = 0.698, 95 % CI = 0.419 0 . 881 ) and breeding swans ( = 0.850, 95 % CI = 0.686 0 . 936 ) w as slightly less but near reported values for native range. Breeding female mute swans remained on or close to nesting t erritories year - round and were furthest from territories during winter ( = 11.3 km). Juvenile - marked female swans tended to move f a rther from natal area s than juvenile - marked males during their first 2 years of life; however, juvenile - marked females were closer to natal territories than juvenile - marked males at the end of the study. We parameterized a density - dependent matrix population model for Michigan using estimated values from this research. This model suggests that survival rates for juvenile, non - breeding , and breeding swans should be reduced by 26 % annually to achieve the long - term goal of fewer than 2,000 mute swans statewide by 2030. This requires a 17 % removal of the annual population to reach the long - term goal (12,760 swans removed 2018 2029). Importantly, removals mu st be spread evenly across all life stages. Targeting removal across all adult population segments is the most efficient control strategy , as 94% of mute swan nests would need to be destroyed annually (15,748 nests destroyed 2018 2029) to achieve the same goal. Copyright by RANDALL THOMAS KNAPIK 2019 v ACKNOWLEDGEMENTS Many individuals and organizations contributed to success of this research effort. I wo uld like to acknowledge those with foresight to establish the Federal Aid in Wildlife Restoration Act in 1937 , which funded many conservation and research projects for the benefit of consumptive and non - consumptive users alike, including this effort. Fundi ng was provided by the Federal Aid in Wildlife Restoration program (W - 155 - R) administered jointly by the U.S. Fish & Wildlife Service and the Michigan Department of Natural Resources. Additional support from Michigan Involvement Committee of Safari Club In ternational , the Joseph G. Schotthoefer Memorial Student Award , the George J. Wallace and Martha C. Wallace Endowed Scholarship, and the Joseph Laurence Maison Fellowship furthered efficiency of field data collection and allowed the research to be shared with academic and non - academic audiences. Field assistance and logistical support from Michigan State University, USDA APHIS Wildlife Services and both Wildlife Division and Parks and Recreation Division of the Michigan Department of Natural Resources were crucial to success of this research. I would like to personally thank (alphabetically) Tony Aderman, Dusty Arsnoe, Barb Avers, Don Avers, Steve Beyer, Tom Bissett, Aaron Bowden, John Darling, Chris Dohrmann, T ony Duffiney, Dr. Michael Eichholz, Dr. Robert Gates, John Hummel, Earl Krom, Officer Ben Lasher, George Lauinger, Corey Lucas, Ben Luukkonen, Dave Marks, Dr. Russ Mason, Terry McFadden, Jake Nave, Nate Newman, Melissa Nichols, Jeffrey Owens, Joe Robison, Barry Sova, Michael Wegan, Dr. David Williams, Anthony Wilson , and Tim Wilson for their logistical and field support. I would like to especially thank and acknowledge John Hummel, Dusty Arsnoe, and Earl Krom for their hard work, flexibility, and dedication to capture and recovery of many mute vi swans during this research. Tom Cooley provided expert assistance in identifying cause - specific mortality for waterlogged, scavenged, and otherwise unpleasant mute swan carcasses. Bryant Dossman and Dr. Tom Miller prov ided valued assistance in data analysis. Keith Norris and Coree Brooks always len t a sympathetic ear when GPS - transmitters stopped connecting, comprehensive exams were impending, or when I needed to relive memories afield to escape from graduate student li fe. I would like to thank Jill Cruth, Marcia Baar, Jenna Bursley, Sharon Reasoner , and graduate student s of the Fisheries and Wildlife Department for cheer and help in navigating the many policies and procedures of Michigan State University. I thoroughly e njoyed my time on the water and in the air throughout Michigan during this research. The skillful flying of Derek DeRuiter (Northwoods Aviation Inc.) and Sgt. Jerry King (Michigan State Police) allowed for safe execution of project objectives without undes irable effects that tend to accompany small planes and repeated circling. I would also like to thank landowners who allowed me to launch the research boat on their property while tracking down mute swans that ignored study site boundaries. I would like to thank the many citizens who resighted, photographed, and reported neck collared mute swans. Their efforts bolstered the research by keeping tabs on individual swans , especially when GPS collars malfunctioned. I ha ve enjoyed the ability to swap fishing stor ies with and explain the research to many anglers and water recreationists at DNR - maintained boat ramps across the state. I ha ve fielded more hunting and fishing regulation questions at gas pumps and boat launches than I can remember. It did no t take long a m actually not a Conservation Officer , whenever the shiny black truck and odd - looking mud motor lured in the public. I tip my hat to all the Conservation Officers that stay abreast of regulations and succinctly answer those wide - ranging questions. vii The trust, guidance, and professional development opportunities offered by my co - advisors, Dr. Scott Winterstein and Dr. Dav id Luukkonen, made my Ph.D. experience fulfilling, worthwhile, and enjoyable. They created a supportive environment and never hesitated to share a personal anecdote to help me navigate a certain research task or personal conundrum. Dave made sure that I in tegrated within and was exposed to many aspects of Michigan DNR and its waterfowl research and management staff from my very first week in Michigan . Most importantly , Scott and Dave ensured that I was still able to escape from my Ph.D. duties on that perfe ctly - crisp October morning, listen to toms gobble on opening day, and float a slip bobber above late summer bluegills they allowed me to continue pursuing the passions that define me and have defined my career path. I a m also grateful for the 2 additiona l members of my research committee, Dr. Charles Nelson and Dr. Gary Roloff, who provided encouragement to an anguished graduate student during bouts of transmitter failure, helped craft spatial analyses, and provided valuable comments and edits on this dis sertation. I would not have ventured down the path of wildlife research and management if it were no t for the conscious decisions by my father, uncle, and grandfather to take me afield in pursuit of fins and fowl. S teadfast encouragement by my parents, sis ters, aunts and uncles, and grandparents ensured that I was successful in maintaining focus while on the journey to accomplishing this degree. Special praise is needed for my wife, Lauren, who gave unwavering support, love, and trust throughout all aspect s of my professional journey thus far. She perfected the ability to not only appear interested when I ramble continuously during our outdoor excursions about which species of bird is calling overhead or what type of plant I just stepped on, but she somehow viii recalls that information three weeks later when she sees the plant again or hears a bird singing in the distance for that I a m truly grateful. ix TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ...................... xii LIST OF FIGURES ................................ ................................ ................................ ................... xiv CHAPTER 1: INTRODUCT ION ................................ ................................ ................................ 1 Management in North America ................................ ................................ .............................. 2 Demogr aphic and life history comparison ................................ ................................ ............. 4 STUDY DESIGN ................................ ................................ ................................ ........................ 8 Study Area ................................ ................................ ................................ ................................ 8 Focal site selection ................................ ................................ ................................ .................. 8 Focal site descriptions ................................ ................................ ................................ ........... 11 DISSERTATION CONTENT ................................ ................................ ................................ . 14 LITERATURE CITED ................................ ................................ ................................ ............ 16 CHAPTER 2: NESTING E COLOGY OF MUTE SWAN S IN MICHIGAN, USA ............. 21 INTRODUCTION ................................ ................................ ................................ .................... 21 STUDY AREA ................................ ................................ ................................ .......................... 23 Study Site Selection ................................ ................................ ................................ ................ 23 Study Site Descriptions ................................ ................................ ................................ .......... 24 METHODS ................................ ................................ ................................ ................................ 25 Field Techniques ................................ ................................ ................................ .................... 25 GPS - marking ................................ ................................ ................................ ......................... 26 Pre - hatch nest monitoring ................................ ................................ ................................ ..... 27 Post - hat ch investigation ................................ ................................ ................................ ........ 28 Data analysis methods ................................ ................................ ................................ ........... 29 Nesting parameters ................................ ................................ ................................ ................ 29 Modeled nest survival ................................ ................................ ................................ ............ 30 Egg survival ................................ ................................ ................................ ........................... 31 RESULTS ................................ ................................ ................................ ................................ .. 31 GPS - marking ................................ ................................ ................................ .......................... 33 Egg survival ................................ ................................ ................................ ............................ 34 Nest survival ................................ ................................ ................................ ........................... 35 Post - hatch parameters ................................ ................................ ................................ ........... 36 DISCUSSION ................................ ................................ ................................ ............................ 38 MANAGEMENT IMPLICATI ONS ................................ ................................ ....................... 41 LITERATURE CITED ................................ ................................ ................................ ............ 43 CHAPTER 3: DENSITY D EPENDENCE IN PRODUCT IVIT Y OF A NORTH AMERICAN MUTE SWAN POPULATION ................................ ................................ .......... 48 INTRODUCTION ................................ ................................ ................................ .................... 48 STUDY AREA ................................ ................................ ................................ .......................... 51 Site Selection ................................ ................................ ................................ ........................... 51 x Site Descriptions ................................ ................................ ................................ ..................... 52 METHODS ................................ ................................ ................................ ................................ 53 Field methods ................................ ................................ ................................ .......................... 53 Nest density ................................ ................................ ................................ ............................ 53 Breeding productivity ................................ ................................ ................................ ............ 54 Characteristic nesting cover ................................ ................................ ................................ .. 54 Data analysis ................................ ................................ ................................ ........................... 55 Nest spacing ................................ ................................ ................................ ........................... 55 Digitization of characteristic nesting cover ................................ ................................ .......... 55 Saturation of nesting cover ................................ ................................ ................................ .... 56 Comparison of observed breeding productivity among sites ................................ ................ 56 RESULTS ................................ ................................ ................................ ................................ .. 57 DISCUSSION ................................ ................................ ................................ ............................ 65 MANAGEMENT IMPLICATI ON S ................................ ................................ ....................... 69 LITERATURE CITED ................................ ................................ ................................ ............ 73 CHAPTER 4: LIFE - STAG E SPECIFIC SURVIVAL AND MOVEMENTS OF MUT E SWANS IN MICHIGAN, U SA ................................ ................................ ................................ .. 78 INTRODUCTION ................................ ................................ ................................ .................... 78 STUDY AREA ................................ ................................ ................................ .......................... 80 METHODS ................................ ................................ ................................ ................................ 80 GPS - marking ................................ ................................ ................................ .......................... 81 Displacement from capture location ................................ ................................ ..................... 83 Life - stage specific survival analyses ................................ ................................ ..................... 84 Juvenile survival modeling ................................ ................................ ................................ .... 85 Survival modeling for breeding and non - breeding swans ................................ ..................... 85 RESULTS ................................ ................................ ................................ ................................ .. 85 Displacement from capture location ................................ ................................ ..................... 86 Life - stage specific survival analyses ................................ ................................ ..................... 90 Juvenile survival modeling ................................ ................................ ................................ .... 90 Non - juvenile survival modeling ................................ ................................ ............................. 92 DISCUSSION ................................ ................................ ................................ ............................ 93 MANAGEMENT IMPLICATI ONS ................................ ................................ ....................... 97 A PPENDICES ................................ ................................ ................................ ........................... 100 APPENDIX A: MOVEMENT S OBSERVED THROUGH P ILOT NECK COLLARING OF MUTE SWANS CONDUCTED BY MICHIGA N DNR AND USDA APHIS WILDLIFE SERVICES 20 14 2018 ................................ ................................ .................... 101 APPENDIX B: FLIGHT C HARACTERISTICS ON MU TE SWANS DETERMINE D THROUGH GPS - GSM TRAN SMITTERS ................................ ................................ ......... 105 APPENDIX C: DETAILED MOVEMENT OBSERVED T HROUGH GPS - GSM TRANSMITTERS ON MUTE SWANS ................................ ................................ ............... 106 LITERATURE CITED ................................ ................................ ................................ ............ 113 CHAPTER 5: A DENSITY - DEPENDENT M ATRIX POPULATION MOD EL TO INFORM MUTE SWAN MAN AGEMENT IN MICHIGAN, USA ................................ .... 117 INTRODUCTION ................................ ................................ ................................ .................. 117 xi STUDY AREA ................................ ................................ ................................ ........................ 119 METHODS ................................ ................................ ................................ .............................. 120 Comparison of native - and Michigan - parameterization ................................ .................. 121 Formulation of the native - parameterized model ................................ ................................ . 121 Formulation of the Michigan - parameterized model ................................ ............................ 122 Life table response experiment between native - and MI - parameterization ........................ 123 Development of a Density - Dependent Matrix Population Model for Michigan ............ 124 Estimation of removal rates needed to achieve long - term population goal .................... 127 RESULTS ................................ ................................ ................................ ................................ 129 Comparison of native - and Michigan - parameterization ................................ .................. 129 Density - dependent matrix population model for Michigan ................................ ............. 131 Estimation of removal rates needed to achieve long - term population goals .................. 133 Proportional reduction in survival across all swan life stages ................................ ........... 133 Comparison of life - stage specific removals needed to achieve long - term goal .................. 135 Egg and nest destructi on needed to reach long - term goal ................................ .................. 135 DISCUSSION ................................ ................................ ................................ .......................... 136 MANAGEMENT IMPLICATI ONS ................................ ................................ ..................... 138 LITERATURE CITED ................................ ................................ ................................ .......... 140 C HAPTER 6 : M ANAGEMENT IMPLICATIONS ................................ .............................. 144 Review of mute swan management in Michigan ................................ ............................... 144 Overview of pertinent results from demographic and movement study ........................ 145 Management scenarios to reach long - term abundance goal ................................ ............ 146 Practical considerations for future management ................................ .............................. 148 Egg and nest destruction ................................ ................................ ................................ ..... 148 Life stage - specific removal of mute swans 149 LITERATURE CITED ................................ ................................ ................................ ......... 152 xii LIST OF TABLES Table 2.1. Number of detected nesting pairs and number of fledglings per pair 2016 2018 for 6 study sites located in the Lower Peninsula of Michigan, USA. ................................ .................... 32 Table 2.2. Accuracy of flotation methods for estimating hatch date of successful mute swan nests ( n = 82) in Michigan 2016 2018 where hatch and incubation initia tion dates could be determined. ................................ ................................ ................................ ................................ .... 33 Table 2.3. Mean and standard deviation for egg length, width, and volume for eggs ( n = 748) in mute swan nests in Michigan, USA, during 2016 2018. ................................ ........................... 34 Table 2.4. Model selection for logistic regression on egg hatchability of mute swans in Michigan, USA, during 2017 and 2018. ................................ ................................ ....................... 35 Table 2.5. Model selection for mute swan nest survival 2016 Peninsula. ................................ ................................ ................................ ................................ ...... 36 Table 2.6. Apparent hatch to fledge survival estimates by year and color morph for mute swa n cygnets 2016 initial sighting post - hatch (i.e., initial observed brood) and from true initial brood size (i.e., brood size calculated from all hatched eggs). ................................ ................................ ......................... 37 Table 3.1. Reproductive parameters for 6 equal - sized study areas (36 km 2 each) ordered by increasing latitude in the Lower Peninsula of Michigan, USA, 2016 2018. ............................. 58 Table 3.2. Model selection for linear regression on breeding productivity of mute swans in Michigan, USA, during 2016 2018. ................................ ................................ ........................... 60 Table 4.1. Physiographic measurements f or juvenile - and adult - marked mute swans 2016 2018 in the Lower Peninsula of Michigan, USA. ................................ ................................ .................. 86 Table 4.2. Model selection results for a priori candidate model set to explain temporal and morphometric variation in survival for juvenile - marked mute swans 1 September 31 March 2016 and 2017 in the Lower Peninsula of Michigan, USA. ................................ ......................... 92 Table B.1. Summarized flight speeds and altitudes estimated through flexible duty cycles (i.e., FlightMode ) available on neck collar - mounted GPS - GSM transmitte rs ( n = 13,897) on mute swans marked within the Lower Peninsula of Michigan, USA, in 2016 2018. ....................... 105 Table 5.1. Input probabilities for transition matrices of the Michigan - parameterized stage - based deterministic matrix population model and native - parameterized model with input values adapted from Ellis and Elphick (2007). ................................ ................................ ................................ .... 122 xiii Table 5.2. Annual removal needed by life stage to achieve long - term goal of fewer than 2,000 mute swans in Michigan, USA, by the year 2030 using the 100,000 K density - dependent matrix population model with an assumed 26% reduction in survival for all life stages over baseline rates. ................................ ................................ ................................ ................................ ............ 134 Table 5.3. Number of eggs and nests that must be destroyed annually to achieve long - term goal of fewer than 2,000 mute swans in Michigan, USA, by the year 2030 usi ng the 100,000 K density - dependent matrix population model with an assumed 88% reduction egg to fledge survival over baseline rates. ................................ ................................ ................................ ........ 136 xiv LIST OF FIGURES Figure 1.1. Location of Michigan (highlighted in blue) within the 4 administrative flyways of the conterminous United States of America. ................................ ................................ ........................ 3 Fig ure 1.2. Conceptual flow diagram showing integrated nature of research methods and how they were used to refine mute swan management in Michigan, USA. ................................ ........... 7 Figure 1.3. Ten preliminary study sites (white squares outlined in black) located within 5 of the 0 km of shoreline (areas shaded red). ................................ ... 9 Figure 1.4. Six study sites (each 36 km 2 ) located in Antrim, Bay, Cass, Kent, Oakland, and St. for mute swans 2016 2018. ................................ ................................ ................................ ........ 10 Figur e 1.5. Cover type composition of study sites in the Lower Peninsula of Michigan, USA, where intensive nest monitoring occurred for mute swans 2016 2018. ................................ ..... 12 Figure 1.6. Land cover map for Juno study site in Cass County Michigan, USA, showing consolidated NLCD 2011 cover classes. ................................ ................................ ...................... 13 Figure 2.1. ........ 24 Figure 2.2 . Incubation window (shown in red) where all successful nests (gray bars) contained 2018. .................... 33 Figure 2.3. Model predicted hatching probability as a function of egg volume with 95% CI region across range of observed egg volumes (250.4 399.1 cm 3 ) for successful mute swan nests 2016 - ................................ ................................ ................ 35 Figure 2.4. Left - skewed distribution of brood sizes at estimated fledging (1 September) for all monitored mute swan nests with eggs 2016 2018 in the Lower Pe ninsula of Michigan, USA. 37 Figure 3.1. Preliminary and final study sites 2016 2018 were located throughout the Lower Peninsula o f Michigan, USA, within selected public land survey system (PLSS) townships that .............................. 52 Figure 3.2. Example determination of potential nesting locations (white circles) within characteristic nesting cover for mute swans (black line) at the transition of open water and adjacent cover types for focal waterbodies i n the Lower Peninsula of Michigan, USA. ............. 57 Figure 3.3. Comparison of breeding productivity to number of pairs per site (A) and breeding productivity to estimated nesting pair saturation ratio (B) 2016 2018 for 6 equal - sized study sites in the Lower Peninsula of Michigan, USA. ................................ ................................ .......... 59 xv Figure 3.4. Comparison of nest locations 2016 2018 within characteristic and non - characteristic nesting cover for select waterbodies within the Juno site in Cass County, MI, USA (A) and Pontiac site in Oakland County, MI, USA (B) which had nesting pairs establishin g territories outside of characteristic nesting cover and low observed productivity. ....................... 62 Figure 3.5. Comparison of nest locations 20 16 2018 within characteristic and non - characteristic nesting cover for select waterbodies within the Wabasis site in Kent County, MI, USA (A) and Tobico site in Bay County, MI, USA (B) which had most nesting pairs use territories in characteristic nestin g cover and higher observed productivity. ............................... 64 Figure 4.1. Mean weekly displacement from capture location (i.e., nesting territory) for adult - marked female mute swans captured within 6 study sites in the Lower Peninsula of Michigan, USA, 2016 2018 pooled across years with sample size for weekly displacement ave rages indicated on the secondary y - axis. ................................ ................................ ................................ 88 Figure 4.2. Mean weekly displacement since capture (range: 1 109 weeks) for juvenile - marked mute swans captured at 6 study sites in the Lower Peninsula of Michigan, USA, 2016 2018 with sample size for weekly displacement averages indicated on the secondary y - axis. ............. 89 Figure 4.3. Total GPS - derived movements for adult - and juvenile - marked mute swans from 6 capture locations (purple rectangles) April 2016 August 2018 in the Lower Peninsula of Michigan, USA ................................ ................................ ................................ ............................. 90 Figure A.1. Overall movements coded by region of capture for plastic neck collared mute swans during a pilot research effort of the Michiga n Department of Natural Resources and the Wildlife Services Department of the U.S. Department of Agriculture Animal Plant Health Inspection Service in the Lower Peninsula of Michigan, USA, 2014 2018. ................................ ............. 101 Figure A.2. Overall movement for plastic neck collared mute swans captured in southcentral Michigan during a pilot research effort of the Michigan Department o f Natural Resources and the Wildlife Services Department of the U.S. Department of Agriculture Animal Plant Health Inspection Service in the Lower Peninsula of Michigan, USA, 2014 2018. ........................... 102 Figure A.3. Overall movement in southeast Michigan for plastic neck collared mute swans during a pilot research effort of the Michigan Department of Natural Resources and the Wildlife Services De partment of the U.S. Department of Agriculture Animal Plant Health Inspection Service in the Lower Peninsula of Michigan, USA, 2014 2018. ................................ ............. 103 Figure A.4. Overall movement in central Michigan for plastic neck collared mute swans during a pilot research effort of the Michigan Department of Natural Resources and the Wildlife Services Department of the U.S. Department of Agriculture Ani mal Plant Health Inspection Service in the Lower Peninsula of Michigan, USA, 2014 2018. ................................ ................................ .... 104 Figure C.1. Detailed movement of mute swans in southwestern Michigan, USA, as determined by GPS - GSM transmitters. ................................ ................................ ................................ ......... 106 xvi Figure C.2. Detailed move ment of mute swans in southeast Michigan, USA, as determined by GPS - GSM transmitters. ................................ ................................ ................................ .............. 107 Figure C.3. Detailed movement of mute swa ns in east central Michigan, USA, as determined by GPS - GSM transmitters. ................................ ................................ ................................ .............. 108 Figure C.4. Detailed movement of mute swans in west cen tral Michigan, USA, as determined by GPS - GSM transmitters. ................................ ................................ ................................ .............. 109 Figure C.5. Detailed movement of mute swans in the northwest Lower Peninsula of Michigan, USA, as determined by GPS - GSM transmitters. ................................ ................................ ........ 110 Figure C.6. Detailed movement of mute swans in the northern Lower Peninsula of Michigan, USA, as determined by GPS - GSM transmitters. ................................ ................................ ........ 111 Figure C.7. Detailed movement of mute swans in the northern Lower Peninsula and southeastern Upper Peninsula of Michigan, USA, as determined by GPS - GSM transmitters. .. 112 Figure 5.1. Comparison of observed mute swan abundance (blue points) to predicted mute swan abundance in Michigan, USA, for a 150 - year simulation (1949 2098) between the native - (bl ack line) and Michigan - parameterized (red line) deterministic matrix population model. .... 130 Figure 5.2. Comparison of parameter elasticity between deterministic density - independent native - and Michigan - parameterized matrix population models and the de terministic density - dependent Michigan - parameterized model. ................................ ................................ ................ 131 Figure 5.3. Comparison of native - parameterized density - independent matrix population model and a Michigan - parameterized density - dependent matrix population model under 3 simulated levels of carrying capacity for mute swans in Michigan, USA. ................................ ................. 133 Figure 5.4. Comparison of density - dependent modeled reduction in survival needed across all mute swan life stages to achieve the long - term goal of fewer than 2,000 mute swans in Michigan, USA, by the year 2030 . ................................ ................................ ................................ ............... 134 1 CHAPTER 1: INTRODUCT ION Mute swans ( Cygnus olor ) are a large swan species native to northern and central Eurasia (Allin et al. 1987) , but became established in North America during the 20 th century. Concerns regarding over abundance of mute swans exist throughout their native (Wood et al. 2014 ) and introduced ranges (Reese 1975, Petrie and Francis 2003) . Population expansion in their native range resulted from milder winters, protection from harvest, banning of lead fishing weights in portions of their range, and creation of artificial nesting habitat through urban and agricultural expansion (Kirby et al. 1994, Fouque et al. 2007) . Expansion in their introduced range resulted from abundant submerged aquatic vegetation (SAV), protection from harvest, supplemental feeding (Gelston and Wood 1982) , and translocation of mute swans by humans. Early records indicate that mute swans were brought to North Americ a in the late 1800s to adorn city parks and estates (Baldassarre 2014) . The first reports of feral breeding populations in North America o ccurred in the Atlantic flyway along the Hudson river in 1910 (Baldassarre 2014) . Mute swans first established in Michigan in 1919 when a breeding pair was transferred from a private estate in Iowa to Round Lake in Charlevoix County, Michigan, amid concerns of aggression toward children at the Iowa estate (Gelston and Wood 1982) . Populations along the Atlantic coast and throughout the Great Lakes region continued to increase throughout the 20 th century. Population estimates in the Atlantic flyway indicate that the populati on of feral mute swans reached 14,000 in 2002 (Atlantic Flyway Council 2003) . Estimates of mute swan populations in Michigan indicate that the population grew rapidly through 2010 with a long - term annual growth rate of 9.3 % but peaked near 2013 ( n = 17,520) with onset of heightened control efforts (D. R. Luukkonen, Michigan Department of Natural Resources, unpublished data). 2 Management in North America Management of mute swans in North America is implemented to alleviate conflicts with native wildlife, aquatic ecosystems, and humans; however, reductions of mute swans is often met with opposition from segments of the public (Allin and Husband 2004, Blac kburn et al. 2010, Jager et al. 2016) . However, management of mute swans is supported by a broad variety of environmental organizations (U. S. Department of Agricu lture 2012) . Opposition to culling mute swans and subsequent litigation le d to the United States Court of Appeals for the D.C. Circuit granting mute swans federal protection under the Migratory Bird Treaty Act on 28 December 2001. Congress passed the Migratory Bird Treaty Reform Act of 2004 which require s the U. S. Fish and Wildlife Serv ice to publish an official list of bird species to which the Migratory Bird Treaty Act did not apply (U. S. Department of Agriculture 2012) . This list, which included mute swans, was publi shed in the federal register on 15 March 2005 (U. S. Fish and Wildlife Service 2005) . This clarification removed ambiguity in federal protection of nonnative birds and relegated mute swan management to individual states. Migra tory bird populations in North America, specifically those with consumptive use, are managed cooperatively across political borders to ensure equitable access and biological sustainability. This manifest s in a system w ith 4 administrative flyway regions (i .e., Atlantic Flyway, Mississippi Flyway, Central Flyway, and Pacific Flyway; Nichols et al. 1995; Figure 1.1 ). Mute swans can be found in all administrative flyw ays; however, highest abundance of mute swans has historically occurred in the Atlantic and Mississippi Flyways (Mississippi Flyway Council 2012) . 3 Figur e 1.1. Location of Michigan (highlighted in blue) within the 4 administrative flyways of the conterminous United States of America. Coordination of management goals is partially accomplished through administrative flyway councils t hat establish flyway - wid e population objectives . H owever, state - level management of mute swans in the United States varies in scope and urgency. The Atlantic Flyway Council established their first formal mute swan management plan in 2003 that sought to reduce abundance of mute sw ans in the Chesapeake Bay area to fewer than 3,000 individuals by 2011 (Atlantic Flyway Council 2003) . This plan was revised in 2015 after initial control efforts by partner states failed to achieve the flyway - wide reduction goal set i n 2003 (Costanzo et al. 2015) despite localized success in some regions such as Chesap eake Bay (L. Hindman, unpublished data). The Michigan DNR established a policy for managing mute swans in 2006 with a short - term goal of reducing the mute swan population to 3,500 individuals by 2010 and a long - term goal of no more than 2,000 mute swans in Michigan by 2030 (Michigan Department 4 of Natural Resources 2006) . The Mississippi Flyway Council formally established a mute swan manageme nt plan in 2012 with a goal of no more than 4,000 mute swans in the flyway by 2030 . Concomitantly, the Michigan DNR updated their mute swan management policy with revised short - and long - term goals which, respectively, were to: 1) remove all mute swans on DNR - administered lands and reduce statewide population growth to zero, and 2) maintain fewer than 2,000 mute swans statewide by 2030 (Michigan Department of Natural Resources 2012) . The Wildlife Services section of the U. S. Department of Agricul ture Animal and Plant Health Inspection Service conducted an environmental assessment to review management options and potential environmental impacts of their involvement in mute swan damage management activities in Michigan in 2012 (U. S. Department of Agriculture 2012) which culminated in a finding of no significant impact (FONSI) of mute swan damage management in Michigan. The revised policy program and procedures published by the Michi gan DNR in 2012 and FONSI in the environmental assessment by USDA - APHIS Wildlife Services established a foundation for mute swan management in Michigan . H owever, the level of removal needed to achieve the short - and long - term goals was uncertain due to amb iguity in mute swan demographics and particularly the sub - adult life cycle of mute swans in Michigan. Demographic and life history comparison Stochastic population models developed by Ellis and Elphick (2007) demonstrated that reduction in adult survival was likely the most socially acceptable and biologically efficient strategy to reduce mute swan populations in the short te rm; however, the model structure assumed absence of density - dependence in vital rates and was parameterized using research conducted in the native range of mute swans. An attempt to use the Ellis and Elphick (2007) model structure to predict observed abundance estimates of mute swans in Michigan suggested 5 that the mute swan population should have experienced exponential growt h in the 1970s and 1980s; however, exponential growth of the mute swan population was not observed until the 1990s and 2000s (D. R. Luukkonen, unpublished data). Basing management scenarios o n ill - fitting population model s lead to unrealistic projections o f population abundance under management scenarios; therefore, a need to understand variation in demographics across the geographic range of mute swan s exists . Vital rates likely differ for a species between native range where evolutionary forces shaped lif e - history traits and an introduced range where the species is subject to varying resource availability, different clim a tic factors, different interspecies interactions, and c hanging interplay between human and natural systems. Conover and Kania (1999) estimated clutch sizes in mute swans of Chesapeake Bay as slightly higher (6.6 ± 0.1 eggs/clutch) than estimates from native range (5.9 ± 2.2 eggs/clutch ; P errins and Reynolds 1967 ) , although Wood and Gelston (1972) found lower (4.5 eggs/clutch) clutch sizes in a semi - captive flock in northern Michigan when swan abundance was below current levels. Reese (1980) estimated cygnet survival at 82% in Maryland while Brown and Brown (2002) est imated survival of cygnets at 69% in the United Kingdom. Conover et al. (2000) found cygnet survival varied from 53% to 87% on the Atlantic Coast and depended on cygnet color morph. Two color morph s of mute swan cygnets exist in varying proportions throughout introduced and native ranges. Variation exists due to preferential selection of the recessive white color morph in swan propagation programs (Munro et al. 1968, Nelson 1976, Enright 1994) . Historically, white color morph individuals made up only 1% of the population in Britain, but nearly 20% of the sub - populations in eastern Europe are composed of white morphs (Bacon 1980) . Leucistic cygnets have lower survival rates than those exhibiting gray juvenile plumage 6 (Conover et al. 2000) . Conover et al. (2000) also found white plumage males and females more likely nest ed earlier than gray morph individuals of the same age. R atio of color morphs in Michigan was unknown at the initiation of this study; however, leucistic morph individuals were prevalent in populations nearby in Ontario (Lumsden 2016) . V ital rates also exhibit temporal variation within a geographic extent due to biotic and abiotic factors. Interannual variation in survival and reproductive productivity may result from factors such as winter severity (i.e., ice coverage) or seasonal food a vailability. Birkhead et al. (1983) found egg laying date and clutch size related to mean winter temperature (i.e., December - March) prior to the breeding season although the preceding mean winter temperature was not ultimately related to number of cygnets fledged. Czapulak and Wieloch (1991) and Czapulak (2002) found that clutches initiated later in the nest season were smaller and contained smaller eggs (i.e., egg volume) while also finding that mean egg size in clutches influenced cygnet survival to 100 days. Scott and Birkhead (1983) determined that mute swans with high quality territories (i.e., abundant aquatic vegetation) laid earlier clutches and had larger clutch siz es compared to swans in other territories; however, they did not find a relationship between territory quality and number of fledged young per pair or cygnet weight. These findings along with potential for density dependence in reproductive parameters of mute swans (McCleery et al. 2002, Nummi and Saari 2003) suggest that vital rates and life history strategies for mute swans vary between native and introd uced ranges as well as among populations originated through discrete origins (i.e., translocation of a few individuals). Interannual, genetic, and geographic variation must be accounted for when implementing control strategies for invasive species, such as mute swan s . Much is known about demographics, movement, and population trends in the native range of mute swans; however, comparatively 7 little is known about vital rates and movements in North America, especially in the Great Lakes region. This study was implemented in Michigan with the expressed goals to 1) estimate survival rates for breeding, non - breeding , and immature swans, 2) document reproductive parameters and breeding productivity, 3) understand natal and seasonal movements of juvenile and adult s wans, 4) develop population projection models using derived demographic parameters, and 5) provide strategies to achieve short - and long - term mute swan management goals (Figure 1.2) . Figure 1.2. Conceptual flow diagram showing integrated nature of resear ch methods and how they were used to refine mute swan management in Michigan, USA. 8 STUDY DESIGN Study A rea Michigan is biologically and administratively located in the Mississippi Flyway ( Figure 1.1; U. S. Fish and Wildlife Service 1959, Boere and Stroud 2006) and the Upper Mississippi River and Great Lakes Region Joint Venture (NAWMP Committee 1999) . We focused research in the L ower P eninsula of Michigan (centroid 43° 29 ' 19 . 2 " , - 84° 37 ' 34.2 " ) where most mute swan detections recently occurred during breeding wat erfowl surveys (Michigan DNR, L ower P eninsula represents a s outh to n orth gradient of deciduous hardwoods (i.e., oak [ Quercus spp.], beech [ Fagus grandifolia ], and maple [ Acer spp.]) interspersed with agriculture to mixed forest (i.e., pines [ Pinus spp.], spruces [ Picea spp.], firs [ Abies spp.], maples, oaks, and aspen [ Populus spp.]; Pugh et al. 2017 ). There are over 26,000 individual lakes in Michigan in size with just over 18,000 occurring in the L ower P eninsula (Breck 2004) . Focal site selection Annual surveys of breeding and wintering waterfowl by the Michigan Department of Natural Resources (DNR) , surveillance efforts by U.S. Department of Agriculture Wildlife Services, and broad - scale habitat suitability guidelines by Weaver et al. (2012) helped guide selection of focal study sites . We used a geographic information system (GIS; ArcGIS 10.3.1, ESRI, Redlands, CA, USA) to stratify the Lower Peninsula into physiographic regions (Schaetzl et al. 2013) . We overlaid hydrologic and public land survey system (PLSS) township shapefiles (Center for Shared Solutions and Technology Partnerships 2015) to further stratify physiographic regions . We summarized total area (ha) and shoreline length (km) of lakes within each township. We link ed locations where per square mile was estimated during the breeding 9 season of 201 5 via Michigan DNR spring waterfowl surveys to PLSS townships. We used me an amount of shoreline in PLSS townships where estimated spring density was per square mile wa s 2011 2015 ( 40 km) to identify townships where adequate shoreline edge existed to likely have multiple nesting mute swan pairs during the first breeding season of this research (2016; Figure 1. 3 ). Figure 1. 3 . Ten preliminary study sites (white squares outlined in black) located within 5 of the L ower P eninsula with inland sites occurring within townships 40 km of shoreline (ar eas shaded red) . We selected 10 sites (6 km x 6 km) in areas with potential concentrated nesting habitat (Figure 1. 3 ). All 10 sites included publicly - accessible and privately - owned waterbodies. These sites, located in 5 of the 8 physiographic regions of the L ower P eninsula, were surveyed 11 December 2015 via fixed - wing aircraft (Cessna 185; Northwoods Aviation Inc, Cadillac, MI, USA) to record the amount of suitable nesting habitat for nesting pairs the following spring (i.e., 10 emergent aquatic vegetation along shorelines) and number of mute swans within proposed boundaries to aid in study site determination. We detected 1,111 mute swans within site boundaries with mute swan counts among the 10 sites ranging from 9 to 28 6 . We surveyed 5 of 10 preliminary sites and 3 additional areas with fixed - wing aircraft on 12 and 14 April 2016 to count breeding pairs, approximate nest locations, and estimate number of non - breeding mute swans within site boundaries. We chose 5 of these sites for studying nesting ecology 2016 - 18. We added an additional study site in the northern L ower P eninsula in 2017; therefore, 6 sites had nest monitoring in 2017 and 2018 (Figure 1. 4 ). Figure 1. 4 . Six study sites (each 36 km 2 ) located in Antrim, Bay, Cass, Kent, Oaklan d, and St. L ower P eninsula where intensive nest monitoring was conducted for mute swans 2016 2018. 11 Focal site descriptions We chose 6 equal - sized (36 km 2 ) study sites of varying land cover composition for intensive nest moni toring (Figure s 1. 4 and 1. 5 ). Study sites occurred in 6 counties (Antrim, Cass, Bay, Kent, Oakland, and St. Clair) across 5 of 8 physiographic regions of the L ower P eninsula (Schaetzl et al. 2013) . We summarized l a nd cover data within study sites using data from the 2011 National Land Cover Database ( 2011 NLCD; Homer et al. 2015) . The 2011 NLCD classified la yer recognizes 16 land cover classes (Homer et al. 2015) . We found 15 cover classes within 1 study site. We consolidated the 15 cover classes into 7 cover classes (i.e., agriculture, developed, early successional, emergent herbaceous wetlands, forest, open water , and woody wetlands). Open water contained areas of water with 25 % soil or vegetation (Homer et al. 2015) . Developed included areas with residential or commercial development designated as either developed, open space; developed, low intensity; developed, medium intensity; or developed, high intensity (Homer et al. 2015) . Forest included deciduous forest, evergreen forest, or mixed forest. Early successional inclu ded shrub/scrub or grassland/herbaceous. Agriculture included pasture/hay or cultivated crops. Woody wetlands included seasonally wet or flooded areas with 20 % forest or shrub cover whereas emergent herbaceous wetlands were seasonally wet or flooded area s with 80 % coverage in perennial herbaceous vegetation (Homer et al. 2015) . Stu dy sites (Figure 1. 4 ) contained areas of open water ( = 29.0 % , [range: 9.6 - 85.3 % ]) varying between chains of inland lakes and a portion of freshwater delta, St. Clair Flats (Figure 1. 5 ). Human influences were prevalent across the study sites with areas dominated by agriculture ( = 21.3 % , [range: 0 45.9 % ]) and development ( = 14.7 % , [range: 2.0 - 41.7 % ]) . Much of the developed areas were adjacent to open water (Figure 1. 6 ). Percent age of emergent herbaceous 12 wetlands ( = 4.2 % , [range: 0.6 12.0 % ]) and woody wetlands ( = 11.9 % , [range: 0.6 16.1 % ]) also varied among study sites; however, waterbodies on all sites contained areas of developed and natural shoreline. Figure 1. 5 . C over type composition of study sites in the L ower P eninsula of Michigan, USA, where intensive nest monitoring occurred for mute swans 2016 2018. 1 EHW = emergent herbaceous wetlands; 2 ES = early successional 0 0.2 0.4 0.6 0.8 1 Juno St. Clair Pontiac Wabasis Tobico Clam Proportion Open Water EHW¹ Woody Wetlands Forest ES² Agriculture Developed 13 Figure 1. 6 . Land cover map for Juno study site in Cass County Michigan, USA, showing consolidated NLCD 2011 cover classes. Emergent herbaceous wetlands were dominated by narrow - leaved cattail ( Typha angustifolia ), broad - leaved cattail ( T . latifolia ), and their hybrid ( T. glauca ). Monotypi c stands of phragmites ( Phragmites australis ) occurred across all sites; however, they were especially abundant on the St. Clair study site. Many other aquatic plant species (e.g., American lotus [ Nelumbo lutea ], bulrush [ Schoenoplectus spp.], sweet - scente d water lily [ Nymphaea odorata ], yellow pond - lily [ Nuphar spp.]) also occurred within and alongside areas designated emergent herbaceous wetlands. Shallow open water areas contained beds of submerged aquatic vegetation (SAV) with coontail ( Ceratophyllum de mersum ), Eurasian watermilfoil ( Myriophyllum spicatum ), Sago pondweed ( Stuckenia pectinatus ), s lender naiad ( Naja flexilis ), and other pondweeds ( Potamogeton spp.) . Woody wetlands contained buttonbush ( Cephalanthus 14 occidentalis ), dogwoods ( Cornus spp.), willows ( Salix spp.), and alders ( Alnus spp.) in addition to emergent herbaceous plants like cattail, phragmites, and reed canary grass ( Phalaris arundinacea ). Agriculture was primarily row crop cultivation of corn ( Zea mays ) and soybean ( Glycine m ax ) with wheat ( Triticum spp.) occasionally planted as a cover crop or for harvest. DISSERTATION CONTENT This dissertation is organized into this introductory chapter, four primary research chapters, and a concluding chapter. I intend to submit individu al chapters for publication in the scientific literature with coauthors; therefore, I wrote these chapters using plural pronouns even though I take full responsibility for the work presented herein. In Chapter 2 I report on nest survival, cygnet survival, brood survival, cygnet color morph ratios, and mean clutch size for This was accomplished by 1) counting number of eggs per clutch, 2) estimating incubation initiation and hatch dates, 3) tracking individual egg su rvival, and 4) examining number of hatched and fledged cygnets of both color morphs. In Chapter 3 I examined influence of nest density on breeding productivity. This was completed by 1) c ounting nesting pairs and estimating breeding productivity with fixed - wing aircraft, 2) estimating total available nesting cover using aerial imagery and boat surveys, and 3) by comparing current nesting density to a theoretical maximum density derived through spatial optimization procedures. In Chapter 4 I document seasona l movement s and generat e life stage - specific survival estimates for use in demographic modeling. I accomplished this by 1) affixing GPS - GSM transmitters to breeding females, 2) GPS - marking cygnets of known origin, and 3) monitoring swan movements in relati on to their annual cycle (e.g., brood rearing, molting, etc.) and abiotic factors such as winter severity or disturbance. In Chapter 5 I incorporate d derived demographic parameters and density - dependent influences on breeding productivity into a matrix 15 pop ulation model that guides future management of mute swan populations in Michigan. The four primary research chapters, collectively, will advance mute swan management within the Great Lakes region of the United States while also contributing to general understanding of mute swan biology. Finally, in C hapter 6 , I offer data - driven strategies to h elp state and federal agencies achieve management goals for managing this charismatic invasive species. 16 LITERATURE CITED 17 LITERATURE CITED Allin, C., G. Chasko, and T. P. Husband. 1987. Mute Swans in the Atlantic flyway: a review of the history, population growth and management needs. Northeast Section of The Wildlife Society 44:32 - 42. Allin, C., and T. Husband. 2004. An evaluation of 22 years of mute swan management in Rhode Islan d. Atlantic Flyway Council. 2003. Atlantic flyway Mute Swan management plan 2003 - 2013. Atlantic Flyway Council, Laurel, Maryland, USA. Bacon, P. J. 1980. A possible advantage for the Polish morph of the Mute Swan Cygnus olor. Wildfowl 31:51 - 52. Baldassa rre, G. 2014. Ducks, geese, and swans of North America. Volume 1.J ohns H opkins U niversity Press. Baltimore, MD, USA. Birkhead, M., P. J. Bacon, and P. Walter. 1983. Factors affecting the breeding success of the Mute Swan Cygnus olor. Journal of Animal Eco logy 52:727 - 741. Blackburn, T. M., N. Pettorelli, T. Katzner, M. E. Gompper, K. Mock, T. W. J. Garner, R. Altwegg, S. Redpath, and I. J. Gordon. 2010. Dying for conservation: eradicating invasive alien species in the face of opposition. Animal Conservatio n 13:227 - 228. around the world. Eds. G.C. Boere, C.A. Galbraith & D.A. Stroud. The Stationery Office, Edinburgh, UK. pp. 40 - 47 Breck, J. E. 2004. Compilation of databases on Michigan lakes. Michigan Department of Natural Resources, Fisheries Division. Brown, A. W., and L. M. Brown. 2002. Prefledging survival of Mute Swan Cygnus olor cygnets in the Lothians, UK: Survival rates 1981 98 were quantifie d between four growth stages, and related to habitat type and altitude. Bird Study 49:97 - 104. Center for Shared Solutions and Technology Partnerships. 2015. Michigan GIS open data portal. < http://gi s.michigan.opendata.arcgis.com/ >. Conover, M. R., and G. S. Kania. 1999. Reproductive success of exotic Mute Swans in Connecticut. Auk 116:1127 - 1131. 18 Conover, M. R., J. G. Reese, and A. D. Brown. 2000. Costs and benefits of subadult plumage in Mute Swans : Testing hypotheses for the evolution of delayed plumage maturation. American Naturalist 156:193 - 200. Costanzo, G., C. Davies, M. DiBona, J. Fuller, L. J. Hindman, M. Huang, J. Lefebvre, T. Nichols, J. Osenkowski, P. Padding, C. Poussart, E. Reed, and D. Sausville. 2015. Atlantic flyway Mute Swan management plan. Czapulak, A. 2002. Egg size variation in mute swans: Its influence on egg hatchability, cygnet body size and cygnet survival. Waterbirds 25:250 - 257. Czapulak, A., and M. Wieloch. 1991. The breeding ecology of the Mute Swan Cygnus olor in Poland - preliminary report. Wildfowl:161 - 166. Ellis, M. M., and C. S. Elphick. 2007. Using a stochastic model to examine the ecological, economic and ethical consequences of population control in a charism atic invasive species: Mute Swans in North America. Journal of Applied Ecology 44:312 - 322. Enright, L. 1994. Ecological significance of the white and grey colour morphs of the Mute Swan. Ontario Birds 12:19 - 26. Fouque, C., M. Guillemain, M. Benmergui, G. Delacour, J. Y. Mondain - Monval, and V. Schricke. 2007. Mute Swan (Cygnus olor) winter distribution and numerical trends over a 16 - year period (1987/1988 - 2002/2003) in France. Journal of Ornithology 148:477 - 487. Gelston, W., and R. Wood. 1982. The Mute Sw an in northern Michigan. Traverse City, MI , USA . Homer, C., J. Dewitz, L. Yang, S. Jin, P. Danielson, G. Xian, J. Coulston, N. Herold, J. Wickham, and K. Megown. 2015. Completion of the 2011 National Land Cover Database for the conterminous United States representing a decade of land cover change information. Photogrammetric Engineering & Remote Sensing 81:345 - 354. Jager, C., M. P. Nelson, L. Goralnik, and M. L. Gore. 2016. Michigan Mute Swan management: a case study to understand contentious natural reso urce management issues. Human Dimensions of Wildlife 21:189 - 202. Kirby, J., S. Delany, and J. Quinn. 1994. Mute Swans in Great - Britain: a review, current status and long - term trends. Hydrobiologia 279:467 - 482. Lumsden, H. G. 2016. Colour morphs, downy an d juvenile plumages of Trumpeter and Mute Swans. Ontario Birds 34:198 - 204. McCleery, R. H., C. Perrins, D. Wheeler, and S. Groves. 2002. Population structure, survival rates and productivity of Mute Swans breeding in a colony at Abbotsbury, Dorset, Englan d. Waterbirds:192 - 201. 19 Michigan Department of Natural Resources. 2006. DNR Mute Swan management and control program procedures. _____. 2012. Mute Swan management and control program policy and procedures. < www.michigan.gov/documents/dnr/2012_Mute_Swan_Policy_378701_7.pdf >. Accessed 01 March 2016. Mississippi Flyway Council. 2012. Mississippi Flyway Council policy management of Mute Swans. < https://www.michigan.gov/documents/dnr/Mississippi_Flyway_Council_Mute_Swan_p olicy_364885_7.pdf >. Munro, R. E., L. T. Smith, and J. J. Kupa. 1968 . Genetic basis of color differences observed in Mute Swan (Cygnus olor). Auk 85:504 - 505. NAWMP Committee. 1999. North American waterfowl management plan, 1998 update; expanding the vision. North American Waterfowl and Wetlands Office, US Fish and Wildlif e Service, Arlington, Virginia. 32pp. Nelson, C. H. 1976. Color phases of downy Mute Swans. Wilson Bulletin 88:1 - 3. Nichols, J. D., F. A. Johnson, and B. K. Williams. 1995. Managing North American waterfowl in the face of uncertainty. Annual review of ecology and systematics 26:177 - 199. Nummi, P., and L. Saari. 2003. Density - dependent decline of breeding success in an introduced, increasing Mute Swan Cygnus olor population. Journal of Avian Biology 34:105 - 111. Perrins, C. M., and C. M. Reynolds. 1967. A preliminary study of the Mute Swan, Cygnus olor. The Wildfowl Trust. 18 th Annual Report. p. 74 - 84. Petrie, S. A., and C. M. Francis. 2003. Rapid increase in the lower Great Lakes population of feral Mute Swans: a review and a recommendation. Wildlife S ociety Bulletin 31:407 - 416. Pugh, S. A., D. C. Heym, B. J. Butler, D. E. Haugen, C. M. Kurtz, W. H. McWilliams, P. D. Miles, R. S. Morin, M. D. Nelson, and R. I. Riemann. 2017. Michigan forests 2014. Resour. Bull. NRS - 110. Newtown Square, PA: US Departmen t of Agriculture, Forest Service, Northern Research Station. 154 p. 110:1 - 154. Reese, J. G. 1975. Productivity and management of feral Mute Swans in Chesapeake Bay. The Journal of Wildlife Management 39:280 - 286. _____. 1980. Demography of European Mute S wans in Chesapeake Bay. Auk 97:449 - 464. 20 Schaetzl, R. J., H. Enander, M. D. Luehmann, D. P. Lusch, C. Fish, M. Bigsby, M. Steigmeyer, J. Guasco, C. Forgacs, and A. Pollyea. 2013. Mapping the physiography of Michigan with GIS. Physical Geography 34:2 - 39. Sc ott, D. K., and M. E. Birkhead. 1983. Resources and reproductive performance in Mute Swans Cygnus olor. Journal of Zoology 200:539 - 547. U. S. Department of Agriculture, A. a. P. H. I. S., Wildlife Services,. 2012. Final environmental assessment: Mute Swan damage management in Michigan. U. S. Fish and Wildlife Service. 2005. Final list of bird species to which the Migratory Bird Treaty Act does not apply. Federal Register 70. U. S. Fish and Wildlife Service. 1959. The waterfowl counci ls: a conservation partnership. Circular 78. Weaver, J., T. Conway, and M. - environmental variables changes across multiple spatial scales. Landscape ecology 27:1351 - 1362. Wood, K. A., R. A. Stillman protect aquatic plants from herbivore grazing? using behavioural ecology to inform wildlife management. PloS one 9:e104034. Wood, R., and W. L. Gelston. 1972. Preliminary report: the Mute S wans of Michigan's Grand Traverse Bay region. State of Michigan, Department of Natural Resources, Wildlife Division. 21 CHAPTER 2: NESTING E COLOGY OF MUTE SWANS IN MICHIG AN , USA INTRODUCTION Mute swans ( Cygnus olor ) expanded their geographic distribution in North American since their introduction by humans in the late 1800s (Baldassarre 2014) . This expansion occurred through natural dispersal and human - assisted translocations. The first recorded breeding pairs in North America occurred in the early 1900s along the Hudson river in the e astern United States (Baldassarre 2014) . Feral populations were first noted in the Great Lakes region in 1919 following introduction of a breeding pair to an inlet waterbody of Lake Michigan , Ro und Lake, that connects the Boyne River to Lake Michigan in Charlevoix County, Michigan (Wood and Gelston 1972) . Mute swans in northern Michigan originated as a semidomesticated flock under the care of employees of the Chicago Club and local citizens (Gelston and Wood 1982) . C old winters of northern Michigan coupled with lack of a n established migratory pattern necessitated human assistance through supplemen tal feeding and ice clearing to ensure mute swan survival through winter months. Gelston and Wood (1982) documented dis tribution, nesting ecology, movement, and mortality during the early years of mute swan establishment in Michigan; however, no such formal effort occurred since the statewide expansion of mute swans. Nesting ecology for mute swans has been investigated th roughout their native range (Perrins and Reynolds 1967, Birkhead et al. 1983, Czapulak 2 002) and in areas of introduced range (Nummi and Saari 2003) , including the United States (Willey and Halla 1972, Reese 1975, Conover and Kania 1999) . Breeding parameters likely differ between introduced and native ranges due to variation in nest predator communities, food availability, human disturbance, and varying levels of intra - and interspecific competition. Characteristic nesting cover in na tive range consists of emergent vegetation with 46 % of nests placed adjacent to flowing water (Campbell 22 1960) . Small waterbodies (<25 ha) used in aquaculture (i.e., fish ponds) are also colonized by breeding pairs due to emergent nesting cov er along the banks and shallow (<1 m) water (Czapulak 2002) . Breeding in the Atlantic coastal states of the United States were generally limited to estuaries and tidal rivers early during mute swan invasion; however, pairs began nesting on inland waterbodies in the late 1970 s (Conover and Kania 1999) . Wieloch (1991) and Gayet et al. (2011) suggested that mute swans exhibit plasticity in nest site selection which could contribute to range expansion. Variability in nest site composition, intraspecific competition, physical geography, and genetic lineage could also result in differing clutch sizes, egg size, nest survival, and overall productivity. Reese (1975) and Conover and Kania (1999) estimated mean clutch sizes as slightly higher (6.1 [Chesapeake Bay] and 6.6 ± SE 0.1 eggs [Connecticut] , respectively) than reported values within native range (5. 9 eggs , Perrins and Reynolds [1967]); however, Gelston and Wood (1982) documented lower mean clutch size (4.3 eggs ) in a northern Michigan mute swan sub - population. Conover and Kania (1999) documented higher nest survival, egg survival, and overall breeding productivity compared to areas of their native range. Reese (1975) found a Chesapeake Bay sub - population exhibited higher cygnet survival and more young fledged per pair compared to other populations in native (Eltringham 1966, Perrins and Reynolds 1967) and introduced ranges (Willey 1 968, Gelston and Wood 1982) . Additionally, Conover et al. (2000) found varying cygnet survival rates among the two cygnet color morphs (i.e., white or leucistic [Polish] and gray [royal]) controlled by a sex - linked recessive gene (Munro et al. 1968) . Variation in nesting parameters across the ge ographic range of mute swans and potential for difference s related to genetic composition underscores need for regional estimation of nesting ecology parameters. This is especially true if those parameters are being incorporated into 23 modeling for future po pulation management. We conducted this study to estimate region - specific nesting ecology parameters of mute swans while also investigating phenotypic distribution and demographic consequences of leucistic and gray morph individuals. STUDY AREA Study Site Selection We studied (Figure 2.1) . T o ensure that study sites captured geographic and physiographic variability within the nesting range of mute swan s in the Lower Peninsula , we used physiog raphic regions (Schaetzl et al. 2013) and a geographic information system (GIS; ArcGIS 10.3.1, ESRI, Redlands, CA, USA) to delineate potential study sites based on waterbod y availability and topography. We overlaid results from a 2015 survey of breeding waterfowl (Michigan Department of Natural Resources, unpublished data) with hydrography and public land survey system data (PLSS; Center for Shared Solutions and Technology Partnerships 2015 ) to further separate the Lower Peninsula into discrete blocks. PLSS townships with estimated spring density per square mi le during the 2011 2015 waterfowl breeding season had me an shoreline distance of 40 km for inland lakes and rivers; therefore, we used this as a threshold to identify PLSS townships where mute swan presence was likely during the 2016 breeding season (Fig ure 2.1). We subsequently flew 14 6 by 6 km study sites with fixed - wing aircraft (Cessna 185; Northwoods Aviation Inc, Cadillac, MI, USA) in December of 2015 or April 2016 to estimate mute swan abundance and determine suitability as study sites (Figure 2.1 ). We chose 5 study sites in Bay, Cass, Kent, Oakland, and St. Clair Counties for investigation of nesting ecology in 2016 2018 and added 1 site in Antrim County in 2017 2018 (Figure 2.1). 24 Figure 2.1. Preliminary and final study sites were Study Site Descriptions Six study sites were located in 5 of the 8 physiographic regions (Sc haetzl et al. 2013) . Land cover composition varie s latitudinally across the Lower Peninsula of Michigan (Homer et al. 2015) and, therefore, varied among our four inland and two coastal - oriented sites . Inland sites contained a mix of private and publicly - accessible waterbodies with moderate to heavily developed shorelines. Inland wate rbodies contained areas of developed and undeveloped shoreline with permanently - flooded open water ( cover type L1UBH based on the National Wetland Inventory Classification System ; U. S. Fish and Wildlife Servic e 2015). I nland sites also contained areas of freshwater emergent wetland (PEM) adjacent to open water areas (U. S. 25 Fish and Wildlife Service 2015) . The coastal - or iented sites varied in composition. One contained the Tobico Marsh wetland complex in Bay County with a reas of persistent emergent vegetation that was semi - permanently flooded (PEM1F) and open water (PABG; U. S. Fish and Wildlife Service 2015) among areas of agriculture, human development, and forest cover (Homer et al. 2015) . The second coastal study site contained a portion of a freshwat er delta, St. Clair Flats, in St. Clair County. This site was primarily open water (L1UBH and L2UBH) with large areas of freshwater emergent wetlands (PEM) that consisted of native emergent vegetation (e.g., broad - leaved cattail [ Typha latifolia ], bulrushe s [ Schoenoplectus spp.], etc.) and nonnative emergent vegetation (narrow - leaved cattail [ Typha angustifolia ], phragmites [ Phragmites australis ]; U. S. Fish and Wildlife Service 2015). Small (< 5 ha e ach) developed islands are also interspersed within this study area. METHODS Field Techniques We conducted a erial surveys annually during incubation (12 April 1 May) to locate nesting pairs within each study site ( n = 5 sites , 2016 ; n = 6 sites , 2017 2018). Incubating females on nests are easily identifiable from aerial surveys due to conspicuous plumage and large ( 1 m) nest structure against a backdrop of senesced emergent vegetation and open water (Conover and Kania 1999) . Two observers worked together to detect nests on either side of the aircraft. All detected nests were marked on orthophotographs and later transferred to digital format through a GIS (ArcGIS Pro, ESRI, Redlands, CA, USA). We logged f light paths using a cellphone application that records GPS coordinates at 1 second intervals (Strava, Inc., San Francisco, CA USA). Flights were recorded using wing strut - mounted high - definition video camera system s in 2016 2017 (MotoCam 360, Bothell, WA, USA) and 2018 ( GoPro Hero 4 26 Silver, San Mateo, California, USA ). We assumed that detection probability of mute swan nests was near 1 as r eliable detection of nests and determination of nesting locations was possi ble by using 2 aerial observers, recording a detailed flight path, and use of flight video. GPS - marking Uniquely identifiable alphanumeric neck collars were placed on a subset of nesting females ( 5 per site) to aid in detection of broods throughout the brood - rearing cycle, estimate between - year nesting constancy, and estimate survival and movement (Chapter 4). N esting females were captured during incubation or brood - rearing using a modified shephe (Coleman and Minton 1979) or shoulder - fired netgun (CODA Enterprises Inc. Mesa, AZ, USA). We fit nesting female mute swans with green and white plastic neck collars (56 mm diameter; S pinner Plastics, Inc., Springfield, Illinois, USA) that included a GPS - GSM transmitter (CTT - 1070 BT3 ; Cellular Tracking Technologies, Inc., Rio Grande, New Jersey, USA) and weighed 117 - 121.5 g when deployed (< 1.4 % of body weight) . Select male mute swans paired with GPS - collared females were also captured and fit with uniquely - coded plastic neck collars that did not include a GPS - GSM transmitter. All captured swans were weighed, sexed, and fitted with rivet - lock aluminum leg bands (28.5 mm diameter (9C); National Band and Tag Co., Newport, KY, USA) and we measured tarsus, wing, and skull length using dial Vernier calipers or a stopped wing ruler . Ca pture and handling of live mute swans was led by staff of the U.S. Department of Agriculture Wildlife Service s section of the Animal Plant Health Inspection Service (USDA APHIS WS). Michigan State University (MSU) Institutional Animal Care and Use Committee (IACUC) granted an animal - use exemption for MSU personnel throughout this project since capture and marking efforts were led by staff of USDA APHIS Wildlife Services. 27 This work was also partially supported by salary support for Scott R. Winterstein from the USDA National Institute of Food and Agriculture (Project No. MICL02588). Pre - hatch nest monitoring We v isited a subset of nests ( 12 annually per site; n = 110) detected via aerial surveys ( n = 251) to intensively monitor aspects of mute swan nesting ecology. We used a flat - bottomed boat with longtail mud motor (S.W.O.M.P. 26.5, Backwater, Inc., Freeport, MN, USA) to approach mute swan nests which were typically located adjacent to shallow (< 1 m) water. Nests with completed clutches (i.e., warm eggs and/or an actively incubating female) had incubation stage determined by floating all eggs using methods outlined by Westerskov (1950) and Walter and Rusch (1997) adjusted for mean incubation length of mute swans (36 days; Re ese 1975, Baldassa r re 2014). We estimated incubation initiation date for each nest by averaging estimated egg age across the entire clutch and subtract ing mean egg age (in days) from date of observation. Laying order was estimated for all eggs in each clutch based on variati on in egg staining (dirtiest eggs were assumed laid earliest and cleanest eggs last) , and eggs were numbered uniquely using colored markers (Sharpie, Newell Brands, Inc., Atlanta, GA, USA). All eggs had the number corresponding to laying order marked on th eir shells in ~10 locations due to the large size of mute swan eggs. Nests that were encountered during the laying stage were subsequently revisited to mark newly laid eggs, obtain estimated hatch dates, and determine clutch size. We measured length and wi dth of all eggs ( n = 748) using a dial Vernier caliper (Flexbar Machine Corporation, Islandia, NY USA) to the nearest tenth of mm. We placed coated (Plasti Dip, Plasti Dip International, Blaine, MN, USA) temperature logging iButtons (DS1921G - F5#; Maxim Int egrated San Jose, CA, USA) beneath each clutch of eggs to monitor nest fate by comparing in - nest temperature to ambient temperatures recorded by an auxiliary iButton placed on the edge of 28 or in cover near a representative nest. Temperature logging technolo gy has successful ly determin ed h atching and nest failure dates with ground nesting avifauna (Hartman and Oring 2006, O'Connor and Ritchison 2013) , especially when ambient temperatures during the nesting cycle are 29 ° C (Schneider and McWilllams 2007 ) . Mute swans incubate nests constantly (Conover and Kania 1999) ; therefore, changes in nest temperature relative to ambient temperature indicate s nest failure or h atching of young. We coated iButtons in clear Plasti Dip which safely waterproof s the devices to prevent data loss with minimal influence on temperature readings (Roznik and Alford 2012, MacNeil and Williams 2014) . iButtons were programed to turn on at a predetermined date prior to nesting season and were set to record temperature readings every 60 minutes so that onboard memory storage ( n = 2,048 readings) w ould not be fill before termination of the nesting season. Post - hatch investigation We visited nests near estimated hatching dates to recover iButtons, ascertain nest status, count hatched or depredated eggs, and count cygnets hatched and color morph, if successful. Hatched and depredated eggs were distin guished by status of egg shell membranes (Klett et al. 1986) . We determined fate of i ndividual eggs in 2017 and 2018 based on presence of eggshells with uniquely identifiable markings (i.e., colored numbers). Marked eggs absent in eggshell fragments upon nest hatch were assumed to have failed through predation or removal from nest. We located the nesting pair and broods near the nest site and counted cygnets in each color morph. The time needed for mute swans to fledge varies with environmental factors, but is typically 120 15 4 days (Willey and Halla 1972, Reese 1975) . We added 120 days to actual hatching dates of monitored nests to conservatively estimate fledging date for all hatched cygnets. 29 B oat and aerial surveys were conducted near estimated fledging dates to again count cygnets in each color morph for intensively monitored and non - monitored nests. Aerial surveys were flown in fixed - wing aircraft with flight paths and flight video recorded in the same manner as during the spring nest detection surveys. We counted white (i.e., adult or white - morph cygnets) and gray swans and recorded approximate location of broods on aerial p hotos. We used boats to locate pairs detected through aerial surveys and confirm brood size and color morph ratio (i.e., count white - morph cygnets separately from adults) when necessary. Data analysis methods Nesting parameters N esting ecology parameters and use of terminology varies widely within the scientific literature; therefore, we felt it useful to clarify our terminology and methods for ease in comparing these estimates with those in the published literature. We defined ap parent nest survival as proportion of all nests under observation with eggs that were successful (i.e., hatched 1 egg; Conover and Kania 1999) . Modeled nest survival (see below) was estimated through progra m MARK (White and Burnham 1999) . Further, nest survival has been reported in multiple ways which either includes survival of the nest through the egg laying period (Conover and Kania 1999) , excludes the egg laying period, or does not note the technique used (King et al. 2013) . We reported modeled nest survival both ways. Daily sur vival rate follows Dinsmore et al. (2002) as the probability a nest survives 1 day. Hatching rate was de fined as proportion of eggs that hatched from all eggs laid including those from nests that did no t hatch 1 egg (i.e., failed). Apparent egg survival was defined as the proportion of eggs that hatched from successful nests (Johnson and Shaffer 1990) . Estimates of hatched young per nest are typically based on an initial brood size at first re - sighting of the brood post - hatch; however, this may be biased low if hatched 30 cygnets die or disappear before the first nest check after hatch. Further, estimates of apparent cygnet survival will be biased high if initial youn g observed is used rather than number of eggs that hatched from each nest. We reported number of eggs that hatched (i.e., true initial brood size) and observed initial brood sizes (i.e., number of young at first brood re - sighting) for comparison among stud ies in the literature. Additionally, cygnet survival was calculated using both estimates of initial brood size. Cygnet survival was defined as proportion of cygnets that survived from hatch or first brood re - sighting (see above) to estimated fledge for suc cessful nests . Apparent brood survival was estimated by taking number of nests that fledged young and Brood size at fledging or productivity is also reported i n multiple ways which may or may not account for pairs that failed to hatch or fledge young. We followed Conove r and Kania (1999) by reporting number of young per pair for pairs with young at fledging (i.e., mean brood size at fledging) , and number of young fledged per nesting pair (i.e., overall productivity). Nesting ecology parameters were summarized by year and then used as sampling units to obtain a grand mean and standard error with sites pooled unless otherwise noted. Modeled nest survival We used the nest survival approach (Dinsmore et al. 2002) in Program MARK (ver. 8.1; White and Burnham 1999) to estimate annual nest survival across study sites. Use of iButton temperature loggers allowed us to summarize nest fate using daily intervals during the nesting season without regular nest checks thereby avoiding dist urbance that accompanies those activities (Boellstorff et al. 1988, Sedinger 1990) . However, n ests were typically visited once during incubation to asses s status of the clutch a nd to ensure iButtons remained just below eggs in the nest bowl. We included terms for temporal and spatial variability in nest survival analyses 31 based on a priori models of variation in nest survival between year, site, and an additive model of year and s ite. Egg survival We used a generalized linear mixed - effects model (GLMM) within Program R (R Development Core Team 2018) to understand the relationship between egg hatching and its size. We used logistic regression to estimate the probability of survival for eggs ( n = 603) in successful nests as a function of egg volume. Egg volume was calculated using the formula follo wing (Hoyt 1979) where is a shape consta nt (0.512), L = egg length, and B = egg breadth. Egg volume and site were included in the logistic regression models as fixed effects while nest was included as a random effect. We Information Criterion corrected f or small sample sizes (AIC c ; Anderson and Burnham 2002 ). RESULTS Two - hundred twenty - nine pairs with nests were detected within study site boundaries from 2016 2 018 (Table 2.1) using aerial and boat surveys. We intensively monitored 109 nests (Juno = 25, St. Clair = 24, Pontiac = 28, Wabasis = 17, Tobico = 7, and Clam = 8) for clutch sizes, nest survival, egg survival, initial brood sizes, and cygnet survival. Thi s total includes nests of two GPS - marked females that nested beyond study site boundaries in 2018 but were captured and nested within site boundaries in previous study years. We documented 10 instances where territories in typical nesting cover (i.e., resi dual cattails, reed, or phragmites) were filled with new individuals following dissolution of a mating pair through death of one o r both members. Mean incubation initiation date was 8 April and mean hatch date was 12 May; however, we noted egg s in nests as early as 16 March and as late as 7 June. E arliest recorded hatch occurred on 20 April 2017 and latest observed hatch occurred on 4 June in 2016 and 2017. 32 Ninety - three percent of all successful nests with known hatch dates ( n = 81) had eggs throughout the period of 2 0 30 April (Figure 2.2) . Mean clutch size pooled across years and sites was 7.0 ± 0.15 eggs per clutch (range 1 - 10). One GPS - marked female renested 172 m from the original nest and began incubation on a new clutch ( n = 4 eggs) 25 days after failure of the first clutch ( n = 5 eggs; failed on 13 April). Two failed nests with small clutches of infertile eggs were incubated beyond estimated hatch dates. One nest containing one infertile egg was incubated 22 days past its estimated hatch date. A s econd nest containing 2 eggs was incubated an additional 20 days. Neither female renested following these failed nesting attempts. Hatch date was typically estimated to within 1 day (0.93 ± 4.22 days ) of actual hatch date for nests ( n = 82) using flotation methods (Westerskov 1950, Walter and Rusch 1997) although flotation slightly overestimated nest age early and late in the in cubation period and underestimated age at median incubation (Table 2.2). Table 2.1. Number of detected nesting pairs and number of fledglings per pair 2016 2018 for 6 study sites located in the Lower Peninsula of Michigan, USA. *Total does not include nests that were influenced by investigators or nests of pairs culled during incubation in official removal efforts . 33 Figure 2 . 2 . Incubation window (shown in red) where all successful nests (gray bars) contained eggs for monitored mute swan nests in 2018. Table 2. 2 . Accuracy of flotation methods for estimating hatch date of successful mute swan nests ( n = 82 ) in Michigan 2016 2018 where hatch and incubation initiation dates could be determined . Float Category a Estimated Age a (days) Actual Age (days) (days) 1 3 2.1 _ 0.9 2 9 6.3 _ 2.7 3 15 15.5 - 0.5 4 21 24.4 - 3.4 5 27 26.3 _ 0.7 6 33 30.8 _ 2. 2 a Nest age was estimated using floatation methods and categories outlined by (Westerskov 1950) and Walter and Rusch (1997) that were adjusted for mute swan incubation interval of 36 days. GPS - marking Thirty - two breeding females were GPS - marked 2016 2018. Seven male mates of GPS - marked females were also fit with plastic neck collars. We documented 23 instances of mute swans nesting in consecutive years and 7 instances when females only nested in 1 yea r (76.7 % nesting constancy). Incubation length (35.5 days; range 32 37) was estimated for 7 GPS - 34 marked females in 2017. Median distance between successive nests (i.e., breeding dispersal) was 123.1 m (range: 0.91 20,342. 38 m ; n = 20 ) . Seven a dult femal es that dispersed (i.e., moved 300 m between successive nests; Wlodarczyk et al. 2013) had a me dian dispersal of 2.14 km (mean = 5.48 km) . One instance was noted in which a GPS - marked female nested successfully with an unmarked male on a new territory even though the previous mate (neck collared) was alive and defending their historic territory. Egg survival M ean egg length ( 11.32 ± 0.45 cm ), width (7.52 ± 0.22 cm ), and volume (328.2 ± 26.6 cm 3 ) was similar among years (Table 2.3) . Mean hatching rate 2016 2018 was 0.66 ± 0. 0 4 1 . Apparent egg survival was 0.804 ± 0.015. The most parsimonious model for egg survival included a fixed effect for egg volume ( ^ = 0.610, z = 3.23, p = 0.001 ) and a random effect of nest (Table 2. 4 ). Probability of hatching increased with egg volume (Figure 2. 3 ). Table 2 . 3. Mean and standard deviation for egg length, width, and volume for eggs ( n = 748) in mute swan nests in Michigan, USA, during 201 6 2018 . 2016 201 7 2018 Length (cm) 11.26 ± 0.4 11.35 ± 0. 5 11.33 ± 0. 5 Width (cm) 7.51 ± 0.2 7.52 ± 0.2 7.5 3 ± 0.2 Volume (cm 3 ) 325.9 ± 24.4 329. 2 ± 2 7 . 8 32 9 . 1 ± 26. 9 35 Table 2 . 4 . Model selection for logistic regression on egg hatchability of mute swans in Michigan, USA, during 2017 and 2018. Model k a AIC c a c a w i a (1 | Nest) + Egg Volume 1 370.687 0 0.991 (1 | Nest) 0 380.244 9.558 0.008 (1 | Nest) + Site 1 384.469 13.782 0.001 a k = number of parameters in model; AIC c c = change in AIC c from lowest AIC c model; w i = Akaike weight. Figure 2. 3 . Model predicted hatching probability as a function of egg volume with 95 % CI region across range of o bserved egg volumes (250.4 399.1 cm 3 ) for successful mute swan nests 2016 - Nest survival Mean apparent nest survival ( n = 1) was 0.817 ± 0.044. Nest survival was modeled for 98 nests using a 72 - day nesting interval (28 March 7 June). The null model was the most parsimonious model (Table 2. 5 ) . We averaged estimated daily survival rate across all 4 models according to their Akaike weight ( w i ) . We used the model - averaged daily survival rate (0.9925, 95 % CI [0.9878 0.9971]) to calculate a nest sur vival estimate for the 36 - day incubation period ( = 0.761) and for the total nesting period (47 days) that included an estimated laying period ( 36 = 0.701). We used mean clutch size and an egg laying interval of 36 hours to estimate a mean laying period of 11 days for mute swans in Michigan. Failure of nests to hatch ( n = 16) were caused by predation of eggs ( n = 3), mortality of the incubating fema le ( n = 2), small clutches of infertile eggs ( n = 2), flooding and nest destruction caused by storm surge ( n = 2), abandonment ( n = 2), or unknown ( n = 5). Table 2. 5 . Model selection for mute swan nest survival 2016 Peninsula. Model k a AIC c a c a w i a N ull 1 190.267 . 0.587 Year 2 191.490 1.224 0.319 Site + Year 6 195.002 4.736 0.055 Site 6 195.699 5.433 0.039 a k = number of parameters in model, AIC c = , c = difference between AIC c of best fitting model and current model, w i = . Post - hatch parameters Mean true initial brood size was 5.80 ± 0.34 cygnets per successful nest ( n = 55) 2017 2018. Observed initial brood size was 4.9 ± 0.10 cygnets per successful nest (2016 2018) . Median interval between observed hatch date and first nest visit where youn g could be observed was 4 days (range : 0 75 ) . Gray morph cygnets made up 36.9 % of all young observed during first brood re - sightings; however, percentage of cygnets in each color morph varied by site (0 80 % gray morph). Apparent cygnet survival calcula ted with the true initial brood size was 0.27 ± 0.01. Apparent cygnet survival using initial observed brood sizes was 0.33 ± 0.03. Cygnet survival did no t vary with regard to color morph (Table 2. 6 ). Mean estimated survival for eggs to fledging was 0.19 8 w hen sites and years were pooled . Mean estimated survival probability for eggs to fledging calculated from mean fledged per pair among sites (1.42; Table 2.1) and mean clutch size was 0.203. Mean brood survival (0.58 ± 0.03) was calculated for monitored nests that hatch ed young. Mean brood size at fledging was identical across all years at 3.1 fledged 37 cygnets/pair. Overall pooled breeding productivity was 1.2 cygnets/pair. Brood size distribution for monitored nests at fledging w as left - skewed ( Figure 2. 4 ). Mean estimated fledge date for mute swan nests using a 120 - day brood - rearing period was 8 September. Table 2. 6 . Apparent hatch to fledge survival estimates by year and color morph for mute swan cygnets 2016 Peninsula calculated from observed brood size at initial sighting post - hatch (i.e., initial observed brood ) and from true initial brood size (i.e., brood size calculated from all hatched eggs). 2016 2017 2018 SD p * SD p * SD p * Initial Observed Brood 0.31 0.37 0.38 0.35 0.29 0.3 Leucistic 0.35 0.69 0.84 0.35 0.37 0.52 0.30 0.32 0.99 Gray 0.38 0.35 0.43 0.38 0.29 0.34 True Initial Brood 0.26 0.30 0.28 0.29 0.26 0.3 * p - value for 2 sample t - test between year - specific survival of leucistic and gray morph cygnets Figure 2. 4 . Left - skewed distribution of brood sizes at estimated fledging (1 September) for all monitored mute swan nests with eggs 2016 2018 in the Lower Peninsula of Michigan, USA. 38 DISCUSSION Nesting ecology studies typically estimate many parameters throughout nesting and brood - rearing cycle s by conducting repeated site visits and nest checks. This study used new research methods to investigate nesting ecology across mute swan Peninsula , with an experimental design that repre sented spatial variability in parameter estimates without personnel or equipment usually needed for comparable investigations. We monitored fate of individual eggs throughout the incubation period by uniquely marking eggs with colored markers. We documente d fate for eggs and ascertain ed whether eggs hatched, failed, or were predated/lost during incubation (Reynolds et al. 1965) . iButton temperature loggers placed be low eggs in nests determined hatch or failure dates for most nests; however, determination of nest status change was most easily measured for semi - terrestrial nests that were less prone to water saturation of nesting material which buffered fine - scale chan ges in nest temperature. Additionally, breeding females continually built and repaired nesting mounds during incubation which required adjustment of iButtons to ensure they remained just below eggs in the nest bowl. Interestingly, this observed behavior to continually rebuild nests resulted in accidental burial of viable eggs by females on several occasions. Proper use of iButtons in mute swan nests remains a practical way to estimate hatch and failure timing despite initial setbacks experienced based on ne st positioning and female behavior. Our e stimates of nest survival are comparable to those reported in the published literature for mute swans in introduced range on the eastern coast of North America (Reese 1980, Conover and Kania 1999, Hindman et al. 2014) ; however, a few key differences relating to productivity and cygnet survival exist . Mute swans tend to have higher clutch sizes in introduced ranges (Ciaranca et al. 1997, Conover and Kania 1999) , compared to their native range (Perrins and 39 Reynolds 1967) and this trend held in Michigan. Additionally, mean volume for mute swan eggs observed i n this study were between values reported for their native range (Bacon and Mountford 1990, Czapulak 2002) and we noted that smaller eggs tended to remain unhatched in the nest more often (Figure 2. 3 ) consistent with observation s by Czapulak (2002) . Czapulak (2002) found that larger eggs, due to their proportionately higher increase in yolk and lipid stores (Birkhead 1984) , influenced cygnet survival to 100 days post - hatch, but did no t explain variation in the first month of life when cygnets are more susceptible to influences of weather extremes or potential predators. We w ere unable to investigate this relationship in Michigan because we did not individually mark cygnets according to their res pective egg. Our p ost - hatch parameters deviate d from ranges in the published literature. Mean initial observed brood size after hatch (4.9 cygnets/pair) was slightly higher than values noted for Connecticut (Conover and Kania 1999) and Maryland (Reese 1980) , but was identical to values obtained from Long Point in Ontario, Canada (Knapton 1993) . This likely results from similar nest survival estimates coupled with a slightly elevated clutch size and egg survival rates. True initial brood size, not reported in most investigations, wa s higher (5.8 cygnet/pair). Uniquely marking all eggs in nests with different colored markers allowed us to accurately account for potential disparity in number of egg s know n to hatch and brood size at initial resighting. Importantly, the method we used removed bias in apparent egg survival related to eggs that were removed or lost from the nest prior to hatch (i.e., their uniquely - marked eggshells were not present duri ng post - hatch nest checks). C ygnet color morph ratio was easily determined for broods at first visit after hatch. Gray morph cygnets were characterized by gray downy plumage and gray feet and bills whereas white or leucistic morph cygnets had brownish - whi te plumage and flesh - colored feet and bills (Nelson 40 1976) . M ean ratio of gray to white young (36.9 % ) varied Lower Peninsula (0 80 % gray ) but was stable across years likely due to genetic similarity of breeding pairs among years. P differs fro m many native (Bacon 1980, Wieloch and Czapulak 1991) and introduced populations (Conover et al. 2000) which are dominated by gray morph individuals , with the ex ception of mute swans in nearby Ontario, Canada (Knapton 1993) . Conover et al. (2000) documented effects of gray and leucistic morph cygnets on population demographics. Pre - fledge survival was lower for leucistic morp h cygnets and differences in parental behavior toward cygnets were noted between both color morphs (Conover et al. 2000) . Additionally, Conover et al. (2000) found that leucistic males had lower survival rates for first 2 years of life, but were able to pair and mate earlier than gray morph ma les. We did no t find statistical differences in cygnet survival between color morphs ( Table 2. 6 ); however, gray cygnets had higher apparent survival than leucistic morph cygnets in the first two years, but n ot in the final year ( which also had the lowest overall cygnet survival ) . No juvenile - marked swan nested at 1 or 2 years of age. The first nesting attempt s for GPS - marked cygnets occurred after the conclusion of the study at 3 years of age (R. Knapik, unpublished data). E stimated survival was lower for leucistic individuals from fledging to their first April although differences were not statistically significant (Chapter 4). Lower cygnet survival and brood survival ( = 0.58) in Michigan resulted in a lower egg to fledge survival estimate ( = 0.19 8 ) when compared to Connecticut ( = 0.41; Conover and Kania 1999) , Maryland, ( = 0.48; Reese 1975), and England ( = 0.48; Reyno lds et al. 1965). Subsequently, overall productivity in Michigan (1.2 cygnets/pair) was below the 2.7 cygnets per nest ing pair reported by Conover and Kania (1999) and 2.2 cygnets/pair reported by Reese 41 (1980) ; however, it was identical to estimates already reported for Michigan mute swans (1.2 cygnets/p air; Wood and G elston 1972). Conover and Kania (1999) conclude d that their higher breeding productivity may result from an expanding mute swan population that was not near carrying capacity. Interestingly, despite lower cygnet survival and brood survival, mean brood size at fledging in our research (3.1 cygnets/pair) was near values re ported for Connecticut (3.2 cygnets/pair; Conover and Kania 19 99). These findings suggest that lower overall breeding productivity in Michigan resul ts from increased brood mort ality for some pairs rather than homogenously lower cygnet survival across all broods. Therefore, heterogeneity in territory quality or available brood - rearing habitat between pairs is likely influencing overall productivity through brood and cygnet surviv al rates. MANAGEMENT IMPLICATI ONS Managers are increasingly interested in reducing abundance of introduced mute swan populations . A s abundance increases, effects o n wetlands and native wildlife are realized, and public tolerance for wildlife conflict de creases. Concomitantly, the public desires science - based natural resources management policies that are carried out by trained professionals (Reiter et al. 1999) . Management plans for non - native invasive species, such as mute swan s , require scientific information on their basic biology to inform future strategies. Region - specific information on survival or movement is typicall y unavailable due to novelty of the species in the area or because of the need for early detection and decision making (Mack et al. 2000, Edelaar and Tella 2012) . Therefore, management or eradication plans typically use demographic and movement data from other introduced regions or from native range. This practice may lead to un desirable results if dynamics differ between native and intr oduced range s due to varying climactic conditions, predator assemblages, or interspecific competition. This research focused on quantifying 42 management - relevant nesting ecology parameters for mute swans in introduced range in Michigan. Oiling eggs or destr uction of nests can be effective for reduc ing the number o f mute swans that enter the non - breeding population at local scales , and is sometimes preferable to culling of breeding adults in high density areas since pairs remain on territories despite the now - infertile clutch (Hin dman et al. 2014) . Managers interested in reliably locating nests for oiling or for culling of adult pairs should utilize low - level aircraft surveys of target areas when most females are incubating eggs (20 30 April) due to the conspicuous nature of m ute swan nests. Egg flotation methods (Westerskov 1950, Walter and Rusch 1997) adapted for the 36 day mute swan incubation period can be used to determine estimated hatch date for nests in a specific region , or management efforts could be scheduled prior t o mean estimated hatch date (12 May). Broad - scale use of egg oiling or other techniques aimed at reducing number of hatch ed young per nest may be of limited benefit at current levels of mute swan abundance in Michigan. Widespread egg oiling for large populations of mute swans is labor and cost intensive (Hindman et al. 2014) . Additionally, realized benefit of egg oiling in Michigan will be low due to high natural cygnet and brood mortality rate s. Our results indicate that egg oiling will be most effective for pairs with highest pred icted cygnet survival. H ighest cygnet survival rates for a population likely occur in areas where access to nesting and brood - rearing resources is high and competition from other nesting pairs is low. Such areas are likely to be small secluded wetlands or isolated shallow lakes with few breeding pairs in the core of their distribution; however, these could include larger wetland complexes and lakes along an invasion front where intraspecific competition is low. 43 LITERATURE CITED 44 LITERATURE CITED Anderson, D. R., and K. P. Burnham. 2002. Avoiding pitfalls when using information - theoretic methods. Journal of Wildlife Management 66:912 - 918. Bacon, P., and M. Mountford. 1990. The effects of genotypic and yearly variations on the egg volumes of the Mute Swan Cygnus olor. Wildfowl 41:7 - 12. Bacon, P. J. 1980. Status and dynamics of a Mute Swan Cygnus olor population near Oxford England UK between 1976 and 1978. Wildfowl 31:37 - 50. Baldassarre, G. 2014. Ducks, geese, and swans of North America. Volume 1.J ohns H opkins U niversity Press. Baltimore, MD, USA. Birkhead, M. 1984. Variation in the weight and composition of Mute Swan (Cygnus olor) eggs. The Condor 86:489 - 490. Birkhead, M., P. J. Bacon, and P. Walter. 1983. Factors affecting the breeding success of the Mute Swan Cygnus olor. Journal of Animal Ecology 52:727 - 741. Boellstorff, D. E., D. W. Anderson, H. M. Ohlendorf, and E. J. O'Neill. 1988. Reproductive effects of nest - marking studies in an American White Pelican colony. Colonial Waterbirds:215 - 219. Campbell, B. 1960. The Mute Swan census in England and Wales 1955 56. Bird Study 7:208 - 223. Center for Shared Solutions and Technology Partnerships. 2015. Michigan GIS open data p ortal. < http://gis.michigan.opendata.arcgis.com/ >. Ciaranca, M. A., C. C. Allin, and G. S. Jones. 1997. Mute Swan. Birds of North America 273:1 - 28. Coleman, A., and C. Minton. 1979. Pairing and bre eding of Mute Swans in relation to natal area. Wildfowl 30:27 - 30. Conover, M. R., and G. S. Kania. 1999. Reproductive success of exotic Mute Swans in Connecticut. Auk 116:1127 - 1131. Conover, M. R., J. G. Reese, and A. D. Brown. 2000. Costs and benefits o f subadult plumage in Mute Swans: Testing hypotheses for the evolution of delayed plumage maturation. American Naturalist 156:193 - 200. 45 Czapulak, A. 2002. Egg size variation in mute swans: Its influence on egg hatchability, cygnet body size and cygnet surv ival. Waterbirds 25:250 - 257. Dinsmore, S. J., G. C. White, and F. L. Knopf. 2002. Advanced techniques for modeling avian nest survival. Ecology 83:3476 - 3488. Edelaar, P., and J. L. Tella. 2012. Managing non - native species: don't wait until their impacts are proven. Ibis 154:635 - 637. Eltringham, S. K. 1966. Survival of Mute Swan cygnets. Bird Study 13:204 - 207. Gayet, G., C. Eraud, M. Benmergui, J. Broyer, F. Mesleard, H. Fritz, and M. Guillemain. 2011. Breeding Mute Swan habitat selection when accounting for detectability: a plastic behaviour consistent with rapidly expanding populations. European Journal of Wildlife Research 57:1051 - 1056. Gelston, W., and R. Wood. 1982. The Mute Swan in northern Michigan. Traverse City, MI , USA . Hartman, C. A., and L. W. Oring. 2006. An inexpensive method for remotely monitoring nest activity. Journal of Field Ornithology 77:418 - 424. Hindman, L. J., W. F. Harvey IV, and L. E. Conley. 2014. Spraying corn oil on Mute Swan Cygnus olor eggs to preven t hatching. Wildfowl 64:186 196. Homer, C., J. Dewitz, L. Yang, S. Jin, P. Danielson, G. Xian, J. Coulston, N. Herold, J. Wickham, and K. Megown. 2015. Completion of the 2011 National Land Cover Database for the conterminous United States representing a d ecade of land cover change information. Photogrammetric Engineering & Remote Sensing 81:345 - 354. Hoyt, D. F. 1979. Practical methods of estimating volume and fresh weight of bird eggs. Auk:73 - 77. Johnson, D. H., and T. L. Shaffer. 1990. Estimating nest s uccess - when Mayfield wins. Auk 107:595 - 600. King, R. S., J. J. Trutwin, T. S. Hunter, and D. M. Varner. 2013. Effects of environmental stressors on nest success of introduced birds. The Journal of Wildlife Management 77:842 - 854. Klett, A. T., H. F. Due bbert, C. A. Faanes, and K. F. Higgins. 1986. Techniques for studying nest success of ducks in upland habitats in the prairie pothole region. DTIC Document. Knapton, R. W. 1993. Population status and reproductive - biology of the Mute Swan, Cygnus - olor, at Long Point, Lake Erie, Ontario. Canadian Field - Naturalist 107:354 - 356. 46 Mack, R. N., D. Simberloff, W. Mark Lonsdale, H. Evans, M. Clout, and F. A. Bazzaz. 2000. Biotic invasions: causes, epidemiology, global consequences, and control. Ecological applicati ons 10:689 - 710. MacNeil, J. E., and R. N. Williams. 2014. Effects of timber harvests and silvicultural edges on terrestrial salamanders. PloS one 9:27. Munro, R. E., L. T. Smith, and J. J. Kupa. 1968. Genetic basis of color differences observed in Mute S wan (Cygnus olor). Auk 85:504 - 505. Nelson, C. H. 1976. Color phases of downy Mute Swans. Wilson Bulletin 88:1 - 3. Nummi, P., and L. Saari. 2003. Density - dependent decline of breeding success in an introduced, increasing Mute Swan Cygnus olor population. Journal of Avian Biology 34:105 - 111. O'Connor, R. S., and G. Ritchison. 2013. Notes on the incubation, brooding, and provisioning behavior of Chuck - will's - widows. Wilson Journal of Ornithology 125:815 - 819. Perrins, C. M., and C. M. Reynolds. 1967. A preliminary study of the Mute Swan, Cygnus olor. The Wildfowl Trust. 18 th Annual Report. p. 74 - 84. R Development Core Team. 2018. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. < https://www.R - project.org/ >. Reese, J. G. 1975. Productivity and management of feral Mute Swans in Chesapeake Bay. The Journal of Wildlife Management 39:280 - 286. _____. 1980. Demography of European Mute Swans in C hesapeake Bay. Auk 97:449 - 464. Reiter, D. K., M. W. Brunson, and R. H. Schmidt. 1999. Public attitudes toward wildlife damage management and policy. Wildlife Society Bulletin:746 - 758. Reynolds, C., G. Stanfield, and N. Elkins. 1965. The survival of Mute Swan cygnets. Bird Study 12:128 - 134. Roznik, E. A., and R. A. Alford. 2012. Does waterproofing Thermochron iButton dataloggers influence temperature readings? Journal of Thermal Biology 37:260 - 264. Schaetzl, R. J., H. Enander, M. D. Luehmann, D. P. Lusch , C. Fish, M. Bigsby, M. Steigmeyer, J. Guasco, C. Forgacs, and A. Pollyea. 2013. Mapping the physiography of Michigan with GIS. Physical Geography 34:2 - 39. Schneider, E. G., and S. R. McWilllams. 2007. Using nest temperature to estimate nest attendance o f piping plovers. Journal of Wildlife Management 71:1998 - 2006. 47 Sedinger, J. S. 1990. Effects of visiting Black Brant nests on egg and nest survival. The Journal of Wildlife Management:437 - 443. U. S. Fish and Wildlife Service. 2015. National wetlands inventory website. U.S. Department of the Interior, Fish and Wildlife Service, Washington, D.C. < http://www.fws.gov/wetlands/ >. Walter, S. E., and D. H. Rusch. 1997. Accuracy of egg flotation in determ ining age of Canada Goose nests. Wildlife Society Bulletin 25:854 - 857. Westerskov, K. 1950. Methods for determining the age of game bird eggs. Journal of Wildlife Management 14:56 - 67. White, G. C., and K. P. Burnham. 1999. Program MARK: survival estimati on from populations of marked animals. Bird Study 46:S120 - S139. Wieloch, M. 1991. Population trends of the Mute Swan Cygnus olor in the Palearctic. Wildfowl:22 - 32. Wieloch, M., and A. Czapulak. 1991. Cygnus olor immutablis in Poland. Wildfowl:304 - 309. W illey, C. H. 1968. The ecological significance of the Mute Swan in Rhode Island. Transactions of the Northeast Wildlife Conference v. 25,p. 121 - 134 Willey, C. H., and B. F. Halla. 1972. Mute Swans of Rhode Island. Rhode Island Department of Natural Resour ces, Division of Fish and Wildlife. Wlodarczyk, R., M. Wieloch, S. Czyz, P. T. Dolata, and P. Minias. 2013. Natal and breeding dispersal in Mute Swans Cygnus olor: influence of sex, mate switching and reproductive success. Acta Ornithologica 48:237 - 244. Wood, R., and W. L. Gelston. 1972. Preliminary report: the Mute Swans of Michigan's Grand Traverse Bay region. State of Michigan, Department of Natural Resources, Wildlife Division. 48 CHAPTER 3: DENSITY DEPENDENCE I N PRODUCTIVITY OF A NORTH AMERI CAN MUTE SWAN POPULA TION INTRODUCTION Lack (1954) noted that wildlife populations tend to fluctuate around a certain number rather than growing indefinitely at a sustained rate. This observation, density - dependent regulation of populations, has become a fundamental underpinning of many wildlife populatio n models although its pervasiveness and generalizability ha s sparked debate in the scientific literature (Hanski et al. 1993, Berryman 2004, White 2007) . Definitive de monstration of density dependence in wild populations has historically proven difficult (Lack 1966) due to the need for long - term datasets on population demographics (Hassell et al. 1989, Godfray and Hassell 1992) although density dependence is implicated for many species (Woiwod and Hanski 1992, Lima and Jaksic 1998, Nummi and Saari 2003, Gunnarsson et al. 2013) . Newton (1998) note d that density dependence likely manifests through principle demographic factors (e.g., birth, death, and emigration rates) and could be regula ted by intraspecific resource competition. Lebeuf and Giroux (2014) and Sedinger et al. (1998) suggest ed that density dependence wa s important in determining outcome s of reproductive effort for territorial waterbirds and manifest ed through several mechanisms (Ferrer and Donazar 1996) . Lack (1966) and Fretwell and Lucas (1969) proposed that overall productivity would decrease and variation between individuals would remain stable with increasing density. They hypothesized that increased agonistic interactions would accompany increased density and, therefore, would reduce mean territory quality for all individuals regardless of realized quality at lower densities. Kadmon (1993) suggested that negative density dependence in productivity could also be explained by heterogeneity in habitat quality for pairs. T his would manifest in a 49 population in which productivity declines with increased density, but variation in breeding performance among individuals increases rather than be equal as proposed by Lack (1966) . Variation in individual performance relating to habitat heterogeneity follows the hypothesis that the best habitats are filled first (Hildén 1965, Ferrer and Donazar 1996, Rodenhouse et al. 1997, Lovette and Fitzpatrick 2016) . Ferrer and Donazar (1996) summarize d and investigate d these 2 hypotheses (i.e., habitat heterogeneity hypothesis, [hereafter, HHH]; and interference hypothesis, [hereafter, IH]) for a population of Spanish imperial eagles ( Aquila adalberti ) and conclude d that HHH is the regulating mechanism for population growth rates. However, Sergio and Newton (2003) note d the impo rtance and difficulty of distinguishing between quality territories and quality individuals when investigating support for or differentiation between HHH and IH. Understanding mechanisms triggering density - dependent relationships is important for managing harvested (Gunnarsson et al. 2013) , reintroduced (Armstrong et al. 2005) , special concern (Carrete et al. 2006) , and invasive species (Nummi and Saari 2003) . Nummi and Saari (2003) conducted a longitudinal study (1976 1998) to analyze reproductive parameters f or mute swans ( Cygnus olor ) in an introduced range of the Finnish archipelago. They hypothesized that breeding success would differ between territories of varying quality , with quality measured as length of time it has been occupied. Density of breeding pa irs was inversely related to clutch size, brood size, and fledged young per pair. Nummi and Saari (2003) found several cygnets with crushed skulls in high density areas which is evidence for IH; however, the y also found that site s occupied longest (i.e., were of highest quality) produced more young , and that coefficient of variation for brood size increased with increasing density. Their results, much like Ferrer and Donazar (1996) , predominantly provide support for HHH which results in lower survival of young due to increased variation in territory quality of nesting pairs . 50 Their findings and accounts published elsewhere in the scientific literature (Lack 1954, Sedinger et al. 1998) signif ied that dynamics in the brood rearing phase for birds become increasingly important as densities increase (Ferrer et al. 2008) . We assessed influence of density dependence on productivity for mute swans in ounty in 1919. Population growth was slow throughout the 1900s; however, their population grew to an estimated 17,520 individuals in 2013 (D. Luukkonen, unpublished data) making it the largest population of mute swans in North America. The Michigan Departm ent of Natural Resources (DNR) formalized their management goals and objectives for mute swans in 2012 (Michigan Department of Natural Resources 2012) . This policy established a long - term goal of no more than 2,000 mute swans in Michigan by 2030 as determined via their annual breeding waterfowl survey; however, this policy did not outline levels of control needed to accomplish that goal due to uncertainty in demographic parameters for this introduced population. Furthermore, there w ere no empirica l data to examine the role of density as it relates to breeding productivity within North America despite it being demonstrated in other introduced populations (Nummi and Saari 2003) . Ellis and Elphick (2007) mention ed that density dependenc e was likely occurring in mute swan populations that have been established for more than three decades; therefore, investigation of density dependence is warranted for well - established sub - populations in Michigan. W e observed productivity under a range of breeding densities within the core mute swa n range in Michigan and compared densities relative to amount of characteristic nesting cover to identify whether incorporation of density dependence in breeding productivity is appropriate for p opulation modeling. 51 STUDY AREA Site Selection The core of breeding range for Michigan mute swans is found in the Lower Peninsula ( Michigan Department of Natural Resources , unpublished data). We overlaid mean estimated spring de nsity of mute swans in 2011 2015 with hydrography data (Center for Shared Solutions and Technology Partnerships 2015) in a geographic information system (GIS; ArcGIS 10.3.1, ESRI, Redlands, CA, USA ) to identify waterbodies that may harbor breeding pairs of mute swans. We stratified the Lower Peninsula of Michigan using township boundaries as outlined Technolo gy Partnershi ps 2015). We performed a query in GIS to identify all PLSS townsh ips where mean estimated spring density was per 259 ha (roughly 1 per square mile ) during 2011 2015 a nd took the mean shoreline distance within those identified townships (40 km; a coarse proxy for nesting cover potential) to estimate where mute swan presence was likely in the Lower Peninsula during the following breeding season (i.e . , 2016). We identified 15 p reliminary study sites (6 x 6 km each ) using these methods (Figure 3.1). We used fixed - wing aircraft (Cessna 185; Northwoods Aviation Inc, Cadillac, MI, USA) to survey p reliminary study areas for mute swan presence and nesting habitat in December 2015 or April 2016. We chose 5 study sites in Bay (Tobico) , Cass (Juno) , Kent (Wabasis) , Oakland (Pontiac) , and St. Clair (St. Clair) Counties to investigate density dependence in breeding productivity 2016 2018 (Figure 3.1). One additional site (6 by 6 km) in Antrim County (Clam) was included in this investigation for 2017 2018 (Figure 3.1). 52 Figure 3 .1. Preliminary an d final study sites 2016 2018 were located throughout the Lower Peninsula of Michigan, USA, within selected public land survey system (PLSS) townships that Site Descriptions Land cover composition varies latitudinally in the Lower Peninsula of Michigan, USA, (Homer et al. 2015) , and concomitantly varied across our 4 inland (Juno, Pontiac, Wabasis, Clam) and 2 coastal study sites (Tobico, St. Clair) . Inland waterbodies contained a mix of natural and developed shoreline with moderate to heavily developed upla nd areas adjacent to areas of permanently - flooded open water (L1UBH; U. S. Fish and Wildlife Servic e 2015). Natural shoreline consisted of characteristic mute swan nesting cover (PEM; U. S. Fish and Wildlife Service 2015 ) such as broad - leaved cattail ( Typha latifolia ), bulrushes ( Schoenoplectus spp.), narrow - leaved cattail ( Typha angustifol ia ), and phragmites ( Phragmites australis ) or woody 53 vegetation (buttonbush [ Cephalanthus occidentalis ] , willow [ Salix spp.], ash, [ Fraxinus spp.], maple [ Acer spp.], cottonwood [ Populus deltoides ], and oak [ Quercus spp.]). The 2 coastal sites included areas of persistent emergent vegetation that is semi - permanently flooded (PEM1F) and open water (PABG; U. S. Fish and Wildlife Service 2015) among areas of agriculture, human development, and forest cover (Homer et al. 2015) . The St. Clair study site was primarily open water (L1UBH and L2UBH) with large areas of freshwater wetlands that consisted of em ergent vegetation ( PEM; e.g., broad - leaved cattail, bulrushes, narrow - leaved cattail, and phragmites). The St. Clair study site also had small developed islands (< 5 ha each) dispersed within the matrix of open water and emergent vegetation. METHODS Field methods Nest density We used boat and aerial surveys to detect actively nesting pairs within study sites and determine GPS coordinates of all observed nests. Annual aerial surveys of nesting mute swans were conducted (12 April 1 May) for sites ( n = 5 , 2016 ; n = 6, 2017 2018). Two observers worked together to detect incubating females on the large ( 1 m) conspicuous nests (Conover and Kania 1999) against the backd rop of senesced emergent vegetation . Cooper (1979) and Kear ( 1972) estimated that incubation recesses for female swans were < 30 minutes daily; therefore, detection probability of active nests was likely near 1 during low - level aerial surveys . All detected nests were recorded on orthophotographs and were later transferred to a GIS. We recorded flights using wing strut - mounted video camera systems in 2016 2017 ( MotoCam 360, Bothell, WA, USA ) and 2018 ( GoPro Hero 4 Silver, San Mateo, California, USA ) to aid in 54 determining exact physical location of nesting pairs. We documented flight paths with 1 second GPS fix intervals using a cellphone application (Strava, Inc., San Francisco, CA USA). Breeding productivity Aerial and boat surveys were used to estimate breeding productivity per pair near estimated fledging (1 September) for all sites. Aerial surveys were flown in fixed - wing aircraft with flight paths and flight video recorded in the same manner as the spring nest detection surveys. We counted white (i.e., adult or leucistic - morph cygnets) and gray swans (i.e., gray - morph cygnets) and recorded approximate location of all pairs and broods on aerial photos which were later transferred to a GIS. We used boats with in study sites to confirm brood size and color morph ratio (i.e., count leucistic - morph cygnets separately from adults). N umber of fledged cygnets per site w as compared to total number of nesting pairs to derive an estimate of productivity that includes failed nests and failed broods. Characteristic nesting cover We document ed extent and location of characteristic nesting cover (i.e., cattails, bulrush, and phr agmites ; Ciaranca et al. 1997 ) available to mute swans ( i.e., dense vegetation adjacent to water; Baldassarre 2014 ) for all study sites during the 2018 nesting season. We recorded where stands of c haracteristic cover bordered water on recent (2014 2016) orthophotographs and then conducted in - field surveys to verify detection of characteristic cover using orthophotographs . Location of characteristic nesting cover was transferred from orthophotograp hs to digital polyline features using a GIS ( Arc GIS Pro 2.1.2 , ESRI, Redlands, CA, USA) . 55 Data analysis Nest spacing We determined median distance to closest conspecific nest for all detected nests within study site boundaries 2016 2018. We used a GIS t o determine Euclidean distance between mute swan nests that resided on the same waterbody ( n = 143 comparisons). Calculating nest spacing in this manner eliminated measurements of nest spacing between adjacent but distinct waterbodies which may be biologic ally irrelevant since mute swans are nearly fully aquatic (Sousa et al. 2008) and are likely not directly influenced by presence of pairs on adjacent waterbodies during the nesting period. Nest spacing measurements were summarized for each year with sites pooled because we wished to understand the typical conspecific nest spacing across sites. Digitization of characteristic nesting cover We used a GIS and recent (2014 2016) high - resolution (< 1 m per pixel) leaf - off (April) orthophotos to manually digitize transition between open water and other cover types (e.g., emergent herbaceous vegetation, forests, developed land). Digitizing transition between open water and adjacent cover types allowed us to use this border for predicted placement of characteristic nesting cover. We conver ted open water polygons to line features using a GIS. We then subset the open water line features into 2 categories (i.e., areas adjacent to characteristic nesting cover and areas adjacent to other cover types). We created a new line feature that contained all the segments whe re characteristic nesting cover was immediately adjacent to open water. We created equally - spaced points (10 m spacing) along areas of characteristic nesting cover to represent potential nest locations for the optimization process (Fi gure 3.2). 56 Saturation of nesting cover We were interested in understanding maximum nesting pair density possible (i.e., saturation) for the 6 study sites given arrangement of characteristic nesting cover and observed spacing of conspecific nests. Our esti mates of saturation assume that mute swans optimally space nests in characteristic cover to obtain the highest possible number of nests. Mute swans are likely not optimally spacing nests in this manner (see results); however, this method provides a liberal estimate of nest ing pair saturation to which actual density can be compared. Actual pair densities that are near or exceed estimated saturation densities (i.e., saturation ratios of 7 ) and instances of swans establishing nesting territories outside of characteristic nesting cover will indicate that site nesting density is likely at saturation especially since mute swan nest ing pairs are likely spaced sub optimally within characteristic cover . We used a manual spatial optimization approach through GIS to estimate saturation of characteristic nesting cover. Our methods were conceptually similar to the anti - covering location problem (ACLP; Moon and Chaudhry 1984 , Murray and Church 1997 ) approach employed by Downs et al. (200 8) to estimate nesting carrying capacity for territorial sandhill cranes. Our manual optimization methods likely approximate estimates that could be derived through mathematical optimization in this system due to the discrete patches of characteristic n esting cover (Figure 3.2); however, the exact placement of nests within the cover, which we are uninterested in, may slightly differ between a manual and mathematical optimizatio n. Comparison of observed breeding productivity among sites The functional re lationship between breeding productivity (i.e., the number of fledglings per nesting pair) and nesting pair density i s important in determining density dependence . We conducted a linear regression in program R (R Development Core Team 2018) to examine 57 relationship of productivity to observed pair density and to the ratio of observed pairs to estimate saturation . We ranked our competing small sample sizes (AIC c ; Anderson and Burnham 2002 ). This allowed us to see if adjusting the observed number of nest ing pairs on each site by a vailability of characteristic nesting cover (i.e., calculating an ecological density) furthered our understanding of the relationshi p between nesting pair densit y and productivity. Figure 3.2. Example d etermination of potential nesting locations (white circles) within characteristic nesting cover for mute swans (black line) at the transition of open water and adjacent cover types for focal waterbodies in the Lower Peninsula of Michigan, USA. RESULTS We detected 228 pairs of mute swans with nests within study site boundaries 2016 2018 (Table 3.1) using aerial and boat surveys. Median distance between closest conspecific nest was 58 418.1 m ( = 495.1, SD = 388.7; range : 22.6 2959.2). Mean number of nests and number of fledged young per pair 2016 2018 varied among the 6 equal - sized study areas (3.5 26 nests per site and 0.6 2.3 fledglings per pair ; Table 3.1). Table 3.1. Reproductive parameters for 6 equal - sized study areas (36 km 2 each) ordered by increasing latitude in the Lower Peninsula of Michigan, USA, 2016 2018. *Total does not include nests that were influenced by investigators or nests of pairs culled during incubation in official removal efforts , Fl. = Fledge, Sat. = Saturation M ean number of fledged young per pair tended to increase with decreasing pair density (Adjusted R 2 = 0.1798; Figure 3.3a); however, the site with the lowest number of nesting pairs a lso fledged the fewest young (Table 3.1). Converting observed pair densities to saturation ratios (i.e., ratio of actual nesting pairs to estimated saturation) provided a slightly better fit to the data (Adjusted R 2 = 0.1989) although variation among sites was still evident (Figure 3.3b). The most parsimonious model in the linear regression of breeding productivity (Table 3.2) included a fixed effect for ratio of observed nesting pairs to estimated saturation ( ^ = - 0.9792 , p = 0.04). The next competing model was within 2 AIC c units of the top model and contained a fixed effect for number of nesting pairs ( ^ = - 0.0392 , p = 0.05; Table 3.2). C orrelation between observed number of nesting pairs and estimated saturation ratio was 0.6273 ; therefore, we did not include the se covariates together in an additive model. 59 (A) Figure 3.3. Comparison of breeding productivity to number of pairs per site (A) and breeding productivity to estimated nesting pair saturation ratio (B) 2016 2018 for 6 equal - sized study sites in the Lower Peninsula of Michigan, USA. 60 Figure 3. 3 (B) Table 3.2. Model selection for l inear regression on breeding productivity of mute swans in Michigan, USA, du ring 2016 2018 . Model k a AIC c a c a w i a Saturation Ratio 1 38.085 0 0.372 Number of Pairs 1 38.486 0.401 0.304 N ull Model 1 40.952 2.867 0.089 a k = number of parameters in model, AIC c = Akaike's Information Criterion adjusted c = difference between AIC c of best fitting and current model w i = Akaike's weigh t Nesting pairs were not spaced optimally in characteristic nesting cover (Figure s 3.4 and 3.5). Sites nearest saturation of nesting pair density tended to have pairs nesting in non - characteristic nesting cover (Figure 3.4) , whereas pairs almost exclusively nested in characteristi c cover on less saturated sites (Figure 3.5). Uncorrected estimates of pair density (i.e., pairs per km 2 ) provided a suitable estimation of pair saturation (i.e., saturation ratios) for 4 61 of the 6 sites (Juno, Pontiac, Wabasis, and Tobico); however, interp retation of pair saturation changed on 2 sites (Clam and St. Clair) when adjusting observed pair density by expected saturation of pairs (Table 3.1). The St. Clair site had a low saturation ratio (Table 3.1) due to the large amount of characteristic cover present despite having a high number of nesting pairs. The Clam site had few nesting pairs, but also had a paucity of well - spaced characteristic nesting cover resulting in a site that was near saturation even though the observed number of pairs was low (Ta ble 3.1). 62 (A) Figure 3.4. Comparison of nest locations 2016 2018 within characteristic and non - characteristic nesting cover for select waterbodies within the Juno site in Cass County, MI, USA (A) and Pontiac site in Oakland County, MI, USA (B) which had nesting pairs establishin g territories outside of characteristic nesting cover and low observed productivity. 63 Figure 3. 4 (B) 64 (A) Figure 3.5. Comparison of nest locations 2016 2018 within characteristic and non - characteristic nesting cover for select waterbodies within the Wabasis site in Kent County, MI, USA (A) and Tobico site in Bay County, MI, USA (B) which had most nesting pairs use terr itories in characteristic nesting cover and higher observed productivity. 65 Figure 3. 5 (B) DISCUSSION Effects of density dependence in territorial avian species typically first manifest in aspects of breeding productivity like hatching success (Lebe uf and Giroux 2014) , growth of young (Sedinger e t al. 1998) , or nearly all breeding parameters (Nummi et al. 2015) . Longitudinal studies of breeding parameters under naturally fluctuating species abundance s are typically used to demonstrate presence or absence of dens ity dependence (Godfray and Hassell 1992) ; however, the mechanisms causing this pattern are not always identified. Two hypotheses have been proposed to explain the mechanism by which density of territorial species effects breeding performance. Kadmon (1993) and Rodenhouse et al. (1997) argue d that heterogeneity in habitat suitability (i.e., HHH) influences reproductive performance for populations resulting in 66 a lower mean productivity and increased variance in productivity at higher densities . A second hypothesis (i.e., IH) asserts that agonistic interactions between conspecifics at higher densities lowers overall productivity for all pairs resulting in similar variance under high and low densities (Lack 1966, Sutherland 1996) . Our approach all owed us to look for the presence of and mechanism by which density dependence is acting on Lower Peninsula. We found evidence for density dependence in breeding productivity for mute swans in Michigan. Mean productivit y per pair declined as number of breeding pairs per site increased (Figure 3.3a). Our range in observed nesting density across (0.1 nesting pairs/km 2 0.7 nesting pairs/km 2 ) was similar to values estimated by Nummi and Saari (2003) in early ( 0.1 nesting pairs/km 2 ) and late (0.6 nesting pairs/km 2 ) stages of invasion for part of a Finnish archipelago. Nummi and Saari (2003) note d that density of nesting pairs during late stages of invasion was the highest rec orded density in Europe for a non - colonial population. We observed similar extreme nesting densities and low productivity on 2 inland sites and 1 coastal site (Table 3.1) in the fewest young. We found that this site (Clam) was near saturation of characteristic nesting cover despite the low number of nesting pairs (Table 3.1); therefore, it was similar to sites with higher nesting pair densities . Interestingly, a site (St. Clai r) with many nesting pairs that fledged few young was not close to estimated saturation of nesting cover using this methodology (Table 3.1). These results for St. Clair could be due to its uniqueness among the study sites (i.e., an open water site located in the largest contained freshwater delta in North America) which could potentially have other extrinsic factors limiting cygnet survival not encountered on inland sites nor captured in this analysis (e.g., storm surges, cooler water temperatures). Additio nally, our estimated 67 saturation level using these methods may be unrealistic in that other factors may become limiting before nesting habitats become saturated . Density of nesting pairs under our estimated saturation for St. Clair (Table 3.1) would be 1.86 pairs per km 2 which is 3 times higher than any reported mute swan pair density outside of colonial populations (Nummi and Saari 2003) . Therefore, we considered St. Clair to be a high - density site since its observed density of nesting pairs (Table 3.1) is near the maximum density reported in the literature (Nummi and Saari 2003) . Further, nesting pair density was reduced on the St. Clair site under permit by the U.S. Department of Agriculture Wildlife Services during the final year of investigation to reduce human wildlife conflict. Nesting pairs that remained fledg ed more cygnets per pair and utilized larger areas during brood rearing than in previous years (R. Knapik, unpublished data). This suggests that density impacts were realized at observed nesting pair density in previous years even though it was below the e stimated saturation. Our optimization methods to estimate saturation of characteristic nesting cover aligned well with observed density and productivity f or the other sites (5 of 6 total sites; Table 3.1). Our spatial comparison of nesting pair density to characteristic nesting cover provided insight s into the mechanism by which density is influencing breeding productivity. Our results support the HHH (Andrewartha and Birch 1954, Kadmon 1993, Rodenhouse et al. 1997) . Pairs almost exclu sively nested in characteristic cover on sites with few nesting pairs and unfilled characteristic nesting cover remained ( e.g., Wabasis, Tobico, and Clam; Figure 3.5) whereas pairs filled characteristic nesting cover and nested , presumably, in suboptimal a reas on sites at or near estimated saturation ( e.g., Juno, St. Clair, and Pontiac; Table 3.1, Figure 3.4). Additionally, mean brood survival was lower ( 0.58 ± 0.03; Chapter 2 ) in Michigan when compared to other introduced populations (Conover and Kania 1999) despite a normal mean brood size at fledging 68 for pairs that fledged young (3.1 cygnets/pair; Chapter 2) . This means that pairs which successfully fledged young did so with brood sizes comparable to other areas of their introduce d range (Conover and Kania 1999) despite a n overall lower mean breeding productivity across sites (1. 4 2 cygnets/pair ; Table 3.1 ) compared to low - density areas of their introduced range (Reese 1975, Conove r and Kania 1999, Nummi and Saari 2003) . Our d istribution of brood sizes at fledging (Chapter 2) is expected for territorial long - lived bird species exhib iting density dependence due to heterogeneity in nesting habitat (Ferrer et al. 2008) . These findings lend further credit to the HHH which expects increased variance in mean brood size produced per pair (Andrewartha and Birch 1954, Rodenhouse et al. 1997) rather than a uniform reduction in number of fledged young across all pairs (Lack 1966) . We cannot claim that agonistic intera ctions (i.e., IH) had no effect on productivity because both HHH and IH can simultaneously occur (Ferrer and Donazar 1996, Krüger et al. 2012) ; however, our observed patterns provide the most support for HHH rather than IH in Michigan mute swans . We could not control for all factors potentially influencing productivity o n sites . Furthermore, w e could not separate effects of individual and territory quality in this short - term study because not all nesting individuals within sites were uniquely marked (Chapter 2). We noted 10 instances on sites with highest observed nesting pair densities where territories and exact nesting mounds in characteristic nesting cover were immediately taken over by new pairs following dissociation of a nesting pair (i.e., through death of a mate or pair; Chapter 2) . T herefore, we could not assume that unmarked individuals observed on territories were constant between years. The s urviving member of pairs joined non - breeding flocks and did not nest throughout the remainder of the study (R. Knapik, unpublished data) signaling high competition for territories in characteristic cover. Additionally , presence and abundance of non - breeding 69 flocks could have also influenced productivity; however, non - breeding flocks were successfully excluded from areas with actively nesti ng pairs especially after hatching of cygnets (R. Knapik, personal observation). Birkhead et al. (1983) demonstrated that inexperienced pairs had slightly lower breeding product ivity than experienced pairs; however, inexperienced pairs still produced young. Therefore, while individual quality may be partly confounded with territory quality for mute swans in this study, variation in individual quality is no t likely to be the mecha nism driving our observed support for the HHH. Our evidence for habitat - mediated density dependence in breeding productivity of an introduced North American mute swan population is similar to findings reported by Nummi and Saari (2003) for an introduced population in a Finnish archipelago. Further, we demonstrated that strength of density dependence in breeding productivity varies spatially within Michig an, USA, based on local dynamics of nesting pairs and coverage of characteristic nesting habitat. We argue that density - mediated breeding productivity should be considered when developing demographic models for North American mute swan populations especial ly when using regionally - estimated demographic parameters. MANAGEMENT IMPLICATI ONS The Michigan DNR has chosen to pursue a wildlife damage management approach to invasive mute swans in the Great Lakes region rather than a targeted eradication program (Michigan Department of Natural Resources 2012) . A wildlife damage management program will likely be a more successful form of mute swan management in the short term since immigration is likely still occurring from neighboring provinces and states , and not all reproductive swans can be targeted for management (Bomford and O'Brien 1995) due to als. While management of invasive species should be initiated when abundance is low (Usher 1989, Edelaar and Tella 2012) , wildlife 70 damage management programs for established invasive species can be successful if they incorporate local demography and proclivities of the species (Bomford and O'Brien 1995) . Such programs must also be cognizant of potential density - dependence in demographic rates (Newton 1998, Nummi and Saari 2003) . We d erive d region specific demographic rates for mute swans through this research (Chapter 2 and 4) which will parameterize matrix population models (Chapter 5) aimed at guiding in - field management. Further, this analysis not only provides evidence that density dependence should be incorporated into breeding productivity of matrix population models (Table 3.1), but it provides practical in - field guidelines for efficient methods to perform culling of s wans or destroying of nests. Targeted removal of breeding pairs should be first prioritized for areas which are likely to be most productive. These are lakes and wetlands where the number of breeding pairs is low but characteristic nesting cover (i.e., cat tails, phragmites, or reeds next to shallow open water) is abundant . Our results indicate that these pairs are recruiting the most immature swans into the population; however, we also recognize that targeting low density areas increases cost s needed to rem ove each swan in the short term, but the long - term costs of management should be lower since fewer total swans would need to be removed (Chapter 5; Ellis and Elphick 2007) . Post - removal surveillance of these areas should occur to ensure that pairs do not return because newly colonizing swans will likely have high breeding productivity while nesting pair density is low. Removal of swans in summer or winter concentration areas rema ins a viable management strategy; however, these efforts should also be focused in areas that have low nesting densities but large amounts of characteristic nesting cover represent ing high potential productivity areas. Although mute swans are capable of la rge movements, they tend to move to the closest open 71 water areas during summer molting and winter which is particularly true for established pairs (Chapter 4). Therefore, immature or non - breeding swans in summer flocks will likely try to settle in the regi on where they are flocking. Winter flocks of mute swans can contain established breeding pairs, their current and former cygnets, and non - breeding swans (Chapter 4); therefore, these efforts should also be first focused in regions where overall swan densit y is low. This is because local (Chapter 4) established breeding pairs may be culled during winter removal efforts which then effects local breeding density and dynamics the following spring. Anything short of a complete removal of breeding pairs within or adjacent to an area of saturated suitable nesting cover could allow remaining pairs to be more productive due to lower nesting density (Table 3.1; Nummi and Saari 2003 ). Our evidence for habitat - mediated density dependence in breeding productivity also has implications for lethal management options targeted during incubation stage , such as egg oiling. Oiling mute swan eggs during incubation is a highly effect ive method for preventing hatch (Hindman et al. 2014) ; however, it is no t effective at reducing the overall population in the short term (Ellis and Elphick 2007, Hindman et al. 2014) . Nevertheless, it is a method that can have localized impact, reduce summer population of cygnets (Hindma n et al. 2014) , and is sometimes the only management option desired by landowners and lake associations. Our results show that egg oiling will be most effective when the probability of cygnet survival is high (i.e., when breeding pair densities are low) ; therefore, egg oiling procedures should first be focused on lakes and wetlands with few pairs or with pairs that have proven ability to produce and fledge cygnets. Egg oiling in areas where intraspecific competition is high and where pairs are nesting in non - characteristic nesting cover (Figure 3.4) will be inefficient since most pairs would no t have fledged young anyway. We demonstrated that management aimed at achieving long - term 72 population goals (Michigan Department of Natural Resources 2012) should be cognizant that density dependence in breeding productivity is occurring in Michigan. In - field and demographic modeling should account for this relationship as populations are reduced . 73 LITERATURE CITED 74 LITERATURE CITED Anderson, D. R., and K. P. Burnham. 2002. Avoiding pitfalls when using information - theoretic methods. Journal of Wildlife Management 66:912 - 918. Andrewartha, H. G., and L. C. Birch. 1954. The distribution and abundance of animals. University of Chicago Press. Chicago, IL, USA. Armstrong, D. P., R. S. Davidson, J. K. Perrott, J. Roygard, and L. Buchanan. 2005. Density - dependent population growth in a reintroduced population of North Island Saddlebacks. Journal of Animal Ecology 74:160 - 170. Baldassarre, G. 2014. Ducks, geese, and swans of North America. Volume 1. J ohns H opkins U niversity Press. Baltimore, MD, USA. Berryman, A. A. 2004. Limiting factors and population regulation. Oiko s 105:667 - 670. Birkhead, M., P. J. Bacon, and P. Walter. 1983. Factors affecting the breeding success of the Mute Swan Cygnus olor. Journal of Animal Ecology 52:727 - 741. Bomford, M., and P. O'Brien. 1995. Eradication or control for vertebrate pests? Wild life Society Bulletin 23:249 - 255. Carrete, M., J. A. Sánchez Zapata, J. L. Tella, J. M. Gil Sánchez, and M. Moleón. 2006. Components of breeding performance in two competing species: habitat heterogeneity, individual quality and density dependence. Oikos 112:680 - 690. Center for Shared Solutions and Technology Partnerships. 2015. Michigan GIS open data portal. < http://gis.michigan.opendata.arcgis.com/ >. Ciaranca, M. A., C. C. Allin, and G. S. Jones. 1997. Mute Swan. Birds of North America 273:1 - 28. Conover, M. R., and G. S. Kania. 1999. Reproductive success of exotic Mute Swans in Connecticut. Auk 116:1127 - 1131. Cooper, J. A. 1979. Trumpeter Swan nesting behaviour. Wildfowl 30:55 - 71. Kear, J. 1972. Reproduction and fami l y life. Pages 79 - 124 in P. Scott and The Wildfowl Trust, editor. The swan. Michael Joseph. London, England. Downs, J. A., R. J. Gates, and A. T. Murray. 2008. Estimating carrying capacity for Sand hill Cranes using habitat suitability and spatial optimization models. Ec ological M odelling 214:284 - 292. 75 Edelaar, P., and J. L. Tella. 2012. Managing non - native species: don't wait until their impacts are proven. Ibis 154:635 - 637. Ellis, M. M., and C. S. Elphick. 2007. Using a stochastic model to examine the ecological, economic and ethical consequences of population control in a charismatic invasive species: Mute Swans in North America. Journal of Applied Ecology 44:312 - 322. Ferrer, M., and J. A. Dona zar. 1996. Density dependent fecundity by habitat heterogeneity in an increasing population of Spanish Imperial Eagles. Ecology 77:69 - 74. Ferrer, M., I. Newton, and E. Casado. 2008. Density dependence hypotheses and the distribution of fecundity. Journal of Animal Ecology 77:341 - 345. Fretwell, S. D., and H. L. Lucas. 1969. On territorial behavior and other factors influencing habitat distribution in birds. Acta biotheoretica 19:16 - 36. Godfray, H. C. J., and M. P. Hassell. 1992. Population biology - long time series reveal density dependence. Nature 359:673 - 674. Gunnarsson, G., J. Elmberg, H. Poysa, P. Nummi, K. Sjoberg, L. Dessborn, and C. Arzel. 2013. Density dependence in ducks: a review of the evidence. European Journal of Wildlife Research 59:305 - 321 . Hanski, I., W. Ian, and J. Perry. 1993. Density dependence, population persistence, and largely futile arguments. Oecologia 95:595 - 598. Hassell, M. P., J. Latto, and R. M. May. 1989. Seeing the wood for the trees - detecting density dependence from exi sting life - table studies. Journal of Animal Ecology 58:883 - 892. Hildén, O. 1965. Habitat selection in birds: a review. Annales Zoologici Fennici. Vol. 2. No. 1. Finnish Zoological and Botanical Publishing Board. Hindman, L. J., W. F. Harvey IV, and L. E. Conley. 2014. Spraying corn oil on Mute Swan Cygnus olor eggs to prevent hatching. Wildfowl 64:186 196. Homer, C., J. Dewitz, L. Yang, S. Jin, P. Danielson, G. Xian, J. Coulston, N. Herold, J. Wickham, and K. Megown. 2015. Completion of the 2011 Nation al Land Cover Database for the conterminous United States representing a decade of land cover change information. Photogrammetric Engineering & Remote Sensing 81:345 - 354. Kadmon, R. 1993. Population dynamic consequences of habitat heterogeneity: an experim ental study. Ecology 74:816 - 825. Krüger, O., N. Chakarov, J. T. Nielsen, V. Looft, T. Grünkorn, B. Struwe - Juhl, and A. P. Møller. 2012. Population regulation by habitat heterogeneity or individual adjustment? Journal of Animal Ecology 81:330 - 340. 76 Lack, D . 1954. The natural regulation of animal numbers. The Clarendon Press , Oxford , UK . Lack, D. L. 1966. Population studies of birds. The Clarendon P ress, Oxford, UK . Lebeuf, A. P., and J. F. Giroux. 2014. Density - dependent effects on nesting success of temp erate - breeding Canada Geese. Journal of Avian Biology 45:600 - 608. Lima, M., and F. M. Jaksic. 1998. Population variability among three small mammal species in the semiarid Neotropics: the role of density - dependent and density - independent factors. Ecograph y 21:175 - 180. Lovette, I. J., and J. W. Fitzpatrick. 2016. Handbook of bird biology. John Wiley & Sons. Hoboken, NJ , USA. Michigan Department of Natural Resources. 2012. Mute Swan management and control program policy and procedures. < www.michigan.gov/documents/dnr/2012_Mute_Swan_Policy_378701_7.pdf >. Accessed 01 March 2016. Moon, I. D., and S. S. Chaudhry. 1984. An analysis of network location problems with distance constraints. Management Science 30:290 - 307. Murray, A. T., and R. L. Church. 1997. Solving the anti - covering location problem using Lagrangian relaxation. Computers & Operations Research 24:127 - 140. Newton, I. 1998. Population limitation in birds. Academic P ress. San Diego, CA, USA. Nummi, P., S. Holopainen, J. Rintala, and H. Poysa. 2015. Mechanisms of density dependence in ducks: importance of space and per capita food. Oecologia 177:679 - 688. Nummi, P., and L. Saari. 2003. Density - dependent decline of breeding success in an introduced, increasing Mute Swan Cygnus olor population. Journal of Avian Biology 34:105 - 111. R Development Core Team. 2018. R: A language and environment for statistical c omputing. R Foundation for Statistical Computing, Vienna, Austria. < https://www.R - project.org/ >. Reese, J. G. 1975. Productivity and management of feral Mute Swans in Chesapeake Bay. The Journal of Wildlife Manag ement 39:280 - 286. Rodenhouse, N. L., T. W. Sherry, and R. T. Holmes. 1997. Site dependent regulation of population size: a new synthesis. Ecology 78:2025 - 2042. Sedinger, J. S., S. L. Mark, B. T. Person, M. W. Eichholz, M. P. Herzog, and P. L. Flint. 1998 . Density - dependent effects on growth, body size, and clutch size in Black Brant. Auk 115:613 - 620. 77 Sergio, F., and I. Newton. 2003. Occupancy as a measure of territory quality. Journal of Animal Ecology 72:857 - 865. Sousa, C. M., R. A. Malecki, A. J. Lemb o Jr, and L. J. Hindman. 2008. Monitoring habitat use by male Mute Swans in the Chesapeake Bay. Proceedings of the Southeastern Association of Fish and Wildlife Agencies 62:88 - 93. Sutherland, W. J. 1996. From individual behaviour to population ecology. Vo lume 11.Oxford University Press on Demand. U. S. Fish and Wildlife Service. 2015. National wetlands inventory website. U.S. Department of the Interior, Fish and Wildlife Service, Washington, D.C. < http://www.fw s.gov/wetlands/ >. Usher, M. 1989. Ecological effects of controlling invasive terrestrial vertebrates. in Drake, J. A.; Mooney, H. A.; di Castri, F.; Groves, R.H.; Kruger, F.J.; Rejmanek, M. and Williamson, M. (eds.). Biological Invasions. A global perspec tive, p. 463 - 484. John Wiley & Sons. NY, USA. White, T. 2007. Resolving the limitation regulation debate. Ecological R esearch 22:354 - 357. Woiwod, I. P., and I. Hanski. 1992. Patterns of density dependence in moths and aphids. Journal of Animal Ecology 61 :619 - 629. 78 CHAPTER 4: LIFE - STAG E SPECIFIC SURVIVAL AND MOVEMENT S OF MUTE SWANS IN MICHIGAN , USA INTRODUCTION Mute swans ( Cygnus olor ) are large waterbirds native to portions of central Europe (e.g., Poland, Sweden, Denmark), the Baltic States, and regions in western Russia and central Asia (Allin et al. 1987, Wieloch 1991) , but recently introduced into many areas by humans a s early as 1186 (Reeber 2015) . Mute swans were translocated to areas beyond their native range as an ornamental and in some instances as a source of readily available protein (Perrins and Ogilvie 1981, Baldassarre 2014) . The long history of mute swans and humans throughout Europe, Asia, and North America makes it difficult to definitively outline their na tive range; however, mute swans present in North America are the direct result of human - assisted movement s beginning in the late 1800s (Alison and Burton 2008, Elphick 2009) . The close association of mute swans with humans, their semi - domestication (Austin 1961, Ogilvie 1967) , and selective breeding for leucistic - morph individuals (Bacon 1980) has potentially altered aspects of their biology from their naturally occurring state . Mute swans are non - migrato ry across their range in western Europe and North America; however, some populations in central Europe (i.e., Poland) exhibit migratory behaviors between breeding and wintering grounds (Wieloch and Remisiewicz 2001) . Individuals in sedentary populations depend on anthropogenic food resources year - round (Wlodarczyk et al. 2013) and infrequently move more than 50 km from location of marking (Collins 2002) , with movements typically following water courses or coastlines (Ogilvie 1967) . Non - migratory populations of mute swans in temperate climates will move to nearby areas of shallow ice - free water during winter (Mathiasson 1993) , including urban areas (Campbell 1960) . Migratory and sedentary 79 populations in their native ran ge have historically been monitored via swan counts, coded leg bands, and neck collars (Ogilvie 1967, Wieloch 1991, Kirby et al. 1994, Wlodarczyk et al. 2013) . Over 80 % of individuals in some regions have been marked with combinations of metal or plastic - coded leg bands and plastic neck collars (Coleman et al. 2001, Watola et al. 2003) , resulting in very large sample size s and studies that last a decade or more . These long - term studies (Coleman et al. 2001, Collins 2002) contributed to our understanding of mute swan across their range; however, derived demographic parameters and observed movement patterns may not be universally applicable across native and introduced range s due to varying climactic, genetic, and anthropogenic histories. R esearchers acknowledge that relying on resightings to infer mortality or moveme nt s has potential to introduce movement - biased variation because individuals that emigrate from study regions may be lost to follow - up (Colem an et al. 2001, Collins 2002, Watola et al. 2003) . Additionally, few survival studies use quantitative analyses that estimate detection probability of individually - marked swan s (Watola et al. 2003) . Parameterizing demographic models with data not representative of the population under study will lead to unrealistic abundance estimates and, in the case of actively - managed invasive populations, potentially inappro priate management prescriptions that may not allow for achievement of population objectives. We studied survival and movement patterns for mute swans in the Lower Peninsula of Michigan. Our intention was to describe intrastate variation in survival and m ovement s while deriving region - specific demographic parameters that could be incorporated into a population model. We also aimed to understand seasonal movement patterns of sub - adult and adult swans and hoped to identify summer molting areas and wintering concentrations that could be targeted for management. 80 STUDY AREA We investigated survival and movement s for mute swans originating from six study areas across 5 of 8 physiographic regions found in the Lower Peninsula of Michigan, USA (Schaetzl et al. 2013). Terrestrial land cover and wetland density varied latitudinally across four inland and two coastal study sites (Homer et al. 2015) . Inland waterbodies had areas of natural and developed shoreline adjacent to moderate ly or heavily developed upland areas . Most open water areas on these sites wer e permanently - flooded (L1UBH based on the National Wetland Inventory Classification System ; U. S. Fish and Wildlife Service 2015 ). Two coastal sites were characterized by semi - permanently flooded persistent emergent vegetation (PEM1F) and open water (PABG; U. S. Fish and Wildlife Service 2015 ) adjacent to areas of agriculture, human development, and forest cover (Homer et al. 2015) . A detaile d description of study sites and their selection process can be found in Chapter 3. METHODS A pilot marking project was undertaken by staff of the Michigan Department of Natural Resources ( DNR ) and U.S. Department of Agricultures (USDA) Animal Plant Hea lth Inspection Service (APHIS) Wildlife Services (WS) in 2014 to gather preliminary movement data (Appendix A.) . Plastic neck collars were affixed to non - breeding mute swans within summer molting flocks ( n = 26) , although a few actively nesting swans were also marked ( n = 5). Observed inter - and intrastate movement s during the pilot study ( 98 citizen reports of 23 individuals 2014 2018; Appendix A.) confirmed the need for tracking individual swans with remotely accessible GPS technology rather than tradit ional very high frequency (VHF) telemetry or coded plastic neck collars. 81 GPS - marking We targeted actively breeding (i.e., incubating or brooding) female mute swans ( n = 32) within study sites for capture 2016 2018 using boats equipped with longtail (S.W .O.M.P. 26.5, Backwater, Inc., Freeport, MN, USA) or surface - drive mud motors (GTR35, Gator - Tail Outboards, Loreauville, LA, USA). We captured adult females using an extendable aluminum catch pole ( n = 25; Coleman and Minton 1979 ) or a shoulder - fired netgun ( n = 7; CODA Enterprises Inc. Mesa, AZ, USA ). We fit breeding female mute swans with alphanumeric plastic neck collars (56 mm internal diameter X 90.5 mm tal l; Spinner Plastic s, Inc., Springfield, Illinois, USA) that included a GPS - GSM (Global System for Mobile Communication) transmitter (CTT - 1070 BT3; Cellular Tracking Technologies, Inc., Rio Grande, New Jersey, USA) and weighed 117 - 121.5 g total (< 1.4 % of body weight). We also placed plastic neck collars on select male mute swans paired with GPS - collared females ( n = 7). We captured pre - fledge mute swans ( n = 40) from broods hatched within study sites in 2016 and 2017 using an extendable aluminum catch pole ( n = 39; Coleman and Minton 1979 ) or shoulder - fired netgun ( n = 1). Pre - fledge mute swans were captured from separate broods except on two occasions in 2016 where siblings were accidentally GPS - marked. We placed neck col lar - mounted GPS - GSM transmitters (117 - 121.5 g total) on pre - fledge swans (< 1.4 % of body weight). GPS - GSM transmitters were set to record a GPS fix every 15 minutes 24 hours per day during spring and summer but were reduced to a GPS fix every 15 minutes during daylight only for winter due to reduced solar charging capacity. The 2 - way communication of the GPS transmitter u nits allowed us to occasionally set a more restrictive temporary duty cycle (1 fix/hour) for select units to let solar - charged batteries recover. Some units, primarily those deployed on juvenile - marked swans, were set to use an accelerometer - triggered duty cycle (i.e., FlightMode) that began collecting a 82 GPS fix every 10 seconds when flight was detected (Appendix B) . For the purpose of movement and survival analys es , we considered swans juveniles from approximate fledge date (1 September) to their 1st spring (31 March). Juvenile s wan s transitioned to being considered juvenile - marked non - breeding swans (i.e., non - breed ing swans originally marked as juveniles) beginning with their 1 st spring (1 April) when they were approximately 11 months old . All swans other than juveniles were characterized as breeders if they initiated a nest with eggs in the current year or non - bree ders otherwise. No juvenile - marked swans nested within the duration of this study (Chapter 2). We weighed (40 kg digital, PESOLA Präzisionswaagen AG , Schindellegi, Switzerland) , cloacally sexed , and banded (28.5 mm diameter (9C); National Band and Tag Co., Newport, KY, USA) all captured swans. We also measured tarsus, wing, and skull length using a dial Vernier caliper (300 mm, Flexbar Machine Corporation, Islandia, NY USA) or stopped wing ruler (Brown et al. 2003) . Eleven mute swans were recaptured 2016 2018 to replace malfunctioning GPS - GSM transmit ters. Recovery of carcasses was attempted for all mute swans where mortality was indicated based on GPS coordinates. Recovered carcasses were transferred to the Michigan DNR for necropsy. All capture and handling of live mute swans was led by staff of the U.S. Department of Agriculture Wildlife Services section of the Animal Plant Health Inspection Service (USDA APHIS WS). A s such, the Michigan State University (MSU) Institutional Animal Care and Use Committee (IACUC) granted an animal - use exemption for MSU personnel during this project. This work was also partially supported by salary support for Scott R. Winterstein from the USDA National Institute of Food and Agriculture (Project No. MICL02588). 83 D isplacement from capture location We documented annual movement patterns for adult - and juvenile - marked mute swans with interest in the timing and distance of displacement from cap ture location (i.e., nesting territory for adult - marked females and natal area for juvenile - marked swans) . We compiled GPS locations for all swans from first initial capture in this study (19 April 2016) until the end of the study (31 August 2018). We remo ved GPS locations with horizontal dilution of precision (HDOP) > 4 m, fixes that were no t three - dimensional, and where elapsed time until GPS fix (a measure of signal strength) was > 118 seconds based off manufacturer recommendations (A. McGann, Cellular T racking Technologies, personal communication). (Hijmans 2016) (Wickham et al. 2018) packages in Program R (R Development Core Team 2018) to calculate distance between each GPS location an d capture location for individual mute swans using a Vincenty ellipsoid representation of the earth (Vincenty 1975) . We calculated a mean weekly displacement distance from capture location for each individual . We also wanted to understand how far swans move d from their nesting territory or natal area during winter; therefore, we also calculated maximum distance from capture location during winter (i.e., December March) for adult - and juve nile - marked swans with working transmitters that survived the entire winter period . We structured movement analyses differently for juvenile - marked swans and those marked as adult females. Movement patterns likely change between years for juvenile - marked swans as they mature, find mates, and establish territories ; therefore, we pooled swans marked at estimated fledging in 2016 and 2017 and analyzed weekly displacement from capture location (i.e., natal territory). Adult breeding females typically have a se asonal pattern to displacement since they have already paired and established nesting territories. W e pooled adult females 84 captured in different years (i.e., 2016, 2017, or 2018) into one dataset that represented their mean weekly displacement from capture location (i.e., nesting territory) for the annual period. Adult s that were GPS - marked for multiple years (i.e., had displacement distances for a given week in multiple years) were included in this analysis, but had their displacement averaged by week acro ss multiple years. Summarizing movement s using weekly intervals only for individuals that had GPS data for each week lessened potential bias that could result from GPS transmitters that malfunctioned . Life - stage specific survival analyses We used the known fates approach in Program MARK (ver. 8.1; White and Burnham 1999 ) to model life - stage specific survival estimates derived from GPS - collared swans . Survival was estimated separately for each life - stage o f interest (e.g., juvenile, breeding, and non - breeding swans) using 7 - day intervals which are typically adequate for providing unbiased estimation of survival parameters (Murray 2006) . Murray (2006) suggest ed that wildlife telemetry studies modeling survival should have a baseline of at least 30 mortalities with 10 additional observed mortalities per variable of interest , while also noting that species with low mortality rates of ten need larger sample sizes. Our moderate sample size for each life stage and relatively low mortality rates limited our ability to fit c omplex models with many variables of interest ; therefore , we were conservative in our a priori model development and o nly considered including variables that directly related to study - oriented questions. We were still able to generate estimates of life - stage specific survival despite relatively small sample sizes for each life - stage. 85 Juvenile survival modeling We summarized fates for 40 juvenile mute swans using a 30 - week period from estimated fledging (1 September) through 31 March of the following year (i.e., their 1 st spring) . Juveniles marked in 2016 ( n = 22) and 2017 ( n = 18) were pooled for this analysis. We included effects for month and color morph (i.e., leucistic or gray) in a priori models since we were interested in the influence of color morph frequency (see Chapter 2) and temporal variability on juvenile survival probability . We ranked competing mod small sample sizes (AIC c ; Anderson and Burnham 2002 ). We averaged the derived survival estimate across all a priori candidate models according to their Ak aike weights. S urvival modeling for breeding and non - breeding swans We were primarily interested in generating annual survival estimates for breeding and non - breeding swans through this analysis; therefore, we did no t test effects of competing a priori models. We included non - breeding and breeding swans 1 year of age into an annual (i.e., 52 week) survival analysis and examined derived annual survival estimates for breeding and non - breeding swans. We were unable to generate separate estimates for immature non - breeding (i.e., 1 year s old, but < 2 year s old) and adult non - breeding (i.e., 2 year s old) swans as was accomplished by Watola et al. (2003) ; however, the similarity of estima tes obtained by Watola et al. (2003) for these two stages (immature = 0.73 ± 0.02, nonbreeder = 0.71 ± 0.02) provides biological justification for combining these life stages of non - breeding swans in our analyses . RESULTS Seventy - two mute swans (53 females, 18 males, 1 unknown) were marked with neck collar - mounted GPS - GSM transmitters 2016 2018 on the Juno ( n = 14 ), St. Clair ( n = 18), Pontiac ( n = 16), Wabasis ( n = 15), Tobico ( n = 3), and Cl am ( n = 6) study sites. Seven adult 86 male mute swans were marked with plastic neck collars on the Juno ( n = 3), Pontiac ( n = 1), Wabasis ( n = 2), and Tobico ( n = 1) study sites. We obtained 1,853,771 locations through GPS - GSM transmitters deployed on juveni le and adult mute swans resulting in 1,553,253 GPS fixe s with acceptable error and fix quality ( i.e., HDOP - dimensional location, and time to GPS Average HDOP for all acceptable locations was 1.97 m (range: 0.7 4 m) . Captured males were heavier a nd morphologically larger than females when comparing between sexes for both age classes although females of both age classes tended to have larger wings than males (Table 4.1). However, some adult males were captured during th eir annual flight feather molt which biased wing length low and increased standard deviation of wing length (Table 4.1). Cloacal sexing can be difficult for subadult mute swans (Brown et al. 2003) ; however, s imilarity of sex - specific tarsus length between juvenile - and adult - marked swans indicates that captured fledglings likely had the correct sex assigned since the tarsus length is nearly maximized at fledging and remain s constant in adults (Mathiasson 1981) . Table 4.1. Physiographic measurements for juvenile - and adult - marked mute swans 2016 2018 in the Lower Peninsula of Michigan, USA. * Th is table does not include 1 unknown sex juvenile - marked swan. D isplacement from capture location Largest mean weekly d isplacement from capture occurred during winter for adult - marked female mute swans (Figure 4.1) . Largest mean weekly displacement from capt ure location for juvenile - marked mute swans was observed during their first summer; however, the two other times of peak weekly displacement occurred during their 1 st and 2 nd winters (Figure 4.2). Mean 87 weekly displacement was typically higher for juvenile - marked females when compared to juvenile - marked males; however, juvenile - marked males had larger displacement values for the last 15 weeks of the study (i.e., their second summer after fledging; Figure 4.2). Mean maximum winter displacement for juvenile - ma rked mute swans ( 49.4 km ) was 37.1 km in 2016 - 17 and 72.5 km in 2017 - 18 . Juvenile - marked females tended to move f a rther (72.0 km) from natal area s during winter than males (28.7 km) , but difference was not statistically significant ( t 21 = 1.58, P = 0.21 ). Mean maximum winter displacement for adult - marked females ( 11.3 km ) was 11.9 km in 2016 - 17 and 10.6 km in 2017 - 18. Adult - marked females typically left nesting territory by week 51 (i.e., mid - December) and returned to nesting territories in week 8 (i.e., late February ; Figure 4.1 ). Juvenile - marked swans moved further from place of capture (i.e., natal/nesting territory) than adult - marked female mute swans (Figure 4.3). It is unknown if movement s of juvenile - or adult - marked swans occurred in flocks or by themselves . Regional maps of mute swan movement s can be found in Appendix C. 88 Figure 4.1. Mean weekly displacement from capture location (i.e., nesting territory) for adult - marked female mute swans captured within 6 study sites in the Lower Peninsula of Michigan, USA , 2016 2018 pooled across years with sample size for weekly displacement averages indicated on the secondary y - axis. 89 Figure 4.2. Mean weekly displacement since capture (range: 1 109 weeks) for juvenile - marked mute swans captured at 6 study sites in the Lower Peninsula of Michigan, USA , 2016 2018 with sample size for weekly displacement averages indicated on the secondary y - axis . 90 Figure 4.3. Total GPS - derived movement s for adult - and juvenile - marked mute swans from 6 capture locations ( purple rectangles ) April 2016 August 2018 in the Lower Peninsula of Michigan, USA Life - stage specific survival analyses Juvenile survival m odeling Juvenile survival was estimated from 40 individuals captured in 2016 ( n = 22) and 2017 ( n = 18). An average of 26 GPS - marked juveniles were at risk for each weekly survival interval. Thirteen individuals had encounter histories partially censored due to transmitter failure . Observed transmitter failures were independent of mortality (10 of 13 censored juvenile s were reported alive after transmitter failure) . Fourteen mortality events were documented for swans in the juvenile life stage. On ly 3 causes of mortality were able to be diagnosed by necropsy (T. Cooley, Michigan D epartment of Natural Resources, personal communication) for juvenile 91 swans due to location of mortality (i.e., near open water pockets on ice - covered waterbodies) and the ability of carcasses to sink or wash away with melting of ice. The three mortality diagnoses for GPS - collared juveniles were p ulmonary c ongestion/ p ulmonary e dema, v erminous h emorrhagic u lcerative e nteritis caused by infestation of Sphaeridiotrema globulus , and canid predation facilitated by malnutrition. We document ed mortality by v erminous h emorrhagic u lcerative e nteritis ( n = 10) , acute lead poisoning (ingestion of lead fishing weights ; n = 1 ), trauma/predation ( n = 1) , and pulmonary congestion/pulmonary edema ( n = 1) for other unmarked juvenile mute swans in areas used by juvenile - marked swans ; however, it is important to note that proportional cause of mortality in recovered carcasses of unmarked swans many not equal proportional cau se of mortality realized by juvenile swans (i.e., unmarked swans killed through predation are unlikely to be found and necropsied). The most parsimonious model for juvenile survival included additive effects for month and color morph (Table 4.2). This mod el was ranked within a c of 2 from the next competing model which only included the month effect . The model - averaged 30 - week maximum likelihood survival estimate was = 0.526, 95 % CI = 0.342 0 . 703. Leucistic - morph individuals typically had lower estimated survival ( = 0.400, 95 % CI = 0.202 0 . 637) than gray morph juveniles ( = 0.685, 95 % CI = 0.397 0 . 878) , but the confidence interval of the coefficient contained 0 ( ^ l eucistic = - 0. 908 , 95 % CI = - 2.086 0 . 269 ) . Weeks with lowest estimated survival occurred in December and January. 92 Table 4.2. Model selection results for a priori candidate model set to explain temporal and morphometric variation in survival for juvenile - marked mute swans 1 September 31 March 2016 and 2017 in the Lower Peninsula of Michigan, USA. Model k a AIC c a c a w i a Month + Color Morph 8 125.447 0 0.561 Month 7 125.940 0.493 0.439 Color Morph 2 150.127 24.680 0 Constant (null) 1 150.753 25.306 0 a k = number of parameters in model, AIC c c = difference between AIC c of best fitting and current model, w i = Akaike's weight. Non - juvenile survival modeling Juvenile - marked swans that survived to their first spring (i.e., 1 April) with working GPS - transmitters ( n = 12) and a dult - marked ( n = 29) swans were pooled across years for estimation of breeding and non - breeding survival rates. Forty - one unique individuals were included in the pooled analysis. Individuals that survived between years ( n = 16) were entered as new individuals in the 2 nd Fourteen individuals had encounter histories partially censored due to transmitter failure ( n = 12) or incidental culling by USDA APHIS Wildlife Services ( n =1) and the Michigan DNR ( n = 1). Two of the swans with failed transmitters had un its replaced within the year they failed and were entered into the analysis as new individuals from point of recapture . As noted above with the juvenile survival analysis , transmitter failure was independent of mortality ( 11 of 12 individuals with malfunct ioning transmitters were resighted alive after failur e of GPS units). Nine mortality events were documented in 2016 - 17 ( n = 1) and 2017 - 18 ( n = 8). Cause of mortality was determined by in - field evidence or laboratory necropsy ( T. Cooley, M ichigan Departmen t of Natural Resources, personal communication ) f or 5 of 9 mortality events (Predation = 3, Hepatitis = 1, West Nile Virus =1, Unknown = 4). V erminous h emorrhagic u lcerative e nteritis , lead poisoning, avian predation (likely by bald eagle [ Haliaeetus leucocephalus ] or great horned owl [ Bubo virginianus ] ), mammalian predation (likely by red fox [ Vulpes vulpes ] or coyote [ Canis latrans ]) , collisions with trees and 93 powerlines, starvation, and drowning were implicated in mortalities of non GPS - marked mute swans encountered while performing field work in areas with GPS - marked individuals . An average of 42 breeders or non - breeders were at risk for each weekly survival interval. Breeding swans typically had higher annual surv ival rates ( = 0.850, 95 % CI = 0.686 0 . 936) than non - breeding swans ( = 0.698, 95 % CI = 0.419 0 . 881) , but the confidence interval of the coefficient overlapped 0 ( ^ b reeder = 0.793 , 95 % CI = - 0.525 2.111 ) . Five of 9 mortalities occurred in January or February, two occurred in May, and one was recorded in each of June and July. DISCUSSION We confirmed that mute swans found in Michigan are non - migratory but can move as necessary to shallow open water areas during periods of ice cover. Most adult - marked female swans remain ed relatively close to nesting territories during winter when compared to juveniles (Figure 4.1 and 4.2 ). Some adult - marked females in southeast and northern Michigan exhibited movements to open water areas associated with the Great Lakes ( i.e., Detroit River or Grand Traverse Bay; Figure 4.3), but not all adult - marked female s in southeast or northern Michigan moved large distances. Adult - marked f emales from all sites that remained near nesting territories typically relocated to shallow streams or rivers that were connected to the waterbody of their territory or were near ( < 1 1.3 km ) their nesting waterbody. Generally, these shallow areas of flowing water contained a wintering flock of mute swans, Canada geese ( Branta canadensis ) , and several species of dabbling and diving ducks in addition to GPS - collared mute swans , but not a ll areas used by GPS - marked swans during winter were visited by researchers. Winter movement s or displacement from nesting territory has no t been rigorously investigated for many non - migratory populations of mute swans. Although our data analysis was 94 not directly comparable to t hat of Collins and Whelan (1994) , our mean maximum displacement from nesting territory for adult female mute swans is reas onable given the movement rates observed for banded mute swans in their Irish population ( 32 % of all marked swan s regardless of age moved ) . Additionally, peak movement s for their population occurred during October March (Collins and Whelan 1994) , which is comparable to peak observed movement for adult - marked swans in Michigan. Juvenile - marked swans moved f a rther from their natal area than adult - marked femal es moved from their nesting territory (Figure 4.1 and 4.2). Mean displacement from natal area for juvenile - marked swans observed in this study peaked in their 1 st summer after fledging and during winter (Figure 4.2) . Displacement of juvenile - marked swans from natal areas during winter was likely related to the same factor (i.e., ice coverage of natal territory) causing adult - marked females to move from nesting territories; however, juvenile - marked swans moved f a rther from their natal areas during bot h winter s than adult - marked females moved from their nesting territories during winter . Th e peak in juvenile - marked swan movements from natal territories during their 1 st summer likely result ed from non - breeding swans being excluded from typical nesting ar eas while searching for locations to complete their annual flight feather molt with less disturbance from breeding pairs and humans (Holm 2002) . M olt migration of non - breeding individuals has been observed for some mute swan populations (Mathiasson 1993) . W e did no t detect widespread migration to a few select molting sites although juvenile - marked non - breeding mute swans tended to com plete their molt on medium (170 ha) to large ( 800 ha ) inland lakes , coastal lakes of the Great Lakes (e.g., White Lake, Muskegon Lake, Mona Lake, Pentwater Lake), or secluded shallow water areas of the Great Lakes themselves (e.g., Saint Martin Bay 95 of La ke Huron, nearshore areas around the Beaver Islands, nearshore areas of Lake Huron between Grindstone City, MI, and Port Hope, MI). We were unable to observe true natal dispersal through this research since age at first nesting is likely near 5 or 6 years of age for established populations (Collins 1991, Coleman et al. 2001, Wlodarczyk et al. 2013) ; however, we were able to document movement of juvenile - marked swans for the first 28 months of life. Female juvenile - marked swans moved f a rther from their natal area than subadult males in nearly all weeks of the analysis except for the final 15 weeks of this study. Mute swans exhibit male - biased natal dispersal (Coleman et al. 2001, Wlodarczyk et al. 2013) that is typical for waterfowl (Anderson et al. 1992) . However, we found that juvenile - marked females tended to move f a rther from their location of capture in early life than did juvenile - marked male s (Figure 4.2 ; also see Collins 2002 ). Collins (2002) found that non - breedin g females tended to have a 46.9 km mean maximum displacement from capture location , whereas non - breeding males had a 39.8 km mean maximum displacement from capture. We found that juvenile - marked females tended to move f a rther from their point of capture du ring winter than juvenile - marked males ( 72.0 km and 28.7 km , respectively) , although differences between sexes were not statistically significant. It is possible that mean maximum distances moved during winter observed in this study are not the maximum win ter displacement that will occur for juvenile - marked non - breeding swans before they breed since they will likely not establish nesting territories for at least another year, but most pre - nesting movement for swans does occur in their second year of life (Collins and Whelan 1994) . Most mortality of juvenile, non - breeding , and breeding adult swans occurred during the winter months when they were furthest from their natal or nesting territory , although two adult - marked females were killed on their nests . The principle documented cause of mortality for all 96 swans found dead during this res earch was v erminous h emorrhagic u lcerative e nteritis caused by infection of Sphaeridiotrema globulus , although mortality from lead poisoning, predation, and collision with fixed obstacles (i.e., power lines or trees) , hepatitis, and drowning w ere also obse rved in marked and unmarked swans . V erminous h emorrhagic u lcerative e nteritis was the likely cause for most mortalities where carcasses could not be retrieved for diagnosis (based on location s of mortality) . Most mortality for juvenile - and adult - marked sw ans occurred during winter. This contrasts with populations in their native range where most documented mortality occurs during movement peaks in spring and autumn (Perrins and Reynolds 1967) by collisions with power lines or other fixed objects (Ogilvie 1967, Mathiasson 1993, Coleman et al. 2001, Collins 2002) . This could partly be explained by prevalence of Sphaeridiotrema globulus and lack of anthropogenic food resources specifically offered to mute swans in Michigan compared to areas of their native range where such supplemental feeding can b e common (Scott and Birkhead 1983, Sears 1989) . Estimated survival for juvenile swans was slightly lower in Michiga n ( = 0.526) than in other populations ( = 0.68, Watola et al. 2003; = 0.66, Perrins and Reynolds 1967 ) , with a caveat that expos ure periods for which survival was estimated are not equivalent (30 weeks in this study ; 12 weeks, Watola et al. 2003; ~35 weeks, Perrins and Reynolds 1967 ). Survival for juvenile swans in Michigan seemed to be biologically related to color morph ( leucistic = 0.400, gray = 0.685) , but the relationship was no t statistically significant. Lower observed survival for leucistic morph individuals during 30 - weeks post - fledging could result from lower parental care and increased exposure to environmental hazards (e.g., predation, intraspe cific agonistic interactions, or feeding locations where infection by Sphaeridiotrema globulus can occur) compared with gray - morph juveniles. Conover et al. (2000) found that parents would dissociate 97 with leucistic juveniles during their first winter or force them from the natal territory , but allowed gray - morph cygnets to remain with them during this period . Conover et al . (2000) also found that leucistic juvenile males had lower survival than did gray - morph males for their 1 st two years of life. Overall l ower juvenile survival rates observed in this study could be partly explained by the higher percentage of leucistic morph individuals in Michigan (Chapter 2) coupled with seemingly lower survival for leucistic individuals. Estimated survival rates for breeding ( = 0.850) and non - breeding ( = 0.698) individuals were not statistically different, but were each within re ported ranges for those life - stages (Watola et al. 2003) . Our decision to pool juvenile - marked non - breeding swans (i.e., first year) and adult - marked non - breeding mute swans was likely justified given the similar (and overlapping) estimates for these two life - stages (Watola et al. 2003) . It should be noted that we estimated breeding adult survival using only adult females rather than a sample of males and females. Annual mortality for adults is typically low and normally not estimated separately for females and males (Watola et al. 2003) due to linked behavior of mated pairs throughout the year . Collins (2002) found no difference in cause - specific mortality between males and females . Additionally, major causes of mortality noted in this re search (e.g., v erminous h emorrhagic u lcerative e nteritis , lead poisoning) are not sex - specific mortality factors. Therefore, we believe that our estimated survival rate for breeding adults is representative of both breeding males and breeding females. MANA GEMENT IMPLICATIONS Juvenile - and adult - marked mute swans are capable of long - distance movements; however, we observed different movement patterns between these two life - stages . The largest displacement from nesting territory for adult - marked mute swans occurred during winter; 98 however, subadult mute swans moved furthest from their natal territory during their 1 st summer even though they also moved large distances during the ir 1 st two winter periods. Juvenile - marked swans moved f a rther throug hout the year than did adult - marked females. Adult - marked female mute swans typically stayed close to their nesting territory during winter but did make local movements to shallow ice - free areas which were typically within 11.3 km of their territory. Some, but not all, adult - marked female mute swans nesting in southeastern and northern Michigan joined wintering flocks on waterbodies associated with the Great Lakes (i.e., Grand Traverse Bay or the Detroit River). This suggests that wintering flocks found in those areas primarily consist of non - breeding individuals with fewer breeding pairs that nest nearby . Observed survival rates for juvenile, non - breeding , and breeding mute swans suggest that culling efforts targeted at breeding mute swans (i.e., highest su rvival rate) will be most effective at reducing the overall population although removal of non - breeding individuals should also be considered due to their relatively high survival and ability to replace breeding mute swans when territories are vacated. R ecommendations for location and timing of culling vary between subadult and adult mute swans due to differences in observed movement patterns. Efforts to remove adult breeding pairs must either target pairs during the incubation or brood - rearing period or focus on small wintering flocks near their nesting territories. Territorial behavior of mute swans typically results in many small lakes, ponds, or wetlands that have few nesting pairs. Therefore, targeting breeding pairs during the nesting season requires more effort than during winter when pairs are concentrated in wintering flocks . Unfortunately, inland wintering locations typically have poor access (i.e., are privately - owned or cannot be accessed by boat due to ice - coverage on connected lakes) or are lo cated in urban environments ; however, these locations must be targeted if removal 99 of breeding adults is desired since the majority of breeding females (thereby breeding pairs) in Michigan do not go to waters of the Great Lakes during winter (Figure 4.3) . Removal of non - breeding mute swans can occur at summer molting sites or at wintering areas. Our observed movement s for non - breeding mute swans suggests that summer culling programs will be most effective after swans have settled into molting areas, but before early fall ( 15 July 31 August ; Figure 4.2). Breeding adult mute swans could also be easily removed during this same period (i.e., immediately before young fledge) if molting flocks are in proximity to breeding pairs . Removal of non - bre eding swans at wintering locations can also be highly effective and removal efforts at these locations may also incidentally cull breeding adults as well. Winter r emoval efforts for non - breeding and breeding swans will be most effective after inland ice co verage prompts movement to wintering locations (late December), but before adult females attempt to return to nesting territories (late February). 100 APPENDICES 101 APPENDIX A: MOVEMENT S OBSERVED THROUGH PI LOT NECK COLLARING OF MUTE SWANS CONDUC TED BY MICHIGAN DNR AND US DA APHIS WILDLIFE SERVICES 2014 2018 Figure A.1. Overall movements coded by region of capture for plastic neck collared mute swans during a pilot research effort of the Michigan Department of Natural Resources and the Wildlife Services Department of the U.S. Department of Agriculture Animal Plant Health Inspection Service in the Lower Peninsula of Michigan, USA, 2014 2018. 102 Figure A. 2 . Overall movement for plastic neck collared mute swans captured in southcentral Michigan during a pilot research effort of the Michigan Department of Natural Resources and the Wildlife Services Department of the U.S. Department of Agriculture Animal Plant Health Inspection Service in the Lower Peninsula of Michigan, USA, 2014 2018. 103 Figure A. 3 . Overall movement in southeast Michigan for plastic neck collared mute swans during a pilot research effort of the Michigan Department of Natural Resources and the Wildlife Services Department of the U.S. Department of Agriculture Animal Plant Health Inspe ction Service in the Lower Peninsula of Michigan, USA, 2014 2018. 104 Figure A. 4 . Overall movement in central Michigan for plastic neck collared mute swans during a pilot research effort of the Michigan Department of Natural Resources and the Wildlife Se rvices Department of the U.S. Department of Agriculture Animal Plant Health Inspection Service in the Lower Peninsula of Michigan, USA, 2014 2018. 105 APPENDIX B: FLIGHT C HARACTERISTICS ON MU TE SWANS DETERMINED THROUGH GPS - GSM TRAN SMITTERS Tabl e B.1. Summarized flight speeds and altitudes estimated through flexible duty cycles (i.e., FlightMode ) available on neck collar - mounted GPS - GSM transmitters ( n = 13,897) on mute swans marked within the Lower Peninsula of Michigan, USA, in 2016 2018. 106 APPENDIX C: DETAILED MOVEMENT OBSERVED T HROUGH GPS - GSM TRANSMITTERS ON MUTE SWAN S Figure C.1. Detailed movement of mute swans in southwestern Michigan, USA, as determined by GPS - GSM transmitters. 107 Figure C.2. Detailed movement of mute swans in southeast Michigan, USA, as determined by GPS - GSM transmitters. 108 Figure C.3. Detailed movement of mute swans in east central Michigan, USA, as determined by GPS - GSM transmitters. 109 Figure C.4. Detailed movement of mute swans in west central Michigan, USA, as determined by GPS - GSM transmitters. 110 Figure C.5. Detailed movement of mute swans in the northwes t Lower Peninsula of Michigan, USA, as determined by GPS - GSM transmitters. 111 Figure C.6. Detailed movement of mute swans in the northern Lower Peninsula of Michigan, USA, as determined by GPS - GSM transmitters. 112 Figure C.7. Detailed movement of mute swans in the northern Lower Peninsula and southeastern Upper Peninsula of Michigan, USA, as determined by GPS - GSM transmitters. 113 LITERATURE CITED 114 LITERATURE CITED Alison, R., and K. Burton. 2008. New evidence of early presence of Cygnus olor. Picoides 21:36 - 45. Allin, C., G. Chasko, and T. P. Husband. 1987. Mute Swans in the Atlantic flyway: a review of the history, population growth and m anagement needs. Northeast Section of The Wildlife Society 44:32 - 42. Anderson, D. R., and K. P. Burnham. 2002. Avoiding pitfalls when using information - theoretic methods. Journal of Wildlife Management 66:912 - 918. Anderson, M. G., J. M. Rhymer, and F. C. Rohwer. 1992. Philopatry, dispersal, and the genetic structure of waterfowl populations. in Batt, D. J., Ecology and management of breeding waterfowl. University of Minnesota Press, Minneapolis, MN , USA. p. 365 - 395. Austin, O. L. 196 2 . Birds of the world ; a survey of the twenty - seven orders and one hundred and fifty - five families. Hamlyn Publishing. London, England. Bacon, P. J. 1980. A possible advantage for the Polish morph of the Mute Swan Cygnus olor. Wildfowl 31:51 - 52. Baldassarre, G. 2014. Ducks, geese, and swans of North America. Volume 1.J ohns H opkins U niversity Press. Baltimore, MD, USA. Brown, A. W., L. M. Brown, and P. I. Stevick. 2003. Sexing Mute Swans Cygnus olor by discriminant analysis. Ringing & Migration 21:174 - 180. Campbell, B. 1960. The Mute Swan census in England and Wales 1955 56. Bird Study 7:208 - 223. Coleman, A., and C. Minton. 1979. Pairing and breeding of Mute Swans in relation to natal area. Wildfowl 30:27 - 30. Coleman, A. E., J. T. Coleman, P. A. Coleman, and C. D. T. Minton . 2001. A 39 year study of a Mute Swan Cygnus olor population in the English Midlands. Ardea 89:123 - 133. Collins, R. 1991. Breeding performance of an Irish Mute Swan Cygnus olor population. Wildfowl:144 - 150. Collins, R. 2002. Sex differences in the movements and mortality of Mute Swans. Waterbirds 25:157 - 161. 115 Collins, R., and J. Whelan. 1994. Movements in an Irish Mute Swan Cygnus olor population. Ringing & Migration 15:40 - 49. Conover, M. R., J. G. Reese, and A. D. Brown. 2000. Costs and benefits o f subadult plumage in Mute Swans: Testing hypotheses for the evolution of delayed plumage maturation. American Naturalist 156:193 - 200. Elphick, C. S. 2009. Evidence that Mute Swans are native to North America is lacking. Picoides 22:20 - 23. Hijmans, R. J. 2016. geosphere: Spherical Trigonometry. R package version 1.5 - 5. < https://cran.r - project.org/package=geosphere >. Holm, T. E. 2002. Habitat use and activity patterns of Mute Swans at a molting and a wintering site in Denmark. Waterbirds 25:183 - 191. Homer, C., J. Dewitz, L. Yang, S. Jin, P. Danielson, G. Xian, J. Coulston, N. Herold, J. Wickham, and K. Megown. 2015. Completion of the 2011 National Land Cover Database for the conterminous United States representing a decade of land cover change information. Photogrammetric Engineering & Remote Sensing 81:345 - 354. Kirby, J., S. Delany, and J. Quinn. 1994. Mute Swans in Great - Britain: a review, current status and long - term trends. Hydrobiol ogia 279:467 - 482. Mathiasson, S. 1981. Weight and growth rates of morphological characters of Cygnus olor Pages 379 - 389 in Proceedings of the Second International Swan Symposium, Sapporo 1980 (GVT Mathews and M. Smart, Eds.). International Waterfowl Resea rch Bureau, Slimbridge , UK. Mathiasson, S. 1993. Mute Swans, Cygnus olor, killed from collision with electrical wires, a study of 2 situations in Sweden. Environmental Pollution 80:239 - 246. Murray, D. L. 2006. On improving telemetry - based survival estima tion. The Journal of Wildlife Management 70:1530 - 1543. Ogilvie, M. 1967. Population changes and mortality of the Mute Swan in Britain. Wildfowl 18:64 - 73. Perrins, C., and M. Ogilvie. 1981. A study of the Abbotsbury mute swans. Wildfowl 32:35 - 47. Perrins, C. M., and C. M. Reynolds. 1967. A preliminary study of the Mute Swan, Cygnus olor. The Wildfowl Trust. 18 th Annual Report. p. 74 - 84. R Development Core Team. 2018. R: A language and environment for statistical computing. R Foundation for Statist ical Computing, Vienna, Austria. < https://www.R - project.org/ >. 116 Reeber, S. 2015. Wildfowl of Europe, Asia and North America. Princeton University Press . Princeton, NJ, USA. Schaetzl, R. J., H. Enander, M. D. Lueh mann, D. P. Lusch, C. Fish, M. Bigsby, M. Steigmeyer, J. Guasco, C. Forgacs, and A. Pollyea. 2013. Mapping the physiography of Michigan with GIS. Physical Geography 34:2 - 39. Scott, D. K., and M. E. Birkhead. 1983. Resources and reproductive performance in Mute Swans Cygnus olor. Journal of Zoology 200:539 - 547. Sears, J. 1989. Feeding activity and body condition of Mute Swans Cygnus olor in rural and urban areas of a lowland river system. Wildfowl 40:88 - 98. U. S. Fish and Wildlife Service. 2015. National wetlands inventory website. U.S. Department of the Interior, Fish and Wildlife Service, Washington, D.C. < http://www.fws.gov/wetlands/ >. Vincenty, T. 1975. Direct and inverse solutions of geodesics on the ellipsoid with application of nested equations. Survey review 23:88 - 93. Watola, G. V., D. A. Stone, G. C. Smith, G. J. Forrester, A. E. Coleman, J. T. Coleman, M. J. Goulding, K. A. Robinson, and T. P. Milsom. 2003. Analyses of two Mute Swan populations a nd the effects of clutch reduction: implications for population management. Journal of Applied Ecology 40:565 - 579. White, G. C., and K. P. Burnham. 1999. Program MARK: survival estimation from populations of marked animals. Bird Study 46:S120 - S139. Wickh am, H., R. Francois, L. Henry, and K. Müller. 2018. dplyr: a grammar of data manipulation. R package version 0.7. 6. Wieloch, M. 1991. Population trends of the Mute Swan Cygnus olor in the Palearctic. Wildfowl:22 - 32. Wieloch M., Remisiewicz M. 2001. Changes in wintering areaof the Mute Swan Cygnus olor. In: waterfowl in central and eastern Europe. OMPO SpecialPubl., Vilnus, pp. 94 103 Wlodarczyk, R., M. Wieloch, S. Cz yz, P. T. Dolata, and P. Minias. 2013. Natal and breeding dispersal in Mute Swans Cygnus olor: influence of sex, mate switching and reproductive success. Acta Ornithologica 48:237 - 244. 117 CHAPTER 5: A DENSITY - DEPENDENT MA TRIX P OPULATION MODEL TO I NFORM MUTE SWAN MANAGEMENT IN MICHIGAN , USA INTRODUCTION Demographic modeling through life tables, matrix population models, and other techniques ha ve been used for nearly a century to model factors influencing abundance change in human and wildlife popu lations. Early efforts estimated mortality rates by following individuals or cohorts through time using life table s (Cox 1972) . Matrix population models (MPMs) have been used extensively by ecologists since the 1970s (Leslie 1945, Caswell 2001) ; however, extensions of MPMs (integral projection models; IPMs) have been developed to better handle continuous ecological data (Easterling et al. 2000, Besbeas et al. 2002) . IPMs offer analytical flexibility for groups like plants that are obs erved across continuous scales (i.e., growth measurements) rather than discrete stag es (Miller et al. 2009) . However, m atrix population models remain useful for species that are organized into and can be measured at naturally occurring life stages such as mute swans ( Cygnus olor ) . Mute swans were introduced to North America by humans in the late 1800s and first (Wood and Gelston 1972) . Abundance in Michigan remained low throughout the 20 th century despite additional introductions (Wood and Gelston 1972) , but quickly gr ew in the late 1990s and early 2000s to a peak estimated abundance of 17,520 in 2013 (M ichigan Department of N atural R esources , unpublishe d data). The Michigan Department of Natural Resources ( M DNR) formalized their mute swan management plan in 2012 which aimed to have fewer than 2,000 mute swans statewide by 2030 (Michigan Department of Natural Resources 2012) . Mute swan control c onducted by the MDNR and the U.S. Department of Agriculture Wildlife Services removed < 650 swans annually prior to 2010, 118 but efforts were expanded after 2011 partially due to increased funding through the Great Lakes Restoration Initiative (Arsnoe and Duffine y 2018) . Expanded control reduced mute swan abundance in Michigan ( MDNR , unpublished data) ; however, annual abundance stabilized despite continuing removal efforts (Arsnoe and Duffiney 2018) . Uncertainty of regional demographic parameters translated to u ncertainty in ho w to most effectively and efficiently reach the long - term goal established by the Michigan DNR. D emographic models exist for introduced populations of mute swans in North America (Allin et al. 1987, Hindman and Harvey IV 2004, Ellis and Elphick 2007) ; however, these models were parameterized exclusively or in part with demographic parameters estimated from native ran ge of mute swans . Long - established populations in native range likely have different demographic rates than newly established populations in introduced range . Watola et al. (2003) also found that native sub - populations in rel ative proximity had varying demograph ics . Variation in demographic rates may arise due to semi - domestication which has altered genetic composition (Munro et al. 1968) , or introduced populations could b e responding to different predator communities and environmental factors not found in the native range. Demogra phic models for introduced populations should incorporate regionally estimated demographic parameters where possible otherwise life stage - specific management strategies derived from poorly parameterized model s may be unrealistic or inappropriate. Our goal was development of a stage - based MPM represent ing mute swan population that could be used to inform future management . We used th e model structure developed by Ellis and Elphick (2007) to predict mute swan abundance using native - and Michigan - estimated demographic parameters with in a deterministic framework . We then refined the Michigan - specific demographic model by incorporating density dependence in 119 breeding productivity based on observed variation in site - level productivity between varying nesting densities (Chapter 3) . We used the density - dependent matri x population model to estimate number and proportion of each life stage that must be removed annually from 2018 20 29 to reach the long - term goal set forth by the Michigan DNR (Michigan Department of Natural Resources 2012) . We also simulated le vel of alternative management strategies (i.e., egg oiling or nest destruction ) needed to obtain the same level of control by 2030. STUDY AREA We model ed population demographics for mute swans in the core of their Michigan distribution which occurs in the Lower Peninsula and eastern Upper Peninsula (Michigan DNR, unpublished data). Demographic data used in parameterization of the Michigan - specific matrix p (Chapter s 2, 3, and 4). Land cover varies latitudinally across Michigan from a combination of developed farmland and hardwoods (i.e., oak [ Quercus spp.], beech [ Fagus grandifo lia ], and maple [ Acer spp.]) in the southern portions to more homogeneously - distributed mixed forests (i.e., pines [ Pinus spp.], spruces [ Picea spp.], firs [ Abies spp.], maples, oaks, and aspen [ Populus spp.] ) in the northern Lower Peninsula and eastern Upper Peninsula (Pugh et al. 2017) . Michigan has several distinct geographic regions with coastlines on four of the five Great Lakes (Schaetzl et al. 2013) . These coastlines contain many inlet waterbodies, river deltas, and coastal wetlands (Sommers 1984) . There are more than 46,000 lakes >2.02 ha in Michigan (Institute for Fisheries Research 2013) , with 18,000 individual lakes in Michi Lower Peninsula that are a in size (Breck 2004) . Detailed descriptions of the study sites used to estimate demographic paramete rs can be found in Chapter s 1 and 3 of this dissertation . 120 METHODS We replicated the life - stage structure used by Ellis and Elphick (2007) package (Stubben and Milligan 2007) in Program R (R Development Core Team 2018) . The incorporates much of the MATLAB ( MathWorks, Natick, MA , USA) code and foundational concepts found in Caswell (2001) within the Program R computing framework (Stubben and Milligan 2007) . The transition matrices used within MPMs in this research contain ed the 6 x 6 structure found in the transition matrix of Ell is and Elphick (2007) . This structure has varying survival and transition probabilities for six life stages: juvenile ( 0 - 1 immature swans ; fledging to 1 st April), immature first - year non - breeding swans (1 - 2 - year - old immature ) , inexperienced non - breeding swans ( are no t currently breeding and never ha ve nested) , experienced non - breeding swans ( are no t currently breeding but have previously nested), inexperienced breeding swans ( are currently breeding, but never ha ve previously nested), and expe rienced breeding swans ( are currently breeding swans that have nested previously ; Ellis and Elphick [ 2007 ] ) . A graphical representation of model s tructure can be found in Figure 1 of Ellis and Elphick (2007) . We wanted to utilize the structure found in Ellis and Elphi ck (2007) because 1) it captures the life cycle of mute swans and 2) we wanted to perform a life table response experiment (LTRE) to understand how differen ce s in parameters from native and introduced ranges contribute to changes in estimated population gr owth rate ( ) . We used the transition matrices described above and an initial population vector that corresponded to estimated mute swan population in 1948 (Gelston and Wood 1982) to conduct projection analyses for 150 years (1949 20 98 ) for each of the models described belo w, and compare d model - breeding waterfowl survey. 121 Comparison of n ative - and Michigan - p arameterization Formulation of the native - parameterized model We parameterized the transition matrix of the native model using the survival and transition values found in Table 1 of Ellis and Elphick (2007) which are replicated in Table 5.1 . This essentially simulated the MPM format used by Ellis and Elphick (2007) but used initial population values for Michigan in a deterministic framework . We used estimated population size of mute swans in Michigan in 1948 (i.e., 47 swans ; Gelston and Wood 1982 ) to create the initial population vector (N) within the mode l , where is the population projection matrix. We rounded up the estimated population size in 19 48 to 48 individuals so that we could divide the total population by two (since matrix models only model the female population component) , and assign integer values to stages within the initial population vector. Ellis and Elphick (2007) began modeling with all individuals placed into the experienced breeder stage; however, we placed 8 individuals into the 0 - 1 immature stage which represents juvenile swans from fledging to their 1 st April. We pla ced the remaining 16 individuals into the experienced breeder category. This distribution of life stages likely approximates the distribution of life stages in the northern Michigan population when nesting cover was abundant, and density of swans was low. It is important to note that variation in input values of the initial conditions vector will influence time o f transience, but initial conditions will not affect ultimate model behavior (Caswell 2001) . 122 Table 5.1. Input probabilities for transition matrices of the Michigan - parameterized stage - based deterministic matrix population model and native - parameterized model with input values adapted from Ellis and Elphick (2007). *We used Birkhead and Perrins (1986) estimate for the probability of first breeding in both the Michigan - and native parameterized models. Formulation of the Michigan - parameterized model We used the same model structure and initial population vector as t he native - parameterized model when creating a Michigan - parameterized matrix population model. All parameter values in the transition matrix for th e Michigan - parameterized model were estimated in Michigan with exception of the probability of first breeding (Table 5.1). Most mute swans typically do no t pair and establish breeding territories until at least 3 or 4 years of age (Reese 1980, McCleery et al. 2002) . We did not observe fi rst nesting attempts for juvenile - marked swans marked through this research; therefore, we relied on the probability of first nesting reported for other populations in the literature (0.45; Birkhead and Perrins 1986) also used by Ellis and Elphick (2007) . Estimation of demographic parameters described in previous chapters and whose values are included herein involved the capture and handling of live mute swans led 123 by staff of the U.S. Department o f Agriculture Wildlife Services section of the Animal Plant Health Inspection Service (USDA APHIS WS). Michigan State University (MSU) Institutional Animal Care and Use Committee (IACUC) granted an animal - use exemption for MSU personnel throughout this pro ject. This work was also partially supported by salary support for Scott R. Winterstein from the USDA National Institute of Food and Agriculture (Project No. MICL02588). We removed Ellis and Elphick (2007) clutch size penalty (0.81) for inexperienced breeders in Michigan - specific models since differentiation of experienced and inexperienced pairs could not be determined in this study. M ean clutch size for pairs with known breedi ng experience was equal to the overall estimated clutch size (7.0 eggs/clutch) , whereas mean clutch size for pairs with unknown experience ( likely a mix of experienced and inexperienced pairs) was 7.2 eggs/clutch. Therefore, removal of the clutch size penalty was justified in this modeling . We also used identical survival rates for immature non - breeding (1 - 2 - year - old immature) and adult non - breeding swans since those two life stages were pooled in survival analyses due to sample size constraints (Chapte r 4) . Ellis and Elphick (2007) used survival estimates for these two stages derived from Watol a et al. (2003) . S urvival estimates for 1 - 2 - year - old immature non - breeders (0.73) and adult non - breeders (0.71 ; Watola et al. 2003 ) were similar and near our estimated survival for these two life stages pooled (0.698); therefore, we used the same surviv al estimate for both life stages in our Michigan - parameterized model s . Life t able r esponse e xperiment b etween n ative - and MI - p arameterization Life table response experiments are useful for comparing matrix population models of identical structure but different input parameters (Caswell 1989) . R esults of LTREs show proportional change in population growth rate attributed to each input parameter of the transition 124 matrix (Caswell 2001) . We compared relative contributions of demographic parameters to change in lambda between native - and Michigan - parameterized models. To conduct the LTRE, w e averaged transition matrices for the native - and Michigan - parameterized models and then derived sensitivity for each parameter for the mean transition matrix. We then multiplied sensitivities for each parameter in the mean transition matrix by change in input values between the transitio n matrices of the native - and Michigan - parameterized models to determine the influence that the difference in input values had on the overall observed change in lambda between the two models. Development of a Density - Dependent Matrix Population Model for M ichigan We were able to estimate life stage - specific demographic parameters for mute swans within Michigan (Chapters 2 and 4); however, we noted that site - level breeding productivity was influenced by number of nesting pairs and saturation of characterist ic nesting habitat (Chapter 3) , als o demonstrated in another introduced population (Nummi and Saari 2003) . Our six stud y sites (see Figure 3.1 in Chapter 3) . S ites also varied in nesting density , abundance of nesting cover, and breeding productivity (Chapter 3). Our estimated egg to fledge survival rates (0.198), which we ha ve demonstrated is influenced by nesting pair density (Chapter 3), is one of the lowest egg to fledge survival rates reported in the literature and nearly approach ed the extreme low productivity observed by Nummi and Saari (2003) in the late stages of invasion in a Fin nish archipelago. Our s ite selection criteria sought study sites with multiple nesting pairs within a localized area . This criterion helps ensure that adequate sample sizes of nesting pairs could be obtained for intensive nesting ecology investigation whil e representing variability present geographically in 125 . Selection criteria required that at least 5 nesting pairs were located within a 36 km 2 area at the start of the nesting ecology study in 2016. Observed density dependence in b reeding productivity resulted in our selection criteria being suboptimal for estimating mean productivity across all mute swans in Michigan since areas with few (i.e., 1 or 2) nesting pairs were not represented in our sample of sites . Density - dependent bre eding productivity represents the respective nesting pair densities of these sites; however, distribution of nesting pair density across our sites may not represent the distribution of nesting pair densities across Michigan (i.e., there are likely many are as with few nesting pairs but abundant nesting cover since observed population growth is no t indicating that the population is at carrying capacity ; MDNR, unpublished data ). Further, our estimation of breeding productivity, which is known to be density - dep endent (Chapter 3 , Nummi and Saari 2003), is no t likely applicable across all levels of mute swan abundance from the first stages of invasion to present day ; therefore, it was appropriate to incorporat e density dependence in productivit y at varying levels of mute swan abundance . We incorporate d density dependence in productivity within the Michigan - parameterized MPM using a penalty term that changed breeding productivity according to abundance in previo us year. Th e penalty term proportionally lowered breeding productivity as mute swan abundance increased in relation to a hypothesized carrying capacity ( K ). This penalty term for density specific breeding productivity ( F t ) took the form where F is the initial breeding productivity rate (i.e., mean clutch size multiplied by an estimated egg to fledge survival rate) under low nesting pair densities, K is the hypothesized carrying capacity, and N t - 1 is the total population of females in the previous time step (Jensen 1995) . The penalty equation describes a linear relation between productivity and pair density we observed among local study 126 areas (Chapter 3); however, K in this equation refers to statewide mute swan abundance. The penalty term resulted in equal productivity for experienced and inexperienced breeders since we did no t detect differences in breeding productivity related to prior breeding experience (see above) . Egg to fledge survival rates have been reported in th e literature for native and introduced populations. We used an initial egg to fledge survival rate that minimized influence of density dependence; therefore, we reported egg to fledge survival rates for newly established introduced populations that represe nted vital rates for populations below carrying capacity. Nummi and Saari (2003) reported an egg to fled ge survival rate of 0.44 during the early stages of invasion in a Finnish archipelago whereas values for introduced populations in North America were slightly different (Willey 1968, 0.49; Reese 1975, 0.46; Reese 1980, 0.40; Conover and Kania 1999, 0.41). We chose an initial egg to fledge survival rate of 0.46 reported by Reese (1975) for a population that had nested in the wild for ~10 years. A 0.46 egg to fledge survival rate coupled with our observed mean clutch size in Michigan ( 7.0 eggs per pair ; Chapter 2) resulted in an estimated 3.2 fledged young per pair (1.6 fledged females per breeding female assuming a 1:1 sex ratio at birth; Willey 1968) during the early stages of invasion. Therefore, the penalty term started with a breeding productivity value ( F ) of 1.6 and was reduced in proportion to population abundance according to the penalty term described above. Other parameters in the transition matrix for the density - dependent matrix population model were identical to those listed for the Michigan - parameterized model in Table 5.1. Use of the penalty term required a preset value of estimated carrying capacity for mute swans in Michigan. from 1949 2010 was 9.3 % ; however, population growth appeared to slow from 1991 to 2010 127 (4.5 % annual growth) indicating that density may have begun to influence vital rates (D Luukkonen, unpublished data) . This period of observed slower growth corresponded to an estimated abundance of 4,069 swans in 1991 to 15,532 swan s in 2010 (D . Luukkonen, unpublished data ) . While the carrying capacity of mute swans in Michigan is unknown, it seems reasonable that up to 3 1 , 25 0 nesting pairs (i.e., under the 1 25, 000 K simulation) could establish territories on the more than 46,000 inl and lakes (Institute for Fisheries Research 2013) or wit hin wetlands associated with waters of the Great Lakes. This would assume that about half of the total estimated population (i.e., 62,500 individuals ) would be found in mixed flocks of immature non - breeding or adult non - breeding swans which has been noted for established populations in their native range (Baker et al. 2006) . We evaluated sensitivity of the density - dependent matrix population model using three different hypothetical values for carrying capacity ( K ) of mute swans in Michigan set within the penalty ter m (75,000 individuals, 100,000 individuals, and 125,000 individuals) . Estimation of removal rates needed to achieve long - term population goal We used the Michigan - parameterized density - dependent MPM for a hypothesized carrying capacity of 100,000 individuals to estimate reduction in demographic parameters needed to achieve the long - term goal of no more than 2,000 mute swans by 2030 (Michigan Department of Natural Resources 2012) . We began s imulations with the 2018 estimated abundance (12,048 individuals) per the annual breeding waterfowl survey. We used an initial population vector f or 2018 that consisted of 6,024 total females distributed across life stages according to the 100,000 K densit y - dependent model (1949 2098 simulation) : 2085 juveniles (0 - 1 - year - old immatures), 1055 immature non - breeders (1 - 2 - year - old immatures), 501 128 inexperienced breeders, 613 inexperienced non - breeders , 1358 experienced breeders, and 412 experienced non - breeder s ). We conducted simulations that proportionally reduced survival for all age classes (i.e., juveniles, immature non - breeders , non - breeders , and breeding swans) in 10 % increments and observed estimated abundance in the year 2030. We reduced survival rates by 1 % increments once estimated abundance was near acceptable levels in 2030 to determine the minimum percent reduction in survival needed to achieve the long - term goal. We chose to evenly reduce the survival of all age classes by the same percentage beca use determini ng life stages prior to removal during culling efforts is difficult especially during winter when all life stages may be present in the same flock. It is important to note that estimation of the reduction in survival through these methods assu mes that annual removals of each life stage exhibit complete additi ve mortality. We also estimated reduction in survival and removal needed to achieve the long - term goal by only targeting a specific life stage (i.e., juvenile, nonbreeder, or breeder). This allowed of removing individuals in a specifi c stage with regard to reaching the 2030 goal . We accomplished this by systematically adjust ing survival , as in previous simulations, but we only adjusted survival rates for one life stage in each simulation and held survival for the other stages at their values listed in Table 5.1 . We were able to use these simulations to estimate the number (and proportion) of each life stage that must be removed annually to achieve the long - term goal if the remaining life stages were unavailable for removal. Th ese simula tion s assumed that no artificial manipulation of clutch sizes or egg to fledge survival (i.e., nest destruction or egg oiling) occurred during the modeling period. 129 We also used the Michigan - parameterized density - dependent MPM to determine level of egg an d nest destruction needed to achieve the same level of mute swan abundance by the year 2030. Similar to the simulation s above that incrementally reduced adult survival, we incrementally reduced egg to fledge survival rates by 10 % for each modeling iteratio n and then incrementally reduced egg to fledge survival by 1 % once near threshold of acceptable abundance in 2030 to identify the minimum reduction needed to achieve the long - term goal (Michigan Department of Natural Resources 2012) . These simula tions allowed us to calculate the number of eggs and nests that must be destroyed annually to reach long - term objectives. RESULTS Comparison of native - and Michigan - parameterization The native - parameterized matrix population model using an initial population vector from Michigan in 1948 overestimated abundance in Michigan during the 1991 2018 period - wing aerial abundance survey ( Figure 5. 1). The estimated intrinsic rate of population growth for the native - parameterized model was 1.142. The Michigan - parameterized density - independent matrix population model underestimated future abundance in Michigan (Figure 5.1) and resulted in a negative i ntrinsic growth rate ( = 0.979). Generation time was similar between the native - (8.79 years) and Michigan - parameterized model (8. 11 years). Survival of experienced breeders was most elastic in both models followed by survival of juvenile swans (0 - 1 - year - old immature swans) and 1 - 2 - year - old immature non - breeders (Figure 5.2). The life table response experiment demonstrated that much of the observed change in lambda between the native - and Michigan - parameterized density - independent MPMs ( - 0.163 ) resulted fr om reduction in the probability 130 that an experienced breeder survives and continues to breed ( - 0.07 5) although the reduction in reproduction of experienced breeders was also important ( - 0.059 ). Figure 5.1. Comparison of observed mute swan abundance (blue points) to predicted mute swan abundance in Michigan, USA, for a 150 - year simulation (1949 2 098 ) between the native - (black line) and Michigan - parameterized (red line) deterministic matrix population model. 131 Figure 5. 2. Comparison of parameter elasticity between deterministic density - independent native - and Michigan - parameterized matrix population models and the deterministic density - dependent Michigan - parameterized model. Density - d ependent m atrix p opulation m odel for Michigan The Michigan - parameterized density - dependent MPM fit the observed mute swan population estimates better than the native - parameterized density - independent model (Figure 5.3). Predicted population size near observed population abundance estimates in 2018 (12,04 7 swans; MDNR, unpublished data) w ere similar between all three levels of K used within the penalty term (Figure 5.3) demonstrating that the model was not overly sensitive to the K chosen when abundance is not near K . Mute swan populations co uld have be en expected to grow to 31,249 individuals, 39,608 individuals, or 47,131 individuals by 2030 under the three density - dependent simulations using varying levels of K in the penalty term , respectively, if expanded control efforts were not initiated in 2011. If control efforts were to stop in 2018, abundance of mute swan s in Michigan could expect to grow to 26,034 individuals by 2030. Incorporation of density - dependence into the model caused the intrinsic rate of population growth and distri bution 0.0 0.1 0.2 0.3 0.4 Survival of experienced breeders Survival of inexperienced breeders Survival of experienced nonbreeders Survival of inexperienced nonbreeders Survival of juveniles Survival of immature nonbreeders Reproduction of experienced breeders Reproduction of inexperienced breeders Elasticity Native (density-independent) Michigan (density-independent) Michigan (density-dependent) 132 of life stages to vary throughout the simulations until the population was near a realized carrying capacity. During the s imulation periods, abundance did not reach the carrying capacity set in the penalty term due to input values in more elastic pa rameters in the model (e.g., experienced breeder and juvenile [0 - 1 - year - old immature] survival). The relative elasticity ranking for parameters was similar in all density - dependent models although specific values varied among MPMs (Figure 5.2) . Experienced breeder survival, juvenile survival (0 - 1 - year - old immature), and immature nonbreeder (1 - 2 - year - old immature) survival were the most elastic parameters ( Figure 5. 2 ). The penalty term in the 100,000 K density - dependent model varied from 3.2 fledg ed cygnets/pair at low pair abundance to 1.6 fledged individuals per pair when abundance was near estimated carrying capacity. Areas where pair productivity is below the value of a stable population (1.6 fledglings per pair ; i.e., egg to fledge survival is 0.229 or lower ) were not contributing to overall population growth at current estimated levels of life - stage specific survival. 133 Figure 5. 3. Comparison of native - parameterized density - independent matrix population model and a Michigan - parameterized density - dependent matrix population model under 3 simulated levels of carrying capacity for mute swans in Michigan, USA. Estimation of removal rates needed to achieve long - term population goal s Proportional reduction in survival acr oss all swan life stages A 26 % reduction in survival for all life stages was needed for mute swan s in Michigan to reach the long - term goal of a population estimate of fewer than 2,000 mute swans based on the 2030 breeding survey conducted by the MDNR (Fig ure 5.4). A 26 % reduction in survival for all life stages would require removal of 12,760 swans from 2018 2029 which requires a mean annual removal rate of 17.2 % of the estimated annual population (Table 5.2). Annual removals would need to be distributed across the three condensed life stages represented in the model to achieve the long - term goal (Table 5.2) . Thirty - three percent of annual removals should be juvenile swans (i.e., post - fledging swans not yet 1 year old), 35 % should non - breeding swans, and 3 2 % should be breeding swans (Table 5.2) . 134 Figure 5.4. Comparison of density - dependent modeled reduction in survival needed across all mute swan life stages to achieve the long - term goal of fewer than 2,000 mute swans in Michigan, USA, by the year 2030. Table 5. 2. Annual removal needed by life stage to achieve long - term goal of fewer than 2,000 mute swans in Michigan, USA, by the year 2030 using the 100,000 K density - dependent matrix population model with an assumed 26 % reduction in survival for al l life stages over baseline rates. Annual Removal by Life Stage Year Juveniles Non - breeders Breeders Total 2018 570 755 822 2147 2019 716 583 640 1939 2020 562 593 497 1652 2021 445 504 426 1374 2022 388 412 362 1163 2023 334 351 303 989 2024 283 301 256 840 2025 241 255 217 713 2026 206 217 184 607 2027 176 185 156 518 2028 150 158 133 441 2029 128 135 113 377 Total 4199 4451 4110 12760 A maintenance level of 11 % reduction in survival across all swan life stages is needed o nce the long - term goal is met in 2030 to keep annual abundance fewer than 2,000 mute swans. 135 This requires about 7.3 % of the population to be removed annually which would be a mean annual removal of at least 146 swans with a similar distribution of removal across the three condensed life stages. Comparison of life - stage specific removals needed to achieve long - term goal Juvenile (0 - 1 immature) survival would need to be reduced by 96 % to achieve the long - term goal in 2030 if all other life stage survival rates remained at values listed in Table 5.1. This would require an annual removal of 50.5 % of the juvenile cohort whic h totals to a removal of 20,445 juvenile swans 2018 2029. Survival for non - breeding swans (i.e., 1 - 2 - year - old immature non - breeders , experienced non - breeders , and inexperienced non - breeders ) needs to be reduced by 65 % annually to reach the long - term goal which would require 45.4 % of the non - breeding swans in Michigan to be removed annually totaling 12,188 swans 2018 2029. About 60 % of breeding swans would need to be removed annually (i.e., 70 % reduction in survival) to achieve the same long - term goal. T his would require a total of 10,036 breeding swans removed 2018 2029 assuming no removals occurred for other life - stages. Less than half of the total swans would need to be removed if culling efforts were targeted solely on breeding swans when compared t o solely on juvenile swans. Egg and nest destruction needed to reach long - term goal Baseline levels of breeding productivity would need to be reduced by 88 % annually to achieve the long - term population goal in 2030 . This requires at least 94.5 % of all mut e swan nests in the state of Michigan to be destroyed annually 2018 2029 to reach the long - term goal which would require removal of at least 15,748 nests and approximately 110,237 eggs (Table 5.3). Low natural egg to fledge survival under optimal nesting conditions (0.46) and the inability 136 to know which remaining eggs can be expected to fledge will result in destroying approximately 63,000 eggs (61.6 % of total) that would no t have produced a fledged cygnet if left untouched. Table 5.3. Number of eggs and nests that must be destroyed annually to achieve long - term goal of fewer than 2,000 mute swans in Michigan, USA, by the year 2030 using the 100,000 K density - dependent matrix population model with an assumed 88 % reduction egg to fledge survival over ba seline rates . Total Present Total to Remove Year Nests Eggs Nests Eggs 2018 1859 13013 1756 12295 2019 1957 13696 1849 12940 2020 2054 14379 1941 13585 2021 1845 12917 1743 12204 2022 1617 11318 1528 10693 2023 1411 9879 1333 9334 2024 1230 8612 1162 8137 2025 1072 7506 1013 7092 2026 935 6543 883 6182 2027 815 5706 770 5391 2028 711 4977 672 4702 2029 620 4342 586 4102 2030 541 3789 511 3580 Total 16668 116677 15748 110237 DISCUSSION Our Michigan - specific density dependent MPM , like many analyses that examine dynamics of long - lived species (Watola et al. 2003, Ellis and Elphick 2007, Alisauskas et al. 2011) , suggests that adult survival (combined elasti city of 0.773) is the most influential parameter when compared to juvenile survival or breeding productivity (Figure 5.2) . We found that the breeder subsegment of the adult population was most influential (0.477) , although the non - breeding segment was also important (0.296). Results of the Life Table Response Experiment show that the change in lambda observed between the two density - independent models was caused by our lower observed survival rate and probability of continuing to breed for breeding adults. We successfully incorporate d density - dependence into breeding productivity of 137 our matrix population model althoug h strength of density dependence in productivity was low to moderate at our observed population abundance (Figure 5.3) . However, incorporation of density dependence was helpful in estimating future trajectories of abundance if management of mute swans became infeasible due to political will or public desire. Our simulation modeling using a density - dependent matrix population model shows that efforts that reduce productivity of breeding swans is relatively inefficient compared to removal of adults for reducing the overall mute swan population in Michigan. Howev er, destroying nests or eggs through removal of nest vegetation o r oiling of eggs remains a useful technique for locally reducing the number of cygnets during the summer months , alleviating aggressive behavior of brooding pairs , and retaining public support for comprehensive management programs (Watola et al. 2003, Allin and Husband 2004) . Our modeling also show s that most eggs destroyed through egg oiling o r nest removal efforts would no t have produced a fledged cygnet if management actions were not taken further demonstrating its inefficiency as a population management strategy . Our simulations of management alternatives that target differential culling of juvenile and adult swans provide s flexible management options to agencies tasked with mute swan management . All life stages of mute swans are available for removal throughout Michigan; however, not all individuals in a life stage share an equal probability of being removed. Some breeding adults remain on private waterbo dies or in heavily developed areas year - round (Chapter 4) protecting them from most removal efforts. Additionally, flocks of non - breeders may summer in remote shallow water areas of the Great Lakes and escape removal due to lack of detection or the inabili ty for crews to reach the area. O ur simulations also show that not all life stages of mute swans e qually contribute to population demographics (i.e., culling a juvenile swan does no t 138 provide the same population - level impact as removing an actively breeding swan). Simulations show that m anagement agencies w ould need to cull approximately twice as many juvenile swan s as breeding swans to reach the long - term goal if culling efforts focu s ed only on the juvenile life stag e instead of the breeding life stage . MANAGEMENT IMPLICATI ONS Our density - dependent model provides managers with a tool for evaluating alternatives for mute swan population control. Simulations through our matrix population model suggest that management should focus on culling juvenile and adult swans rather than focusing on reducing breeding output. Additionally, culling efforts should target all life stages that are available for removal (i.e., juvenile, adult non - breeding , and adult breeding) rather t han only focusing on one life stage . Removing swans from all life stages results in a management program that is robust to circumstances where the solely targeted life stage is unavailable for removal (i.e., breeding swans remaining on territories year - rou nd due to less ice coverage under climate change). S tratifying annual removals among life stages also allow s fo r temporal targeting of specific stages. Molting flocks of non - breeding swans congregate in mid - to late - summer on large inland lakes and open w ater areas of the Great Lakes. These flocks are nearly entirely consisted of immature non - breeding and adult non - breeding swans. They could be targeted for removals to ensure that the culling quota for non - breeding swans is met or exceeded annually. A simi lar, although less efficient, strategy exists for ensuring that the annual quota of breeding swans is reached or exceeded. Culling efforts can target breeding pairs in early spring (i.e., April) through early summer (i.e., June) when they are either actively incubating eggs or brooding cygnets. Breeding swans that failed nesting or lost hatched cygnets early in the brood rearing period are still likely to be near their nesting locati on (R. Knapik, personal obs.). 139 Distinguishing between adult life stages during winter culling efforts can be impossible although juvenile swans (i.e., swans < 1 year old) can be reliably distinguished from adults. Temporally targeting specific life stages throughout the year can help ensure that removal quotas for each life stage are met. Annual removal goals for each life - stage under a proportional reduction in survival ( 26 % ; Table 5.2) should be considered conservative minimum goals since slight annual va riation in input parameters (e.g., stochasticity in adult survival, juvenile survival, or reproduction) can lead to changes in the absolute number of swans that must be removed to be certain of a decline (Ellis and Elphick 2007) . Further, it should be noted that predicted annual removals to achieve the long - term goal assume that removal goals are met annually. Additional swans will need to be removed in subsequent years following the failure to achieve annual removal goals ( Table 5.2 ) . Alternatively, a strategy of removing more swans early in population control compared to a policy of an annually - constant removal rate would reduce the overal l number of swans needed to be removed to achieve population goals by 2030. We hav e presented demographic outcomes of several realistic management strategies that target var i ous life stages (i.e., reducing reproduction, proportional culling of all life sta ges, or targeting specific life stages); however, the management strategy that works best for a given agency is one that can be reliably accomplished annually given constraints of funding, staff time, and public cooperation. Our density - dependent matrix po pulation model can be used to model additional management strategies that may work best for management agencies. 140 LITERATURE CITED 141 LITERATURE CITED Alisauskas, R. T., R. F. Rockwell, K. W. Dufour, E. G. Cooch, G. Zimmerman, K. L. Drake, J. O. Leafloor, T. J. Moser, and E. T. Reed. 2011. Harvest, survival, and abundance of midcontinent lesser snow geese relative to population reduction efforts. Wildlife Monographs 179:1 - 42. A llin, C., G. Chasko, and T. P. Husband. 1987. Mute Swans in the Atlantic flyway: a review of the history, population growth and management needs. Northeast Section of The Wildlife Society 44:32 - 42. Allin, C., and T. Husband. 2004. An evaluation of 22 year s of mute swan management in Rhode Island. Arsnoe, D., and A. Duffiney. 2018. From beauty to beast. Pages 40 - 44 in The Wildlife Professional 12:3. The Wildlife Society Bethesda, MD. Baker, H., D. A. Stroud, N. J. Aebischer, P. A. Cranswick, R. D. Gregor y, C. A. McSorley, D. G. Noble, and M. M. Rehfisch. 2006. Population estimates of birds in Great Britain and the United Kingdom. British Birds 99:25. Besbeas, P., S. N. Freeman, B. J. T. Morgan, and E. A. Catchpole. 2002. Integrating mark - recapture - recove ry and census data to estimate animal abundance and demographic parameters. Biometrics 58:540 - 547. Birkhead, M., and C. M. Perrins. 1986. The Mute Swan. Croom Helm. Breck, J. E. 2004. Compilation of databases on Michigan lakes. Michigan Department of Natural Resources, Fisheries Division. Caswell, H. 1989. Analysis of life table response experiments I. Decomposition of effects on population growth rate. ecological modelling 46:221 - 237. _____. 2001. Matrix population models. Wiley Online Library. Conover, M. R., and G. S. Kania. 1999. Reproductive success of exotic Mute Swans in Connecticut. Auk 116:1127 - 1131. Cox, D. R. 1972. Regression models and life - tables. Journal of the Royal Statistical Society Series B - Statistical Methodology 34 :187. Easterling, M. R., S. P. Ellner, and P. M. Dixon. 2000. Size specific sensitivity: applying a new structured population model. Ecology 81:694 - 708. 142 Ellis, M. M., and C. S. Elphick. 2007. Using a stochastic model to examine the ecological, economic a nd ethical consequences of population control in a charismatic invasive species: Mute Swans in North America. Journal of Applied Ecology 44:312 - 322. Gelston, W., and R. Wood. 1982. The Mute Swan in northern Michigan. Traverse City, MI , USA . Hindman, L. J., and W. F. Harvey IV. 2004. Status and management of feral Mute Swans in Maryland. Institute for Fisheries Research. 2013. Facts about Michigan's lakes. < https://www.michigan.gov/dnr/0,4570,7 - 350 - 79135_81276_82887 - 160092 -- ,00.html >. Jensen, A. 1995. Simple density - dependent matrix model for population projection. E cological M odelling 77:43 - 48. Leslie, P. H. 1945. On the use of matrices in certain p opulation mathematics. Biometrika:183 - 212. McCleery, R. H., C. Perrins, D. Wheeler, and S. Groves. 2002. Population structure, survival rates and productivity of Mute Swans breeding in a colony at Abbotsbury, Dorset, England. Waterbirds:192 - 201. Michigan Department of Natural Resources. 2012. Mute Swan management and control program policy and procedures. < www.michigan.gov/documents/dnr/2012_Mute_Swan_Policy_378701_7. pdf >. Accessed 01 March 2016. Miller, T. E., S. M. Louda, K. A. Rose, and J. O. Eckberg. 2009. Impacts of insect herbivory on cactus population dynamics: experimental demography across an environmental gradient. Ecological Monographs 79:155 - 172. Munro, R. E., L. T. Smith, and J. J. Kupa. 1968. Genetic basis of color differences observed in Mute Swan (Cygnus olor). Auk 85:504 - 505. Nummi, P., and L. Saari. 2003. Density - dependent decline of breeding success in an introduced, increasing Mute Swan Cygnus olor population. Journal of Avian Biology 34:105 - 111. Pugh, S. A., D. C. Heym, B. J. Butler, D. E. Haugen, C. M. Kurtz, W. H. McWilliams, P. D. Miles, R. S. Morin, M. D. Nelson, and R. I. Riemann. 2017. Michigan forests 2014. Resour. Bul l. NRS - 110. Newtown Square, PA: US Department of Agriculture, Forest Service, Northern Research Station. 154 p. 110:1 - 154. R Development Core Team. 2018. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vien na, Austria. < https://www.R - project.org/ >. 143 Reese, J. G. 1975. Productivity and management of feral Mute Swans in Chesapeake Bay. The Journal of Wildlife Management 39:280 - 286. _____. 1980. Demography of European Mute Swans in Chesapeake Bay. Auk 97:449 - 464. Schaetzl, R. J., H. Enander, M. D. Luehmann, D. P. Lusch, C. Fish, M. Bigsby, M. Steigmeyer, J. Guasco, C. Forgacs, and A. Pollyea. 2013. Mapping the physiography of Michigan with GIS. Physical Geography 34:2 - 39. Sommers, L. M. 1984. Michigan: a geography. Westview Press. Stubben, C., and B. Milligan. 2007. Estimating and analyzing demographic models using the popbio package in R. Journal of Statistical Software 22:1 - 23. Watola, G. V., D. A. Stone, G. C. Sm ith, G. J. Forrester, A. E. Coleman, J. T. Coleman, M. J. Goulding, K. A. Robinson, and T. P. Milsom. 2003. Analyses of two Mute Swan populations and the effects of clutch reduction: implications for population management. Journal of Applied Ecology 40:565 - 579. Willey, C. H. 1968. The ecological significance of the Mute Swan in Rhode Island. Transactions of the Northeast Wildlife Conference v. 25,p. 121 - 134 Wood, R., and W. L. Gelston. 1972. Preliminary report: the Mute Swans of Michigan's Grand Traverse Bay region. State of Michigan, Department of Natural Resources, Wildlife Division. 144 CHAPTER 6: MANAGEMEN T IMPLICATIONS Human - assisted movement s of mute swans to North America resulted in feral populations along the U.S. East Coast, in the Great Lakes region, and in isolated pockets throughout the western U.S. (Allin et al. 1987, Sullivan et al. 2009) . Populations can grow quickly once estab lished and persist for long periods due to relatively high survival for adult life stages (Chapter 4, Watola et al. 2003) . Management a gencies are often tasked with creating polices withou t understanding of regional dynamics and movement s of mute swans since early detection an d decision making is typically needed to combat invasive species (Mack et al. 2000, Edelaar and Tella 2012) . Using poorly parameterized demographic models to project outcomes of future management can produce undesirable results if control levels are insufficient to reduce overall abundance. Additionally, management programs may not be initiated if there is a perceived inabili ty to control the invasive species given resources available for management. Agencies also face public scrutiny over management plans due to the charismatic nature of some invasive species such as mute swans and their close association with humans (Allin et al. 1987) ; however, many individuals, at least in Michigan, are concerned with specific management techniques and not necessarily opposed to all control methods (Jager et al. 2016) . The public wants to ensure that science - based management policies are selective , effective, considered nonlethal options, and carried out by trained biologists (Reiter et al. 1999) . Review of mute swan management in Michigan The Michigan Department of Natural Resources (DNR) formally outlined their management policy for mute swans in 2012 along with stating short - and long - term goals which were to reduce population growth initially and to have fewer tha n 2,000 mute swans statewide by 2030 (Michigan Department of Natural Resources 2012) . Concomitantly, the Mississippi Flyway 145 Council also outlined flyway - wide abundance goals of fewer than 4,000 mute swans by 2030. Management in Michigan prior to 2011 primarily focused on addressing human - wildlife conflict although some larger removal efforts occurred . Conflict resolution typically resulted in oiling eggs, removing of nest material, or culling of individuals. The Wildlife Services division of the U .S. Department of Agriculture Animal and Plant Health Inspection Service (USDA APHIS) began receiving additional funding from the Great Lakes Restoration Initiative (GLRI) in 2011 to increase mute swan surveillance and intensify their culling efforts (Marks 2015) . The additional funding provided by the GLRI and sustain ed funding from the Michigan D NR allowed for a more robust statewide culling program which removed over 9,700 individuals since 2011 (Arsnoe and Duffiney 2018) . However, the mute swan population in Michigan appears to have stabilized despite an initi al decrease in abundance (MDNR, unpublished data). Population stabilization and increased difficulty in accessing areas to remove mute swans prompted research to estimate Michigan - specific demographic parameters and understand regional movement for sub - adu lt and adult swans. It was unknown if removal levels achieved 2011 present were sufficient to meet the long - term goal ; however, results from this research suggest that removal rates were insufficient in all years except in 2012 when 2,628 mute swans were culled (1 7 % of the 2012 estimated population ; Marks 2015, MDN R, unpublished data ) . Our study of demographics and movement across the full life cycle of the mute swan has improved demographic models and has also provided practical in - field guidance for mute swan management. Overview of pertinent results from demograp hic and movement study Our investigation of mute swan nesting ecology found productivity varied across Michigan with density of nesting pairs. Mean clutch size in Michigan (7.0 eggs/pair) was higher 146 than in other areas of native and introduced range s ; howe ver, mean productivity was lower especially when considering sites with many pairs and saturated nesting cover (Chapter 3). Reduced cygnet survival was related to variation in territory quality for saturated sites as a result of high competition for charac teristic territories (Chapter 3) . Survival for immature and adult non - breeding swans (Chapter 4) was similar to values reported for other areas of their range (Watola et al. 2003) . Importantly, juvenile and breeding adult survival (Chapter 4) was slightly lower than values reported in the literature (Watola et al. 2003) . Slightly lower values for adult survival resulted in a large change in the population growth rate ( ) when comparing a native - parameterized model to a Michigan - specific model especially since survival and probability of continuing to breed for adult breeding swans was most elastic in our demographic model (Chapter 5). Our density - dependent matrix population model allowed comparison of management scenarios aimed at reaching the long - term goal established by the Michigan DNR given the estimated abundance in 2018 and observed demographics from this study. Management sc enarios to reach long - term abundance goal Demographic modeling suggests that a 26 % reduction in survival across all life stages is needed to reduce mute swan abundance to goal levels by 2030. This will require removal of at least 17 % of the annually estim ate d population; however, annual removals must be distributed across 3 primary life stages (32 % juvenile, 35 % non - breeding swan, and 33 % breeding swan) to be effective . A total of at least 12,760 swans would need to be removed before 2030 with annual remov al goals listed in Table 5.2. It is important to note that agencies tasked with mute swan management in Michigan should consider annual removal goals listed in Table 5.2 to be conservative minimums since they assume complete additive mortality for those li fe stages . 147 S tochasticity in demographic rates could also result in slight variation in the number of swans that n eed to be removed yearly to reach the long - term goal (Ellis and Elphick 2007) . We also found that increased management effort is needed if only one swan life stage is targeted for removal or if egg and nest destruction was the chosen management method. An estimated 20,445 juvenile swans (i.e., fledged swans not yet 1 year old) would need to be removed from 2018 2029 to achieve the long - term goal. This would require annual fall and winter culling of at least 50 % of the juveniles statewide . About 40 % of non - breeders would need to be removed annually (12,188 total removed non - breeders 2018 2029) to achieve the same goal. At least 60 % of breeding swans (10,036 total removed breeding swans 2 018 2029) would need to be removed annually for mute swan abundance in Michigan to be fewer than 2,000 mute swans by 2030. The long - term abundance goal could only be met through reduction in breeding productivity if at least 94.5 % of all mute swan nests and eggs were destroyed annually in Michigan (15,748 nests and approximately 110,237 eggs destroyed 2018 2029) . Further, the inefficiency of egg and nest destruction is highlighted in the fact that about 61 % of the total eggs destroyed would no t have pro duced a fledged cygnet anyway. It is impossible for managers to know which eggs w ill produce fledged cygnets during incubation ; therefore, a ll eggs must be destroyed regardless of potential fate . Watola et al. (2003) evaluate d the use of clutch size reduction in reducing overall abundance but found partial clutch removal to be ineffective at reducing ultimate productivity and overall mute swan abundance. Simulations of our density - dependent matrix population model show ed that the population of mute swans in Michigan stabilizes when mean statewide breeding productivity is near 1.6 fledg lings per pair (22.8% egg to fledge survival ; mean clutch size of 7). Further, w e ha ve shown that there is spatial variation in productivit y throughout Michigan that results from 148 differing nesting pair densities and competition for characteristic nesting territories. These two findings suggest that productivity within areas where density dependence is strongest (i.e., where characteristic nes ting cover is saturated, and pairs are nesting in suboptimal cover) is not contributing to overall population growth in Michigan . A t least 81 % of all eggs produced in these areas fails to yield a fledged cygnet during our study . Areas where mean productivi ty is near 1.6 fledglings per pair are contributing to overall population stability but may not be contributing to growth. Conversely, breeding pairs in areas where nesting cover is abundant, pair density is low, and productivity is in excess of 1.6 indivi duals per pair are contributing to overall population growth in Michigan. Practical considerations for future management Egg and nest destruction O bserved density - dependence in breeding productivity and low elasticity for egg to fledge survival in demogra phic modeling shows that destruction of nests and oiling of eggs is inefficient at reducing overall mute swan abundance . However, egg oiling and nest destruction remain useful management tool s to locally reduce summer cygnet abundance, limit aggression by nesting pairs, and to retain support for the management program by concern ed citizens. If nest destruction operations must only focus on some nests within a given area, nests in typical nesting cover or those which have been present for multiple years shou ld be prioritized for destruction as those are the most likely to produce fledged cygnets . N ests that are being targeted for egg oiling and nest destruction should be checked between 20 30 April when most breeding females are 149 incubating and at least a we ek prior to our observed mean hatch date of 12 May. Low - level surveys in fixed - wing aircraft are invaluable at locati ng nests for such operations. Life stage - specific removal of mute swans The long - term statewide abundance goal of fewer than 2,000 mute swans in Michigan by 2030 can be achieved if removal efforts target all juvenile or adult life stages . Targeting swans by life stage allows different management objectives to be met year - round since some stages (e.g., juveniles and breeders) are harder to target during some seasons. Flocks of non - breeding swans can be accessed in summer once they settle on large inland bodies of water or on waters of the Great Lakes (typically by 15 July) to co mplete their annual flight feather molt. Non - breeding swans begin to move from molting areas by the end of August. Juvenile swans and breeding pairs are most readily accessible when ice coverage forces them to regional wintering sites (typically late Decem ber) . Removals in these areas should be conducted before breeding females start returning to nesting territories in mid - February . Mean displacement from nesting location for breeding females was about 11 km; therefore, winter culling efforts targeted at br eeding females will likely need to occur at inland wintering concentration sites in additional to waters of the Great Lakes. It is important to note that targeting mute swans during periods of widespread ice coverage may have lethal or sublethal impacts on non - target species like native waterfowl that also congregate in those regions during periods of ice coverage ; however, the magnitude of impact to native species is unknown. Although o ur demographic model is spatially invariant and assumes that life stag e - specific removals have a uniform effect on survival and reproduction statewide , our knowledge of mute swan distribution and density dependence in breeding productivity can further leverage the effectiveness of life stage - specific culling . The current goa l of the Michigan DNR is to 150 reduce abundance to fewer than 2,000 individuals by 2030 (Michigan Department of Natural Resources 2012) . This damage management approach i s preferable to complete eradication in the short term since immigration is pos sible from nearby mute swans in Ontario and could hinder efforts for complete removal (Bomford and O'Brien 1995) . Therefore, r emoval of breeding swans will be most effective at reducing overall abundance when conducted in areas where breeding productivity is above average. Removal of juvenile and non - breeding swans will also be most effective at reducing overall abundance when in proximity to areas of unoccupied typical nesting cover. Conversely, removal of breeding pairs that nest in areas of saturated nesting cover may not cause a n overall reduction in local productivity due to effects of density dependence in productivity. We s uggest that culling efforts , especially for breeding swans, should first be focused in regions where regional density is low to maximize reduction in breeding output and reduce the ease of which non - breeders can occupy unfilled characteristic nesting terri tories . We acknowledge that targeting removals in areas with fewer total swans is likely to increase the cost per swan removed; however, fewer total swans would need to be removed using this strategy due to density - dependent productivity and potential immi gration from other established populations (i.e., Ontario). A management strategy such as this would temporarily leave breeding pairs intact in areas such as southeast Michigan where breeding productivity is likely low and competition for nesting territori es are high. Additionally, developed inland lakes and wetland areas of southeast Michigan are likely not of highest conservation priority for other wetland - dependent species at this time. However, c ulling mute swans in high - quality wetlands and stopover ar eas used by migrating waterfowl in these regions should remain a priority under this strategy. A spatially stratified culling approach such as this , which temporarily leaves breeding mute swans in high - density areas of southeast Michigan could 151 also serve a s a physical buffer to dispersing swans from adjacent regions where mute swans are present but widespread management programs are not being performed (i.e., Ontario). 152 LITERATURE CITED 153 LITERATURE CITE D Allin, C., G. Chasko, and T. P. Husband. 1987. Mute Swans in the Atlantic flyway: a review of the history, population growth and management needs. Northeast Section of The Wildlife Society 44:32 - 42. Arsnoe, D., and A. Duffiney. 2 018. From beauty to beast. Pages 40 - 44 in The Wildlife Professional 12:3. The Wildlife Society Bethesda, MD. Bomford, M., and P. O'Brien. 1995. Eradication or control for vertebrate pests? Wildlife Society Bulletin 23:249 - 255. Edelaar, P., and J. L. Tella. 2012. Managing non - native species: don't wait until their impacts are proven. Ibis 154:635 - 637. Ellis, M. M., and C. S. Elphick. 2007. Using a stochastic model to examine the ecological, economic and ethical consequences of population control in a charismatic invasive species: Mute Swans in North America. Journal of Applied Ecology 44:312 - 322. Jager, C., M. P. Nelson, L. Goralnik, and M. L. Gore. 2016. Michigan Mute Swan management: a case study to understand contentious natural resource management issues. Human Dimensions of Wildlife 21:189 - 202. Mack, R. N., D. Simberloff, W. Mark Lonsdale, H. Evans, M. Clout, and F. A. Bazzaz. 2000. Biotic invasions: causes, epidemiology, global consequences, and control. Ecological applications 10:689 - 710. Mark s, D. R. 2015. Management of the invasive Mute Swan in Michigan. Lansing, Michigan, USA. Michigan Department of Natural Resources. 2012. Mute Swan management and control program policy and procedures. < www.michigan.gov/documents/dnr/2012_Mute_Swan_Policy_378701_7.pdf >. Accessed 01 March 2016. Reiter, D. K., M. W. Brunson, and R. H. Schmidt. 1999. Public attitudes toward wildlife damage management and policy. Wildlife Society Bulletin:746 - 758. Sullivan, B. L., C. L. Wood, M. J. Iliff, R. E. Bonney, D. Fink, and S. Kelling. 2009. eBird: a citizen - based bird observation network in the biological sciences. Biological Conservation 142:2282 - 2292 . Watola, G. V., D. A. Stone, G. C. Smith, G. J. Forrester, A. E. Coleman, J. T. Coleman, M. J. Goulding, K. A. Robinson, and T. P. Milsom. 2003. Analyses of two Mute Swan 154 populations and the effects of clutch reduction: implications for population management. Journal of Applied Ecology 40:565 - 579.