A MULTIDIMENSIONAL PERSPECTIVE ON WILDLIFE CONSERVATION AND MANAGEMENT: PEOPLE, POPULATIONS, AND HABITAT By Talesha Janill Dokes 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 A MULTIDIMENSIONAL PERSPECTIVE ON WILDLIFE CONSERVATION AND MANAGEMENT: PEOPLE, POPULATIONS, AND HABITAT By Talesha Janill Dokes As natural resource managers the decisions we make today can affect our tomorrow. Therefore, we have to have a diverse skillset to address a variety of issues to susta in our natural resources for future generations. As a manager there are three primary responsibilities 1) people, 2) habitat, and 3) populations. My dissertation is a combination of research topics that captures all three responsibilities of a natural reso urce manager. In Chapter 1, I assessed current natural resource students surveys to determine what motivated them to choose natural resources as a career. I analyzed their family backgrounds and the students current interest to see what motivating factors influenced their career decision. I found that students whose families participated in consumptive activities were most influential for the student in pursuing a natural resource career and that these students tended to choose natural resources as a major between the 11 th and 12 th grade. In Chapter 2, I investigated factors that affect trapping success of the American marten ( Martes americanus ) . I wanted to determine what weather and ecological factors had an influence on trapper success. I found that more old) and harvest of marten per year varied substantially. Temperature and precipitation during the months of October, November, and December was found to have a correlation with marten harvest. North American h ardwoods forest was the only existing vegetation type from LANDFIRE found significant. I also found that density of roads had an effect on trapper harvest due to accessibility into certain forested areas. Lastly, in Chapter 3 I evaluated songbird species r esponses to timber harvest. The extent of my dissertation research encapsulates the three responsibilities of a natural resource manager (people, habitat, and populations) and why it is important for managers to become trained in a diverse set of skills an d have a certain level of understanding of all aspects of our natural resources. Results from this research assist colleges and managers in their decision making processes of building the next generation of managers and sustaining our natural resources. iv ACKNOWLEDGEMENTS I would like to thank God who has blessed me in ways I could not even imagine. I would like to thank my graduate advisor, Dr. Gary Roloff for all of his time, support, and guidance, during my time at MSU. What you have instilled in me and the lessons that I have learned will continue to help me throughout my life and career, and for that I am forever grateful. I would like to acknowledge and thank my graduate committee: Drs. Kelly Millenbah, Ashton Shortridge, and David Williams for their time, energy, and effort in the successful completion of my degree. I would like to specially thank the Fisheries and Wildlife Department (AFWEL colleagues, fellow graduate students, staff and faculty) I really enjo yed getting to know you all. I am forever grateful to my MANRRS colleagues for all of your love, support and guidance, you were my family away from home and I will miss you all dearly. Lastly, I would like to appreciate my family and loved ones, particular ly my beautiful mother (Jocelyn Dokes), fiancé (Terrance Dumas), sister (Tina Dokes) and brothers (Dennis and Phillip Dokes). Thank you for believing in me, pushing me to do great things, and being tough on me when I needed it the most. I am blessed to hav e you all in my life, WE DID IT! There are numerous amounts of people that I could list that have helped me in thi s journey and I just cannot thank you all enough . You know who you are and I am grateful for your prayers, friendship, and encouragement . This dissertation would not have been possible without funding provided by the G raduate S chool and teaching assistantship program at Michigan State University . I would like to thank Michigan DNR, Inland Wildlife Department of Sault Ste. Marie Tribe of Chippewa Indians, Safari Club International Michigan Involvement Committee , John Humphreys, Clint Otto, Kelly Millenbah and all my co - authors for their research contributions. v PREFACE Each chapter within this dissertation was drafted as a stand - alone manuscript for publication in a peer - reviewed journal. Chapter 1 is currently in press with The Wildlife Society Bulletin . Chapter 2 and Chapter 3 have not been submitted for publication. D ue to copyright restrictions, Chapter 1 is not included in this dissertation. For chapter 1, I provide a brief summary of the article. Although I am listed as the sole author and use the pronoun I throughout this dissertation, each chapter was a collaborat ive effort and all associated manuscripts and chapters will include one or more co - authors when submitted for peer - review. vi TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ....................... viii LIST OF FIGURES ................................ ................................ ................................ ...................... xii INTRODUCTION ................................ ................................ ................................ .......................... 1 LITERATURE CITED ................................ ................................ ................................ ................... 5 CHAPTER 1 ................................ ................................ ................................ ................................ ... 8 NATURAL RESOURCE UNDERGRADUATE STUDENTS IN THE NEW MILLENNIUM .. 8 CHAPTER 2 ................................ ................................ ................................ ................................ . 10 FACTORS AFFECTING TR APPING SUCCESS FOR A MERICAN MARTEN ...................... 10 Abstract ................................ ................................ ................................ ................................ ..... 10 Introduction ................................ ................................ ................................ ............................... 11 Study Area ................................ ................................ ................................ ................................ . 14 Methods ................................ ................................ ................................ ................................ ..... 14 Results ................................ ................................ ................................ ................................ ....... 18 Landfire Factors ................................ ................................ ................................ .................... 18 Weather Factors ................................ ................................ ................................ .................... 19 Road Factor ................................ ................................ ................................ ........................... 20 Pelt Prices ................................ ................................ ................................ ............................. 20 Model Fitting ................................ ................................ ................................ ......................... 20 Discussion ................................ ................................ ................................ ................................ . 21 APPENDIX ................................ ................................ ................................ ................................ ... 26 LI TERATURE CITED ................................ ................................ ................................ ................. 50 CHAPTER 3 ................................ ................................ ................................ ................................ . 59 MULTISCALE RELATIONSHIPS OF SONGBIRD OCCUPANCY IN A MANAGED FOREST LANDSCAPE ................................ ................................ ................................ ............... 59 Abstract ................................ ................................ ................................ ................................ ..... 59 Introduction ................................ ................................ ................................ ............................... 60 Methods ................................ ................................ ................................ ................................ ..... 62 Study Area ................................ ................................ ................................ .............................. 62 Site Selection ................................ ................................ ................................ .......................... 63 Bird Sampling ................................ ................................ ................................ ........................ 64 Habitat Covariates ................................ ................................ ................................ ................ 64 Model Generation and Analysis ................................ ................................ ............................ 65 Results ................................ ................................ ................................ ................................ ....... 67 Discussion ................................ ................................ ................................ ................................ . 69 APPENDIX ................................ ................................ ................................ ................................ ... 73 LITERATURE CITED ................................ ................................ ................................ ................. 92 vii CONCLUSION ................................ ................................ ................................ ............................. 97 LITERATURE CITED ................................ ................................ ................................ ............... 100 viii LIST OF TABLES Table 2.1. Deviance Information Criteria (DIC) and Watanabe - Akaike Information Criteria (WAIC) for models predicting likelihood of harvesting a marten in a township section of the Upper Peninsula of Michigan, 2001 - 2018. One model included a spatial random effect term (Spatial) and one did not (No Spatial). k = effective Table 2.2. Parameter estimates (standard deviation (SD)) and 95% credible intervals for fixed effects used to predict the likelihood of trapping a marten in a township section in the Upper Peninsula of Michigan, 2001 Table 2. S1. Vegetation disturbance (VDISTURB) category as a majority cover type in sections where trappers harvested marten, and comparable background data points used for m odeling. Sections represent counts, and percent represents percentage of the total (calculated from total number of registered marten (n=3,412) and background locatio Table 2. S2. The majority canopy bulk density (CBD) w ithin a section where trappers harvested marten, and comparable background data points used for modeling. Sections represent counts, and percent represents percentage of the total (calculated from total number of registered marten (n=3,412) and background locations (n Table 2. S3 . The majority existing vegetation cover class (EVC) in sections where trappers harvested marten, and comparable background data points used for modeling. Sections represent counts, and percent re presents percentage of the total (calculated from total number of registered marten (n=3,412) and background locations 38 Table 2. S4 . Majority existing vegetation height (EVH) in sections where trappers harvested marten, and comparable background data points used for modeling. Sections represent counts, and percent represents percentage of the total (calculated from total number of registered marten (n=3,412) and background locations ..41 Table 2. S5 . Majority existing vegetation type (EVT) in sections where trappers harvested marten, and comparable background data points used for modeling. Sections represent counts, and percent represents percentage of the total (calculated fro m total number of registered marten (n=3,412) and background locations (n=24, ix Table 2. S6 . Average (standard error) and range of monthly maximum temperatures and monthly maximum precipitation for October, November, and D ecember based on PRISM data, and average (standard error) and range of road densities (based on All Roads layer; Center for Shard Solutions and Technology Partnerships 2014) for the Upper Peninsula, Michigan from 2001 - Table 3.1 . Number o f plots sampled by retention canopy cover (in 50m radius plot) and .79 Table 3.2 . Average (SE) and range (Min, Max) of cover (%) of landscape variables surrounding bird survey points a t 3 radii in northwest Lower Peninsula of Michigan, 201 Table 3.3 . Candidate models for ovenbird occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - dif ference in AIC value from top model, AIC wt = weight of evidence, and k = number of model Table 3.4 . Candidate models for red - eyed vireo occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - 2011. AIC = Akaike Information Criteria, wt = weight of evidence, and k = number of model par Table 3.5 . Candidate models for Rose - breasted grosbeak occupancy at points in the northwes tern Lower Peninsula of Michigan, 2010 - 2011. AIC = Akaike wt = 83 Table 3.6 . Candidate models for indigo bunting occupan cy at points in the northwestern Lower Peninsula of Michigan, 2010 - 2011. AIC = Akaike Information Criteria, wt = weight of evidence, and Table 3.7 . Candidate models for American redstart occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - 2011. AIC = Akaike Information Criteria, wt = weight of evidence, and k = number of mode l par Table 3.8 . Candidate models for chestnut - sided warbler occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - 2011. AIC = Akaike x wt = weight of evidence, and k = numbe Table 3.9 . Candidate models for eastern towhee occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - 2011. AIC = Akaike Information Criteria, alue from top model, AIC wt = weight of evidence, and k = number of mode Table 3.10 . Candidate models for Nashville warbler occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - 2011. AIC = Akaike Information Criteria, wt = weight of evidence, and ..88 Table 3. S1 . Candidate univariate landscape - level models for ovenbirds surveyed in 2010 - 2011, northw est Lower Peninsula of Michigan. AIC = Akaike Information AIC wt = weight of Table 3. S2 . Candidate univariate landscape - level models for r ed - eyed vireos surveyed in 2010 - 2011, northwest Lower Peninsula of Michigan. AIC = Akaike Information AIC wt = weight of Table 3. S3 . Candi date univariate landscape - level models for chestnut - sided warblers surveyed in 2010 - 2011, northwest Lower Peninsula of Michigan. AIC = Akaike AIC wt = weight of evidence, and k = numbe r of Table 3. S4 . Candidate univariate landscape - level models for indigo buntings surveyed in 2010 - 2011, northwest Lower Peninsula of Michigan. AIC = Akaike Information AIC wt = weight of evidence, and k = number of Table 3. S5 . Candidate univariate landscape - level models for Nashville warblers surveyed in 2010 - 2011, northwest Lower Peninsula of Michigan. AIC = Akaike Information = difference in AIC value from top model, AIC wt = weight of evidence, and k = number of Table 3. S6 . Candidate univariate landscape - level models for American redstarts surveyed in 2010 - 2011, northwest Lower Peninsula of Mi chigan. AIC = Akaike Information AIC wt = weight of evidence, and k = number of xi Table 3. S7 . Candidate univariate landscape - level models for rose - breasted grosbeaks sur veyed in 2010 - 2011, northwest Lower Peninsula of Michigan. AIC = Akaike AIC wt = Tables 3. S8 . Candidate univariate lands cape - level models for eastern towhees surveyed in 2010 - 2011, northwest Lower Peninsula of Michigan. AIC = Akaike Information AIC wt = weight of evidence, and k = number of xii LIST OF FIGURES Figure 2.1 . The Upper Peninsula of Michigan, USA showing marten harvest data (recorded to the nearest 2.6 km 2 ) collected by the Michigan Department of Natural Resources from trappers between 20 01 - Figure 2. S1 . Age at capture of marten relative to the global capture mean (horizontal line). Data collected by the Michigan Department of Natural Resources from registered marten in the Upper Peninsula, Michigan, USA between 2001 - 2018 (n=3,412)..28 Figure 2. S2 . Sex of captured marten relative to the global capture mean (horizontal line). Data collected by the Michigan Department of Natural Resources from registered marten in the Upper Peninsula, Michigan, USA betwee n 2001 - 2018 (n=3,412)..29 Figure 2. S3 . Year of marten harvest relative to the global mean (horizontal line). Data collected by the Michigan Department of Natural Resources from registered marten trappers in the Upper Peninsula, Michigan, USA captured bet ween 2001 - 30 Figure 2. S4 . Crown Bulk Density (CBD) and marten harvest probability relative to the global mean (horizontal line). Data collected by the Michigan Department of Natural Resources from registered mart en in the Upper Peninsula, Michigan, USA 31 Figure 2. S5 . Influence of prior year harvest from the same section on marten harvest probability. Data collected by the Michigan Departme nt of Natural Resources from registered marten in the Upper Peninsula, Michigan, USA between 2001 - 32 1 INTROD U C TION Decisions made by wildlife managers today have long - lasting effects. Wildlife management in the 21 st century is highly complex (Ascher 2001; Cilliers et al.2013), requiring diverse skills for effective movement of conservation and sustainability in a positive direction. Broadly, wildlife managers have three primary responsibilities 1) people, 2) habitat, and 3) animal populations. In North America the public plays a critical, active role in wildlife conservation by providing funding (through taxation and license sales; Organ et al. 2012), interacti ng with public agencies that serve as wildlife trustees (Organ et al. 2012), and by voting (Kilpatrick and Walter 1997). Habitat is the foundation of wildlife population performance, and managers frequently manipulate habitats to affect populations (Morri son et al. 1992; Messmer 2009). The ultimate indicator of successful wildlife conservation and sustainable management is population performance, best expressed as long - term population growth rate (Lindenmayer 2000). Managers coordinate the actions of peopl e, and manipulate habitats and populations to affect long - term population growth rate to meet some objective. For overabundant wildlife causing property damage, the objective is likely to reduce populations and mitigate damage (e.g., Conover 2001). For rar e species, the objective is likely to increase distribution, numbers, and population growth rate (e.g., Wydeven et al. 2009). Collectively, people, habitat, and animal populations form the - My diss ertation is a combination of research topics that include components of the - have with their environments. In Chapter 1, I assessed what motivated current natural reso urce students to choose natural resources as a career, recognizing that younger generations in the United States may not relate to the North American Model of Wildlife Conservation. Younger 2 generations in the United States are increasingly urbanized (Manfr edo et al. 2003), often at the expense of utilitarian connections to wildlife and under - appreciation for some tools used to manage animal populations like hunting and trapping (Manfredo et al. 2003). However, younger generations have a close non - utilitaria n connection to wildlife and the environment (Manfredo et al. 2003), offering a substantial conservation opportunity. This places organizations relying on hunting, trapping, and fishing license sales to implement wildlife conservation (e.g., state resource agencies) in a difficult position. On one hand, funding for the organization is tied to an increasingly outdated interest in wildlife (for example) so implementation of programs and activities must maintain or attempt to increase those interests. Converse ly, those programs may alienate younger generations, potentially missing a critical opportunity to engage the broader public in conservation. Ultimately, wildlife management organizations recognize that employees must represent diverse and value public int erests to remain relevant in the 21 st century. In my first chapter, I analyzed family backgrounds and current interests of student enrolled in natural resource programs in the United States to understand motivating factors that influenced their apparent ca reer decision. The premise was to lay a foundation for understanding the future employee pool responsible for implementing wildlife conservation, guide student recruiting into the profession, and offer suggestions to improve college natural resource course offerings. Managers use harvest regulations to achieve habitat or animal population objectives and to influence public participation and interest (e.g., Riley et al. 2002; Lauber et al. 2012). Factors affecting participation and effort in wildlife harves t by the public are multi - faceted and complex in space, time, and circumstance (Riley et al. 2002; Enck 2006). For example, weather conditions (Obbard et al. 1999), state of the economy (Obbard et al. 1999), and social or cultural demographics (Miller and Vaske 2003) affect hunting participation and effort. Given that harvest 3 regulations are a key element of many wildlife conservation programs, increased understanding of factors that motivate people to participate and be successful benefit management organi zations. In Chapter 2, I investigated factors that effected trapping success of American marten ( Martes americanus ) in Michigan. I sought to determine what factors could potentially be manipulated by wildlife managers to affect harvest success. I evaluated factors directly controlled by managers (e.g., distance from maintained roads), those related to socio - economic forces beyond the management organization (e.g., pelt prices), and factors that were uncontrollable (e.g., weather). As such, this chapter cont - legged effort) influenced marten populations under varying habitat conditions. Wildlife conservation programs often include some form of habitat management. In some instances, wildlife conservation can be included in practices commonly used for resource extraction like timber harvest. In forested regions of North America, managers commonly use timber harvest purposefully to provide wildlife habitat (e.g., Linden and Roloff 2013). In other instances, timber extraction is the primary management objective but wildlife considerations are included (Blinn and Kilgore 2001; Demarais et al. 2017). One way to include wildlife in timber harves t objectives is through retention forestry, where managers retain elements of the pre - harvest forest to increase structural complexity (Fedrowitz et al. 2014; Mori and Kitagawa 2014). Retention forestry is particularly relevant in silvicultural systems lik e clearcutting, where managers remove all merchantable trees. Clearcutting is a common practice used on aspen ( Populus spp.) forests in Michigan, and foresters are required to retain unharvested trees to provide wildlife habitat (Bielecki 2012). Retention of these trees comes at a cost through lost timber revenues, potentially increased safety hazards for equipment operators, and potential loss 4 of forest regeneration. Hence, knowing that retention forestry is having a positive effect on wildlife populations is a critical information need. Otto and Roloff (2012) found that retention forestry in aspen clearcuts of Michigan had minimal effect on bird occupancy probability, and they surmised that landscape context was an important consideration. In Chapter 3, I evaluated how songbird occupancy related to structural retention in aspen clearcuts using a hierarchical model that included patch - and landscape - factors, with the goal of better understanding how landscape context affected the function of retained structu res as bird habitats. Although this chapter focuses on habitat management and how it affected a population parameter (i.e., occupancy), the results inform decisions made by managers and policy - makers (i.e., people). My dissertation research encapsulated th e three responsibilities of a wildlife manager (people, habitat, and populations), highlight the importance of multi - dimensional training and experiences for managers. I also used sound sampling designs and a suite of modeling approaches to generate scient ific evidence, consistent with efforts to infuse science into natural resources decision - making (Mills and Clark 2001). Results from my research offer insights into how people decide to embark on wildlife careers, how people respond to socio - economic and e nvironmental factors to manipulate wildlife populations, and how habitat management decisions by people can influence wildlife populations. 5 LITERATURE CITED 6 LITERATURE CITED Ascher, W. 2001. Coping with complexity and organizational interests in natural resource management. Ecosystems 4:742 - 757. Bielecki, J., J. Ferris, K. Kintigh, M. Moss, D. Kuhr, S. Mackinnon, S. Throop, L. Visser, M. Walters. 2012. Within - stand retentio n guidance. Michigan Department of Natural Resources, Forest Resources Division . 241 - 250 pp. Blinn, C. R. and M. A. Kilgore. 2001. Riparian management practices, a summary of state guidelines. J ournal of Forest ry 99 (8): 11 - 17. Chilliers, P., H. C. Biggs, S. Blignaut, A. G. Choles, JH. S Hofmeyer, G. P. W. Jewitt, and D. J. Roux. 2013. Complexity, modeling, and natural resource management. Ecology and Society 18( 3 ):1 . Conover, M. R. 2001. Effects of hunting & trapping on wil dlife damage. Wildlife Society Bulletin 29(2) : 521 - 352. Demarais, S., J. P. Verschuyl, G. J. Roloff, D. A. Miller, and T. B. Wigley. 2017. Tamm review: T errestrial vertebrate biodiversity and intensive forest management in the U.S. Forest Ecology and Manag ement 385 : 308 - 330. Enck, J. W., D. J. Decker, S. J. Riley, L. H. Carpenter, and W. F. Siemer. 2006. Integrating ecological and human dimensions in adaptive management of wildlife - related impacts. Wildlife Society Bulletin 34(3):689 - 705. Fedrowitz, K., J. Koricheva, S. C. Baker, D. B. Lindenmayer, B. Palik, R. Rosenvald, W. Beese, J. F. Franklin, J. Kouki, E. Macdonald, C. Messier, A. Sverdrup - Thygeson and L. Gustafsson. Can retention forestry help conserve biodiversity? A meta - analysis. J ournal of Applied Ecology 51 : 1669 - 1679. Kilpatrick, H. J. and W. D. Walter. 1997. Urban deer management: A community vote. Wildlife Society Bulletin 25 ( 2 ) :388 - 391. Lauber, T. B., D. J. Decker, K. M. Leong, L. C. Chase, and T. M. Schusler. 2012. Stakeholder engagement in wildlife management. Pages 139 - 156 in D. J. Decker, S. J. Riley, and W. F. Siemer. 2 nd edition. Human dimensions of wildlife management. The John Hopkins University Press, Baltimore, Maryland. Leopold, A. 1987. Game m anagement. University of Wisconsin Press , Madison. 520pp. Linden, D. W. and G. J. Roloff. Retained structures and bird communities in clearcut forests of the Pacific Northwest, USA. Forest Ecology and Management 310 : 1045 - 1056. 7 Lindenmayer, D. B., C. R. Margules, and D. B. Botkin. 2000. In dicators of biodiversity for ecologically sustainable forest management. Conservation Biology 14 ( 4 ): 941 - 950. Manfredo, M. J., T. L. Teel, A. D. Bright. 2003. Why are public values towards wildlife changing? Human Dimensions of Wildlife 8:287 - 306. Messme r, T. A. Human - wildlife conflicts: emerging challenges and opportunities. Human - Wildlife Conflicts 3(1):10 - 17. Miller, C. A. and J. J. Vaske. 2003. Individual and situational influences on declining hunter efforts in Illinois. Human Dimensions of Wildli fe 8:263 - 276. Mills, T. J. and R. N. Clark. 2001. Roles of research scientists in natural resource decision - making. Forest Ecology and Management 153(1 - 3) : 189 - 198. Mori, A. S. and R. Kitagawa. 2014. Retention forestry as a major paradigm for safeguardi ng forest biodiversity in productive landscapes: A global meta - analysis . Biological Conservation 175 : 65 - 73. Morrison, M. L., B. G. Marcot and R. W. Mannan. 1992. Wildlife - habitat relationships: Concepts and applications. University of Wisconsin Press, M adison . 343 pp . Obbard, M. E., J. G. Jones, R. Newman, A. Booth, A. J. Satterthwaite, and G. Linscombe. 1999. Furbearer harvests in North America. Pages 1007 1034 in M. E. Obbard, J. G. Jones, R. Newman, A. Booth, A. J. Satterthwaite, and G. Linscombe Wildlife furbearer management and conservation in North America Section VII: Technology, Techniques, and Harvests Organ, J.F., V. Geist, S.P. Mahoney, S. Williams, P.R. Krausman, G.R. Batcheller, T.A. Decker, R. Carmichael, P. Nanjappa, R. Regan, R.A. Medellin, R. Cantu, R.E. McCabe, S. Craven, G.M. Vecellio, and D.J. Decker. 2012. The North American Model of wildlife conservation. The Wildlife Society Technical Review 12 - 04. The Wildlife Society, Bethesda, Maryland, USA. 47 pp. Otto, C. R. V., and Roloff, G. J. 2012. Songbird response to green - tree retention prescriptions in clearcut forests. For est Ecol ogy and Management 284:241 - 250. Riley, S. J., D. J. Decker, L. H. Carpenter, J. F. Organ, W. F. Siemer, G. F. Mattfeld, an d G. Parsons. 2002. The essence of wildlife management. Wildlife Society Bulletin 30 ( 2 ) :585 - 593. Wydeven, A. P., J. E. Wiedenhoeft, R. N. Schultz, R. P. Thiel, R. L. Jurewicz, B. E. Kohn, and T. R. Van Deelen. 2009. History, population growth, and manag ement of wolves in Wisconsin in Wydeven A. P., T. R. Van Deelen, E. J. Heske. (eds) Recovery of grey wolves in the Great Lakes region of the United States. Springer, New York, NY Chapter 6 87 - 19 pp . 8 CHAPTER 1 NATURAL RESOURCE UNDERGRADUATE STUDENTS IN THE NEW MILLENNIUM Historically, undergraduate college students enrolled in natural resources programs came from rural backgrounds and regularly participated in fishing, hunting, and trapping (i.e., consumptive activities). Student demographics shifted c onsiderably over the last 30 years, with more natural resources students coming from urban backgrounds with lower levels of engagement in consumptive activities. Some stakeholders and employers are concerned that misalignment between student participation in consumptive and nonconsumptive wildlife - related activities and priorities of natural resource management authorities might result in contradicting views on consumptive activities. We sought to understand the background, participation in wildlife - related activities, and career decision - making process of undergraduate college students currently enrolled in natural resources programs in the early 21 st century. We conducted an online survey of students enrolled in member universities and colleges of the Nati onal Association of University Fisheries and Wildlife Programs (NAUFWP). We examined the ways in which demographic information (including personal and family characteristics) and participation in outdoor activities shaped student; i ) decisions to pursue na tural resources training in college, and ii ) career choices. We received 1,376 undergraduate respondents (570 males, 806 females) representing universities and colleges in 29 US states. Responding students were primarily Caucasian with the majority between 18 and 22 years old. Most identified that they regularly (>11 hours/month) spent time outdoors, and just over half acknowledged participating in hunting and fishing. Participation in hunting, fishing, and farming were lower among students identifying as f emale compared to those identifying as male. We also found that family participation in hunting and fishing was most influential on both student involvement in consumptive activities 9 and their decisions to pursue natural resources careers. Students with fa mily participation in consumptive activities made natural resources career decisions in high school compared to students whose families did not participate in consumptive activities who waited until early college. Our study indicates that 50% of college st udents enrolled in natural resource programs do not participate in consumptive activities. Our study also highlights the importance of family participation in consumptive activities in determining school and career outcomes, with implications for student r ecruitment and retention practices for natural resource academic programs and employers in the early 21 st century. This work is currently in press and will appear in the Wildlife Society Bulletin. 10 CHAPTER 2 F ACTORS AFFECT ING TRAPPING SUCCESS FOR AMERICA N MARTEN A bstract Wildlife managers often rely on harvest data (e.g. , number of animals harvested, trapper effort) collected from trappers to make population - level inferences and establish regulations for furbearers. I used data from marten harvested across the upper peninsula of Michi gan from 2001 - 2018 to determine if individual demographic s ( sex ratio and age) , weather, land cover types, pelt prices, or roads played a role in successful capture of marten. I used a Bay esian geostatistical approach that accounted for spatial and temporal autocorrelation . My sample included 3,882 marten s (2008 males and 825 females) with known year of harvest , gender , and harvest location to the nearest 2.6 km 2 . I used PRISM and LANDFIRE to compile weather and land cover information , respectively. Relative to the global mean the harvest sample included significantly more - old) , fewer females, and harves t varied substantially by year (differed from global mean 11 of 18 years) . Pelt price and distance to maintained roads were not significant predictors, but I found that temperature and precipitation in October, November, and December correlated with marten harvest. Warmer temperatures in October and November, yet colder temperatures in December associated with higher marten harvests. More precipitation in October and December, but drier Novembers associated with higher marten harvests. C rown bulk density (C BD) in the 2.6 km 2 containing the harvest location generally had a positive effect on harvest ; no other LANDFIRE variables were strongly influential. Harvest of marten from a section the prior year was a strong positive predictor of marten harvest. Pelt pr ice and distance to maintained roads were not important. My results indicate that weather conditions leading up to and during the trapping season for marten 11 are important determinants of harvest success, offering some short - term forecasting potential. My r esults also indicate that the coarse resolution of harvest location limits the ability to model harvest success from readily available land cover data. I ntroduction Harvest data from game species are important sources of ecological and evolutionary information to wildlife managers (Mysterud et al. 2001; Yoccoz et al. 2002; Bonenfant et al 2003; Carranza et al. 2004; Martinez et al. 2005). M anagers often use harvest data to assess population dynamics (Getz and Haight 1989; Solberg et al., 1998; Nielsen et al., 2004; Allen et al. 2018). Anglers, hunters, and trappers tend to select wildlife for particular phenotypic characteristics (Pelletier et al. 2012; Allen et al. 2018), such as large fish (JØrgensen et al. 2007), or male animals during hunting (Coltman et al. 2003; Martinez et al. 2005; Mysterud et al. 2006; Garel et al. 2007 ; Mysterud 2011; Pelletier et al. 2012 ). Preferential selection of sexually dimorphic mammals is common among ungulate hunters, but less common for c ertain furbearer species (Allen et al. 2018). Selective h arvest s affect wildlife population size s , age structures , and sex ratios (Markgren 1969; Crete, Taylor & Jordan 1981; Bubenik 1987; Ginsberg and Milner - Gulland 1994; Saether et al. 2003), and reduce s life expectancy for selected cohorts (Langvatn and Loison 1999). For example, h unting can increase mortality and reduce life expectancy of older animals , making them rare in the population (Bubenik 1987; Courtois 1989), with ramifications for reproductio n (Proaktor et al. 2007). Hunting can also play a significant role in skewing sex ratios in favor of females for some wildlife groups ( e.g., ungulates; Markgren 1969; Crête et al. 1981; Bubenik 1987), potentially indirectly affecting female space use , surv ival , and reproductive success . For example, in the absence of sexually mature or dominant males, 12 females may increase home range size during mating season potentially exposing them to higher risk (Claveau and Courtois 1992). Additionally, m ale s for some species (e.g., bears ( Ursus spp. ) ) may be more susceptible to harvest because of relatively larger home range sizes and risky behaviors compared to females (Mysterud et al. 2006; Bischof et al. 2009; Mysterud 2011). Ballard et al. (1991) concluded that lack of males reduces reproduction in moose (Thomason 1991, Stephenson et al. 1995), thereby decreasing fecundity, recruitment, and ultimately population growth. Low numbers of mature males can lead to sociobiological problems (Bubenik 1987; Crichton 1992), such as breakdown of dominance hierarchies (Laurian 2000). Insights into factors affecting selection of specific animals should inform harvest regulations (Mysterud et al. 2011; Allen et al. 2018). Little information exists on preferential selection of furbearers or environmental factors that aff ect composition of furbearer harvests. Activity of trappers and hunters who pursue furbearers (collectively called furharvesters) and composition of animals harvested is a complex expression of intrinsic (e.g., physical, experience, or equipment capabiliti es of individual trappers or hunters) and extrinsic (e.g., market prices, furbearer populations, weather) factors, complicating management. The primary reason to pursue furbearers is to sell or use the fur or other body parts (Elsken - Lacy et al. 1999; Hill er et al. 2011; Kapfer and Potts, 2012). E ffort expended by furharvesters directly links to market prices (Gehrt et al. 2002; Poole 2003; Hil l er et al. 2011; Elsken - Lacy et al. 1999), but maybe mediated by weather and landscape factors. In northern regions , for example, snow depth, temperature, and restricted road access hinders trapper success (Strickland 1994; Hiller et al. 2011; Kapfer and Potts 2012). Understanding spatio - temporal variation in furbearer harvests as a function of weather and landscape fa ctors can assist managers in spatially allocating trapping effort or predicting harvest amounts. 13 American marten ( Martes americanus ) are high - valued furbearer s trapped across North America. In some areas (e.g., northern Michigan and Wisconsin) , marten popu lations are highly sensitive to habitat and harvest pressures (Hiller et al. 2011) . Marten are habitat specialist s and indicator species (Buskirk and McDonald 1989; Watt et al. 1996; Fecske et al. 2002) , sensitive to amounts of mature forest cover (Buskirk and Powell 1994; Fecske et al., 2002 ; Cushman et al. 2011 ) , anthropogenic activity (Davis 1983; Soukkala 1983; Hodgman et al. 1997; Robitaille and Aubry 2000), and forest cover type (Cushman et al., 2011) . American marten are lean, with small body mass resulting in home range s three times the size predicted for terrestrial carnivores based on allometrics (Lindstedt et al. 1986). Additionally, males travel farther than females ( Chapin et al. 1998; Silet 2017 ), making males more vulnerable to harvest (Souk kala 1983; Archibald and Jessup 1984; Fortin and Cantin 1994) . Little i nformation on a ge specific mortality rates among sexes exists for marten (Hodgman et al. 1994 ; Bull and Heather 2001 ). Trappers catch marten with body - grip trap s (e.g. , conibears), foot hold trap s, and cage traps (Frawley 2017), resulting in opportunities for selective harvest pressures ( primarily with cage traps). F actors influenc ing trapper efforts (Daigle et al. 1998) and success remain poorly understood by wildlife managers and should be further investigated (Ahlers 2016). I studied spatial and temporal dimensions of harvested marten across the Upper Peninsula (UP) , Michigan, USA. My goal was to understand factors affecting locations of marten harvests, assuming that these locations represent a decision by trappers on where to focus effort. I sought to understand demographics of harvested marten and weather , landscape , and pelt price as factors affecting harvests. I predicted that marten harvest would be higher in areas with less forest disturbance, in mature forest types (particularly those dominated by conifers), in areas with high canopy cover and canopy bulk density , and closer to maintained roads . I modeled these 14 relationships for an 18 - year harvest data set. My results improve understanding of marten harvest dynamics with implications for managing this valuable furbearer. S tudy A rea My study area encompassed the Upper Peninsula (UP) of Michigan ( approximately 42,600 km 2 ; Fig. 1 ) , which is approximately 85% forested and <5% agricultural lands (United States Department of Agriculture, Natural Resources Conservation Service 2006), with 40% in public ownership (state and federal) (Skalski et al. 2011). Major forest types include aspen ( Populus spp.), hemlock ( Tsuga canadens is ) , maple ( Acer spp.), spruce ( Picea spp.), birch ( Betula spp.) and fir ( Abies balsamea; Sommers 1977 , Hiller et al. 2011 ). Approximately 308,000 people lived in the UP in 2009 (U.S. Census Bureau 2009). Annual precipitation in the western portion general ly ranges from 660 to 940 mm , and from 762 to 914 mm in the eastern portion (United States Department of Agriculture, Natural Resources Conservation Service 2006) , and substantially varies in space during the trapping season (Hiller et al. 2011) from lake effect snows . Annual temperatures in the west vary from ~3 C to ~7 C and in the east from ~4 C to ~6 C . The western portion has greater topographic relief, with elevations ranging from 184m to 606m , whereas the eastern portion has low relief, is poorly drained, and include s peat bogs and extensive swamps (Albert 1995; Hiller et al. 2011) . M ethods Marten occur throughout the Upper Peninsula, with the majority of trapping records coming from northerly latitudes (Fig. 1). Legal trapping methods in clude foothold, body - gripping and cage traps (Hiller et al. 2011; Skalski et al. 2011) . O nce a marten is harvested trappers are required by state and Tribal regulations to have the fur sealed (Skalski 2011). Upon registration additional data are collected including harvest date, sex (estimated at the time of registration), 15 and location (to the nearest 2.6 km 2 ). A tooth is also extracted for age estimation using cementum annuli (Poole et al. 1994). I obtained marten harvest records from the MDNR from 2001 - 201 8 . Marten reduce movement s during extremely low temperatures and high precipitation ( Carroll 2007; Hiller et al. 2011) , t herefore, I calculated monthly maximum temperature s and precipitation f or October, November , and December from data provided by th e PRISM Climate Group (PRISM 2018). These months correspond to the early part of the Tribal trapping season (runs from October to March), and the non - Tribal season that is restricted to December. The MDNR started consistently documenting Tribal marten harvests in 200 9 , and since then 1 8 % (n=523) were registered on Tribal licenses . I predicted that low monthly maximum temperatures and high precipitation would limit trap - line preparation , scouting , and correspond to reduced marten harvest. I used rasters downloaded from the LANDFIRE Program (LANDFIRE 2010) to evaluate vegetation variables that potentially contributed to marten captures . These variable s included v egetation disturbance (VDISTURB) , vegetation canopy cover (EVC) , canopy bulk density (CBD) , vegetation height (EVH) , and vegetation type (EVT) modeled and mapped at 30m resolution. V egetation disturbance is a composite (over last 10 years) of disturbance type, severity on vegetation, and time since the disturbance occurred (LANDFIRE 2019). Vegeta tion canopy cover portrays the vertically projected cover of live canopy, accounting f or trees, shrubs, and herbaceous cover (LANDFIRE 2019) . Canopy bulk density describes the density of available canopy, defined as canopy fuel for fire per canopy volume ( kg/m 3 ; LANDFIRE 2019). Vegetation height portrays average height of dominant vegetation, and vegetation type describes complexes of plant communities (LANDFIRE 2019). Given the temporal extent of the marten 16 data, I used five LANDFIRE products (2001, 2008, 2010, 2012, and 2014), and matched the LANDFIRE data closest to year of marten harvest I used the zonal statistics tool in ArcGIS spatial analyst to identify the majority value for each LANDFIRE layer within the 2.6km 2 area containing the marten harvest. R oads negatively affect marten as direct (hit by vehicles) and indirect (access by trappers) sources of mortality ( Soukkala 1983; Hiller et al. 2011; Cervinka et al., 2015; Ruette et al., 2015). However, marten use certain road types for traveling and hunting (Robitaille and Aubry 2000 ; Silet 2017 ). Michigan maintains an all roads geographic framework layer that categorizes roads (Center for Shared Solutions and Technology Partnerships 2014) . Using this spatial framework , I extracted t he maintained roads , which describ e roads that are permanent and accessible year round ( maintained ) . I buffered maintained road s by 8 m and summed the road areas within each 2.6km 2 containing a harvested marten. A buffer was important to account for roads that occurred on township section boundaries ; the buffer allowed a single road to influence multiple section s. I performed analysis using the R language version 3.4.1 for statistical computing ( R Core Team 2 017 ). To model the probability of marten harvest, I used a Bayesian geostatistical approach that accounted for spatial and temporal autocorrelation. Harvests we re modeled as a Bernoulli point process such that Z(s ,t ) signifies the occurrence (1) or absence (0) of a marten harvest in a township section s ( s 17276 ) , 2018), with a probability of marten harvest given as . Generally, this relationship is: 17 where is the intercept, X s ,t represents a vector of fixed covariates and random effects with corresponding coefficients , and is the space - time structured effect that quantifies spatial and temporal correlation between sections (s) and across years (t) (Humph reys et al. 2017 ; Humphreys et al. 2019 ). The spatial components of were modeled using - regression for time following the SPDE approach (Lindgren2011). Because the model generates a separate realization of the spatial field for each time step, it was necessary to reduce model temporal resolution by aggregating time into two year groupings or bins (i.e., time knots) (Blangiardo and Cameletti 2015). Lowering the temporal res olution enabled us to model the entire marten harvest record (2001 - 2018) with minimal effect to estimated spatial - temporal correlation. I scaled and centered predictor variables, and matched monthly t emperature and precipitation data to the corresponding harvest year (200 1 - 2018) . I checked for correlation among predictor variables using the colldiag function from the perturb package in R, which calculates condition indices and variance decomposition proportions (Hendrick s 2019). Condition indices >30 paire d with variance decomposition proportions >0.50 suggest collinearity and warrant further exploration (Belsley et al. 1980). I evaluated two model structures , with and without spatial and temporal autocorrelation, using Deviance Information Criteria (DIC) a nd Watanabe - Akaike Information Criteria (WAIC). We deemed model parameters significant if 95% credible intervals did not overlap zero. To help ensure identifiability and avoid over - parameterization, we enforced a zero mean constraint on all random effects ( Schrodle and Held 2011, Bivand et al., 2015). 18 R esults T rappers registered 4,4 4 9 marten from 2001 2018 ; 3, 412 included year and location ( ~190 /year) . Marten h arvest was highest (n=353) in 2018, and lowest in 2001 (n=63). On average, <1.5 marten were harvested from the same township section within a year, indicating that captures were spread across space . Most harvested marten were young - of - the - year ; 0 .5 ( 39 %; n =1 , 3 37 ) and 1.5 years of age (1 4 %; n= 487 ). The oldest marten were 12.5 years of age (0. 09 %; n=3) . Gender was recorded for 2, 833 marten; males dominated harvest ( 71 %; n=2 ,008 ) followed by females ( 2 9 %; n=8 25 ). L andfire F actors When summarized as majority types within 2.6 km 2 areas, I found that few categories dominated most LANDIFRE variables. I found that vegetation disturbance was not a majority characteristic of sections in the Upper Peninsula (99% of background) , or where marten were harvested ( 99% of harvests; Table S 1 ). Low numbers of marten were captured in sections dominated by disturbance ( 0. 6 % of captures), and almost all of these (95%) were captured > 1 year after the disturbance (Table S 1 ). M ost (70%) marten were harvested in sections domi nated by low CBD (0.01kg/m 3 ), and some (11%) in sections dominated by non - forested areas ( that presumably contained forest patches; Table S 2 ). Overall, CBD in the Upper Peninsula is low, with 34% of the sections dominated by non - forested areas (0.00 CBD) , and 49% dominated by sparse overstory canopies (0.01 CBD; Table S 2 ). A t coarse resolution (i.e., 2.6 km 2 ), CBD is low across this landscape , likely reflecting dominance of deciduous and mixed deciduous - coniferous forests . I found that tree cover ranging from 5 0 to 9 0% was a dominate characteristic of sections ( 70 % of background) in the Upper Peninsula , and most marten ( 93 % ) were harvested in these areas ( Table S 3 ) . Low numbers of marten ( <9 %) were harvested in section s dominated by shrub 19 cover an d herb aceous cover , presumably from forested or wetland patches embedded in this matrix (Table S 3 ). Trappers caught most (98%) marten in sections dominated by taller forests (10 - 25 m), and this forest stature was most ubiquitous (78% of section) in the Upp er Peninsula (Table S 4 ). My height data, when combined with cover data, suggest that marten from sections dominated by shrubs tended to come from taller shrub communities, like alder ( Alnus spp.) or scrub - shrub wetlands >10m tall. The LANDFIRE layer contained 60 cover types in the Upper Peninsula, dominated by variations of hardwood forests ( at least 63 % of sections) and boreal w hite s pruce - f ir (30% of sections; Table S 5 ). We note that the No Data class dominated 18% of sections in the Upper Peninsula (Table S 5 ); lakes Superior, Michigan, or Huron dominated these sections. The m ajority of harvested marten came from section s dominated by hardwood forests (68%; Table S 5 ). Surprisingly, l ow numbers of marten (13%) came from sections dominated by conifer - hardwood swamp forest s , but sections dominated by this type were uncommon (17% background) . Other sections where marten were trapped more frequently (i.e., >1% of the sections) were dominated by boreal a cidic p eatland systems (collectively 4% of sections; Table S 5 ). At coarse resolution, our results highlight the positive relationship between hardwood forest communities and marten harvests in the Upper Peninsula . W eather F actors Average maximum monthly temperat ures for October , - S 6). On average, sections where trappers harvested marten were warmer in October and November, but colder in December than average Upper Peninsula conditions (Table S 6). Average maximum precipitation was highest in October (102mm), and declined into November (65mm) and December (60mm) across the Upper 20 Peninsula (Table S 6). Trappers harvested marten in sections that were considerably drier in October when compared to the Upper Peninsula, but precipitation amounts were comparable to the broader landscape in November and December (Table S 6). R oad F actor On average, 108 m 2 /section of maintained roads occurred in sections whe re marten were harvested, compared to 91m 2 of maintained roads across the Upper Peninsula (Table S 6 ). P elt P rices I compiled annual marten pelt prices from a Michigan Department of Natural Resources furbearer harvest report ( Frawley 2018 ) as a proxy for trapper effort . I assumed that effort would positively relate to pelt price. From 2001 - 2018, average pelt price was $46 USD for marten, with a range of $29 USD (2018) to $72 USD (2016), showing high inter - annual variability ( sd = $13 USD). M odel F itting I did not find any predictor variables that were collinear (condition indices <10), so I used the complete suite of variables for modeling. The spatial and non - spatial models ranked higher than the non - spatial model (Table 1 ), so I used the spatial model for inference. For random effects, I found that marten age, sex, trapping year, CBD , and whether trappers harvested marten from the same section the previous year significantly influenced our model. Relative to the global mean harves t probability, harvest probability was higher for 4 .5 years old and lower for 6 .5 years old (Fig. S1 ). On average, h arvest probability was lower for females (Fig. S2), and significantly fluctuated by year (Fig. S3). Harvest probability was below average in areas with low CBD, except one higher class of CBD (0.22 kg/m3) also corresponded to low harvest 21 probability (Fig. S4). Trappers appeared to revisit the same sections where the re were successful ly harvested marten the prior year , as this was a strong predictor of harvest (Fig. S 5). I did not find strong significant effect s of vegetation disturbance, existing vegetation cover or height , or vegetation type as portrayed by majority LANDFIRE classes within a section on marten harvest. I found that only weather fixed effect s were significant ( Table 2 ). Warmer October ( = 1. 03 , SD = 0.07 ) and November ( = 4. 89, SD = 0.13 ) temperatures positively affected marten harvest , and colder December ( = - 2. 37, SD = 0.08 ) temperatures positively af fect ed harvest ( Table 2 ). The most influential month for temperature was November, followed by December (Table 2). I found that drier Octobers ( = - 0. 72, SD = 0.06 ) and Decembers ( = - 0. 27, SD = 0.05 ) positively affected marten harvest probability , but w etter Novembers were better for harvest ( = 0.6 7, SD = 0.05 ). Density of maintained roads, pelt price, and prior harvest of marten at any time prior did not affect harvest probability. (Table 2 ). Discussion Predicting harvest success for game animals is difficult because so many factors (some stochastic) affect the outcome (e.g., Johnson et al. 1997; Strickland et al. 2013 ). Regulations often established months or years in advance of harvest seasons are critical tools for wildlife population management, thus any ability to predict harvest outcomes is beneficial. I analyzed land cover, weather, roads, and pelt prices that potentially affect ed marten harvest in the Upper P eninsula of Michigan from 2001 - 2018. I found that weather was the most important correlate of harvest at the township section level , after accounting for age, sex, year, CBD, and whether a marten was harvest from the same section the prior year . Marten are more vulnerable to capture during years with warm er temperatures ( Octobers and Novembers ) , but with colder December 22 temperatures. Drier Octobers and Decembers, and w etter Novembers also positively corresponded to marten harvests. Based on the magnitude of standardized model coefficients, November and December temperatures were most influential. Changes in these annual weather patterns resulted in a strong year effect, underscoring the complexities in trying to make long - term predic tions of harvest outcomes for marten. Furthermore, I found that LANDFIRE covariates summarized at the section level (i.e., majority value within a section) were generally poor predictors of marten harvests, with the exception of CBD . For predicting marten harvest outcomes, my results suggest that temperature and precipitation for the months leading up the harvest season hold promise as short - term predictors. Generally, I found that harvest probability of fe male marten was lower than males ( see also Strick land 1994) . I also found that harvest were skewed to young marten (see also Harris et al. 1997; Ruette et al 2015) , consistent with a harvest strategy that focuses on young animals and preserves breeding stock (Strickland 1994). Differences in trapping vul nerability among age and sex cohorts in Martes often relates to differences in home range size s , distances traveled, and behaviors (Krohn et al. 1994 , Powell 1994 ) , with males having large home ranges and traveling further than females (Powell 1994) , and j uveniles traveling substantial distances during dispersal (Strickland et al. 1994) . Contrary to Hiller et al. (2011), I found a strong weather effect on where and when trappers captured marten. My observed effects varied annually, likely representing an in tegrated response to weather, trapping effort as mediated by weather and fur prices, and annual harvest regulations (Hiller et al. 2011). Differences in how researchers summarized weather data between studies likely explains some of the contradicting resul ts. The Hiller et al. (2011) study summarized weather during the trapping season, which ranged from 11 - 15 days in December, 23 whereas I evaluated a longer temporal extent that offered the model a broader range of values. Additionally, Hiller et al. (2011) used data from proximate weather stations, whereas I relied on interpolated values that corresponded to township sections . Together, the results suggest that weather during the trapping season is not overly influential o n marten harvest, but that longer - term, monthly weather patterns matter. Longer - term weather may affect trap line scouting (e.g., patterning marten from November snows) or access to trapping areas (e.g., snow accumulation from November to December). I hypothesized that marten would be trapped in areas with less forest disturbance, mature forest types (dominated by conifers), areas that were high in canopy cover and bulk density , and in areas that were closer to roads . At coarse resolution (i.e., majorit y type within 2.6 km 2 ) I found few remotely sensed vegetation variables that significantly explained marten harvests. However, several patterns emerged from summaries of capture counts within township sections that provide insights into the Upper Peninsula landscape. Most marten were caught in sections dominated by minimal disturbance (also see Ruggiero et al. 1994; Zielinski, Kucera 1995) , tree (also see Bushirk et al. 1989, Cushman et al. 2011) , and with low er crown bulk densities ( likely refle cting the dominance of hardwood forest types). Others demonstrated that higher crown bulk density, characteristic of conifers, positively affected marten use ( Williams et al. 2007, Cushman et al. 2011, Bridger et al. 2017). Marten tend to have a strong eco logical relationship with certain prey species that are associated with conifer forests (Tevis 1956; Clough 1987; Nordyke and Buskirk 1991; Pearson and Ruggiero 2001), but hardwoods and mixed hardwoods and conifer forests also support marten (Williams et a l. 2007). I suspect that conifers are an important component of marten habitat in the Upper Peninsula, particularly in 24 lowland or along riparian areas, but the coarse resolution of our land cover data failed to identify these potentially important areas. Lastly, I hypothesized that roads would positively affect marten harvest s. I failed to find a road effect on harvest probability, but caution that my analysis was limited to maintained roads. I suspect that seasonal roads play a greater role in trapper acc ess and ultimately harvest success (Soukkala 1983; Robitaille and Aubry 2000; Hiller et al. 2011), but accurate maps of seasonal roads were unavailable. The strong weather effect on marten harvest probability likely relates to access, with drier Decembers potentially allowing vehicular access to a broader landscape. I conducted this research using a large data set accumulated from martens registered by trapper s , presumably representing marten populations across the Upper Peninsula of Michigan . However, har vest data can contain inaccuracies and potential biases (Pelletier et al. 2012). I assume d that location biases in the data were minimal because trappers reported at a coarse resolution (2.6 km 2 ). I also acknowledge potential shortcomings in how I used LAN DFIRE data . Given the section - level resolution of harvest data, I aligned resolution s by summarizing the majority LANDFIRE class in each township section. Therefore, this approach reduced my ability to identify important, less common features or fine scale heterogeneity that may a ffect trapping success (e.g., conifer dominated riparian areas) . For example, one township section may have 51% hardwood and 49% lowland conifer ; i f a marten was trapped i n the lowland conifers my analysis would miss t his relationship . My results indicate that w ildlife managers have minimal control over distribution of marten harvests at the township section level. Instead, harvest probability seems most related to temperature and precipitation leading up to and during the harvest season. Marten t rappers understand habitat , often identify ing areas occupied by marten prior to the trapping season , using 25 this information to their advantage (Wiebe et al. 2013) , if weather conditions cooperate . My data also indicate that t ra ppers return to areas where they harvested marten the prior year , partially explaining the preponderance of younger animals in the harvest population . 26 A PPENDIX 27 FIGURE & TABLE CAPTIONS Figure 2. 1. The Upper Peninsula of Michigan, USA showing marten harvest data (recorded to the nearest 2.6 km 2 ) collected by the Michigan Department of Natural Resources from trappers between 2001 - 2018 (n=3,412). 28 Figure 2. S1. A ge at capture of marten relative to th e global capture mean (horizontal line). Data collected by the Michigan Department of Natural Resources from registered marten in the Upper Peninsula, Michigan, USA between 2001 - 2018 (n=3, 412 ). 29 Figure 2. S2. Sex of captured marten relative to the glo bal capture mean (horizontal line). Data collected by the Michigan Department of Natural Resources from registered marten in the Upper Peninsula, Michigan, USA between 2001 - 2018 (n=3, 412 ). 30 Figure 2. S3. Y ear of marten harvest relative to the global mean (horizontal line). Data collected by the Michigan Department of Natural Resources from registered marten trappers in the Upper Peninsula, Michigan, USA captured between 2001 - 2018 (n=3, 412 ). 31 Figure 2. S4 . Crown Bulk Density (CBD) and marten har vest probability relative to the global mean (horizontal line). Data collected by the Michigan Department of Natural Resources from registered marten in the Upper Peninsula, Michigan, USA (n=3,412). To convert CBD c lass code to kg/m3, multiply by 0.01. 32 Figure 2. S5. Influence of p rior year harvest from the same section on marten harvest probability . Data collected by the Michigan Department of Natural Resources from registered marten in the Upper Peninsula, Michigan, USA between 2001 - 2018 (n=3, 412 ). 33 Table 2. 1. Deviance Information Criteria (DIC) and Watanabe - Akaike Information Criteria (WAIC) for models predicting likelihood of harvesting a marten in a township section of the Upper Peninsula of Michigan, 2001 - 2018. One model included a spatial random effect term (Spatial) and one did not (No Spatial). k = effective number of parameters. Model DIC k DIC WAIC k WAIC Spatial No Spatial 5,411.71 5,416.92 95.16 92.33 5,397.06 5,403.20 77.58 75.71 34 Table 2. 2. Parameter estimates (standard deviation (SD)) and 95% credible intervals for fixed effects used to predict the likelihood of trapping a marten in a township section in the Upper Peninsula of Michigan, 2001 2018. Variable a Estimate (SD) 95% Credible Intervals Temperatures October November December Precipitation October November December Maintained Roads Pelt Price Prior Harvest 1.03 (0.07) 4.89 (0.13) - 2.37 (0.08) - 0.72 (0.06) 0.67 (0.05) - 0.27 (0.05) 0.04 (0.03) - 0.04 (0.03) - 0.01 (0.01) 0.90 1.16 4.64 5.15 - 2.52 - 2.23 - 0.85 - 0.61 0.58 0.77 - 0.37 - 0.18 - 0.03 0.10 - 0.10 0.03 - 0.02 0.01 a Temperature and precipitation represent mean maximum monthly values, derived from PRISM. Roads represent those not maintained throughout the year (seasonal) and those maintained (e.g., plowed). Prior harvest indicates whether a trapper harvested a marten in the same section at any time previously. 35 Table 2. S1. Vegetation disturbance (VDISTURB) category as a majority cover type in sections where trappers harvested marten, and comparable background data points used for modeling. Sections represent counts, an d percent represents percentage of the total (calculated from total number of registered marten (n=3,412) and background locations (n=24,570). LANDFIRE Vegetation Disturbance Marten Background Code Type Sections Percent Sections Percent 0 No disturbance 3,361 99 24,595 99 111 Fire, Low, 1 yr ago 0 0 4 <1 112 Fire, Low, 2 - 5 yrs ago 1 <1 18 <1 113 Fire, Low, 6 - 10 yrs ago 0 0 10 <1 121 Fire, Medium, 1 yr ago 1 <1 2 <1 122 Fire, Medium, 2 - 5 yrs ago 1 <1 16 <1 123 Fire, Medium, 6 - 10 yrs ago 0 0 7 <1 132 Fire, High, 2 - 5 yrs ago 3 <1 11 0 133 Fire, High, 6 - 10 yrs ago 2 <1 7 0 212 Mechanical Add, Low, 2 - 5 yrs ago 0 0 8 <1 213 Mechanical Add, Low, 6 - 10 yrs ago 3 <1 19 <1 311 Mechanical Remove, Low, 1 yr ago 0 0 2 <1 312 Mechanical Remove, Low, 2 - 5 yrs ago 8 <1 24 <1 313 Mechanical Remove, Low, 6 - 10 yrs ago 0 0 4 <1 36 Table 2. S1 ( c ont d ) LANDFIRE Vegetation Disturbance Marten Background Code Type Sections Percent Sections Percent 322 Mechanical Remove, Medium, 2 - 5 yrs ago 1 <1 11 <1 323 Mechanical Removal, Medium, 6 - 10 yrs ago 0 0 5 0 332 Mechanical Remove, High, 2 - 5 yrs ago 1 <1 7 <1 37 Table 2. S2. The majority canopy bulk density (CBD) within a section where trappers harvested marten, and comparable background data points used for modeling. Sections represent counts, and percent represents percentage of the total (calculated from total number of re gistered marten (n=3,412) and background locations (n=24,570). LANDFIRE Canopy Bulk Density Marten Background Code Kg/m 3 Sections Percent Sections Percent 0 0.00 379 11 8,431 34 1 0.01 2,375 70 12,216 49 6 0.06 2 <1 0 0 8 0.08 5 <1 13 <1 9 0.09 10 <1 78 <1 11 0.11 60 2 541 2 16 0.16 306 9 2,021 9 22 0.22 231 7 1,450 5 30 0.30 14 <1 0 0 38 Table 2. S3. The majority existing vegetation cover class (EVC) in sections where trappers harvested marten, and comparable background data points used for modeling. Sections represent counts, and percent represents percentage of the total (calculated from total number of registered marten (n=3,412) and background locations (n=24,570). LANDFIRE Vegetation Cover Marten Background Code Type Sections Percent Sections Percent 0 None 7 <1 4643 19 11 Open water 38 1 430 2 16 Developed - upland herbaceous 6 <1 18 <1 17 Developed - upland shrubland 0 0 7 <1 22 Developed - low intensity 1 <1 9 <1 23 Developed - medium intensity 0 0 18 <1 25 Developed roads 3 <1 94 <1 31 Barren 0 0 72 <1 32 Quarries - strip mines - gravel pits 0 0 47 <1 64 NASS - row crop 0 0 27 <1 65 NASS - close grown crop 1 <1 75 <1 66 Nass - pasture and hayland 1 <1 33 <1 81 Pasture/hay 2 <1 12 <1 82 Cultivated crops 16 <1 171 <1 39 Table 2. S3 . ( c ont d ) LANDFIRE Vegetation Cover Marten Background Code Type Sections Percent Sections Percent 95 Herbaceous wetlands 19 <1 253 1 100 Sparse vegetation canopy 0 0 9 <1 101 Tree cover >= 10 and < 20% 0 0 12 <1 102 Tree cover >=20 and <30% 2 <1 37 <1 103 Tree cover >=30 and <40% 29 <1 168 <1 104 Tree cover >=40 and <50% 5 <1 79 <1 105 Tree cover >=50 and <60% 119 4 1087 5 106 Tree cover >=60 and <70% 489 14 4164 17 107 Tree cover >=70 and <80% 2139 63 10805 42 108 Tree cover >=80 and <90% 389 12 1583 6 111 Shrub cover >=10 and <20% 55 2 359 1 112 Shrub cover >=20 and <30% 4 <1 4 <1 113 Shrub cover >=30 and <40% 3 <1 34 <1 114 Shrub cover >= 40 and <50% 0 0 29 <1 115 Shrub cover >=50 and <60% 2 <1 50 <1 121 Herb cover >=10 and <20% 17 <1 91 <1 122 Herb cover >=20 and <30% 8 <1 96 <1 40 Table 2. S3 . ( c ont d ) LANDFIRE Vegetation Cover Marten Background Code Type Sections Percent Sections Percent 123 Herb cover >=30 and <40% 0 0 4 <1 124 Herb cover >=40 and <50% 9 <1 96 <1 125 Herb cover >=50 and <60% 18 <1 134 <1 41 Table 2. S4 . Majority existing vegetation height (EVH) in sections where trappers harvested marten, and comparable background data points used for modeling. Sections represent counts, and percent represents percentage of the total (calculated from total number of registered marten (n=3,412) and background locations (n=24 ,570). LANDFIRE Vegetation Height Marten Background Code Type Sections Percent Sections Percent 0 No Data 6 <1 4626 19 11 Open water 11 <1 237 <1 16 Developed - upland herbaceous 2 <1 0 <1 25 Developed - roads 0 0 63 <1 31 Barren 0 0 72 <1 32 Quarries - strip mines - gravel pits 0 0 36 <1 64 NASS - row crop 0 0 9 0 65 NASS - close grown crop 0 0 14 <1 81 Pasture/hay 0 0 4 <1 82 Cultivated crops 1 <1 36 <1 95 Herbaceous wetlands 6 <1 67 <1 101 Herb height 0 to 0.05 meters 5 <1 32 <1 102 Herb height 0.5 to 1.0 meters 18 <1 118 <1 103 Herb height > 1.0 meter 0 0 2 <1 42 Table 2. S4 . ( c ont d ) LANDFIRE Vegetation Height Marten Background Code Type Sections Percent Sections Percent 104 Shrub height 0 to 0.5 meters 12 <1 108 <1 107 Shrub height >3.0 meters 11 <1 77 <1 108 Forest height 0 to 5 meters 4 <1 12 <1 110 Forest height 10 to 25 meters 3306 98 19237 78 43 Table 2. S5. Majority existing vegetation type (EVT) in sections where trappers harvested marten, and comparable background data points used for modeling. Sections represent counts, and percent represents percentage of the total (calculated from total nu mber of regist ered marten (n=3,412) and background locations (n=24,570). LANDFIRE Vegetation Type Marten Background Code Type Sections Percent Sections Percent 0 No data 6 <1 4626 18 11 Open water 12 <1 189 <1 16 Developed - upland herbaceous 2 <1 9 <1 25 Developed - roads 0 0 27 0 31 Barren 0 0 36 <1 32 Quarries - strip mines - gravel pits 0 0 27 <1 64 Agricultural - row crop 0 0 9 0 65 Agricultural - grown crop 0 0 5 0 66 Agricultural - fallow/idle cropland 1 <1 24 <1 44 Table 2. S5. ( c ont d ) LANDFIRE Vegetation Type Marten Background Code Type Sections Percent Sections Percent 81 Agricultural - pasture and hay 2 <1 4 <1 82 Agricultural - cultivated crops and irrigated agriculture 6 <1 118 <1 95 Caribbean 9 0 144 1 2191 Recently logged - herb and grass cover 17 <1 101 <1 2195 Recently burned - herb and grass cover 5 <1 10 <1 2198 No data 0 0 5 0 2301 Boreal aspen - birch forest 1 <1 27 <1 2302 Laurentian - Acadian North American hardwoods forest 860 25 5650 22 2344 Boreal jack pine - black spruce forest 21 1 180 1 2362 Laurentian - Acadian northern pine forest 4 <1 27 <1 2365 Boreal white spruce - fir forest 36 1 469 30 45 Table 2. S5 . (c ont d ) LANDFIRE Vegetation Type Marten Background Code Type Sections Percent Sections Percent 2366 Laurentian - Acadian pine - hemlock forest 45 1 153 <1 2407 Laurentian pine barrens 2 <1 18 <1 2444 Eastern boreal floodplain 3 <1 9 <1 2466 Great Lakes wooded dune and swale 0 0 9 <1 2475 Laurentian - Acadian floodplain systems 11 <1 36 <1 2477 Boreal acidic peatland systems 72 2 386 2 2481 Laurentian - Acadian alkaline conifer - hardwood swamp 169 5 2200 8 2492 Great Lakes coastal marsh systems 0 0 8 <1 2534 Managed tree plantation - northern and central hardwood 21 <1 182 <1 3191 Recently logged - herb and grass cover 7 <1 31 <1 3195 Recently burned - herb and grass cover 1 <1 13 0 46 Table 2. S5 . (c ont d) LANDFIRE Vegetation Type Marten Background Code Type Sections Percent Sections Percent 3240 Laurentian - Acadian hardwood forest 28 1 61 <1 3241 Laurentian - Acadian pine - hemlock - hardwood forest 4 <1 0 <1 3242 Laurentian - oak barrens 0 0 9 0 3245 Boreal white spruce - fir - hardwood forest 0 0 27 <1 3272 Eastern boreal floodplain shrubland 1 <1 0 0 3277 Laurentian - Acadian floodplain shrubland 1 <1 0 0 3279 Boreal acidic peatland shrubland 13 <1 99 <1 3281 Laurentian - Acadian alkaline conifer - hardwood swamp shrubland 3 <1 63 <1 3285 Laurentian - Acadian shrub wetlands 14 <1 144 1 3292 Open water 36 1 216 1 3294 Barren 0 0 36 <1 47 Table 2. S5 . (c ont d) LANDFIRE Vegetation Type Marten Background Code Type Sections Percent Sections Percent 3494 Laurentian - Acadian forested wetlands 2 <1 0 0 3295 Quarries - strip mines - gravel pits 0 0 20 <1 3299 Developed - roads 1 <1 36 <1 3301 Boreal aspen - birch forest 3 <1 36 <1 3302 North American hardwoods forest 1438 43 5969 24 3344 Boreal jack pine - black spruce forest 50 1 224 <1 3362 Laurentian - Acadian northern pine forest 1 <1 9 <1 3365 Boreal white Spruce - fir forest 58 2 250 1 3366 Laurentian - Acadian pine - hemlock forest 5 <1 36 <1 3407 Laurentian pine barrens 21 <1 36 <1 3444 Eastern boreal floodplain woodland 3 <1 9 0 48 Table 2. S5 . (c ont d ) LANDFIRE Vegetation Type Marten Background Code Type Sections Percent Sections Percent 3466 Great Lakes wooded dune and swale 0 0 9 <1 3475 Laurentian - Acadian floodplain forest 5 <1 36 <1 3477 Boreal acidic peatland forest 76 2 234 1 3481 Laurentian - Acadian alkaline conifer hardwood forest swamp 263 8 2016 8 3534 Managed tree plantation - northern and central hardwood and conifer plantation group 20 1 151 1 3909 Eastern cool temperate urban shrubland 0 0 7 <1 3974 Eastern cool temperate row crop 0 0 4 <1 3975 Eastern cool temperate close grown crop 3 <1 52 <1 3977 Eastern cool temperate pasture and hayland 20 1 229 1 49 Table 2. S6 . Average (standard error) and range of monthly maximum temperatures and monthly maximum precipitation for October, November, and December based on PRISM data, and average (standard error) and range of road densities (based on All Roads layer; Center for Shard Solutions and Technology Partnerships 2014) for the Upper Peninsula, Michigan from 2001 - 2018. Marten Background Category Variable a (SE) Range (SE) Range Temperature ( Oct 13.8 (0.03) 9.7, 17.8 12.3 (0.01) 6.1, 19.2 Nov 6.4 (0.04) 2.0, 12.6 4.6 (0.02) - 3.3, 11.3 Dec - 3.3 (0.07) - 10.1, 4.0 - 2.0 (0.02) - 11.0, 5.7 Precipitation (mm) Oct 65 (.71) 6, 200 102 (.27) 22, 285 Nov 71 (.24) 32, 133 65 (.20) 5, 203 Dec 59 (.36) 18, 162 60 (.16) 4, 165 Roads ( m 2 ) Maintained 108 (2.84) 0, 1,038 91 (1.10) 0, 1,731 a Temp = average monthly maximum temperature, Precip = average monthly maximum precipitation, Seasonal = density of unmaintained roads, Maintained = density of roads maintained throughout the year (Center for Shared Solutions and Technology Partnerships 2004). 50 LITERATURE CITED 51 LITERATURE CITED Ahlers, A. A., J. Edward, J. Heske and C. A. Miller. 2016. Economic influences on trapper participation and per capita harvest of muskrat. The Wildlife Society Bulletin 40(3):548 - 553. Albert, D. A. 1995. Regional landscape ecosystems of Michigan, Minnesota, and Wisconsin: a working map and classification. General Technical Report NC - 178. U. S. Department of Agriculture, Forest Service, North Central Forest Experiment Station, St. Paul, Minnesota, USA. Allen, A. W. 1984 Habitat suitability index models: marten. U.S. Fish and Wildl. Serv. FWS/OBS - 82/10.11. Washington, D.C. Revised. 13pp Allen, M. L., N. M. Roberts, and T. R. Van Deelen. 2018. Hunter selection for larger and older male bobcats affects annual harvest demography. R. Soc. open sci. 5:180668. Archibald, W. R., and R. H. Jessup. 1984. Population dynamics of the pine marten ( Martes americaca ) in the Yukon Territory. Northern Ecology and Resource Management (eds R. Olson, R. Hastings & F. Geddes), University of Alberta Press, Edmonton, Canada, p. 81 - 97. Baddeley, A., and R. Turner. 2005. Spatstat: an R package for anal yzing spatial point patterns. J. Stat. Softw. 12(6) 1 - 42 http://www.jstatsoft.org/v12/i06/. Belsley, D., E. Kuh, and R. Welsch. 1980. Regression diagnostics: Identifying influential data and sources of collinearity. John Wiley & Sons, Hoboken, New Jersey. Bischof R., J. E. Swenson, N. G. Yoccoz, A. Mysterud, O. Gimenez. 2009. The magnitude and selectivity of natural and multiple anthropogenic mortality causes in hunted brown bears. J. Anim. Ecol. 78: 656 - 665. Bivand, R. S., Gomez - Rubio, V., and Rue, H. ( 2015). Spatial data analysis with R - INLA with some extensions. Journal of Statistical Software, 63(20):1 - 31. Blangiardo, M., and M. Cameletti. 2015. Spatial and SpatioTemporal Bayesian Models with R - INLA. Chichester, UK: John Wiley & Sons, Ltd. Bryan H. 1977. Leisure value systems and recreational specialization: the case of trout fisherman. J. Leis. Res. 9:174 - 187. Bubenik, A. B. 1987. Behavior of moose ( Alces alces ssp.) of North America. Swedish Wildlife Resources Supplement 1: 333 - 365. 52 Bull, E. L . and T. W. Heather. 2001. Survival, causes of mortality, and reproduction in the American marten in northeastern Oregon. Northwestern Naturalist 82:1 - 6. Buskirk, S. W. and L. L. McDonald. 1989. Analysis of Variability in home - range size of the American m arten. J. Wildl. Manag. 53(4): 997 - 1004. Buskirk, S. W. and S. L. Lindstedt. 1989. Sex biases in trapped samples of Mustelidae. J. Mamm. 70(1): 88 - 97. Buskirk, S. W., and R. A. Powell. 1994. Habitat ecology of fisher and American martens. Pages 283 - 296 in S. W. Buskirk, A. S. Harestad, M. G. Raphael, and R. A. Powell, eds. Martens, sables, and fishers: biology and conservation. Cornell University Press, Ithaca, New York. Carroll, Carlos. 2005. Carnivore restoration in the Northeastern U.S. and Southeastern Canada: A regional - scale analysis of habitat and population viability for wolf, lynx, and marten (Report 2: lynx and marten viability analysis). Wildlands project special paper No. 6. Richmond, VT: Wildlands Project. pp. 46. Center for Shared Solution and Technology Partnerships. 2014. CSS_SDE_ADMIN.ROADS_ALL_MGF_V17. Center for Shared Solution and Technology Partnerships 15a - scale evaluation of carnivore road density mortality: the effects of landscape and local scale characteristics. Mammal Research 60(3): 233 - 243. pour la mise en evidence de spermatozoides dans le tractus genital. Canadian Journal of Zoology 10:804 - 809. Clough, G. C. 1987. Relations of small mammals to forest management in northern Maine. Canadian Field - Naturalist. 101: 40 - 48. - Bianchet, M. 2003 U ndesirable evolutionary consequences of trophy hunting. Nature 426:655 - 658. Courtois, R. 1989. Analyse du systeÁme de suivi de l'orignal. MinisteÁre du Loisir, de la Chasse et de la Peà che. Direction de la gestion des espeÁces et des habitats. Publicatio n 1770. Gouvernement du QueÂbec, QueÂbec, Canada. Crête, M., R. J. Taylor, and P. A. Jordan. 1981. Optimization of moose harvest in southern Quebec. J. of Wildl. Mange. 45:598 - 611. Crichton, V. 1992. Six years (1986/87 - 1991/92) summary of in utero produc tivity of moose in Manitoba Canada. Alces. 28:203 - 214. 53 2011. Limiting factors and landscape connectivity: the American marten in the Rocky Mountains. Landscape Ec ology 26:1137 - 1149. Daigle, J. T., R. M. Muth, R. R. Zwick, and R. J. Glass. 1998. Sociocultural dimensions of trapping: a factor analytic study of trappers in six northeastern states. Wildlife Society Bulletin 26:614 - 625. Davis, M. H. 1983. Post - relea se movements of introduced marten. J. Wildl. Manage. 47(1):59 - 66. Department of Natural Resources. 2016. Forest roads inventory. https://www.michigan.gov/dnr/0,4570,7 - 3 50 - 79119_79148_80679 --- ,00.html . Accessed 1 April 2019. Elsken - Lacy, P., A. M. Wilson, G. A. Heidt, and J. H. Peck. 1999. Arkansas gray fox fur price - harvest model revisited. J. Arkansas Acad. Sci. 53:50 - 54. ESRI 2018. ArcGIS desktop: release 10. Redlands, CA: Environmental Systems Research Institute. Fecske, D. M., J. A. Jenks, V. J. Smith. 2002. Field evaluation of a habitat - relation model for the American marten. Wildlife Society Bulletin. 30:3. Fortin, C. and M. Cantin. 1994. The effects of trapping on a newly exploited American marten population. Pages 179 - 191 in S. W. Buskirk, A. S. Harestad, M. G. Raphael, and R. A. Powell, eds. Martens, sables, and fishers: biology and conservation. Cornell Univ. Press, Ithaca, N. Y. Frawley, B.J. 2018 . 2018 marten and fisher harvest survey. Michigan Department of Natural Resources Wildlife Division Report No. 3683. https://www.michigan.gov/documents/dn r/2015_fisher_marten_harvest_report_678263_7 .pdf . Accessed 20 March 2019. Frawley, B.J. 2017. 2015 Marten and fisher harvest survey. Michigan Department of Natural Resources Wildlife Division Report No. 3629. https://www.michigan.gov/documents/dnr/2015_fisher_marten_harvest_report_548413_7 .pdf . Accessed 8 March 2019. Frawley, B. J., and D. Etter. 2008. 2007 bobcat hunter and trapper opinion survey. Michig an Department of Natural Resources, Wildlife Division Report No. 3486. https://www.michigan.gov/documents/dnr/bobcat_harvest_report_2017_642965_7.pdf Accessed 10 March 2019. 54 Garel, M., Cugnasse, J. M., Maillard, D., Gaillard, J. M., Hewison, A. J. M. and Dubray, D. 2007. Selective harvesting and habitat loss produces long - term life history changes in a mouflon population. Ecological Applications 17:1607 - 1618. Ge hrt, S. D., G. F. Hubert, and J. A. Ellis. 2002. Long - term population trends of raccoons in Illinois. Wildlife Soc. Bull. 30:457 - 463. Ginsberg, J. R. and E. J. Milner - Gulland. 1994. Sex - biased harvesting and population dynamics in ungulates: implications for conservation and sustainable use. Conserv. Biol. 8:157 - 166. Harris, J. E. and C. V. Ogan., Eds. 1997. Mesocarnivores of Northern California: Biology, Management, and Survey Techniques, Workshop Manual. August 12 - 15, 1997, Humboldt State Univ., Arcata, CA. The Wildlife Society, California North Coast Chapter, Arcata, CA. 127 p. Hendricks, J. 2019. perturb: Tools for evaluating collinearity. R package version 2.10. Hiller, T. L., J, R. Etter, and A. J. Tyre. 2011. Factors affecting harvests of fisher and American martens in northern Michigan. J. Wildl Manage. 75:1399 - 1405. Hodgman, T. P., D. J. Harrison, D. M. Phillips, K. D. Elowe. 1997. Survival of American marten in an untrapped forest preserve in Maine. In: Proulx, G., H. N. Bryant, P. M. Woodard , editors. Martes : taxonomy, ecology, techniques, and management. Edmonton, AB: Provincial Museum of Alberta. p. 86 - 99. Hodgman, T. P., D. J. Harrison, D. D. Katnik, and K. D. Elowe. 1994. Survival in an intensively trapped marten population in Maine. J. Wildl. Manage. 58(4):593 - 600. Humphreys, J. M., Elsner, J. B., Jagger, T. H., Pau, S. 2017. A Bayesian geostatistical approach to modeling global distributions of Lygodium microphyllum under projected climate warming. Ecological Modelling 363, 192 - 206. Humphreys, J. M., Murrow, J. L., Sullivan, J. D., & Prosser, D. J. 2019. Seasonal occurrence and abundance of dabbling ducks across the continental United States: Joint spatio - temporal modelling for the Genus Anas. Diversity and Distributions, 25(9), 1497 1508. https://doi.org/10.1111/ddi.12960 Johnson, F. A., Moore, C. T., Kendall, W. L., Dubovsky, J. A., Caithamer, D. F., Kelley, J. R., and Williams, B. K. 1997. Uncertainty and the management of mallard h arvest. The Wildlife Society 61(1) 202 - 216 pp. JØrgensen, C., Enberg, K., Dunlop, E. S., Arlinghaus, R., Boukal, D. S., Brander, K., Ernande, B., Gardmark, A., Johnston, F., Matsumura, S., Pardoe, H., Raab, K. Silva, A., Vainikka, A., Dieckmann, U., Heino , M. and Rijnsdorp, A. D. 2007. Managing evolving fish stocks. Science 318: 1247 - 1248. 55 Kapfer, P. M., and K. B. Potts. 2012. Socioeconomic and ecological correlates of bobcat harvest in Minnesota. J. Wild. Manage. 76:237 - 242. Krohn, W.B., S.M. Arthur, and T.F Paragi. 1994. Mortality and vulnerability of a heavily trapped fisher population. Pages 137 146 in S.W. Buskirk, A.S. Harestad, M.G. Raphael, and R.A. Powell, editors. Martens, sables, and fishers: Biology and conservation. Cornell University Press, Ithaca, New York. LANDFIRE. 2010. U.S. Department of Agriculture and U.S. Department of Interior. Accessed 15 October 2018 at http://landfire.cr.usgs.gov/viewer/ . LANDFIRE. 2019. Data prod ucts overview. Online at https://www.landfire.gov/data_overviews.php , Last accessed 6 September 2019. Langvatn, R. and Loison, A. 1999. Consequences of harvesting on age structure, sex, ratio and population dynamics of red deer Cervus elaphus in central Norway. Wildlife Biology 5:213 - 223. Laurian, C., J. P. Ouellet, R. Courtois, L. Breton, and S. St - Onge. 2000. Effects of intensive harvesting on moose reproduction. J. Appl. Ecol. 37:515 - 531. Lin dgren, F., H. Rue, J. Lindström. 2011. An explicit link between Gaussian fields and Gaussian Markov random field: The stochastic partial differential equations approach. Journal of the Royal Statistical Society. Series B: Statistical Methodology 73, 423 - 49 8. Lindstedt, S. L., B. J. Miller, and S. W. Buskirk. 1986. Home range, time, and body size in mammals. Ecology, 67(2):413 - 418. Markgren, G. 1969. Reproduction of moose in Sweden. Viltrevy 6:129 - 299. Martinez M., C. Rodriguez - Vigal, O. R. Jones, T. Coulson, A. San Miguel. 2005. Different hunting strategies select for different weights in red deer. Bio. Lett. 1:353 - 356. McLellan, B. N., F. W. Hovey, R.D. Mace, J. G. Woods, D. W. Carney, M. L. Gibeau, W. L. Wakkinen, W. F. Kasworm. 1999. Rates and causes of grizzly bear mortality in the interior mountains of British Colombia, Alberta, Montana, Washington, and Idaho. J. Wildl. Manage. 63:911 - 920. Moors, P. J. 1980. Sexual dimorphism in the body size of mustelids (Carnivora): the roles of food habitats and breeding systems. Oikos, 34:147 - 158 Mysterud A., P. Tyjanowski, M. Panek. 2006. Selectivity of harvesting differs between local and foreign roe deer hunters: trophy stalkers have the first chot at th e right place. Biol. Lett. 2:632 - 635. 56 Mysterud A. 2011. Selective harvesting of large mammals: How often does it result in directional selection? J. Appl. Ecol. 48:827 - 834. Nordyke, K. A. and Buskirk, S. W. 1991. Southern red - backed vole, Clethrionomys gapperi , populations in relation to stand succession and old - growth character in the central Rocky Mountains. Canadian Field - Naturalist. 105:330 - 334. Noss, A. J. 1999. Censuring rainforest game species with communal net hunts. Afr. J. Ecol. 37:1 - 11. Pear son, D. E. and Ruggiero, L. F. 2001. Test of the prey - base hypothesis to explain use of red squirrel midden sites by American martens. Canadian J. of Zoology 79(8) pg. 1372 - 1379. Pelletier, F., Festa - Bianchet, M., and Jorgenson, J. T. 2012. Data from sele ctive harvests underestimate temporal trends in quantitative traits. Bio Lett. 8:878 - 881. Poole, K. G., G. M. Matson, M. A. Strickland, A. J. Magoun, R. P. Graf, and L. M. Dix. 1994. Age and sex determination for American martens and fishers. Pages 204 - 22 3, in S. W. Buskirk, A. S. Harestad, M. G. Raphael, and R. A. Powell, editors. Marten, sables, and fishers: biology and conservation. Cornell University Press, Ithaca, New York, USA. Poole, K. G. 2003. A review of the Canada Lynx, Lynx Canadensis, in Cana ds. Can. Field - Naturalist 117:360 - 376. Powell, R.A. 1994. Structure and spacing of Martes populations. Pages 101 - 121 in S.W. Buskirk, A.S. Harestad, M.G. Raphael, and R.A. Powell, editors. Martens, sables, and fishers: Biology and conservation. Cornell Un iversity Press, Ithaca, New York. Powell, R. A., S. W. Buskirk, and W. J. Zielinski. 2003. Fisher and marten ( Martes pennant ) and ( Martes Americana ). Pages 758 - 786 in G. A. Feldhamer, B. C. Thompson, and J. A. Chapman, editors. Wild mammals of North Ameri ca: biology, management and conservation. Second edition. John Hopkins University Press, Baltimore, Maryland, USA. Proaktor, G., T. Coulson and E. J. Milner - Gulland. 2007. Evolutionary responses to harvesting in ungulates. J. Anim. Eco. (76) 4:669 - 678. PRISM. 2018. Precipitation and minimum temperature climate data. See http://prism.oregonstate.edu . R Core Team. 2017. R: A language and environment for statistical computing. R Foundation for Statistical Compu ting, Vienna, Austria. URL https://www.r - project.org/ . RStudio Team. 2015. RStudio: Integrated Development for R. Studio, Inc, Boston, MA URL http://www.rstudio.com/. 57 Robitaille, J. F., and K. Aubry. 2000. Occur rence of activity of American martens ( Martes Americana ) in relation to roads and other routes. Acta Theriologica 45:137 - 143. Ruette, S., J. M. Vandel, M. Albaret, S. Devillard. 2015. Comparative survival patterns of the syntopic pine and stone martens in a trapped rural area in France. Journal of Zoology 295(3): 214 - 222.Ruggiero, L. F., Aubry, K. B., Buskirk, S. W., Lyon, L. J., and Zielinski, W. J. 1994. The scientific basis for conserving forest carnivores American marten, fisher, lynx, and wolverine in the western United States. Gen. Tech. Rep RM - 254. Ft. Collins, CO: U. S. Department of Agriculture, Forest Service, Rocky Mountain Forest and Range Experiment Station. Pg. 184. Saether, B. E., Solberg, E. J. and Heim, M. 2003. Effects of altering sex rati o structure on the demography of an isolated moose population. Journal of Wildlife Management 67: 455 - 466. Schrodle B and Held L (2011). Spatio - Temporal Disease Mapping Using INLA. Environmetrics, 22(6), 725 - 734. Skalski, J. R., J. T. Millspaugh, M. V. C lawson, J. L. Belant, D. R. Etter, B. J. Frawley, and P. D. Friedrich. 2011. Abundance trends of American martens in Michigan based on statistical population reconstruction. The Journal of Wildlife Management 75:1767 - 1773. Soukkala, A. M. 1983. The effect s of trapping on marten populations in Maine. Thesis, University of Maine, Orono, USA. Stephenson, T. H., J. W. Testa, G. P. Adams, R. G. Sasser, C. C. Schwartz, and K. J. Hundertmark. 1995. Diagnosis of pregnancy and twinning in moose by ultrasonography and serum assay. Alces . 31:167 - 172. Strickland, M. A. 1994. Harvest management of fishers and American martens. Pages 149 - 164 in S. W. Buskirk, A. S. Harestad, M. G. Raphael, and R. A. Powell, editors. Martens, sables, and fishers: biology and conservati on. Cornell University, Ithaca, New York, USA. Strickland, B. K., S. Demarais, P. D. Jones, and C. M. Dacus. 2013. Phenotypic and reproductive variation in female white - tailed deer: the role of harvest and environment. The Journal of Wildlife Management. 77(2): 243 - 253 pp. Tevis, L. Jr. 1956. Responses of small mammal populations to logging of Douglas fir. Journal of Mammalogy. 37:189 - 196. Thomas, R. N. 1991. An analysis of the influence of males age and sex ratios on reproduction in British Columbia moose ( Alces alces L.) populations. MS Thesis, University of British Colombia, Victoria Canada. U. S. Census Bureau. 2009. Population estimates, county population estimates. http://w ww.census.gov/popest/counties/CO - EST2008 - 01.html . Accessed 15 Apr 2019. 58 United States Department of Agriculture, Natural Resources Conservation Service. 2006. Land resource regions and major land resources areas of the United States, the Caribbean, and th e Pacific Basin. U.S. Department of Agriculture, Handbook 296. < ftp://ftpfc.sc.egov.usda.gov/NSSC/Ag_Handbook_269/handbook_296_low.pdf >. Accessed 5 March 2019. Watt, W. R., J. A. Baker, D. M. Hogg, J. G. McNicol, and B. J. Naylor. 1996. Forest Sault Ste. Marie. Weibe, P. A., J. M. Fryxell, I. D. Thompson, L. Börger and J. A. Baker. 20 13. Do trappers understand marten habitat? J. Wild. Mngmt. 77(2):379 - 391. Williams, B. W., Gilbert, H. J., and Zollner, P. A. 2007. Historical perspective on the reintroduction of the fisher and American marten in Wisconsin and Michigan. Gen. Tech. Rep. N RS - 5. Newtown Square, PA: USDA forest service. Zielinski, W. J. and Kucera, T. E. 1995. American marten, fisher, lynx, and wolverine: survey methods for their detection. Gen. Tech. Rep. PSW - GTR - 157. Albany, CA: Pacific Southwest Research Station, Fore st Service, U. S. Department of Agriculture: pg. 163. 59 CHAPTER 3 MULTISCALE RELATIONSHIPS OF SONGBIRD OCCUPANCY IN A MANAGED FOREST LANDSCAPE A bstract I mpact s of forest management on biodiversity is a major conservation concern. In response, forest managers implement practices designed to mitigate negative impacts of some timber harvest techniques, like clearcutting. These practices include protection of waterbodies and unique ecosystems, protection of uncommon species, and retention of forest structure elements. Oftentimes, information on effectiveness of these practices is lacking, yet managers increasingly must defend implementation. I evaluated the effects of stand - and landscape - level variables across varying levels of structural retention on songbird occupancy in clearcut aspen stand s in the northwestern Lower Peninsula, Michigan . Field technicians surveyed birds during the breeding season of 2010 and 2011 at 250 sites. I built 5 candidate occupancy models for 8 common species. M y candidate models included stand - and landscape - level covariates; stand - level covariates included clearcut age, canopy cover of structural retention, understory woody stem density, and distance to the nearest plot. Landscape - level variables included amoun t of 5 LANDFIRE cover types surrounding survey points; Riparian, Short Conifer (0 - 5m tall), Tall Conifer (>5m tall), Short Hardwood, and Tall Hardwood. I calculated the amount of each cover type within 150, 250, and 400 m buffers from plot center. I found that landscape level variables poorly described occupancy probability for mature forest birds, and that occupancy of only red - eyed vireo negatively related to a plot level covariate (i.e., woody stem density, mostly in the understory). Conversely, occupanc y probability for some birds associated with dense understories or riparian areas related to landscape covariates, including American redstart, 60 chestnut - sided warbler, and Nashville warbler. Eastern towhee occupancy probability associated with a plot level covariate; age of the clearcut containing the plot. I found highly variable predictors of bird occupancy among species even within guilds, highlighting challenges associated with identifying widespread prescriptions that account for all species. My resear ch highlights the importance of assessing stand - and landscape - level factors when quantifying effectiveness of wildlife conservation practices like structural retention. I ntroduction For decades, concerns regarding the role of forest management in declin es of songbird populations across North America have persisted ( Tittler et al. 2001 ; Griesser et al. 2007 ; Moore et al. 2010 ). Clearcutting (a silviculture prescription) is one forest management technique resulting in significant changes to bird habitats. Clearcutting is an effective management tool for encouraging early successional habitat and for increasing habitat heterogeneity ( Annand and Thompson 1997; Costello et al. 2000; King and DeGraaf 2000; Thompson and DeGraaf 2001 ; Franklin et al. 2002 ; DeGraa f and Yamasaki, 2003 ) , but results in loss of older forest habitats ( Wallendorf et al. 200 7 ). Some birds are disturbance - sensitive species ( DeGraaf and Yamasaki 2003 ), meaning that occupancy, abundance, and demographics are readily impacted by forest management. Given current focus on conservation of biodiversity in managed forest landscapes (e.g., Sustainable Forestry Initiative 2010 ; Forest Stewardship Council 2012 ) , improved understanding of how birds respond to varying silvicultural prescriptions and landscape context is important . Clearcutting effects on birds varies by species, regional landscape characteristics, size of clearcut openings, and forest type ( Thomp son and Curran 1995 ; Wallendorf et al. 200 7 ) . To ameliorate the negative effects of clearcutting on mature forest dependent birds, retention of 61 unharvested forest p atches or individual structures (e.g., green trees, snags) is suggested (reviewed by Demarai s et al. 2017 ). The idea is that these remnant forest structures, if embedded in a larger matrix of managed land s, could function as refuges allowing species sensitive to disturbance to persist until forest regeneration occurs ( Thomas 2007 ). A problem with retained structures is that patterning rarely emulate s natural patch - disturbance patterns, instead representing anthropogenic - imposed patch structure on the landscape ( Freedman et al. 1994 ) . For example in eastern boreal mixed - wood forest s conversions of large - scale mature mixed - wood forests to deciduous forests ( Carleton and McLellan 1994 ) , short er timber rotations ( that truncate successional trajectories; Spies and others 1994 ; Gauthier et al. 1996 ) , and timber stand improvement practices ( that simplify forest structure; Wallendorf et al. 200 7 ) resulted in anthropogenic patchiness and vegetation structure and species composition that threaten some bird communities ( Drapeau et al. 2000 ) . Responses of birds to structural retention vary depending on patch sizes, clearcut sizes, condition of surrounding landscapes, and age of the surrounding clearcut forest. Forest structure changes rapidly within the first 10 - 15 years post - clearcutting in most managed ecosystems ( DeGraaf and Yamasaki 2003 ). In landscapes do minated by older forests and small clearcut sizes, retention effects on songbirds are minimal, with age of the surrounding regenerating forest a primary determinant of the bird community (e.g., Otto and Roloff 2012 ). In landscapes dominated by intensively managed younger forests with larger retention patch sizes, bird responses to retention are more pronounced (e.g., Linden et al. 2012). These studies collectively indicate that landscape context has a strong influence on function of retention patches as bir d habitat (Aubry et al. 1999; Tittler et al. 2001). 62 Aspen ( Populus spp.) is an important component of timber production in Michigan. Aspen are highly invasive after disturbances and can aggressively reproduce under certain circumstances ( Haeussler 2004 ) F r om 1900 to 1966 aspen increased from 117,400 ha to 1.7 million ha (Dickmann and Leefers 2016). From 1966 to 1993, a s pen forest s in Michigan declined by an estimated 600,000 ha, largely because some landowners allowed aspen to convert to shade tolerant spe cies (Dickmann and Leefers 2016). Managers primarily use clearcutting and subsequent natural regeneration to manage aspen, with prescriptions on state - owned lands requiring retention of green trees to provide wildlife habitat (Bielecki et al. 2006). My goa l was to quantify the roles that retention patches and landscape context play in bird use of clearcut aspen forests. Previously, Otto and Roloff (2012) found minimal support for retention effects on bird occupancy dynamics in clearcut aspen so I predicted that landscape context would better describe bird occupancy. Furthermore, I predicted that birds would respond to different landscape - level components depending on general habitat affinities, with later successional species more dependent on older forest a mount around the clearcuts. Lastly, I predicted that landscape context closer to clearcuts would have a greater effect on songbird use than more distal contexts. My results provide species - specific guidance for managing bird assemblages in aspen clearcuts, as influenced by three elements under direct management control: 1) structural retention, 2) clearcut age, and 3) landscape context. M ethods S tudy A rea My study used data collected from Otto and Roloff (2012) , and the study area description is consistent with that publication. Field data were collected in the northwestern Lower Peninsula of Michigan, USA, from 2010 - 2011. The last glaciation strongly influenced topography and soils 63 in this area, with survey sit es located on glacial outwash - plains with sandy porous soils, and on more fertile ice - contact zones and moraines ( Albert 1995 ). All sampling occurred on state - owned forest lands managed for production of aspen ( Populus spp). Clearcut stands ranged from 1 t o 15 years post - harvest, and each received a green - tree retention prescription ( Bielecki et al. 2006 ). These prescriptions included low (<3%), medium (3 - 10%), and high (>10%) retention of canopy cover or basal area, generally of species that reflected preh arvest stand conditions (but uncommon or mast - producing species were favored; Bielecki et al. 2006). Species that occurred in retention patches included oak ( Quercus spp.), white pine ( Pinus strobus ). red maple ( Acer rubrum ), black cherry ( Prunus serotine ) , American beech ( Feagus grandifolia ), and aspen. Understories were dominated by woody plants that included aspen, red maple, blackberry ( Rubus spp.), black cherry, downy serviceberry ( Amelanchier arborea ), witch - hazel ( Hamamelis virginiana ), American beec h, and ironwood ( Ostrya carpinifolia ). Average high temperatures range from - 3.3°C in January to 25.6°C in July, and monthly precipitation ranges from 3.7cm (February) to 10.0cm in September (U.S. Climate Data 2020). Snowfall mostly occurs from November th rough March, averaging 65.0cm per month (U.S. Climate Data 2020). S ite S election I used study sites selected by Otto and Roloff (2012) , consisting of state - owned aspen stands >8 ha in size, between 1 and 15 years post - harvest. Otto and Roloff (2012) used a Geographic Information System (GIS; ArcGIS 9.1; Environmental Systems Research Institute, Redlands, CA) to superimpose a 60 x 60m lattice over each aspen stand, excluding lattice cells that intersected or included unharvested forest edge, active logging r oads, off - road recreational vehicle trails, or wetlands visible from aerial imagery (2005 National Agricultural Imagery Program (NAIP); Michigan Department of Information Technology, 2007). The remaining 64 lattice cells represented candidates for bird sampli ng. Otto and Roloff (2012) digitized retention canopy cover for the areas encompassed by candidate cells and then randomly selected cells that represented 3 levels of structural retention (<3%, 4 - 10%, >10% canopy cover) and 3 levels of regenerating clearcu t age (1 - 4, 5 - 9, 10 - 15 years post - harvest). B ird S ampling Otto and Roloff (2012) surveyed birds at the center of each selected lattice. Bird sampling occurred in 2010 and 2011. At each site location, an individual observer conducted a 9 - min point - count, sub - divided into 3,3min sub - counts (Otto and Roloff 2012). P oint counts were conducted 30 mins after sunrise and no later than 3.5 hours after sunrise on days with appropriate weather conditions (i.e., no rain, no strong winds; Otto and Roloff 2012). Songbirds seen or heard within a 50 - m sampling radius of the poin t - count center were recorded (Otto and Roloff 2012), with the 50m sampling distance reducing the likelihood of sampling birds in adjacent stands (i.e., reduce edge effects; Otto and Roloff 2012). Observers were trained in estimating songbird detection dist ance prior to the start of the study. Initial site visits occurred in late - May, then every 10 - 14 days thereafter. This sampling method continued until all surveyors visited all sites at least 4 times within a 2 - month period within the breeding season (late - May to mid - July; Otto and Roloff 2012). H abitat C ovariates I identified two levels of habitat covariates for analysis: plot and landscape. Tree crowns of structural retention in each harvested aspen stand were digitized from NAIP imagery (1998 - 2010, depe nding on stand age) in ArcGIS, and I used percent of retained canopy cover within 50 - m of bird survey points as a plot - level covariate (Otto and Roloff 2012). I also included age of 65 the clearcut as a plot - level covariate given the importance of that variab le in previously describing occupancy dynamics of birds from these data (Otto and Roloff 2012). Density of trees >1m tall was calculated for each plot along 4 transects (4 - m x 25 - m) radiating outward in cardinal directions from plot center. I checked thes e covariates for collinearity using condition 1980) in R (R Core Team 2019), package perturb version 3.4.1 (Gates et al. 2019). For landscape - level variables, I use d GIS layers available from the 2012 LANDFIRE Program (LANDFIRE 2012a, b). I first extracted raster cells labeled as forests, and then assigned each to conifer (included conifer - hardwood mixes), hardwoods, and riparian using the EVT_PHYS field in the exist ing vegetation type (EVT) layer (LANDFIRE 2012a). Subsequently, I identified short (0 - 5m) and tall (>5m) forests from the existing vegetation height (EVH) layer (LANDFIRE 2012b) and merged the two queries together to produce 5 landscape - level variables: Ri parian, Short Conifer, Tall Conifer, Short Hardwood, and Tall Hardwood. I calculated percent of each variable within 150, 250, and 400 m of plot center. The 400m radii (~50 ha circle) encompasses multiple breeding territories for focal bird species. M odel G eneration and A nalysis I selected 8 of the more common forest bird species from Otto and Roloff (2012) for modeling that represented varying habitat requirements. Ovenbird ( Seiurus aurocapilla ), red - eyed vireo ( Vireo olivaceus ), and rose - breasted grosbea k ( Pheucticus ludovicianus ) represented mature forest species (Brewer et al. 199 2 ), whereas indigo bunting ( Passerina cyanea ), eastern towhee ( Pipilo erythrophthalmus ), Nashville warbler ( Leiothlypis ruficapilla ), chestnut - sided warbler ( Setophaga pensylva nica ), and American redstart ( Setophaga ruticilla ) represented species with affinity for denser forest understories and riparian areas (Brewer et al. 199 2 ). I generated 66 detection histories for each species consistent with the format needed for a dynamic oc cupancy model ( MacKenzie et al. 200 6). For example, the detection history for 2 site locations was: Site 1: 100 111 000 101 Site 2: 010 000 110 000 means not detected. Each group of 3 numbers represents a 9 - min point count where sites were assumed closed to changes in occupancy (i.e., unoccupied sites did not become occupied, or occupied sites did not become unoccupied; Otto and Roloff 2012). The 9 - m in point count constituted a primary - period within a dynamic occupancy model, whereas the 3 sub - counts were temporal sampling replicates used for estimating detection probability (MacKenzie et al. 2006). For each bird species, I first determined which spa tial extent and landscape - level covariate best explained occupancy. Landscape covariates in my analyses were highly collinear by definition (e.g., if the value of one variable was high in a buffer, the other variables had to be low), so I constructed 15 un ivariate candidate models and used Akaike Information Criteria (AIC) to rank those models. I subsequently used the landscape - level covariate and spatial extent in the top - ranking model in subsequent models that included stand - and landscape - level covariate s to describe bird occupancy. I constructed 5 candidate models for each bird species that included an intercept - only model (i.e., Null Model), models based only on stand or landscape level covariates (Stand and Landscape models, respectively), and a global model that included stand and landscape covariates (Global Model). A distance to nearest plot neighbor was included in all models to account for spatial autocorrelation in plots close to each other (e.g., some plots occurred in the 67 same clearcut stand). L astly, I included a global model without the spatial term (Global Model No Space). For detection, colonization and extinction components of the dynamic occupancy model I used intercept - only terms. Otto and Roloff (2012) previously found this was a reasonab le approach for the detection parameter, and inference on colonization or extinction was not part of my objective. For each species I ranked occupancy models using AIC, and model averaged parameter estimates for competing models ( 2 AIC). I evaluated 95% confidence intervals from model averaged parameter estimates and those not including 0 were deemed significant correlates of occupancy. I used model averaging and 95% confidence intervals as described for the occupancy model to determine parameter signific ance. I plotted relationships of significant occupancy parameters by holding other covariates at mean values. Results Field crews surveyed 250 sites; 157 sites in 2010, 93 different sites in 2011. Each stratum contained samples with more old (10 - 15 years since harvest) and high (>10% cover; 64 plots) plots compared to the other groupings (Table 1). The least sampled stratum was old - aged (10 - 15 years) and mid - retention (4 - 10% cover; 10 plots). Imbalance in the sample design reflects field implementation of retention practices (more likely at low and high amounts) and random lattice placement. Stratifying the sample achieved representation across a broad range of covariates as continuous predictors in bird models. Retained cover across all sites ranged from 0 - 100% and stand ages from 1 - 15 years. Regenerating stem densities ranged from 27 to 1,775/100m 2 . None of these variables exhibited collinearity. Areas within buffers around bird survey points were primarily (>52% on average) tall hardwood forest, follow ed by riparian and tall conifer forests (>7%; Table 2). Thus, on average recent timber harvests were uncommon in surrounding forest areas, indicating a landscape 68 context dominated by taller forests. However, short forests (up to 52%) surrounded some points (Table 2). Across the spatial extents I evaluated, cover type percentages were relatively invariant, except for tall hardwoods that ranged from 0 - 100% (Table 2). For evaluating landscape variables across varying buffer sizes, I failed to find a clear top - ranking model (AIC wt ranged from 0.09 0.49); most bird species except American redstart and chestnut - landscape covariates did not greatly differentiate in explaining occupa ncy probability (Tables S1 S8). I found that the top ranked landscape variable for ovenbirds, red - eyed vireos, and, chestnut - sided warblers was cover of tall conifer forest within 250m of the survey point (250m for ovenbird, 150m for red - eyed vireos and chestnut - sided warblers; Tables S1, S2, S3). Riparian cover was the most influential landscape variable for indigo bunting, Nashville warbler, American redstart, and rose - breasted grosbeak (Tables S4, S5, S6, S7). Riparian was most influential at 150m for indigo buntings (Table S4), 250m for Nashville warblers (Table S5, and 400m for American redstarts and rose - breasted grosbeaks (Tables S6, S7) The only species apparently affected by landscape - level hardwoods was eastern towhee at 400m radii (Table S8). For combined stand - and landscape - level models for mature forest species, I found that the Null model received the most support (AIC wt = 0.53; Table 3) for ovenbirds, indicating that none of the variables I explored described occupancy probability better t han random. For red - eyed vireos I found that the Landscape model (AIC wt = 0.41) competed with the Global No Space AIC wt = 0.25) and Null models (AIC wt = 0.21; Table 4). Model averaging indicated that only plot - - 1.64, 95% CI = - 0.003, - 7.651), with occupancy probability declining as understory stem density increased. For rose - breasted grosbeaks I found that the Landscape model (AIC wt = 0.53) competed with the Null model 69 (AIC wt = 0.35; Table 5). I failed to find any sig nificant parameters for rose - breasted grosbeak occupancy. Collectively, none of the landscape variables significantly correlated with occupancy probability for mature forest bird species, and only plot - level stem density was important to red - eyed vireos. For bird species associated with denser forest understories or riparian areas, I found that the Null model ranked highest for indigo buntings (AIC wt = 0.48; Table 6). For American redstart occupancy, I found that the Landscape model ranked highest (AIC wt = 0.68), with no - 2.81, 95% CI = - 5.265455, - 0.373560), with occupancy declining as the amount of riparian habitat within 400m increased. Similarly, the Landscape model ranked highest for chestnut - sided warblers (AIC wt = 0.71; Table 8), with no competing models (Table 7). For rufous - sided towhee occupancy, I found the Global No Space model (AIC wt = 0.37) competed with the Stand (AIC wt = 0.27) and Null models (AIC wt = 0.17; Table 9). After model averaging, I found that age of that towhee occupancy increased as clearcut age increased (up to 15 years). For Nashville warbl er occupancy, I found that the Global No Spatial model (AIC wt = 0.44) competed with Landscape (AIC wt = 0.38) and Global (AIC wt = 0.16) models (Table 10). After model averaging, , 95% CI = 0.635, 4.13201), indicating the occupancy probability increased as riparian cover type increased. D iscussion I quantif ied the relative contributions of stand - (e.g., amount of retention , understory stem density, age of the surrounding clearcut) and landscape - level (e.g., proportion of surrounding area in different forest structures) attributes on bird occupancy probability in aspen clearcut s 70 across the northwestern Lower Peninsula of Michigan . I found more supporting evidence for occupancy models based on landscape - level variables compared to stand - based models, but evidence varied by guild. Occupancy probability of birds associated with mature forest conditions did not relate to landscape - level factors, whereas birds associate d with dense understories and riparian areas did. Collectively, my results indicate that landscape - level variables are more influential on occupancy than stand - level variables for early - seral birds using clearcuts with retention in northern Michigan. This finding contradicts my prediction that landscape context would have a greater effect on birds associated with mature forest conditions. Characteristics of the matrix surrounding forest patches are important parts of bird habitat assessments (Saab 1999, R enjifo 2001, Brotons et al. 2003, Heartsill - Scalley and Aide 2003). In my study, significant landscape variables included amount of riparian cover type within 400m (for American redstarts, negative relationship) and 250m (Nashville warblers, positive rela tionship) of survey plots. Riparian types in my study area are highly varied structurally, including dense thickets of alder and scrub - shrub. American redstarts breed in moist, deciduous second - growth woodlands with abundant shrubs, often situated with th ickets (Brewer et al. 199 2 ). Compared to occupying a recent aspen clearcut, the negative association with riparian areas is intuitive as these areas are often occupied (Brewer et al. 199 2 ). My results suggest that abundant riparian cover type around aspen clearcuts decreases occupancy probability of the clearcuts by redstarts, potentially because redstarts are selecting riparian instead. Nashville warblers generally breed in second - growth woodlands with shrubby undergrowth (Brewer et al. 199 2 ). In contrast to American redstarts, my results indicated that riparian cover type around clearcuts increases occupancy probability of the clearcut, potentially because more suitable forest structure occurs in clearcuts. 71 I predicted that landscape context closer to surv ey points would be more influential on bird occupancy than plot - level variables . This prediction held true for 3 of 8 bird species, but none of the landscape variables were significant at the smallest spatial extent (i.e., 150m). My study area generally co nsists of a forested matrix, with interspersed agriculture and recent clearcuts. Lack of a consistently strong landscape - level signal in bird occupancy in my study may relate to the relatively low level of large - scale disturbance. The influence of landscap e context on bird assemblages increases as use of surrounding lands intensifies (Austen et al. 2001, Martin et al. 2006). Similar to my results, Edenius and Sjöberg (2006) found no matrix effect on bird species richness in a landscape dominated by forest. For conservation of birds in managed forest landscapes, managers can leave unharvested reserves (i.e., structural retention) or use selective cuts to retain general stand structure (Tittler, 2001). Structural retention undoubtedly affects wildlife use of timber harvest areas (Verner et al. 1986, DeGraaf et al. 1998, Hagan and Meehan 2002, Linden et al. 2012, Otto and Roloff 2012). I found no support for retention affecting bird occupancy in relatively small (~8 ha) clearcuts embedded in a forest - dominated matrix. My landscape context was dominated by tall forest conditions, with <10% in recently harvested (or short) status and small harvest unit sizes. In contrast, bird response to stand - level retention in landscapes from industrial forest ownerships in the Pacific Northwest showed stronger effects, particularly in areas with larger clearcut sizes and overall younger forest structure across the landscape (Linden et al. 2012). I caution that my data represent a limited number of bird species that potentiall y use clearcuts with retention. I restricted analysis to species that required an intercept - only detection parameter (Otto and Roloff 2012), and those that were relatively ubiquitous and represented varying habitat requirements. Furthermore, my choice of l andscape variables including 72 categories derived from LANDFIRE and extent of radii evaluated undoubtedly affected results. It is plausible that I selected spatial extents that did not coincide with important landscape patterns (Mitchell et al. 2001), or veg etation classifications too coarse to coincide with bird habitat selection (Roloff et al. 2009). As global demands for wood products continue to increase, wildlife conservation in managed forest landscapes will depend on effective practices aimed at ret aining habitat elements as part of operational forest management. The concept of structural retention as a means for mitigating the negative effects of timber harvest is well established (reviewed by Demarais et al. 2017), but effectiveness monitoring is l acking. As forest managers balance demands for increased timber production with activities that are viewed as a cost (e.g., structural retention), effectiveness monitoring becomes critical to defend conservation activities. My results indicate that an impo rtant component of monitoring effectiveness of structural retention for bird conservation (and potentially other taxon) includes consideration of landscape context. 73 APPENDIX 74 FIGURES & T ABLES Table 3. 1 . Number of plots sampled by retention canopy cover (in 50m radius plot) and clearcut age in northwest Lower Peninsula of Michigan, 2010 and 2011. Retained Canopy Cover (%) Clearcut Age (yrs) <3 4 - 10 >10 Row Total 1 - 4 25 16 30 71 5 - 9 19 20 35 64 10 - 15 31 10 64 105 Column Total 75 46 129 75 Table 3. 2 . Average (SE) and range (Min, Max) of cover (%) of landscape variables surrounding bird survey points at 3 radii in northwest Lower Peninsula of Michigan, 2010 and 2011. Radii (m) 150 250 400 Landscape Variable a Mean (SE) Range Mean (SE) Range Mean (SE) Range Short Conifer Forest Tall Conifer Forest Short Hardwood Forest Tall Hardwood Forest Riparian 2 (0.45) 8 (0.89) 6 (0.69) 68 (1.73) 10 (1.16) 0,47 0, 98 0, 52 0, 100 0, 97 2 (0.25) 9 (0.82) 4 (0.44) 56 (1.56) 11 (1.12) 0, 27 0, 63 0, 44 0, 99 0, 84 2 (0.20) 11 (0.81) 3 (0.28) 53 (1.54) 13 (1.24) 0, 18 0, 66 0, 26 0, 96 0, 82 a Derived from 2012 LANDFIRE Program (LANDFIRE 2012a, b) by first extracting grid cells labeled as forests, and then assigning each to conifer (included conifer - hardwood mixes), hardwoods, and riparian using the EVT_PHYS field in the existing vegetation type (EVT) layer (LANDFIRE 2012a). Subsequently, we identified short (0 - 5m) and tall (>5m) forests from the existing vegetation h eight (EVH) layer (LANDFIRE 2012b) and merged the two queries together. 76 Table 3. 3 . Candidate models for o venbird occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - ce in AIC value from top model, AIC wt = weight of evidence, and k = number of model parameters. Occupancy Model a AIC AIC wt k Null Landscape Stand Global Global (No Space) 2009.53 2010.89 2012.48 2013.96 2015.22 0.00 1.36 2.95 4.43 5.69 0.53 0.27 0.12 0.06 0.03 5 6 8 9 8 a Null = only distance to nearest neighbor as a parameter; Landscape = top - ranked landscape - level variable (Supplement A); Stand = years since timber harvest, percent cover of retention, and understory stem density; Global = Landscape + Stand; Global (No Sp ace) = Global model without the nearest neighbor term. 77 Table 3. 4 . Candidate models for r ed - eyed v ireo occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - AIC value from t op model, AIC wt = weight of evidence, and k = number of model parameters. Occupancy Model a AIC AIC wt k Landscape Global (No Space) Null Global Stand 1468.44 1469.47 1469.79 1471.21 1473.64 0.00 1.03 1.35 2.76 5.20 0.41 0.25 0.21 0.10 0.03 6 8 5 9 8 a Null = only distance to nearest neighbor as a parameter; Landscape = top - ranked landscape - level variable (Supplement A); Stand = years since timber harvest, percent cover of retention, and understory stem density; Global = Landscape + Stand; Global (No Space) = Global model without the nearest neighbor term. 78 Table 3. 5 . Candidate models for Rose - breasted grosbeak occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - difference in AIC va lue from top model, AIC wt = weight of evidence, and k = number of model parameters. Occupancy Model a AIC AIC wt k Landscape Null Global Global (No Space) Stand 1989.58 1990.41 1994.26 1994.99 1995.66 0.00 0.83 4.68 5.41 6.08 0.53 0.35 0.05 0.04 0.03 6 5 9 8 8 a Null = only distance to nearest neighbor as a parameter; Landscape = top - ranked landscape - level variable (Supplement A); Stand = years since timber harvest, percent cover of retention, and understory stem density; Global = Landscape + Stand; Global (No Sp ace) = Global model without the nearest neighbor term. 79 Table 3. 6. Candidate models for i ndigo b unting occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - AIC value from t op model, AIC wt = weight of evidence, and k = number of model parameters. Occupancy Model a AIC AIC wt k Null Landscape Stand Global (No Space) Global 1901.30 1901.83 1905.55 1905.87 1906.32 0.00 0.53 4.25 4.57 5.02 0.48 0.37 0.06 0.05 0.04 5 6 8 8 9 a Null = only distance to nearest neighbor as a parameter; Landscape = top - ranked landscape - level variable (Supplement A); Stand = years since timber harvest, percent cover of retention, and understory stem density; Global = Landscape + Stand; Global (No Space) = Global model without the nearest neighbor term. 80 Table 3. 7 . Candidate models for American redstart occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - difference in AIC value f rom top model, AIC wt = weight of evidence, and k = number of model parameters. Occupancy Model a AIC AIC wt k Landscape Global (No Space) Null Global Stand 1353.96 1356.56 1358.46 1359.02 1364.07 0.00 2.60 4.50 5.06 10.11 0.68 0.19 0.07 0.05 0.00 6 8 5 9 8 a Null = only distance to nearest neighbor as a parameter; Landscape = top - ranked landscape - level variable (Supplement A); Stand = years since timber harvest, percent cover of retention, and understory stem density; Global = Landscape + Stand; Global (No Sp ace) = Global model without the nearest neighbor term. 81 Table 3. 8 . Candidate models for c hestnut - sided warbler occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - difference in AIC valu e from top model, AIC wt = weight of evidence, and k = number of model parameters. Occupancy Model a AIC AIC wt k Landscape Null Global Global (No Space) Stand 1417.38 1419.70 1423.35 1424.00 1425.78 0.00 2.32 5.97 6.62 8.40 0.71 0.22 0.04 0.03 0.01 6 5 9 8 8 a Null = only distance to nearest neighbor as a parameter; Landscape = top - ranked landscape - level variable (Supplement A); Stand = years since timber harvest, percent cover of retention, and understory stem density; Global = Landscape + Stand; Global (No Sp ace) = Global model without the nearest neighbor term. 82 Table 3. 9 . Candidate models for eastern towhee occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - AIC value from t op model, AIC wt = weight of evidence, and k = number of model parameters. Occupancy Model a AIC AIC wt k Global (No Space) Stand Null Global Landscape 1590.95 1591.57 1592.58 1593.29 1594.22 0.00 0.61 1.62 2.33 3.26 0.37 0.27 0.17 0.12 0.07 8 8 5 9 6 a Null = only distance to nearest neighbor as a parameter; Landscape = top - ranked landscape - level variable (Supplement A); Stand = years since timber harvest, percent cover of retention, and understory stem density; Global = Landscape + Stand; Global (No Space) = Global model without the nearest neighbor term. 83 Table 3. 10 . Candidate models for Nashville warbler occupancy at points in the northwestern Lower Peninsula of Michigan, 2010 - difference in AIC value f rom top model, AIC wt = weight of evidence, and k = number of model parameters. Occupancy Model a AIC AIC wt k Global (No Space) Landscape Global Null Stand 1947.08 1947.40 1949.08 1953.95 1955.20 0.00 0.31 1.99 6.87 8.12 0.44 0.38 0.16 0.01 0.01 8 6 9 5 8 a Null = only distance to nearest neighbor as a parameter; Landscape = top - ranked landscape - level variable (Supplement A); Stand = years since timber harvest, percent cover of retention, and understory stem density; Global = Landscape + Stand; Global (No Sp ace) = Global model without the nearest neighbor term. 84 SUPPLEMENTAL TABLES Table 3. S1 . Candidate univariate landscape - level models for ovenbirds surveyed in 2010 - 2011, difference in AIC value from top model, AIC wt = weight of evidence, and k = number of model parameters. Variable, Radii AIC c c AIC wt k Tall Conifers, 250m Tall Conifers, 400m Riparian, 250 Short Hardwoods, 400m Tall Hardwoods, 400m Short Conifers, 250m Riparian, 400m Short Hardwoods, 150m Riparian, 150m Short Conifers, 150m Tall Conifers, 150m Tall Hardwoods, 250m Short Hardwoods, 250m Tall Hardwood, 150m Short Conifer, 400m 2012.69 2012.69 2013.55 2013.72 2013.80 2013.82 2013.84 2013.91 2014.03 2014.03 2014.03 2014.03 2014.07 2014.09 2014.10 0.00 0.27 0.86 1.03 1.11 1.13 1.15 1.22 1.34 1.34 1.34 1.34 1.38 1.40 1.41 0.11 0.10 0.07 0.07 0.06 0.06 0.06 0.06 0.06 0.06 0.06 0.06 0.06 0.06 0.06 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 85 Table 3. S2 . Candidate univariate landscape - level models for red - eyed vireos surveyed in 2010 - difference in AIC value from top model, AIC wt = weight of evidence, and k = number of model parameters. Variable, Radii AIC c c AIC wt k Tall Conifers, 150m Tall Conifer, 250m Short Conifer, 150m Tall Conifers, 400m Short Conifers, 250m Short Conifer, 400m Short Hardwood, 250m Short Hardwood, 400m Tall Hardwood, 250m Riparian, 400m Riparian, 250m Tall Hardwood, 150m Short Hardwood, 150m Riparian, 150m Tall Hardwood, 400m 1466.67 1467.27 1468.26 1468.78 1468.79 1469.37 1469.69 1469.85 1470.08 1470.14 1470.18 1470.19 1470.21 1470.22 1470.24 0.00 0.60 1.59 2.10 2.12 2.70 3.02 3.18 3.41 3.47 3.51 3.52 3.54 3.55 3.57 0.21 0.16 0.09 0.07 0.07 0.05 0.05 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.04 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 86 Table 3. S3. Candidate univariate landscape - level models for chestnut - sided warblers surveyed in 2010 - 2011, northwest Lower Peninsula of Michigan. AIC = Akaike Information Criteria, AIC wt = weight of evidence, and k = number of model parameters. Variable, Radii AIC c c AIC wt k Tall Conifers, 150m Tall Conifers, 400m Tall Hardwood, 150m Tall Hardwood, 400m Tall Conifers, 250m Short Hardwood, 250m Tall Hardwood, 250m Short Hardwood, 150m Riparian, 400m Riparian, 250m Riparian, 150m Short Conifers, 250m Short Conifer, 150m Short Conifers, 400m Short Hardwood, 400m 1418.28 1418.35 1419.75 1420.31 1420.49 1420.72 1420.83 1421.20 1421.30 1421.63 1421.75 1421.82 1421.85 1421.93 1422.06 0.00 0.07 1.48 2.03 2.21 2.45 2.55 2.92 3.02 3.35 3.47 3.55 3.57 3.66 3.79 0.19 0.19 0.09 0.07 0.06 0.06 0.05 0.04 0.04 0.04 0.03 0.03 0.03 0.03 0.03 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 87 Table 3. S4 . Candidate univariate landscape - level models for indigo buntings surveyed in 2010 - difference in AIC value from top model, AIC wt = weight of evidence, and k = number of model parameters. Variable, Radii AIC c c AIC wt k Riparian, 150m Short Hardwood, 250m Riparian, 400m Tall Hardwood150m Tall Hardwood, 400m Short Hardwood, 150m Tall Hardwood, 250m Short Hardwood, 400m Riparian, 250m Tall Conifers, 400m Tall Conifers, 150m Short Conifers, 150m Tall Conifers, 250m Short Conifers, 400m Short Conifers, 250m 1901.39 1901.73 1902.15 1902.15 1902.31 1902.32 1902.38 1902.46 1902.59 1902.80 1902.89 1902.93 1902.95 1902.95 1902.96 0.00 0.34 0.76 0.76 0.93 0.93 0.99 1.08 1.20 1.42 1.50 1.54 1.56 1.57 1.57 0.11 0.09 0.08 0.08 0.07 0.07 0.07 0.06 0.06 0.05 0.05 0.05 0.05 0.05 0.05 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 88 Table 3. S5 . Candidate univariate landscape - level models for Nashville warblers surveyed in 2010 - = difference in AIC value from top model, AIC wt = weight of evidence, and k = number of model parameters. Variable, Radii AIC c c AIC wt k Riparian, 250m Riparian, 400m Tall Conifers, 400m Riparian, 150m Tall Conifers, 250m Tall Conifers, 150m Short Hardwood, 400m Short Hardwood, 250m Short Conifers, 400m Short Conifers, 250m Tall Hardwood, 400m Short Conifers, 150m Tall Hardwood, 150m Short Hardwood, 150m Tall Hardwood, 250m 1945.62 1947.55 1948.39 1950.90 1951.15 1951.21 1951.27 1952.63 1952.88 1953.31 1953.43 1953.67 1954.07 1954.11 1954.14 0.00 1.93 2.77 5.28 5.54 5.59 5.65 7.01 7.26 7.69 7.82 8.05 8.45 8.49 8.52 0.49 0.19 0.12 0.03 0.03 0.03 0.03 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 89 Table 3. S6 . Candidate univariate landscape - level models for American redstarts surveyed in 2010 - = difference in AIC value from top model, AIC wt = weight of evidence, and k = number of model parameters. Variable, Radii AIC c c AIC wt k Riparian, 400m Tall Hardwood, 400m Riparian, 250m Short Conifers, 400m Short Hardwood, 400m Riparian, 150m Tall Conifers, 150m Short Hardwood, 250m Short Hardwood, 150m Short Conifers, 150m Tall Conifers, 250m Short Conifers, 250m Tall Hardwood, 250m T all Hardwood, 150m Tall Conifers, 400m 1351.69 1353.03 1354.33 1355.61 1355.66 1355.69 1356.97 1357.01 1357.26 1357.28 1357.37 1357.39 1357.77 1357.84 1357.96 0.00 1.34 2.64 3.92 3.98 4.01 5.29 5.32 5.57 5.59 5.68 5.71 6.09 6.15 6.27 0.37 0.19 0.10 0.05 0.05 0.05 0.03 0.03 0.02 0.02 0.02 0.02 0.02 0.02 0.02 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 90 Table 3. S7 . Candidate univariate landscape - level models for rose - breasted grosbeaks surveyed in 2010 - 2011, northwest Lower Peninsula of Michigan. AIC = Akaike Information Criteria, AIC wt = weight of evidence, and k = number of model parameters. Variable, Radii AIC c c AIC wt k Riparian, 400m Riparian, 250m Riparian, 150m Tall Conifers, 150m Short Conifers, 400m Tall Hardwood, 400m Short Conifers, 150m Tall Hardwood, 150m Tall Hardwood, 250m Short Hardwood, 400m Short Hardwood, 150m Tall Conifers, 250m Short Conifers, 250m Ta ll Conifers, 400m Short Hardwood, 250m 1990.36 1991.14 1992.51 1992.54 1992.70 1992.85 1993.14 1993.22 1993.29 1993.29 1993.41 1993.60 1993.65 1993.66 1993.66 0.00 0.79 2.16 2.18 2.34 2.50 2.78 2.86 2.93 2.93 3.06 3.24 3.30 3.30 3.31 0.20 0.14 0.07 0.07 0.06 0.06 0.05 0.05 0.05 0.05 0.04 0.04 0.04 0.04 0.04 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 91 Table 3. S8 . Candidate univariate landscape - level models for eastern towhees surveyed in 2010 - difference in AIC value from top model, AIC wt = weight of evidence, and k = number of model parameters. Variable, Radii AIC c c AIC wt k Short Hardwood, 400m Short Conifers, 400m Riparian, 400m Short Hardwood, 250m Tall Conifers, 150m Short Conifer, 150m Tall Hardwood, 400m Riparian, 250m Tall Hardwood, 150m Short Conifer, 250m Tall Conifers, 250m Tall Conifers, 400m Tall Hardwood, 25 0m Riparian, 150m Short Hardwood, 150m 1591.88 1591.95 1592.14 1592.21 1592.24 1592.40 1592.49 1592.52 1592.54 1592.55 1592.61 1592.64 1592.65 1592.65 1592.65 0.00 0.07 0.26 0.34 0.36 0.52 0.62 0.64 0.66 0.67 0.74 0.76 0.77 0.77 0.77 0.09 0.08 0.08 0.07 0.07 0.07 0.06 0.06 0.06 0.06 0.06 0.06 0.06 0.06 0.06 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 92 LITERATURE CITED 93 LITERATURE CITED Albert, D.A. 1995. Regional landscape ecosystems of Michigan, Minnesota, and Wisconsin: A working map and classification. U.S. Forest Service General Technical Report NC - 178. 250pp. Annand, E., and F. Thompson. 1997. Forest bird response to regeneration p ractices in c entral h ardwood f orests. Journal of Wildlife Management 61 (1) : 159 - 171. Austen, M.J.W., C.M. Francis, D.M. Burke, and M.S.W. Bradstreet. 2001. Landscape context and fragmentation effects on forest birds in southern Ontario. Ornithological Applications 103(4):701 - 714. Belsey, D., E. Kuh, and R. Welsch. 1980. Regression diagnostics: Identifying influential data and sources of collinearity. Wiley & Sons, New York. 292 pp. Bielecki, J., J. Ferris, K. Kintigh, M. Koss, D. Kuhr, S. MacKinnon, S. Throop, L Visser, and M. Walters. 2006. Within stand retention guida nce. Michigan Department of Natural Resources, Lansing, MI. http://www.michigan.gov/dnr/ . Brewer, R., G.A. McPeek, and R.J. Adams, Jr. 1992. Ornithological Literature -- The atlas of breeding birds of Michigan . A journal of Ornithology. The Wilson Bulletin 104.3 561pp. Brotons, L., M. Mönkkönen, J.L. Martin. 2003. Are fragments islands? Landscape context and density - area relationships in boreal forest birds. American Naturalist 162(3):343 - 357. Carleton, T. J. , and P. McLellan. 1994. Woody vegetation responses to fire versus clearcutting logging: A comparative survey in the central Canadi an boreal forest. Ecoscience 1: 141 - 152. Costello, C.A., M. Yamasaki, P.J. Perkins, W.B. Leak, and C.D. Neefus, C. D. 2000. So ngbird response to group selection harvests and clearcuts in a New Hampshire northern hardwood forest. Forest Ecology and Management 127:41 - 54. DeGraaf, R.M., J.B. Hestbeck, and M. Yamasaki. 1998. Associations between breeding abundance and stand struct ure in the White Mountains, New Hampshire and Maine, USA. Forest Ecology Management 103:217 - 233. DeGraaf, R.M., M. Yamasaki, W.B. Leak, and J.W. Lanier. 1992. New England wildlife: Management of forested habitats. U.S. Forest Service General Technical Report NE - 144. 271 pp. DeGraaf, R.M., and M. Yamasaki, M. 2003. Options for managing early - successional forest and shrubland bird habitats in the northeastern United States. Forest Ecology and Management 186:179 - 191. 94 Dickmann, D.L., and L.A. Leefers. 2016 . The forests of Michigan, revised edition. University of Michigan Press, Ann Arbor. 312pp. Drapeau, P., A. Leduc, J.F. Giroux, J.P.L. Savard, Y. Bergeron, and W.L. Vickery. 2000. Landscape - scale disturbances and changes in bird communities of boreal mixe d - wood forest. Ecological Monographs 70(3):423 - 444. Edenius, L., and K. Sjöberg. 2006. Distribution of birds in natural landscape mosaics of old - growth forests in northern Sweden: Relations to habitat area and landscape context. Ecography 20(5):425 - 431. Forest Stewardship Council. 2012. FSC p rinciples and c riteria. Online at: https://my.fsc.org/en - my/certification/principles - and - criteria . Last accessed 5 May 2020. Franklin, J. F., T. A. Spies, R. Van Pelt, A.B. Carey, D.A. Thornburgh, D.R. B erg, D.B. Lindenmayer, M.E. Harmon, W.S. Keeton, D.C. Shaw, K. Bible, and J.Q. Chen . 2002. Disturbances and structural development of natural forest ecosystems with silvicultural implications , using Douglas - fir for ests as an example. Forest Ecology and Management 155 : 399 423. Freedman, B., S. Woodley, and J. Loo. 1994. Forestry practices and biodiversity, with particular reference to the Maritime Provinces of eastern Canada. Environmental Rev iew 2:33 - 77. Gates, K.M., Z. Fisher, and C. Arizmendi. 2019. perturb: Random Perturbation of Count Matrices. R package version 0.1.3. Griesser, M., M. Nystrand, S. Eggers, and J. Ekman. 2007. Impact of forestry practices on fitness correlates and populat ion productivity in an open - nesting bird species. Conservation Biology 21:767 - 774. Gauthier, S., A. Leduc, and Y. Bergeron. 1996. Forest dynamics modeling under natural fire cycles: A tool to define natural mosaic diversity for forest management. Pages 417 - 434 i n Sims R.A., Corns I.G.W., Klinka K. (eds) . Global to Local: Ecological Land Classification. Springer, Dordrecht . Haeussler, S., and Y. Bergeron. 2004. Range of variability in Boreal Aspen plant communities after wildlife and clear - cutting. Canadian Journal of Forest Research 34.2 pg. 274 - 288. Hagan, J.M. and Meehan, A. L. 2002. The effectiveness of stand - level and landscape - level variables for explaining bird occurrence in an industrial forest. Forest Science 48(2) 231 - 243 pp. Heartsill - S calley, T., and T.M. Aide. 2003. Riparian vegetation and stream condition in a tropical agriculture - secondary forest mosaic. Ecological Applications 13:225 - 234. 95 King, D.I. and R.M. DeGraaf. 2000. Bird species diversity and nesting success in mature, clear cut and shelterwood forest in northern New Hampshire, USA. Forest Ecology and Management 129(1 - 3):227 - 235. LANDFIRE. 2012a. Existing Vegetation Type (EVT). U.S. Department of Agriculture and U.S. Department of Interior. Online resource at: https://www.landfire.gov/vegetation.php . Last accessed 3 January 2020. LANDFIRE. 2012b. Existing Vegetation Type (EVT). U.S. Department of Agriculture and U.S. Department of Interior. Online resource at: https://www.landfire.gov/vegetation.php . Last accessed 3 January 2020. Linden, D.W., G.J. Roloff, and A. J. Kroll. 2012. Conserving avian richness through structure retention in managed forests of the Pacific Northwest, USA. Forest Ecology and Management 284:174 - 184. MacKenzie, D.I., J.D. Nichols, A.A. Royle, K.H. Pollock, L.L. Bailey, and J.E. Hines. 2006. Occupancy estimation and modeling: Inferring patterns and dynamics of species occurrence. Academic Press, New York, NY. 324 pp. Martin, T.G., S. McIntyre, C.P. Catterall, and H.P Possingham. 2006. Is landscape context important for riparian conservation? Birds in grassy woodland. Biological Conservation 127(2):201 - 214. Michigan Department of Informati on Technology. 2007. National agricultural imagery program data (NAIP 2005). URL http://www.mcgi.state.mi.us/mgdl/ . Mitchell, M.S., R.A., Lancia, and J.A. Derwin. 2001. Using landscape - level data to pred ict the distribution of birds on a managed forest: effects of scale. Ecological Applications 11(6):1692 - 1708. Moore, L.C., B.J.M. Stutchbury, D.M. Burke, and K.A. Elliott. 2010. Effects of forest management on post fledging survival of rose - breasted gros beak ( Pheucticus ludovicianus ). The Auk 127(1) 185 - 194 pp. Otto, C.R.V., and G.J. Roloff. 2012. Songbird response to green - tree retention prescriptions in clearcut forests. Forest Ecology and Management 284: 241 - 250. R Core Team. 2019. R: A language an d environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R - project.org/ . Renjifo, L.M. 2001. Effect of natural and athropogenic landscape matrices on the ab undance of subandean bird species. Ecological Applications 11:14 - 31. Roloff, G.J., M. L. Donovan, D. W. Linden, and M. L. Strong. 2009. Lesson learned from using GIS to model landscape - level wildlife habitat. Pages 287 - 320 in J.J. Millspaugh and F.R. 96 Thompson, III (eds). Models for Planning Wildlife Conservation in Large Landscapes. Academic Press, Burlington, MA. Saab, V. 1999. Importance of spatial scale to habitat use by breeding birds in riparian forests: A hierarchical analysis. Ecological Appli cations 9:135 - 151. Spies, T.A., W.J. Ripple, and G.A. Bradshaw. 1994. Dynamics and pattern of a managed coniferous forest landscape in Oregon. Ecological Applications 4(3):555 - 568. Sustainable Forestry Initiative, 2010. Requirements for the SFI 2010 2014 Program, Section 2: Sustainable Forestry Initiative 2010 2014 standard. Sustainable Forestry Initiative, Washington, DC . 14pp. Thomas, L. 2007. Green - tree retention in harvest units: Noon or burst for biodiversity? Science Findings, Pacific Northwest Res earch Station. Online resource at: https://www.fs.fed.us/pnw/sciencef/scifi96.pdf . Last a ccessed 11 Oct 2019. Thompson, F.R., and R.M. DeGraaf. 2001. Conservation approaches for woody, early s uccessional communities in the eastern United States. Wildl ife Society B ulletin 29 : 483 494. Thompson, I.D., J.A. Baker, and M. Ter - Mikaelian. 2003. A review of the long - term effects of post - harvest silviculture on vertebrate wildlife, and predictive mode ls, with an emphasis on boreal forests in Ontario, Canada. For est Ecology and Management 177 : 441 469. Thompson, I. D. and W.J. Curran. 1995. Habitat suitability for marten of second - growth balsam fir forest in Newfoundland. Canadian Journal of Zoology 73( 11):2059 - 2064. Tittler, R., S.J. Hannon, and M.R. Norton. 2001. Residual tree retention ameliorates short - term effects of clear - cutting on some boreal songbirds. Ecological Applications 11(6):1656 - 1666. U.S. Climate Data. 2020. Climate Kalkaska, Michigan. Online at https://www.usclimatedata.com/climate/kalkaska/michigan/united - states/usmi0444 , Last accessed May 5, 2020. Verner, J. and L.V. Ritter. 1986. Hourly variation in monitoring point counts of birds. Auk 103(1):117 - 124. Wallendorf, M.J., P.A. Porneluzi, W.K. Gram, R.L. Clawson, and J. Faaborg. 2007. Bird response to clear cutting in Missouri Ozark Forest s. Journal of Wildlife Management 71:6 1899 - 1905 pp. 97 CONCLUSION My research offers insights into factors affecting student decisions to pursue a wildlife career, how socio - economic and environmental factors affect wildlife harvest (in my case, American marten), and how habitat management decisions by people can influence wildlife populations. Collectively, my dissertation chapters represent the major components of wildlife management in North America; people, wildlife populations, and wildlife habitat ( Decker et al. 2012 ). Each of these individual research endeavors sought to add practical knowledge to wildlife management. In Chapter 1, I assessed 21 st century students to gain a better understanding of what motivated them to choose natural resources as a career. The perception of current generations of students (e.g., Millennials, Generation Z) is that friends, social media, and other outside influences play an important role in decision - making. I showed that family strongly affects interests in consump tive activities (like hunting, trapping) and decisions on natural resource careers for current natural resources students. I also found that students with consumptive family backgrounds and current consumptive interests decided on natural resources career earlier then students without those backgrounds or interests. My results also indicated that about 50% of natural resources students were not currently interested in consumptive activities, suggesting that natural resource values among the next generation of workers is diversifying. I make the argument that student training in consumptive and non - consumptive values is important for current college curricula as both values will shape the future of wildlife management. In Chapter 2, I assessed factors potent ially affecting harvest of American marten in harvest success is difficult because of complexities associated with individual trapper effort and 98 personal circumstances . I sought to model focal areas of marten harvest from land cover, roads, pelt price, and weather as a means to forecast annual harvest outcomes. I found no evidence that land cover (within the constraints of how I modeled it), proximity to maintained road s, or pelt price could be used to predict marten harvest. Rather, weather conditions preceding and during the December harvest season were most influential on harvest outcomes. I posited that combinations of temperature and precipitation in October through December affected scouting and trap - line preparation, access to trapping areas, movements and vulnerability of marten, and ultimately where trappers focused effort. My results indicate short - term forecasting potential based on weather that may help antici pate harvest outcomes and ultimately regulation setting. Lastly, in Chapter 3 I explored stand - and landscape - level variables influencing songbird occupancy in clearcut forests. I posited that landscape - level variables would be more important for describi ng occupancy that stand - level variables, given that previous research failed to find strong stand - level effects (Otto and Roloff 2012). I failed to find consistent effects for birds associated with mature forest conditions, but for some birds associated wi th dense, shrubby undergrowth and riparian areas I found landscape effects. Lack of a consistent response to stand - or landscape - level variables likely related to condition of the matrix environment; predominately forested. I posit that birds affected by r elatively small - scale disturbances (~8 ha in this study) in heavily forested matrices temporarily displace, occupying adjacent areas until the clearcut regains suitable forest structure for occupancy. Even practices for retaining forest structure in clearc ut areas does not appear to improve occupancy (e.g., see also Otto and Roloff 2012) of clearcuts by birds. My results indicate that in landscapes dominated by forests few consistent, reliable predictors of bird occupancy exist (within the domain of variabl es I assessed), though I found some species - specific exemptions. Furthermore, my results indicate that general 99 prescriptions for entire bird communities may not exist, instead highlighting the importance of species - specific approaches. Overall, my resea rch highlighted various topics that wildlife managers currently face. I provide information that can aid managers in understanding these topics further. My research integrated across people, wildlife populations, and wildlife habitat; focuses on the three pillars that has guided wildlife management for almost 100 years. 100 LITERATURE CITED 101 LITERATURE CITED Decker, D. J., S. J. Riley, and W. F. Siemer. 2012. Human dimensions of wildlife management 2 nd edition. The John Hopkins University Press, Baltimore, Maryland 3 - 14 pp. Otto, C. R. and G. J. Roloff. 2012. Songbird response to green - tree retention prescriptions in clearcut forest. Journal of Forest Ecology and Management 284 241 - 250 pp.