UNDERSTANDING MANAGEMENT AND LANDSCAPE INFLUENCES ON THE HARVEST O F MALE WHITE - TAILED DEER ACROSS A LARGE GEOGRAPHIC REGION By Rebecca L ynne Cain A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Fisheries and Wildlife Doctor of Philosophy 2020 ABSTRACT UNDERSTANDING MANAGEMENT AND LANDSCAPE INFLUENCES ON THE HARVEST OF MALE WHITE - TAILED DEER ACROSS A LARGE GEOGRAPHIC REGION B y Rebecca Lynne Cain The North American Model of Wildlife Conservation relies on the active participation of citizen hunters to achieve management goals. One factor that motivates hunters to become active participants is an opportunity to harvest a mature white - tailed deer ( Odocoileus virginianus ) with large antlers, especially the case for achievement - oriented wildlife recreation ists. Variation in antler conformation and size among white - tailed deer is noticeable across landscapes. Moreover, when mapped , there is obvious spatial heterogeneity in the harvests of record deer (e.g., deer with large antler s that qualify for entry in t he Boone and Crockett records) across the United States, with the majority of entries coming from the Midwestern region. This dissertation should engage the interests of wildlife biologists and researchers. Chapter 1 focuses on testing hypotheses about ha rvest outcomes for antler point restrictions in the state of Michigan. Chapter 2 evaluates spatially explicit trends in antler sizes of record deer across the Midwestern United States. Chapter 3 evaluates the degree to which management regulations influenc ed the harvest of record deer in the Midwest United States. Chapter 4 focuses on potential issues related to reporting bias and proposes an adaptation of N - mixture models to account for imperfect detection. Findings from this research include : 1) the impo rtance of spatial context when evaluating trends in harvest data across a large geographic region; 2) antler point restrictions do indee d protect yearling males from harvest and advance the age structure of male harvest; 3) implementing antler point restri ctions did not increase antlerless harvest or change the trajectory in hunter numbers; 4) antler sizes of record deer in the Midwest showed increasing trends; 5) harvests of record deer were greater in areas with management regulations that restricted the buck harvest; 6) more record deer were reported when at least 1 record deer was reported the previous year ; 7) detection of harvests of record deer do not follow any spatial or temporal pattern. As interest in quality deer management and harvesting adult males with large antlers increases , it is important for wildlife managers and hunters to understand how regulations can influence harvests of record deer. My work offers insights into the relationships between management strategies and harvest outcomes. T h is research provides managers important information about factors affecting harvests of record deer, outcomes of management regulations, and inherent differences in record deer harvests and characteristics among ecoregions . Managers can draw on the insight s gained from this dissertation research during the decision - making process when setting annual hunting regulations, as well as communicating reasonable expectations for deer populations to hunters and other interested stakeholder groups. iv ACKNOWLEDGEMENTS Undertaking this Ph .D. has been a life - changing experience for me. This dissertation is the culmination of years of hard work that would not have been possible without the encouragement and support I received from many people. I would par ticularly like to thank my supervisors, Bill Porter and David Williams, for their professional and personal guidance and for con tinual ly pushing me to reach my potential. I am especially grateful to them for allowing me to grow not only as a scientist but also as a teacher, leader, and manager. I would also like to thank my advisory committee, Mike Everett, Shawn Riley, and Gary Roloff, for asking challenging questions that forced me to think about this research from various perspectives and for providing h elpful feedback on ideas and written documents . I w ant to offer my special thanks to Garrett Knowlton for his assistance in collecting and processing data used in this research. Without the Official Measurers of the Boone and Crockett Club, this dissertati on project would not have been possible. I w ant to express my gratitude to these Official Measurers for their dedication to measuring, which contributed to the data necessary to answer the questions posed in this research. I owe many thanks to Justin Sprin g for his interest and involvement in this research . I am thankful to Chad Stewart and Brent Rudolph for their constructive comments, time, and encouragement. I am grateful to the Michigan Department of Natural Resources, the U.S. Fish and Wildlife Service through the Pittman - Robertson Wildlife Restoration Act Grant MI W - 155 - R, the Boone & Crockett Quantitative Wildlife Center at Michigan State University, and the v Michigan Involvement Committee of Safari Club International for financial support and assistan ce in completing this project. I w ant to offer sincere appreciation to Rose Stewart and Joanne Crawford for providing support, mentoring, critique, an d a much - needed perspective at critical times during my Ph .D. program. My thanks and appreciation to my g raduate student colleagues in the Fisheries and Wildlife D epartment, and especially my fellow lab members in the Boone and Crockett Quantitative Wildlife Center, for meaningful discussions, constructive feedback, constant encouragement, and wonderful camar aderie. Most importantly, thank you to my family. Words cannot describe how grateful I am for the unconditional love and support of all my relatives during my time at Michigan State University. I must express my heartfelt appreciation to my parents for the encouragement and life lessons that made me who I am. I thank my grandparents for their unwavering belief in me and my brother, Patrick, for keeping me grounded. I am indebted to Andrew Dennhardt, Amanda Dolinski, Samantha Jefferson, and Lauren Phillips f or being there when I needed you. Lastly, I owe my deepest gratitude to my dog, Dixie, for her endless devotion, companionship, and comfort during the research and writing process. vi TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ .................... viii LIST OF FIGURES ................................ ................................ ................................ ...................... x PROLOGUE ................................ ................................ ................................ ................................ . 1 LITERATURE CITED ................................ ................................ ................................ ................. 4 CHAPTER 1: EVALUATING THE EFFECTIVENESS OF ANTLER POINT RESTRICTIONS TO ACHIEVE WHITE - TAILED DEER MANAGEMENT GOALS ............ 7 INTRODUCTION ................................ ................................ ................................ .................. 7 STUDY AREA ................................ ................................ ................................ ....................... 9 METHODS ................................ ................................ ................................ ........................... 13 Data Collection ................................ ................................ ................................ ............... 13 Statistical Analysis ................................ ................................ ................................ .......... 14 RESULTS ................................ ................................ ................................ ............................. 17 Age Structure of Male H arvest ................................ ................................ ....................... 17 Antlerless CPUE ................................ ................................ ................................ ............. 22 Hunter Numbers ................................ ................................ ................................ .............. 25 DISCUSSION ................................ ................................ ................................ ....................... 28 APPENDIX ................................ ................................ ................................ ................................ . 34 LITERATURE CITED ................................ ................................ ................................ ............... 36 CHAPTER 2: ANTLER SIZES IN RECORD DEER ARE INCREASING IN A REGION OF HIGH - QUALITY HABITAT ................................ ................................ ............................... 42 INRODUCTION ................................ ................................ ................................ ................... 42 STUDY AREA ................................ ................................ ................................ ..................... 44 METHODS ................................ ................................ ................................ ........................... 47 RESULTS ................................ ................................ ................................ ............................. 51 DISCUSSION ................................ ................................ ................................ ....................... 62 LITERATURE CITED ................................ ................................ ................................ ............... 68 CHAPTER 3: MANAGEMENT INFLUENCES ON THE NUMBER OF WHITE - TAILED DEER IN T HE BOONE AND CROCKETT RECORDS ................................ ........... 73 INRODUCTION ................................ ................................ ................................ ................... 73 STUDY AREA ................................ ................................ ................................ ..................... 75 METHODS ................................ ................................ ................................ ........................... 78 Boone and Crockett Record Deer ................................ ................................ ................... 78 Zero - inflated Poisson Model ................................ ................................ ........................... 81 RESULTS ................................ ................................ ................................ ............................. 83 DISCUSSION ................................ ................................ ................................ ....................... 89 APPENDI X ................................ ................................ ................................ ................................ . 94 LITERATURE CITED ................................ ................................ ................................ ............... 96 vii CHAPTER 4: ADDRESS ING UNCERTAINTY IN THE REPORTING PROCESS OF TH E BOONE AND CROCKETT RECORDS ................................ ................................ ......... 102 INRODUCTION ................................ ................................ ................................ ................. 102 STUDY AREA ................................ ................................ ................................ ................... 104 METHODS ................................ ................................ ................................ ......................... 105 Observed Data Independent Counts of Record Deer ................................ ................. 105 Model C ovariates ................................ ................................ ................................ .......... 107 Modeling Framework ................................ ................................ ................................ .... 108 RESULTS ................................ ................................ ................................ ........................... 110 DISCUSSION ................................ ................................ ................................ ..................... 119 APPENDICES ................................ ................................ ................................ .......................... 123 APPENDIX A: ZERO - INFL ATED POISSON N - MIXTURE MODEL ........................... 124 APPENDIX B: TRACE PLOTS FOR PARAMETERS OF N - MIXTURE MODEL ........ 125 APPENDIX C: MAP OF ECOREGION CLASSIFICATIONS FOR COUNTIES IN WISCONSIN ................................ ................................ ................................ ...................... 134 APPENDIX D: DIAGRAM OF PROCESS GENERATING RECORDS DATA ............. 135 LITERATURE CITED ................................ ................................ ................................ ............. 136 EPILOGUE ................................ ................................ ................................ ............................... 142 LITERATURE CITED ................................ ................................ ................................ ............. 145 viii LIST OF TABLES Table 2.1 Model comparisons evaluating random effects structure for identifying geographical areas where antler sizes of record deer have been similar in the Midwest United States from 1973 2014. Boone and Crockett records of white - tailed deer from Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, an d Wisconsin w ere included in the analysis. ................................ ................................ ................................ ..... 55 Table 2.2 Model comparisons relating climate/weather, landscape composition, landscape configuration, and soil characteristics to the gross score (cm) of record white - tailed deer. Boone and Crockett records of white - tailed deer harvested in Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin during years of study (1973 2014) were included in the analysis. ................................ ................................ ............... 57 Standard Error (SE) for the Habitat Model hypothesis, which states that the driving factors of antler size among record deer are habitat features and year. Boone and Crockett records of white - tailed deer harvested in Illinois, Indiana, Iowa, Kentucky, Mic higan, Minnesota, Missouri, Ohio, and Wisconsin during years of study (1973 2014) were included in the analysis. ................................ ................................ ...... 58 Standard Error (SE) for the Landscape Model hypothesis, which states that the driving factors of antler size among record deer are the configuration and composition of habitat and year, whereas the influences of climate and the longer - term impacts of nutrient cycling in the soil are negligible. Boone and Crockett records of white - tailed deer harvested in Ill inois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin during years of study (1973 2014) were included in the analysis. ................................ ................................ ................................ .............. 58 Standard Error (SE) for the Recent Model Hypothesis, which states tha t the driving factors of antler size among record deer are recent changes in habitat, weather, and year, whereas the longer - term impacts of nutrient cycling in the soil and climate are negligible. Boone and Crockett records of white - tailed deer harvested i n Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin during years of study (1973 2014) were included in the analysis. ................................ ............. 59 Standard Error (SE) for the Full Model, which states that all co variates of interest, considered in this analysis, are driving antler size in the Midwest United States (i.e., Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin) during years of study (1973 2014). ................................ ........ 59 Table 3.1 Parameter estimates and associated 95% credible intervals (LCI: Lower Credible Interval, UCI: Upper Credible Interval) from zero - inflated Poisson model for the number of record deer harvested in the Midwest United States (1973 2014). - values indicated that the mode l successfully converged. Data for analysis included record deer harvested in Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin. ....... 86 ix Table 3.2 The number of counties included and estimated intercepts for each ecoregion based on a random effect for ecoregion with a mean of zero. I assigned an ecoregion to each county by determining which ecoregion covered the greatest amount of area within the county. Boone and Crockett records of white - tailed deer harvested in Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin during years of study (1973 2014) were included in the analysis. ................................ ................................ ...... 87 x LIST OF FIGURES Figure 1.1 Study area for examining how harvest outcomes changed after the implementation of mandatory antler point restrictions. The 23 counties across the northern Lower Peninsula of Michigan were differentiated into categories based on their regulation history. The NW12 counties (light gray) had mandatory antler point restrictions, which prohib ited hunters from harvesting any antlered white - tailed deer with fewer than 3 points on one side. The non - APR counties (dark gray) were where the first buck tag remained unrestricted during all the years of the study (1987 2016). ................................ ....................... 11 Figure 1.2 Timeline of important events related to deer management regulations in Michigan. This timeline also includes the beginning and ending years of the study period for the analyses in this chapter. ................................ ................................ ................................ ... 12 Figure 1.3 Deviance plot for the proportion of 1.5 - year old deer in the male harvest (1987 2016) for 12 counties i n the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013 . Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. This plot shows that the model with the smallest deviance value was with a breakpoint in 2013 ( ). ................................ ................................ ................................ ................................ ......... 19 Figure 1.4 Piecewise trends (orange lines) in the proportion of 1.5 - year old deer in male harvest from 1987 2016, with a breakpoint in 2013. Open circles (purple) are the calculated proportions for each of the 12 counties i n the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013 . ................................ ................................ ................................ ................................ ............ 19 Figure 1.5 Deviance plot for the proportion of 2.5 - year old deer in the male harvest (1987 2016) for 12 counties in the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013. Smaller deviance values indicate better fit of trend li nes when the breakpoint occurs in that year. This plot shows that the model with the smallest deviance value was with a breakpoint in 2013 ( ) . ................................ ................................ ................................ ................................ ......... 20 Figure 1.6 Piecewise trends (orange lines) in the proportion of 2.5 - year old deer in male har vest from 1987 2016, with a breakpoint in 2013. Open circles (purple) are the calculated proportions for each of the 12 counties in the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013. ................................ ................................ ................................ ................................ ............ 20 xi Fig ure 1.7 - year old deer in the male harvest (1987 2016) for 12 counties in the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013. Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. This plot shows that the model with the smallest deviance value was with a breakpoint in 2013 ( ). ................................ ................................ ................................ ............... 21 Figure 1.8 Piecewise trends (orange lines) in the proportion - year old deer in male harvest from 1987 2016, with a breakpoint in 2013. Open circles (purple) are the calculated proportions for each of the 12 counties in the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemente d within the NW12 in 2013. ................................ ................................ ................................ ................................ ............ 21 Figure 1.9 Deviance plot for antlerless CPUE (2001 2016) for 12 counties in the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013. Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. This plot shows that the model with the smallest deviance value was with a breakpoint in 2007 ( ). ................................ .................. 23 Figure 1.10 Piecewise trends (orange lines) in antlerless CPUE from 2 001 - 2016, with a breakpoint in 2007. Open circles (purple) are the calculated CPUE for each of the 12 counties in the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013. ................................ ................................ ............ 23 Figure 1.11 Deviance plot for antlerless CPUE (2001 2016) for 11 counties in the northern Lower Peninsula (non - APR) of Michigan. Antler point restrictions were never implemented in these counties over the course of my study. Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. This plot shows that the model with the smallest deviance value was with a breakpoint in 2005 ( ). ................................ ................................ ................................ ............. 24 Figure 1.12 Piecewise trends (orange lines) in antlerless CPUE from 2001 - 2016, with a breakpoint in 2005. Open circles (purple) are the calculated CPUE for each of the 11 counties in the northern Lower Peninsula (non - APR) of Michigan. Antler point restrict ions were never implemented in these counties over the course of my study. ................................ ... 24 Figure 1.13 Deviance plot for number of hunters in the NW12 counties from 2001 - 2016. Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. The NW12 is the 1 2 - county area in the northwest Lower Peninsula of Michigan where antler point restrictions were implemented in 2013. This plot shows that the model with the smallest deviance value was with a breakpoint in 2005 ( ). .......... 2 6 Figure 1.14 Trend (orange line) of the number of hunters in the NW12 counties from 2001 - 2016. Open circles (purple) are the total number of hunters for each county in the NW12. The NW12 is the 1 2 - county area in the northwest Lower Peninsula of Michigan w here antler point restrictions were implemented in 2013. ................................ ................................ ... 26 xii Figure 1.15 Deviance plot for hunter numbers (2001 2016) for 11 counties in the northern Lower Peninsula (non - APR) of Michigan. Antler point restrictions were never implemented in these counties o ver the course of my study. Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. This plot shows that the model with the smallest deviance value was with a breakpoint in 2007 ( ). ................................ ................................ ................................ ................ 27 Figure 1.16 Trend (orange line) of the number of hunters in the non - antler point restriction counties from 2001 - 2016. Open circles (purple) are the total number of hunters for each of the 11 counties in the northern Lower Peninsula (non - APR) of Michigan. Antler point restrictions were never implemented in these counties over the course of my study. ................ 27 Figure 1.17 A Map of the areas the Michigan Department of Natural Resources (MDNR) used to summarize deer harvest data in the state for the annual hunting seasons. The f igure was copied from the Michigan Deer Harvest Survey Report for the Hunting Seasons in 2000 (MDNR Wildlife Report No. 3344, Frawley 2001). ................................ .......................... 35 Figure 2.1 Map of ecoregion classification for each county in study area (Omernik 1987, Bailey 1995). I assigned an ecoregion to each county by determining which ecoregion covered the majority area within the county. These data are from 9 Midwestern states in the United States (Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin). ................................ ................................ ................................ ........................... 4 6 Figure 2.2 Variograms and county - specific prediction maps for years 1984 and 2004 for the number of days where snow depth was 2.54 cm or more (SNOW). The darker shade indicates fewer days where snow depth was one or more inches. These data are from 9 Mid western states in the United States (Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin). (a) 1984 variogram is a Gaussian model with nugget = 26.54, sill = 187.48, and range = 6.87. (b) 2004 variogram is a Gaussian model with nugget = 42.43, sill = 93.07, and range = 3.15. (c) Plot shows predicted number of days in 1984. (d) Plot shows predicted number of days in 2004. ................................ ............... 53 Figure 2.3 Variograms and county - specific prediction maps showing total precipitation for years 1984 and 2004 during the spring and summer months (PRECIP). The darker shades indicate greater precipitation during the spring and summer months. These data are from 9 Midwestern states in the United States (Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesot a, Missouri, Ohio, and Wisconsin). (a) 1984 variogram is a Spherical model with nugget = 2,678.59, sill = 8,876.58, and range = 6.53. (b) 2004 variogram is a Spherical model with nugget = 3,618.99, sill = 8,035.93, and range = 4.97. (c) Plot shows predicte d amount of precipitation in 1984. (d) Plot shows predicted amount of precipitation in 2004. .... 54 Figure 2.4 Distribution of density (deer/km 2 ) in record deer among ecoregions in the Midwest United States from 1973 2014. These data are from 9 Midwestern states in the United States (Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin). ................................ ................................ ................................ ........................... 56 xiii Figure 2.5 Trends in antler size of record deer for each ecoregion in across 9 Midwestern states in the United States (Illinois, Indiana , Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin). Each line has a positive slope (0.16 cm/year) with a unique intercept that represents the trend in the size of antlers for individual ecoregion from 1973 2014. ................................ ................................ ................................ ......................... 60 Figure 2.6 Magnitude a nd direction of change in mean antler sizes for each county relative to the average antler size of entire study area (1973 2014). Counties are grouped by which ecoregion covered the greatest amount of area within the county . This map shows that deer harves ted in the western regions, southern Illinois, and eastern Ohio have larger antlers relative to other parts of the Midwest. ................................ ................................ ........................ 61 Figure 3.1 Ecoregion classification (Omernik 1987, Bailey 1995) for the counties of 9 states in the Midwestern United States incl uded in my study area: Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin. I assigned an ecoregion to each county by determining which ecoregion covered the greatest amount of area within the county. ................................ ................................ ................................ ................................ ... 77 Figure 3.2 The num ber of Boone and Crockett record deer harvested in each county in the Midwestern United States from 1973 2014. My study area covered 9 states: Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin. .................... 80 Figure 3.3 Map showing differences in the number of record deer among the ecoregions across the Midwest United States (i.e., Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin) during study (1973 2014). I assigned an ecoregion to each county by determining which ecoregion covered the greatest amount of area within the county. Purple areas are ecoregions with fewer record deer than expected from the zero - inflated Poisson model, whereas the orange areas correspond to ecoregions with more record de er than predicted given the model. Ecoregions that did not have a significant influence (i.e., credible interval included zero) on the number of record deer are white. ................................ ................................ ................................ ................................ ........... 88 Figure 3.4 Infographic to show the sequence of critical steps required for a male whi te - tailed deer that has record - sized antlers to become a data point in the Boone and Crockett outcome (Yes/No) of each step are given in the gray boxes. ................................ ...................... 95 Figure 4.1 The mean detection probability for each county in Wisconsin from 1981 2014 estimated from the N - mixture model. ................................ ................................ ....................... 112 Figure 4.2 Detection probability of each county in Wisconsin from 1981 2014. Each line represents the detection probability f or a single county through time. ................................ ..... 113 Figure 4.3 Mean number of record deer harvested from 1981 2014 in each county of Wisconsin estimated using the N - mixture model. ................................ ................................ .... 114 Figure 4.4 The standard deviation in number of record deer harvested estimates for each county in Wisconsi n from 1981 2014 estimated from the N - mixture model. ......................... 115 xiv Figure 4.5 Posterior distribution of percent agriculture through time (1981 2014) in Wisconsin. ................................ ................................ ................................ ................................ . 116 Figure 4.6 Posterior distribution of the Contrast Weighted Edge Density (CWED) metric to evaluate t he influence of landscape configuration on the number of record deer harvested through time (1981 2014) in Wisconsin. ................................ ................................ ................. 117 Figure 4.7 Posterior distribution of antlered harvest through time (1981 2014) in Wisconsin. ................................ ................................ ................................ ................................ . 118 Figure 4.8 Trace plots of the values generated from the N - mixture model in Chapter 4 with the value of sample from MCMC process (y - axis) versus the iteration number (x - axis). Each plot represents the sampling histories of a single model parameter. These plots show that the chains for each par ameter are mixing well over the parameter space. ........................ 125 Figure 4.9 Map of ecoregion classification for each county in study area (Omernik 1987, Bailey 1995). I assigned an ecoregion to each county by determining which ecoregion covered the majority area within the county. ................................ ................................ ............ 134 Figure 4.10 Infographic to show the sequence of critical steps required for a male white - tailed deer that has record - sized antlers to become a data point in the Boone and Crockett outcome (Yes/No) of each step are given in the gray boxes. ................................ .................... 135 1 PROLOGUE This dissertation brought toge ther two of my passions, white - tailed deer and statistics. When I first began working on this research, the focus of the project was to investigate the spatial and temporal distribution in the harvest of white - tailed deer in the Boone and Crockett ecords of North American Big Game. Over the years, my dissertation evolved to include 4 distinct chapters, with antlers becoming the common theme. Therefore, I have dedicated most of this prologue to an overview of antlers. Antlers are secondary sexual cha racters exclusively found in males of the Cervidae family, except in reindeer where they are seen in both sexes (Landete - Castillejos et al. 2012) . Deer grow and shed their antlers every year, requiring large amounts of nutrients and energy (Banks 1974, Ditchkoff et al. 2001) , making antlers costly to produce. Antlers grow from the pedicle, which is located on the frontal bone of the skull. Moreover, these secondary sex (Landete - Castillejos et al. 2012) (Grasman and Hellgren 1993) . In general, deer antlers occur in an extensive diversity of sizes and forms, which depend (Strickland and Demarais 2000, Demarais and Strickland 2011) . Moreover, other factors such as condition of the mother, date of birth, health of the individual, and weather conditions can a f fect antler development (Schultz and Johnson 1995, Monteith et al. 2009) . Strickland and Demarais (2008) found that antler sizes of white - tailed deer are influenced by landscape composition in Mississippi. Their model suggests a positive influence on antler size in land - use types that promote growth of early 2 successional herbaceous plant communities. R esearchers and state agencies use antler measurements to evaluate deer populations because a close relationship between antler size and the nutritional state of the white - tailed deer population exists (McCullough 1982) . Wildlife management agencies typically use a variety of antler characteristics to define the minimum harvest criteria for hunted populations (Strickland et al. 2001). These selective harvest criteria attem pt to protect the younger age class to recruit more males into older age classes. However, criteria designed to protect these younger males may have an impact on harvest of the older males (Strickland et al. 2001) . Given variability in habitat in which white - tailed deer occur, it is important that harvest criteria be designed based on antler characteristics specific to the population. In the literature, effects of selective harvesting practices on antler size s of p opulations vary . Some studies suggest selective harvesting of yearling males is not likely to influence the genetic potential of antler growth ( i.e., low hereditability , Lukefahr and Jac obson 1998, Webb et al. 2012, Hewitt et al. 2014, Webb et al. 2014) , usually by citing the complex interactions of environmental factors and various injuries potentially affecting antler development. Other studies suggest that selective harvest at young ages can affect antler size of deer remaining in the cohort at later ages (Strickland et al. 2001, Lockwood et al. 2007, Hewitt et al. 2014, Ramanzin and Sturaro 2014) , because of the positive relationship between yearling antler size and antler size at later ages. Allendorf et al. (2008) recommend that managers assume that some genetic change will occur due to selective harvesting and appl ication of basic genetic principles to management strategies for harvested species. Th e goal of this dissertation should be to engage interests of wildlife biologists and researchers. The organization of the dissertation is as follows. Chapter 1 uses harvest data of 3 white - tailed deer to investigate harvest outcomes of antler point restrictions in the state of Michigan. This chapter is designed to understand if changes in harvest data are driven by implementation of antler point restrictions. Chapter 2 describes the results of a spatially explicit analysis to investigate trends in antler sizes of white - tailed deer with large antlers. This chapter was inspired by the work of Monteith et al. (2013) that investigated the tempo ral trends in horn and antler sizes of animals in the book, Records of North American Big Game. Chapter 3 is an investigation into the spatial and temporal distribution in harvests of white - tailed deer with large antlers. The final chapter of this disserta tion speaks to potential issues related to reporting bias and proposes an adaptation of N - mixture models (Royle 2004) to account for imperfect detection. 4 LITERATURE CITED 5 LITERATURE CITED Allendorf, F. W., P. R. England, G. Luikart, P. A. Ritchie, and N. Ryman. 2008. Genetic effects of harvest on wild animal populations. Trends in Ecology & Evolution 23:327 337. Banks, W. 1974. The ossification process of the developing antler in the white - tailed deer ( Odocoileus virginianus ). Calcified Tissue R esearch 14:257 274. Demarais, S., and B. K. Strickland. 2011. Antlers. Pages 107 145 in D. G. Hewitt, editor. Biology and management of white - tailed deer. CRC Press, Boca Raton, Florida, USA. Di deer ( Odocoileus virginianus ): e Evolution 55:616 625. Grasman, B. T., and E. C. Hellgren. 1993. Phosophorus nutriti on in white - tailed deer: n utrient balance, physiological responses, and antler growth. Ecology:2279 2296. Hewitt, D. G., M. W. Hellickson, J. S. Lewis, D. B. Wester, and F. Management 78:976 984. Landete - Castillejos, T., J. A. Estevez, F. Ceacero, A. J. Garcia, and L. Gallego. 2012. A review of factors affect ing antler composition and mechanics. Frontiers in Bioscience E 4:2328 2339. Lockwood, M. A., D. B. Frels, W. E. Armstrong, E. Fuchs, and D . E. Harmel. 2007. Genetic and environmental i nter eer. The Journal of Wildlife Management 71: 2732 2735. Lukefahr, S. D., and H. A. Jacobson. 1998. Variance component analysis and heritability of antler traits in white - tailed deer. The Journal of Wildlife Management 62:262 268. McCullough, D. R. 1982. Antler characteristics of George Reserve white - tailed deer. The Journal of Wildlife Management 46:821 826. Monteith, K. L., R. A. Long, V. C. Bleich, J. R. Heffelfinger, P. R. Krausman, and R. T. Bowyer. trophy ungulates. Wildlife Monographs 183:1 28. Monteith, K. L., L. E. Schmitz, J. A. Jenks, J. A. Delger, and R. T. Bowyer. 2009. Growth of male white - tailed deer: c onsequences of maternal effects. Journal of Mammalogy 90:651 660. 6 Ramazin, M., and E. Sturaro. 20 selectivity in roe deer. European Journal of Wildlife Research 60:1 10. counts. Biometrics 60:108 115. Schultz, S. R., and M. K. Johnson. 1995. Effects of birth date and body mass at birth on adult body mass of male white - tailed deer. Journal of Mammalogy 76:575 579. Strickland, B. K., and S. Demarais. 2000. Age and regional differences in antlers and mass of white - tailed deer. Journal of Wildlife Management 64:903 911. Strickland, B. K., and S. Demarais. 2008. Influence of landscape composition and structure on antler size of white - tailed deer. Journal of Wildlife Management 72:1101 1108. Strickland, B . K., S. Demarais, L. E. Castle, J. W. Lipe, W. H. Lunceford, H. A. Jacobson, D. Frels, and K. V. Miller. 2001. Effects of selective - harvest strategies on white - tailed deer antler size. Wildlife Society Bulletin 29:509 520. Webb, S. L., S. Demarais, B. K. Strickland, R. W. DeYoung, B. P. Kinghorn, and K. L. Gee. a modeling approach. The Journal of Wildlife Management 76:48 56. Webb, S. L., K. L. Gee, R. W. DeYoung, and S. M. Harju. 2014 . Variance component analysis of body mass in a wild population of deer ( Odocoileus virginianus ): r esults from two decades of research. Wildlife Research 40:588 598. 7 CHAPTER 1: EVALUATING THE EFFECTIVENESS OF ANTLER POINT RESTRICTIONS TO ACHIEVE WHITE - TAILED DEER MANAGEMENT GOALS INTRODUCTION Managers of white - tailed deer ( Odocoileus virginianus ) set hunting regulations based on science and stakeholder input with the intent that harvest outcomes are appropriate for the management goals of an area (Smith and Coggin 1984, Geist et al. 2001, Riley et al. 2002, Hansen 2011, Organ et al. 2012). Fiftee n state agencies indicated using some form of antler restriction to help achieve management goals (Quality Deer Management Association [QDMA] Staff 2018). Antler point restrictions are intended to protect younger males from harvest by restricting take to a ntlered deer with a minimum number of antler points (Carpenter and Gill 1987, Hamilton et al. 1995 a , Hansen et al. 2017, Wallingford et al. 2017). Support for antler point restrictions is mixed among hunters ( Decker et al. 1980, Schroeder et al. 2014) and there have been few tests of the effectiveness of these restrictions in achieving management goals. (Decker et al. 2013, Mason and Rudolph 2015). Antler point restrictions were first mentioned in the literature by Carpenter and Gill (1987) in their discu ssion about trade - offs and knowledge gaps associated with these regulations for mule deer ( Odocoileus hemionus ) and elk ( Cervus elaphus ) harvest systems. More recently, alleged costs and benefits of mandatory antler point restrictions have been debated in the popular literature with vocal stakeholders on both sides of the issue (Pinizzotto 2017, YoungeDyke et al. 2017). In theory, the reduced harvest pressure on yearling males under antler point restrictions will result in higher recruitment of male deer i nto older age classes (Carpenter and Gill 1987, 8 Hansen et al. 2017, Wallingford et al. 2017). Schroeder et al. (2014) reported that hunters targeting large antlered deer were supportive of antler point restrictions at first, but their support of the regula tion decreased over time. The decline in support for the antler point restrictions may reflect unmet expectations that these hunters had for antler point restrictions as a tool for producing large antlered deer (Decker et al. 1980). Although antler point r estrictions are designed to protect the majority of yearling males from being harvested by hunters (Hamilton et al. 1995 a ) and advance the age structure of male deer (Frawley 2012), 2 indirect outcomes of antler point restrictions are also hypothesized. Th e first is that antler point restrictions will increase the harvest of antlerless deer where implemented. Intuitively, when there is reduced availability of yearling males for harvest, hunters will be more likely to harvest female deer under antler point restrictions (Hamilton et al. 1995 b , Cornicelli et al. 2011, Hansen et al. 2017, Wallingford et al. 2017, Hansen et al. 2018). The ability to control and stabilize deer populations by increasing antlerless permits or quotas is limited (Curtis et al. 2000, Schroeder et al. 2014), and additional or alternative regulations may add incentives to shift harvest pressure to female deer (Decker and Connelly 1989, Cornicelli et al. 2011). The second indirect outcome of antler point restrictions is improving hunter recruitment and retention due to increases in perceived opportunities to harvest mature bucks with large antlers (Hansen et al. 2018). White - tailed deer managers rely heavily on active hunter participation to achieve management goals, but declining hunter numbers lead to questions about the effectiveness of hunters in controlling white - tailed deer populations in the future (Brown et al. 2000, Winkler and Warnke 2013). Moreover, an increasing number of hunters are interested in management regulations that co uld improve their opportunity to harvest mature bucks 9 (Connelly et al. 2012). However, little is known about the effects of antler point restrictions on hunter recruitment and retention. Therefore, an investigation is warranted because antler point restri ctions may influence hunters to move into an area if they perceive that there are more abundant opportunities available. To date, only a few studies have assessed the harvest outcomes of antler point restrictions in white - tailed deer (Hansen et al. 2017, Wallingford et al. 2017, Hansen et al. 2018), but none have evaluated temporal trends in harvest outcomes leading up to and after antler point restrictions were implemented. Thus, the question remains, do antler point restrictions cause the trajectory of h arvest outcomes to change. My objective was to test 3 hypotheses that antler point restrictions caused a change in harvest levels. H 1 Male age structure Antler point restrictions shift harvest pressure to older aged males H 2 Antlerless harvest Antler point restrictions increase harvest of antlerless deer H 3 Hunter numbers The decline in hunter numbers is less severe under antler point restrictions STUDY AREA Harvest data from 23 counties in the Northern Lower Peninsula of Michigan off ered an ideal case study for investigating how harvest outcomes (male age structure, antlerless harvest, and hunter numbers) changed after implementation of a mandatory antler point restriction because 12 of the counties recently implemented mandatory antl er point restrictions (Figure 1.1, Figure 1.2, Frawley 2017). I classified the 23 counties into categories based on regulation history 10 (Figure 1.1). Since 1991, Michigan hunting regulations have limited hunters to a maximum harvest of 2 antlered deer (i.e. Michigan enacted a statewide regulation placing an antler point restriction on the second tag of harvested antlered white - tailed deer. For hunters that harvested 2 bucks per year, the antler poi nt hunters who purchased a single buck tag were not required to follow the antler point restriction, thus any legal antlered deer could be taken. In 2013, mandatory antler point restrictions, which prohibited hunters from harvesting any antlered white - tailed deer with fewer than 3 points on one side, were enacted in 12 counties in the Northwest Lower Peninsula (hereafter referred to as NW12, Figure 1.1) base d on a proposal by t he Northwest Michigan Chapter of the Quality Deer Management Association (Frawley 2012). One goal of the NW12 antler point restriction was to advance the buck age structure (Frawley 2012). Eleven adjacent counties (referred to as non - an tler point restriction [non - APR]) served as a control treatment for comparison (Figure 1.1). Counties of NW12 and non - APR were characterized by similar landscapes (e.g., mostly forested with some agriculture, little residential or commercial development). More importantly, the NW12 and non - APR counties were under similar regulations (e.g., season lengths, weapons permitted, disease controls) with exception that the first buck tag remained unrestricted during the years of the study in the non - APR counties. T he similarities between NW12 and non - APR counties allowed for reasonable comparisons of harvest outcomes between the two groups. 11 Figure 1.1 Study area for examining how harvest outcomes changed after the implementation of mandatory antler point restrict ions. The 23 counties across the northern Lower Peninsula of Michigan were differentiated into categories based on their regulation history. The NW12 counties (light gray) had mandatory antler point restrictions, which prohibited hunters from harvesting an y antlered white - tailed deer with fewer than 3 points on one side. The non - APR counties (dark gray) were where the first buck tag remained unrestricted during all the years of the study (1987 2016). 12 Figure 1.2 Timeline of important events related to deer management regulations in Michigan. This timeline also includes the beginning and ending years of the study period for the analyses in this chapter. 13 METHODS Data Collection I acquired harvest data from 2 sources: deer check - station records (1987 2016) and annual harvest surveys (2001 2016). I used check - station data to obtain information about the age structure of harvested deer and harvest survey data to determine changes in a population index and hunter numbers. For over 50 years, Michigan Department of Natural Resources (MDNR) personnel collected information from hunter - harvested deer at voluntary check stations, including the county in which the deer was harvested as well as the sex and age of the deer. Data from voluntary check s tations are used by the state to monitor composition and health of the deer herd (MDNR 2016). M DNR personnel estimated the age of each checked deer using the tooth - wear and replacement technique (Severinghaus 1949), and in cases where the age could not be my interest in understanding trends in buck age structure, I excluded deer that could not be properly aged (e.g., unable to extract jawbone) from my analyses. Fo r each year and county in the study area, I calculated the proportion of 3 different age classes (e.g., 1.5 - year, 2.5 - year, and - year classes) in the total male harvest that was reported at voluntary check stations from 1987 to 2016. I extracted count y - level data about estimated number of hunters, hunter effort (i.e., days afield), and harvest of antlerless deer from 2001 to 2016 from the Michigan Department of Natural Resources annual harvest survey reports. These reports are generated from hunter res ponses to a mailed survey about their deer hunting experience. Given that antler point restrictions may alter hunter effort (Miller and Vaske 2003, Seng et al. 2017), I calculated an 14 antlerless catch per unit effort (CPUE) index by dividing total antlerles s harvest by hunter effort for each county in each year. Antlerless catch in this case is synonymous with antlerless harvest. Prior to the 2001 hunting season, the state was di vided into 8 regions (Appendix ) related to administration units for wildlife man agement of the Department of Natural Resources (Frawley 2001). Each region spanned multiple counties and harvest data were collected and summarized according to the region where the hunt occurred (Frawley 2001). Consequently, harvest data from the surveys were not available at smaller spatial scales, such as individual counties. To ensure this change in spatial scales did not influence my analysis, I only used harvest data from the annual survey reports for years when county - specific information was availab le (i.e., 2001 2016). Statistical Analysis I used the proportion of 1.5 - year, 2.5 - - year old male deer in the harvest as response variables to test the hypothesis about influence of antler point restrictions on age structure of the male harve st (H 1 ). I used antlerless CPUE index to test the hypothesis that antler point restrictions influence antlerless harvest (H 2 ). Lastly, I used number of hunters from 2001 - 2016 to test the hypothesis that antler point restrictions influence hunter recruitmen t and retention where they are implemented (H 3 ). For each hypothesized harvest outcome, I analyzed county - level data of the response variables, so I had multiple data points for each year. I used piecewise regressions to investigate the influence of antle r point restrictions on different harvest outcomes (Flora 2008, Crawley 2013). For each hypothesis, I analyzed the trend in each response variable (proportion of male age classes, antlerless CPUE, number of hunters) across years. I identified changes throu gh time in longitudinal data by fitting a trend line (or line 15 of best fit) to understand the relationship between my response variable and time (Stasinopoulos and Rigby 1992, Flora 2008). However, the trend may nonlinear through time (e.g., declining trend initially but increasing later), so I tested different breakpoints to identify the year for which the trends before and after were most different (Stasinopoulos and Rigby 1992, Crawley 2013). I identified the breakpoint for each hypothesized trend (Stasin opoulos and Rigby 1992) and calculated the slope and intercept of the trends for the timeframes before and after the breakpoint (Flora 2008, Crawley 2013). Therefore, instead of 1 linear trend through time, which assumes a consistent trend in the response, I created a nonlinear regression model with 2 trends, split at the year where the 2 lines would be most different. Rather than visually selecting a breakpoint based on a scatter plot of the data, I let the data inform where the breakpoint should occur (St asinopoulos and Rigby 1992, Crawley 2013). I used a piecewise regression model to identify the year to serve as the optimal break ( ) for the response variables of each hypothesis . For each response variable, I considered a range of possible years as the optimal breakpoint and compared the deviance values of models with different breakpoints to select the break that fit the data best. I then used the year from the model with the smallest deviance value as the optimal breakpoint . If antler point restrictions influenced the harvest outcome , I would expect = 2013. Although the piecewise regression models for each hypothesis had different response variables, they followed the same general format of a 2 - segment piecewise regression (Crawley 2013). Data were specified for each county for , 16 where the relationship between the response and year is different before and after the break ( . When , th e relationship is linear with slope ( ) and y - intercept ( ), whereas when , the relationship, while still linear, has a different slope and y - intercept . Once the breakpoint was identified, I performed an analysis of variance (ANOVA) comparing the 2 - trend model with the breakpoint to a null model, which was a simple linear regression with no breakpoint, to determine if the breakpoint model was a significant improvement over the null model (Crawley 2013). If the model with 2 - trends did not provide significantly more information than the null model, then only the trend of the null model was interpreted. If the piecewise regression model was a significant improvement over the null model, I determined if the trend was increasing, decreasin g, or constant over the years before and after the break by looking at slopes of lines. I performed the analyses of data from the NW12 counties separately from analyses with data from the non - APR counties. This distinction allowed comparison of the results among the counties had that recently implemented antler point restrictions and counties that had not. I compared slopes of trends between the NW12 and non - APR county groups. I used this comparison to determine if the trends of the NW12 differed from trend s in surrounding counties without antler point restriction. If the confidence intervals of the slopes overlapped, I concluded that t he slopes between the NW12 and n on - APR groups were not different. 17 RESULTS Age Structure of Male Harvest Male deer registered at voluntary check stations among the NW12 counties averaged 214.89 deer/county ( SD = 146.39, range: 24 887 deer; n = 77,362 male deer) across all years of this study. The age structure of male harvest varied among counties and across years. From 1987 201 6, average proportion of male deer registered at check stations in the 1.5 - year age class was 0.60 deer/county ( SD = 0.17, range: 0.06 0.84; n = 49,050 yearling males), 2.5 - year age class was 0.19 deer/county ( SD = 0.08, range: 0.05 0.51; n =13,844 male d eer in 2.5 - year age - year age class was 0.13 deer/county ( SD = 0.11, range: 0.00 0.58; n = 7,650 male deer 3.5 - years and older). Piecewise regression models on the effects of antler point restrictions on age structure of checked male deer f it the data best (i.e., had the smallest deviance value) with a break when (Figure 1.3, Figure 1.5, Figure 1.7). Moreover, results from the ANOVA comparisons showed that each of the piecewise regression models with the optimal break was an improvem ent over simple linear regression over the entire timeframe (1.5 - year olds: ; 2.5 - year olds: - year olds: ). There were noticeable changes in the proportion of males harvested in each age class between the 2012 and 2013 hunting seasons. The proportion of 1.5 - year old males estimated for the 2013 hunting season was 59.7% less than the estimated harvest of the 2 012 season (Figure 1.4). Conversely, there was an 81.3% increase in the proportion of 2.5 - year old males (Figure 1.6) and a 120.2% - year old males harvested in 2013 than in 2012 (Figure 1.8). During the years after the bre ak, the trend in the harvest of 1.5 - year old males was stable (i.e., 18 not increasing or decreasing; ) . The trend in the proportion of 2.5 - year olds in male harvest was also stable ( ) over the years following the break, - year age class increased ( ) during the same years. 19 Figure 1.3 Deviance plot for the proportion of 1.5 - year old deer in the male harvest (1987 2016) for 12 counties i n the northwest Lower Penins ula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013 . Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. This plot shows that the model with the smallest deviance value was with a breakpoint in 2013 ( ). Figure 1.4 Piecewise trends (orange lines) in the proportion of 1.5 - year old deer in male harvest from 1987 2016, with a breakpoint in 2013. Open circles (purple) are the calculated proportions for each of the 12 counties i n the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013 . 20 Figure 1.5 Deviance plot for the proportion of 2.5 - year old deer in the male harvest (1987 2016) for 12 coun ties in the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013. Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. This plot shows that the model with the smallest deviance value was with a breakpoint in 2013 ( ) . Figure 1.6 Piecewise trends (orange lines) in the proportion of 2.5 - year old deer in male harvest from 1987 2016, with a breakpoint in 2013. Open circles (purple) are the calculated proportions for each of the 12 counties in the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013. 21 Figure 1.7 - year old deer in the male harvest (1987 2016) for 12 counties in the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013. Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. This plot shows that the model with the smallest deviance value was with a breakpoint in 2013 ( ). Figure 1.8 - year old deer in male harvest from 1987 2016, with a breakp oint in 2013. Open circles (purple) are the calculated proportions for each of the 12 counties in the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013. 22 Antlerless CPUE From 2001 2016, antlerless harvest in the NW12 averaged 1,712 deer ( SD = 1,230; range: 114 6,903; n= 328,767 antlerless deer), whereas antlerless harvest in the surrounding counties where there were no antler - point restrictions averaged 2,279 deer ( SD = 1,390; range: 63 5 ,180; n = 401,277 antlerless deer). Over the same timeframe, hunters in the NW12 spent an average of 98,558 days afield ( SD = 37,589; range: 39,301 217,131; n = 18,923,219 days). The average number of days spent afield (119,341 days) was greater in the sur rounding counties where there were no antler - point restrictions ( SD = 43,106; range: 48,185 207,733; n = 21,004,065 days). For the NW12, the best fitting model for antlerless catch per unit effort (CPUE) included a breakpoint at (Figure 1.9). Model s that were least supported by the data were those with a break in any year after antler point restrictions were implemented in the NW12 (Figure 1.9). The piecewise regression model was a significant improvement over the null model ( ). The estimated antlerless CPUE was a 56.4% greater in 2007 than it was in 2006 (Figure 1.10). The data for the interval after the breakpoint suggested a positive relationship between antlerless catch per unit effort and year ( ). For the non - AP R counties, the best fitting model for antlerless CPUE was with a breakpoint in the year 2005 (Figure 1.11). The antlerless CPUE model showed a break in 2005 and was significantly different from the null model for data from the group of non - APR counties ( ). The trend in antlerless CPUE was stable over the years following the break ( , Figure 1.12). 23 Figure 1.9 Deviance plot for antlerless CPUE (2001 2016) for 12 counties in the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013. Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. This plot shows that the model with the smallest deviance value was with a breakpoint i n 2007 ( ). Figure 1.10 Piecewise trends (orange lines) in antlerless CPUE from 2001 - 2016, with a breakpoint in 2007. Open circles (purple) are the calculated CPUE for each of the 12 counties in the northwest Lower Peninsula (NW12) of Michigan. Antler point restrictions were implemented within the NW12 in 2013. 24 Figure 1.11 Deviance plot for antlerless CPUE (2001 2016) for 11 counties in the northern Lower Peninsula (non - APR) of Michigan. Antler point restrictions were never impleme nted in these counties over the course of my study. Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. This plot shows that the model with the smallest deviance value was with a breakpoint in 2005 ( ). Figure 1.12 Piecewise trends (orange lines) in antlerless CPUE from 2001 - 2016, with a breakpoint in 2005. Open circles (purple) are the calculated CPUE for each of the 11 counties in the northern Lower Peninsula (non - APR) of Michigan. Ant ler point restrictions were never implemented in these counties over the course of my study. 25 Hunter Numbers From 2001 2016, the number of hunters averaged 9,700 hunters ( SD = 4,085; range: 4,372 24,288) in the NW12, whereas hunter numbers in the surrounding counties that have not implemented antler point restrictions averaged 10,885 hunters ( SD = 3,882; range: 2,167 19,410). The piecewise regression model for number of hunt ers that fit the NW12 data the best included a breakpoint in the year 2005 (Figure 1.13), whereas a break in 2007 was the optimal breakpoint for data from the non - antler point restriction counties (Figure 1.15). However, the piecewise regression models wit h 2 - trends did not provide significantly more information than the null model (i.e., simple linear regression) for the same data (NW12: ; non - APR: ). Therefore, the null model was used to interpret trends in the number of hu nters. The number of hunters in NW12 and non - antler point restriction counties decreased since 2001 (Figure 1.14, Figure 1.16). Furthermore, overlapping confidence intervals of the slopes suggest the rate of decline in hunter numbers was not statistically different between the NW12 and non - antler point restriction counties . 26 Figure 1.13 Deviance plot for number of hunters in the NW12 counties from 2001 - 2016. Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. The NW12 is the 1 2 - county area in the northwest Lower Peninsula of Michigan where antler point restrictions were implemented in 2013. This plot shows that the model with the smalle st deviance value was with a breakpoint in 2005 ( ). Figure 1.14 Trend (orange line) of the number of hunters in the NW12 counties from 2001 - 2016. Open circles (purple) are the total number of hunters for each county in the NW12. The N W12 is the 1 2 - county area in the northwest Lower Peninsula of Michigan where antler point restrictions were implemented in 2013. 27 Figure 1.15 Deviance plot for hunter numbers (2001 2016) for 11 counties in the northern Lower Peninsula (non - APR) of Michiga n. Antler point restrictions were never implemented in these counties over the course of my study. Smaller deviance values indicate better fit of trend lines when the breakpoint occurs in that year. This plot shows that the model with the smallest deviance value was with a breakpoint in 2007 ( ). Figure 1.16 Trend (orange line) of the number of hunters in the non - antler point restriction counties from 2001 - 2016. Open circles (purple) are the total number of hunters for each of the 11 counties in the northern Lower Peninsula (non - APR) of Michigan. Antler point restrictions were never implemented in these counties over the course of my study. 28 DISCUSSION The increased proportion of older male deer harvested starting in 2013 suggests tha t mandatory antler point restrictions implemented in the NW12 increased the age structure of bucks harvested in that area. I found no evidence that antler point restrictions changed antlerless deer harvest or the number of hunters. My analytical approach d iffers from most analyses looking at outcomes after regulation changes. In general, evaluations compare the means between 2 timeframes (i.e., mean before event, mean after event). Although these comparisons can be useful, this approach assumes that any dif ferences between the means were caused by the regulation change. Conversely, the piecewise regressions used in this paper do not assume that the new regulation caused changes in harvest. Rather, my piecewise regression approach tests this assumption by eva luating trends in the data for the entire timeframe. This ensures that I did not falsely attribute a change in my harvest variables to the new mandatory antler point restrictions. My finding that antler point restrictions successfully advanced the age stru cture of harvested male deer differs from harvest outcomes under antler point restrictions on elk ( Cervus canadensis ) and mule deer ( Odocoileus hemionus ) populations in the western USA (Hansen 2011). The regulation in western USA was successful in protecti ng male deer and elk for 1 year, but these animals generally did not survive the following hunting season (Weigand and Mackie 1987, Biederbeck et al. 2001). If harvest pressure was a major barrier for advancing that age structure of deer and elk in western states, then how can harvest pressure not be an issue in recruitment of older - aged bucks for NW12 populations? One reason is that western hunters, for the most part, are unable to harvest female deer and elk (Hansen 2011). Therefore, hunters in the NW12, in having the opportunity to harvest antlerless deer, are not limited to harvesting only 29 legal males. I hypothesize that the additional harvest opportunities available to NW12 hunters potentially reduce the intensity of hunting pressure on legal males ther eby allowing recruitment of older - aged bucks. Selective - harvest criteria, like antler point restrictions that are designed to protect young males have consequences for harvest of older males (Strickland et al. 2001). A concern of employing antler point res trictions is that male deer will only live one additional year and not survive to older ages because hunting pressure will be high on the 2.5 - year age class (Carpenter and Gill 1987). My findings showed an increasing trend in harvest of 3.5 - year and older males (Figure 1.8) despite an expected increase in harvest pressure on older males when antler point regulations are in place. Investigations from other states have reported similar increases in the harvest of older male deer (Hansen et al. 2017, Wallingfo rd et al. 2017, Gulsby et al. 2019). Although antler point restrictions are intended only to influence survival of yearling males, there is evidence that harvest vulnerability may decrease with age (Ditchkoff et al. 2001). Therefore, in protecting yearling males, antler point restrictions may indirectly enhance the survival of male deer in older age classes. I did not find evidence to support the hypothesis that antler point restrictions caused a change in the antlerless harvest. If antler point restriction s influenced my index of antlerless harvest, we would expect the optimal break in 2013. However, my results suggest a positive trajectory in antlerless harvest since the abrupt change in 2007 (Figure 1.9). Therefore, the trend in antlerless harvest after a ntler point restrictions is actually a consequence of something that occurred around 2007, rather than an outcome of implementing antler point restrictions. The increase in antlerless harvest in 2007 may reflect increases in hunter cooperatives practicing quality deer management (Hamilton et al. 1995 b ), increased hunter willingness to harvest 30 antlerless deer (Adams and Hamilton 2011, deCalesta 2012), or some combined influence of these actions. The theory underlying the antlerless harvest hypothesis is that if hunters are unable to harvest yearling males due to implementation of an antler point restriction, hunters will focus their harvest on the female segment of the deer population (Hansen et al. 2018). Previous studies suggested that antler point restrict ions do not reduce opportunities for hunters to hunt antlerless deer (Wallingford 2012), and there is evidence that hunters may shift their focus, at least in part, to harvesting antlerless deer under these regulations (Hansen et al. 2017). In the NW12, an increase in antlerless harvest occurred during the first year of antler point restrictions but this change was not held through time. Antlerless CPUE in NW12 showed a 15.69% increase from 2012 to 2013, which supports the hypothesis that hunters will harve st more deer that are antlerless under an antler point restriction. However, this increase in antlerless harvest was followed by a 15.17% decrease from 2013 to 2014, so fewer antlerless deer were harvested during second year under antler point restrictions than were harvested before the regulation was implemented. The temporary increase in antlerless harvest during the first year of antler point restrictions aligns with previous suggestions these regulations are a short - term solution for skewed deer populat ions (Gulsby et al. 2019). In Missouri, hunters perceived that adult males were available in greater numbers under antler point restrictions (Hansen et al. 2018). Therefore, I hypothesize that the decrease in antlerless harvest during the second year is d ue to changes in hunter perceptions of the availability of adult males that re - shifts hunter focus back to bucks. Antler point restrictions may show different harvest outcomes for deer populations from differing habitats, even within the same state (Hansen et al. 2017). Hansen et al. (2017) reported that antler point restrictions in Missouri did not increase harvest of female deer in an area of 31 poorer quality habitat, but there was an increase in antlerless harvest in an area of better quality habitat. Habi tat quality influenced deer population density and the percentage of 1.5 - and 2.5 - year old males that could be legally harvested, with inferior habitat having a lower population density and lower percentage of younger males attaining legal status. Deer in the NW12 are impacted by severe winters but have been relatively stable in recent years (MDNR 2019). My findings that antler point restrictions had no influence on antlerless harvest and the contrasting results from 2 different habitats in Missouri speak t o the importance of developing restrictions specific to the characteristics of the deer herd that will be affected (Hamilton et al. 1995 a , Hamilton et al. 1995 b , Strickland et al. 2001, QDMA Staff 2018). My results indicated that the number of hunters in the NW12 declined at rates similar to the 11 non - APR counties ( Figure 1.14, Figure 1.16) . Thus, there was no evidence for a significant change in the number of hunters during the years of this analysis, regardless of whether antler point restrictions were implemented. It is interesting to note that this decline occurred despite survey findings that hunters would hunt the same amount or more often under antler point restrictions, if Michigan DNR implemented them where they hunt (Seng et al. 2017). This discr epancy is consistent with Stedman et al. (2004) that found a ctual hunter behavior in the field might differ from what has been reported in survey responses. Although regulations can constrain hunter participation (Miller and Vaske 2003), the majority of hu nters (about 77%) in the NW12 supported the mandatory antler point restriction (Frawley 2017). Moreover, when asked what could be done to get participants to hunt more frequently, several respondents suggested that the Michigan Department of Natural Resour ces increase the number of mature bucks and improve herd health; expanding antler point restrictions was mentioned specifically (Seng et al. 2017). Hunters also exhibit high - site fidelity (Cornicelli et al. 2011), which could 32 help explain why hunter number s did not change in the NW12 relative to non - APR counties. Thus, despite support of antler point restrictions indicated by hunters via survey results, attitudes of hunters toward these regulations were not sufficiently positive to influence hunter movement s into the area. My results suggest that additional harvest regulations, beyond antler point restrictions, are necessary to decrease the size of a deer population. Alternative harvest opportunities must be available (e.g., antlerless tags) for the possibility that hunters shift harvest pressure to antlerless deer when antler point restrictions reduce the availability of harvestable yearling males (Hansen et al. 2017). Hunting participation may inadvertently decrease if implementing these regulation is perceived by constitue nts to decrease available harvest opportunities (Fulton and Manfredo 2004), especially for those hunting regulations with competing interests (Decker et al. 2015, Hansen et al. 2017). Enhancements to the quality of the deer herd increases hunter satisfacti on, specifically satisfactions related to achievement (Elbeling - Schuld and Darimont 2017). In this chapter, I tested hypotheses related to how deer harvest outcomes change after implementation of antler point restrictions. My findings are of particular imp ortance to deer managers because they often seek to achieve multiple objectives with harvest regulations (Robinson et al. 2019, Fuller et al. 2020). Moreover, there is a growing interest among hunters for regulations designed to produce more mature bucks ( Ozoga et al. 1995, Cornicelli and Grund 2011, Connelly et al. 2012, Harper et al. 2012). Thus, if antler point restrictions help achieve multiple management objectives, they would be an excellent tool for managers. Based on the results from this chapter, I conclude that antler point restrictions would be a useful tool where the management goal is to advance the age structure of the male segment of the white - tailed deer herd. However, my results also indicated that antler point restrictions alone would not a chieve 33 management goals relating to increasing antlerless harvest or improving hunter recruitment and retention. Although, the arguments for antler point restrictions influencing population growth rate and hunter interest have foundations in logic, I was u nable to find definitive empirical support to substantiate these claims. 34 APPENDIX 35 F igure 1 .17 A Map of the areas the Michigan Department of Natural Resources (MDNR) used to summarize deer harvest data in the state for the annual hunting seasons. The figure was copied from the Michigan Deer Harvest Survey Report f o r the Hunting Seasons in 2000 (MDN R Wildlife Report N o. 3344 , Frawley 2001 ). 36 LITERATURE CITED 37 LITERATURE CITED Adams, K. P., and R. J. Hamilton. 2011. Management history. Pages 355 377 in D. G. Hewitt, editor. Biology and management of white - tailed deer. CRC Press, Boca Raton, Fl orida, USA. Biederbeck, H. H., M. C. Boulay, and D. H. Jackson. 2001. Effects of hunting regulations on bull elk survival and age structure. Wildlife Society Bulletin 29:1271 1277. Brown, T. L., D. J. Decker, S. J. Riley, J. W. Enck, T. B. Lauber, P. D. Cu rtis, and G. F. Mattfeld. 2000. The future of hunting as a mechanism to control white - tailed deer populations. Wildlife Society Bulletin 28:797 807. Carpenter, L. H., and R. B. Gill. 1987. Antler point regulations the good, the bad, and the ugly. Transacti ons of the Western Association of Game and Fish Commissioners 67:94 107. Connelly, J. W., J. H. Gammonley, and T. W. Keegan. 2012. Harvest management. Pages 202 - 231 in N. J. Silvy, editor. The wildlife techniques manual: management. Volume 2. John Hopkins University Press, Baltimore, Maryland, USA. Cornicelli, L., D. C. Fulton, M. D. Grund, and J. Fieberg. 2011. Hunter perceptions and acceptance of alternative deer management regulations. Wildlife Society Bulletin 35:323 329. Cornicelli, L., and M. D. Grund. 2011. Assessing deer hunter attitudes toward regulatory change using self - selected respondents. Human Dimensions of Wildlife 16:174 182. Crawley, M. J. 2013. The R book. John Wiley and Sons, Chichester, West Sussex, United Kingdom. Curtis, P. D., A. N. Moen, J. W. Enck, S. J. Riley, D. J. Decker, and G. F. Mattfeld. 2000. Human Dimensions Research Unit: IV. Approaching the limits of traditional hunter harvest for managing wtd populations at the landscape level. Cornell University Department of Natura l Resources, Ithaca, New York, USA. deCalesta, D. 2012. Collaboration among scientists, managers, landowners, and hunters: the kinzua quality deer cooperative. Pages 192 209 in J. P. Sands, S. J. DeMaso, M. J. Schnupp, L. A. Brennan, editors. Wildlife scie nce: connecting research with management. CRC Press, Boca Raton, Florida, USA. Decker, D. J., T. L. Brown, and R. J. Gutiérrez. 1980. Further insights into the multiple - satisfactions approach for hunter management. Wildlife Society Bulletin 8:323 331. Decker, D. J., and N. A. Connelly. 1989. Motivations for deer hunting: implications for antlerless deer harvest as a management tool. Wildlife Society Bulletin 17:455 463. 38 Decker, D. J., A. B. Forstchen, E. F. Pomeranz, C. A. Smith, S. J. Riley, C. A. Jaco bson, J. F. Organ, and G. R. Batcheller. 2015. Stakeholder engagement in wildlife management: does the public trust doctrine imply limits? Journal of Wildlife Management 79:174 179. doi:10.1002/jwmg.809 Decker, D. J., S. J. Riley, and W. F. Siemer. 2013. H uman dimensions of wildlife management. Pages 34 50 in P. R. Krausman and J. W. Cain III, editors. Wildlife management and conservation: contemporary principles and practices. John Hopkins University Press, Baltimore, Maryland, USA. Ditchkoff, S. S., E. R. Welch, Jr., R. L. Lochmiller, R. E. Masters, and W. R. Starry. 2001. Age - specific causes of mortality among male white - tailed deer support mate - competition theory. Joutnal of Wildlife Management 65:552 559. Elbeling - Schuld, A. M., and C. T. Darimont. 2017 . Online hunting forums identify achievement as prominent among multiple satisfactions. Wildlife Society Bulletin 41:523 529. Flora, D. B. 2008. Specifying piecewise latent trajectory models for longitudinal data. Structural equation modeling: a multidisci plinary journal 15:513 533. Frawley, B. J. 2001. Michigan deer harvest survey report 2000 seasons. Michigan Department of Natural Resources Wildlife Division Report 3344, Lansing, Michigan, USA. Frawley, B. J. 2012. A ntler point restrictions survey: deer m anagement units in the northwest lower peninsula. Michigan Department of Natural Resources Wildlife Division Report 3555, Lansing, Michigan, USA. Frawley, B. J. 2017. Evaluation of antler point restrictions in deer management units in the northwest Lower P eninsula. Michigan Department of Natural Resources Wildlife Division Report 3638, Lansing, Michigan, USA. Fuller, A. K., D. J. Decker, M. V. Schiavone, and A. B. Forstchen. 2020. Ratcheting up rigor in wildlife management decision making. Wildlife Society Bulletin 44:29 41. Fulton, D. C., and M. J. Manfredo. 2004. A panel design to assess the effects of regulatory Wildlife 9:35 55. Geist, V., S. P. Mahoney, and J. F. Organ. 2001. Why hunting has defined the North American Model of wildlife conservation. Transactions of the North American Wildlife and Natural Resources Conference 66:175 185. Gulsby, W. D., C. H. Killmaster, J. W. Bowers, and K. V. Miller. 2019. Effects of res trictive harvest criteria on antler size of hunter - harvested male white - tailed deer and hunter opportunity. Wildlife Society Bulletin 43:213 221. 39 Hamilton, J., W. M. Knox, and D. C. Guynn. 1995 a . Harvest strategies. Pages 47 57 in K. V. Miller and R. L. Ma rchinton, editors. Quality whitetails: the why and how of quality deer management. Stackpole Books, Mechanicsburg, Pennsylvania, USA. Hamilton, J., W. M. Knox, and D. C. Guynn. 1995 b . How quality deer management works. Pages 7 18 in K. V. Miller and R. L. Marchinton, editors. Quality whitetails: the why and how of quality deer management. Stackpole Books, Mechanicsburg, Pennsylvania, USA. Hansen, L. 2011. Extensive management. Pages 409 451 in D. G. Hewitt, editor. Biology and management of white - tailed dee r. CRC Press, Boca Raton, Florida, USA. Hansen, L. P., E. B. Flinn, J. A. Sumners, X. Gao, and J. J. Millspaugh. 2017. Effects of an antler 41:516 522. Hansen, L. P., J. A. Sumners, R. Reitz, Y. Bian, X. Gao, and J. J. Millspaugh. 2018. Effects of an antler point restriction on deer hunter perceptions and satisfaction in Missouri. Wildlife Society Bulletin 42:607 615. Harper, C. A., C. E. Shaw, J. M. Fly, and J. T. Beav er. 2012. Attitudes and motivations of Tennessee deer hunters toward quality deer management. Wildlife Society Bulletin 36:277 285. Mason, R., and B. Rudolph. 2015. The value of science to state wildlife management. Proceedings of the National Wild Turkey Symposium 11:39 44. Michigan Department of Natural Resources [MDNR]. 2016. Michigan deer management plan. Wildlife Division Report 3626, Lansing, Michigan, USA. Michigan Department of Natural Resources [MDNR]. 2019. DMU management info. < http ://www.michig an.gov/dnr/0,4570,7 - 350 - 79136_79608_81471 - 428859 -- ,00.html>. Accessed 1 December 2019. Miller, C. A., and J. J. Vaske. 2003. Individual and situational influences on declining hunter effort in Illinois. Human Dimensions of Wildlife 8:263 276. 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 con servation. The Wildlife Society Technical Review 12 - 04. The Wildlife Society, Bethesda, Maryland, USA. Ozoga, J. J., E. E. Langenau Jr., and R. V. Doepker. 1995. The north - central states. Pages 210 237 in K. V. Miller and R. L. Marchinton, editors. Quality whitetails. Stackpole Books, Mechanicsburg, Pennsylvania, USA. 40 Pinizzotto, N. 2017. Antler restrictions among most divisive hunting issues. National Deer Alliance. 11 April 2017 . Accessed 11 December 2018. Quality Deer Management Association [QDMA] Staff. 2018. Quality Deer Management Association's whitetail report. Quality Deer Management Association, Bogart, Georgia, USA. Riley, S. J., D. J. Decker, L. H. Carpenter, J. F. Organ, W. F. Siemer, G. F. Mattfeld, and G. Parsons. 2002. The essence of wildlife management. Wildlife Society Bulletin 30:585 593. Robinson, K. F., A. K. Fuller, R. C. Stedman, W. F. Siemer, and D. J. Decker. 2019. Integration of s ocial and ecological sciences for natural resource decision making: challenges and opportunities. Environmental Management 63:565 573. Schroeder, S. A., L. Cornicelli, D. C. Fulton, and M. D. Grund. 2014. What predicts support for antler point restrictions ? Human Dimensions of Wildlife 19:301 308. Seng, P., D. Escher, and M. Harlow. 2017. Understanding the barriers to hunter retention in Michigan: results of focus groups and online survey research. Michigan Department of Natural Resources, and DJ Case & Ass ociates. Lansing, Michigan, USA. Severinghaus, C. W. 1949. Tooth development and wear as criteria of age in white - tailed deer. Journal of Wildlife Management 13:195 216. Smith, R. L., and J. L. Coggin. 1984. Basis and role of management. Pages 571 600 in L . K. Halls, editor. White - tailed deer: ecology and management. Stackpole Books, Harrisburg, Pennsylvania, USA. Stasinopoulos, D. M., and R. A. Rigby. 1992. Detecting break points in generalised linear models. Computational statistics and data analysis 13:4 61 471. Stedman, R., D. R. Diefenbach, C. B. Swope, J. C. Finley, A. E. Luloff, H. C. Zinn, G. J. San Julian, and G. A. Wang. 2004. Integrating wildlife and human - dimensions research methods to study hunters. Journal of Wildlife Management 68:762 773. Str ickland, B. K., S. Demarais, L. E. Castle, J. W. Lipe, W. H. Lunceford, H. A. Jacobson, D. Frels, and K. V. Miller. 2001. Effects of selective - harvest strategies on white - tailed deer antler size. Wildlife Society Bulletin 29:509 520. Wallingford, B. D. 201 2. White - tailed deer antler point restrictions, harvest and survival rates, and deer hunter support: perception versus reality. Dissertation, The Pennsylvania State University, University Park, Pennsylvania, USA. Wallingford, B. D., D. R. Diefenbach, E. S. Long, C. S. Rosenberry, and G. L. Alt. 2017. Biological and social outcomes of antler point restriction harvest regulations for white - tailed deer. Wildlife Monographs 196:1 26. 41 rview to approaches to management for quality hunting. Preceedings of the Western Association of Fish and Wildlife Agencies 67:69 76. Winkler, R., and K. Warnke. 2013. The future of hunting: an age - period - cohort analysis of deer hunter decline. Population and Environment 34:460 480. YoungeDyke, D., R. P. Smith, and J. Brauker. 2017. Mandatory antler point restrictions: pro & con. Michigan Out of Doors. Spring 2017:58 65. 42 CHAPTER 2: ANTLER SIZES IN RECORD DEER ARE INCREASING IN A REGION OF HIGH - QUALITY HABITAT INTRODUCTION Recent evidence suggests antlers of trophy white - tailed deer ( Odocoileus virginianus ) have been getting smaller across North America (Monteith et al. 2013). The hypothesized reason for this trend is that harvest by hunters is non - rando m, and hunters are increasingly interested in opportunities to harvest mature bucks (Connelly et al. 2012). Antler characteristics of white - tailed deer are heritable (Harmel 1983, Allendorf and Hard 2009, Webb et al. 2012), thus nonrandom harvest by hunter s for these characteristics could have genetic implications via artificial selection. Examples of artificial selection have been reported in terrestrial and aquatic systems (Allendorf and Hard 2009). In bighorn sheep ( Ovis canadensis ) selection and harvest of trophy animals has led to smaller horn characteristics (Allendorf and Hard 2009, Monteith et al. 2013). Moreover, exploitation of fisheries tends to impose selection that alters fitness and population viability characteristics (e.g., smaller body sizes and lower reproductive productivity) by removing the older and larger fish (Allendorf and Hard 2009). Strickland et al. (2001) simulated the effects of selective harvest criteria based on antler characteristics on antler size. They found that antler chara cteristics can be used as selective harvest criteria, but widespread application of these criteria may have differing efficiencies across landscapes, due to changes in social and environmental conditions affecting antler growth and development (Strickland et al. 2001). However, genetics is not the only factor affecting antler size. Antler growth in white - tailed deer depends primarily on the age, nutritional intake, and genetic potential of the 43 individual deer (Demarais and Strickland 2011) . Other factors such as condition of the mother, date of birth, health of the individual, and weather conditions may affect body condition and influence antler development (Garroway and Broders 2005, Monteith et al. 2009, Simard et al. 2014). These other factors may have a stronger influence at regional and local geographic scales. Deer are highly adaptable to a variety of landscapes and do well in fragmented habitats (Stewart et al. 2011). Previous studies have shown differences in conditions and characteristics of deer populations from diverse habitats (Hewitt 2011, Demarais and Strickland 2011 and references therein). In Mississippi, antler development was greatest in regions of greater soil fertility (Strickland an d Demarais 2000). Areas of greater nutrition allowed deer to grow larger antlers at younger ages relative to other soil regions (Strickland and Demarais 2000). Variation in antler size has also been attributed to differences in composition of land cover ty pes (Strickland and Demarais 2008). Patterns of variation in antler size and conformation of deer are noticeable when considering geographic regions (Demarais and Strickland 2011), suggesting that processes influential to antler formation vary spatially. H owever, regional analyses of antler sizes in white - tailed deer with antlers that qualify for entry in the Boone and Crockett records (hereafter referred to as record deer) are lacking. Moreover, it is unclear whether the declining trend in antler size of a ll North American records of white - tailed deer reflects trends at smaller geographic scales. Therefore, the goal of this chapter was to understand regional trends and influences on antler sizes of record deer in the Midwest United States. The objectives we re to 1) identify geographical areas where antler sizes of record deer were similar, 2) assess how antler sizes of record deer have changed through time in the Midwest United States, and 3) evaluate ecological influences on antler sizes of record deer. 44 STU DY AREA This research included 9 Midwestern states (857 counties): Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin (Figure 2.1). The region covered 4 ecosystem provinces (Bailey 1983, Bailey 1995) . Much of the Midwest consists of mixed agriculture and forested lands, which is high - quality habitat for an edge species such as white - tailed deer (Alverson et al. 1988). The topography across most of the study area is rolling, but there are some areas with irregular, more rug ged terrain (Bailey 1995). The ecoregions with rugged terrain within study area include Central Appalachians, Driftless Area, Interior Plateau, Ozark Highlands, and Western Allegheny Plateau (Omerni k 1987, Wiken et al. 2011). The Laurentian mixed forest pr ovince characterized the north, the east was characterized by the eastern broadleaf forest (oceanic) province, the south - central region was portrayed by the eastern broadleaf forest (continental) province, and the west - central region was characteristic of the prairie parkland (temperate) province. Furthermore, vegetation of the north, east, and south central were characterized by forests, while the west - central vegetation was described as forest - steppe. A variety of forest species are found throughout the s tudy area including maples ( Acer spp.), oaks ( Quercus spp.), hickories ( Carya spp.), and spruce ( Picea spp.) trees (Omerni k 1987, Bailey 1995, Pierce et al. 2011, Wiken et al. 2011). The climate of counties within the study area is a product of latitude and position relative to the Great Lakes. In general, as locations move farther away from the equator and closer to the poles winters become more severe. Average temperatures in the Northern Lakes and Forests ecoregion range from 2°C to 6°C, with an avera ge of 10°C in the winter (Wiken et al. 2011). Annual temperatures are warmer in the Ozark Highlands where the average ranges from 12°C to 15°C (Wiken et al. 2011). The time available for crop production is known as the frost - free 45 period. The northern exte nt of my study area has the fewest number of days available for crop production, with as few as 95 100 days available in the Lake Manitoba and Lake Agassiz Plain and Northern Lake and Forests ecoregions (Wiken et al. 2011). The Central Irregular Plains and Ozark Highlands ecoregions have some of the longest frost - free periods at 165 235 days and 140 230 days, respectively (Wiken et al. 2011). Nearly all the natural vegetation in the Eastern Corn Belt Plains, Central Corn Belt Plains, and Western Corn Belt P lains has been converted into cropland that mainly produces corn and soybeans. The Western Corn Belt Plains ecoregion is one of the most productive areas in the world for corn and soybeans (Wiken et al. 2011). Additional descriptions of the study area can be found in VerCauteren and Hygnstrom (2011). 46 Figure 2.1 Map of ecoregion classification for each county in study area (Omernik 1987, Bailey 1995). I assigned an ecoregion to each county by determining which ecoregion covered the majority area within the county. These data are from 9 Midwestern states in the United States (Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin). 4 7 METHODS Data on record deer were obtained from the Boone and Crockett Records of North Americ an Big Game. The Boone and Crockett Club established a standardized system for measuring and scoring big game in North America and maintained the data to serve as a baseline for future studies that investigate trends in record animals (Nesbit and Wright 20 16). The system was designed to emphasize bilateral symmetry by penalizing the net score based on the amount of asymmetry; the non - typical category was developed to recognize deer with unusually large amounts of abnormal growth. Detailed measurements of an tler characteristics and specific calculations produce a numerical net score that serves to rank the animals of a particular category (Reneau et al. 2011). I focused on records of white - tailed deer from 2 categories recognized by the Boone and Crockett Clu b, typical white - tailed deer and non - typical white - tailed deer that were harvested in the Midwestern U nited States from 1973 2014. In general, typical record deer have very few or no abnormal points, whereas the antlers of nontypical deer are characterized by numerous abnormal points (Nesbit and Wright 2016). I limited my assessment to records from this time frame because 1973 was the first year that the Boone and Crockett Club began quality control measures for the records (personal communication, Jack Ren eau, Director of Records for the Boone and Crockett Club , 30 June 2015 ). Thus, any records submitted after 1972 were verified for accuracy before being accepted into the record book. Measurements of typical deer and nontypical deer are the same; however, t he 2 categories differ in how they incorporate abnormal points into the final score (Nesbit and Wright 2016). Unlike the final score, the calculations for gross score are indistinguishable between the categories of record deer. The gross score is the sum o f the antler measurements and does not include penalties (i.e., score deductions) for non - symmetry (Nesbit and Wright 2016). Therefore, 48 I used gross score as my response variable because it is more representative of the total amount of antler grown. The In stitutional Animal Care and Use Committee at Michigan State University determined that the acquisition of harvest data from the Boone and Crockett Club was exempt from protocols by the Animal Care and Use Committee. I followed guidelines outlined by the Bo one and Crockett Club for securely processing and storing the data supplied for this research. Climate data from weather stations were obtained from the National Oceanic and (NCEI). I used monthly summaries datasets to portray total precipitation (mm) during the spring and summer months (PRECIP) and number of days with 2.54 cm or more snow on the ground (SNOW). These data were provided as points for each weather station. Howe ver, I do not know the exact point location where each record deer was harvested, so I used block kriging to interpolate the climate data for each county. I fit a variogram model to the climate data and predicted climate values from a kriged model (for pre diction results see Figure 2.2 and Figure 2.3). This process was repeated for each year and both climate variables (package gstat v1.1 - 0, Program R v3.2.2, R Development Core Team .) I used ArcGIS to reclassify the 2001, 2006, and 2011 National Land Cover Database (Fry et al. 2011) from 16 to 7 classes (i.e., agriculture, forest, rangeland, developed, wetlands, water, and other). Using the reclassified raster, I calculated landscape me trics in program FRAGSTATS (v4.1, University of Massachusetts, Amherst). The percent cover of the 7 land cover classes for each county was used to characterize the composition of the landscape, and the percent of forest cover (FOR) and percent of agricultu re cover (AG) were used in analyses. The interspersion - juxtaposition index (IJI) and contagion value (CONTAG) for each county characterized the configuration of the 7 land cover classes. 49 I used the Soil Survey Geographic Database (SSURGO) to quantify the National Commodity Crop Productivity Index (CPI) model for each county. Crop productivity is an interpretation of the capacity for soils, landscapes, and climates to produce non - irrigated commodity crops such as corn, soybeans, grains, and cotton (Dobos et al. 2008) . I used the coefficient of variation (CV) for the crop productivity value within each county to represent variation in the productivity index. I assumed that variation in crop productivity was important at the county - level because suitable habit at for deer includes mixing of agricultural and non - agricultural (i.e., forested) land (Cain et al. 2019). Moreover, previous research has shown that differences in soil attributes can lead to variation in antler characteristics (Strickland and Demarais 20 00). I calculated the variance inflation factor (VIF) for the full set of potential variables to assess which variables were highly related (Zuur et al. 2007) and used the backward selection process to remove variables until VIF values of all remaining va riables were low ( , Zuur et al. 2009). I also used correlation matrices to determine the degree of collinearity in explanatory variables. I considered 2 explanatory variables to be collinear when the correlation coefficient was high ( , Pr ogram R v3.1.3, R Development Core Team). The first step I took toward understanding how sizes of antlers change through time and vary across the Midwest was to determine if there were areas with similar antler sizes. To assess trends in identifying the ap propriate grouping method for record deer in the Midwest, I categorized counties using three a priori variables: County (no grouping), State (IA, IL, IN, KY, MI, MN, MO, OH, or WI), and Level 3 Ecoregion (Bailey 1995). For counties crossing ecoregion bound aries, I used the majority ecoregion type within the county (Figure 2.1). For each category, I evaluated 2 random effects structures for variable groupings, the random 50 intercept model and the random intercept and slope model (Zuur et al. 2009, Table 2.2). Models were fit using the lme4 package in R (version 3.6.1). The random intercept model assumes that the groups follow the same trend in antler size over time but there is variation in antler size among the groups. The random intercept and slope model assumes that groups vary in antler size and follow different trends in antler size over time. If the random intercept and slope structure fits best, it would provide evidence of variation in the magnitude or direction of trends in antler rion with correction (AIC c ) for small sample size to evaluate the set of model structures for each grouping method (Gelman and Rubin 1992). c values between 0 2 were competing models for explaining the underlying structure of the data ( AICc modavg package, R version 3.2.2). To test hypothesized relationships between antler sizes and environmental characteristics, I developed 9 linear mixed - effects models using the random effects structure with the most support. The hypotheses I tested were di fferent combinations of year, soil, climate (snow days and precipitation), and landscape (percent forest, interspersion - juxtaposition index) covariates (see Table 2.2). I fit the 9 models using the lme4 package in R (version 3.2.2). I considered any model 2 to be competing models for explaining the data ( AICcmodavg package, R version 3.2.2). I created a line graphs to display changes in antler size through time, and to visualize the spatial relationships with results from the top - ranking model. Additionally, a map was produced to demonstrate group - level differences in mean antler sizes of record deer. 51 RESULTS I analyzed records for 2,900 nontypical deer and 2,846 typical deer from the Boone and Crockett Records of North American B ig Game from 1973 2014. Gross scores of all record deer included in this analysis ranged from 434.0 863.6 cm, with an average antler size of 505.7 cm ( SD = 46.6). The percent of agriculture cover (AG) was omitted from further analysis due to VIF value abov e the cutoff. The 2 landscape configuration variables were collinear, thus the contagion value (CONTAG) was also omitted from further analysis. The following 6 variables were included in subsequent analyses: year of harvest (YEAR), percent of forest cover (FOR), interspersion - juxtaposition index (IJI), number of days where snow depth was 2.54 cm or more (SNOW), total precipitation during the spring and summer months (PRECIP), and variation in the crop productivity index (CPI_CV). I found that the intercept - only random effect structure for the ecoregion model was the c = 0.00, weight = 0.87), and the state and county models received minimal support (Table 2.1). Among the 6 candidate models used to evaluate an appropriate random effects str ucture, I found consistent support for models using the Intercept - only temporal random effects structure regardless of grouping method (Table 2.1). Therefore, I used the intercept - only structure with a random effect with grouping by ecoregion to evaluate e nvironmental factors influencing gross scores of record deer through time. Records were not evenly distributed among ecoregions (Figure 2.4). For the environmental analysis, I found 4 competing models, indicating some support for 4 of the hypotheses that c = c c = 0.15, c = 0.29, weight = 0.23, Table 2.3). Variables included in 52 all competing models (i.e., YEAR [ ] , FOR [ ] , and IJI [ ] ) had consistent effects on the response variable (Table 2.3, Table 2.4, Table 2.5, Table 2.6). I found a positive relationship between gross scores and YEAR ( , but the proportion of forest ( and degree of agriculture - forest interspersion ( were negatively associated with antler sizes for ecoregions (Table 2.3). Moreover, the best - supported models included the FOR and IJI covariates, whereas these covariates are not part of the remaining models. Mean antler sizes differed among ecoregions in the Midwest United States ( , Figure 2.6). The Driftless ecoregion had the smallest mean antler sizes ( ) and the Erie Drift Plain ecoregion had the largest mean antler sizes ( ). 53 Figure 2.2 Variograms and county - specific prediction maps for years 1984 and 2004 for the number of days where snow depth was 2.54 cm or more (SNOW). The darker shade indicates fewer days where snow depth was one or more inches. These data are from 9 Midwe stern states in the United States (Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin). (a) 1984 variogram is a Gaussian model with nugget = 26.54, sill = 187.48, and range = 6.87. (b) 2004 variogram is a Gaussian mod el w ith nugget = 42.43, sill = 93.07, and range = 3.15. (c) Plot shows predicted number of days in 1984. (d) Plot shows predicted number of days in 2004. 54 Figure 2.3 Variograms and county - specific prediction maps showing total precipitation for years 1984 and 2004 during the spring and summer months (PRECIP). The darker shades indicate greater precipitation during the spring and summer months. These data are from 9 Midwestern states in the United States (Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin). (a) 1984 variogram is a Spherical model with nugget = 2,678.59, sill = 8,876.58, and range = 6.53. (b) 2004 vario gram is a Spherical model with nugget = 3,618.99, sill = 8,035.93, and range = 4.97. (c) Plot shows predicted a mount of precipitation in 1984. (d) Plot shows predicted amount of precipitation in 2004. 55 Table 2.1 Model comparisons evaluating random effects structure for identifying geographical areas where antler sizes of reco rd deer have been similar in the Midwest United States from 1973 2014. Boone and Crockett records of white - tailed deer from Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin were included in the analysis. 56 Figure 2.4 Distribution of density (deer/km 2 ) in record deer among ecoregions in the Midwest United States from 1973 2014. These data are from 9 Midwestern states in the United States (Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wiscon sin). 57 Table 2.2 Model comparisons relating climate/weather, landscape composition, landscape configuration, and soil characteristic s to the gross score (cm) of record white - tailed deer. Boone and Crockett records of white - tailed deer harvested in Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin during years of study (1973 2014) were included in the analysis. ** ** Model Abbreviations: Gross score (GS), year of harvest (YEAR), percent of forest cover (FOR), interspersion - juxtaposition index (IJI), number of days where snow depth was 2.54 cm or more (SNOW), total precipitation during the spring and summer months (PRECIP), and variation in the crop productiv ity index (CPI_CV) 58 Standard Error (SE) for the Habitat Model hypothesis, which states that the driving factors of antler size among record deer are habitat features and year. Boone and Crockett records of white - tailed deer harvested in Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin during years of study (1973 2014) were included in the analysis. 0.13 0.39 0.001 0.18 Standard Error (SE) for the Landscape Model hypothesis, which states that the driving factors of antler size among record deer are the configuration and composition of habitat and year, whereas the influences of climate an d the longer - term impacts of nutr ient cycling in the soil are negligible. Boone and Crockett records of white - tailed deer harvested in Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin during years of study (1973 2014) were included in the analysis. 0.13 0.001 0.21 59 Standard Error (SE) for the Recent Model Hypothesis, which states that the driving factors of antler size among record deer are recent changes in habitat, weather, and year, whereas the longer - term impacts of nutrient cycling in the soil and climate are n egligible. Boone and Crockett records of white - tailed deer harvested in Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin during years of study (1973 2014) were included in the analysis. 0.10 0.34 0.001 0.23 Standard Error (SE) for the Full Model, which states that all covariates of interest, considered in this analysis, are driving antler size in the Midwest United States (i.e., Illinois, Indiana, Iowa, Kentucky, Michigan, Minne sota, Missouri, Ohio, and Wisc onsin) during years of study (1973 2014). 0.23 0.56 0.37 0.41 0.001 0.18 60 Figure 2.5 Trends in antler size of record deer for each ecoregion in across 9 Midwestern states in the United States (Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin). Each line has a positive slope (0.16 cm/year) with a uni que intercept that represents the trend in the size of antlers for individual ecor egion from 1973 2014. 61 Figure 2.6 Magnitude and direction of change in mean antler sizes for each county relative to the average antler size of entire study area (1973 2014). Counties are grouped by which ecoregion covered the greatest amoun t of area within the county . This map shows that deer harvested in the western regions, southern Illinois, and eastern Ohio have larger antlers relative to other parts of the Midwest. 62 DISCUSSION My findings underscored the importance of considering spat ial and environmental context when investigating trends in wildlife populations. My results showed that antler sizes of record deer were most similar at scales relevant to ecoregions in the Midwestern United States. By grouping observations according to ec oregion and using the random intercept model, I was able to evaluate factors influencing trends in antler sizes of record deer and inherent differences in mean antler sizes among ecoregions. I found that antler sizes of record deer have been increasing in the Midwest since 1973. My findings also suggest that landscape factors were important for explaining antler size variability in record deer of the Midwestern United States. The aggregation of antler sizes at the ecoregion - level suggests evidence of varia tion in antler sizes of record deer between different geographical contexts (Duncan et al. 1998). Ecoregions are characterized as areas with similar habitat and climate and can be useful toward addressing environmental issues and questions across large sca les because they transcend political boundaries (Bailey 1995, Omernik 1995). I found no evidence that similarities in antler size of record deer in the Midwest were due to state - or county - level management strategies. My findings demonstrate that the spati al clustering of similar antler sizes through ti me is related to the ecoregion context rather than a management context. The resources available for antler growth influence the trends in antler size of record deer. Record deer within the context of an ecoregion experience similar influences on antler size through time. Given the various patterns and processes that characterize the delineation of each ecoregion (Bailey 1995), I expected to find that antler sizes in so me areas were increasing while antler sizes in other areas were decreasing. If trends in antler sizes were increasing in some areas of the Midwestern US while decreasing in others, I would have found greater support for the 63 random slope and random intercep t model. Instead, my results suggest that the random intercept model fit the data the best, indicating that trends in antler sizes among areas in the Midwestern US are trending at the same rate in the same direction (Figure 2.5). My finding that there is n o evidence to support differing trends in antler size across the Midwest could be because (1) trends in antler size are not different or (2) sample sizes in individual ecoregions are small. The latter is an artifact of the modeling framework in which there is not enough data to detect varying trends among ecoregions. Therefore, the parallel trends across all ecoregions in the Midwest are a manifestation of the rigid structure imposed on the data under random intercept models (Duncan et al. 1998). There is no single scale at which ecological phenomena should be studied (Levin 1992), and conclusions drawn from analyses of organisms at one scale may not be applicable at other scales (Wiens 1976, Turner 1989, Wiens 1989, Turner et al. 1995). At the continental scale, antlers of record deer appear to be getting smaller (Monteith et al. 2013), whereas my findings suggest that in the Midwest United States, antlers of record deer across all ecoregions are getting larger through time (Figure 2.5). Therefore, trends i n antler sizes of record deer are not consistent across changing spatial scales. Similar conclusions were reported by Festa - Bianchet et al. (2015) that found the broad - scale conclusions in the Monteith et al. (2013) study did not reflect trends in horn siz es of a local population of big horn sheep ( Ovis canadensis ). If antler sizes are getting larger in the Midwest but declining overall, as suggested by Monteith et al. (2013), then antler sizes of record deer must be getting smaller elsewhere in North Ameri ca. For a complete look at the trends across North America, spatially explicit analyses, like the one I present here, must be conducted for the remaining areas of the continent. 64 The increasing trend in antler size of record deer requires something to have been changing in the Midwest to cause antlers to get bigger. Antler growth in deer is driven by the genetic code, nutritional condition, and age of an individual (Monteith et al. 2009, Demarais and Strickland 2011). The harvesting of free - ranging white - tai led deer by hunters is a phenotypically nonrandom selection and previous studies disagree on how far the impacts reach. Some studies suggest selective harvesting of yearling males is not likely to influence the genetic potential of antler growth (low heritability , Lukefahr and Jacobson 1998, Webb et al. 2012, Hewitt et al. 2014, Webb et al. 2014) , usually by citing the complex interactions of the envir onmental factors and the various injuries affecting antler development. Other studies suggest that selective harvest at young ages can impact the antler size of deer remaining in the cohort at later ages (Strickland et al. 2001, Lockwood et al. 2007, Hewitt et al. 2014, Ramanzin and Sturaro 2014) , because of the relationship between yearling antler size and antler size at later ages. Given that record deer are harvested animals, I speculate that the increase in antler size is related to changes in cultural practice that interact with the high - quality habitat of the Midwest resulting in larger antlers. For example, there is a growing interest among hunters for regulations designed to produce more mature bucks (Ozoga et al. 1995, Cornicelli and Grund 2011, Connelly et al. 2012, Harper et al. 2012). Quality Deer Management (QDM) is a management paradigm focused on reducing year ling buck harvest and maintaining appropriate antlerless harvests to improve herd health and quality (Hamilton et al. 1995, Adams and Hamilton 2011). The decline of yearling males in the harvest is an indication of the spread of the management paradigm (Ad ams and Hamilton 2011). The percentage of yearlings in the male harvest is decreasing in the Midwest and other areas in the United States ( Quality Deer Management Association [ QDMA ] Staff 2017). Moreover, forage quality for white - tailed deer is heterogeneous across space (Hewitt 2011), with a greater 65 proportion of the Midwest providing high quality forage for white - tailed deer, relative to the proportion of suitable habitats across a ll of North America (VerCauteren and Hygnstrom 2011). Therefore, the interaction between practices and habitat (e.g., QDM practices in conjunction with good deer habitat) that leads to the increasing trend in antler sizes of record deer in the Midwest. Rec ord deer harvested in Iowa, southern Minnesota, southern Illinois, and eastern Ohio, have larger antlers relative to other parts of the Midwest (Figure 2.6). However, the greatest density of record deer was harvested from the Driftless ecoregion (Figure 2. 4), which had the smallest antler sizes relative to other ecoregions. Therefore, areas where record deer were harvested regularly were not necessarily the areas producing deer with the largest antlers. Differences in average antler sizes could represent ho w energy demands of deer can vary among ecoregions, and what it means for offspring in area, as evidenced by the long - lasting consequences of deficient maternal investment. The environmental conditions experienced by male deer during gestation and early in life can have life - long influences on growth and development (Verme and Ullrey 1984, Monteith et al. 2009). Energy demands are greatest for female deer during pregnancy and lactation (Hewitt 2011). Deer are ruminants, and physical limitations of their dig estive track impose limitations on forage intake ( Ditchkoff 2011, Hewitt 2011). The daily energy requirements of lactation exceed the amount of food that females can ingest. To meet the energy costs for reproduction despite intake limitations, lactating fe males must sacrifice body condition and metabolize stored nutrients to provide offspring with resources for growth (Hewitt 2011). Although it is uncommon for females to produce antlers, there is evidence that the nutritional condition of females (i.e., mat ernal effects) in a population can be a driving factor in the body and antler sizes of males (Monteith et al. 2009). Monteith et al. (2009) showed that deer born to females in poorer condition showed reduced antler sizes. Deficient nutrition early in life 66 can have lasting impacts on the growth and development of deer, even if those deer have access to high - quality forage. Therefore, the differences in antler sizes among ecoregions may represent disparities in the initial condition of male offspring. Any mod el that included landscape configuration (IJI) and composition (FOR) covariates was competitive, which suggests that landscape factors were important for explaining antler size variability. In Mississippi, Strickland and Demarais (2008) reported that lands cape composition explained variation in antler characteristics. My results support these findings, because IJI was not statistically significant. The negative relationship between proportion of forest and antler size suggests some interesting possibilities : (1) deer antlers are not able to grow as large on deer occupying densely forested areas or (2) deer with large antlers are less likely to be harvested in heavily forested areas. Forests provide woody vegetation that deer use for cover (Stewart et al. 201 1), which protects deer from predators and hunters. Forest cover has been shown to decrease the amount of forage biomass available to deer (Stransky 1969, Conroy et al. 1982). In addition, my forest cover covariate shared an inverse relationship with perce ntage of agriculture in Midwest ( ). Therefore, the negative relationship between antler sizes and the amount of forest cover could represent that deer antlers grow larger when an area is dominated by agriculture. This interpretation would align with previous work showing that early successional forage improves phenotypic quality of deer (Strickland and Demarais 2008, Simard et al. 2014). In this chapter, I sought to understand how antler sizes of record deer changed through time and across space. Thi s spatial component produced findings potentially of interest to managers and ecologists. My analysis of record deer harvests in the Midwest United States showed that antler sizes have been increasing across this region of high - quality deer habitat since 1 973. This result contrasts with findings of previous work that reported a decline in the antler sizes of record 67 deer in North America over the last 100 - years (Monteith et al. 2013). These conflicting results can be explained by differences related to scale and habitat between the studies. Managers may find the findings from this chapter useful when communicating with hunters or landowners about their expectations for the deer in the area. Moreover, because the differences in average antler sizes followed ec ological delineated boundaries, rather than political boundaries, future studies using record deer might consider including these measures of ecological patterns and processes. The findings I report in this chapter are useful beyond the ecology and managem ent of record deer, as support for the critical importance of scale considerations in ecological research. 68 LITERATURE CITED 69 LITERATURE CITED Adams, K. P., and R. J. Hamilton. 2011. Management history. Pages 355 377 in D. G. Hewitt, editor. Biology and management of white - tailed deer. CRC Press, Boca Raton, Florida, USA. Allendorf, F. W., and J. J. Hard. 2009. Human - induced evolution caused by unnatural selection through harvest of wild animals. Proceedings of the National Academy of Science s 106:9987 9994. Alverson, W. S., D. M. Waller, and S. L. Solheim. 1988. Forests too deer: e dge effects in northern Wisconsin. Conservation Biology 2:348 358. Bailey, R. G. 1983. Delineation of ecosystem regions. Environmental Management 7:365 373. Bailey, R. G. 1995. Description of the ecoregions of the United States. Second edition. Forest Service Miscellaneous Publication 1391, Washington, D.C., USA. Cain, R. L., N. P. Snow, J. C. Crawford, D. M. Williams, and W. F. Porter. 2019. Spatial distribution and landscape associations of large - antlered deer. The Journal of Wildlife Management 83:1762 1772. Connelly, J. W., J. H. Gammonley, and T. W. Keegan. 2012. Harvest management. Pages 202 231 in N. J. Silvy, editor. Th e wildlife techniques manual. John Hopkins University Press, Baltimore, Maryland, USA. Conroy, M. J., R. G. Oderwald, and T. L. Sharik. 1982. Forage production and nutrient concentrations in thinned loblolly pine plantations. The Journal of Wildlife Management 46:719 727. Cornicelli, L., an d M. D. Grund. 2011. Assessing deer hunter attitudes toward regulatory change using self - selected respondents. Human Dimensions of Wildlife 16:174 182. Demarais, S., and B. K. Strickland. 2011. Antlers. Pages 107 145 in D. G. Hewitt, editor. Biology and ma nagement of white - tailed deer. CRC Press, Boca Raton, Florida, USA. Ditc hkoff, S. S. 2011. Anatomy and p hysiology. Pages 43 73 in D. G. Hewitt, editor. Biology and management of white - tailed deer. CRC Press, Boca Raton, Florida, USA. Dobos, R., H. Sinclair, and K. Hipple. 2008. User guide national commodity crop productivity index (NCCPI) version 1.0. National Soil Survey Center, Lincoln, Nebraska, USA. Duncan, C., K. Jones, and G. Moon. 1998. Context, composition, and heterogeneity: using multileve l models in health research. Social Science & Medicine 46:97 117. 70 Festa - Bianchet, M., S. Schindler, and F. Pelletier. 2015. Record books do not capture population trends in horn length of bighorn sheep. Wildlife Society Bulletin 39:746 750. Fry, J., G. Xia n, S. Jin, J. Dewitz, C. Homer, L. Yang, C. Barnes, N. Herold, and J. Wickham. 2011. Completion of the 2006 national land cover database for the conterminous United States. Photogrammetric Engineering and Remote Sensing 77:858 864. Garroway, C. J., and H. G. Broders. 2005. The quantitative effects of population density and winter weather on the body condition of white - tailed deer ( Odocoileus virginianus ) in Nova Scotia, Canada. Canadian Journal of Zoology 83:1246 1256. Gelman, A., and D. B. Rubin. 1992. Inf erence from iterative simulation using multiple sequences. Statistical Science 7: 457 472. Hamilton, J., W. M. Knox, and D. C. Guynn. 1995. How quality deer management works. Pages 7 18 in K. V. Miller and R. L. Marchinton, editors. Quality whitetails: the why and how of quality deer management. Stackpole Books, Mechanicsburg, Pennsylvania, USA. Harmel, D. E. 1983. Effects of genetics on antler quality and body size in white - tailed deer. Pages 339 348 in R. D. Brown, editor. Antler d evelopment in Cervidae. Caesar Kleberg Wildlife Research Institute, Kingsville, Texas, USA. Harper, C. A., C. E. Shaw, J. M. Fly, and J. T. Beaver. 2012. Attitudes and motivations of Tennessee deer hunters toward quality deer management. Wildlife Society Bulletin 36:277 285. Hewi tt, D. G. 2011. Nutrition. Pages 75 105 in D. G. Hewitt, editor. Biology and management of white - tailed deer. CRC Press, Boca Raton, Florida, USA. patterns of a Management 78:976 984. Levin, S. A. 1992. The problem of pattern and scale in ecology: the Robert H. MacArthur Award Lecture. Ecology 73:1943 1967. Lockwood, M. A., D. B. Frels, W . E. Armstrong, E. Fuchs, and D. E. Harmel. 2007. Genetic and 71:2732 2735. Lukefahr, S. D., and H. A. Jacobson. 1998. Variance component analysis and heritability of antler traits in white - tailed deer. The Journal of Wildlife Management 62:262 268. Monteith, K. L., R. A. Long, V. C. Bleich, J. R. Heffelfinger, P. R. Krausman, and R. T. Bowyer. ctures in trophy ungulates. Wildlife Monographs 183:1 28. 71 Monteith, K. L., L. E. Schmitz, J. A. Jenks, J. A. Delger, and R. T. Bowyer. 2009. Growth of male white - tailed deer: consequences of maternal effects. Journal of Mammalogy 90:651 660. Nesbit, W. H., and P. L. Wright. 2016. How to score north american big game: Boone and Crockett Club's official measurers manual. Fourth edition. Boone and Crockett Club, Missoula, Montana, USA. Omernik, J. M., 1987. Map supplement: ecoregions of the conterminous United States. Annals of the Association of American Geographers 77:118 125. Omernik, J. M., 1995. Ecoregions: a spatial framework for environmental management. Pages 49 62 in W. S. Davis, and T. P. Simon, editors. Biological assessment and criteria: tools for water resource planning and decision making. Lewis Publishers, Boca Raton, Florida, USA. Ozoga, J. J., E. E. Langenau Jr., and R. V. Doepker. 1995. The north - central stat es. Pages 210 237 in K. V. Miller and R. L. Marchinton, editors. Quality whitetails. Stackpole Books, Mechanicsburg, Pennsylvania, USA. Pierce, R. A. II, J. A. Sumners, and E. B. Flinn. 2011. Ecology and management of white - tailed deer in Missouri. Univers ity of Missouri Cooperative Extension, Columbia, USA. Quality Deer Management Association [QDMA] Staff. 2017. Quality Deer Management Association's whitetail report. Quality Deer Management Association, Bogart, Georgia, USA. Ramanzin, M., and E. Sturaro. selectivity in roe deer. European Journal of Wildlife Research 60:1 - 10. Reneau, J., J. Spring, and E. L. Buckner. 2011. Records of North American big game. Thirteenth edition. Boone and Cro ckett Club, Missoula, Montana, USA. Simard, A., J. Huot, S. De Bellefeuille, and S. D. Côté. 2014. Influences of habitat composition, plant phenology, and population density on autumn indices of body condition in a ldlife Monographs 187:1 28. Stewart, K. M., R. T. Bowyer, and P. J. Weisberg. 2011. Spatial use of landscapes. Pages 181 217 in D. G. Hewitt, editor. Biology and management of white - tailed deer. CRC Press, Boca Raton, Florida, USA. Stransky, J. J. 1969. De er habitat quality of major forest types in the South. Proceedings of a symposium: white - tailed deer in the southern forest habitat. Southern Forest Experiment Station, U.S. Department of Agriculture, Nacogdoches, Texas, USA. Strickland, B. K., and S. Demarais. 2000. Age and regional differences in antlers and mass of white - tailed deer. Journal of Wildlife Management 64:903 911. 72 Strickland, B. K., and S. Demarais. 2008. Influence of landscape composition and structure on antler size of white - tailed deer . Journal of Wildlife Management 72:1101 1108. Strickland, B. K., S. Demarais, L. E. Castle, J. W. Lipe, W. H. Lunceford, H. A. Jacobson, D. Frels, and K. V. Miller. 2001. Effects of selective - harvest strategies on white - tailed deer antler size. Wildlife S ociety Bulletin 29:509 520. Turner, M. G. 1989. Landscape ecology: the effect of pattern on process. Annual Review of Ecology and Systematics 20:171 197. BioScience 45 :S29 S35 VerCauteren, K. C., and S. E. Hygnstrom. 2011. Managing white - tailed deer: Midwest North America. Pages 501 535 in D. G. Hewitt, editor. Biology and management of white - tailed deer. CRC Press, Boca Raton, Florida, USA. Verme, L. J., and D. E. Ullr ey. 1984. Physiology and nutrition. Pages 91 118 in L. K. Hal ls, editor. White - tailed deer: e cology and management. Stackpole Books, Harrisburg, Pennsylvania, USA. Webb, S. L., K. L. Gee, R. W. DeYoung, and S. M. Harju. 2014. Variance component analysis of body mass in a wild population of deer ( Odocoileus virginianus ): results from two decades of research. Wildlife Research 40:588 598. Webb, S. L., S. Demarais, B. K. Strickland, R. W. DeYoung, B. P. Kinghorn, and K. L. Gee. 2012. Effects of selective harve approach. The Journal of Wildlife Management 76:48 56. Wiens, J. A. 1976. Population responses to patchy environments. Annual Review of Ecology and Systematics 7:81 120. Wiens, J. A. 1989. Spatial scaling in ecology. Functional Ecology 3:385 397. Wiken, E., F. J. Nava, and G. Griffith. 2011. North American terrestrial ecoregions level III. Commission for Environmental Cooperation. Montreal, Canada. Zuur, A. F., E. N. Ieno, and G. M. Smith. 2007. Analysing ecological data. Springer Science+Business Media, New York, New York, USA. Zuur, A. F., E. N. Ieno, N. J. Walker, A. A. Saveliev, and G. M. Smith. 2009. Mixed effects models and extensions in ecology with R. Springer Science+Business Media, New York, New York, USA. 73 CHAPTER 3: MANAGEMENT INFLUENCES ON THE NUMBER OF WHITE - TAILED DEER IN THE BOONE AND CROCKETT RECORDS INTRODUCTION Harvest of deer by hunters is a non - random selection process and constrained by hunting regulations (Connelly et al. 2012). Vari ation in antlers can be observed throughout the distribution of white - tailed deer ( Odocoileus virginianus ), where differences in antler conformation and size are noticeable across landscapes (Demarais and Strickland 2011). White - tailed deer with antlers that qualify for entry in the Boone and Crockett records (hereafter referred to as record deer) are conceiv ably the fittest and highest quality mates that females seek (Verme and Ullrey 1984, Pierce et al. 2012, Morina et al. 2018). Antler growth is condition - dependent (Andersson 1986) and primarily influenced by age, nutrition, and genetics of the individual d eer (Harmel 1982, Goss 1983, Scribner et al. 1989, Brown 1990, Demarais and Strickland 2011). Antlers are energetically expensive to produce (Brown 1990), so large antlers may serve as an honest indicator of male quality (Zahavi 1975). Moreover, findings f rom previous research provide evidence that antler characteristics are a visual representation of an a , Ditchkoff et al. 2001 b , Demarais and Strickland 2011, Landete - Cas tillejos et al. 2012). Males that can afford the physiological cost to produce large antlers are selected by females over males with smaller antlers (Morina et al. 2018). Hunter selectivity is affected by management regulations (Mysterud et al. 2006, Fest a - Bianchet and Mysterud 2018), and hunting older male deer seems to be increasing in popularity (Adams et al. 2011, Heffelfinger 2013) as indicated by expanding interest in Quality Deer 74 Management (QDM) programs (Connelly et al. 2012, Harper et al. 2012). Hunters and other stakeholders have become more politically active in the decision - making process of setting hunting regulations (Nie 2004). Hunters can encourage managers to implement regulations perceived to have a positive effect on their opportunity to hunt mature deer with large antlers, such as antler restrictions that are often used to minimize harvest of young males and increase average age of male deer in a population (Miller and Marchinton 1995, Connelly et al. 2012). Habitat composition (Strickla nd and Demarais 2008), soil quality (Strickland and Demarais 2000), and land - use configuration (Cain et al. 2019, Cain 2020 Chapter 2) also have effects on antler size, thus harvest criteria based on antler morphology must be specific to the area where i mplemented to be effective. Hunters that want to harvest a record deer tend go where deer with large antlers are expected (Adams et al. 2009, Barrientos 2014, Hayworth 2014). The pursuit of deer with large antlers has a unique place in the hunting communit y (Messner 2011). Every year, numerous local, statewide, and national antler - size contests occur (Bauer 1993). Antlered specimens are measured, scores are calculated, animals are ranked by local and national organizations, and deer with large antlers (e.g. , antler sizes that meet the minimum requirement for Boone and Crockett record deer) are prized in these contests. Although the Boone and Crockett records can be used to locate counties, states, and regions with frequent entries (Demarais and Strickland 20 11, Barrientos 2014, Spring 2014), research to understand the spatial and temporal influences on the distribution of record deer is limited. Therefore, the question that remains is how management regulations have influenced the number of record deer. Moreo ver, it is unclear whether the occurrence and frequency of record deer reflects variation in site - specific conditions (e.g., soil fertility, composition of land - use types, winter severity), a temporal lag from the presence of a record 75 deer in previous year , or both. Therefore, my goal for this chapter was to understand how geographic location, temporal autocorrelation, and management influence the number record deer in the Midwest United States. The objectives of my analysis were to 1) evaluate the relative change in harvests of record deer among ecologically relevant areas, 2) estimate the impact of having a record deer in the previous year, and 3) evaluate the relationship between management strategies and the number of record deer. The spatially comprehen sive and long - term nature of the data collected and maintained by the Boone and Crockett Club provides an opportunity to quantify relationships between management and harvest of record deer. STUDY AREA My research included record deer that were harvested across a large spatial extent (129,087,856 - ha region) of 856 counties from 9 Midwestern states: Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin (Fig. 3.1). Menominee County, Wisconsin (FIPS: 55078), and St Louis City, Missouri (FIPS: 29510) were not included in the analysis. Menominee County, Wisconsin was excluded because the Native American tribes have authority to manage the deer in this area and are not required t o report harvest or hunter information to the state agency. The study area extended over 23 ecoregions (Figure 3.1, Omernik 1987, Bailey 1995), with forest cover being dominant in the north and south, and agriculture dominating the central portion of my s tudy area (Fry et al. 2011). Most agriculture practices focused on corn and soybean production. The topography across most of the study area is rolling, but there are areas with irregular, more rugged terrain (Bailey 1995). The ecoregions with rugged terra in within study area include Central Appalachians, Driftless Area, Interior Plateau, Ozark Highlands, and 76 Western Allegheny Plateau (Omernik 1987, Wiken et al. 2011). The climate of counties within the study area is a product of latitude and position relat ive to the Great Lakes. For additional information about climate and seasonality of my study area, see descriptions by Kunkel et al. (2013). 77 Figure 3.1 Ecoregion classification (Omernik 1987, Bailey 1995) for the counties of 9 states in the Midwestern U nited States included in my study area: Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin. I assigned an ecoregion to each county by determining which ecoregion covered the greatest amount of area within the county. 78 METH ODS Boone and Crockett Record Deer The Boone and Crockett Club established a standardized system for measuring and scoring big game in North America and maintained the data to serve as a baseline for future studies to investigate trends in record animals ( Nesbit and Wright 2016). The system was designed to emphasize bilateral symmetry by penalizing the final score based on the amount of asymmetry; detailed measurements of antler characteristics and specific calculations produce a numerical final score serve s to rank the animals of a particular category (Nesbit and Wright 2016). For this analysis, I focused on records of white - tailed deer from 2 categories recognized by the Boone and Crockett Club, typical white - tailed deer and nontypical white - tailed deer, t hat were harvested in the Midwestern US from 1973 2014. In general, typical record deer have very few or no abnormal points, whereas the antlers of nontypical deer are characterized by numerous abnormal points (Nesbit and Wright 2016). White - tailed deer re Big Game are examples of rare animals. The record deer that are reported represent a subset of all white - tailed deer with large antlers in the population. The Boone and Crockett Club relies on hunters to self - report and a network of trained volunteers (i.e., Official Measures) to generate biological data on harvested animals (Nesbitt and Wright 2016). Therefore, if social, economic, or other types of barriers impede hunters from registering a harvested deer, then that animal is less likely to be reported (see Appendix B). I obtained data for the record deer harvested during 1973 2014 from The Records of North American Big Game (Reneau et al. 2011). I limited my assessment to records from this time frame because 1973 was the first year that the Boone and 79 Crockett Club began quality control measures for the records (personal communications; Jack Reneau, Director of Records for the Boone and Crockett Club, 30 June 2015). Thus, any records submitte d after 1972 underwent a more rigorous verification process to ensure the accuracy of biological data (e.g., measurements, calculations) before being accepted into the record book. I calculated the number of record deer reported annually in each county and used this information as my response variable in my spatially explicit models (Fig. 3.2). 80 Figure 3.2 The number of Boone and Crockett record deer harvested in each county in the Midwestern United States from 1973 2014. My study area covered 9 states: Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin. 81 Zero - inflated Poisson Model To evaluate the relationships between management actions and occurrence of record deer I used a zero - inflated Poisson (ZIP) model (Lambert 1 992) via the jagsUI package (version 3.6.1, R Core Team 2019). Zero inflation is appropriate when data contain more counts of zero than would be expected for a Poisson distribution (Zuur et al. 2009). Given how rare the harvest and reporting of record deer to the Boone and Crockett Club can be, the observed number of record deer harvests for each county and year is more likely to be zero than any other number. Under the ZIP model, counts of record deer were modelled as a mixture of a Bernoulli distribution and a Poisson distribution, and zeros were possible at both levels. I chose the ZIP model over the hurdle (i.e., zero - truncated) model, because I wanted to account for the issues surrounding imperfect detection of record deer by the Boone and Crockett reco rds. I presumed that some of the zero - counts of record deer were due to characteristics of the county that made it unsuitable for producing record deer (i.e., true or structural zeros). However, I also recognized that some of the zeros could be due to impe rfect detection (i.e., false zeros). The ZIP model helped account for unknown factors that led to a record deer going undetected by allowing zero - counts under the Poisson process (Kéry and Schaub 2012). The general format of the ZIP model I used to relate the number of record deer ( ) of county i to a linear predictor of covariates: 82 where is the outcome of a Bernoulli distribution that determines if the observation i (county during a particular year) is suitable for record deer with probability . If the observation is suitable for record deer ( , th en the number of record deer harvested for a county during a particular year is determined by next level in the hierarchy. is the expected mean of suitable sites as a function of covariates on the link scale. To assess the spatial distribution in the number of record deer harvested I included a random effect for ecoregion of the county ( ) . To measure the temporal autocorrelation in the harvest of record deer, I used a covariate ( TAuto ) to represent the status of records in the p revious year. TAuto was a binary variable, and a value of 1 was given if there was a record deer reported in the Boone and Crockett records the year before, otherwise a value of 0 was assigned for the county and year. The variable was included to account f or possible sociological influences in reporting patterns through time. I predicted that a correlative temporal relationship existed in record deer by county, dependent on the presence of a record deer appearing in the previous year. To evaluate the effe cts management regulations on the harvest of record deer in the Midwest United States I included three covariates representing various hunting regulations ( Buck_limit , antler point restrictions [ APR ], and Season_length ). The Buck_limit variable represents the maximum number of antlered males that a hunter may take annually. In some cases, the maximum number of antlered deer a hunter could harvest depended on the type of license they purchased. In such cases, the Buck_limit was always th e greatest number of antlered deer that a hunter could harvest annually. Hunters were able to harvest 1 to 4 antlered deer annually depending on the county and year they hunted. APR was a binary variable. A value of 1 was given to counties during years tha t an antler point restriction was implemented, and a value of 83 zero was given when antler point restrictions were not implemented in the county during the year. The Season_length was the length in days of the gun and muzzleloader seasons ( range = 6 33 days) . I included this variable, because the length of the season has a direct effect on total hunting opportunity. Shorter seasons do not provide as many opportunities (hunting days) as longer seasons. Moreover, the presence of hunters in the field influences deer movements (Little et al. 2014). To make model - fitting process more efficient, I standardized the values for Season Length and Buck Limit. Estimates from the model were based on 3 MCMC chains of 40,000 iterations. After a burn - in of 3,000 iterations wi th a thinning rate of 10 iterations yielded 11,100 total samples from the joint posterior. The Institutional Animal Care and Use Committee at Michigan State University determined that the acquisition of harvest data from the Boone and Crockett Club and sta te agencies was exempt from protocols by the Animal Care and Use Committee. I followed guidelines outlined by the Boone and Crockett Club for securely processing and storing the data supplied for this research. RESULTS A total of 8,236 record deer were har vested and reported to the Boone and Crockett Club over the course of the study period (1973 2014). There were 29,949 (83.3%) county years with 0 counts. The other 6,013 (16.7%) county years had at least 1 record deer harvested and averaged 1.37 record dee r ( ). In 2010, Buffalo County, Wisconsin had 13 record deer, which is the largest count record for a single year. values indicated that the model successfully converged (Table 3.1), because all the values were < 1.1 (Kéry 2010, Gelman et al. 2 013). 84 When I assessed the spatial distribution of record deer, I found that the number of record deer in a county year differed by ecoregion with 6 ecoregions above average (positive intercept) and 6 ecoregions below average (negative intercept) when the a verage number of record deer across the entire Midwest was set to zero (Fig. 3.3, Table 3.2). There were 8 ecoregions with numbers of record deer less than the Midwest average and 14 ecoregions with greater harvests of record deer than the Midwest average. Counties in the Central Appalachians ecoregion had the fewest record deer harvested (0.23 deer/year, credible interval [CI]: 0.10 0.45) relative to the Midwest average, whereas the average number of record deer was greatest in counties of the Driftless Ar ea ecoregion (3.26 deer/year, CI: 2.49 4.29). The status of records in the previous year proved to be a positive predictor of record deer in the Midwest United States. When I measured the temporal autocorrelation of record deer in the Midwest, I found a po sitive relationship between the number of record deer and the existence of an entry in the previous year (Table 3.1). The number of record deer was on average 2.76 more deer (CI: 2.62 2.90) when there was an entry the previous year. When I evaluated the e ffects of management regulations on the harvest of record deer, I found evidence for a significant effect (Table 3.1). Negative predictors of record deer included the length of the hunting season ( Season_length , mean change = 0.71 record deer, CI: 0.68 0.7 4) and the annual limit of antlered deer ( Buck_limit , mean change = 0.82 record deer, CI: 0.79 0.85) for each county. For every 1 - day increase in season length, the number of record deer decreased by 0.71 on average. For every 1 - buck increase in the limit, the number of record deer decreased by 0.82 on average. The magnitude of predictive strength for Season_length was greater than the strength of the relationship between Buck_limit and record deer. The results from my model suggested that fewer record deer would be harvested from areas with longer seasons 85 or more liberal bag limits for antlered deer. The presence of antler point restrictions in an area proved to be a positive predictor of record deer (Table 3.1). The number of record deer harvested from are as with antler point restrictions averaged 1.60 more record deer (CI: 1.40 1.81 deer) than areas that did not have antler point restrictions. 86 Table 3.1 Parameter estimates and associated 95% credible intervals (LCI: Lower Credible Interval, UCI: Upper C redible Interval) from zero - inflated Poisson model for the number of record deer harvested in the Midwest United States (1973 2014). - values indicated that the model successfully converged. Data for analysis included record deer harvested in Illinois, Ind iana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin. Parameters * LCI UCI n.eff Intercept - 1.63 - 1.89 - 1.40 1.02 270 0.62 0.59 0.64 1.00 4,994 TAuto ** 1.014 0.96 1.06 1.00 11,100 APR ** * 0.468 0.34 0.60 1.00 11,100 Buck_limit - 0.197 - 0.24 - 0.16 1.00 4,932 Season_length - 0.341 - 0.38 - 0.30 1.00 3,389 * TAuto represents the status of records in the previous year, Antler Point Restrictions (APR), Buck_limit represents the maximum number of antlered males that a hunter may take annually, Season_length was the length in days of the gun and muzzleloader seasons * * TAuto was a binary variable, and a value of 1 was given if there was a record deer reported in the Boone and Crockett records the year before, otherwise a value of 0 was assigned for the county and year *** APR was a binary variable, and a value of 1 was gi ven to counties during years that an antler point restriction was implemented, and a value of zero was given when antler point restrictions were not implemented in the county during the year 87 Table 3.2 The number of counties included and estimated intercepts for each ecoregion based on a random effect for ecoregion with a mean of zero. I assigned an ecoregion to each county by determining which ecoregion covered the greatest amount of area within the county. Boone and Crockett records of white - taile d deer harvested in Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin during years of study (1973 2014) were included in the analysis. Ecoregion Classification a No. Counties Central Appalachians* 16 - 1.472 Northern Lakes and Forests* 71 - 0.807 Southern Michigan/Northern Indiana Drift Plains* 37 - 0.760 Western Allegheny Plateau* 36 - 0.676 Interior River Valleys and Hills* 96 - 0.507 Northern Minnesota Wetlands* 3 - 0.473 Ozark Highlands 47 - 0.135 Mississippi Alluvial Plain 6 - 0.025 Huron/Erie Lake Plains 21 0.038 Eastern Corn Belt Plains 84 0.046 Western Corn Belt Plains 119 0.057 Southeastern Wisconsin Till Plains 22 0.065 Mississippi Valley Loess Plains 9 0.066 Central Corn Belt Plains 46 0.095 Lake Agassiz Plain 10 0.137 Erie Drift Plain 19 0.262 Central Irregular Plains* 49 0.451 Northern Glaciated Plains* 6 0.515 North Central Hardwood Forests* 44 0.623 Southwestern Appalachians* 6 0.694 Interior Plateau* 85 0.805 Driftless Area* 26 1.191 a Ecoregions with an asterisk ( * ) are significant. Significance was determined by a credible interval (CI) that did not include 0. 88 Figure 3.3 Map showing differences in the number of record deer among the ecoregions across the Midwest United States (i.e., Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, and Wisconsin) during study (1973 2014). I assigned an ecoregion to each county by determining which ecoregion covered the greatest amount of area within the county. Purp le areas are ecoregions with fewer record deer than expected from the zero - inflated Poisson model, whereas the orange areas correspond to ecoregions with more record deer than predicted given the model. Ecoregions that did not have a significant influence (i.e., credible interval included zero) on the number of record deer are white. 89 DISCUSSION My findings provide information about the regional trends and influences on the number of record deer in the Midwestern United States. By grouping observations accor ding to ecoregion, I found that characteristics within ecoregions influenced the number of record deer harvested and reported. My results showed that the average number record deer was greatest in the Driftless ecoregion and smallest in the Central Appalac hians ecoregion. I found evidence for temporal autocorrelation in the records, with reporting the previous year having a positive influence on the number of record deer in the following year. My findings also suggest that management regulations were import ant for explaining variation in the number of record deer harvested and reported. I focused on management variables in the model for this chapter but included a random effect for each ecoregion to account for the ecological and environmental differences am ong these areas. My results demonstrate that when the management covariates (i.e., Buck_limit , Season_length , and APR ) are held constant, there is an ecoregion - level effect. This ecoregion effect could be related to the relationship between habitat quality and antler size demonstrated in previous studies. Antler growth and size are influenced by soil fertility (Strickland and Demarais 2000, Jones et al. 2010) and land - use types that promote or suppress early successional plants (Strickland and Demarais 2008 , Cain 2020 Chapter 2). Much of the Midwest consists of mixed agriculture and forested lands, which is high - quality habitat for an edge species such as white - tailed deer (Alverson et al. 1988). Ecoregions with positive influences on the number of record deer harvested may represent areas where environmental characteristics promote record deer production and harvest. Although record deer harvested in the Driftless ecoregion possess antlers smaller than the average for the Midwest (Cain 2020 Chapter 2), t he number of record deer 90 harvested was greatest in the Driftless ecoregion relative to other ecoregions in the Midwest. Conversely, areas with fewer harvests of record deer than expected may represent places where poorer quality habitats have resulted in f ewer harvest opportunities. Moreover, hunters in the Central Appalachians, for example, may have fewer opportunities to harvest, less success at harvesting, or infrequent reporting of successful harvests when a record deer is present in the area (Appendix B). Therefore, the ecoregion effect may also represent sociological differences among the hunting community. My results suggest that a record deer was more likely to be harvested and reported when at least one record deer was reported in the previous year. This could be a cultural artifact in that a hunter may be more likely to report a record deer in areas where record deer have been reported recently. Given the large amounts of private land in the Midwest, it is common for hunters to pay farmers and other landowners for access to their land for hunting opportunities (Hansen 2011, VerCauteren and Hygnstrom 2011). The landowner may charge any amount that they see fit, thus hunting leases vary in costs. In areas where hunters have the opportunity to harvest m ale deer with large antlers, hunting leases may be more expensive because the opportunity to hunt record deer seems to be important to hunters (Eliason 2008, Whittington 2014). Consequently, if hunters are concerned that reporting the harvest of a record d eer will lead landowners to increase the cost of hunting leases in the area, then the hunter may decide not to report a record deer (Adams et al. 2011). Thus, positive temporal autocorrelation makes sense because hunters may be more likely to report a reco rd deer when there was a record deer reported the year before. My results suggest that implementing management strategies focused on the male segment of the population have the potential to influence record deer harvests in the area. First, 91 the number of record deer was greater in counties that had antler point restrictions (1.5 deer/year) compared to the harvest in counties that did not have these regulations (1 deer/year). Management regulations restricting the harvest of yearling males, such as antler p oint restrictions, increase the probability that they will survive the hunting season and reach older ages. Previous research shows that antler point restrictions protect sub adult males from harvest with a greater proportion of the male harvest consisting of individuals from older age classes (Hansen et al. 2017, Wallingford et al. 2017, Cain 2020 Chapter 1). Antler size increases with the age of the individual (Demarais and Strickland 2011), and counties with antler point restrictions have experienced g reater harvests of adult males with larger antlers (Wallingford et al. 2017). Second, limiting the number of antlered deer that hunters could harvest per year had a positive influence on the number of record deer. Implementing a limit on buck harvest decr eases the overall harvest of male deer, which increases the number of males that survive the hunting season. However, unlike antler point restrictions, limiting the number of bucks does not protect a certain age class. Hunters in the Midwest generally harv est 1 2 deer each year (VerCauteren and Hygnstrom 2011). The values derived from or assigned to a wildlife resource vary from person to person (Conover 1997), but hunters want an opportunity to hunt bucks every year (Cornicelli et al. 2011). By limiting hu nters to 1 buck per year, hunters may become more selective in the buck they choose to harvest. My results suggest a negative relationship between the number of record deer and the length of the hunting season. The harvest under 33 - day hunting season (0.68 deer) was smaller than the expected harvest under a 6 - day hunting season (1.76 deer). This finding makes sense given previous work showing that deer change their movement behaviors to minimize harvest 92 risk during the hunting season (Whitman 2012, Little e t al. 2014, Marantz et al. 2016). The vulnerability of deer to harvest is influenced by the habitat and deer movements (Whitman 2012). Alterations made to the length of the hunting season influence the activity and timing of hunters in the field, and deer respond rapidly to their presence by changing their movement to avoid harvest risk (Little et al. 2016, Marantz et al. 2016). Longer seasons afford hunters more opportunities to get out in the field by offering hunters more days doing so, but my results su ggest that there is not necessarily a corresponding increase in harvests of record deer. This is likely due to the influence of hunting pressure on deer movements, which affects the vulnerability of deer to harvest (Whitman 2012). Roseberry and Klimstra (1 974) reported that hunters tend to select for adults over fawns and males over females when harvesting deer. Diekert et al. (2016) used ideas from economic co sts (e.g., license fees, lease payments, opportunity costs of traveling) associated with hunting and these investments made by the hunter influence their individual threshold value for choosing to shoot an animal. From the perspective of the hunter, any de er valued at or above the threshold are subject to hunting while animals valued below are safe from being shot. The higher the opportunity cost incurred by the hunter, the lower their reservations to harvesting the deer that they see (Diekert et al. 2016). herd, investments made by the individual (Diekert et al. 2016), and the relative importance the hunter places on various characteristics of the animal (e.g., antler size, sex, age). Therefor e, similar to beauty, the value of the deer is in the eye of the beholder. The decision to shoot an animal is unique to the hunter, the value placed on the animal trait, and harvest restrictions and regulations (Mysterud 2011, Ramazin and Sturaro 2014). 93 As interest in quality deer management and harvesting adult males with large antlers increases (Connelly et al. 2012, Harper et al. 2012), it is important for wildlife managers and hunters to understand how regulations can influence harvests of record deer. Although research has been conducted on factors affecting antler growth and development (Demarais and Strickland 2011 and references therein), no studies have investigated the relationship between record deer and management regulations. My analysis demonst rates that changes in management regulations can influence the harvest of record deer in the Midwest with inherent differences among ecoregions. In areas of highly suitable habitat for white - tailed deer, management regulations, such as shorter season and l imiting the harvest of antler deer, can provide enhancements to survival of antlered deer that may result in the additional harvest of record deer. 94 APPENDIX 95 Figure 3. 4 Infographic to show the sequence of critical steps required for a male white - tailed deer that has record - sized antlers Records of North American Big Game. Influential factors that may direct the outcome (Yes/No) of each step are given in the gray boxes . 96 LITERATURE CITED 97 LITERATURE CITED Management Association, Bogart, Georgia, USA. Adams, K. P., J. Management Association, Bogart, Georgia, USA. Alverson, W. S., D. M. Waller, and S. L. Solheim. 1988. Fores ts too deer: e dge effects in northern Wisconsin. Conservation Biology 2:348 358. Andersson, M. 1986. Evolution of condition - dependent sex ornaments and mating preferences: sexual selection based on viability differences. Evolution 40:804 816. Bauer, E. A. 1993. Whitetails: behavior, ecology, and conservation. Voyageur Press, Stillwater , Minnesota, USA. Bailey, R. G. 1995. Description of the ecoregions of the United States. Second edition. Forest Service Miscellaneous Publication 1391, Washington, D.C., USA. Barrientos, R. R. 2014. The elusive whitetail in Texas. Pages 147 167 in J. Cowa n, editor. Arbor, Michigan, USA. Brown, R. D. 1990. Nutrition and antler development. Pages 426 441 in G. A. Bubenik, and A. B. Bubenik, editors. Horns, pronghorns, and ant lers. Springer, New York, New York, USA. Cain, R. L. 2020. Understanding management and landscape influences on the harvest of male white - tailed deer across a large geographic region . Dissertation, Michigan State University, East Lansing , USA. Cain, R. L., N. P. Snow, J. C. Crawford, D. M. Williams, and W. F. Porter. 2019. Spatial distribution and landscape associations of large - antlered deer. Journal of Wildlife Management 83:1762 1772. Connelly, J. W., J. H. Gammonley, and T. W. Keegan. 2012. Harvest mana gement. Pages 202 - 231 in N. J. Silvy, editor. The wildlife techniques manual: management. Volume 2. John Hopkins University Press, Baltimore, Maryland, USA. Cornicelli, L., D. C. Fulton, M. D. Grund, and J. Fieberg. 2011. Hunter perceptions and acceptance of alternative deer management regulations. Wildlife Society Bulletin 35:323 329. Conover, M. R. 1997. Monetary and intangible valuation of deer in the United States. Wildlife Society Bulletin 25:298 305. 98 Demarais, S., and B. K. Strickland. 2011. Antlers. Pages 107 145 in D. G. Hewitt, editor. Biology and management of white - tailed deer. CRC Press, Boca Raton, Florida, USA. Diekert, F. K., A. Richter, I. M. Rivrud, and A. Mysterud. 2016. How constraints affect the ngs of the National Academy of Sciences 113:14450 14455. Ditchkoff, S. S., R. L. Lochmiller, R. E. Masters, S. R. Hoofer, and R. A. Bussche. 2001 a . Odoco ileus virginianus Evolution 55:616 625. Ditchkoff, S. S., R. L. Lochmiller, R. E. Masters, W. R. Starry, and D. M. Leslie Jr. 2001 b . Does fluctuating asymmetry of antlers in white - tailed deer ( Odocoileus virginianus ) follow patterns predicted for sexually selected traits? Proceedings of the Royal Society of London B 268:891 898. Eliason, S. L. 2008. A statewide examination of hunting and trophy nonhuman animals: perspectives of Montana hunters. Society and Animals 16 :256 278. Festa - Bianchet, M., and A. Mysterud. 2018. Hunting and evolution: theory, evidence , and unknowns. Journal of Mamma logy 99:1281 1292. Fry, J., G. Xian, S. Jin, J. Dewitz, C. Homer, L. Yang, C. Barnes, N. Herold, and J. Wickham. 2011. Completion of the 2006 national land cover database for the conterminous United States. Photogrammetric Engineering and Remote Sensing 77:858 864. Gelman, A., J. B. Carlin, H. S. Stern, D. B. Dunson, A. Vehtari, and D. B. Rubin. 2013. Bayesian Data Analysis. Third edit ion. CRC Press, Boca Raton, Florida, USA. Goss, R. J. 1983. Deer antlers: regeneration, function, and evolution. Academic Press, New York, New York, USA. Hansen, L. 2011. Extensive management. Pages 409 451 in D. G. Hewitt, editor. Biology and management o f white - tailed deer. CRC Press, Boca Raton, Florida, USA. Hansen, L. P., E. B. Flinn, J. A. Sumners, X. Gao, and J. J. Millspaugh. 2017. Effects of an antler 41:516 52 2. Harmel, D. E. 1982. Effects of genetics on antler quality and body size in white - tailed deer. Pages 339 348 in R. D. Brown, editor. Antler development in Cervidae. Caesar Kleberg Wildlife Research Institute, Kingsville, Texas, USA. Harper, C. A., C. E. Shaw, J. M. Fly, and J. T. Beaver. 2012. Attitudes and motivations of Tennessee deer hunters toward quality deer management. Wildlife Society Bulletin 36:277 285. 99 Hayworth, K. 2014. Whitetails north of the border. Pages 169 177 in J. Cowan, editor. Boone a Michigan, USA. Heffelfinger, J. R. 2013. Hunting and trapping. Pages 130 143 in P. R. Krausman and J. W. Cain III, editors. Wildlife management and conservation: contempora ry principles and practices. Johns Hopkins University Press, Baltimore, Maryland, USA. Jones, P. D., B. K. Strickland, S. Demarais, B. J. Rude, J. P. Muir, and S. L.Edwards. 2010. Soils and forage quality as predictors of white - tailed deermorphometrics. Wi ldlife Biology 16:430 439. Kéry, M. 201 0. Introduction to WinBUGS for e cologis ts: a Bayesian approach to regression, ANOVA, mixed models and related analyses. Academic Press, San Diego, California, USA. Kéry, M., and M. Schaub. 2012. Bayesian population an alysis using WinBUGS: a hierarchical perspective. Academic Press, San Diego, California, USA. Kunkel, K. E., L. E. Stevens, S. E. Stevens, L. Sun, E. Janssen, D. Wuebbles, S. D. Hillberg, M. S. Timlin, L. Stoecker, N. E. Westcott, and J. G. Dobson. 2013. R egional climate trends and scenarios for the U.S. National Climate Assessment Part 3: climate of the Midwest U.S. United States Department of Commerce and National Oceanic and Atmospheric Administration Technical Report NESDIS 142 - 3, Washington, D.C., USA. Lambert, D. 1992. Zero - inflated Poisson regression, with an application to defects in manufacturing. Technometrics 34:1 14. Landete - Castillejos, T., J. A. Estevez, F. Ceacero, A. J. Garcia, and L. Gallego. 2012. A review of factors affecting antler composition and mechanics. Frontiers in Bioscience 4:2328 2339. Little, A. R., S. L. Webb, S. Demarais, K. L. Gee, S. K. Riffell, and J. A. Gaskamp. 2016. Hunting intensity alters movement behaviour of white - tailed deer. Basic and Applied Ecology 17:360 36 9. Little, A. R., S. Demarais, K. L. Gee, S. L. Webb, S. K. Riffell, J. A. Gaskamp, and J. L. Belant. 2014. Does human predation risk affect harvest susceptibility of white - tailed deer during hunting season? Wildlife Society Bulletin 38:797 805. Marantz, S . A., J. A. Long, S. L. Webb, K. L. Gee, A. R. Little, and S. Demarais. 2016. Impacts of human hunting on spatial behavior of white - tailed deer ( Odocoileus virginianus ). Canadian Journal of Zoology 94:853 861. Messner, T. C. 2011. White - tailed deer managem ent strategies and domestication processes. Human Ecology 39:165 178. 100 Miller, K. V., and L. Marchinton. 1995. Quality whitetails: the why and how of quality deer management. Stackpole Books, Mechanicsburg, Pennsylvania, USA. Morina, D. L., S. Demarais, B. K. Strickland, and J. E. Larson. 2018. While males fight, females choose: male phenotypic quality informs female mate choice in mammals. Animal Behaviour 138:69 74. Mysterud, A. 2011. Selective harvesting of large mammals: ho w often does it result in directional selection? Journal of Applied Ecology 48:827 834. Mysterud, A., P. Tryjanowski, and M. Panek. 2006. Selectivity of harvesting differs between local and foreign roe deer hunters: trophy stalkers have the first shot at t he right place. Biology Letters 2:632 635. Nesbit, W. H., and P. L. Wright. 2016. How to score north american big game: Boone and Crockett Club's official measurers manual. Fourth edition. Boone and Crockett Club, Missoula, Montana, USA. Nie, M. 2004. Stat e wildlife policy and management: the scope and bias of political conflict. Public Administration Review 64:221 233. Omernik, J. M., 1987. Map supplement: ecoregions of the conterminous United States. Annals of the Association of American Geographers 77:11 8 125. Pierce, R. A. II, J. A. Sumners, and E. B. Flin n. 2012. Antler development in white - tailed deer: i mpl ications for m anagement. University of Missouri Cooperative Extension, Columbia, USA. R Core Team. 2019. R: a language and environment for statistic al computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.r - project.org. selectivity in roe deer. European Journal of Wildlife Research 60:1 10. Reneau, J., J. Spring, and E. L. Buckner. 2011. Records of North American big game. Thirteenth edition. Boone and Crockett Club, Missoula, Montana, USA. Roseberry, J. L., and W. D. Klimstra. 1974. Differential vulnerability during a controlled dee r harvest. Journal of Wildlife Management 38:499 507. Scribner, K. T., M. H. Smith, and P. E. Johns. 1989. Environmental and genetic components of antler growth in white - tailed deer. Journal of Mammalogy 70:284 291. Spring, J. 2014. The statistics behind t he trophies. Pages 203 231 in J. Cowan, editor. Boone and Michigan, USA. 101 Strickland, B. K., and S. Demarais. 2000. Age and regional differences in antlers and mass of white - ta iled deer. Journal of Wildlife Management 64:903 911. Strickland, B. K., and S. Demarais. 2008. Influence of landscape composition and structure on antler size of white - tailed deer. Journal of Wildlife Management 72:1101 1108. VerCauteren, K. C., and S. E. Hygnstrom. 2011. Managing white - tailed deer: Midwest North America. Pages 501 535 in D. G. Hewitt, editor. Biology and management of white - tailed deer. CRC Press, Boca Raton, Florida, USA. Verme, L. J., and D. E. Ullrey. 1984. Physiology and nutrition. Pages 91 118 in L. K. Hal ls, editor. White - tailed deer: e cology and management. Stackpole Books, Harrisburg, Pennsylvania, USA. Wallingford, B. D., D. R. Diefenbach, E. S. Long, C. S. Rosenberry, and G. L. Alt. 2017. Biological and social outcomes of antler point restriction harvest regulations for white - tailed deer. Wildlife Monographs 196:1 26. Wiken, E., F. J. Nava, and G. Grif fith. 2011. North American terrestrial ecoregions level III. Commission for Environmental Cooperation. Montreal, Canada. Whitman, B. J. 2012. White - tailed deer movement and habitat interactions prior to death in central New York. Thesis, State University of New York, Syracuse, New York, USA. Whittington, G. 2014. This way to paradise. Pages 111 117 in J. Cowan, editor. Boone and Michigan, USA. Zahavi, A. 1975. Mate selection a selection for a handicap. Journal of Theoretical Biology 53:205 214. Zuur, A. F., E. N. Ieno, N. J. Walker, A. A. Saveliev, and G. M. Smith. 2009. Zero - truncated and zero - inflated models for count data. Pages 261 293 in M. Gail, K. Krickeberg, J. Samet, A . Tsiatis, and W. Wong, editors. Mixed effects models and extensions in ecology with R: statistics for biology and health. Springer Science + Business Media, New York, New York, USA. 102 CHAPTER 4: ADDRESSING UNCERTAINTY IN THE REPORTING PROCESS OF THE BOONE AND CROCKETT RECORDS INTRODUCTION Quantifying the abundance and distribution of animals in a free - ranging population is one of the principal goals of wildlife research (Williams et al. 2002, Kellner and Swihart 2014). A major challenge in obtaining measures of abundance is that some individuals in the population may avoid detection (Link and Sauer 1997). Consequently, a complete count of every individual in a population is usually impossible (Dice 1941, Dénes et al. 2015), and monitoring is generally based on a subset of the total population. The probability that an animal is counted given its presence in the population is known as the detection probability ( , Williams 2001). Ecological models that use counts o f individuals as proxies for abundance assume the detection probability is perfect ( ) or proportionally related to abundance by a constant (Williams 2001, Kéry et al. 2005). In most situations, species are imperfectly detected (i.e., detection probabil ity < 1) and this assumption is violated (Ficetola et al. 2018). Similar to data collected from field observations, data collected from hunter harvest surveys are also subject to issues of imperfect detection (Rosenberry et al. 2004, Goddard and Miller 200 9), because the probability that a hunter reports a successful harvest varies through time and across space (Roseberry and Wool f 1991, Rosenberry et al. 2004). For game species, such as white - tailed deer ( Odocoileus virginianus ), harvest data represent an important source of information that are commonly used by managers to monitor trends in wildlife populations subjected to hunter harvest (Roseberry and Woolf 1991, Brown et al. 2000, Kilpatrick et al. 2005, Goddard and Miller 2009, Monteith et al. 2013). E stimating 103 annual harvests of deer and inferences drawn about harvest trends are essential for informing and appraising management decisions (Rupp et al. 2000). Failing to account for variation in the detection probability (or reporting rate) in harvest dat a can lead to biased estimates and erroneous inferences about trends in harvest (Williams et al. 2002, Rosenberry et al. 2004, Fiske and Chandler 2011, Guillera - Arroita 2016). Issues of imperfect detection, when in reference to harvest estimates, may be ma gnified when animals are rare leading to small sample sizes. The white - tailed deer recorded in the Boone Reneau et al. 2011) are examples of rare animals. Moreove r, the record deer that are reported represent an unknown subset of white - tailed deer with large antlers harvested from a population. The Boone and Crockett Club does not solicit this information, but rather the process of entering a record deer is initiat ed by the individual hunter (Nesbitt and Wright 2016). An anticipated consequence of social, economic, or other types of barriers that impede a hunter from registering a harvested deer is that the animal is less likely to be reported. Therefore, analytical approaches using harvest data reported by hunters must account for imperfect detection to avoid inaccurate inferences about trends in record deer harvest (Mackenzie et al. 2005, Ryan et al. 2019). One of the more recent developments in accounting for imp erfect detection in abundance estimation is the N - mixture modeling framework (Royle 2004). This framework explicitly accounts for imperfect detection by using a hierarchical model to estimate parameters for abundance and detection probability from spatiall y and temporally replicated counts of unmarked individuals (Royle 2004). One level of the hierarchy serves to describe the variation in abundance, while another describes the observations conditional on the abundance (Royle and Dorazio 2008). A third level can be added to the hierarchical model to describe the suitability of 104 sites, which is a useful extension when zero - inflat ion in the data is a concern (Ké ry and Schaub 2012). The flexibility of this modeling framework allows the inclusions of covariates, w hich are linked through a generalized linear model link function, that are believed to influence population abundance, site suitabili ty, or detection probability (Ké ry and Schaub 2012). Despite the flexibility inherent in the n - mixture modeling framework, no one has investigated the applicability of using this modeling framework to evaluate trends in harvest data. Therefore, in this chapter I seek to evaluate whether n - mixture models could be used to address concerns related to imperfect detection (or repor ting rates) for harvest data. My goal is adapt the zero - inflated n - mixture modeling framework to model data on harvests of record deer in Wisconsin from 2 independent record keeping organizations (i.e., Boone and Crockett Club, Wisconsin Buck and Bear Club ). My specific objective for this chapter were to: 1) determine if reporting rates (detection probabilities) show biases in space or time and 2) evaluate the influence of land cover and harvest characteristics on the harvest of record deer. Wisconsin provi des a good place to look at the potential of applying this method because both sources of record deer data have been collecting for many years. STUDY AREA The study area consisted of 71 of the 72 counties in Wisconsin. I excluded Menominee County from this analysis because most of the county is under Menominee Tribal jurisdiction with deer hunting regulated by the Tribal government (Wisconsin Department of Natural Resources 2019). The advancing and retreating of glaciers across the Upper Midwest shaped the landscape of Wisconsin and fundamentally influenced development of ecosystems in the study area. The Driftless Area ecoregion (Bailey 1983, Bailey 1995) covers about 20% of Wisconsin, 105 primarily along the Mississippi River, and denotes an area that escaped the most recent glaciation. Wisconsin is characterized by a mixture of agriculture and forested land. While agriculture occurs statewide, production is more dominant in the southern portion of the state (Wisconsin Department of Natural Resources 2012). The most extensive deciduous forests are found in Northcentral Wisconsin (Wisconsin Department of Natural Resources 2012). During the years of my analysis, the Wisconsin Department of Natural Resources modified some of the regulations for hunting white - taile d deer in the state, including implementing Earn - A - Buck (EAB) regulations intending to increase antlerless harvest, and subsequently decrease population growth rate (McCullough 1984). An annual limit of 2 bucks was imposed on all hunters in Wisconsin for t he duration of this study. METHODS Observed Data Independent Counts of Record Deer For my response variable in this chapter, I obtained records of white - tailed deer for 1981 2014 from 2 independent record - keeping organizations: the Boone and Crockett Club and the Wisconsin Buck and Bear Club (WBBC). The timeframe was determined by data a vailability, and 1981 was the first year that population estimates and detailed summaries of season frameworks were available for each deer management unit (personal communication; Robert Rolley, Population Ecologist for the Wisconsin Department of Natural Resources). The Boone and Crockett Club established a standardized system for measuring and scoring big game in North America and maintained the data to serve as a baseline for future studies that investigate trends in record animals (Nesbit and Wright 20 16). The system was 106 designed to emphasize bilateral symmetry by penalizing the final score based on the amount of asymmetry; detailed measurements of antler characteristics and specific calculations produce a numerical final score that serves to rank the a nimals of a particular category (Reneau et al. 2011). All measurements are the same between typical deer and nontypical deer; however, the 2 categories differ in how they incorporate abnormal points into the final score (Nesbit and Wright 2016). To be entered into the Boone and Crockett records, the animal must be scored by an Official Measurer that was trained by the Club. In addition to the scoring details (e.g., individual measurements, calculated scores), each entry included the county of harvest an d the year it was taken. I organized my observed count data using this information on the location and year of harvest. The WBBC began in 1965 to enhance the information on the harvest of record deer in Wisconsin (WBBC 2019). While the WBBC adheres to the same scoring system as Boone and Crockett, they have their own measurers. That said, 31 (55.4%, Wisconsin Buck and Bear Club 2020) of the state certified official measures are also Boone and Crockett official measures, but not all. Volunteers for the WBBC travel across the state and attend different sporting shows to measure harvested animals. Although the Wisconsin Buck and Bear Club has a lower minimum entry score, I limited the data to those records that met or exceeded the minimum score required by the Boone and Crockett Club (406.4 cm for typical and 469.9 cm for nontypical). This cut off ensured that I did not count any records that would not qualify for entry in the Boone and Crockett Club records. For each county and year, I used the data sets to pro duce two independent counts of record deer from the same population. 10 7 Model C ovariates To evaluate the influence of land cover on the harvest of record deer, I used ArcGIS (version 10.1; Environmental Systems Research Institute, Redlands, California, USA) to reclassify the 2001, 2006, and 2011 National Land Cover Database (Fry et al. 2011) to correspond to important land cover for white - tailed deer (Alverson et al. 1988, Williams et al. 2012, Dechen Quinn et al. 2013, Snow et al. 2018 , Cain et al. 2019 ). I used the reclassified data and program FRAGSTATS (version 4.1, University of Massachusetts, Amherst, Massachusetts, USA) to calculate 2 landscape metrics meaningful to deer in each county (i.e., contrast weighted edge density [CWED] and percentage of agric ulture [AG_LAND]). CWED represented the sum of the borders between cover types multiplied by a corresponding contrast - weight (i.e., weight = 1 for agriculture and forest cover types, and weight = 0 for all other cover types) divided by the area of the coun ty (km/km 2 ). AG_LAND denoted the percentage of the area within a county that was classified as agriculture cover. Before a deer on the landscape can be entered as a record deer, it must first be successfully harvested (Appendix D). The vulnerability of ant lered deer to harvest is influenced by hunting regulations, environmental conditions, and the behavior and density of deer and hunters (Roseberry and Klimstra 1974, Roseberry and Woolf 1998, Brown et al. 2000, Stewart et al. 2011). The Wisconsin Department of Natural Resources provided annual harvest data for antlered white - tailed deer. To evaluate the influence of harvest characteristics on the harvest of record deer, I calculated the total number of antlered deer taken in each county for the year (ANTLERE D) by adding together the number of antlered deer harvested across all the hunting seasons (e.g., archery, crossbow, gun) from 1981 2014. I evaluated collinearity in my covariates by assessing if the values were correlated ( ). 108 The Institutional Anima l Care and Use Committee at Michigan State University determined that the acquisition of harvest data from the Boone and Crockett Club, Wisconsin Buck and Bear Club, and Wisconsin Department of Natural Resources was exempt from protocols by the Animal Care and Use Committee. I followed guidelines outlined by the Boone and Crockett Club and Wisconsin Buck and Bear Club for securely processing and storing the data supplied for this research. Modeling Framework I setup the model to allow the probability of detecting a record deer to vary across space and over time. I used a zero - inflated Poisson N - mixture modeling framework to develop a hierarchical model to estimate the number of record deer harvested in each county (Appendix A). I estimate county - level det ection probabilities for record deer by treating the Boone and Crockett and WBBC records as independent double count data. I modeled the observation process as: , where was the total count fo r each sampling unit (i.e., county) during replicate and year . This binomial process rendered the identification of individuals irrelevant because the detections of record deer were random events (Royle and Dorazio 2008 ). N i,k was the parameter representing the estimated number of record deer after corr ecting for imperfect detection . The harvest of record deer in county i during a given year k was the outcome of a latent process, because it cannot be directly observed in the data (Royle et al. 2005, Kéry and Schmidt 2008). Given how the harvest of a record deer is a rare event, the data exhibit an excessively greater number of zero - counts than would be expected from a Poisson distribution. 109 Consequently, each county is more likely to have a value of zero than any other number. Therefore, I included an additional hierarchical level that characterized the suitability of a county for record deer that was able to deal with the excess zero - counts as part of a Bernoulli distri bution process (Kéry and Schaub 2012). This binary level of the hierarchical model was: . A uniform distribution was used as an uninformative prior for the proportion of suitable sites because parameter could only take on values between 0 and 1 (Link and Barker 2010). The site suitability followed a Bernoulli random distribution with probability . It was necessary to include this additional hierarchical level in the model, because the zeros coul d represent either a void of record deer harvested in the county (e.g., true zeros, unsuitable sites) or an omission of record deer that were harvested successfully but never reported (e.g., false zeros, reporting bias). Given that a county was suitable (i .e., when ), the estimated number of record deer harvested was modeled following a Poisson process: , where the is the true harvest for county i during year k . The i,k is an intensity param eter, which is conditioned on covariates, for county i and year k and modeled as: , where the is the intercept and is the slope that quantifies the log - linear relationship between lambda and covariates . For this analysis, I included the AG_LAND, CWED, and ANTLERED in each county as independent covariates to estimate the true harvest of record deer for each county and year. I chose these covariates because of their influences on antler sizes of male deer and o n the vulnerability of white - tailed deer to harvest. 110 I used a Bayesian framework to estimate parameters for the N - mixture model using software program JAGS (Plummer 2003) in program R (R Core Team 2019). I assumed vague prior distributions for estimated pa rameters and used R package jagsUI (version 1.5.0) to streamline the analysis. I evaluated posterior distributions, MCMC trace plots, and values for model convergence. RESULTS From 1981 to 2014, there were 1,350 records for white - tailed deer in the Boone and - tailed deer based on the minimum score set by the Boone and Crockett Club, I included 3,679 records reported to the WBBC. The N - mixture model performed well with values below 1.1 for all years and counties (Gelman and Rubin 1992). My model demonstrated successful convergence of all parameters in a model that used an uninformative prior for detection. There was adequate mixing and convergence of parameters with visual inspection of the trace plots (see Appendix B; Gelman and Rubin 1992, Link and Barker 2010). My results show spatial and temporal variability in the detection (Figure 4.1, Figure 4.2). The mean detection probability (Figure 4.1) as determined by the posterior distribution was highest for Dodge County (0.55, CI: 0.15 0.93) and Buffalo County (0.55, CI: 0.31 0.78), whereas the smallest mean detection probability was found in R ichland County, Wisconsin (0.30, CI: 0.06 0.72). Although the detection probabilities of each county vary across Wisconsin (Figure 4.1), there was no indication of spatial clustering; instead, the mean detection probability appeared to vary randomly across Wisconsin, with an average 111 detection of 0.43 (range: 0.30 0.55). Similarly, the detection probabilities of the counties displayed no notable patterns in values through time (Figure 4.2). The mean harvest of record deer was highest for Buffalo County (12. 50 deer/year, CI: 10.41 16.85 deer/year) and lowest for Calumet County (0.80 deer/year, CI: 0.24 3.21 deer/year). Unlike, the distribution of detection probability values, the mean estimates for the true harvest of record deer appeared spatially distribute d across Wisconsin (Figure 4.3). The number of record deer harvested across Wisconsin averaged 3.07 deer/year (range: 0.81 12.50). There was variation in the estimated number of record deer across the state (Figure 4.4). Dodge County, Wisconsin had the low est amount of variation on average with a standard deviation of 0.73, while Crawford County showed the greatest variation in the estimated number of record deer with a standard deviation of 5.36 (Figure 4.4). None of the covariates (i.e., AG_LAND, CWED, AN TLERED) that I included in the model had a significant influence on the number of record deer during every year of this study (Figure 4.5, Figure 4.6, Figure 4.7). Of the three covariates used to model the true harvest of record deer, only CWED was influen tial to the harvest of record deer for the majority of years. The posterior distribution for CWED showed that the influence CWED on the harvest of record deer was significant during 26 years (76.5%) of the study (Figure 4.6). Therefore, my results suggest CWED has a positive effect on the presence of record deer. 112 Figure 4.1 The mean detection probability for each county in Wisconsin from 1981 2014 estimated from the N - mixture model. 113 Figure 4.2 Detection probability of each county in Wisconsin from 1981 2014. Each line represents the detection probability for a single county through time. 114 Figure 4.3 Mean number of record deer harvested from 1981 2014 in each county of Wisconsin estimated using the N - mixture model. 115 Figure 4.4 The standard deviation in number of record deer harvested estimates for each county in Wisconsin from 1981 2014 estimated from the N - mixture model. 116 Figure 4.5 Posterior distribution of percent agriculture through time (1981 2014) in Wisconsin. 117 Figure 4.6 Posterior distribution of the Contrast Weighted Edge Density (CWED) metric to evaluate the influence of landscape configuration on the number of record deer harvested through time (1981 2014) in Wisconsin. 118 Figure 4.7 Posterior dist ribution of antlered harvest through time (1981 2014) in Wisconsin. 119 DISCUSSION My findings provide a methodological proof of concept using the N - mixture modeling framework (Royle 2004) to estimate the number of record deer harvested with 2 independent sour ces of data. My results showed that detection probabilities do not appear to follow obvious spatial or temporal patterns, and randomness of detection suggests that factors related to detection of record deer are not influential. By using the N - mixture mode ling framework to investigate record deer harvests, I was able to evaluate factors influencing trends in the harvest of record deer. I found that the number of record deer harvested was positively influenced by the CWED metric during most years covered in this analysis. My study is the first application of the N - mixture modeling framework to generate detection corrected estimates of harvest from voluntarily reported harvest data. Traditional applications of N - mixture models have used data from repeated cou nts of animals observed on the landscape to estimate population abundance (Ké ry et al. 2005, Keever et al. 2017, Christensen 2018). The detectability of individual animals is important to consider when estimating harvest of record deer because animals are detected (or reported) imperfectly. Researchers have long recognized that imperfect detection is pervasive in wildlife data, and limits our ability to draw conclusions from analyses (Williams 2001). While analysis does not get at the likelihood of an indiv idual hunter deciding to report, the analysis does quantify the probability that the record deer is detected by the records. Moreover, my findings that the detection probability values vary randomly across time and space provide evidence that biases associ ated with reporting rates of record deer are random. 120 Multiple independent observations are required for the N - mixture modeling approach to be successful (Royle 2004). For my analysis of record deer, the multiple observations were count data from the Boon e and Crockett Club and the Wisconsin Buck and Bear Club. One of the challenges remaining with the approach is that state - record programs do not exist for every state. Without this information, we cannot implement the N - mixture modeling framework to estima te harvests of record deer in states where these record programs do not exist. The difference in the number of records, with WBBC having more records than Boone and Crockett, was expected. WBBC actively seeks animals for their record book, and the Boone an d Crockett Club relies on hunters to initiate the process for submitting a record deer. The spatial distribution of record deer may serve as an indicator of high - quality deer habitat because large antler sizes are associated with deer in good condition ( Ditchkoff et al. 2001). The southwestern portion of the state appears to have higher harvests of record deer on average (Figure 4.3). These counties are within and adjacent to the Driftless Area ecoregion (Appendix C), suggesting that the ecological charac teristics that define Driftless Area influenced the harvest of record deer. N - mixture models can provide new information about how a covariate influences the response through time. To illustrate this point, my results showed that none of the covariates (i. e., AG_LAND, CWED, ANTLERED) included in the model influenced the number of record deer during every year of this study (Figure 4.5, Figure 4.6, Figure 4.7). Therefore, not only was I able to quantify the effects of each habitat covariate for the entire ti meframe, I could also evaluate how the effect has changed over the years. Habitat quality influences a variety of demographic and behavioral characteristics in cervids, including survival and recruitment (Ginnett and Young 2000, Hurley et al. 2014), body m ass and condition (Strickland et al. 2001, 121 Pettorelli et al. 2002), and timing of seasonal migration (Mysterud et al. 2017). Given the temporal nature of these dynamics, changes in the effect of habitat over time maybe of greater interpretative value than the average influence of habitat across the years of study. My finding that CWED was positively associated with the number of record deer aligns with current understanding of deer ecology, because a mixture of agriculture and forested lands (VerCauteren an d Hygnstrom 2011) is excellent habitat for an edge species like white - tailed deer (Alverson et al. 1988). High CWED values indicate areas with greater interspersion of the high - quality forage and cover types that are preferred by white - tailed deer (Roseber ry and Woolf 1998, Walter et al. 2009, Dechen Quinn et al. 2013). Moreover, my results align with previous research from Cain et al. (2019). They analyzed data of record deer harvested from 9 states of Midwestern United States and found that more record de er were harvested in counties with greater amounts of high - contrast edges. From this application of the N - mixture modeling framework, I found that the issues of reporting bias in analyses using harvests of record deer might not be as concerning as I expect ed. Instead, my results suggest that reporting rates vary randomly across space and do not follow any obvious temporal trends. Given our limited knowledge on the factors influencing the reporting rates of harvested deer, the N - mixture modeling framework is a good path forward for analyses using harvest data. The framework is flexible enough to allow the detection to vary (Royle 2004) and general enough that successful convergence of all parameters in a model that used an uninformative prior for detection wa s possible. Moreover, analyzing harvest data under the N - mixture modeling framework may provide new opportunities to understand the functional relationships between deer harvests and environmental covariates. This modeling framework might be successfully a pplied to other collections of harvest data that have information spanning 122 multiple decades and large areas to explore relationships between harvests of the species and ecological patterns that have changed through time. 123 APPENDICES 124 APPENDIX A: ZERO - INFLATED POISSON N - MIXTURE MODEL , , 125 APPENDIX B: TRACE PLOTS FOR PARAMETERS OF N - MIXTURE MODEL F igure 4.8 Trace plots of t he values generated from the N - mixture model in Chapter 4 with the value of sample from MCMC process (y - axis) versus the iteration number (x - axis). Each plot represents the sampling histories of a single model parameter. Th ese plots show that the chain s for each parameter are mixing well over the parameter space. 126 Figure 4.8 127 Figure 4.8 128 Figure 4.8 129 Figure 4.8 130 Figure 4.8 131 Figure 4.8 132 Figure 4.8 133 Figure 4.8 134 APPENDIX C: MAP OF ECOREGION CLASSIFICATIONS FOR COUNTIES IN WISCONSIN Figure 4.9 Map of ecoregion classification for each county in study area (Omernik 1987, Bailey 1995). I assigned an ecoregion to each county by determining which ecoregion covered the majority area within the county. 135 APPENDIX D: DIAGRAM OF PROCESS GENERATING RECORDS DATA F igure 4 . 10 Infographic to show the sequence of critical steps required for a male white - tailed deer that has record - Records of North American Big Game. Influential factors that may direct the outc ome (Yes/No) of each step are given in the gray boxes. 136 LITERATURE CITED 137 LITERATURE CITED Alverson, W. S., D. M. Waller, and S. L. So lheim. 1988. Forests too deer: e dge effects in northern Wisconsin. Conservation Biology 2:348 358. Bailey, R. G. 1983. Delineation of ecosystem regions. Environmental Management 7:365 373. Bailey, R. G. 1995. Description of the ecoregions of the United States. Second edition. Forest Service Miscellaneous Publication 1391, Washington, D.C., USA. Brown, T. L., D. J. Decker, S. J. Riley, J. W. Enck, T. B. Lauber, P. D. Curtis, and G. F. Mattfeld. 2000. The future of hunting as a mechanism to control white - tailed deer populations. Wildlife Society Bulletin 28:797 807. Cain, R. L., N. P. Snow, J. C. Crawford, D. M. W illiams, and W. F. Porter. 2019. Spatial distribution and landscape associations of large - antlered deer. The Journal of Wildlife Management 83:1762 1772. Christensen, S. A. 2018. Population impacts and disease patterns of hemorrhagic disease in white - taile d deer: understanding the role of an emergent disease. Dissertation. Michigan State University, East Lansing, USA. Dechen Quinn, A. C., D. M. Williams, and W. F. Porter. 2013. Landscape structure influences space use by white - tailed deer. Journal of Mammal ogy 94:398 407. Dénes, F. V., L. F. Silveira, and S. R. Beissinger. 2015. Estimating abundance of unmarked animal populations: accounting for imperfect detection and other sources of zero inflation. Methods in Ecology and Evolution 6:543 556. Dice, L. R. 1 941. Methods for estimating populations of mammals. Journal of Wildlife Management 5:398 407. deer ( Odocoileus virginianus ): e 55:616 625. Ficetola, G. F., A. Romano, S. Salvidio, and R. Sindaco. 2018. Optimizing monitoring schemes to detect trends in abundance over broad scales. Animal Conservation 21:221 231. Fiske, I. J., and R. B. Chandler. 2011. unmarked: an R package for fitting hierarchical models of wildlife occurrence and abundance. Journal of Statistical Software 43(10):1 23. Fry, J., G. Xian, S. Jin, J. Dewitz, C . Homer, L. Yang, C. Barnes, N. Herold, and J. Wickham. 2011. Completion of the 2006 national land cover database for the conterminous United States. Photogrammetric Engineering and Remote Sensing 77:858 864. 138 Gelman, A., and D. B. Rubin. 1992. Inference fr om iterative simulation using multiple sequences. Statistical science 7: 457 472. Ginnett, T. F. and E. L. Butch Young. 2000. Stochastic recruitment in white - tailed deer along an environmental gradient. Journal of Wildlife Management 64:713 720. Goddard, J . L. and C. A. Miller. 2009. Assessment of deer harvest reporting systems in the eastern United States. Human Dimensions of Wildlife 14:458 459. Guillera - Arroita, G. 2016. Modelling of species distributions, range dynamics and communities under imperfect d etection: advances, challenges and opportunities. Ecography 40:281 295. Hurley, M. A., M. Hebblewhite, J. M. Gaillard, S. Dray, K. A. Taylor, W. K. Smith, P. Zager, and C. Bonefant . 2014. Functional analysis of n ormalized difference vegetation i ndex curves reveals overwinter mule deer survival is driven by both spring and autumn phenology. Philosophical Transactions of the Royal Society B 369:20130196. Keever, A. C., C. P. McGowan, S. S. Ditchkoff, P. K. Acker, J. B. Grand, and C. H. Newbolt. 2017. Efficacy of N - mixture models for surveying and monitoring white - tailed deer populations. Mammal Research Institute 62:413 422. Kellner, K. F., and R. K. Swihart. 2014. Accounting for imperfect detection in ecology: a quantitative review. PLoS One, 9: e111436. Kéry , M., J. A. Royle, and H. Schmid. 2005. Modeling avian abundance from replicated counts using binomial mixture models. Ecological Applications 15:1450 1461. Kéry, M., and M. Schaub. 2012. Bayesian population analysis using WinBUGS: a hierarchical perspective. Academic Press, San Diego, California, USA. Kéry, M. and B. R. Schmidt. 2008. Imperfect detection and its consequences for monitoring for conservation. C ommunity Ecology 9:207 216. Kilpatrick, H. J., A. M. LaBonte, and J. S. Barclay. 2005. Factors affecting harvest - reporting rates for white - tailed deer. Wildlife Society Bulletin 33:974 980. Link, W. A., and R. J. Barker. 2010. Bayesian inference with ecolo gical applications. First edition. Academic Press, San Diego, California, USA. Link, W. A., and J. R. Sauer. 1997. Estimation of population trajectories from count data. Biometrics 53:488 497. Mackenzie, D. I., J. D. Nichols, N. Sutton, K. Kawanishi, and L . L. Bailey. 2005. Improving inferences in population studies of rare species that are detected imperfectly. Ecology 86:1101 1113. 139 McCullough, D. R. 1984. Lessons from the George Reserve, Michigan. Pages 211 242 in L. K. Halls, editor. White - tailed deer: e cology and management. Stackpole Books, Harrisburg, Pennsylvania, USA. Monteith, K. L., R. A. Long, V. C. Bleich, J. R. Heffelfinger, P. R. Krausman, and R. T. Bowyer. in trophy ungulates. Wildlife Monographs 183:1 28. Mysterud, A., B. K. Vike, E. L. Meisingset, and I. M. Rivrud. 2017. The role of landscape characteristics for forage maturation and nutritional benefits of migration in red deer. Ecology and Evolution 7:4 448 4455. Nesbit, W. H., and P. L. Wright. 2016. How to score North A merican big game: Boone and Crockett Club's official measurers manual. Fourth edition. Boone and Crockett Club, Missoula, Montana, USA. Omernik, J. M., 1987. Map supplement: ecoregions of the conterminous United States. Annals of the Association of American Geographers 77:118 125. Pettorelli, N., J - M. Gaillard, G. V. Laere, P. Duncan, P. Kjellander, O. Liberg, D. Delorme, and D. Maillard. 2002. Variations in adult body mass in roe deer: th e effects of population density at birth and of habitat quality. Proceedings of the Royal Society London B 269:747 753. Plummer, M. 2003. JAGS: a program for analysis of Bayesian graphical models using Gibbs sampling. Proceedings of the International Works hop on Distributed Statistical Computing 124: 125. R Core Team. 2019. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.r - project.org. Reneau, J., J. Spring, and E. L. Buckner. 2011. Records of North American big game. Thirteenth edition. Boone and Crockett Club, Missoula, Montana, USA. Roseberry, J. L., and W. D. Klimstra. 1974. Differential vulnerability during a controlled deer harvest. Journal of Wildlife Management 38:499 507. Ro seberry, J. L., and A. Wool f. 1991. A comparative evaluation of techniques for analyzing white - tailed deer harvest data. Wildlife Monographs 117:3 59. Roseberry, J. L., and A. Woolf. 1998. Habitat - population density relationships for white - tailed deer in I llinois. Wildlife Society Bulletin 26:252 258. Rosenberry, C. S., D. Diefenbach, and B. D. Wallingford. 2004. Reporting - rate variability and precision of white - tailed deer harvest estimates in Pennsylvania. The Journal of Wildlife Management 68:860 869. 140 Ro counts. Biometrics 60: 108 115. Royle, J. A., and R. M. Dorazio. 2008. Hierarchical modeling and inference in ecology: the analysis of data from populations, metapop ulations and communities. Academic Press , San Diego, California, USA . Royle, J. A., J. D. Nichols, and M. Kéry. 2005. Modelling occurrence and abundance of species when detection is imperfect. Oikos 110:353 359. Rupp, S. P., W. B. Ballard , and M. C. Wallace . 2000 . A nationwide evaluation of deer hunter harvest techniques. Wi ldlife Society Bulletin, 28: 570 578. Ryan, G. E., E. Nicholson, J. C. Eames, T. N. E. Gray, R. Loveridge, S. P. Mahood, P. Sum, and M. A. McCarthy. 2019. Simultaneous - count models to estimate abundance from counts of unmarked individuals with imperfect detection. Conservation Biology 33:697 708. Snow, N. P., Z. Zhang, A. O. Finley, B. A. Rudolph, W. F. Porter, D. M. Williams, and S. R. Winterstein. 2018. Regional - based mitigation t o reduce wildlife - vehicle collisions. Journal of Wildlife Management 82:756 765. Stewart, K. M., R. T. Bowyer, and P. J. Weisberg. 2011. Spatial use of landscapes. Pages 181 217 in D. G. Hewitt, editor. Biology and management of white - tailed deer. CRC Press, Boca Raton, Florida, USA. Strickland, B. K., S. Demarais, L. E. Castle, J. W. Lipe, W. H. Lunceford, H. A. Jacobson, D. Frels, and K. V. Miller. 2001. Effects of selective - harve st strategies on white - tailed deer antler size. Wildlife Society Bulletin 29:509 - 520. VerCauteren, K. C., and S. E. Hygnstrom. 2011. Managing white - tailed deer: Midwest North America. Pages 501 535 in D. G. Hewitt, editor. Biology and management of white - t ailed deer. CRC Press, Boca Raton, Florida, USA. Walter, W. D., K. C. VerCauteren, H. Campa III, W. R. Clark, J. W. Fischer, S. E. Hygnstrom, N. E. Matthews, C. K. Nielsen, E. M. Schauber, T. R. Van Deelen, and S. R. Winterstein. 2009. Regional assessment on influence of landscape configuration and connectivity on range size of white - tailed deer. Landscape Ecology 24:1405 1420. Williams, B. K. 2001. Uncertainty, learning, and the optimal management of wildlife. Environmental and Ecological Statistics 8:269 288. Williams, B. K., J. D. Nichols, and M. J. Conroy. 2002. Analysis and management of animal populations. Academic Press, San Diego, California, USA. Williams, D. M., A. C. Dechen Quinn, and W. F. Porter. 2012. Landscape effects on scales of movement by white - tailed deer in an agricultural forest matrix. Landscape Ecology 27:45 - 57. 141 Wisconsin Buck & Bear Club [W BBC]. 2019. History of the Wisconsin Buck & Bear Club. . Accessed 14 October 2019. Wisconsin Buck & Be ar Club [WBBC]. 2020. Locate a m easure r. . Accessed 6 June 2020. Wisconsin Department o f Natural Resources. 2012. Ecological landscapes of Wisconsin. Handbook 1805.1, Wisconsin Department of Natural Resources, Madison, USA. Wisconsin Department of Natural Resources. 2019. Deer hunting regulations. Wisconsin Department of Natural Resources PUB - WM - 431 2019, Madison, USA. 142 EPILOGUE In my dissertation, I sought to test hypothesized outcomes of antler point restrictions, evalua te spatially explicit trends in antler sizes of record deer, investigate the relationship between management and record deer harvests, and apply modeling framework to account for imperfect detection and assess trends in record deer harvests. By analyzing t rends in longitudinal harvest data and effects of environmental, management, and spatial contexts, this dissertation has shown how harvest outcomes relate to characteristics and regulations of the area. In this final section of my dissertation, I briefly r eview the findings and contributions from each chapter and suggest relevant research for future study. Chapter 1 sought to determine whether antler point restrictions brought about changes in the age structure of the male harvest, antlerless harvest, or number of hunters when implemented in Michigan , USA . Antler point restrictions are designed to protect the majority of yearling males from harvest (Hamilton et al. 1995). In areas where antler point restrictions were implemented the proportion of yearling harvests decreased and greater proportion of the male s harvest ed were from older age classes. However, antler point restrictions did not ap pear to cause a significant or lasting change in antlerless harvest or the nu mber of hunters. The findings from this chapter can be used to help managers and hunters alike set reasonable expectations for changes in harvest outcomes under antler point restrictions. Chapter 2 sought to evaluate the spatially explicit trends in antler sizes of record deer across the Midwest United States. The findings from this chapter underscore d the importance of considering the spatial context when analyzing trends across large geographic areas. Accounting for space is important because global traje ctories may not reflect trends happening at smaller 143 spatial scales. For example, declining trends in antler sizes of deer across all records in North America (Monteith et al. 2013) versus the increasing trends in antler sizes of record deer in the Midwest e rn United States . The findings from this chapter also demonstrated that the spatial clustering of similar antler sizes through time is related to the ecoregion context rather than a management context. Managers may find the results from this chapter usefu l when communicating with hunters or landowners about their expectations for the deer in the area. Moreover, because the differences in average antler sizes followed ecologically delineated boundaries, rather than political boundaries, future studies using record deer might consider including these measures of ecological patterns and processes. Chapter 3 sought to evaluate the degree to which management regulations influenced the harvest of record deer in the Midwest United States and review evidence for sp atial and temporal biases in reporting. Although some ecoregions seem to have inherently more record entries than others, management regulations do appear to have some influence on the harvest of record deer. In areas of highly suitable habitat for white - t ailed deer, management regulations, such as shorter season and limiting the harvest of antler deer, can provide enhancements to survival of antlered deer that may result in the additional harvest of record deer. Chapter 4 sought to incorporate detectabilit y in the modeling framework to make inferences about the harvest of record deer. In this chapter, I demonstrated the applicability of the N - mixture modeling framework (Royle 2004) to evaluate harvests of record deer in Wisconsin. The results suggest that r eporting rates vary randomly across space and do not follow any obvious temporal trends. Analyzing harvest data under the N - mixture modeling framework may also provide new opportunities to understand the functional relationships between deer harvests and e nvironmental covariates. 144 As interest in quality deer management and harvesting adult males with large antlers increases (Connelly et al. 2012, Harper et al. 2012), it is important for wildlife managers and hunters to understand how regulations and habitat can influence harvests of record deer. Further research is needed to determine the causes of imperfect detection, including information on existing barriers in reporting record harvests and factors affecting the probability that a hunter reports a record deer. Addressing the need for information related to imperfect detection will require studies in human dimensions and hunter behaviors. 145 LITERATURE CITED 146 LITERATURE CITED Connelly, J. W., J. H. Gammonley, and T. W. Keegan. 2012. Harvest management. Pages 202 231 in N. J. Silvy, editor. The wildlife techniques manual: management. Volume 2. John Hopkins University Press, Baltimore, Maryland, USA. Hamilton, J., W. M. Knox, and D. C. Guynn. 1995. Harvest strategies. Pages 47 57 in K. V. Miller and R. L. Marchinton, editors. Quality whitetails: the why and how of quality deer management. Stackpole Books, Mechanicsburg, Pennsylvania, USA. Harper, C. A., C. E. Shaw, J. M. Fly, and J. T. Beaver. 2012. Attitudes and motivations of Tennessee deer hunters toward quality deer management. Wildlife Society Bulletin 36:277 285. Monteith, K. L., R. A. Long, V. C. Bleich, J. R. Heffelfinger, P. R. Krausman, and R. T. Bowyer. 2013. Effects of trophy ungulates. Wildlife Monographs 183:1 28. counts. Biometrics 60: 108 115.