WATERFOWL USE AND HUNTER SUCCESS ON MANAGED WATERFOWL AREAS IN MICHIGAN By Herman David McClinton III A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Fisheries and Wildlife – Master of Science 2021 ABSTRACT WATERFOWL USE AND HUNTER SUCCESS ON MANAGED WATERFOWL AREAS IN MICHIGAN By Herman David McClinton III Michigan is located at the center of the Great Lakes Region that supports more than 3 million autumn waterfowl migrants annually. Beginning in the 1940s and 1950s, the Michigan Department of Natural Resources (DNR) and the United States Fish and Wildlife Service (USFWS, hereafter) began consolidating large wetland complexes in Michigan’s Lower Peninsula. Since the 1970s, Michigan DNR and USFWS staff have maintained records on autumn waterfowl use, as well as harvest associated with managed hunting programs. I analyzed archived count data in conjunction with various measures of climate, weather, hydrology, and stock. I determined that the total number of ducks that these areas support in a given year have largely been in decline since the early 1990s. I analyzed timings of species-specific waterfowl abundance on areas and observed trends in seasonal timings. Finally, I measured changes in relative abundance of mallards as a function of weather variable. In my second chapter, I determined the influence of time of day, season progression, habitat type, and disturbance levels on a measure of habitat selection, using a novel method for surveying waterfowl. In my third chapter, I used archived data dating back to the 1970s to evaluate annual and seasonal measures of hunting program success. These results provide insights on multiple levels of waterfowl use and associated recreation and will help inform future management on the study sites. ACKNOWLEDGEMENTS First and foremost, I would like to thank my co-advisors Drs. David Luukkonen and Daniel Hayes. Dave, thank you for believing in my potential as a scientist and recruiting me to this project at Michigan State. Additionally, thank you for your mentorship on the world of waterfowl research and management. Dan, thank you for consistently being there at the drop of a hat to answer my various “philosophical” questions and for always encouraging my scheming (hunting or otherwise). Next, thank you to Dr. Charles Nelson for serving on my guidance committee. Also, thank you for pushing me to always consider the human component of natural resource management, as well as the insights that history can provide. Additionally, thank you to all three for putting up with my stubbornness. Thank you to my funding sources; Michigan Department of Natural Resources, Michigan State University, and the US Fish and Wildlife Service through the Pittman-Robertson Wildlife Restoration Act Grant MI W-155-R. Also, thank you to the MSU Fisheries and Wildlife Graduate Students Organization, the Green Creek Wildlife Society, Safari Club International – Michigan Involvement Committee, and The Wildlife Society – Wetlands Working Group for provide additional financial assistance along the way. I am thankful for the support I received from Michigan DNR and USFWS field staff at my study sites. Particularly, Pat Brickel, Zach Cooley, John Darling, Eric Dunton, Brandy Dybas-Berger, Tammy Giroux, Jeremiah Hiese, Nik Kalejs, Don Poppe, Mike Richardson, and Adam Shook who helped me track down various bits of information. Other Michigan DNR folks who deserve specific mentioning include Melissa Nichols, Steve Beyer, and Barb Avers. Melissa and Steve thanks for all the help with logistic coordination for this project and for suffering through my camera trap difficulties with me. Barb, thank you iii for always being willing to chat and for letting me sit in on meetings with the Waterfowl Working Group, Managed Waterfowl Hunt Area Staff, and Citizens Waterfowl Advisory Committee. I am also incredibly grateful for the seven research technicians who worked on various aspects of my project: Therin Bradshaw, Molly Engelman, Tim Hawley, Hannah Landwerlen, Connor Smeader, Evan Sluja, and Julia Whyte. My thesis would have been impossible to finish if y’all hadn’t time and energy y’all contributed to help look through those 1.4 million camera trap photos. A special thanks goes out to the Park Lake Posse, the PHLRM Lab, and all other great friends I have made at Michigan State. Finally, thank you to all my family and friends back home in Texas for supporting me and my desire to travel across the country chasing ducks. iv TABLE OF CONTENTS LIST OF TABLES ...................................................................................................................... vii LIST OF FIGURES .................................................................................................................... xii CHAPTER 1: AN ASSESSMENT OF WATERFOWL MIGRATION IN MICHIGAN .......1 INTRODUCTION..............................................................................................................1 METHODS .........................................................................................................................4 Study areas .............................................................................................................4 Waterfowl abundance data ...................................................................................6 Climate, weather, and population data ................................................................7 Data analyses ..........................................................................................................8 Total autumn duck abundance .....................................................................8 Migration timing ........................................................................................11 Season progression and weather influence on mallard abundance ..........13 RESULTS .........................................................................................................................15 Total autumn duck abundance ...........................................................................15 Migration timing ..................................................................................................17 Season progression and weather influence on mallard abundance.................27 DISCUSSION ...................................................................................................................30 Total autumn duck abundance ...........................................................................30 Migration timing ..................................................................................................33 Season progression and weather influence on mallard abundance.................35 MANAGEMENT IMPLICATIONS ..............................................................................37 APPENDICES ..................................................................................................................41 APPENDIX 1.1: SPECIES-SPECIFIC PEAK ABUNDANCE DATES .........42 APPENDIX 1.2: AVERAGE SPECIES-SPECIFIC PEAK ABUNDANCES .... ................................................................................................................................51 APPENDIX 1.3: MALLARD PROPORTION OF CORE PERIOD TOTALS ................................................................................................................................52 LITERATURE CITED ...................................................................................................53 CHAPTER 2: AUTUMN WATERFOWL USE OF INTENSELY MANAGED HUNTING AREAS IN SOUTHEASTERN MICHIGAN ............................................................................60 INTRODUCTION............................................................................................................60 METHODS .......................................................................................................................64 Study areas ...........................................................................................................64 Habitat classification and field methods ............................................................67 Disturbance level classification ...........................................................................69 Photo processing...................................................................................................70 Data work and analyses .......................................................................................71 Species-specific use ....................................................................................72 Seasonal nocturnal and diurnal use ..........................................................73 v RESULTS .........................................................................................................................75 Species-specific use...............................................................................................76 Seasonal nocturnal and diurnal use ...................................................................88 Disturbance levels ......................................................................................88 Habitat strata .............................................................................................93 DISCUSSION ...................................................................................................................97 Species-specific use...............................................................................................97 Seasonal nocturnal and diurnal use .................................................................100 Disturbance levels ....................................................................................100 Habitat strata ...........................................................................................102 Camera traps ......................................................................................................102 MANAGEMENT IMPLICATIONS ............................................................................105 APPENDIX .....................................................................................................................109 LITERATURE CITED .................................................................................................112 CHAPTER 3: SUCCESS TRENDS AND CONTRIBUTIONS OF MANAGED WATEFOWL HUNTING AREAS TO STATEWIDE DUCK HARVESTS IN MICHIGAN ......................................................................................................................................................119 INTRODUCTION..........................................................................................................119 METHODS .....................................................................................................................122 Study areas .........................................................................................................122 Data collection and organization ......................................................................125 Data analyses ......................................................................................................125 Annual use and success ............................................................................125 Seasonal success ......................................................................................126 Managed waterfowl hunt areas harvest contribution ..............................129 RESULTS .......................................................................................................................129 Annual use and success ......................................................................................129 Seasonal success .................................................................................................135 Managed waterfowl hunt areas harvest contribution ....................................141 DISCUSSION .................................................................................................................142 Annual use and success ......................................................................................143 Hunter use ................................................................................................143 Annual harvest totals and success rates ..................................................144 Seasonal success .................................................................................................146 Managed waterfowl hunt areas harvest contribution ....................................148 Other considerations ..........................................................................................148 MANAGEMENT IMPLICATIONS ............................................................................149 APPENDICES ................................................................................................................153 APPENDIX 3.1: ANNUAL TOTALS ..............................................................154 APPENDIX 3.2: MANAGED WATERFOWL HUNT AREAS MAPS ........157 LITERATURE CITED .................................................................................................164 vi LIST OF TABLES Table 1.1. Main effect, additive, and interacting models explaining total annual DUDs. Area = study site where observations took place. Year = numerical linear trend or annual factor. MI Mallard = Michigan breeding mallard population estimate for a respective year. Fall Palmer Z Index (FPI) = average Palmer Z value for the months of September through December for a respective year. Lake = deviation from average lake level for the lake adjacent to or most interacting with a respective Area. NAO = numerical NAO index for a respective year, ENSO = factor variable of the annual classification of ONI ........................................................................10 Table 1.2. Main effect, additive, and interactive models explaining timing of peak species- specific dabbling duck abundance in Michigan. Area = study site where observations took place. Year = numerical linear trend or annual factor. Year2 = quadratic trend. Block = three-year block. AO = numerical AO index for a respective year. NAO = numerical NAO index for a respective year. ENSO = factor variable of the annual classification of ONI ...............................12 Table 1.3. Description of weather variables used to explain rate of change (r) in duck abundance on eight wetland complexes in Michigan’s lower peninsula .........................................................14 Table 1.4. Top 8 performing models among 36 models considered in explaining total annual DUD. Bold values are metrics for the top performing model. K = number of parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value .......................................................................15 Table 1.5. Top 3 performing models for each species explaining peak abundance timing. Bold values are those of the top performing model. Area = study location, Block = 3-year time, Year = linear time trend, Year2 = curvilinear time trend, AO = Arctic Oscillation, NAO = North Atlantic Oscillation, ENSO = El Niño Southern Oscillation. K = number of model parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value .......................................................................17 Table 1.6. Top 8 models for explaining change in relative abundance of mallards. Bold values indicate that they are representative of the top performing model. TEMP = maximum value of - (daily temperature) between survey periods, Julian = numerical day of year, MEANTemp = mean temperature between survey periods, Cumulative WSI = maximum value for TEMP + TEMPDAYS + SNOW + SNOWDAYS between survey periods, WSIMean = maximum value for MEANTemp + TEMPDAYS + SNOW + SNOWDAYS between survey periods. K = number of model parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value ..........................27 vii Table A1.1.1. Numerical day of year of peak mallard abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ...........................................................................................42 Table A1.1.2. Numerical day of year of peak American black duck abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ............................................................................43 Table A1.1.3. Numerical day of year of peak wood duck abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ...........................................................................................44 Table A1.1.4. Numerical day of year of peak American wigeon abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ............................................................................45 Table A1.1.5. Numerical day of year of peak gadwall abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ...........................................................................................46 Table A1.1.6. Numerical day of year of peak northern pintail abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ...........................................................................................47 Table A1.1.7. Numerical day of year of peak northern shoveler abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ............................................................................48 viii Table A1.1.8. Numerical day of year of peak blue-winged teal abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ............................................................................49 Table A1.1.9. Numerical day of year of peak green-winged teal abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ............................................................................50 Table A.1.2.1. Average peak group abundances observed (2006 – 2019) ....................................51 Table 2.1. Managed Waterfowl Hunt Areas and Shiawassee NWR property size and refuge size, as well as area of natural wetlands and cultivated fields flooded for waterfowl. FP = Fish Point, HI = Harsens Island, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. .......................................................................................66 Table 2.2. MWHAs hunt schedule during the two study seasons .................................................66 Table 2.3. Total camera allocation across the 2018 and 2019 field seasons by habitat stratum in location classified as refuge and those open to hunting .................................................................68 Table 2.4. Disturbance level classification of 2018 and 2019 camera trap locations. Refuge location were closed to hunting for the entirety of the waterfowl season. Low disturbance level areas were hunted on average 1 – 4 periods a week. Moderate disturbance level areas were hunted on average 5 – 8 periods a week. High disturbance level areas were hunted on average 9 – 14 periods a week...........................................................................................................................69 Table 2.5. Additive and interacting conditional models explaining overall species/group specific habitat use on an individuals per day scale. The zero-inflation model components set as null (~1) or as identical to the conditional component and both linear and quadratic parameterizations of the negative binomial data distribution family. Habitat Stratum is a factor variable with 4 levels (CUL, CMS, MMS, and OAB). Hunted is a factor variable with 2 levels (Hunted and Not Hunted) ......................................................................................................................................... 72 Table 2.6. Additive and interacting conditional models explaining seasonal progression of diurnal and nocturnal duck and goose habitat use as a function differing disturbance levels. The zero-inflation model components set as null (~1) or as identical to the conditional component and both linear and quadratic parameterizations of the negative binomial data distribution family. DN is a factor variable with 2 levels (Day and Night). Disturbance Level is a factor variable with 4 levels (Refuge, Low, Moderate, and High). Period is a factor variable with 10 levels (Pre-season, Season Weeks 1–8, and Post Season) .......................................................................................... 74 ix Table 2.7. Additive and interacting conditional models explaining seasonal progression of diurnal and nocturnal duck and goose habitat use as a function differing habitat types. The zero- inflation model components set as null (~1) or as identical to the conditional component and both linear and quadratic parameterizations of the negative binomial data distribution family. DN is a factor variable with 2 levels (Day and Night). Habitat Stratum is a factor variable with 4 levels (CUL, CMS, MMS, OAB). Period is a factor variable with 10 levels (Pre-season, Season Weeks 1–8, and Post Season) ................................................................................................................... 74 Table 2.8. Group and species breakdown of waterfowl observed across both field seasons ........75 Table 2.9. Species/group (species alpha codes used by United States Geological Survey’s Bird Banding Laboratory) AIC values for habitat use models that converged without warning. Bolded values indicate best model fit. WAR indicates the model had convergence issues ........................77 Table 2.10. Disturbance level models with associated conditional variables and AIC values when explaining ducks and geese observed per day. WAR indicates that a given model had convergence issues. Bolded AIC values indicates the best fitting model that was used to generate observation estimates .....................................................................................................................88 Table 2.11. Habitat stratum models with associated conditional variables and AIC values when explaining ducks and geese observed per day. WAR indicates that a given model had convergence issues. Bolded AIC values indicates the best fitting model that was used to generate observation estimates .....................................................................................................................93 Table. 3.1. Managed Waterfowl Hunt Areas’ property characteristics and available hunting opportunity ...................................................................................................................................124 Table 3.2. Classification of Managed Waterfowl Hunt Areas. Duck = areas with predominately flooded hunting zones where hunters primarily shoot ducks. Both = areas with predominately flooded hunting zones where hunters harvest large numbers of ducks and geese. Goose = areas with predominately dry hunting zones where hunters primarily shoot geese ..............................127 Table 3.3. Models describing seasonal progression of ducks harvested at select Managed Waterfowl Hunt Areas (1997 – 2019). Bold values are metrics for the top performing model. K = number of parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value, ω = respective model weight, Cumulative ω = weight of respective and all better fitting models, LL = log- likelihood .....................................................................................................................................135 Table 3.4. Models describing seasonal progression of ducks harvested per hunter trip at select Managed Waterfowl Hunt Areas (1997 – 2019). Bold values are metrics for the top performing model. K = number of parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value, ω = respective model weight, Cumulative ω = weight of respective and all better fitting models, LL = log-likelihood ............................................................................................................................137 x Table 3.5. Models describing seasonal progression of geese harvested at select Managed Waterfowl Hunt Areas (1997 – 2019). K = number of parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value, ω = respective model weight, Cumulative ω = weight of respective and all better fitting models, LL = log-likelihood ..............................................................................139 Table 3.6. Models describing seasonal progression of geese harvested per hunter trip at select Managed Waterfowl Hunt Areas (1997 – 2019). K = number of parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value, ω = respective model weight, Cumulative ω = weight of respective and all better fitting models, LL = log-likelihood ......................................................140 xi LIST OF FIGURES Figure 1.1. Locations of the eight wetland complexes in Michigan’s lower peninsula serving as study areas for this research .............................................................................................................5 Figure 1.2. Estimates from the top DUD model (Area * Year). AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. Estimates are not portrayed for Allegan (AL) and Shiawassee SGA (SH) prior to 1998 and 2000, respectively, due to lack of usable data ......................................................................................................................................16 Figure 1.3. Expected change in an area’s respective annual DUDs when considering the added effect of lake levels ........................................................................................................................17 Figure 1.4. Model estimates of peak abundance timing for mallards (Area + Year Block model). AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ........19 Figure 1.5. Model estimates of peak abundance timing for American black ducks (Area + Year Block model). AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ..............................................................................................................................20 Figure 1.6. Model estimates of peak abundance timing for wood ducks (Area * Year model). AL = Allegan, FP = Fish Point, HI = Harsens Island, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge .............................21 Figure 1.7. Model estimates of peak abundance timing for American wigeon (Year2 model) .....22 Figure 1.8. Model estimates of peak abundance timing for gadwall (Area + Year2 model). AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ..................23 Figure 1.9. Model estimates of peak abundance timing for northern pintails (Year2 model) ........24 Figure 1.10. Model estimates of peak abundance timing for northern shovelers (Area + Year2 model). FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ..................25 Figure 1.11. Model estimates of peak abundance timing for blue-winged teal (Area * Year model). FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ..................26 xii Figure 1.12. Model estimates of peak abundance timing for green-winged teal (Area * AO model). AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ..............................................................................................................................27 Figure 1.13. Model output from the best fitting mixed-effects model explaining change in mallard relative abundance (ln(Nt)-ln(Nt-1)) along with individual point observations. – (Temperature) is the inverse of ambient temperatures in degrees Celsius because below freezing is more severe ................................................................................................................................28 Figure 1.14. Outputs from the Week Number * Area model with a loess smoother describing seasonal mallard abundance on the managed areas. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge ............................................................................29 Figure A1.3.1. Area specific estimates (and 95% confidence interval) of the proportion of mallards in annual core period (numerical weeks 40 – 51 of the year) duck totals. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge .............................52 Figure 2.1. Locations of the six wetland complexes in southeastern Michigan serving as study areas for this research.....................................................................................................................65 Figure 2.2. Estimated marginal means and confidence intervals of diurnal mallard observations (Model 5). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water .......................................................................................78 Figure 2.3. Estimated marginal means and confidence intervals of diurnal Canada goose observations (Model 3). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water ...................................................79 Figure 2.4. Estimated marginal means and confidence intervals of diurnal American black duck observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water ...................................................80 Figure 2.5. Estimated marginal means and confidence intervals of diurnal American green- winged teal observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water ...................................................81 Figure 2.6. Estimated marginal means and confidence intervals of diurnal Northern pintail observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water ...................................................82 Figure 2.7. Estimated marginal means and confidence intervals of diurnal American wigeon observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water ...................................................83 xiii Figure 2.8. Estimated marginal means and confidence intervals of diurnal diving duck observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water ...................................................84 Figure 2.9. Estimated marginal means and confidence intervals of diurnal wood duck observations (Model 6). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water ...................................................85 Figure 2.10. Estimated marginal means and confidence intervals of diurnal gadwall observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water .......................................................................................86 Figure 2.11. Estimated marginal means and confidence intervals of diurnal blue-winged teal observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water ...................................................87 Figure 2.12. Estimated marginal means and confidence intervals of diurnal Northern shoveler observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water ...................................................88 Figure 2.13. Estimated marginal means of seasonal diurnal duck observations per day across varying disturbance levels (Model 6). Refuge = location closed to hunting, Low = location hunted on average 1–4 times a week, Moderate = location hunted on average 5–8 times a week, High = location hunted on average 9–14 times a week. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season ...................................................................................................................90 Figure 2.14. Estimated marginal means of seasonal nocturnal duck observations per night across varying disturbance levels (Model 6). Refuge = location closed to hunting, Low = location hunted on average 1–4 times a week, Moderate = location hunted on average 5–8 times a week, High = location hunted on average 9–14 times a week. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season ...................................................................................................................91 Figure 2.15. Estimated marginal means of seasonal diurnal goose observations per day across varying disturbance levels (Model 1). Refuge = location closed to hunting, Low = location hunted on average 1–4 times a week, Moderate = location hunted on average 5–8 times a week, High = location hunted on average 9–14 times a week. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season ...................................................................................................................92 xiv Figure 2.16. Estimated marginal means of seasonal nocturnal goose observations per night across varying disturbance levels (Model 1). Refuge = location closed to hunting, Low = location hunted on average 1–4 times a week, Moderate = location hunted on average 5–8 times a week, High = location hunted on average 9–14 times a week. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season ...................................................................................................................93 Figure 2.17. Estimated marginal means of seasonal diurnal duck observations across varying strata (Model 3). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season ................................................................................................................................94 Figure 2.18. Estimated marginal means of seasonal nocturnal duck observations per day across varying strata (Model 3). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season ...................................................................................................................95 Figure 2.19. Estimated marginal means of seasonal diurnal goose observations per day across varying strata (Model 5). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season ...................................................................................................................96 Figure 2.20. Estimated marginal means of seasonal nocturnal goose observations per day across varying strata (Model 5). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season ...................................................................................................................97 Figure 3.1. Locations of the seven Managed Waterfowl Hunt Areas in Michigan’s lower peninsula ......................................................................................................................................123 Figure 3.2. Area specific hunter trip totals (1974–2019). AL = Fennville Farm Unit, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee ..........................................................................................................................130 Figure 3.3. Annual Managed Waterfowl Hunt Area duck harvest totals (1974–2019). AL = Fennville Farm Unit, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee ..........................................................................131 Figure 3.4. Annual Managed Waterfowl Hunt Area goose harvest totals (1974–2019). AL = Fennville Farm Unit, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee ..........................................................................132 xv Figure 3.5. Annual measures and linear trends in ducks harvested per hunter trip on Managed Waterfowl Hunt Areas (1974–1996). AL = Fennville Farm Unit, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee ..133 Figure 3.6. Annual measures and linear trends in ducks harvested per hunter trip on Managed Waterfowl Hunt Areas (1997–2019). AL = Fennville Farm Unit, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee ..134 Figure 3.7. Annual measures and linear trends in geese harvested per hunter trip on Managed Waterfowl Hunt Areas (1974–2019). AL = Fennville Farm Unit, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee ..135 Figure 3.8. Weekly estimates (Area * Season Week(factor)) of duck harvest at select Managed Waterfowl Hunt Areas (1997 – 2019). FP = Fish Point, HI = Harsens Island, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee ..........................................................................137 Figure 3.9. Model estimates (Area * Season Week3) of seasonal ducks harvested per hunter trip at select Managed Waterfowl Hunt Areas (1997 – 2019). FP = Fish Point, HI = Harsens Island, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee ............................................138 Figure 3.10. Weekly estimates (Area * Week Number(factor)) of goose harvest at select Managed Waterfowl Hunt Areas (1997 – 2019). AL = Fennville Farm Unit, FP = Fish Point, MU = Muskegon, SH = Shiawassee ....................................................................................................140 Figure 3.11. Model estimates (Area * Week Number3) of seasonal geese harvested per hunter trip at select Managed Waterfowl Hunt Areas (1997 – 2019). AL = Fennville Farm Unit, FP = Fish Point, MU = Muskegon, SH = Shiawassee ..................................................................................141 Figure 3.12. The proportion of Michigan’s annual state harvest totals taken on a Managed Waterfowl Hunt Area (1997–2019) .............................................................................................142 Figure A.3.1.1. Total annual hunter trips on Managed Waterfowl Hunt Areas (1974 – 2019). Data available for all areas beginning in 1985 .............................................................................154 Figure A.3.1.2. Total annual duck harvest on Managed Waterfowl Hunt Areas (1974 – 2019). Data available for all areas beginning in 1985 .............................................................................155 Figure A.3.1.3. Total annual goose harvest on Managed Waterfowl Hunt Areas (1974 – 2019). Data available for all areas beginning in 1985 .............................................................................156 Figure A.3.2.1. Fennville Farm Unit managed hunting map .......................................................157 Figure A.3.2.2. Fish Point managed hunting map .......................................................................158 Figure A.3.2.3. Harsens Island Unit managed hunting map ........................................................159 xvi Figure A.3.2.4. Muskegon managed hunting map .......................................................................160 Figure A.3.2.5. Nayanquing Point managed hunting map ...........................................................161 Figure A.3.2.6. Pointe Mouillee managed hunting map ..............................................................162 Figure A.3.2.7. Shiawassee managed hunting map .....................................................................163 xvii CHAPTER 1. AN ASSESSMENT OF WATERFOWL MIGRATION IN MICHIGAN INTRODUCTION Bird migration is characterized by long distance movements of individuals to specific locations at roughly the same times each year (Newton 2008a). In North America, efforts to understand these movements began when John James Audubon, who is credited as the first to “band” birds, tied threads to nestling Eastern phoebes in 1803 and Leon J. Cole is credited with having introduced systematic bird banding as a way to track individual movements in 1901 (Wood 1945). Due to their status as game species, waterfowl became a focus for banding pioneers like Jack Miner (Wood 1945) and projects by the Bureau of Biological Survey (U.S. Fish and Wildlife Service, today; Lincoln 1935). These works helped inform the conceptualization of four distinct North American waterfowl flyways (Lincoln 1935) and finer migratory corridors within the larger flyways (Bellrose 1968). In addition to corridors, staging areas are characterized as perennially available habitats that provide an abundance of available food and space where large flocks of waterfowl can assemble before making migratory movements (Baldassarre and Bolen 2006a) to wintering areas, and are of major ecological importance. Situated along the upper Mississippi and Atlantic flyways, the Great Lakes Region serves as a major migratory funnel with approximately 3 million waterfowl migrating through on an annual basis (Bookhout et al. 1989). Historically waterfowl used the vast staging areas inland and along coastal areas such as Saginaw Bay, Lake St. Clair, and western Lake Erie. Just as there has been great interest in understanding where birds migrate to, understanding the interacting endogenous and exogenous cues that trigger migratory behavior has also been of interest. Birds fall on a spectrum of obligate to facultative migrants. Obligate migrants are those where endogenous cues triggered by photoperiod (i.e., daylength) drive the decision to migrate, while facultative migrants are those who respond to prevailing conditions such as food availability (Newton 2008b). Obligate migrants generally have a high degree of consistency with regard to the annual timing of migrations and tend to be long distance migrants while facultative migrants have a high degree of variability with regard to timing of departure and distance traveled (Newton 2008b). The life history traits of an individual species influence how it reacts to the interacting migratory cues. As with many areas in North America, the Great Lakes Region has experienced major wetland loss threatening regional availability of staging areas for the millions of migrants. In Canada, southern Ontario had approximately 20,266 km2 of wetlands prior to European settlement (Eamer et al. 2010). On the United States’ side, considerable wetland acreage historically existed in Minnesota, Wisconsin, Michigan, and northwestern Ohio (Kaatz 1955, Tiner 1984). By the 1980s, Minnesota and Wisconsin had lost 53 and 32 percent of their respective historic wetland acreage (Tiner 1984), Michigan had lost greater than 50 percent of its wetland acreage (Tiner 1984, Dahl and Allord 1996), and the Great Black Swamp is largely no more (Kaatz 1955). Southern Ontario had lost more than 60 percent of its wetlands (Snell 1987) including 35 percent of its acreage along the shores of lakes Erie, Ontario, and St. Clair (McCullough 1985). In addition to loss of wetland acreage, the habitat conditions in many wetlands have changed over time. The spread of invasive species such narrowleaf cattail (Typha angustifolia), common reed (Phragmites australis), and reed canary grass (Phalaris arundinacea) can lead to extensive monocultures that drastically reduce the quality of remaining 2 wetlands (Baldassarre and Bolen 2006b) diminishing the capacity of the region to support large numbers of migrating waterfowl. Furthermore, the general warming associated with global climate change has been correlated with changes in bird migratory phenology (Cotton 2003). Climate indices such as Arctic Oscillation (AO), North Atlantic Oscillation (NAO), and El Niño Southern Oscillation (ENSO) are commonly used to track climate variation in North American. Negative AO and NAO values have both been associated with more severe winters in the eastern United States (Hurrell 1995, Serreze and Barry 2014), while extremes in the El Niño (warm phase) and La Niña (cool phase) of the Southern Oscillation have been associated with regional shifts in precipitation and temperature patterns (Ropelewske and Halpert 1986, Dracup and Kahya 1994). These indices correlate with spring (Rainio et al. 2006) and autumn (Schummer et al. 2014, Thurber et al. 2020) migratory timings in waterfowl. The effect of these climate indices varies across species, however. Species such as blue-winged teal (Spatula discors) that are obligate in their migration timing have shown less response to weather as their decision to migrate is largely dependent on photoperiod. Mallards (Anas platyrhynchos) however, tend to exhibit far more facultative weather driven characteristics (Schummer et al. 2010, Dalby et al. 2013, Van Den Elsen 2016). Other dabbling ducks largely fall between these two on the spectrum (Baldassarre 2014). Due to the varying degree of facultative migratory strategies, all waterfowl are subject to a degree of influence of climate change with respect to their migratory timing. Delays in autumn waterfowl migratory timing have already been documented both in Europe (Lehikoinen and Jaatinen 2012) and in eastern North America (Thurber et al. 2020) with further delays in the Great Lakes region being predicted by the mid-21st century (Notaro et al. 2016). 3 Due to loss of wetlands, the Michigan Department of Natural Resources (DNR) and the United States Fish and Wildlife Service (USFWS, hereafter) began acquiring land in large wetland complexes in Michigan’s Lower Peninsula during the 1940s and 1950s. These areas were already significant staging areas for migrating waterfowl, but implementation of intense management regimes increased their capacity to support migrating waterfowl. Continued loss and degradation (Dahl 2011) of adjacent nonprotected wetlands further magnified their importance as staging areas. Shifts in migration timing (Thurber et al. 2020), the observed overwintering of late migrant species (Reed 1971, Brook et al. 2009), and predicted expansion of overwintering by species that historically exhibited middle to early migratory characteristics (Notaro et al. 2016) raises questions about potential mismatches between timing of food availability and regional capacity to support overwintering birds. As such, the objective for this chapter was to investigate waterfowl use of key staging areas in Michigan, with a particular focus on examining (1) trends in total number of ducks using key staging areas during the fall migratory period (2) potential shifts in species specific peak abundance timing and (3) the influence of season progression and local weather conditions on abundances of a key species, mallard. Study areas METHODS The areas of interest for this research (Fig. 1.1) include the Fennville Farm Unit of the Allegan State Game Area (Allegan, hereafter), Fish Point State Wildlife Area (Fish Point, hereafter), the Muskegon County Wastewater Treatment Plant (Muskegon, hereafter), Nayanquing Point State Wildlife Area (Nayanquing, hereafter), Pointe Mouillee State Game Area (Pointe Mouillee, hereafter), the Shiawassee National Wildlife Refuge (Shiawassee NWR, 4 hereafter), Shiawassee River State Game Area (Shiawassee SGA, hereafter), and the Harsens Island Unit of the St. Clair Flats State Wildlife Area (Harsens Island, hereafter). Figure 1.1. Locations of the eight wetland complexes in Michigan’s lower peninsula serving as study areas for this research. These areas are collectively referred to as Managed Waterfowl Hunting Areas (MWHAs) by the DNR. They are managed to support a suite of habitat types (e.g., emergent marsh, shrub- scrub swamps, aquatic bed, moist soil management units, flooded agriculture units, etc.), with the intent of supporting a variety of migrating waterfowl species. Given their importance, they 5 have also become centers of intense active wetland management in southern Michigan with an emphasis on attracting and holding waterfowl during autumn. This management consists of water level manipulation during the growing season to promote the prevalence of beneficial wetland plant species (e.g., barnyard grass, smartweeds, nutsedges, etc.) to be flooded in the fall along with planted agriculture grains (corn, buck wheat, Japanese millet, etc.), as well as mitigating the spread of non-desirable invasive species (phragmites reed, narrow-leaf cattail, flowering rush, etc.). Additionally, these areas have designated places of spatial refuge and established hunting plans tailored to each property to maximize quality recreational hunting opportunity while maintaining sizeable waterfowl populations during fall migration. Waterfowl abundance data Managers of these complexes started conducting regular surveys of autumn migrating waterfowl abundances in the 1970s. These surveys generally consist of weekly perimeter ground counts of refuge units. While detection is not perfect on these counts due to high numbers of mingling waterfowl and visual obstruction associated with wetland vegetation, they provide an index of species-specific use. I acquired these count data from Michigan DNR and US Fish and Wildlife Service collaborators and consolidated them into a single dataset. I addressed inconsistences (e.g., date format, week number, species-specific abundances, etc.) through communications with area managers and by referring to annual reports. I standardized each week to start on Monday and ending on Sunday in accordance with international standard. I also created a survey date column depicting the numerical day of the year (Julian date hereafter) on which the survey was completed. These data were not always included within the original datasets but were usually available through discussions with area managers or by referring to 6 archived annual reports. In instances where I could not determine the specific date of the survey, I assigned the Wednesday of each week to be the standard. Climate, weather, and population data I obtained monthly AO (National Oceanic and Atmopheric Administration (NOAA) 2019a), NAO (National Oceanic and Atmopheric Administration (NOAA) 2019b)), and ENSO via the Oceanic Nino Index (ONI;(National Oceanic and Atmopheric Administration (NOAA) 2019c)) from the National Oceanic and Atmospheric Administration (NOAA hereafter). I obtained the yearly AO and NAO values by averaging the October, November, and December monthly values of each respective index. Following the lead of Schummer et al. (2014) and Thurber et al. (2020), I reclassified ONI to represent La Niña and El Niño events. A La Niña event was any year where five consecutive three-month overlapping seasons (June–August– September through January–February–March) had an ONI of ≤ -0.5. Years where there were 5 consecutive three-month periods with an ONI of ≥ -0.5 were classified as El Niño (Tozuka et al. 2005). Neutral years were those where neither of these occurred. For measures of hydrology, I obtained annual average lake levels in meters for Lake Erie, Lake Michigan/Huron, and Lake St. Clair (National Oceanic and Atmopheric Administration (NOAA) 2019d). I centered lake levels relative to their mean level over the study period to capture varying coastal wetland conditions. Positive values indicate higher lake levels than normal. I also obtained autumn (September – December) Palmer Z Index values (NOAA National Centers for Environmental Information 2019) for the state of Michigan. The Palmer Z Index represents short-term drought intensity on a monthly scale, with negative values being associated with drier periods. These indices serve as characterizations of how much water is 7 present on a landscape and thus how reliant waterfowl might be on perennially available habitat associated with the MWHAs and the SNWR. For measures of weather, I obtained daily minimum and maximum temperatures (C), snowfall (cm) and snow depth (cm) at the respective county level for each of my areas of interest (National Oceanic and Atmopheric Administration (NOAA) 2020). I then transformed these data into a format similar to that of Schummer et al. (2010) and Van Den Elsen (2016). Finally, I obtained breeding mallard population estimates for Michigan and the Great Lakes Region to serve as an index of yearly stock for the autumn migratory period (U.S. Fish and Wildlife Service 2019a). Data analyses I analyzed measures of waterfowl use with sets of a priori candidate regression models that represented biologically plausible hypotheses for explaining waterfowl abundance. For all model sets, any continuous explanatory variables with a Pearson correlation coefficient of r ≥ 0.7 were not included in the same model to avoid collinearity issues (Dormann et al. 2013). Individual observations were only excluded if they were an entry error, incomplete observation, or both an outlier and of high leverage. Total autumn duck abundance I utilized data spanning the 28 years from 1992–2019. Though surveys were conducted prior to 1992, more sporadic existence of count data as well as my keen interest in regional breeding population estimates (Michigan surveys initiated in 1992) limited the capacity in which I could utilize data prior to this period. The total ducks observed at each site in a given year was used as an index of annual duck use days (DUDs, hereafter). DUDs are commonly used to track the productivity of a staging area through available food. To get an estimate of annual DUDs for 8 each respective area, I multiplied weekly survey observation totals by 7 to represent actual season totals. To minimize the influence of variability in the number of surveys conducted per year and the timing of surveys between areas, I established a “core survey period” that included week 40 through week 51. This was the twelve-week period where the matrix of data was most complete with consistent observations across areas and years. Although the exact dates associated with these weeks vary, this was generally the end of September through mid- December. In instance where multiple surveys were conducted during a single week, the average of the number of ducks observed for those respective surveys served as the count index for that week. Instances where a weekly survey did not take place, I used simple averaging imputation of weeks preceding and following to generate values. This accounted for only approximately 5% (133 of 2,352) of weekly survey observations. If a respective area’s surveys ended before the end of my core period, it was assumed that freeze up resulted in 0 or a negligible number of birds being present. If an area did not initiate their surveys until after the beginning of the core period, I dropped that year from analyses rather than using more complex data imputation methods. This was done 5 times out of 202 possible area and year combinations. Additionally, Allegan SGA data before 1998 were not included due to concerns that survey methodologies changed. Shiawassee River SGA data was not included prior to 2000, as this was when a surveyable refuge was established (though another refuge unit existed prior to this). Finally count data did not exist or was unobtainable for the following years: Allegan SGA (2003), Fish Point (2004), Harsens Island (1994, 1996, 1998), and the Shiawassee NWR (2009 – 2011). Altogether, the data set consisted of 196 observations of annual duck use day totals, with no area having fewer than 20 observations. 9 I created a candidate set of 37 a priori models based on plausible biological hypotheses (Table 1.1). Table 1.1. Main effect, additive, and interacting models explaining total annual DUDs. Area = study site where observations took place. Year = numerical linear trend or annual factor. MI Mallard = Michigan breeding mallard population estimate for a respective year. Fall Palmer Z Index (FPI) = average Palmer Z value for the months of September through December for a respective year. Lake = deviation from average lake level for the lake adjacent to or most interacting with a respective Area. NAO = numerical NAO index for a respective year, ENSO = factor variable of the annual classification of ONI. Main Effect Models Null Area Additive Models Area + Year Interaction Models Area * Year Area + Year (annual factor) Area * MI Mallard Area + MI Mallard Area + FPI Area + Lake Area + NAO Area + ENSO Area + NAO + ENSO Area + MI Mallard + Lake Area + MI Mallard + FPI Area + MI Mallard + NAO Area + MI Mallard + ENSO Area + MI Mallard + NAO + ENSO Area + Year + Lake Area * FPI Area * Lake Area * NAO Area * ENSO Area + NAO * ENSO Area * NAO + ENSO Area * NAO * ENSO Area * MI Mallard + Lake Area * MI Mallard + FPI Area * MI Mallard + NAO Area * MI Mallard + ENSO Area * MI Mallard + NAO + ENSO Area * MI Mallard + NAO * ENSO Area + MI Mallard + NAO * ENSO Area + MI Mallard * Lake Area * MI Mallard * Lake Area * Year + Lake Area + Year * Lake Area * Year * Lake All models (other than the null) within this candidate set included Area as an explanatory variable due to the known influence of refuge size and configuration on duck observations at my study sites. Even though total duck use days included all species, I included a regional measure of mallard breeding population due to their being by far the most prevalent species. Other variables that I considered for this analysis but did not include in the final candidate model set 10 were the Great Lakes Regional Mallard population estimate, the annual Palmer Z Index, and Arctic Oscillation. These variables were highly correlated (r ≥ 0.70) with other variables (Michigan Mallard population estimate, fall Palmer Z Index, and North Atlantic Oscillation, respectively) and were ultimately excluded to avoid the creation of an excessively large candidate set. I considered the DNR’s Michigan mallard breeding population estimate for this question because it has been shown that 68% of the mallards harvested in Michigan were produced in Michigan, with the other Great Lakes States contributing only an additional 7% (Arnold and de Sobrino, unpublished data). I considered fall Palmer Z Index because it better captures the hydrology during the period of interest than annual Palmer Z Index. I considered NAO, because it has been noted to possibly better represent northern hemisphere conditions (Ambaum et al. 2001) and has been employed others researching autumn waterfowl migration in adjacent regions (Thurber et al. 2020). Migration timing To investigate potential trends in migration timing of waterfowl species, I utilized 14 years of data spanning from 2006 – 2019. Although some species-specific abundance data were available (e.g., mallards and American black ducks) before this time period, 2006 was the first year surveys were standardized to species-specific counts for waterfowl across all areas. I used the day in which peak abundance for each species of dabbling duck was observed as the response variable for linear models, similar to Thurber et al. (2020). Few data were available for diving ducks and observations were sporadic except for select species on select areas (e.g., ruddy ducks (Oxyura jamaicensis) at Muskegon). This is largely due to the nature of the management being targeted towards providing habitat more conducive for dabbling ducks. As such, I only considered dabbling ducks for this objective. 11 Because the day of peak abundance is based on a maximum statistic, I wanted to ensure that estimates for each year would not be unduly influenced by random fluctuation for individual species at low abundance. As such, I required a site’s average peak abundance observed for a respective species across the 14 years of this data set to be ≥ 50 birds to be included in that species timing analyses. This resulted in the exclusion of peak timing data for American wigeon, blue-winged teal, northern pintail, and northern shoveler from Allegan and American wigeon, northern pintail, and wood ducks from the Muskegon WWTP from analyses. The candidate set of models applied to all species for explaining timing of peak abundance contained 31 a priori models (Table 1.2). Table 1.2. Main effect, additive, and interactive models explaining timing of peak species- specific dabbling duck abundance in Michigan. Area = study site where observations took place. Year = numerical linear trend or annual factor. Year2 = quadratic trend. Block = three-year block. AO = numerical AO index for a respective year. NAO = numerical NAO index for a respective year. ENSO = factor variable of the annual classification of ONI. Main Effect Models Additive Models Interaction Models Null Area Year Year2 Year (annual factor) Block AO NAO ENSO Area + Year Area + Year2 Area + Year (annual factor) Area + Block Area + AO Area + NAO Area + ENSO Area + AO + ENSO Area + NAO + ENSO AO + ENSO NAO + ENSO Area * Year Area * Block Area * AO Area * NAO Area * ENSO AO * ENSO NAO * ENSO Area + AO * ENSO Area + NAO * ENSO Area * AO * ENSO Area * NAO * ENSO Although they are correlated, AO and NAO were both included in this analyses because there is some uncertainty in which measure best captures climate anomalies shown to influence migratory movements in the Great Lakes Region (Schummer et al. 2014). 12 Season progression and weather influence on mallard abundance To observe how weather influenced within-year abundance of a key waterfowl species, I used the rate of change in relative abundance of mallards (r = ln(mallard abundancet) – ln(mallard abudnancet-1)) between two surveys (Schummer et al. 2010, Van Den Elsen 2016) in linear mixed-effects models. In addition, I used the observed abundance of mallards for a survey period as the response variable in generalized linear regression models. I chose mallards for this research question because they are the most numerous and consistently observed species on all study sites. Additionally, they are generally classified as facultative migrants that base migratory decisions primarily from weather conditions more so than more obligate migrants such as blue- winged teal. As with the peak timing models, I used the data spanning 14 years between 2006 and 2019. To account for variability in survey initiation date at respective sites, I standardized the survey matrix to only include surveys taking place between September 1st and January 15th. Because of incomplete survey effort, I removed 1, 2, 1, and 5 surveys from the Shiawassee NWR in 2006, 2007, 2008, and 2012, respectively. Even with the removal of these incomplete observations, ≥14 surveys were available each year for the Shiawassee NWR. I explored the change in relative abundance and actual abundance of mallards using a set (Table 1.3) of weather indices similar to those used by Schummer et al. (2010) and Van Den Elsen (2016), specific Area (random effect for the mixed effect models and fixed effect in all general linear regression models), and various measures of season progression that described curvilinear changes in abundance over time (i.e., Julian, Julian2, Julian3, Julian4, and Week Number). . I specified a Poisson distribution for my generalized linear models explaining abundance. 13 Table 1.3. Description of weather variables used to explain rate of change (r) in duck abundance on eight wetland complexes in Michigan’s lower peninsula. Variable TEMP TEMPDAYS SNOW SNOWDAYS Cumulative WSI TEMPMean WSIMean Julian Julian2 Week Numbera Julian3a Julian4a Areab Description –(daily average temperature) Consecutive days wherein TEMP is ≥ 0 (Snow depth in cm) * 0.394 Consecutive days where snow depth is ≥ 2.54 cm TEMP + TEMPDAYS + SNOW + SNOWDAYS Mean of daily average temperatures between surveys TEMPMean + TEMPDAYS + SNOW + SNOWDAYS Numerical day of year Quadratic trend on numerical day of year Numerical week of year 3rd Order Polynomial on numerical day of year 4th Order Polynomial on numerical day of year Study site where observations took place a Only considering in the candidate set explaining observed mallard abundance b Random effect for rate of change analyses and fixed effect for abundance analyses I downloaded weather data from multiple stations within each county. The station closest to my area of interest served as the primary station. I used methods comparable to Schummer et al. (2010) to addressing missing daily weather data. If large chunks of data were missing (i.e., > one week) I supplemented that county’s data frame with data gathered at another station within the respective county. When only a day or two at a time was missing, I used simple linear imputation to address missing data from a county level. If snow depth data was missing for a given day but existed for the day prior, I referred to average temperature. If the temperature was below freezing, I assumed no melt. If temperature was about freezing, I assumed a degree of snow melt and employed simple averaging imputation. If snow depth was missing for a day but snow fall existed and average temperature was below freezing, I added snow fall to the snow depth of the previous day. I calculated each weather variable in Table 1.3 daily then selected the maximum value (with the exception of TEMPMean) during the period preceding a respective survey. To be associated with a survey (t), I considered all values one day after the previous survey (t -1) through the day of survey (t). 14 All data analyses and figure construction were conducted in R ver. 4.0.3 (R Core Team 2020) using packages lme4 (Bates et al. 2015), AICcmodavg (Mazerolle 2020), effects (Fox 2003, Fox and Weisberg 2019), and ggplot2 (Wickham 2016). I compared model fit using Akaikie’s Information Criterion adjusted for small sample size (AICc). Models with an ΔAICc of 2 or less were viewed as competitive (Burnham and Anderson 2002). Total autumn duck abundance RESULTS The top fitting model for DUD contained an interaction between Area, Year treated as a linear covariate and the effect of adjacent lake level (Lake; Table 1.4). This model provided a good fit to the data and had an adjusted R2 value of 0.780. A model containing a three-way interaction between Area, Year, and Lake was also competitive, with a ΔAICc of 2.288 and model weight of 0.242. This model contained nearly twice the number of parameters as the top model (33 to 18) but only provided a small increase in R2 (0.795 vs. 0.780). Table 1.4. Top 8 performing models among 36 models considered in explaining total annual DUD. Bold values are metrics for the top performing model. K = number of parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value. Model Area * Year + Lake Area * Year * Lake Area * Year Area * MI Mallard + Lake Area * MI Mallard * Lake Area * Lake Area * MI Mallard Are * MI Mallard + NAO K 18 33 17 18 33 17 17 18 AICc 5199.626 5201.914 5222.270 5254.009 5256.300 5262.354 5264.663 5266.969 Δ AICc 0.000 2.288 22.644 54.383 56.675 62.728 65.037 67.343 ω Cumulative ω LL 0.758 0.242 0.000 0.000 0.000 0.000 0.000 0.000 0.989 1.000 1.000 1.000 1.000 1.000 1.000 1.000 -2579.836 -2560.856 -2592.377 -2607.028 -2588.049 -2612.418 -2613.573 -2613.508 The interaction between Area and Year is most apparent in the highly negative slope for Harsens Island (HI) and the slightly positive slopes for the Shiawassee SGA (SH) and the 15 Allegan SGA (AL; Fig. 1.2). The direction of trends were significant for Fish Point, Harsens Island, Shiawassee SGA, and the Shiawassee NWR. Figure 1.2. Estimates from the top DUD model (Area * Year). AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. Estimates are not portrayed for Allegan (AL) and Shiawassee SGA (SH) prior to 1998 and 2000, respectively, due to lack of usable data. The top model included a measure of Great Lakes water levels, which had an added effect of increasing DUDs while lake levels increased (Fig. 1.3). Although the highest deviation from average lake level observed during this study period was 0.794 meters, a relatively modest deviation of ± 0.1 would result in ± 19,555 duck use days across the season (i.e., β1 = 195,550). This equates to approximately 233 additional ducks per day for the entirety of established 12- week core period. 16 Figure 1.3. Expected change in an area’s respective annual DUDs when considering the added effect of lake levels. Migration timing I conducted analysis of the timing of peak abundance for nine dabbling duck species, and the top 3 models for each are portrayed in Table 1.5. Top models describing peak duck abundance included the Area variable coupled with measures of time were commonly among the best fitting. However, Area was not included as an explanatory variable in the top model for American wigeon or northern pintail, and no measure of time was included in the top model for green-winged teal (Table 1.5). Table 1.5. Top 3 performing models for each species explaining peak abundance timing. Bold values are those of the top performing model. Area = study location, Block = 3-year time, Year = linear time trend, Year2 = curvilinear time trend, AO = Arctic Oscillation, NAO = North Atlantic Oscillation, ENSO = El Niño Southern Oscillation. K = number of model parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value. Codea MALL Model Area + Block Area + Year2 Area + NAO K 13 11 10 AICc 882.718 887.540 891.935 Δ AICc 0.000 4.822 9.217 ω Cumulative ω LL 0.891 0.080 0.009 0.891 0.971 0.980 -426.402 -431.380 -434.821 17 Table 1.5. (cont’d) ABDU WODU AMWI GADW NOPI Area + Block Area + NAO Area + NAO + ENSO Area * Year Area * AO Area Year2 Year Null Area + Year2 Area + Block Area + ENSO Year2 Year Block NSHO Area + Year2 BWTE AGWT Area Area + Year Area * Year Area + Block Area + NAO Area * AO Area Area + Year 13 10 12 15 15 8 4 6 2 11 13 11 4 3 6 10 8 9 15 12 9 17 9 10 906.641 910.576 912.011 0.000 3.935 5.370 742.134 742.189 745.682 626.086 627.785 628.284 854.018 854.436 855.540 625.428 626.964 627.869 745.522 746.403 747.158 713.249 726.058 728.865 848.977 850.386 852.570 0.000 0.056 3.549 0.000 1.699 2.198 0.000 0.419 1.523 0.000 1.537 2.441 0.000 0.882 1.636 0.000 12.809 15.616 0.000 1.409 3.593 0.709 0.099 0.048 0.374 0.363 0.063 0.381 0.163 0.127 0.288 0.234 0.135 0.448 0.208 0.132 0.241 0.155 0.106 0.997 0.002 0.000 0.458 0.227 0.076 a United States Geological Survey’s species alpha codes 0.709 0.808 0.857 0.374 0.737 0.800 0.381 0.544 0.671 0.288 0.522 0.656 0.448 0.656 0.788 0.241 0.396 0.502 0.997 0.999 0.999 0.458 0.685 0.761 -438.364 -444.142 -442.346 -352.909 -352.937 -363.974 -308.769 -307.301 -312.062 -414.589 -412.218 -415.351 -308.434 -310.320 -307.343 -361.351 -364.302 -363.440 -338.425 -349.029 -354.321 -403.930 -415.236 -415.102 Model predictions for mallard and American black duck were similar (Fig. 1.4 and 1.5) with Pointe Mouillee (PM) and Nayanquing (NP) having some of the earliest timings, while Allegan (AL) was by far the latest. Additionally, after an initial shift towards earlier American black duck peak abundance observations (Fig. 1.5), both species exhibited gradual shifts towards later peak abundance until the most recent time block when timing became earlier (2017–2019). The difference between this time block and the previous one (2014–2016) was approximately 2 weeks. While there are differences in the area specific timing of peak abundances, the shifts between year blocks were generally parallel across areas. 18 Figure 1.4. Model estimates of peak abundance timing for mallards (Area + Year Block model). AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. 19 Figure 1.5. Model estimates of peak abundance timing for American black ducks (Area + Year Block model). AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. The best fitting wood duck model contained an interaction between Area and Year (Table 1.5). Peak abundances for wood ducks generally were between day 260–280 (mid-September to early October) though varying trends in timing occurred among areas (Fig. 1.6). The sharpest trends took place at the neighboring Shiawassee SGA and Shiawassee NWR, with peak abundance converging towards similar timings in recent years. There was some model uncertainty associated with the peak wood duck models, as the Area * AO model was also competitive with a ΔAICc of 0.056 (Table 1.5). Data from Muskegon was not included for the wood duck analysis due to annual peak observations of fewer than 50 birds. 20 Figure 1.6. Model estimates of peak abundance timing for wood ducks (Area * Year model). AL = Allegan, FP = Fish Point, HI = Harsens Island, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. Unlike most of the other dabbler species, the area effect did not add additional value in explaining the variation in observed timing of peak American wigeon abundance at my study sites. The outputs from the best model, which contained only a Year2 effect, show a subtle initial trend towards later peak abundance timing that has returned to be similar to the timings observed in the mid to late 2000s (Fig. 1.7). There was some uncertainty with the American wigeon models as Block was also competitive at a ΔAICc of 1.699 as well as the null model which had a ΔAICc of 2.198 (Table 1.5). Data from both Muskegon and Allegan were not included in American wigeon analyses due to annual peak observation of fewer than 50 birds. 21 Figure 1.7. Model estimates of peak abundance timing for American wigeon (Year2 model). The best fitting gadwall model was Area + Year2, though there was some model uncertainty with Year + Block and Year + ENSO also being competitive (Table 1.5). Gadwall peak abundance timing started around the beginning of November (day 305). Then they shifted to later timings around the year 2012 but have since returned to timings comparable to those observed during the late 2000s. Additionally, peak timings were latest at Pointe Mouille (PM) and earliest at Nayanquing Point (NP). 22 Figure 1.8. Model estimates of peak abundance timing for gadwall (Area + Year2 model). AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. As with American wigeon, northern pintail peak abundances were best explained by the Year2 model with the Year model also being competitive with a ΔAICc of 1.537 (Table 1.5). Estimated peak timings for all areas started mid-October and trended towards later in the month before shifting back towards slightly earlier timings in recent years (Fig. 1.9). Data from both Muskegon and Allegan were not included in northern pintail analyses due to annual peak observation of fewer than 50 birds. 23 Figure 1.9. Model estimates of peak abundance timing for northern pintails (Year2 model). The best fitting model for northern shoveler was Area + Year2, though there was some model uncertainty with the Area and Area + Year models having ΔAICc values of 0.882 and 1.636 (Table 1.5), respectively. Northern shoveler showed a parabolic pattern in peak abundance timings, similar to other species with recent peak abundance timings more comparable those of the late 2000s than the early 2010s (Fig. 1.10). Additionally, peak timings were latest at Pointe Mouille (PM) and earliest at Fish Point (FP). Data from Allegan were not included in the northern shoveler analysis due to annual peak observations of fewer than 50 birds. 24 Figure 1.10. Model estimates of peak abundance timing for northern shoveler (Area + Year2 model). FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. The best fitting model for explaining peak blue-winged teal abundances was the Area * Year model (Table 1.5). The trends in peak timings varied by area in both direction and strength (Fig. 1.11). The latest peak abundance estimate fell during the first week of October at the Shiawassee SGA in 2006, showing the overall early migratory nature of this species. Data from Allegan were not included in the blue-winged teal analysis due to annual peak observations of fewer than 50 birds. 25 Figure 1.11. Model estimates of peak abundance timing for blue-winged teal (Area * Year model). FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. American green-winged teal was the only species where some representation of year was not included in the top model. The Area*AO was the top model for American green-winged teal and the only other competitive model was the Area main effect model with a ΔAICc of 1.409 (Table 1.5). We see that years of negative AO values have greater variability with regard to peak abundance timing between respective areas while peak abundance timings were more similar across areas in years with a more positive AO (Fig. 1.12). 26 Figure 1.12. Model estimates of peak abundance timing for green-winged teal (Area * AO model). AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. Season progression and weather influence on mallard abundance My analyses exploring the influence of weather on changes in the relative abundance of mallards indicated that TEMP2 was the best at explaining variability in the observed data with TEMP also being competitive (Table 1.6). Table 1.6. Top 8 models for explaining change in relative abundance of mallards. Bold values indicate that they are representative of the top performing model. TEMP = maximum value of - (daily temperature) between survey periods, Julian = numerical day of year, MEANTemp = mean temperature between survey periods, Cumulative WSI = maximum value for TEMP + TEMPDAYS + SNOW + SNOWDAYS between survey periods, WSIMean = maximum value for MEANTemp + TEMPDAYS + SNOW + SNOWDAYS between survey periods. K = number of model parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value. Model TEMP2 TEMP Julian Julian2 MEANTemp2 K 5 4 4 5 5 AICc 5267.627 5269.255 5272.359 5272.903 5275.677 Δ AICc 0.000 1.628 4.732 5.276 8.050 27 ω Cumulative ω LL 0.610 0.270 0.057 0.044 0.011 0.610 0.880 0.937 0.981 0.992 -2628.793 -2630.614 -2632.166 -2631.431 -2632.817 Table 1.6. (cont’d) MEANTemp Cumulative WSI WSIMean2 4 5 5 5276.223 5300.800 5301.311 8.596 33.173 33.684 0.008 0.000 0.000 1.000 1.000 1.000 -2634.097 -2645.379 -2645.634 The predicted line from the top model followed the trend in observed data points, with model outputs from the top model as a line in Fig. 1.13 with the associated points representing individual observations. Observed points largely track the line of best fit well, but variability increases as temperatures approach and drop below freezing (Fig. 1.13). Figure 1.13. Model output from the best fitting mixed-effects model explaining change in mallard relative abundance (ln(Nt)-ln(Nt-1)) along with individual point observations. – (Temperature) is the inverse of ambient temperatures in degrees Celsius because below freezing is more severe. The best fitting generalized linear model of mallard abundance contained an interaction term between Week Number and Area and carried a model weight of nearly 1.0. and the best fitting model was Week Number * Area. The outputs from the top model indicate similar 28 patterns of mallard abundances at Harsens Island, Fish Point, and Shiawassee SGA, though the magnitude of abundances differed among these areas (Fig. 1.14). Pointe Mouillee and Nayanquing Point exhibited similar patterns of abundances with slightly earlier timing of peak abundances compared to other areas, but Nayanquing Point’s abundance dropped off more sharply as the season progressed (Fig. 1.14). Though the magnitude of abundance differed, the Shiawassee NWR and Muskegon both exhibited long gradual builds to peak abundance later in the season than the aforementioned sites (Fig. 1.14). Allegan continued to observe building mallard abundances through the entirety of the survey period (Fig. 1.14). Fig. 1.14. Outputs from the Week Number * Area model with a loess smoother describing seasonal mallard abundance on the managed areas. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge 29 DISCUSSION To my knowledge, this work is the first assessment of migration using archived data from the Shiawassee NWR and Michigan DNR’s MWHAs that considered all of those areas simultaneously. Total autumn duck abundance The nearly 3.5 million DUDs observed across my study sites in 2019 represents only approximately 64% of what was observed in 2000, the first year data were available for all locations. With the exception of Shiawassee SGA and Allegan, all areas are observing fewer annual DUDs. The varying degree of total annual DUD decline across areas provides an indication of underlying regional phenomenon and local nuances. Refuge sizes on the managed areas have largely remained the same throughout the period of observation, though some exceptions exist. The Shiawassee SGA added a surveyable refuge (a non-surveyable refuge has been present on the property since the 1970s) in 2000. Hunt plans remained consistent at Fish Point, Harsens Island, Shiawassee SGA, and Nayanquing Point. Pointe Mouillee added 8 additional zones and one additional hunt period (i.e., half day hunt) per week to the managed hunt program in 2014, but wetland units on the property not included in the managed hunt were historically and remain open to hunting 7 days a week. Muskegon and Allegan had some variability in hunt schedules as a function of Goose Management Unit regulations. Muskegon never approached disturbance levels of any areas other than Point Mouillee with regard to hunter trip totals (see Chapter 3) and hunt periods per week. Since Allegan survey methodology was standardized (1998) there have only been two years wherein disturbance levels (as measured through hunter trips) have approached those of Fish Point, Harsens Island, and Shiawassee (see Chapter 3). In recent years, however, managed hunts have been more frequent than during the 30 mid-2000s. The Shiawassee NWR has changed their plan away from field goose hunts to river duck and goose hunts, but this change moved the location of disturbance from actively managed fields to passively managed river (i.e., more food dense to less food dense habitats). Finally, observer changes have not resulted in obviously differing survey results between seasons. Although this phenomenon could be a function of shifting migratory corridors, it seems likely that the general decline in DUD is due in part to the underlying regional mallard population decline (U.S. Fish and Wildlife Service 2019a). The area with the sharpest decline in total DUD, Harsens Island, historically had a species composition more predominately made up of mallards, relative to some areas exhibiting subtle and uncertain directional trends like Muskegon, Pointe Mouillee, and Nayanquing Point (Michigan DNR unpublished data, Appendix A1.3). Refuge habitats at Harsens Island are comprised primarily of seasonally flooded cultivated plants (e.g., corn and buckwheat) that promote use by more granivorous species such as mallards and American black ducks (Baldassarre 2014). As such, observed total duck abundances are less likely to be supplemented by waterfowl species such as northern shovelers or gadwall that predominately forage on wetland dependent invertebrates and submerged aquatic vegetation, respectively (Baldassarre 2014). However, area specific change can influence observations, as another area exhibiting a sharp decline was the Shiawassee NWR. During the observation period, the refuge shifted away from agricultural grains; this change to more natural vegetation may have reduced the energetic capacity of the refuge to support waterfowl (Baldassarre and Bolen 2006c). This shift likely negatively impacted the total ducks the SNWR can support. The unique trend observed at Shiawassee SGA could be potentially due to the shift from hunted units to refuge units starting in 2000. While these units were always under active 31 management when they were emergent marshes open to hunting, they were not focal points on the 3,998-hectare complex as they are today. The positive trend at Allegan SGA, while directional uncertainty remains, seems to be associated with a transition to more overwintering birds. In the early 2000s, the number of observed birds tailed off to 0 or only a few hundred at the end of the “core period” I outlined while, in recent years, a few thousand birds are still being observed past that period and well into the new year (Michigan DNR unpublished data). Although regional population, migratory corridors, and habitat configuration are capable of influencing waterfowl distribution, local hydrology also likely plays an important role (Nichols et al. 1983). The inclusion of the adjacent lake level variable in the top Area * Year + lake and second Area * Year * lake best models showed to be approximately 20 AICc units better performing than the 3rd and simpler Area * Year model. As such, adjacent lake levels are contributing some explanatory power with regard to DUDs. Parameter estimates from my top fitting Area * Year + lake model provides support for positive relation between Great Lakes water levels and observed DUDs (Fig. 1.3). In addition to trends over time in total DUD and difference among areas, the levels of the adjacent Great Lakes had an additional influence on duck use of these management areas. Model fits favored a simpler model where the response of DUD to lake level was a consistent increase in use with higher water levels. Great Lakes water levels are an indicator of hydrological conditions in general for the state. While some uncertainty remains around this, it has been hypothesized that there are more breeding mallards and greater reproduction when there is more water on the landscape (David R. Luukkonen, personal communication). As such, greater DUDs reflect more ducks (dominated by mallards) when Great Lakes levels are high. However, this may be an oversimplified and incomplete picture with regard to my measure of autumn abundance. AIC favors simpler models and penalizes models by 32 2 AIC units for each additional parameter. Given the more complex model that allowed for area- specific responses to lake level contained 15 additional parameters and just missed being deemed competitive with a ΔAICc of 2.288, I would be cautious to conclude a uniform relation between area observations and lake levels. A more complex but direct relationship with adjacent lake levels is plausible given the differing degrees of connectivity to the Great Lakes among respective areas and warrants further research. Migration timing Some measure of year (i.e., Year, Year2, or Block) was present in the top model for each species peak abundance (Table 1.5), with the exception of American green-winged teal. This provides evidence of shifts in migration timing in Michigan; these results compliment the works of others both abroad (Lehikoinen and Jaatinen 2012) and in adjacent regions (Thurber et al. 2020). The lack of a year measure in the top (or in any competitive) American green-winged teal model could be attributed to their protracted migratory phenology (Baldassarre 2014) and the nature of my response variable (i.e., one measure per area per year). This could contribute to the lack of a distinct trend in peak observation timing. That said, Bellrose (1980) noted peak numbers in northern states occurring from early to mid-October. This is largely comparable to what my analyses show in normal to mild years (via AO classification). There was greater variability in peak abundance timing in years of greater winter weather severity (i.e., negative AO), which is likely a component of the areas specific refuge characteristics (e.g., water depth) making some areas more susceptible to freeze up. The only two species where there was not an underlying area effect with regard to trends in their migration timing were northern pintails and American wigeon. Both of these species are described as early migrants (Baldassarre 2014). Previous research in Illinois, USA and Ontario, Canada documented peak wigeon abundance in 33 and pintail movement through the Great Lakes Region occurring from mid to late October (Knapton 1992, Havera 1999, Malecki et al. 2006), similar to what was observed on my areas of interest (Fig. 1.7 and 1.9). The lack of an area effect is potentially due to the largest concentration of these species moving through the state before local conditions influence food availability. The peak abundances of all other species were a function of an area and a year measure, indicating differing timings among sites with some trend over time. Mallards and American black duck abundances peaked latest in the year, which was to be expected given their large size and opportunistic foraging characteristics (Baldassarre 2014). The order in which areas observed their respective peaks was largely consistent with what would be expected given each area’s respective geographic location, habitat types, and hydrology. Peak abundance timing for northern shovelers and gadwall, referred to as early migrants by many in Baldassarre (2014), did not take place until mid-November across my sites. Furthermore, Pointe Mouillee, an area that despite being a more southern location consistently observed some the earliest peak abundance timings of mallards and American black ducks; in contrast this area observed the latest peak abundance dates for northern shovelers and gadwall. This is likely a component of large emergent marsh units on the property remaining open longer than those of the study sites at more northern latitudes. Early freezes in Michigan in 2018 and 2019 (personal observation) likely influenced the recent shift back towards early peak abundances for mallards, American black ducks, American wigeon, gadwall, northern pintails, and northern shovelers. The lack of uniform trend in peak abundance timing of wood ducks is likely due to highly variable sizes of respective local breeding populations and potential difficulties in surveying them due to their selection of dense 34 habitat structures (Dyson et al. 2018). The area and year interaction of blue-winged teal peak abundances may be influenced by that species’ early migratory phenology and the implementation of an early teal season that has varying participation on the respective study sites. Season progression and weather influence on mallard abundance Weather conditions have been shown to influence migration in waterfowl (Schummer et al. 2010, Van Den Elsen 2016, O’Neal et al. 2018). While these works and my analyses all capture aspects of migratory behavior, they represent slightly differing questions. My work parallels that of Schummer et al. (2010) and Van Den Elsen (2016) wherein (weekly) measures of the rate of change in abundance are utilized to index migration, and are characterized by site specific conditions related to energy needs. I observed a quadratic trend of decreasing temperatures best explaining change in relative mallard abundance. Previous researchers found evidence that while temperature was the driving factor in principle component analyses, more complex indices such as those principle component analyses or weather severity indices were better for explaining changes in the relative abundance of dabbling duck species (Schummer et al. 2010, Van Den Elsen 2016). One possible reason for discrepancies in our findings could be related to study site differences. My observations were on smaller systems (~122 – 1400 hectares) with predominately lentic hydrology, whereas Van Den Elsen’s sites consisted of some larger areas with lotic hydrology (e.g., Mississippi river pools). This differing hydrology would presumably allow loafing areas to stay open longer until additional snowfall reduced the availability of dry field foraging. Furthermore, I indexed migration using actual abundance and found strong support for location specific interactions with a measure of season progression (Week Number). The 35 difference in area size and adjacent landscape composition (i.e., proportion agriculture, forested, open water, urban, etc.) inherently influenced mallard abundance, but comparing patterns of abundance provided insight how mallard staging varies by location (Fig. 1.14). These differing patterns are driven by underlying local conditions. A hydrological difference is seen in the Shiawassee properties. The Shiawassee NWR survey route contains a ~3.2 kilometers section of riverine emergent marsh that remains accessible to waterfowl after adjacent diked units on the property and neighboring Shiawassee SGA have frozen over. These hydrological differences could explain the observed variation in mallard relative abundances changes as temperatures approached and passed freezing (Fig. 1.13). Additionally, the timing of when locations receive harsh enough weather conditions can vary geographically. On a fundamental level, low temperatures during the winter fall along a north to south latitudinal gradient. This could explain why Pointe Mouillee does not exhibit as sharp of a decline in mallard abundance as season progresses relative to the northernly adjacent Harsens Island. However, Allegan and Muskegon exhibit later influxes of mallards relative to study sites of comparable latitude on the eastern part of the state (Fig. 1.13). This is likely due the warming influence Great Lakes exhibit adjacent shore areas during the winter (Scott and Huff 1996). The magnitude of influence increases with closer proximity to the most adjacent lake, if the location is downwind relative to the adjacent lake, and with adjacent lake size (Scott and Huff 1996). The Shiawassee SGA and NWR are approximately 35 kilometers inland from Saginaw Bay, while Allegan and Muskegon are only approximately 20 kilometers inland of Lake Michigan. In winter’s western/northwestern winter prevailing winds, Allegan and Muskegon are located downwind of Lake Michigan while Nayanquing Point and Pointe Mouillee are upwind of Saginaw Bay and Lake Erie, respectively. Finally, Lake Michigan was shown to have the a larger warming effect relative to lakes Huron 36 and Erie (Scott and Huff 1996) so the observed effect of Saginaw Bay on Fish Point and Lake St. Clair on Harsens Island is minimal by comparison. Another, yet more subtle, example of potential differences in adjacent lake influence is seen in comparing the timing of Nayanquing Point and Fish Point peak abundance and declines (Fig. 1.13). Both of these areas immediately border Saginaw Bay and are only ~32 kilometers away from each other, however Nayanquing Point is on the windward side of the bay, thus does not receive the same warming effect and exhibits earlier peaks and declines in mallard abundance. The combination of insights from these differing, yet related analyses is evidence that the ultimate factor for migration (with respect to the facultative characteristics of a given species) is associated with maintaining body condition (i.e., energy acquisition/thermal regulation). Once foraging reaches a giving up point (timing of which is location dependent), endogenous fat stores that maintain individuals until favorable flight conditions serve as the proximate cue for departure. This would be consistent with the findings of O’Neal et al. (2018) who utilized finer scale radar data and related discrete migratory movement to weather conditions more associated with flight (e.g., wind, cloud cover, and precipitation). However, local hydrological and weather conditions influence the timing of the aforementioned processes. MANAGEMENT IMPLICATIONS North American mallard population estimates are above the goal set in the North American Waterfowl Management Plan (U.S. Fish and Wildlife Service 2019a). However, declining Great Lakes (U.S. Fish and Wildlife Service 2019b) and Michigan (U.S. Fish and Wildlife Service 2019a) regional mallard breeding populations raise concern given the importance of locally raised mallards to the state. This work’s measure of total autumn duck abundance on key staging areas further corroborates an underlying regional phenomenon. 37 Previous works on mallard breeding the Great Lakes Region have documented hydrology and habitat structure conditions more beneficial to duckling survival/productivity rates (Simpson et al. 2007, Singer et al. 2016). Managers of my study sites should implement practices to promote the beneficial vegetation structures noted by Simpson et al. (2007), when feasible, so breeding birds can take full advantage when the ideal conditions noted by Singer et al. (2016) are present. Additionally, continued agency partnership with private landowners (i.e., landowner incentive programs) has the capacity to promote conditions beneficial to breeding birds on area that far exceeds the size and distributed of the MWHAs and Shiawassee NWR. Coluccy et al. (2008) noted the sensitivity of population growth rate to nonbreeding survival. Mortality during the nonbreeding season can commonly be attributed to hunting and other hazards associated with making migratory movements. Managing harvest associated mortality through continued season/daily limit regulation and by providing places of refuge will ensure that humans are checked from overexploiting this group of species. Additionally, using insights gained through my work on various measures of waterfowl migration timing can help managers plan managed unit flooding schedules to promote to maximize available food for staging waterfowl. This and continuing to work to mitigate the spread of invasive plant species (e.g., phragmites) will ensure the productivity of natural wetland food sources will help promote an optimal body condition for birds as they make migratory movements. While promoting reproductive success and autumn survival of mallards can help with populations, so can promoting hunting. Migratory Bird Hunting Conservation Stamps (i.e., duck stamp) purchases by hunters have historically served as one steady source of revenue for wetland/waterfowl focused land acquisition/conservation easement purchases. However, Vrtiska et al. (2013) documented a disassociation between prospering North American waterfowl 38 numbers of the last two decades and hunter numbers that historically tracked each other, and hypothesized that a significant decline in hunters could manifest if duck populations fall below a certain threshold. While Michigan waterfowl hunters have not declined at as sharp of rates as neighboring states, declines here have negative implications for wetland acquisition and management and should be of concern for waterfowl managers; particularly in a region, Michigan, with historically variable breeding habitat abundances (Soulliere et al. 2007) and high rates of wetland loss (Dahl 2011) relative to those adjacent. Given that harvest is fundamentally still of high importance to waterfowlers (Slagle and Dietsch 2018), promoting opportunities to harvest birds can presumably help retain them. One way of doing this is by ensuring that season/zone dates best align with waterfowl local abundances. There has been some push in the past by the Michigan Department of Natural Resources’ Citizen’s Waterfowl Advisory Committee (CWAC) to split the southern hunting zone into an east and west zone because of perceived differences in timing peak abundances on the western side of the state (Michigan DNR, personal communication). My analyses corroborate this perception with regard to timing of peak abundances at Allegan and Muskegon, relative to my other study sties. Additionally, my analysis provide evidence for broader regional delays in migration timings of some species and are further supported by comparable works with in region (Thurber et al. 2020). While this has not been much of an issue under longer season frameworks (since 1997), considering other hunting seasons (i.e., deer), local nuances, and monitoring regional changes well help regulatory agencies continue to optimize season/zone configurations. This is of importance given the ever- present possibility of future season restrictions. Waterfowl are among the most studied avian groups (Krapu 2007), in part due to their immense socio-economic value generated through recreational activities such as hunting and 39 birding. Because of their value and stakeholder investment, waterfowl and their habitats have been actively managed and studies for nearly a century. As such, relative to other North American avian communities, waterfowl are doing quite well. In a recent review of North American avifauna, it was found that waterfowl as a group are exhibiting a positive change in net abundance (Rosenberg et al. 2019). However, the future prosperity of this group is not certain. Working to promote breeding success of a locally important species in mallards, as well as aligning management regimes so that autumn food availability tracks broader shifts in migration timing will help to promote regional populations. Furthermore, considering how local conditions influence waterfowl abundance on these key areas can inform decisions on season/zone frameworks. Thus, maximizing potential opportunity for recreationalists, a key source of wetland and waterfowl conservation funding. 40 APPENDICES APPENDIX 1.1: SPECIES-SPECIFIC PEAK ABUNDANCE DATES Table A1.1.1. Numerical day of year of peak mallard abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. NP 281 291 277 277 279 274 274 297 283 297 282 317 NA NA 298 291 298 308 286 264 298 300 298 308 323 303 309 308 321 305 290 289 PM 327 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 312 296 280 278 278 313 NA 301 330 343 279 277 318 282 SH NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 294 NA NA 296 310 327 311 321 304 NA 308 314 336 321 305 304 294 SNWR 300 314 334 318 314 288 317 334 319 311 308 326 297 302 315 356 341 290 281 343 329 NA NA NA 333 323 345 352 344 332 318 302 Year 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 AL NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 337 NA 368 367 361 377 367 342 333 357 358 371 349 354 346 331 HI 310 311 309 303 300 301 NA 303 NA 309 NA 304 321 300 315 299 329 321 312 325 296 308 335 327 326 310 323 336 328 305 304 310 FP 315 303 299 305 325 293 287 284 330 323 281 298 281 301 301 297 NA 285 291 325 309 287 291 318 310 300 320 299 318 309 315 314 MU 317 305 310 293 337 328 334 338 318 295 329 300 318 333 317 329 323 306 326 311 324 329 314 334 333 317 337 329 342 333 325 338 Table A1.1.2. Numerical day of year of peak American black duck abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. NP 322 291 284 284 290 274 274 291 283 297 303 296 NA NA 319 326 298 301 307 320 305 300 305 301 323 296 309 322 300 291 297 296 PM 327 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 312 324 315 278 278 313 NA 301 330 308 293 284 318 338 SH NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 294 NA NA 310 310 299 290 321 304 NA 308 300 336 321 298 283 294 SNWR 300 314 334 311 314 308 317 334 332 311 293 333 297 386 350 356 341 339 358 287 301 NA NA NA 305 316 302 354 344 298 325 330 Year 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 AL NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 368 367 302 342 335 334 333 357 351 371 349 347 339 296 FP 315 303 285 305 325 300 308 312 330 302 310 282 281 294 279 308 NA 285 319 311 309 301 291 318 317 286 306 292 304 309 308 321 HI 303 304 309 303 300 301 NA 303 NA 340 NA 304 314 300 315 299 329 313 334 311 317 322 314 320 319 310 323 336 342 305 304 310 MU 301 317 310 293 337 328 306 296 309 335 329 342 318 333 317 329 323 306 305 311 331 357 314 327 312 317 323 336 335 305 325 324 43 Table A1.1.3. Numerical day of year of peak wood duck abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. NP 273 277 277 277 272 274 274 277 283 NA NA NA NA NA NA NA NA NA 279 257 277 264 256 273 266 275 260 259 279 277 248 275 PM 305 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 277 275 280 278 278 278 NA 280 281 280 279 284 283 275 SH NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 275 275 285 290 286 276 NA 273 251 245 286 256 255 245 SNWR 281 270 298 242 246 260 230 237 235 242 259 228 247 239 273 279 292 248 232 231 238 NA NA NA 270 217 238 238 280 215 273 254 Year 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 AL NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 296 NA 277 255 276 268 286 270 284 280 274 273 272 270 283 282 FP NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 279 NA 285 256 283 261 273 270 269 275 272 299 243 283 274 266 293 HI NA NA NA NA NA NA NA NA NA NA NA 262 NA NA NA NA NA NA NA 255 261 266 265 285 270 282 274 280 300 277 276 275 MU NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 257 NA 262 310 308 NA 299 277 254 260 259 244 256 255 254 44 Table A1.1.4. Numerical day of year of peak American wigeon abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. NP 273 277 NA NA 279 282 281 282 264 NA NA NA NA NA NA NA NA NA 286 285 284 300 298 301 309 303 288 287 293 305 276 296 PM NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 270 289 280 278 271 313 NA 280 309 280 321 326 318 303 SH NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 317 275 313 311 307 304 NA 308 286 301 293 284 283 287 SNWR 267 242 305 304 295 288 286 299 291 304 293 291 276 281 280 279 292 269 281 273 294 NA NA NA 313 294 281 287 287 282 290 302 Year 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 AL NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 298 318 290 282 308 320 270 NA 302 294 314 298 290 282 FP NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 291 283 282 294 305 297 303 286 306 299 290 274 280 286 HI NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 290 296 280 279 285 305 289 302 287 314 298 284 282 MU NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 277 NA 317 315 NA 306 NA 317 267 280 258 291 NA 289 45 Table A1.1.5. Numerical day of year of peak gadwall abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. NP 273 277 NA NA 279 258 258 277 283 NA NA NA NA NA NA NA NA NA 286 NA 298 300 305 273 309 296 267 308 300 291 276 296 PM NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 284 338 329 327 278 313 NA 322 316 336 335 305 318 275 SH NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 282 310 313 311 321 304 NA 294 300 294 300 284 304 287 SNWR 300 303 305 304 306 288 317 334 305 297 NA 291 297 295 280 272 292 297 288 287 294 NA NA NA 313 316 307 342 302 298 297 302 Year 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 AL NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 291 304 302 289 302 314 305 294 316 322 307 305 297 275 FP NA NA NA NA NA NA NA NA NA NA NA 282 288 NA NA NA NA NA 326 283 303 301 277 283 303 293 299 299 318 295 294 293 HI NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 290 296 280 293 306 298 296 295 301 300 305 304 303 MU 353 310 289 345 337 340 354 317 358 364 363 361 350 318 352 356 357 320 291 311 317 315 321 313 319 310 309 315 321 298 311 310 46 Table A1.1.6. Numerical day of year of peak northern pintail abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. NP 281 291 277 277 279 282 281 282 283 NA NA NA NA NA NA NA NA NA 295 285 305 286 298 294 309 296 295 322 300 305 276 289 PM NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 284 289 280 278 278 313 NA 294 302 322 300 305 283 310 SH NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 296 310 313 311 314 276 NA 301 300 322 279 291 276 287 SNWR 293 303 305 304 295 299 286 292 298 289 308 298 297 302 315 300 292 290 281 294 294 NA NA NA 326 301 307 300 309 298 304 310 Year 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 AL NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 291 325 386 289 NA 320 NA 357 358 280 286 389 297 275 FP NA NA NA NA NA NA NA NA NA NA NA 312 281 294 287 NA NA NA 284 283 303 301 305 290 310 300 306 264 318 295 315 314 HI NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 283 296 301 272 313 305 303 302 287 321 298 297 310 MU NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 341 NA 290 366 336 300 313 291 261 288 287 300 291 318 338 47 Table A1.1.7. Numerical day of year of peak northern shoveler abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. NP 288 285 302 302 265 294 294 282 283 NA NA NA NA NA NA NA NA NA 307 NA 298 286 298 294 316 289 302 308 328 298 262 296 PM 287 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 277 310 301 327 278 313 NA 315 309 336 321 319 318 310 SH NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 296 310 320 304 307 304 NA 301 300 308 328 312 304 294 SNWR 281 303 312 304 324 308 307 278 319 297 293 291 311 344 280 272 299 283 274 294 294 NA NA NA 326 323 307 323 302 332 312 302 Year 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 AL NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 298 297 NA 282 NA 334 NA NA NA 315 300 305 NA 296 FP NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 277 311 261 287 284 269 317 300 285 292 304 274 294 314 HI NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 297 NA NA 293 285 333 296 295 301 314 298 311 275 MU 337 NA NA 258 311 307 306 324 309 288 273 307 318 298 291 302 299 285 312 325 296 315 279 299 298 302 302 301 286 291 304 282 48 Table A1.1.8. Numerical day of year of peak blue-winged teal abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. NP 260 262 277 277 265 274 274 277 264 NA NA NA NA NA NA NA NA NA 272 257 263 264 256 273 259 260 260 252 279 263 269 268 PM NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 270 254 280 278 271 278 NA 280 281 280 272 277 262 282 SH NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 275 275 285 285 279 262 NA 273 265 252 258 277 248 245 SNWR 238 270 235 214 219 245 258 250 242 220 238 242 255 246 231 272 243 248 225 217 245 NA NA NA 298 273 266 252 250 264 273 268 Year 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 AL NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 305 304 276 282 256 327 270 273 267 266 265 291 276 282 HI NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 276 268 NA 272 278 259 275 253 252 251 256 262 282 FP NA NA NA NA NA NA NA NA NA NA NA 282 281 285 279 279 NA 285 256 276 261 280 270 269 282 265 257 264 255 239 273 258 MU NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 258 NA 257 256 262 268 259 258 271 259 247 246 245 258 242 241 247 49 Table A1.1.9. Numerical day of year of peak green-winged teal abundance observed of Michigan’s Managed Waterfowl Hunt Areas and the Shiawassee National Wildlife Refuge (1988–2019). NA = information not available. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. NP 273 320 277 277 290 287 287 291 283 NA NA NA NA NA NA NA NA NA 300 285 298 293 298 273 300 275 288 280 279 277 276 289 PM 305 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 298 275 301 271 278 313 NA 280 281 280 279 277 297 275 SH NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 275 275 285 297 321 283 NA 287 279 294 286 305 283 273 SNWR 281 303 264 260 268 299 286 334 305 297 293 298 297 281 315 279 299 311 274 301 280 NA NA NA 326 294 307 300 314 282 312 296 Year 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 AL NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 291 NA 290 289 272 NA 291 NA 295 308 272 298 241 289 FP NA NA NA NA NA NA NA NA NA NA NA 286 281 285 279 308 NA 293 305 318 309 301 284 269 303 286 292 292 304 281 308 307 HI 289 290 NA 275 NA NA NA NA NA NA NA 276 NA NA NA NA 308 321 291 283 275 294 272 313 291 310 274 273 265 277 276 310 MU NA NA NA NA NA 265 NA NA 259 NA NA 266 259 258 NA 258 NA 264 270 283 282 266 286 271 270 282 267 315 293 270 262 247 50 APPENDIX 1.2: AVERAGE SPECIES-SPECIFIC PEAK ABUNDANCES Table A1.1.2. Average peak group abundances observed (2006 – 2019). AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. Group Ducks Dabblers MALL ABDU WODU AMWI GADW NSHO NOPI BWTE AGWT AL 8,692 8,338 8,007 339 416 48 133 6 45 29 121 FP 19,864 17,914 12,112 1,970 849 1,117 1,676 354 638 446 1,707 HI 25,349 25,061 19,782 3,382 890 441 432 192 1,258 216 1,593 MU 15,798 6,318 2,322 296 26 12 358 4,450 26 385 95 NP 7,692 7,506 5,491 439 263 363 288 94 275 284 1,139 PM 11,120 9,636 6,183 440 990 183 959 150 189 1193 1,165 SH 14,029 11,706 9,525 970 339 520 303 339 741 326 868 SNWR 14,076 13,299 10,124 497 311 1,432 1,187 389 1040 220 1,791 51 APPENDIX 1.3: MALLARD PROPORTION OF CORE PERIOD TOTALS Figure A1.3.1. Area specific estimates (and 95% confidence interval) of the proportion of mallards in annual core period (numerical weeks 40 – 51 of the year) duck totals. AL = Allegan, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. 52 LITERATURE CITED LITERATURE CITED Ambaum, M. H. P., B. J. Hoskins, and D. B. Stephenson. 2001. Arctic Oscillation or North Atlantic Oscillation? Journal of Climate 14:3495–3507. Baldassarre, G. A. 2014. Ducks, Geese, and Swans of North America. Johns Hopkins University Press, Baltimore, USA. Baldassarre, G. A., and E. G. Bolen. 2006a. Winter. Pages 249–276 in. Waterfowl Ecology and Management. Second. Krieger Publishing Company, Malabar, USA. Baldassarre, G. A., and E. G. Bolen. 2006b. Wetlands and Wetland Restoration. Pages 405–468 in. Waterfowl Ecology and Management. Second. Krieger Publishing Company, Malabar, USA. . Baldassarre, G. A., and E. G. Bolen. 2006c. Feeding Ecology. Pages 143–176 in. Waterfowl Ecology and Management. Second. Krieger Publishing Company, Malabar, USA. Bates, D., M. Maechler, B. Bolker, and S. Walker. 2015. Fitting Linear Mixed-Effects Models Using {lme4}. Journal of Statistical Software 67:1–48. Bellrose, F. C. 1968. Waterfowl Migration Corridors east of the Rocky Mountains in the United States. Bellrose, F. C. 1980. Ducks, Geese, and Swans of North America. Second. Stackpole Books, Harrisburg, Pennsylvania, USA. Bookhout, T. A., K. E. Bednarik, and R. W. Kroll. 1989. The Great Lakes Marshes. Pages 131– 156 in L. M. Smith, R. L. Pederson, and R. M. Kaminski, editors. Habitat management for migrating and wintering waterfowl in North America. Texas Tech University Press, Lubbock. Brook, R. W., R. K. Ross, K. F. Abraham, D. L. Fronczak, and J. C. Davies. 2009. Evidence for Black Duck Winter Distribution Change. Journal of Wildlife Management 73:98–103. Burnham, K. P., and D. R. Anderson. 2002. Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach. Springer, New York, New York, USA. Coluccy, J. M., T. Yerkes, R. Simpson, J. W. Simpson, L. Armstrong, and J. Davis. 2008. Population Dynamics of Breeding Mallards in the Great Lakes States. Journal of Wildlife Management 72:1181–1187. Cotton, P. A. 2003. Avian migration phenology and global climate change. Proceedings of the National Academy of Sciences of the United States of America 100:12219–12222. Dahl, T. E. 2011. Status and Trends of Wetlands in the Conterminous United States 2004 to 2009. Washington, D.C. Dahl, T. E., and G. J. Allord. 1996. History of wetlands in the conterminous United States. Pages 19–26 in J. D. Fretwell, J. S. Williams, and P. J. Redman, editors. National Water Summary on Wetland Resources. U.S. Geological Survey Water-Supply Paper 2425. Dalby, L., A. D. Fox, I. K. Petersen, S. Delany, and J. C. Svenning. 2013. Temperature does not dictate the wintering distributions of European dabbling duck species. Ibis 155:80–88. Dormann, C. F., J. Elith, S. Bacher, C. Buchmann, G. Carl, G. Carré, J. R. G. Marquéz, B. Gruber, B. Lafourcade, P. J. Leitão, T. Münkemüller, C. Mcclean, P. E. Osborne, B. Reineking, B. Schröder, A. K. Skidmore, D. Zurell, and S. Lautenbach. 2013. Collinearity: A review of methods to deal with it and a simulation study evaluating their performance. Ecography 36:27–46. Dracup, J. A., and E. Kahya. 1994. The relationships between U.S. streamflow and La Niña Events. Water Resources Research 30:2133–2141. Dyson, M. E., M. L. Schummer, T. S. Barney, B. C. Fedy, H. A. L. Henry, and S. A. Petrie. 2018. Survival and habitat selection of wood duck ducklings. Journal of Wildlife Management 82:1725–1735. Eamer, J., T. Hayes, and R. Simth. 2010. Canadian Biodiversity: Ecosystem Status and Trends 2010. Federal, Provincial, and Territorial Governments of Canada. . Van Den Elsen, L. M. 2016. Weather and Photoperiod Indices of Autumn and Winter Dabbling Duck Abundance in the Mississippi and Atlantic Flyways of North America. The University of Western Ontario. . Accessed 15 Apr 2020. Fox, J. 2003. Effect Displays in {R} for Generalized Linear Models. Journal of Statistical Software 8:1–27. . Fox, J., and S. Weisberg. 2019. An R Companion to Applied Regression. 3rd edition. Sage, Thousand Oaks, CA. . Havera, S. P. 1999. Waterfowl of Illinois: Status and Management. Illinois: Illinois Natural History Survey, Urbana, Illinois. 55 Hurrell, J. W. 1995. Decadal trends in the North Atlantic oscillation: Regional temperatures and precipitation. Science 269:676–679. Kaatz, M. R. 1955. The black swamp: A study in historical geography. Annals of the Association of American Geographers 45:1–35. Knapton, R. W. 1992. The American Wigeon at Long Point: a species on the increase? Long Point Bird Observatory Newsletter 24:15. Krapu, G. 2007. Review of Waterfowl Ecology and Management by Guy A. Baldassare and Eric G. Bolen. The Auk 124:724–725. . Accessed 15 Apr 2020. Lehikoinen, A., and K. Jaatinen. 2012. Delayed autumn migration in northern European waterfowl. Journal of Ornithology 153:563–570. Lincoln, F. C. 1935. The waterfowl flyways of North America. Washington, D.C. . Accessed 8 Jun 2020. Malecki, R., S. Sheaffer, D. Howell, N. Carolina, W. Resources, and T. Strange. 2006. Northern Pintails in Eastern North America. Mazerolle, M. J. 2020. AICcmodavg: Model selection and multimodel inference based on (Q)AIC(c). R package version 2.3-1. . McCullough, G. B. 1985. Wetland threats and losses in Lake St. Clair. Pages 201–208 in H. P. Prince and F. M. D’ltri, editors. Coastal Wetlands. Lewis Publishers, Chelsea, MI. National Oceanic and Atmospheric Administration (NOAA). 2019a. Climate Prediction Center: Arctic Oscillation. . National Oceanic and Atmospheric Administration (NOAA). 2019b. Climate Prediction Center: North Atlantic Oscillation. . National Oceanic and Atmospheric Administration (NOAA). 2019c. Climate Prediction Center: El Nino Southern Oscillation. . National Oceanic and Atmospheric Administration (NOAA). 2019d. Great Lakes Dashboard. . Accessed 10 May 2020. National Oceanic and Atmospheric Administration (NOAA). 2020. Climate Data Online. . Nelms, K., B. Ballinger, and A. Boyles. 2007. Wetland management for waterfowl. 56 Newton, I. 2008a. Introduction. Pages 1–18 in. The Migration Ecology of Birds. First. Academic Press, San Diego, USA. Newton, I. 2008b. Control mechanisms. Pages 333–365 in. The Migration Ecology of Birds. First. Academic Press, San Diego, USA. . Nichols, J. D., K. J. Reinecke, and J. E. Hines. 1983. Factors Affecting the Distribution of Mallards Wintering in the Mississippi Alluvial Valley. The Auk 100:932–946. . Accessed 5 Jun 2020. NOAA National Centers for Environmental Information. 2019. Climate at a Glance: Statewide Mapping. . Accessed 10 May 2020. Notaro, M., M. Schummer, Y. Zhong, S. Vavrus, L. Van Den Elsen, J. Coluccy, and C. Hoving. 2016. Projected Influences of Changes in Weather Severity on Autumn-Winter Distributions of Dabbling Ducks in the Mississippi and Atlantic Flyways during the Twenty-First Century. . Accessed 15 Apr 2020. O’Neal, B. J., J. D. Stafford, R. P. Larkin, and E. S. Michel. 2018. The effect of weather on the decision to migrate from stopover sites by autumn-migrating ducks. Movement Ecology 6. . Accessed 28 Jan 2019. R Core Team. 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. . Rainio, K., T. Laaksonen, M. Ahola, A. V Vaihatalo, E. Lehikoinen, and A. V Vkhitalo. 2006. Climatic Responses in Spring Migration of Boreal and Arctic Birds in Relation to Wintering Area and Taxonomy Author (s): Kalle Rainio , Toni Laaksonen , Markus Ahola , Anssi V . Vähätalo and Esa Lehikoinen Source: Journal of Avian Biology 37:507–515. Reed, L. W. 1971. Use of western Lake Erie by migratory and wintering waterfowl. Michigan State University. Ropelewske, C. F., and M. S. Halpert. 1986. North American precipitation and temperature patters associated with the El Nino/Southern Oscillation (ENSO). Monthly Weather Review 114:2352–2362. Rosenberg, K. V., A. M. Dokter, P. J. Blancher, J. R. Sauer, A. C. Smith, P. A. Smith, J. C. Stanton, A. Panjabi, L. Helft, M. Parr, and P. P. Marra. 2019. Decline of the North American avifauna. Science 366:120–124. American Association for the Advancement of Science. 57 Schummer, M. L., J. Cohen, R. M. Kaminski, M. E. Brown, and C. L. Wax. 2014. Atmospheric teleconnections and Eurasian snow cover as predictors of a weather severity index in relation to Mallard Anas platyrhynchos autumn-winter migration. Wildfowl 4:451–469. Schummer, M. L., R. M. Kaminski, A. H. Raedeke, and D. A. Graber. 2010. Weather-Related Indices of Autumn–Winter Dabbling Duck Abundance in Middle North America. Journal of Wildlife Management 74:94–101. Scott, R. W., and F. A. Huff. 1996. Impacts of the Great Lakes on regional climate conditions. Journal of Great Lakes Research 22:845–863. Elsevier. . Serreze, M. C., and R. G. Barry. 2014. The Arctic Climate System. Second. Cambridge University Press, Cambridge, UK. Simpson, J. W., T. Yerkes, T. D. Nudds, and B. D. Smith. 2007. Effects of Habitat on Mallard Duckling Survival in the Great Lakes Region. Journal of Wildlife Management 71:1885– 1891. . Singer, H. V, D. R. Luukkonen, L. M. Armstrong, and S. R. Winterstein. 2016. Influence of Weather, Wetland Availability, and Mallard Abundance on Productivity of Great Lakes Mallards (Anas platyrhynchos). Wetlands. . Accessed 26 Nov 2018. Slagle, K., and A. Dietsch. 2018. National Survey of Waterfowl Hunters: Summary Report Mississippi Flyway 2018. Report to the National Flyway Council from the Minnesota Cooperative Fish and Wildlife Research Unit, University of Minnesota and The Ohio State University. St. Paul, MN. . Accessed 2 Dec 2018. Snell, E. A. 1987. Wetland distribution and conversion in southern Ontario. Soulliere, G. J., B. A. Potter, J. M. Coluccy, R. C. Gatti, C. L. Roy, D. R. Luukkonen, P. W. Brown, and M. W. Eichholz. 2007. Upper Mississippi River and Great Lakes Region Joint Venture Waterfowl Habitat Conservation Strategy. Fort Snelling, Minnesota, USA. Thurber, B. G., C. Roy, and J. R. Zimmerling. 2020. Long-term changes in the autumn migration phenology of dabbling ducks in southern Ontario and implications for waterfowl management. Wildlife Biology 2020. Tiner, R. W. 1984. Wetlands of the United States: current status and recent trends. 58 Tozuka, T., J. J. Luo, S. Masson, S. K. Behera, and T. Yamagata. 2005. Annual ENSO simulated in a coupled ocean-atmosphere model. Dynamics of Atmospheres and Oceans 39:41–60. U.S. Fish and Wildlife Service. 2019a. Waterfowl population status, 2019. Washington, D.C. U.S. Fish and Wildlife Service. 2019b. Adaptive Harvest Management: 2020 Hunting Season. Washington, D.C. . Vrtiska, M. P., J. H. Gammonley, L. W. Naylor, and A. H. Raedeke. 2013. Economic and conservation ramifications from the decline of waterfowl hunters. Wildlife Society Bulletin 37:380–388. Wickham, H. 2016. ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York. . Wood, H. B. 1945. The History of Bird Banding. Auk 62:256–265. 59 CHAPTER 2. AUTUMN WATERFOWL USE OF INTENSELY MANAGED HUNTING AREAS IN SOUTHEASTERN MICHIGAN INTRODUCTION Management of habitats is fundamental to successful wildlife conservation and so understanding factors that influence habitat selection of wildlife is a focus of natural resource managers. Habitat selection is described as a hierarchical process with selection taking place across different levels or scales. Johnson (1980) coarsely describes these as the physical or geographical range of a species (1st order), the home range of an individual or social group (2nd order), the usage of various habitat components within the home range (3rd order), and the use of resources available within a 3rd order site (4th order). Waterfowl have varying species-specific life history traits (Baldassarre 2014) which influence all levels of habitat selection. Due to their relatively large abundances and high mobility, the geographic range (i.e., 1st order habitat selection) for most waterfowl species spans multiple migratory flyways. As such, works studying changes in the geographical presence or absence of a species (i.e., local colonization/reintroduction or extirpation) are more common for less numerous, less mobile, or more controversial species (e.g., wolves; (Jimenez et al. 2017)). One example of changes in the geographic range of waterfowl, however, comes from the South Carolina Department of Natural Resources (DNR) introduction of non-migratory mottled ducks (Anas fulvigula), to the state’s coastal marshes during the 1970s (Shipes 2014). The native range of this relatively sedentary species was restricted to discrete populations in Florida and along the western gulf coast (Stutzenbaker 1988, Baldassarre 2014); making their introduction and subsequent expansion 60 across this adjacent state an example of changing 1st order habitat selection in a waterfowl species. The home range of an individual duck constitutes 2nd order habitat selection (Johnson 1980). Initial early work on home range focused on the breeding season (Dzubin 1955) but, by the 1970s, research on non-breeding waterfowl became more prominent (Baldassarre and Bolen 2006a), with numerous studies addressing home range in Europe (Legagneux et al. 2009, Bengtsson et al. 2014) and in North America (Lane 2017, Yetter et al. 2018, Pollander et al. 2019). These works indicate that home range size varies by time of year and species. Furthermore, due to the migratory nature of most waterfowl species, the location of an individual’s home range within its larger geographic species range also varies. As waterfowl make seasonal migratory movements, landscapes with large degrees of connectivity between wetland systems (Cowardin et al. 1979) and associated areas of forage (Davis et al. 2014, Kaminski and Elmberg 2014) serve as cues for stopover and staging (i.e., 2nd order habitat selection). Movements between various habitat components within a stopover/staging (Warnock 2010) individual’s home range constitutes 3rd order habitat selection (Johnson 1980). This can be conceptualized as a wetland or unit level selection. From a resource acquisition perspective, Lane (2017) found unit salinity to be the strongest predictor of wintering waterfowl use by gadwall (Mareca strepera) in a coastal system. Gadwall diets are predominately composed of submerged aquatic vegetation (Baldassarre 2014), so during a period with heavy precipitation, salinity decreased and submerged aquatic vegetation biomass and subsequent duck use increased. Additionally, Lane (2017) found that waterfowl selection of units under moist-soil management and agriculture production to be less than natural systems. However, Hagy and Kaminski (2012), 61 Pearse et al. (2012), Lancaster (2013), Palumbo et al. (2019) found that moist-soil habitat and flooded croplands were important for wintering ducks, particularly the abundant and dietary generalist mallards (Anas platyrhynchos). In addition to the forage provided by habitat units, risk and disturbance also play a role in waterfowl habitat selection (Fox and Madsen 1997, Bregnballe and Madsen 2004, Casazza et al. 2012, Lancaster 2013, Beatty et al. 2014, Lancaster et al. 2015, Lane 2017, Osborn et al. 2017, Palumbo et al. 2019). The role of risk likely varies by species, where those that are longer lived, such as northern pintail (Anas acuta), are less likely to take risks during periods open to hunting relative to a species like cinnamon teal (Spatula cyanoptera; Ackerman et al. 2006) with a shorter life span. These works indicate that species-specific life history traits, with respect to dietary preference and risk-navigating strategies, in conjunction with local management regimes can highly influence selection of 3rd order habitat. The finest level of habitat selection (4th order) deals with bird location within the habitat component that constitute 3rd order selections and is commonly influenced by micro-location variables. From a resource allocation perspective, water depths and food densities are viewed as driving factors. Foraging strategies of dabbling ducks limits them from readily obtaining submerged food resources in water depths ≥ 45 cm (Fredrickson and Taylor 1982). Similarly, Hagy and Kaminski (2012) noted 90% of observed mallards and other dabbling ducks using study plots with average water depths ≤15 and ≤16 cm, respectively. Additionally, Osborn et al. (2017) noted declining densities of dabbling ducks with increasing water depths. Furthermore, food densities and subsequent waterfowl use within a unit or marsh has been shown to be a function of management practices (e.g., water level manipulation and mowing; (Hagy and 62 Kaminski 2012)) and highly variable natural spatial distribution (Hagy et al. 2014, Vonbank et al. 2016, Lane 2017). True habitat selection constitutes an individual choice that is made based off of habitat quality and through weighing the cost and benefits of occupying a location that may result in disproportionate use relative to availability (Fretwell and Lucas 1970, Jones 2001). Due to the individualistic nature of habitat selection, GPS/GSM or VHF transmitters are commonly employed in habitat/resource selection studies (Casazza et al. 2012, Lancaster 2013, Yetter et al. 2018, McDuie et al. 2019, Palumbo et al. 2019, Shirkey et al. 2020). These works provide insights on how individuals of multiple common dabbling duck species navigate all levels of habitat selection. However, due to the highly mobile nature of waterfowl, it is less common (see Lancaster (2013)) for works to measure habitat selection on discrete areas of wetland/waterfowl management (i.e., National Wildlife Refuges, State Game Area, etc.). Furthermore, due to logistical constraints, most studies tracking individuals only have the capacity to deploy dozens of transmitters, while many areas of intense management commonly have thousands of individuals using them at any given time during migratory or wintering periods. As such, it is common for works with particular focus on select areas (St. James et al. 2013, Lane 2017) to employee methods such as ground counts or aerial surveys that document patterns of habitat use. Given that habitat selection process (Thomas and Taylor 1990, Jones 2001, Manly et al. 2002), studying patterns of use can provide insight on individual selection and direction for subsequent management. Michigan is located at the center of a Great Lakes Region that supports approximately 3 million annual migrants (Bookhout et al. 1989). While there have been some works studying waterfowl movements and habitat selection in the state (Prince et al. 1992, Palumbo 2017, 63 Knapik 2019), this information is lacking relative to other parts of the country (e.g., Mississippi Alluvial Valley). Furthermore, as non-protected wetlands continue to be subject to degradation or loss (Dahl 2011), state and federally owned complexes are increasingly responsible for providing habitat for waterfowl. While select areas were considered in Palumbo (2017) and Rice and Hagy (2020), there has not been a comprehensive assessment of waterfowl use of many key complexes similar to the works of St. James et al. (2013) or Lane (2017). As such the objective for this chapter was to explore how various local conditions influence waterfowl distribution with particular focus on 1) species-specific habitat use of wetland units managed for autumn waterfowl use and where hunting access was controlled, and 2) how season progression influences diurnal and nocturnal waterfowl use of different habitat types and areas of varying levels of hunting disturbance. Study areas METHODS As with the previous chapter, this research took place on Michigan DNR operated Managed Waterfowl Hunting Areas (MWHAs) and the Shiawassee NWR (Figure 2.1). I excluded two MWHAs (Muskegon County Wastewater Treatment Plant and the Fennville Farm Unit of the Allegan State Game Area) from this aspect of the research. While marsh units existed on the properties, zones under the managed hunting framework exclusively consisted of dry fields for geese. 64 Figure 2.1. Locations of the six wetland complexes in southeastern Michigan serving as study areas for this research. These properties vary in size and provide variety of natural and man-made wetland habitats (Table 2.1). The natural wetlands on the areas consist primarily of emergent cattail (Typha spp.) marsh, while buttonbush (Cephalanthus occidentalis) dominated shrub-scrub swamps, and non-mast producing forested wetlands also occur. The flooded fields on these properties consist of areas without perennial vegetation and are only seasonally flooded. Fields on the Shiawassee NWR exclusively consisted of areas under moist-soil management. While 65 some fields on state operated areas are also under moist-soil management regimes, the majority contained cultivated grains planted by Michigan DNR staff or sharecroppers, with some degree of moist-soil management incorporated. Table 2.1. Managed Waterfowl Hunt Areas and Shiawassee NWR property size and refuge size, as well as area of natural wetlands and cultivated fields flooded for waterfowl. FP = Fish Point, HI = Harsens Island, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee, SNWR = Shiawassee National Wildlife Refuge. Area FP HI NP PM SNWR SH Property Size (ha) Refuge Area (ha) Marsh/swamps (ha) Flooded Fields (ha) 1,002 1,358 609 2,102 4,047 3,997 291 121 161 381 3,455 190 486 769 302 872 1,322 1605 217 465 248 81 244 433 In addition to providing a suite of wetland and associated waterfowl habitats, the areas of interest for this work also provided an abundance of managed hunting opportunity for recreationalists (Table 2.2). Table 2.2. MWHAs hunt schedule during the two study seasons. Area 2018–2019 Hunt Schedule 2019–2020 Hunt Schedule Fish Point Harsens Island Nayanquing Pointe Pointe Mouillee Shiawassee NWR Entire Season AM/PM Entire Season AM/PM Entire Season AM/PM Opening Day AM/PM Sundays AM/PM Tuesdays AM Thursdays AM/PM Sundays AM Tuesdays AM Thursdays AMa Saturdays AM Entire Season AM/PM Entire Season AM/PM Entire Season AM/PM Opening Day AM/PM Sundays AM/PM Tuesdays AM Thursdays AM/PM Sundays AM Tuesdays AM Thursdays PMa Saturdays AM Shiawassee Entire Season AM/PM Entire Season AM/PM a Only change between years 66 Habitat classification and field methods In 2018 and 2019, I deployed 80 and 77 Browning Strike Force 850 HD game cameras, respectively, across my six study sites (Figure 2.1) to serve as fixed monitoring points for the entirety of the autumn migratory period. I determined locations for each camera through a modified stratified random selection process. I allocated cameras to habitat stratum rather than to study site. This was due to objectives of comparing waterfowl use of habitat types rather than use of different managed areas. The strata taken into consideration for camera allocation included a coarse habitat type classification and whether the location was open to hunting or not. Although the details of habitat conditions vary continuously, I developed four broad classes for use in stratifying habitat and for subsequent use as a categorical variable in statistical modeling. The classes I developed included cultivated (CUL), cultivated – moist-soil (CMS), open water – aquatic bed (OAB), and marsh – moist soil (MMS). I classified locations in units that were drawn down on an annual basis and planted with grains such as corn (Zea mays.), sorghum (Sorghum bicolor), and buckwheat (Fagopyrum esculentum) as Cultivated (CUL). If locations within these cultivated fields also contained moist-soil plants (Fredrickson and Taylor 1982), whether through active moist-soil management or volunteer growth, I classified them as Cultivated – Moist-soil (CMS). I classified wetland locations with semi-permanent hydrology as Open Water – Aquatic Bed (OAB). These areas consisted of emergent marshes and swamps that were generally inundated during previous growing season to promote the growth of wetland obligate species such as duckweeds (Lemna spp.) and pondweeds (Potamogeton spp.). If a semi- permanent wetland had been drawn down during the prior growing season to promote the growth of moist-soil plants, such as smartweeds (Polygonum spp.), barnyard grass (Enchinochloa crus- 67 galli), and beggarticks (Bidens spp.), I classified locations as Marsh – Moist-soil (MMS). I also classified locations that were in fields subject to moist-soil management but not planted with cultivated grains as MMS. This was because the vegetative communities in these fields more closely resemble the natural plants of MMS relative to CUL and CMS strata. The number of cameras allocated to each stratum was dependent on the total area of that habitat class and variability within that stratum (Table 2.3.). More available and heterogenous habitat stratum received more cameras those that were less available or of more homogeneous composition. While camera allocation was not based off of study sites, if a stratum existed on an area, I allocated at least two cameras to the stratum on the respective area. Once I completed camera allocation, I generated approximate camera location. For cameras allocated to hunted areas, I used a random number generator to select zone placement. For refuge or scramble hunting zones without discrete numbering, I used Google Earth Pro to create grids over imagery of units then subsequently numbered and selected grids using a random number generator. Table 2.3. Total camera allocation across the 2018 and 2019 field seasons by habitat stratum in location classified as refuge and those open to hunting. Disturbance Refuge Hunted Total Habitat Stratum Cultivated Grains Cultivated Grains and Moist Soil Marsh/Swamp – Moist Soil Marsh/Swamp – Open Water Subtotal Cultivated Grains Cultivated Grains and Moist Soil Marsh/Swamp – Moist Soil Marsh/Swamp – Open Water Subtotal Cameras 12 5 17 23 57 25 11 8 56 100 157 I began deployment during mid-August. This process ran through September and into the first week of October because some fields were not flooded until shortly before the opener of 68 general waterfowl season in Michigan’s southern zone (second Saturday in October each year). I mounted cameras on 5–8-foot u-posts and secured them with stainless steel hose clamps. Given that this was a novel employment of camera trapping methodology, I set the cameras to a “timelapse +” setting. In this setting, the cameras took a picture at set temporal intervals during daylight hours (based on the camera’s ambient light sensor) or any time (day or night) that something triggered the camera’s passive infrared (PIR) sensor. Given the novel nature of employing camera traps in a waterfowl use study, the pros and cons of each photo trigger method was unclear. I ultimately set my cameras to take a picture every 10 minutes, with a standard 2- shot (additional photo seconds after the first) with a one-minute delay on motion triggered photos. Factory specification of the cameras’ PIR sensor range was 80ft (~24 m). In an effort to minimize counting birds in clock triggered images that were beyond the bounds of what could be detected by motion triggered photos, I established distance markers using either natural wetland features or posts. I maintained cameras through monitoring period by swapping batteries and SD cards every 2-3 weeks, depending on each camera’s activity and logistic constraints. Cameras remained deployed each season through the New Year and I retrieved them once safe ice was over all the wetlands. I repeated the site selection process prior to the second field season, as the same camera locations were not reused during the second field season. Disturbance level classification The hunting schedules on my study sites were very similar (Table 2.2). In an effort to classify disturbance at a finer scale than area wide hunt schedules, I classified levels of disturbance based on the mean number of periods a week that an individual zone was hunted (Michigan DNR unpublished data). I classified Low level disturbance as zones that were hunted 1 – 4 periods a week. I classified Moderate level disturbance as zones that were hunted 5–8 69 periods a week. I classified High level disturbance as zones that were hunted 9 – 14 periods a week. Finally, Refuge areas were those that were closed to hunting for the entire season. In some instances, study sites had larger “scramble hunting areas” (i.e., units with a maximum hunter/party capacity rather than discrete zone) where hunter use data was coarser. To inform my disturbance classifications for cameras allocated to these units, I used data on hunter trips for the unit as a whole, along with insights from fieldwork interactions and discussions with area managers. This was necessary for 36 camera locations out of 157. Table 2.4. Disturbance level classification of 2018 and 2019 camera trap locations. Refuge location were closed to hunting for the entirety of the waterfowl season. Low disturbance level areas were hunted on average 1 – 4 periods a week. Moderate disturbance level areas were hunted on average 5 – 8 periods a week. High disturbance level areas were hunted on average 9 – 14 periods a week. Disturbance Level Classification Cameras Refuge Low Moderate High Total Photo processing 57 52 23 25 157 Six research technicians and I processed photos using Browning’s proprietary “Time- lapse Viewer Plus” software to view compressed files containing all diurnal photos from a given day. We initially processed all photos taken by the camera traps, but after assessing the processing rate and how much data the camera traps gathered through the first field season, we adapted the methodology to only process time-lapse photos at the rate of every 20 minutes. We continued processing all motion triggered photos. However, even though motion triggered two photos, it only counted as a single instance. I used the second photo only to help with species- specific ID count accuracy of large groups swimming/flying through the field of view. We manually recorded observations in a Microsoft Excel spreadsheet for each camera. The 70 formatting of compressed files limited our capability to efficiently use other common interfaces for processing camera trap photos/video (e.g., CPW Photo Warehouse; Ivan and Newkirk 2016), because metadata for individual still images was not retained. As such, within each data sheet we recorded numerical day of the year (i.e., Julian date), hour of day (0–23), minute of hour (0–59) corresponding with each respective image’s information stamp. In addition to this temporal metadata, we recorded image and site metadata. Image metadata consisted of trigger method (motion or time-lapse) as well as whether a fogged lens or lack of light kept the image field of view from being discernible. The site metadata we recorded noted the presence/absence of water, ice, and humans, as well as if the cameras field of view was completely iced up. We recorded counts of 23 waterfowl species/groups (Appendix 2.1) in the photos as opposed to simple presence/absence. During the early stages of processing, we also counted 4 non-target waterbird (Appendix 2.1) species and white-tailed deer (Odocoileus virginianus), but this was discontinued to help expedite processing. Finally, after all processing was completed, I went through images for all cameras to establish a field of view (FOV) measure (0-1) representing the portion of the 0–24 m monitoring site where vegetation was not obstructing view to the point where a bird could be missed. I did this to help account for differing levels of visibility between some locations. Because of the subjective nature of this measure, I completed all classifications for consistency. Data work and analyses For all of my analyses in this chapter, I fit generalized linear mixed models using the glmmTMB package (Brooks et al. 2017) in R ver. 4.0.3 (R Core Team 2020). This package is capable of handling datasets with a zero-inflated negative binomial distribution while also incorporating random effects. Additionally, users can specify influences of the zero-inflation and 71 different negative binomial family parameterization to facilitate model convergence. For all analyses, I filtered out all photos with humans present and when the FOV could not be discerned due to fog or darkness. Additionally, I filtered out all photos taken prior to inundation at a respective monitoring site. Finally, while we did get sporadic photos of waterfowl standing on ice, if the camera FOV was completely iced up I did not consider those days for analyses due to food resources being unavailable. Species-specific use I modeled species-specific habitat association using total number of individual birds present per day in time-lapse triggered images. Data were filtered to include only time-lapse images, and then summing the total individuals of each species/group for each unique year/camera/day combination. Additionally, I adjusted for differing field of views by dividing duck totals by the respective day’s FOV score and rounding to the nearest whole number. I created a candidate set of 8 models (Table 2.5) to test biological hypotheses. Due to my objective of determining species-specific use associations as a function of habitat type and whether a location of open to hunting, all models contained the combination (either additive or interacting) of these explanatory variables within the conditional component. I used combinations of null and explanatory variables in the zero-inflation component of the model and specified differing parameterizations of the negative binomial family to facilitate model convergence (Table 2.5). Table 2.5. Additive and interacting conditional models explaining overall species/group specific habitat use on an individuals per day scale. The zero-inflation model components set as null (~1) or as identical to the conditional component and both linear and quadratic parameterizations of the negative binomial data distribution family. Habitat Stratum is a factor variable with 4 levels (CUL, CMS, MMS, and OAB). Hunted is a factor variable with 2 levels (Hunted and Not Hunted). Model Conditional Variables Zero-Inflation Variables SPP-1 Habitat Stratum + Hunted ~1 Negative Binomial Linear Parameterization 72 Table 2.5. (cont’d) SPP-2 Habitat Stratum + Hunted SPP-3 Habitat Stratum * Hunted SPP-4 Habitat Stratum * Hunted ~1 ~1 ~1 Quadratic Parameterization Linear Parameterization Quadratic Parameterization SPP-5 Habitat Stratum + Hunted Habitat Stratum + Linear Hunted Parameterization SPP-6 Habitat Stratum + Hunted Habitat Stratum + Quadratic SPP-7 Habitat Stratum * Hunted Habitat Stratum * Linear SPP-8 Habitat Stratum * Hunted Habitat Stratum * Hunted Hunted Parameterization Quadratic Parameterization Hunted Parameterization Seasonal nocturnal and diurnal use In comparing diurnal and nocturnal waterfowl use as a function of season progression, I only considered data taken from motion triggered images as my camera model did not allow for time-lapse trigger images at night. Additionally, I modeled total duck and total goose use, rather than going down to species level. This was due to difficulties associated with species-specific ID at night. Because the ambient light that determines a standard vs infrared capture can vary with camera orientation, overhead cover, and weather, I used legal shooting hours (i.e., 30 minutes before sunrise to sunset) published in the 2018 and 2019 Michigan Waterfowl Digests to differentiate between day and night for the entirety of the study period. Using these times and the times recorded with each image, I was able to create diurnal and nocturnal sums for each respective date. These sums, also corrected for FOV, served as the response variable for subsequent analyses. Additionally, I established weeklong periods beginning on opening day through the entirety of the regular waterfowl season in order to explore temporal patterns in waterfowl use. I classified everything prior to the regular season opener as preseason and everything after the regular season closer as post-season. I created two candidate sets of 8 models 73 to explain seasonal diurnal and nocturnal waterfowl use across locations subject to varying levels of disturbance (Table 2.6) and different habitat types (Table 2.7). Table 2.6. Additive and interacting conditional models explaining seasonal progression of diurnal and nocturnal duck and goose habitat use as a function differing disturbance levels. The zero-inflation model components set as null (~1) or as identical to the conditional component and both linear and quadratic parameterizations of the negative binomial data distribution family. DN is a factor variable with 2 levels (Day and Night). Disturbance Level is a factor variable with 4 levels (Refuge, Low, Moderate, and High). Period is a factor variable with 10 levels (Pre-season, Season Weeks 1–8, and Post Season). Model Conditional Variables DL-1 (DN * Disturbance Level) + Linear Zero-Inflation Variables Negative Binomial ~1 Period DL-2 (DN * Disturbance Level) + Period DL-3 (DN * Disturbance Level) * Period DL-4 (DN * Disturbance Level) * Period ~1 ~1 ~1 Parameterization Quadratic Parameterization Linear Parameterization Quadratic Parameterization DL-5 (DN * Disturbance Level) + (DN * Disturbance Linear Period Level) + Period Parameterization DL-6 (DN * Disturbance Level) + (DN * Disturbance Quadratic Period Level) + Period Parameterization DL-7 (DN * Disturbance Level) * (DN * Disturbance Linear DL-8 (DN * Disturbance Level) * (DN * Disturbance Period Level) * Period Period Level) * Period Parameterization Quadratic Parameterization As within the species-specific habitat objective, other than possibility of an additive or interacting relationship, explanatory variables within the conditional model component were consistent. Table 2.7. Additive and interacting conditional models explaining seasonal progression of diurnal and nocturnal duck and goose habitat use as a function differing habitat types. The zero- inflation model components set as null (~1) or as identical to the conditional component and both linear and quadratic parameterizations of the negative binomial data distribution family. DN is a factor variable with 2 levels (Day and Night). Habitat Stratum is a factor variable with 4 levels (CUL, CMS, MMS, OAB). Period is a factor variable with 10 levels (Pre-season, Season Weeks 1–8, and Post Season). Model Conditional Variables Zero-Inflation Variables Negative Binomial STRAT-1 (DN * Habitat Stratum) + ~1 Period Linear Parameterization 74 Table 2.7. (cont’d) STRAT-2 (DN * Habitat Stratum) + STRAT-3 (DN * Habitat Stratum) * Period Period STRAT-4 (DN * Habitat Stratum) * Period ~1 ~1 ~1 Quadratic Parameterization Linear Parameterization Quadratic Parameterization STRAT-5 (DN * Habitat Stratum) + (DN * Habitat Stratum) + Linear Period Period Parameterization STRAT-6 (DN * Habitat Stratum) + (DN * Habitat Stratum) + Quadratic Period Period Parameterization STRAT-7 (DN * Habitat Stratum) * (DN * Habitat Stratum) * Linear Period Period STRAT-8 (DN * Habitat Stratum) * (DN * Habitat Stratum) * Period Period Parameterization Quadratic Parameterization For all analyses, I only considered models that converged without warning for AIC comparison. I used the emmeans package (Lenth 2020) to generate estimates (marginal means and confidence intervals) of duck use from best fitting GLMMs (i.e., lowest AIC value) on the response scale and created figures displaying these estimates using the ggplot2 package (Wickham 2016). RESULTS Between the 2018 and 2019 field seasons, my cameras took approximately 1.4 million photos and were active for a total of 9,227 trapping days. In approximately 1,515 worker hours, my technicians and I viewed all photos and recorded data for 477,354 images (double photo motion trigger and time-lapse trigger subsampling), resulting in just under 300,000 birds counted (Table 2.8) Table 2.8. Group and species breakdown of waterfowl observed across both field seasons. Group Dabblers Species Mallard American Black Duck American Green-winged Teal Individuals Observed 128,675 16,875 14,273 75 Table 2.8. (cont’d) Divers/Mergansers Unknown Duck Total Ducks Geese Swans Total Birds Species-specific use Northern Pintail American Wigeon Wood Duck Gadwall Blue-winged Teal Northern Shoveler Ringneck Redhead Greater/lesser Scaup Common Merganser Ruddy Duck Canvasback Hooded Merganser Bufflehead Common Goldeneye Red-breasted Merganser Canada Goose Cackling Goose Snow/Blue Goose White-fronted Goose Subtotal Subtotal Subtotal Trumpeter Swan Tundra Swan Mute Swan Subtotal 13,391 9,786 5,742 4,976 3,346 1,625 198,689 3,999 787 757 402 132 62 35 33 21 <10 6,228 43,836 248,353 44,453 <10 <10 <10 44,480 3,447 2,339 282 6,068 298,901 All species/groups of interest had at least one model in the candidate set converge without warning. The most parsimonious model (Model 1) was the top performing model for 8 of the 11 species/groups (Table 2.9) and converged without warning for all species except mallards. Model 3, Model 5, and Model 6 were the best performing models for Canada geese (Branta canadensis), mallards, and wood ducks (Aix sponsa), respectively (Table 2.9). Models 4 76 and 7 did not converge for any species. Models 2 and 8 only converged for one and three species, respectively and were never selected as the best model. Table 2.9. Species/group (species alpha codes used by United States Geological Survey’s Bird Banding Laboratory) AIC values for habitat use models that converged without warning. Bolded values indicate best model fit. WAR indicates the model had convergence issues. Model Model Model Model Model Model Model Model 2 3 4 5 6 7 8 Species/group MALL CAGO ABDU AGWT NOPI AMWI Divers WODU GADW BWTE NSHO 1 WAR 6,335.0 WAR 6,344.9 WAR 3802.3 WAR 3,799.3 WAR 3,755.5 WAR 3,758.6 WAR 2,480.4 WAR WAR WAR 3,187.7 WAR WAR WAR 2,354.6 2,396.7 WAR WAR 7,676.1 WAR WAR WAR 2,919.1 WAR WAR WAR 2,310.5 WAR WAR WAR 1,627.7 WAR WAR WAR WAR WAR WAR 16,389.7 16,582.5 WAR 16,515.0 6,492.2 3802.6 WAR WAR WAR WAR WAR WAR WAR WAR WAR WAR WAR WAR WAR 6,516.8 WAR 3,803.4 WAR WAR WAR WAR WAR WAR WAR 2363.7 WAR 7,664.4 WAR 2,945.6 WAR WAR WAR WAR WAR 2,357.8 WAR WAR WAR WAR Mallards were by far the most abundant species observed at 128,675 individuals (Table 2.8). Across habitat strata, approximately 5 mallards per day are predicted by model 5, except for in the CMS stratum where approximately 8 individuals a day were observed (Figure 2.2). Habitat strata locations open to hunting resulted in fewer than 0.5 individuals per day while their refuge counterparts observed approximately twenty times that (Figure 2.2). 77 Figure 2.2. Estimated marginal means and confidence intervals of diurnal mallard observations (Model 5). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. Canada geese were the next most observed species at 44,453 observations (Table 2.8). Estimates of daily use was highest in OAB refuge locations at 1.84 observations per day (Figure 2.3). Estimates of Canada Goose use were higher in MMS locations that were open to hunting relative to their refuge counterparts at 0.71 to 0.36 individuals per day, respectively (Figure 2.3). The model-based estimate for refuge locations of the CUL stratum was 0.28 geese per day, while estimated observations per day this stratum’s hunted locations as well as all CMS locations were nearly 0. 78 Figure 2.3. Estimated marginal means and confidence intervals of diurnal Canada goose observations (Model 3). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. American black ducks (Anas rubripes) were the third most observed species and the second most documented duck with 16,875 observations (Table 2.8). As with mallards, model- based estimates of use by this species were similar across the CUL, MMS, and OAB habitat stratum with CMS receiving higher point estimate (Figure 2.4). Additionally, black duck use of refuge locations was consistently higher than those where hunting was permitted (Figure 2.4). 79 Figure 2.4. Estimated marginal means and confidence intervals of diurnal American black duck observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. American green-winged teal (Anas crecca) was the third most documented duck species with 14,373 observations (Table 2.8). Model-based estimates of use by green-winged teal in refuge locations of the CUL, CMS, and OAB strata were all approximately 0.34 observations per day, while estimates for refuge locations of the MMS stratum was 0.14 observations per day (Figure 2.5). Observation estimates for hunted locations were roughly 30 times less than refuge locations with approximately 0.01 birds per day in the CUL, CMS, and OAB habitat strata and <0.01 per day for the MMS stratum (Figure 2.5). 80 Figure 2.5. Estimated marginal means and confidence intervals of diurnal American green- winged teal observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. Northern pintail (Anas acuta) was the fourth most documented duck species with 13,391 observations (Table 2.8). The highest model estimates for northern pintail observations per day were 0.18 and 0.15 for refuge OAB and refuge CMS, respectively (Figure 2.6). Refuge MMS and CUL areas received lower estimates of 0.04 and 0.03 observations per day, respectively. Northern pintail observations in areas open to hunting were well below 0.01 per day, regardless of strata (Figure 2.6). 81 Figure 2.6. Estimated marginal means and confidence intervals of diurnal Northern pintail observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. American wigeon (Mareca americana) was the fifth most documented duck species with 9,786 observations (Table 2.8). Model-based estimates for American wigeon observations were highest in refuge locations of the OAB stratum (0.17; Figure 2.7). Estimates of observations per day for cameras located in the OAB stratum were approximately eight times those of cameras in refuge locations of the other stratum types. Though estimates of observations per day in hunted locations were all <0.01, OAB location were again approximately eight times higher than those of other habitat stratum. 82 Figure 2.7. Estimated marginal means and confidence intervals of diurnal American wigeon observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. Diving ducks (Aythya, Oxyura, Mergues, and Lophodytes spp.) accounted for 6,228 observations (Table 2.8). Model estimates for this group predicted 0.18 observations per day in refuge locations of the OAB stratum, while refuge locations of the CUL, MMS, and CMS stratum received estimates of 0.05, 0.02, and 0.01 diver observations per day, respectively (Figure 2.8). While they both rounded to 0.01, hunted locations within the OAB stratum received higher estimates of daily observations than CMS refuge locations. Estimated diving duck observations per day for all other hunted stratum were <0.01. 83 Figure 2.8. Estimated marginal means and confidence intervals of diurnal diving duck observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. At 5,742 observations, Wood ducks were the next most common species but were only half as abundant as American wigeon (Table 2.8). Estimates from the top fitting model predicted 0.31 and 0.09 wood duck observations per day in the OAB and MMS strata, respectively. The 0.06 wood ducks per day model estimate hunted OAB locations was higher than hunted MMS locations and both CUL and CMS locations, regardless of if they were hunted or not (Figure 2.9). 84 Figure 2.9. Estimated marginal means and confidence intervals of diurnal wood duck observations (Model 6). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. Gadwall (Mareca strepera) were the third least documented duck species with 4,976 observations (Table 2.8). Estimates of observations in hunted locations were all <0.01 gadwall per day, regardless of habitat strata. OAB refuge locations had the highest point estimate at 0.17 observations per day, while MMS refuge location received the highest estimate at 0.06 observations per day. These were noticeably higher than estimated gadwall observations in hunted locations of those strata, as well as all CUL and CMS locations regardless of whether they were hunted or not (Figure 2.10). 85 Figure 2.10. Estimated marginal means and confidence intervals of diurnal gadwall observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. Blue-winged teal were relatively uncommon with 3,346 observations (Table 2.8). Model estimates of blue-winged teal observations were highest for refuge locations of the OAB stratum (Figure 2.11). Refuge location estimates for the strata containing moist-soil plants (i.e., CMS and MMS) were next highest at 0.03. Hunted location model estimates for the moist-soil strata were just <0.01 observations per day but were also identical. The estimate of blue-winged teal observations per day in hunted OAB locations was 0.03. 86 Figure 2.11. Estimated marginal means and confidence intervals of diurnal blue-winged teal observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. Northern shovelers (Spatula clypeata) accounted for the fewest duck observations with only 1,625 (Table 2.8). Estimates in locations closed to hunting were similar across strata type at approximately 0.03 observations per day (Figure 2.12). Estimates for locations open to hunting were <0.01 regardless of strata type. 87 Figure 2.12. Estimated marginal means and confidence intervals of diurnal Northern shoveler observations (Model 1). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. Seasonal nocturnal and diurnal use Disturbance levels For ducks combined, four model converged without warning and DL-6 was the best fitting model (Table 2.10). For geese, three models converged without warning and DL-1 was the best fitting (Table 2.10). Table 2.10. Disturbance level models with associated conditional variables and AIC values when explaining ducks and geese observed per day. WAR indicates that a given model had convergence issues. Bolded AIC values indicates the best fitting model that was used to generate observation estimates. Model DL-1 DL-2 DL-3 DL-4 DL-5 DL-6 DL-7 Conditional Variables (DN * Disturbance Level) + Period (DN * Disturbance Level) + Period (DN * Disturbance Level) * Period (DN * Disturbance Level) * Period (DN * Disturbance Level) + Period (DN * Disturbance Level) + Period (DN * Disturbance Level) * Period Ducks 38,684.930 39,032.190 WAR WAR 38,483.770 38,360.780 WAR Geese 17,073.250 18,071.810 WAR WAR WAR 17,509.210 WAR 88 Table 2.10. (cont’d) DL-8 (DN * Disturbance Level) * Period WAR WAR Diurnal duck use was highest in locations closed to hunting (i.e., refuge) for the entirety of the monitoring period (Figure 2.13). The difference was less prior to the start of the regular waterfowl season (pre) but the difference increased immediately at the onset (Figure 2.13). The difference between waterfowl observed in refuge units relative to hunted units remained similar through the end of October (Period 3), before diverging in the beginning of November (Periods 4 and 5) towards greater numbers in refuge units. After peak abundances during the third week in November (Period 6), the number of birds observed per day as well as the difference between refuge and hunting units declined until the last week of the general waterfowl season. Upon the close of the general waterfowl season, the differences between various disturbance levels remained similar to differences observed during the first week of December (Period 8). Moderately disturbed locations received the second largest estimates of duck observation per day, followed low level disturbance locations. High level disturbance locations had minimal estimates of duck observation per day for the entire monitoring period (Figure 2.13). 89 Figure 2.13. Estimated marginal means of seasonal diurnal duck observations per day across varying disturbance levels (Model 6). Refuge = location closed to hunting, Low = location hunted on average 1–4 times a week, Moderate = location hunted on average 5–8 times a week, High = location hunted on average 9–14 times a week. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season. Estimates of nocturnal duck use (Figure 2.14) were highest in moderately disturbed locations. Estimates of nocturnal use in refuge and highly disturbed locations were similar throughout the monitoring period. Model-based estimates of low disturbance locations indicated relatively little use at night (<2.00) throughout the entire monitoring period, even when estimates for all other disturbance level’s locations increased to greater than 13 ducks per night during the third week in November (Period 6; Figure 2.14). Throughout the monitoring period, total estimates of ducks per night (Figure 2.14) were higher than observations per day (Figure 2.13). 90 Figure 2.14. Estimated marginal means of seasonal nocturnal duck observations per night across varying disturbance levels (Model 6). Refuge = location closed to hunting, Low = location hunted on average 1–4 times a week, Moderate = location hunted on average 5–8 times a week, High = location hunted on average 9–14 times a week. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season. Estimates of diurnal goose use were consistently low in hunted locations; all had an average of less than 0.06 geese per day. The highest estimates of goose use (1.61) per day was observed at refuge locations during the third week in October (Period 1; Figure 2.15). Total diurnal estimates of geese per day (Figure 2.15) were much lower than duck estimates over the same monitoring period (Figure 2.13). However, these waterfowl exhibit some similarities within their bimodal patterns of early peak relative abundance associated with the start of hunting season (Period 1) followed temporary decline and subsequent increase to another eventual peak. 91 Figure 2.15. Estimated marginal means of seasonal diurnal goose observations per day across varying disturbance levels (Model 1). Refuge = location closed to hunting, Low = location hunted on average 1–4 times a week, Moderate = location hunted on average 5–8 times a week, High = location hunted on average 9–14 times a week. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season. Model-based nocturnal goose use estimates indicated similar patterns of abundance as diurnal estimates. This pattern indicates less nocturnal selection of hunted areas by geese relative to ducks that were much more common in hunted than refuge areas at night (Figure 2.15). While estimates of refuge location use declined at night, overall nocturnal use on areas was higher across the monitoring period (Figure 2.16). This was a component of increased use of hunted locations, particularly high and moderate locations, at night. 92 Figure 2.16. Estimated marginal means of seasonal nocturnal goose observations per night across varying disturbance levels (Model 1). Estimated marginal means of seasonal diurnal duck observations per day across varying disturbance levels (Model 6). Refuge = location closed to hunting, Low = location hunted on average 1–4 times a week, Moderate = location hunted on average 5–8 times a week, High = location hunted on average 9–14 times a week. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season. Habitat strata For ducks, four models converged without warning and STRAT-3 was the best fitting (Table 2.11). For geese, three models converged without warning and STRAT-5 was the best fitting (Table 2.11). Table 2.11. Habitat stratum models with associated conditional variables and AIC values when explaining ducks and geese observed per day. WAR indicates that a given model had convergence issues. Bolded AIC values indicates the best fitting model that was used to generate observation estimates. Model STRAT-1 STRAT-2 STRAT-3 STRAT-4 Conditional Variables (DN * Habitat Stratum) + Period (DN * Habitat Stratum) + Period (DN * Habitat Stratum) * Period (DN * Habitat Stratum) * Period Ducks WAR 39,424.050 38,926.130 39,177.430 Geese 17,031.850 WAR WAR WAR 93 Table 2.11. (cont’d) STRAT-5 STRAT-6 STRAT-7 STRAT-8 (DN * Habitat Stratum) + Period (DN * Habitat Stratum) + Period (DN * Habitat Stratum) * Period (DN * Habitat Stratum) * Period WAR 39,133.860 WAR WAR 16,900.090 17,667.790 WAR WAR Estimates of diurnal duck use as a function of habitat stratum lacked a distinct pattern in order or magnitude (Figure 2.17). Prior to season, duck use was highest (5.28 ducks/day) in CUL locations. However, on the onset of season, daily use estimates of that strata were never higher than those observed in all others. Duck estimates in the CMS, MMS, and OAB strata largely tracked each other prior to and during the first four weeks of the hunting season. Weekly estimates for these strata diverged during the firth week of the season and the CMS stratum had the highest daily use estimates for the remainder of the season (Figure 2.17). Figure 2.17. Estimated marginal means of seasonal diurnal duck observations across varying strata (Model 3). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season. 94 Estimates of nocturnal duck use by stratum indicated heavy duck use of areas with cultivated grains (CUL and CMS; Figure 2.18). Additionally, the MMS stratum received consistently higher use than the OAB stratum. The periods of highest total nocturnal use were weeks 4 and 5 (Figure 2.18), which were also period of higher diurnal use (Figure 2.17). However, the magnitude of nocturnal use during these periods, and throughout the season, was much greater than diurnal use. Figure 2.18. Estimated marginal means of seasonal nocturnal duck observations per day across varying strata (Model 3). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season. Model-based estimates of diurnal goose use was less than 2.50 birds per day at observation locations for the entire monitoring period (Figure 2.19). The CUL stratum received the highest estimates, however the other stratum with cultivated grains (CMS) received the lowest estimates (Figure 2.19). 95 Figure 2.19. Estimated marginal means of seasonal diurnal goose observations per day across varying strata (Model 5). CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season. As with model-based estimates of diurnal use, nocturnal goose use was highest in the CUL stratum (Figure 2.20). However, the magnitude of nocturnal use was higher. Nocturnal estimates for the MMS and OAB stratum remained similar, but the CMS stratum exhibited higher nocturnal use (Figure 2.20) than diurnal use (Figure 2.19). 96 Figure 2.20. Estimated marginal means of seasonal nocturnal goose observations per day across varying strata (Model 5 CUL = cultivated, CMS = cultivated and moist soil, MMS = marsh/swamp – moist soil, OAB = marsh/swamp – open water. pre = prior to the start of general waterfowl season, 1–8 = respective week of the hunting season, post = after the closure of the general hunting season. DISCUSSION This work does not directly measure habitat selection of individuals. However, it does provide insights on patterns of habitat use on managed areas, thus contributing to larger conservation plans. While locations throughout migratory corridors may be highly dynamic on an inter-annual basis (Davis et al. 2014), study sites are of particular regional importance given their level of consistency providing a suite of habitat types. Species-specific use My estimates of species/group specific diurnal waterfowl use indicate waterfowl are predominately disturbance influenced on intensely managed wetlands in Michigan. This supports the findings of numerous other (Cox and Afton 1997, Evans and Day 2002, Lane 2017, Palumbo 97 et al. 2019). I also found evidence of varying species/group-specific distribution as a function of habitat type. However, given the gregarious and opportunistic (with respect to forage) nature of most waterfowl, I noted use by all species in all disturbance and strata combinations. Mallards were the most observed species in this study. They are a large and long lived species, thus more risk adverse (Ackerman et al. 2006). This risk adverse behavior is seen in the distinct difference between use patterns of hunted and refuge locations within the same stratum type. Their status as a dietary generalist that will exploit rich environments of both natural and cultivated food resources (Baldassarre 2014) is also supported here, with the highest observed use in a mixed stratum (CMS) but similar use across exclusively cultivated or natural stratum. American black ducks are another species in the “mallard-complex”. The tendency of this species to consume greater proportions of animal matter than mallards is well documented, particularly in coastal systems (Baldassarre 2014). However, given the location (inland freshwater wetland) and scale of this study use patterns comparable to mallards is to be expected. Canada goose was the second most observed species. They displayed the only instance where hunted locations within a given habitat stratum received higher use estimates than that stratum’s refuge location. However, this is likely an artifact of a couple of monitoring points in zones that went largely overlooked by hunters for large stretches of the season due to a perceived lack of hunt potential in non-cultivated zones (Michigan DNR, personal communication). Habitat-based estimates for Canada geese indicated higher diurnal use on the managed hunting areas in the natural vegetation strata (OAB and MMS) relative to the cultivated strata (CUL and CMS). While the aquatic vegetation and moist-soil plants that characterize the natural strata provide food for Canada geese, diurnal use of these locations is likely associated with 98 loafing/watering before or after foraging flights out to agricultural fields to procure the grains that serves as the mainstay of their migration and wintering diet (Baldassarre 2014). Wood ducks are noted to be very opportunistic foragers (Baldassarre 2014). Acorns and cultivated grains are of particular importance in the fall, with seed of obligate wetland plants being readily consumed during late summer (Landers et al. 1977). My study sites lacked large abundances of mast producing trees but aquatic seed producing plants (e.g., American lotus; Nelumbo lutea) and cultivated grains were readily available. The availability of aquatic plants coupled with substantial local breeding populations of an early migrant species make higher use of OAB habitats to be expected, especially given a potential mismatch between peak species abundances and cultivated grain maturity. Blue-winged and green-winged teal are the smallest dabbling ducks in North American. As such, they are characterized as being more disposed to risk taking (Ackerman et al. 2006) and would be presumably more tolerant to hunting disturbance. While refuge location estimates were always higher than hunted locations of a given stratum, I found evidence of this risk tolerance through relatively similar blue-wing teal estimates in refuge versus hunted locations. However, it is unclear if this similarity of use is attributed to risk tolerance, an early migration phenology that has them using areas before the onset of the general waterfowl season, or if it is a function of lower observation totals. Teal species have both been documented field feeding on waste corn (Baldassarre and Bolen 1984). However, many works note the higher prevalence of smaller seeds from non-cultivated plants in their diets (Baldassarre 2014). In locations with natural vegetation (CMS, MMS, and OAB) these species are likely taking advantage of the seeds of moist-soil or submerged aquatic vegetation species. In exclusively cultivated locations (CUL), 99 birds of these species are likely foraging more extensively on buckwheat (Fagopyrum esculentum) or invertebrates as opposed to corn. As the 4th and least most observed dabblers, respectively, northern pintails and northern shovelers are of varying abundance at the study sites. Northern pintail diet studies note the importance of moist-soil plants and submerged aquatic vegetation during fall and winter (Baldassarre 2014), and Cox and Afton (1997) and Baldassarre and Bolen (1984) noted foraging flights to rice and corn fields, respectively. Northern shoveler diets, meanwhile, are primarily comprised of animal matter (Baldassarre 2014). While magnitude of observations varied, both species exhibited similar use across strata. Indicating these birds make use of a wide variety of habitat types when procuring food resources. American wigeon and gadwall held true to their aquatic vegetation specialist nature (Baldassarre 2014) and were most observed in the OAB stratum. Sporadic minimal use of other environments to largely similar estimates across the CUL, CMS, and MMS strata. The diver group exhibited similar estimates to wigeon and gadwall. The dietary preferences and foraging strategies of individual species in this group make any observations in cultivated fields (CUL and CMS) or shallow MMS locations likely being of the gregarious nature. Seasonal nocturnal and diurnal use Disturbance levels As expected, diurnal waterfowl use was highest in locations closed to hunting and lowest in those subject to the highest level of disturbance (Fox and Madsen 1997, Madsen 1998, Baldassarre and Bolen 2006a, Lane 2017). This does not represent a disconnect between hunter understanding of waterfowl habitat preference, but rather shows consistent bird displacement from locations due to disturbance. When all locations are available to waterfowl (i.e., night), 100 locations subject to moderate and high levels of disturbance exhibited use higher than and comparable to that of refuge locations. Additionally, documented waterfowl home ranges much larger than my study sites and distances of day to night movement being larger than those at other times of the day (Yetter et al. 2018, Shirkey et al. 2020) provide support that my estimates of nocturnal waterfowl observations are associated with birds loafing off-site and returning to forage at night is not only feasible but highly likely. Furthermore, my work indicated diurnal use of refuge units was highest regardless of season period. This could indicate that refuge units on areas are more desirable than adjacent hunted zones in a disturbance free landscape. The ever-present potential for logistic constraints makes management of refuge units an annual priority relative to those that are hunted. However, given early teal and goose hunting is allowed on most properties, as well as other non-hunting disturbances (Korschgen and Dahlgren 1992), it is also likely that many areas are not truly disturbance free prior to the general waterfowl season. Thus, birds are already to some degree being displaced to places of refuge. During the final few weeks of hunting season, ice forming conditions are common. While hunters break open small holes in hunted units, due to the tail off in hunter trips at the end of the season (Michigan DNR, unpublished data) and relatively short periods when hunters are there to maintain, these locations do not have the capacity to remain open as long as refuge units holding concentrated densities of waterfowl. St. James et al. (2013) did not find a difference in duck densities between units subject to varying levels of hunting disturbance Our differing results could be a component of our respective study sites’ hunt plans. The two different frequencies they compared (2 or 4 days/week, mornings only) would have both classified as low disturbance locations in my study. Waterfowl altering flight path upon approaching the edge of refuge units is a common 101 occurrence (personal observation and MDNR, personal communication) and indicates strong conditioning to managed area/landscape level disturbance configuration. However, it is unclear if this is a function of gradual conditioning through multiple years of identical refuge and hunted unit configuration, or if it takes place on an annual basis. Habitat strata Diurnal use of habitat strata indicated waterfowl use was highest in locations with cultivated grains (CUL and CMS). These finding are similar to those of Palumbo et al. (2019). Subsequent shifts to high levels of nocturnal use of cultivated strata (CMS and CUL) would be consistent with notes of anthropogenic disturbance on activity budgets (Korschgen and Dahlgren 1992) and evening foraging flights observed by mallards (Turnbull and Baldassarre 1987). Camera traps While camera traps have been commonly utilized in various natural resource applications (O’Connell et al. 2011), they have rarely been employed in studies indexing waterfowl use. To my knowledge, Cowardin (1969), and Firth et al. (2020) are the only instances other than this work. The camera set up used by Cowardin (1969), described in detail by Cowardin and Ashe (1965), was limited by the technology of the time and were not the only method they used to quantify use. Additionally, technological limitations likely explain the lack of camera trap utilization is waterfowl studies during the period after Cowardin (1969) but preceding my work and that of Firth et al. (2020). While my study indexed relative waterfowl use for the purpose of management and hunting of waterfowl centric managed hunting areas, Firth et al. (2020) described the use camera traps to index duck and goose use of rice fields as a part of a larger study on low-external-input sustainable-agriculture (LEISA) for the purpose of rice production 102 (Firth et al. 2020a). Regardless of purpose, both of our works showcased the utility of this method. The key advantage of this methodology is that it provides an inconspicuous way to continuously monitor use of a location. Ground counts and aerial surveys are highly conspicuous and disturb both waterfowl and hunters trying to pursue them. These methods serve as snapshots of use and can be confounded by the time in which they are conducted. Furthermore, counts routinely take place during that day. With nocturnal foraging events commonly take place in highly disturbed landscapes, the lack of information of the magnitude of these events presents a significant knowledge gap that has implications for management. Another method to index waterfowl use of habitat components on hunted areas can be indexed through harvest surveys. Bag checks are labor intensive for researchers and hunt self-check methods can be subject to dishonesty associated with a perceived need to hide “hot spots”. Furthermore, temporal limitations (i.e., no measure of nocturnal use) and the inherent influence of hunter skill can limit conclusions made by researchers. Just as there are advantages associated with employing camera traps in waterfowl use studies, there are disadvantages. Most notably is the post data collection processing time. The ~1,515 hours of photo processing for this study translated to approximately a year and a half long endeavor, that would not have been possible without the employment of research technicians. While recent works have expanded on ways to optimize the deep state machine learning (Schneider et al. 2020) and citizen science (Swanson et al. 2016) that mitigate processing time, the studies commonly employing these methods focus on species that lack the sexual dimorphism and seasonal change in plumage characteristics associated with most 103 waterfowl. Furthermore, theft (noted by Firth et al. (2020)) and flooding events (experienced in this study) can lead to data loss. I did not extrapolate my estimates of waterfowl use to area level abundances, but rather compared relative use across species, time, and habitat structure. This decision was due to my specific research questions and traits associated with my focal species’ appearance and life history traits of my focal species. The largely inconspicuous appearance of individuals within their species makes individual identification necessary for simple mark-recapture analysis impossible. While methods for determining broader abundances of unmarked animals have emerged in recent decades (Gilbert et al. 2020), the mobility of waterfowl makes many of these non-applicable. Some methods such as instantaneous sampling (Moeller et al. 2018) could be considered by waterfowl managers in the certain situations. This is due to their requirement of common species and their use of time-lapse triggered photos that eliminate the variation associated with detection probability, which has been shown to be a function of distance and animal size (Rowcliffe et al. 2011). However, assumptions associated with demographic/geographic closure and animal detection being independent of time and space (i.e., homogenous study area landscape) should be considered when determining applicability for a give research question (Moeller et al. 2018, Gilbert et al. 2020). While my particular focus was on indexing waterfowl use of habitat components (i.e., 3rd order), on intense management areas, camera traps have applicability for indexing other questions at this level (e.g., unit specific productivity) or at finer orders (e.g., water depth influence on foraging locations). 104 MANAGEMENT IMPLICATIONS This chapter provides an extensive study of waterfowl use patterns on Michigan’s Managed Waterfowl Hunt Areas. Furthermore, it represents a test of a novel method to monitor waterfowl use. Waterfowl distribution across a landscape is dependent on a suite of factors both within and outside of control of managers. However, natural resources managers manipulate the state of the wetlands to promote ideal conditions. On the study sites for this research, both the Michigan DNR and the U.S. Fish and Wildlife Service manipulate water-levels during the growing season to promote the growth of annual seed producing moist-soil plants. Additionally, the Michigan DNR staff and sharecroppers plant cultivated grains on state operated areas creating wetland complexes rich in food resources with emphasis on providing high levels of food energy during autumn. This level of intense active management can be costly for operating agencies (and local support organizations). My work indicated that waterfowl use of certain habitats aligned with expected dietary preferences. As such, managers should continue to provide a variety of natural and cultivated food resources within wetland complexes. However, managers should consider implementing more mixed management units, characteristic of the CMS stratum, where cultivated grains and moist-soil plants coincide. This essentially is the concept of “dirty corn” wherein the growth of “weeds” is not mitigated through the application of herbicides. Observation estimates for most species were higher in these locations relative to the purely cultivated locations (CUL). Waterfowl readily exploit agriculture grains due to their high carbohydrate return from a minimal foraging effort (Baldassarre and Bolen 2006b). However, waterfowl cannot persist on agriculture grains alone (Loesch and Kaminski 1989). Given that corn kernels remain after the conclusion of the migratory period (personal observation), it is 105 possible that managed fields supplementing energic needs are dropping below the point where ducks can no longer exploit the food resource (50 kg/ha; Reinecke et al. 1989). Quantifying the energetic carrying capacity of the managed areas using published forage values (Baldassarre and Bolen 2006b) relative to area refuge counts accounting for the nocturnal influx of birds that this work documented, would provide definitive measure of this. If energy is not limiting, there is potential to provide a greater abundance of natural food sources without creating an energy deficient landscape. Assessments of unit specific productivity from an energetic carrying capacity standpoint help inform this beyond anecdotal level, given the large degree of variability between area-specific farming programs (e.g., sharecroppers, post-season harvest, management plant changes, etc.) Exposure to disturbance also influences waterfowl habitat selection. The decision by an individual to return to a disturbed location is impacted by the location’s value as a foraging site (Houston et al. 1993, Bregnballe and Madsen 2004). Palumbo et al. (2019) noted mallards navigating hunting related risks to nocturnal and diurnal select for flooded agriculture throughout an autumn migratory period (pre-post season). This corroborates my results of higher nocturnal and diurnal use of locations with cultivated grains throughout a comparable monitoring period. Furthermore, Lane (2017) provides a comparable work of assessing patterns of use on an area of perennial importance and noted waterfowl selection of units closed to hunting. My comparisons of waterfowl diurnal use also indicate this. St. James et al. (2013) noted no difference between duck densities for locations hunted 2 and 4 times a week. My analyses provided some evidence of differing use of locations subject to differing levels of hunting disturbance, but uncertainty remains around this and warrants further research. Drastically altering hunting opportunity on managed areas would be problematic given their history of providing and abundance of 106 opportunity. However, future land acquisitions could be utilized to test experimental hunting pressures. From a methodological standpoint, camera traps are a viable tool in meeting monitoring goals and objective. Currently, area staff conduct weekly counts of refuge units to index waterfowl migration. From these data, managers get insights into migration timing and relative waterfowl abundances (see Chapter 1). However, these weekly counts generally only document waterfowl use in refuge units, exclusively serve as snapshots of waterfowl abundance when movements in and out are continuous and are subject to disturbing birds. Furthermore, the weekly timing of each respective area’s count differs which limits that capacity to index abundance cohesively. The expansion of analytical frameworks (e.g., instantaneous sampling) have the capacity to monitor migration timings and generate estimates of waterfowl abundance from camera trap data at a finer temporal scale than current weekly estimates. The status of refuge counts (i.e., timing and number observed) are of keen interest to hunters. This presumably stems from hunting purposes. The desire to hunt when local abundances are high is commonplace, but also is the desire to hunt on the morning/afternoon where a mid-day count has the potential to rally birds. My work indicates intensely managed areas in Michigan support a variety of waterfowl species and their specific relative use of habitat types largely expected life history traits. As such, a complexes of differing habitat types shall be retained. Diurnal waterfowl use when compared across disturbance levels was highest but not exclusively in refuge units for the entirety of the monitoring period. This indicated that while birds were displaced diurnally, they are capable of navigating hunting disturbed landscapes. Additionally, diurnal use was commonly highest in food rich habitats (i.e., CMS), particularly mid to late season. However, higher estimates in other 107 stratum (i.e., OAB) further support the presence of and early annual use of managed areas by other waterfowl species (see Chapter 1). Finally, nocturnal use of the managed areas was higher relative to diurnal use, throughout the monitoring period. This work serves as a baseline for the magnitude of differences as a function of season progression and should be in management planning. 108 APPENDIX APPENDIX: SPECIES/GROUP LIST UNKN: Unknown duck MALL: Mallard WODU: Wood duck ABDU: American black duck NOPI: Northern pintail AMWI: American wigeon GADW: Gadwall AGWT: American green-winged teal BWTE: Blue-winged teal TEAL: Unknown teal sp. NSHO: Northern shoveler RNDU: Ring-necked duck SCAUP: Scaup sp. BUFF: Bufflehead REDH: Redhead CANV: Canvasback RUDU: Ruddy duck COGO: Common goldeneye COME: Common merganser HOME: Hooded merganser CAGO: Canada goose AMCO: American coot MUSW: Mute swan TRUS: Trumpeter swan TUND: Tundra swan GREG: Great egret GBHE: Great blue heron GREB: Grebe sp. DEER: White-tailed deer 111 LITERATURE CITED LITERATURE CITED Ackerman, J. T., J. M. Eadie, and T. G. Moore. 2006. Does life history predict risk-taking behavior of wintering dabbling ducks? The Condor 108:530. . Baldassarre, G. A. 2014. Ducks, Geese, and Swans of North America. Johns Hopkins University Press, Baltimore, USA. Baldassarre, G. A., and E. G. Bolen. 1984. Field-Feeding Ecology of Waterfowl Wintering on the Southern High Plains of Texas. The Journal of Wildlife Management 48:63–71. Baldassarre, G. A., and E. G. Bolen. 2006a. Winter. Pages 249–276 in. Waterfowl Ecology and Management. Second. Krieger Publishing Company, Malabar, USA. Baldassarre, G. A., and E. G. Bolen. 2006b. Feeding Ecology. Pages 143–176 in. Waterfowl Ecology and Management. Second. Krieger Publishing Company, Malabar, USA. Beatty, W. S., D. C. Kesler, E. B. Webb, A. H. Raedeke, L. W. Naylor, and D. D. Humburg. 2014. The role of protected area wetlands in waterfowl habitat conservation: Implications for protected area network design. Biological Conservation 176:144–152. Elsevier Ltd. . Bengtsson, D., A. Avril, G. Gunnarsson, J. Elmberg, P. Söderquist, G. Norevik, C. Tolf, K. Safi, W. Fiedler, M. Wikelski, B. Olsen, and J. Waldenström. 2014. Movements, home-range size and habitat selection of mallards during autumn migration. PLoS ONE 9. Bookhout, T. A., K. E. Bednarik, and R. W. Kroll. 1989. The Great Lakes Marshes. Pages 131– 156 in L. M. Smith, R. L. Pederson, and R. M. Kaminski, editors. Habitat management for migrating and wintering waterfowl in North America. Texas Tech University Press, Lubbock. Bregnballe, T., and J. Madsen. 2004. Tools in waterfowl reserve management: Effects of intermittent hunting adjacent to a shooting-free core area. Wildlife Biology 10:261–268. Brooks, M. E., K. Kristensen, K. J. van Benthem, A. Magnusson, C. W. Berg, A. Nielsen, H. J. Skaug, M. Maechler, and B. M. Bolker. 2017. glmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling. The R Journal 9:378–400. Casazza, M. L., P. S. Coates, M. R. Miller, C. T. Overton, and D. R. Yparraguirre. 2012. Hunting influences the diel patterns in habitat selection by northern pintails Anas acuta. Wildlife Biology 18:1–13. Cowardin, L. M. 1969. Use of Flooded Timber by Waterfowl at the Montezuma National Wildlife Refuge. Journal of Wildlife Management 33:829–842. Cowardin, L. M., and J. E. Ashe. 1965. An Automatic Camera Device for Measuring Waterfowl Use. The Journal of Wildlife Management 29:636–640. Cowardin, L. M., V. Carter, F. C. Golet, and E. T. Laroe. 1979. Classification of Wetlands and Deepwater Habitats of the United States. Biological Service Program Report No. FWS/OBS-79/31. Washington, D.C., USA. Cox, R. R., and A. D. Afton. 1997. Use of Habitats by Female Northern Pintails Wintering in Southwestern Louisiana. The Journal of Wildlife Management 61:435–443. Dahl, T. E. 2011. Status and Trends of Wetlands in the Conterminous United States 2004 to 2009. Washington, D.C. Davis, J. B., M. Guillemain, R. M. Kaminski, C. Arzel, J. M. Eadie, and E. C. Rees. 2014. Habitat and resource use by waterfowl in the northern hemisphere in autumn and winter. Wildfowl 17–69. Dzubin, A. 1955. Some evidence of home range in waterfowl. Pages 278–298 in. Transactions of the North American Wildlife Conference. Evans, D. M., and K. R. Day. 2002. Hunting disturbance on a large shallow lake: The effectiveness of waterfowl refuges. Ibis 144:2–8. Firth, A. G., B. H. Baker, J. P. Brooks, R. Smith, R. B. Iglay, and J. Brian Davis. 2020a. Low external input sustainable agriculture: Winter flooding in rice fields increases bird use, fecal matter and soil health, reducing fertilizer requirements. Agriculture, Ecosystems and Environment. Volume 300. Firth, A. G., B. H. Baker, M. L. Gibbs, J. P. Brooks, R. Smith, R. B. Iglay, and J. B. Davis. 2020b. Using cameras to index waterfowl abundance in winter-flooded rice fields. MethodsX 7:101036. Elsevier B.V. . Fox, A. D., and J. Madsen. 1997. Behavioural and Distributional Effects of Hunting Disturbance on Waterbirds in Europe: Implications for Refuge Design. Source: Journal of Applied Ecology 34:1–13. . Accessed 29 Nov 2018. Fredrickson, L. H., and T. S. Taylor. 1982. Management of seasonally flooded impoundments for wildlife. Washington, D.C., USA. Fretwell, S. D., and H. L. Lucas. 1970. On territorial behavior and other factors influencing habitat distribution in birds. Acta Biotheoretica 19:16–36. 114 Gilbert, N. A., J. D. J. Clare, J. L. Stenglein, and B. Zuckerberg. 2020. Abundance estimation of unmarked animals based on camera-trap data. Conservation Biology 35:88–100. Hagy, H. M., and R. M. Kaminski. 2012. Winter waterbird and food dynamics in autumn- managed moist-soil wetlands in the Mississippi Alluvial Valley. Wildlife Society Bulletin 36:512–523. Hagy, H. M., J. N. Straub, M. L. Schummer, and R. M. Kaminski. 2014. Annual variation in food densities and factors. Wildfowl Special Issue:436–450. Houston, A. I., J. M. Mcnamara, and J. M. C. Hutchinson. 1993. General results concerning the trade-off between gaining energy and avoiding predation. Philosophical Transactions: Biological Sciences 341:375–397. Ivan, J. S., and E. S. Newkirk. 2016. CPW Photo Warehouse: A custom database to facilitate archiving, identifying, summarizing and managing photo data collected from camera traps. Methods in Ecology and Evolution 7:499–504. St. James, E. A., M. L. Schummer, R. M. Kaminski, E. J. Penny, and L. W. Burger. 2013. Effect of Weekly Hunting Frequency on Duck Abundances in Mississippi Wildlife Management Areas. Journal of Fish and Wildlife Management 4:144–150. . Jimenez, M. D., E. E. Bangs, D. K. Boyd, D. W. Smith, S. A. Becker, D. E. Ausband, S. P. Woodruff, E. H. Bradley, J. Holyan, and K. Laudon. 2017. Wolf dispersal in the Rocky Mountains, Western United States: 1993–2008. Journal of Wildlife Management 81:581– 592. Johnson, D. H. 1980. The Comparison of Usage and Availability Measurements for Evaluating Resource Preference. Ecology 61:65–71. Jones, J. 2001. Habitat Selection Studies in Avian Ecology: A Critical Review. The Auk 118:557–562. Kaminski, R. M., and J. Elmberg. 2014. An introduction to habitat use and selection by waterfowl in the northern hemisphere. Wildfowl 9–16. Knapik, R. T. 2019. Demographics and movements of mute swans in Michigan, USA. Michigan State University. Korschgen, C. E., and R. B. Dahlgren. 1992. Human Disturbances of Waterfowl: Causes, Effects, and Management. Fish and Wildlife Leaflet 13.2.15. Lancaster, J. D. 2013. Survival, habitat use, and spatiotemporal use of wildlife management areas by female mallards in Mississippi’s Alluvial Valley. Mississippi State University. 115 Lancaster, J. D., J. B. Davis, R. M. Kaminski, A. D. Afton, and E. J. Penny. 2015. Mallard use of a managed public hunting area in Mississippi. Journal of the Southeastern Association of Fish and Wildlife Agencies 2:281–287. Landers, J. L., T. T. Fendley, and A. S. Johnson. 1977. Feeding Ecology of Wood Ducks in South Carolina. The Journal of Wildlife Management 41:118–127. Lane, T. C. 2017. Multi-scale habitat selection by wintering waterfowl on Anahuac National Wildlife Refuge. Texas Tech University. Legagneux, P., C. Blaize, F. Latraube, J. Gautier, and V. Bretagnolle. 2009. Variation in home- range size and movements of wintering dabbling ducks. Journal of Ornithology 150:183– 193. Lenth, R. V. 2020. emmeans: Estimated Marginal Means, aka Least-Squares Means. R package version 1.5.3. Loesch, C. R., and R. M. Kaminski. 1989. Winter Body-Weight Patterns of Female Mallards Fed Agricultural Seeds. 53:1081–1087. Madsen, J. 1998. Experimental refuges for migratory waterfowl in Danish wetlands. II. Tests of hunting disturbance effects. Journal of Applied Ecology 35:398–417. Manly, B. F. J., L. L. McDonald, D. L. Thomas, T. L. McDonald, and W. P. Erickson. 2002. Resource selection by animals: Statistical design and analysis for field studies. Second. Kluwer Press, New York, New York, USA. . McDuie, F., M. L. Casazza, C. T. Overton, M. P. Herzog, C. A. Hartman, S. H. Peterson, C. L. Feldheim, and J. T. Ackerman. 2019. GPS tracking data reveals daily spatio-temporal movement patterns of waterfowl. Movement Ecology 7. . Moeller, A. K., P. M. Lukacs, and J. S. Horne. 2018. Three novel methods to estimate abundance of unmarked animals using remote cameras. Ecosphere 9. O’Connell, A. F., J. D. Nichols, and K. U. Karanth. 2011. Camera Traps in Animal Ecology: Methods and Analyses. Springer, Tokyo. Osborn, J. M., H. M. Hagy, M. D. McClanahan, J. B. Davis, and M. J. Gray. 2017. Habitat selection and activities of dabbling ducks during non-breeding periods. Journal of Wildlife Management 81:1482–1493. 116 Palumbo, M. 2017. Resource selection, survival, and departure of adult female mallards from the Lake St. Clair region during autumn and winter. The University of Western Ontario. Palumbo, M. D., S. A. Petrie, M. Schummer, B. D. Rubin, and S. Bonner. 2019. Mallard resource selection trade‐offs in a heterogeneous environment during autumn and winter. Ecology and Evolution 1–11. John Wiley & Sons, Ltd. . Accessed 20 Feb 2019. Pearse, A. T., R. M. Kaminski, K. J. Reinecke, and S. J. Dinsmore. 2012. Local and landscape associations between wintering dabbling ducks and wetland complexes in Mississippi. Wetlands 32:859–869. Pollander, K. M., A. R. Little, J. W. Hinton, M. E. Byrne, G. D. Balkcom, and M. J. Chamberlain. 2019. Seasonal habitat selection and movements by mottled ducks. Journal of Wildlife Management 83:478–486. Wiley-Blackwell. Prince, H. H., P. I. Padding, and R. W. Knapton. 1992. Waterfowl Use of the Laurentian Great Lakes. Journal of Great Lakes Research 18:673–699. Elsevier. . R Core Team. 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. . Reinecke, K. J., R. M. Kaminski, D. J. Moorhead, J. D. Hodges, and J. R. Nassar. 1989. Mississippi Alluvial Valley. Pages 203–248 in L. M. Smith, R. L. Pederson, and R. M. Kaminski, editors. Habitat management for migrating and wintering waterfowl in North America. Texas Tech University Press, Lubbock. Rice, M. B., and H. M. Hagy. 2020. Analysis of Integrated Waterbird Management and Monitoring Program Data : Effects of unit conditions, weather, and other factors on waterfowl abundance on National Wildlife Refuges. Rowcliffe, J. M., C. Carbone, P. A. Jansen, R. Kays, and B. Kranstauber. 2011. Quantifying the sensitivity of camera traps: an adapted distance sampling approach. Methods in Ecology and Evolution 2:464–476. John Wiley & Sons, Ltd (10.1111). . Accessed 10 Jul 2019. Schneider, S., S. Greenberg, G. W. Taylor, and S. C. Kremer. 2020. Three critical factors affecting automated image species recognition performance for camera traps. Ecology and Evolution 10:3503–3517. Shipes, J. C. 2014. Aspects of the ecology and management of mottled duck in Coastal South Carolina. Mississippi State University. 117 Shirkey, B. T., M. D. Palumbo, and J. W. Simpson. 2020. Land Cover Switching in Autumn by Female Mallards in Ohio. Journal of Wildlife Management 84:968–978. John Wiley & Sons, Ltd. . Stutzenbaker, C. D. 1988. The Mottled Duck: Its Life History, Ecology, and Management. Texas Parks and Wildlife Department Press, Austin, TX. Swanson, A., M. Kosmala, C. Lintott, and C. Packer. 2016. A generalized approach for producing, quantifying, and validating citizen science data from wildlife images. Conservation Biology 30:520–531. Thomas, D. L., and E. J. Taylor. 1990. Study Designs and Tests for Comparing Resource Use and Availability. The Journal of Wildlife Management 54:322–330. Turnbull, R. E., and G. A. Baldassarre. 1987. Activity Budgets of Mallards and American Wigeon Wintering in East-Central Alabama. The Wilson Bulletin 99:457–464. Vonbank, J. A., H. M. Hagy, and A. F. Casper. 2016. Energetic Carrying Capacity of Riverine and Connected Wetlands of the Upper Illinois River for Fall-Migrating Waterfowl. The American Midland Naturalist 176:210–221. University of Notre Dame. Warnock, N. 2010. Stopping vs. staging: The difference between a hop and a jump. Journal of Avian Biology 41:621–626. Wickham, H. 2016. ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York. . Yetter, A. P., H. M. Hagy, M. M. Horath, J. D. Lancaster, C. S. Hine, R. V Smith, and J. D. Stafford. 2018. Mallard survival, movements, and habitat use during autumn in Illinois; Mallard survival, movements, and habitat use during autumn in Illinois. The Journal of Wildlife Management 82:182–191. . Accessed 10 Jul 2019. 118 CHAPTER 3. SUCCESS TRENDS AND CONTRIBUTIONS OF MANAGED WATERFOWL HUNTING AREAS TO STATEWIDE DUCK HARVESTS IN MICHIGAN INTRODUCTION No other region in the world has as many waterfowl hunters as North America (Baldassarre and Bolen 2006a). The heavy pressure this puts on agencies managing public trusts to provide abundant populations and hunting opportunities (Baldassarre and Bolen 2006a) has contributed to over a century of hunting-influenced conservation. The Migratory Bird Treaty Act (MBTA; 1918), though it was not waterfowl focused legislation, provided the first effective conservation measure for waterfowl by ending spring/summer shooting and eliminating market hunting. After severe drought decimated wetland/waterfowl abundance during the 1930s (Baldassarre and Bolen 2006a), waterfowl conservation expanded through the United States Federal Migratory Bird Hunting and Conservation Stamp (i.e., duck stamp), the emergence of non-government organizations (e.g., Ducks Unlimited), and the passing of the Federal Aid in Wildlife Restoration Act of 1937 (i.e., The Pittman-Robertson Act; (Anderson et al. 2018)). Funds generated through duck stamp purchases (required to hunt waterfowl in the United States) are used for acquisition of waterfowl habitat. Pittman-Robertson dollars generated through a nationwide excise tax on firearms and ammunition are allocated to states on a cost-shared basis for natural resource related projects such as the acquisition and improvement of wildlife habitat, research into wildlife problems, and the development of access facilities for public use (Anderson et al. 2018). Since 1986, waterfowl conservation and management have been guided by the North American Waterfowl Management 119 Plan (NAWMP; 1986). During that time, regional joint ventures have protected/restored/enhanced millions of hectares of wetland habitat (Williams et al. 1999) using North American Wetlands Conservation Act (NAWCA) grants funded through leveraged Pittman-Robertson funds, MWTA fines, and annual appropriations (Anderson et al. 2018). Finally, in addition to federal appropriations, state hunting license (including state waterfowl hunting permits/stamps) monies are often matched with Pittman-Robertson funds or other grants, funneling money back into more localized level habitat conservation, further shaping the user- pay, user-benefit basis of the North American Model of Conservation (Organ et al. 2012). Michigan is located in the center of the Great Lakes Region along the upper Mississippi Flyway. With an abundance of wetland habitat associated with the Great Lakes, Lake St Clair., and numerous inland lakes and rivers, this region has been historically significant for migrating waterfowl (Bookhout et al. 1989). However, this region was not immune to the urban sprawl and agriculture expansion that took place across North America during the 20th century (Dahl and Allord 1996). Historic high rates of wetland loss would lead to the state retaining less than 50% of its historic wetland acreage by the 1980s (Tiner 1984), so Michigan Department of Natural Resources (DNR) began purchasing wetland acreage during the 1940s and 1950s in historically important locations (Michigan DNR, personal communication). Upon the increase in waterfowl hunter populations and growing demand for waterfowl hunting opportunity after World War II (Baldassarre and Bolen 2006a), the DNR began offering managed hunts on the consolidated wetlands in the 1960s and 1970s (Michigan DNR, personal communication). These areas would become collectively referred to as Managed Waterfowl Hunt Areas (MWHAs). Public hunting opportunity is quite common throughout the United States. Depending on location, anywhere from 30-77% of waterfowl hunters indicated that most of their hunting took 120 place on public land during a 2017 national survey (Slagle and Dietsch 2018a, b, c, d). Along the upper Mississippi Flyway, the majority of hunters (57%) indicated most of their hunting took place on public land or water (Slagle and Dietsch 2018a). This is to be expected given the public opportunity associated with the Great Lakes, Lake St. Clair, inland lakes and rivers, and state/federal land holdings. However, the goal of the Michigan’s MWHAs is to provide waterfowl hunting opportunity of higher quality than can be obtained in most settings. This is achieved annually by maximizing food availability for autumn migrating waterfowl through planting high energy agricultural foods (Baldassarre and Bolen 2006b) and practicing moist soil management (Fredrickson and Taylor 1982), as well as running controlled hunts throughout the season to reduce crowding and maintain safety. Assessments of waterfowl hunting on similarly operating managed public lands are relatively limited. An early work in Ohio measured reporting bias in public hunters and noted their levels of success over ten hunting seasons (Bednarik 1961). More recently, Stafford et al. (2010) used 21 years of hunter success data across 11 Illinois DNR sites and explored the influence of within-season factors (e.g., season dates) under management control and other factors (e.g., low temperature) outside of manger control and St. James et al. (2015) tested the effects of differing weekly hunt rates on hunter success rates across two hunting seasons in Mississippi. However, neither of these works’ focus was on areas’ specific annual and seasonal trends of success as Bednarik (1961) did. Local trends can be informative for area-specific management and provide indication of underlying regional phenomenon. Largely since the implementation of hunting programs, MWHA staff have collected data on hunter experiences. This provides information on more than four decades of public duck hunting in Michigan. While periodic works have helped to guide local management, it has been 121 many years since a comprehensive assessment of MWHA hunting has been completed. Additionally, a perception of declining hunting quality persists among many area users. As such the goal of this chapter is to examine characteristics of hunter use and success on MWHAs, with particular focus on (1) documenting historical annual harvest, hunter trips, and success rates, (2) determining changes in harvest totals and hunter success rates as a function of season progression, and 3) estimating contribution of MWHA duck and goose harvest to state waterfowl harvest. Study areas METHODS The focus of this chapter is the seven Managed Waterfowl Hunt Areas (MWHAs) operated by the Michigan DNR (Figure 3.1). 122 Figure 3.1. Locations of the seven Managed Waterfowl Hunt Areas in Michigan’s lower peninsula. All of these areas offer varying opportunity for waterfowl hunting (Table. 3.1). Weekly hunt schedules for the Fennville Farm Unit of the Allegan SGA, the Muskegon Wastewater Treatment Plant, and Pointe Mouillee changed at varying rates, while Fish Point, Harsens Island, Nayanquing Point, and Shiawassee have been consistent. While all area seasons are largely in line with Michigan’s southern waterfowl zone, the Fennville Farm Unit and the Muskegon 123 Wastewater Treatment Plant (WWTP) are regulated by the Allegan Goose Management Unit (GMU) and the Muskegon County Wastewater System GMU, respectively. Table. 3.1. Managed Waterfowl Hunt Areas’ property characteristics and available hunting opportunity. Area Hunting Periodsa Property Size (ha) Refuge Size (ha) Managed Hunting Additional Scramble Fennville Farm 1,659 546 Unit (Allegan SGA) Fish Point SWA 1,002 291 Tuesday PM Thursday PM Friday AM Saturday AM Sunday PM Entire Season AM/PM Harsens Island 1,358 121 Entire Season Unit AM/PM Muskegon WWTP Nayanquing Point SWA 1,416 607 Tuesday AM/PM Saturday AM/PM 609 161 Entire Season AM/PM Pointe Mouillee 2,102 381 Opening Day SGA AM/PM Sunday AM/PM Tuesday AM Thursday AM/PM Shiawassee River SGA 3,997 190 Entire Season AM/PM Zone Zones Breakdown 116 dry field zones None** 8 hunter capacity upland field 6 hunter capacity marsh 9 hunter capacity upland field 3 hunter capacity marsh Noneb 40 hunter capacity marsh Noneb 65 hunter capacity marsh 51 flooded field zones 5 marsh zones 29 flooded field zones 33 marsh zones 98 dry field zones 25 flooded field zones 8 flooded field zones 16 marsh zones 45 flooded field zones 22 marsh zones a Hunting schedule as of the 2019-2020 waterfowl season. b Additional marsh scramble areas are available on adjacent State Game Area land for hunting outside of the managed hunt draws. Hunters can have a maximum of 4 individuals (6 at Fennville Farm Unit) in their party. Party site selection order is determined through a lottery draw system. Current shot shell limits 124 are standardized across areas to a maximum of 25, though limits of 15 and 18 were previously implemented. Hunters must hunt in their designated zone (unless they are in a scramble area) and may not enter established waterfowl refuges at any time. Data collection and organization Upon completing a hunt, hunters must provide information on their experience. They either check their harvested birds with DNR staff at a check station or turn in a self-check card. Information on zone hunted, party size, wounding loss, and hours spent afield are also recorded. The Michigan DNR has collected this information since the 1960s, providing a long-term data set to track hunter activity and harvest on its managed areas. I obtained a harvest data set from the Michigan Department of Natural Resources detailing weekly hunter trips, duck harvest, and goose harvest at each area. I addressed inconsistencies in the data structure through conversations with area managers and by referring to annual hunting season reports. Data analyses I conducted all data work in R ver. 4.0.3 (R Core Team 2020). Annual use and success Though harvest data for these areas exist in some capacity dating back to the 1960s, some of my study sites did not exist/begin managed hunts until the mid-1970s. As such, I focused my analyses to those from the 1974–2019 hunting seasons. This is the time period where I was able to acquire hunting season measures for most areas on an annual total basis. Property size, hunting schedules and changes in both can influence total harvest and hunter trips. Thus, making comparisons of those measures alone was problematic. Measuring rates of success (i.e., ducks/geese per hunter trip) at these areas provides a way to better quantify hunting success, however, just observing rates of harvest can potentially blur magnitudes of total 125 harvest due the scaling effect. As such, I calculated measures of total harvest and hunter trips, as well as harvest corrected for hunter trips. I included data on early season goose hunts when harvest and trip totals were available. I did not include hunter trips associated with early goose season in my calculation of annual ducks per hunter trip. Early teal seasons took place during the latter part of my observation period (2014 – 2019). If data on harvest and hunter trips were both available, I included it in analyses. If either of these measures were unavailable, I did not include the other in a given years measurements. I portrayed area-specific duck harvest, goose harvest, and hunter trips using the ggplot2 package (Wickham 2016) in R. Additionally, I calculated annual ducks and geese harvested per hunter trip for each areas and used linear regression to evaluate trends in success rates. During my observation period (1974 – 2019) there was a clear transition from annually variable season lengths and bag limits (1 – 10 birds depending on species and order of harvest) associated with “Point System” (1974 – 1987) and restricted seasons with varying aggregate bags (1988 – 1996) to the largely consistent 60 day/ 6 duck season framework beginning in 1997. As such, I considered these periods (i.e., 1974 – 1996 and 1997 – 2019) separately rather than examining rates of duck harvest across the entire observation period. I did not consider any set of years independently for geese because of more complex and variable season and bag frameworks through the observation period, relative to ducks. Seasonal success Analyses of seasonal trend in area success were truncated to the 1997–2019 hunting seasons in alignment with Michigan’s adoption of a 60-day 6 ducks per day season under the Adaptive Harvest Management (AHM) framework. Additionally, annual reports during this 126 timeframe frequently reported harvest/hunter measures down finer than to an annual total level giving me additional capacity to proof weekly measures. While both ducks and geese were harvested at all of my study sites, for the purpose of these analyses I classified areas as predominately duck, goose, or both (Table 3.2). In creating this classification, I considered the habitat types associated with each managed area’s hunting zones (Table 3.1) as well as the number of ducks and geese harvested. I classified Harsens Island, Nayanquing, and Pointe Mouillee as duck areas. Their hunted zones are predominately flooded, and their goose harvests are generally approximately 100 birds or less. I classified Muskegon and the Fennville Farm (Allegan) as goose areas. These areas consist exclusively of dry field zones (with respect to the MWHA portions) and have noted some hunters actively choosing not to harvest ducks in the decoys while goose hunting (Michigan DNR, personal communication). While Muskegon’s total goose harvest in recent years has been similar to Harsens Island and Nayanquing (Michigan DNR, unpublished data), this area consists of dry fields to target geese and is not hunted at the frequency of Harsens Island and Nayanquing. I classified Fish Point and Shiawassee as both. While their hydrology in hunted zones more resembles that of the duck areas, both of these locations have relatively high goose harvests on an annual basis. I only considered duck and both areas for my duck per hunter trip analysis and I only considered goose and both areas for my geese per hunter trip analysis. Table 3.2. Classification of Managed Waterfowl Hunt Areas. Duck = areas with predominately flooded hunting zones where hunters primarily shoot ducks. Both = areas with predominately flooded hunting zones where hunters harvest large numbers of ducks and geese. Goose = areas with predominately dry hunting zones where hunters primarily shoot geese. Duck Both Goose Harsens Island Unit Fish Point SWA Fennville Farm Unit (Allegan) Nayanquing Point SWA Shiawassee River SGA Muskegon WWTP Pointe Mouillee SGA 127 I used season week and week number as measures of season progression. Season week is the numerical week of the hunting season at a given location, with number ranging from 1 through 15 and week number is the numerical week of the calendar year (i.e., 1 – 52). For season week, I used the period (typically Saturday – Friday) in which the data were aggregated. However, in some instances the season week contained less than or more than 7 days. These instances were commonly associated with season frameworks, splits, or levels of data aggregation. For week number, I used the international standard for describing week (Monday – Sunday). Because most hunting seasons begin on Saturdays, the range that harvest/hunter observations represent contains information from two-week periods. For consistency, I used the beginning date of a hunt period to compute week number. These measures, season week and week number are inherently correlated with each other. However, there were notable distinction that informed their respective use in this work. In determining ducks per hunter trips, I employed season week. All of locations of interest (duck/both areas) for this success metric were designated to Michigan’s Southern Waterfowl Zone throughout the duration of this objective’s observation range (1997–2019). Since all duck/both areas have the same season dates in a given year in the Southern Zone, this allowed season week to be a consistent measure of season progression rather than numerical week of year that is influenced by season start dates and slight annual variation in a given week’s classification. However, in determining geese per hunter day, I employed week number. This was because the Fennville Farm Unit and Muskegon are located within special Goose Management Units (GMUs) that have different season dates relative to the rest of Michigan’s Southern Waterfowl Zone. Additionally, the existence of splits at the county owned Muskegon WWTP creates a disconnect in the timing between its hunt periods and those 128 at the other goose areas. Thus, week number provide greater within and across year consistency compared to season week with regard to goose analyses. Managed waterfowl hunt areas harvest contribution To provide a coarse measure of MWHA harvest contribution, I queried total duck and goose harvest from all seven areas to determine annual MWHA harvest totals. I obtained annual estimates of Michigan’s state waterfowl harvest from the U.S. Fish and Wildlife Service Branch of Migratory Bird Management’s Parts Collection Survey, and species-specific estimates of total duck, total dabbler, and total geese harvest. For total ducks, I did not consider any species designated as Sea Ducks by the Sea Duck Joint Venture due to the dabbling duck focus of my study sites. I did, however, retain diving ducks due to their sporadic harvests at my study location. I then quantified the proportion of annual harvest estimates (i.e., MWHA harvest/USFWS state harvest estimate) that the MWHA’s represented for each group in every given year and portrayed those proportions using the ggplot2 package (Wickham 2016) in R. Annual use and success RESULTS Annual total hunter trips declined across all of Michigan’s MWHAs from 1974 to 2019, except for Pointe Mouillee (Figure 3.2; summed totals, Appendix 3.1). The increase in hunter trips at Pointe Mouillee is gradual but is the most consistent. The Fennville Farm Unit (AL) exhibited not only the sharpest decline in hunter trips from historic peaks to recent lows, but also showed the greatest amount of interannual variability. Fish Point, Harsens Island, and Shiawassee all experienced declines in annual hunter trips during the late 1980s/early 1990s relative to the early 1980s. Subsequent increases in hunter trips led to all-time highs that have 129 since gradually declined (Figure 3.2). Nayanquing Point and Muskegon declines have been relatively gradual over the long term (Figure 3.2). Figure 3.2. Area specific hunter trip totals (1974–2019). AL = Fennville Farm Unit, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee. Total ducks harvested at Fish Point, Harsens Island, Nayanquing Point, and Shiawassee declined from the late 1970s/early 1980s to the early 1990s. During the late 1990s and early 2000s, duck harvest totals at Fish Point and Shiawassee returned to levels comparable to those of the early 1980s, and Harsens Island set new records for annual harvest (Figure 3.3). However, in the past decade and a half, all of these areas have experienced gradual declines in total harvest. Nayanquing never recovered to the peak duck harvests from the early 1980s but has exhibited an uptick in harvest during the past decade (Figure 3.3). Fennville (AL), Muskegon, and Pointe Mouillee do not harvest as many ducks on an annual basis as the other MWHAs but have experienced much less annual fluctuation and lack long-term declines (Figure 3.3). 130 Figure 3.3. Annual Managed Waterfowl Hunt Area duck harvest totals (1974–2019). AL = Fennville Farm Unit, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee. Peak goose harvest for MWHAs took place during the late 1980s (Figure 3.4). The Fennville Farm Unit (Allegan), Fish Point, and Shiawassee all had multiple years with goose harvest greater than 2,000 birds. Since that period, these three areas exhibited declines in annual harvest totals. Muskegon’s goose harvests were relatively constant through the 1990s and early 2000s outside of annual volatility but have declined since then. Goose harvest at Harsens Island, Nayanquing, and Pointe Mouillee have always been low relative to the aforementioned areas (Figure 3.4). 131 Figure 3.4. Annual Managed Waterfowl Hunt Area goose harvest totals (1974–2019). AL = Fennville Farm Unit, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee. Annual measures of ducks per hunter trip for the duck focused areas largely fell between 0.6 and 1.5 for the entirety of the observation period (Figures 3.5 and 3.6). Annual measures of ducks harvested per hunter trip for the goose focused areas were generally less than 0.5. During the period leading up to 60 day/6 duck season, Fennville (β = -0.003), Nayanquing (β = -0.018), and Muskegon (β= -0.030) exhibited evidence of declining ducks per hunter trip with Muskegon and Nayanquing’s linear trends being statistically significant (p = <0.05). Concurrently, Fish Point (β = 0.008), Harsens Island (β = 0.011), Pointe Mouillee (β = 0.027), and Shiawassee (β = 0.010) showed evidence for positive linear trends in ducks harvested per hunter trip through the time period, with the trends at Harsens Island and Pointe Mouillee exhibiting statistical significance (p < 0.05). Though there is some evidence for directional trends in ducks harvested 132 per hunter trip since the transition to a 60 day/6 duck season framework (Figure 3.6), only the direction of Muskegon’s trend is statistically significant (p < 0.05). Figure 3.5. Annual measures and linear trends in ducks harvested per hunter trip on Managed Waterfowl Hunt Areas (1974–1996). AL = Fennville Farm Unit, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee. 133 Figure 3.6. Annual measures and linear trends in ducks harvested per hunter trip on Managed Waterfowl Hunt Areas (1997–2019). AL = Fennville Farm Unit, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee. The number of geese harvested per hunter trip has remained consistently low at Harsens Island, Nayanquing Point, and Pointe Mouillee (Figure 3.7). The Fennville Farm Unit (AL), Fish Point, Muskegon, and Shiawassee had higher rates of goose harvest (Figure 3.7), though these rates of harvest were lower than general rates of duck harvest across the areas (Figure 3.6). The direction of success rate trend was only certain at Fish Point (p = < 0.05). 134 Figure 3.7. Annual measures and linear trends in geese harvested per hunter trip on Managed Waterfowl Hunt Areas (1974–2019). AL = Fennville Farm Unit, FP = Fish Point, HI = Harsens Island, MU = Muskegon, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee. Seasonal success The best fitting model to describe seasonal change in ducks harvested contained an interaction between area and a factor measure of season week (Table 3.3). No model describing the ducks harvested as a function of area interacting with some trend of seasonal progression was competitive (Table 3.3). Table 3.3. Models describing seasonal progression of ducks harvested at select Managed Waterfowl Hunt Areas (1997 – 2019). Bold values are metrics for the top performing model. K = number of parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value, ω = respective model weight, Cumulative ω = weight of respective and all better fitting models, LL = log- likelihood. Model Area * K 47 AICc 111,961.20 Δ AICc 0.000 ω Cumulative ω LL 1.000 1.000 -55,933.46 Season Week (factor) Area * Season Week3 12 125,547.80 13,586.61 0.000 0.000 -62,761.77 135 Table 3.3. (cont’d) Area * Season Week2 Area * Season Week4 Area * Season Week 11 128,624.20 16,662.99 0.000 0.000 -64,300.98 13 129,580.80 17,619.56 0.000 0.000 -64,777.21 10 130,770.90 18,809.67 0.000 0.000 -65,375.34 Weekly estimates of duck harvest from the best fitting Area * Season Week (factor) model indicate highest harvests at Fish Point and Shiawassee at the beginning of the season, that subsequently dropped to 2nd and 3rd, respectively (Figure 3.8). After the 2nd week of season, Shiawassee harvested more birds than Fish Point on a weekly basis for the remainder of the season (Figure 3.8). Harsens Island was estimated to have the third highest opening week harvest, followed by highest weekly estimates for the rest of the season. Nayanquing and Pointe Mouillee were estimated to have the 4th and 5th highest weekly harvest estimates through the entire season (Figure 3.8). 136 Figure 3.8. Weekly estimates (Area * Season Week(factor)) of duck harvest at select Managed Waterfowl Hunt Areas (1997 – 2019). FP = Fish Point, HI = Harsens Island, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee. The model describing seasonal change in ducks per hunter trip as a function of an area specific interaction with a 3rd order polynomial season week trend was the best fitting (Table 3.4). The models containing linear and 4th order polynomial trend interactions with area were both also competitive (Table 3.4). Table 3.4. Models describing seasonal progression of ducks harvested per hunter trip at select Managed Waterfowl Hunt Areas (1997 – 2019). Bold values are metrics for the top performing model. K = number of parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value, ω = respective model weight, Cumulative ω = weight of respective and all better fitting models, LL = log-likelihood. Model Area * Season Week3 Area * Season Week Area * Season Week4 Area * Season Week (factor) Area * Season Week2 K 14 12 15 47 13 AICc 783.203 783.668 784.395 788.426 788.604 Δ AICc 0.000 0.465 1.192 5.223 5.401 ω Cumulative ω LL 0.403 0.319 0.222 0.030 0.027 0.403 0.722 0.943 0.973 1.000 -377.386 -379.674 -376.951 -344.816 -381.116 137 Estimates of ducks per hunter trip from the best fitting (Area * Season Week3) model show that while areas had differing levels of success rates throughout the season, the trend of success across areas were largely similar as a function of seasonal progression (Figure 3.9). Shiawassee showed less of a subsequent dip in success rates following the mid-season leveling off event experience across areas (Figure 3.9). Figure 3.9. Model estimates (Area * Season Week3) of seasonal ducks harvested per hunter trip at select Managed Waterfowl Hunt Areas (1997 – 2019). FP = Fish Point, HI = Harsens Island, NP = Nayanquing Point, PM = Pointe Mouillee, SH = Shiawassee. The best fitting model to describe seasonal change in geese harvested contained an interaction between area and a factor measure of annual week number (Table 3.5). No model describing the seasonal change in ducks harvested as a function of area interacting with an order of linear week number trend was competitive (Table 3.5). 138 Table 3.5. Models describing seasonal progression of geese harvested at select Managed Waterfowl Hunt Areas (1997 – 2019). K = number of parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value, ω = respective model weight, Cumulative ω = weight of respective and all better fitting models, LL = log-likelihood. Model Area * K 50 AICc 28,768.52 Δ AICc 0.000 ω Cumulative ω LL 1.000 1.000 -14,330.16 Week Number (factor) Area * Week Number4 Area * Week Number3 Area * Week Number2 Area * Week Number 11 31,410.23 2641.703 0.000 0.000 -15,693.91 10 32,616.01 3847.490 0.000 0.000 -16,297.84 9 8 33,399.07 4630.551 0.000 0.000 -16,690.40 33,533.02 4764.499 0.000 0.000 -16,758.40 Estimates of weekly goose were highest at Shiawassee during the earliest two-weeks wherein general goose seasons were open (40 and 41), but subsequently declined sharply across later week (Figure 3.10). A similar trend in harvest totals existed at Fish Point. No estimates of goose harvest were available for the Fennville Farm Unit (Allegan) during week 40. However, harvest estimates here were 2nd for week 40 then highest for the remainder of weeks with open seasons (Figure 3.10). Estimates of Muskegon goose harvest were never greater than 100 for any week period but showed more consistency in harvest relative to other locations (Figure 3.10). 139 Figure 3.10. Weekly estimates (Area * Week Number(factor)) of goose harvest at select Managed Waterfowl Hunt Areas (1997 – 2019). AL = Fennville Farm Unit, FP = Fish Point, MU = Muskegon, SH = Shiawassee. The model describing seasonal change in geese harvested per hunter trip as a function of an area specific interaction with a 3rd order polynomial season week trend was the best fitting (Table 3.6). The models containing an area and 4th order polynomial trend on week number was also competitive (Table 3.6). Table 3.6. Models describing seasonal progression of geese harvested per hunter trip at select Managed Waterfowl Hunt Areas (1997 – 2019). K = number of parameters, AICc = Akaike’s information criterion adjusted for small sample sizes, Δ AICc = change in the AICc relative to the model with the smallest AICc value, ω = respective model weight, Cumulative ω = weight of respective and all better fitting models, LL = log-likelihood. Δ AICc 0.000 1.038 18.414 31.133 32.724 Area * Week Number3 Area * Week Number4 Area * Week Number2 Area * Week Number 747.801 746.763 729.387 716.668 715.077 385.101 385.618 374.860 367.470 412.822 0.627 0.373 0.000 0.000 0.000 0.627 1.000 1.000 1.000 1.000 Model AICc ω Cumulative ω LL K 10 11 9 8 50 Area * Week Number (factor) 140 Estimates of geese harvested per hunter trip pulled from the best fitting (Area * Season Week3) model show that the dry field goose hunting areas (AL and MU) have higher rates of success throughout year progression than the areas with flooded hunting zones (FP and SH). Seasonal trends in success rates were very similar between Allegan and Muskegon in regard to timing in magnitude of change, with Fish Point also experiencing a comparable subsequent late season increases after initial declines from early season peak success rates (Figure 3.11). Geese harvested per hunter trip at Shiawassee exhibited a leveling off pattern after declines from early season rates rather than the subsequent increase of the aforementioned areas (Figure 3.11). Figure 3.11. Model estimates (Area * Week Number3) of seasonal geese harvested per hunter trip at select Managed Waterfowl Hunt Areas (1997 – 2019). AL = Fennville Farm Unit, FP = Fish Point, MU = Muskegon, SH = Shiawassee. Managed waterfowl hunt areas harvest contribution Since the implementation of a 60 day/6 duck seasons under the Adaptive Harvest framework (1997), cumulative duck and dabbling duck harvest at MWHAs account for 141 approximately 7–14% and 9–16% on Michigan’s annual harvest, respectfully. Although there is considerable year to year variation, no trend over time was apparent (Figure 3.12). During this same period, goose harvest on MWHAs has accounted for approximately 1–3% of the state’s annual totals (Figure 3.12). Figure 3.12. The proportion of Michigan’s annual state harvest totals taken on a Managed Waterfowl Hunt Area (1997–2019). DISCUSSION In maintaining harvest records at their MWHAs since the1960s, the Michigan DNR has created a robust dataset to track area hunter success. Examining this data provides insights on long term regional trends as well as MWHA harvest success and contribution to statewide harvest totals. 142 Annual use and success Hunter Use Hunter trips have largely declined across these managed areas. This was expected given the national decline in waterfowl hunters, which began during the mid-1970s (Vrtiska et al. 2013). Data on duck stamp sales shows a disproportionally large decline in waterfowl hunting participation among the Great Lakes states (MI, MN and WI; Singer 2014) compared to other Mississippi Flyway states. When duck hunting seasons and daily limits were liberalized during the late 1990s, Fish Point, Harsens Island, and Shiawassee mirrored national duck stamp sales with an initial increase in hunter trips followed by subsequent declines. However, in addition to broader societal trends, local regulations/conditions and season/daily limit frameworks can influence area specific trends. For example, Pointe Mouillee has exhibited a relatively stable positive trend. The stability of this trend is likely a component of the historically limited size (16 zones) and frequency (4 hunts per week) of this area’s managed hunts that likely mitigates annual volatility. The positive direction of this trend can potentially be attributed to an expansion of this areas hunting program (8 new zones and 1 additional hunt period) in 2014 as well as creation of new marsh zones prior to this. In contrast, Nayanquing, which is subject to the same weekly hunting schedule as Fish Point, Harsens Island, and Shiawassee, did not experience a comparable increase in hunter trips during the late 1990s. This could potentially be a function of local climatic factors, as Nayanquing generally experiences the earliest icing conditions that can limit hunter opportunity. The expansion of season lengths largely coincided with Nayanquing transitioning to Michigan’s Southern Waterfowl Zone (1996). Thus, the expansion of season length coupled with later opening dates potentially had a nullifying effect on each other. 143 At both goose focused areas, Fennville Farm Unit (Allegan) and Muskegon, hunting opportunity has been subject to more variable GMU season lengths (relative to the Southern Waterfowl Zones) and at times maximum harvest quotas (through the 2006 season). However, the Fennville Farm Unit has exhibited a stronger negative trend and greater annual variation than what has been observed at Muskegon. This could be a component of Fennville’s weekly hunting schedule being more impacted by loss/gain of hunting days relative to Muskegon’s. The elimination of harvest quotas at goose and duck/goose areas in the 2000’s and adoption of a fixed season length did not seem to increase goose harvest at these areas. Annual harvest totals and success rates Over my observation period (1977–2019), the harvest totals of an areas focal group (i.e., ducks or geese) largely tracked the hunter trips on that area. Fish Point, Harsens Island, and Shiawassee experienced declines in total duck harvest from the early 1980s to the early 1990s. These declines were followed by harvest total increases after expanded season frameworks during the late 1990s, and subsequent declines since the early 2000s. Nayanquing Point and Pointe Mouillee duck harvest totals followed the negative and positive trends in hunter trips for those areas, respectively. On the goose focused areas, harvest and hunter totals at the Fennville Farm Unit were routinely over 2,500 and 8,000, respectively during the late 1980s and early 1990s. Harvest or trip totals have not reached these respective magnitudes since, but continue to track each other in low (e.g., 1998 and 1999) and high (e.g., 2003 and 2006) years of opportunity associated with season length. Concurrently, Muskegon has not seen such drastic changes on an annual basis, but higher hunter trip totals were associated with higher levels of goose harvest (e.g., 1995 and 2006). Both of these areas were historically subject to annual maximum harvest quotas. These quotas were routinely not met and thus not a common limiting factor for harvest. 144 Instances of harvest not tracking hunter trips, however, can be seen in the goose harvest at Fish Point and Shiawassee which did not increase with hunter trips during the late 1990s and 2000s. Hunting schedules have not changed at either of these locations, so this is potentially a function of local conditions. Fish Point is largely surrounded by agriculture and increased efficiency of farming equipment, as noted by Krapu et al. (2004) could be potentially influencing the holding capacity of birds in the local area. While goose harvest at Shiawassee is also potentially impacted by increases in local farming efficiency, the gradual transition away from farming at the neighboring Shiawassee NWR could also be contributing to declines in harvest totals on the state game area due to fewer birds being held immediately adjacent to hunting zones. Annual measures of ducks harvested per hunter trip ranged between 0.5 and 1.5. These rates of harvest are largely comparable to those noted at other hunting areas (Bednarik 1961, Hamer and Arthur 1976, Roetker and Anderson 1977, Thornburg and Allen 1979, Stafford et al. 2010). Annual harvest rates were higher (1.54 ± 0.26 to 1.77 ± 0.25) on managed areas in a recent study in Mississippi (St. James et al. 2015), however, these areas were hunted at lower frequency than my study sites (2 or 4 half day hunts/week) and had harvest compositions predominately comprised of species (e.g., green-winged teal and northern shovelers) more susceptible to harvest relative the mallard dominated harvests at my areas (Michigan DNR, unpublished data). Annual variability in hunter success rates has been associated with breeding populations as well as local temperatures (Stafford et al. 2010). Other locals factors (e.g., drought and flood) can influence waterfowl distribution (Nichols et al. 1983) and in turn hunter harvest. Prior to the transition to 60 day/6 duck seasons, I observed positive linear trends in success at Harsens Island and Pointe Mouillee. This indicates that even though hunter trips and total harvest were declining, the quality of hunting (through a ducks per hunter trip metric) was 145 increasing at these locations. Concurrently, there was declines in success rates at Nayanquing Point and Muskegon. Since the transition to 60 day/6 duck seasons, there has not been evidence for directional trends in harvest at any of the duck focused areas. However, the goose focused Muskegon has exhibited a positive trend in success rates. Across the whole monitoring period, duck harvest rates at Fennville remained low but consistent with regard to lack of a directional linear trend. Measures of duck harvest indicated lack of directional long-term linear trend even while regional breeding populations have declined (U.S. Fish and Wildlife Service 2019) and managers have faced difficulties associated with droughts and floods. With regard to goose harvest per hunter trip, there was strong evidence for a subtle negative linear trend (-0.002) in goose harvest rates at Fish Point. This supports the declines in total goose harvest on this property not just being associated with declining hunter trips, but potentially declining local capacity to hold birds due to changing agricultural practices (Krapu et al. 2004). Finally, while, there was evidence for subtle positive linear trends (<0.002) in goose harvest per hunter trip at Fennville, Muskegon and Shiawassee, the variability in the data led to these trends not being statistically significant, and in my interpretations, not biologically or socially significant. Seasonal success Both total duck harvest and ducks harvested per hunter trip were highest during the first week of seasons. Early season success is to be expected, given the abundance of naive locally raised birds (e.g., mallards, and wood ducks) as well as risk taking early migrants such as blue- winged teal. While some human disturbance occurs on these areas prior to general duck season (e.g., early hunting season, boating, bird watching, etc.), it is not of a comparable degree to the general season disturbance. While magnitudes varied, pattern of success rates were largely 146 consistent among areas. The immediate decline in success rates is a likely a function of rapid conditioning of birds to hunted and non-hunted locations. Mid-season leveling off in total harvest and success rates align with higher abundances of birds on the managed areas (see Chapter 1; Michigan DNR, unpublished data) and when the periods of low temperatures that Stafford et al. (2010) noted increased hunter success are expected. However, the slight leveling off of success rate rather than a noticeable increase, indicates that many birds are not available to hunters (see Chapter 2), due to local conditioning and conditioning that took place up the flyway. This perceived lack of success later in the season could further contribute to the seasonal decline in hunter trips and total duck harvest, as recreationalists must already weigh time allocation between duck hunting and other hunting activities (e.g., deer hunting) as well as non-hunting activities (e.g., major sporting events). The differing seasonal pattern of hunter success at Shiawassee relative to the other areas could potentially be due to the lotic hydrology of the river system which delays the onset of ice formation in many hunting zones. Goose harvest rates on managed areas showed a bimodal pattern of success with regard to season progression. Initial early peaks were likely sustained by the high abundance of locally raised temperate breeding Canada geese (Branta canadensis), and subsequent declines are likely associated with displacement from hunted locations (Jensen et al. 2016). Just as with ducks, patterns of success rates change with season progression, as weather up the flyway pushed non- local birds into the region. Later timing in trend change (mid-November relative to late October) is likely due the prevalent field feeding strategy of geese (Baldassarre 2014) that allows them to persist in locations after other wetland dependent duck species (e.g., gadwall) have moved on. Annual managed hunting programs at Fish Point, Muskegon, and Shiawassee conclude (i.e., harvest of 0) prior to the second peak of the bimodally distributed model-based success rate 147 estimates. However, increases in total weekly goose harvest on the dry fields at Fennville (AL) in December, corroborate and are likely driving these model-based estimates of increased success rates. Managed waterfowl hunt areas harvest contribution My measures of state harvest indicated that the MWHAs provide consistent and important locations for hunters to harvest ducks and geese. Glimer et al. (1989) noted numerous public hunting areas combined to contribute 4–16 percent of California’s annual harvest, and while the percentages overlap with those noted in this study, Michigan’s MWHAs annual contributions to statewide harvests were not as variable. Furthermore, given that duck harvest at Muskegon and the Fennville Farm Unit (Allegan) are relatively low, Michigan’s five duck focused areas are driving the annual proportions. The MWHAs have not contributed as high of a proportion of the state goose harvest. This was to be expected given the dry field feeding life history trait of Canada geese (Baldassarre 2014) and the abundance of agriculture in Michigan. Also, the locally-breeding population of geese in Michigan grew rapidly during the 1990’s and this contributed to high goose harvest off the MWHAs. While dry field areas exist on other MWHAs, the Fennville Farm Unit and Muskegon WWTP these upland areas made up all the managed hunt area. These areas consistently offer hunters, who do not have access to private land, the opportunity to dry field hunt for geese. Other considerations As with annual and seasonal progression, hunter trips and success vary with zone-specific characteristics. All areas have numeric zones (Appendix 3.2) with some common perennial “hot spots”. These locations generally correlate with high food availability (e.g., crops or moist soil) and are in closer proximity to the refuge. However, I did not consider zone specific differences 148 for the purpose of this thesis. This decision was due to my goal of assessing broader trends in annual and seasonal waterfowl harvest across these areas rather than zone specific nuances. Area managers routinely examine zone specific harvest and use data for a given season to inform management for the subsequent year, as such it is unlikely that my findings from coarse annual analyses would provide much additional information specific to zone harvests. I did not directly analyze the influence of weather measures on hunter harvest. The influence of various weather conditions on bird movements/distributions (Nichols et al. 1983, Schummer et al. 2010, Van Den Elsen 2016, O’Neal et al. 2018) and hunter success (Stafford et al. 2010, Johnson and Vrtiska 2014) are well documented and generally coincide with season progression. MANAGEMENT IMPLICATIONS The Michigan DNR, MWHAs are located in historically important staging locations for autumn migrating waterfowl (see Chapter 1; Bookhout et al. 1989), and have offered managed hunting opportunity for over 50 years. Total annual hunter trips on areas that have maintained comparable hunt schedules (Fish Point, Harsens Island, Nayanquing, and Shiawassee) as well as the Fennville Farm Unit have declined over time, coincident with an underlying general decline in hunters (Vrtiska et al. 2013). However, area specific conditions (i.e., climatic) outside of societal and hunt schedule change have the capacity to impact hunter trips. Beginning in the 2021 waterfowl season, Nayanquing and Fish Point will be transitioning to the Middle Zone. The early start to seasons will potentially allow these areas to capitalize on early abundances (see Chapter 1), while minimizing days lost to ice up. 149 Although there has been annual variation and there are area specific nuances that influence success (e.g., Fish Point goose harvest), the present rate of hunter success is largely comparable to that of the late 1970s and early 1980s when the hunting was perceived by some to be drastically better. This indicates that declines in total harvest are more a function of participation rather than success rates (i.e., hunt quality). Analysis of seasonal duck harvest and rates of harvest indicated a disconnect between the timing of harvest metric peaks and peak waterfowl abundances (see Chapter 1). This could be associated with strong conditioning on the ducks to avoid hunter areas during diurnal hours (see Chapter 2). Pointe Mouillee is hunted at a lesser frequency than the other duck focused areas, however it did not exhibit a different pattern of harvest rates. Given that this area has comparable habitat configurations (i.e., flooded ag, moist soil, and emergent marshes) to the other duck focused areas, this could indicate that providing additional hunting opportunity here/restricting other locations would have little influence on hunter success. However, the small size of Pointe Mouillee’s managed hunting area (182 ha) relative to larger adjacent sections (737 ha) of the SGA that were open to hunting seven days a week have a confounding effect on disturbance between this and other duck focused MWHAs. This potential lack of independence between disturbance levels on and off of a game area was noted by St. James et al. (2015). Additional research on this is warranted. Goose harvests rates early in the year indicate high abundance and subsequent conditioning of temperate-breeding and early molt migrant Canada geese. Increases in late season increases success rates align with influx of migrant geese are nearing peak abundances of the goose focused Fennville Farm Unit and Muskegon. This provides additional support for the influence weather has on goose migration and subsequent harvest (Johnson and Vrtiska 2014). 150 Continuing to provide goose hunting opportunity December – early February will allow hunters to take advantage of ideal weather conditions for hunting these dry field systems. Three of the Michigan DNR’s goals are to manage for healthy and sustainable wildlife populations, and manage habitats for sustainable wildlife populations and wildlife-based recreation, and enhance sustainable wildlife-based recreation use and enjoyment (Michigan DNR 2016). The MWHAs meet these goals in that they provide habitat, not only for autumn migrating waterfowl species, but also spring migrating waterfowl, summer breeding birds, and year-round habitat for other non-waterfowl species. Staff on these areas (with the help of sharecroppers and local support clubs) promote the areas’ capacity to support large abundances of autumn migrating waterfowl through their active management. Furthermore, these areas promote wildlife based recreation through managed waterfowl hunting programs, but also commonly support an abundance of opportunity for other wildlife and non-waterfowl hunting recreation (Nelson et al. 2007). The multi-purpose aspect of these areas are important given the growing inaccessibility of places to hunt (Eliason 2020). Given that the MWHAs are meeting DNR goals annually, drastic changes in management plans are not required. Certainly, these areas have provided consistent sources of habitat for wildlife, as well as opportunity for successfully waterfowl hunts. However, if the objective of MWHAs is to maximize quality hunting, providing less frequent opportunity could help contribute to greater opportunity for success. While many aspects (e.g., enjoying nature and the outdoors, seeing good behavior from other hunters, and seeing a lot of ducks and geese) contribute to the quality of experience had by waterfowl hunter (Enck et al. 1993, Schroeder et al. 2006, Frawley et al. 2017), it has been suggested that harvest is an important aspect to hunters, particularly those who are less avid (Brunke and Hunt 2008, Slagle and Dietsch 2018a, 151 Schroeder et al. 2019, Schummer et al. 2019, 2020). Changes in hunt plans, however, should not be based on the insight of this work alone. However, considering the influence of disturbance (among other factors) on waterfowl habitat and in turn hunter success, reduced hunting pressure might contribute to a less disproportionate nocturnal vs diurnal use of the properties (Chapter 2), as well as better aligning the timing of peak harvest rates with peak waterfowl abundance (Chapter 1), that in turn promote greater levels of hunter success. 152 APPENDICES APPENDIX 3.1: ANNUAL TOTALS Figure A.3.1.1. Total annual hunter trips on Managed Waterfowl Hunt Areas (1974 – 2019). Data available for all areas beginning in 1985. Figure A.3.1.2. Total annual duck harvest on Managed Waterfowl Hunt Areas (1974 – 2019). Data available for all areas beginning in 1985. 155 Figure A.3.1.3. Total annual goose harvest on Managed Waterfowl Hunt Areas (1974 – 2019). Data available for all areas beginning in 1985. 156 APPENDIX 3.2: MANAGED WATERFOWL HUNT AREAS MAPS Figure A.3.2.1. Fennville Farm Unit managed hunting map. 157 Figure A.3.2.2. Fish Point managed hunting map. 158 Figure A.3.2.3. Harsens Island Unit managed hunting map. 159 Figure A.3.2.4. Muskegon managed hunting map. 160 Figure A.3.2.5. Nayanquing Point managed hunting map. 161 Figure A.3.2.6. Pointe Mouillee managed hunting map. 162 Figure A.3.2.7. Shiawassee managed hunting map 163 LITERATURE CITED LITERATURE CITED Anderson, M. G., R. T. Alisauskas, B. D. J. Batt, R. J. Blohm, K. F. Higgins, M. C. Perry, J. K. Ringelman, J. S. Sedinger, J. R. Serie, D. E. Sharp, D. L. Trauger, and C. K. Williams. 2018. The migratory bird treaty and a century of waterfowl conservation. Journal of Wildlife Management 82:247–259. Baldassarre, G. A. 2014. Ducks, Geese, and Swans of North America. Johns Hopkins University Press, Baltimore, USA. Baldassarre, G. A., and E. G. Bolen. 2006a. Introduction and historical overview. Pages 1–16 in G. A. Baldassarre and E. G. Bolen, editors. Waterfowl Ecology and Management. Second. Krieger Publishing Company, Malabar, USA. Baldassarre, G. A., and E. G. Bolen. 2006b. Feeding Ecology. Pages 143–176 in. Waterfowl Ecology and Management. Second. Krieger Publishing Company, Malabar, USA. Bednarik, K. 1961. Waterfowl Hunting on a Controlled Public Area. Columbus, Ohio, USA. . Accessed 12 Nov 2018. Bookhout, T. A., K. E. Bednarik, and R. W. Kroll. 1989. The Great Lakes Marshes. Pages 131– 156 in L. M. Smith, R. L. Pederson, and R. M. Kaminski, editors. Habitat management for migrating and wintering waterfowl in North America. Texas Tech University Press, Lubbock. Brunke, K. D., and K. M. Hunt. 2008. Human Dimensions of Wildlife Mississippi Waterfowl Hunter Expectations, Satisfaction, and Intentions to Hunt in the Future. . Accessed 2 Dec 2018. Dahl, T. E., and G. J. Allord. 1996. History of wetlands in the conterminous United States. Pages 19–26 in J. D. Fretwell, J. S. Williams, and P. J. Redman, editors. National Water Summary on Wetland Resources. U.S. Geological Survey Water-Supply Paper 2425. Eliason, S. L. 2020. A place to hunt: some observations on access to wildlife resources in the western United States. Human Dimensions of Wildlife 00:1–11. Routledge. . Van Den Elsen, L. M. 2016. Weather and Photoperiod Indices of Autumn and Winter Dabbling Duck Abundance in the Mississippi and Atlantic Flyways of North America. The University of Western Ontario. . Accessed 15 Apr 2020. Enck, J. W., B. L. Swift, and D. J. Decker. 1993. Reasons for Decline in Duck Hunting: Insights from New. Source. Volume 21. Wildlife Society Bulletin. . Accessed 13 Dec 2018. Frawley, B. J., B. Avers, and B. Rudolph. 2017. Survey of Waterfowl Hunters Using Managed Waterfowl Hunt Areas in Michigan. Fredrickson, L. H., and T. S. Taylor. 1982. Management of seasonally flooded impoundments for wildlife. Washington, D.C., USA. Glimer, D. S., J. M. Hicks, J. P. Fleskes, and D. P. Connelly. 1989. Duck harvest on public hunting areas in California. California Fish and Game 75:155–168. Hamer, V. H., and G. C. Arthur. 1976. Hunter use and harvest on public waterfowl areas during 1975. Illinois Department of Conservation Migratory Bird Section - Periodic Report. Volume 14. Singer, H.V. 2014. Factors affecting productivity and harvest rates of Great Lakes mallards. Michigan State University. St. James, E. A., M. L. Schummer, R. M. Kaminski, E. J. Penny, and L. W. Burger. 2015. Effect of Weekly Hunting Frequency on Rate of Ducks Harvested. Journal of Fish and Wildlife Management 6:247–254. Journal of Fish and Wildlife Management . Accessed 7 Nov 2018. Jensen, G. H., J. Madsen, and I. M. Tombre. 2016. Hunting migratory geese: Is there an optimal practice? Wildlife Biology 22:194–203. Johnson, H. M., and M. P. Vrtiska. 2014. Weather Variables Affecting Canada Goose Harvest in Nebraska. Great Plains Research 24:135–143. Krapu, G. L., D. A. Brandt, and R. R. Cox. 2004. Less waste corn, more land in soybeans, and the switch to genetically modified crops: trends with important implications for wildlife management. Wildlife Society Bulletin 32:127–136. Michigan DNR. 2016. Guiding principles and strategies: wildlife division strategic plan 2016- 2020. Nelson, C., E. Steffey, E. Clark, K. Steger, K. Danforth, and R. Studies. 2007. Michigan State Game Area Use and User Assessment: March 15 - December 15, 2006. East Lansing, Michigan, USA. Nichols, J. D., K. J. Reinecke, and J. E. Hines. 1983. Factors Affecting the Distribution of Mallards Wintering in the Mississippi Alluvial Valley. The Auk 100:932–946. . Accessed 5 Jun 2020. 166 North American Waterfowl Management Plan. 1986. . Accessed 21 Nov 2018. O’Neal, B. J., J. D. Stafford, R. P. Larkin, and E. S. Michel. 2018. The effect of weather on the decision to migrate from stopover sites by autumn-migrating ducks. Movement Ecology 6. . Accessed 28 Jan 2019. Organ, J. F., V. Geist, S. P. Mahoney, S. Williams, P. R. Krausman, G. R. Batcheller, T. A. Decker, R. Carmichael, P. Nanjappa, R. Regan, R. A. Medellin, R. Cantu, R. E. Mccabe, S. Craven, G. M. Vecellio, and D. J. Decker. 2012. The North American Model of Wildlife Conservation. The Wildlife Society Technical Review 12-04. Bethesda, Maryland, USA. . Accessed 21 Nov 2018. R Core Team. 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. . Roetker, F., and W. L. Anderson. 1977. Hunter use and harvest on public waterfowl areas during 1976. Illinois Department of Conservation Migratory Bird Section - Periodic Report 17:1–8. Urbana, Illinois, USA. Schroeder, S. A., D. C. Fulton, L. Cornicelli, S. D. Cordts, and J. S. Lawrence. 2019. Clarifying how hunt‐specific experiences affect satisfaction among more avid and less avid waterfowl hunters. Wildlife Society Bulletin 43:455–467. Schroeder, S. A.., D. C. Fulton, J. S. Lawrence, P. Reviewed, S. a. Schroeder, D. C. Fulton, and J. S. Lawrence. 2006. Managing for Preferred Hunting Experiences: A Typology of Minnesota Waterfowl Hunters. Wildlife Society Bulletin 34:380–387. . Schummer, M. L., R. M. Kaminski, A. H. Raedeke, and D. A. Graber. 2010. Weather-Related Indices of Autumn–Winter Dabbling Duck Abundance in Middle North America. Journal of Wildlife Management 74:94–101. Schummer, M. L., J. Simpson, J. B. Davis, B. Shirkey, and K. E. Wallen. 2020. Balancing Waterfowl Hunting Opportunity and Quality to Recruit, Retain, and Reactivate. Wildlife Society Bulletin 44:391–395. Schummer, M. L., A. M. Smith, R. M. Kaminski, K. M. Hunt, E. St. James, and H. Havens. 2019. Achievement-Oriented Effects on Waterfowl-Hunt Quality at Mississippi Wildlife Management Areas. Journal of the Southeastern Association of Fish and Wildlife Agencies 6:129–135. 167 Slagle, K., and A. Dietsch. 2018a. National Survey of Waterfowl Hunters: Summary Report Mississippi Flyway 2018. Report to the National Flyway Council from the Minnesota Cooperative Fish and Wildlife Research Unit, University of Minnesota and The Ohio State University. St. Paul, MN. . Accessed 2 Dec 2018. Slagle, K., and A. Dietsch. 2018b. National Survey of Waterfowl Hunters: Summary Report Atlantic Flyway 2018. Report to the National Flyway Council from the Minnesota Cooperative Fish and Wildlife Research Unit, University of Minnesota and The Ohio State University. St. Paul, MN. Slagle, K., and A. Dietsch. 2018c. National Survey of Waterfowl Hunters: Summary Report Central Flyway 2018. Report to the National Flyway Council from the Minnesota Cooperative Fish and Wildlife Research Unit, University of Minnesota and The Ohio State University. St. Paul, MN. Slagle, K., and A. Dietsch. 2018d. National Survey of Waterfowl Hunters : Summary Report Pacific Flyway. Report to the National Flyway Council from the Minnesota Cooperative Fish and Wildlife Research Unit, University of Minnesota and The Ohio State University. St. Paul, MN. Stafford, J. D., A. T. Pearse, C. S. Hine, A. P. Yetter, and M. M. Horath. 2010. Factors associated with hunter success for ducks on state-owned lands in Illinois, USA. Wildlife Biology 16:113–122. Thornburg, D., and W. Allen. 1979. Hunter use and harvest on public waterfowl areas during 1978. Illinois Department of Conservation Migratory Bird Section - Periodic Report 25:1–9. Urbana, Illinois, USA. Tiner, R. W. 1984. Wetlands of the United States: current status and recent trends. U.S. Fish and Wildlife Service. 2019. Waterfowl population status, 2019. Washington, D.C. Vrtiska, M. P., J. H. Gammonley, L. W. Naylor, and A. H. Raedeke. 2013. Economic and conservation ramifications from the decline of waterfowl hunters. Wildlife Society Bulletin 37:380–388. Wickham, H. 2016. ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York. . Williams, B. K., M. D. Koneff, and D. A. Smith. 1999. Evaluation of Waterfowl Conservation under the North American Waterfowl Management Plan. The Journal of Wildlife Management 63:417–440. 168