EASTERN MASSASAUGA R ATTLESNAKE POPULATIO N AND HABITAT ECOLOGY IN S OUTHERN MICHIGAN By Stephanie Anne Shaffer A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Fisheries and Wildlife Doctor of Philosophy 2018 ABSTRACT EASTERN MASSASAUGA RATTLESNAKE POPULATION AND HABITAT ECOLOGY IN SOUTHERN MICHIGAN By Stephanie Anne Shaffer The eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) is a federally Threatened species due to factors that include degradation and fragmentation of habita t and corresponding population declines throughout the species range . Further, a fungal pathogen ( Ophidiomyces ophiodiicola ) which affects snake species across North America is a potential threat to massasauga populations. T he ability to identify and assess suitability of current and potential habitats , t o effectively locate massasauga s , and to determine population demographic rates, are warranted for conservation of the species. In 2015 and 2016, w e identified 27 20 - ha study sites of varying habitat quality throughout southern Michigan where massasaugas w ere known to occur historically . At these sites , we applied and validated a habitat suitability index (HSI) mode l by quantifying the habitat attributes defined by the model (% live and dead herbaceous cover, number and DBH of trees and shrubs >3 m in height , % area in early successional deciduous upland and wetland ) . Habitat suitability index scores across all sites ranged from 0.21 to 0.95. We used HSI scores to predict the likelihood that a given site is occupied by massasaugas (i.e., occupancy proba bility). For sites approaching maximum suitability (HSI = 1), predicted occupancy was > 0.5 , while for poor suitability sites (HSI scores approaching 0) predicted occupancy was < 0.2 . We used r esource selection probability function analysis to quantify hab itat use versus availability and found a positive relationship between probability of use and % live and dead herbaceous cover, and a negative relationship between probability of use and number of stems and DBH of trees >3 m in height. To aid researchers and managers in successfully detecting massasaugas in occupied habitat we developed and tested a visual encounter survey method . We quantified detection probability for massasaugas using this method and determined factors important in influencing detection probability (i.e., environmental conditions, surveyor conditions) at 4 study sites . In 2016 , we completed 54 detection surveys with 5 surveyors; we detected 1 massasauga at 11 surveys. Our null detection probabil ity was 0.31. W e found that time spent searching and minimum air temperature were the most important correlates of detection probability of the variables we measured : d etection probability approached 1.00 as searcher time exceeded 90 minutes and approached 0.80 on cooler mornings ( down to 12.8°C) . At s ix of our 27 study sites , we obtained sufficient massasauga locations (i.e., 30) via radio telemetry or random encounter s to estimate survivorship for the study region . In total, we telemetered 22 adults, 6 juveniles, and 10 neonates using both surgical implantation of transmitters and external attachment. For the 137 - day study period during the active season, adult massasauga survivorship estimated using the Mayfield method was 0.767 (SE = 0.016 ) . For juveni les and neonates, apparent survivorship was 0.65 . R adiograph imagery or palpation of 17 gravid females resulted in embryo counts ranging from 5 to 18 ; litter counts in the field ranged from 1 to 15 ( n = 6) . We radio tracked n eonate massasaugas from 2 to 26 days and found that they moved up to 551.2 m from the gestation site . S nake fungal disease is a potential threat to massasauga populations and has been found in Michigan . T o determine the presence of snake fungal disease at our study sites, w e collected skin swab samples of all captured massasaugas . Out of 24 and 46 samples collected in 2015 and 201 6 , respectively, 3 individuals at good and optimal suitability sites tested positive for the pathogen in two counties where the pathogen had not yet been documented . iv For Stella. v ACKNOWLEDGMENTS I would like to thank Michigan State University and the Department of Fisheries and Wildlife for their financial support. I thank my committee members for their assistance and involvement with this research since the beginning . S pecifically, I thank my adviser Dr. Rique Campa for his unwavering support, availability, patience, and trust; Dr. Gary Roloff for his invaluable knowledge and expertise in many aspects of this research ; and Drs. J ean Tsao and Scott Winterstein for their constant support and interest in this work. I thank the Michigan Department of Natural Resources for financial support of this project, and volunteer visits to our field sites . I specifically thank Kristin Bissell, Amy Derosier, Mike Donnelly, Ray Fahlsing, Christine Hanaburgh, A licia Ihnken, D aniel Kennedy, K enneth Kesson, J ulie Oaks, Dr. D aniel ike Parker. From the U.S. Fish and Wildlife Service, I thank J ack Dingledine , S cott Hicks , Matt Ih nken, and Carrie Tansy . I thank the employees and veterinary staff of John Ball Zoo, Grand Rapids, Michigan, for their generous support and the allowing us the opportunity to collaborate . S pecifically, Dr. Ryan Colburn, B ill Flanagan, and Heather Teater our visits to the zoo were always a great experience! I also thank Dr. T ara Meyers Harrison for her assistance with our questions . I thank Mr. and Mrs. Weaver , Mr. Buck , Mr. and Mrs. Sbardella, Mr. and Mrs. Hubbell, and Carson and Associates, Inc , for acc ess to their properties . From the Nature Conservancy, I thank Sara Leavitt, C hris May, Rodolfo Villegas , and everyone at the Nancy Hand Field House. From the Michigan Audubon Society I thank Tom Funke and R achelle Roake . I thank Yu Man Lee from the Michigan Natural Features Inventory for her interest in this work. I thank Dr. M ike Dreslik and Dr. Matt Allender at the University of Illinois for their wealth of knowledge and vi support of this project . I thank Dr. Joshua Millspaugh for his suppo rt with the telemetry aspect of this research. I give a very special and heartfelt thanks to my field technicians T ricia Brockman, B rianna Brodowski, C aleb Burden, G regory Payter, and H annah Reynolds (Britz) . Without their interest in this work and the spe cies, their incredible patience, their willingness to push through dense vegetation and 3 feet of muck, and their ability to laugh when times were tough, none of this could have been accomplished. I thank Valerie Romanek and Alyssa Tarnowski for their inte rest and involvement as well. I thank my fellow lab mates, past and present, David Dressel and Chad Williamson, for volunteering and for listening to me carry on about this work in the office. I thank Amy Bleisch for her support, advice, and interest. I wi sh to give a special thanks to Robyn Bailey for hiring me all those years back I wish to thank my parents, Patricia and Stephen Zimmer, my sister Katie Kazakos, her husband Niko, and my neph ews Nick and Ben, my sister Laura Morrison, her husband Will, and my nieces Anna and Julia, for their love and support, and especially for their willingness to help me make more time to get a bit of work done! I thank my husband Michael Shaffer for his con stant support, patience, confidence, and avid interest in my work someone to explore the outdoors with ! L astly, most importantly, I thank my sweet Stella for putting up with me as I finished this work during the first year of her life a significant portion of this dissertation was written with you either sleeping in my lap or playing nearby, and your cheerful attitude kept me moving forward. vii PREFACE This dissertation was formatted following general guidelines for the Journal of Wildlife M anagement and Chapters 1, 2, and 3 are treated as stand - alone manuscripts; therefore, some repetition of sections occurs between chapters . Chapter 4 offers overall conclusions , recommended management guidelines , and future research directions for the massasauga. Information and findings relating to this work have been presented at local, state, regional, and national conferences, as well as through various publicly - accessible outlets ( Appendix A ). The eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ; hereafter massasauga) is one of three massasauga subspecies in North America and ranges from central New York and southern Ontario in the east to Iowa, Minnesota, and Missouri in the west, and includes southern Wisconsin, Michigan, Ill inois, Indiana, Ohio, and Pennsylvania . Massasaugas are known to be in decline rangewide due to habitat loss and degradation, human persecution, and population fragmentation due to land use change . As a result, massasaugas were designated as federally Thre atened as of October 31, 2016 . M assasauga populations in Michigan have been considered among the most viable rangewide due to a relatively high number of known populations, with adult survivorship estimates for the species ranging from 0.66 to 0.95 within the state (Jones et al. 2012) . Still, little information is available regarding survival of neonate and juvenile stages . Regarding the reduction in habitat quality through loss and degradation is the notion that natural processes, such as fires, grazing, and flooding events that historically maintained early successional vegetation communities and suitable habitat conditions for massasaugas no longer function within their historical range of variability. For example, wildfire once maintained suitable open herbaceous vegetation types (e.g., grasslands) by preventing succession, yet with viii land use change and human development, wildfire has been removed from the landscap e in many massasauga habitats rangewide, resulting in the loss of suitable habitats for the species ( et al. 2013 ) . Thus, habitat assessment and active management is necessary to identify and maintain suitable areas for remaining massasauga populati ons. Further, the ability to detect massasaugas in potential habitats is important for managers as they locate areas of conservation concern, warranting standardized, reproducible methods for detecting massasaugas within their habitats. Finally, the docume ntation of a fungal infection ( Ophidiomyces ophiodiicola ) among free - ranging snakes is a recently emerged and important potential threat to viability of massasauga populations (Allender et al. 2015) . In light of these issues, objectives for this project fo cus on eastern massasauga rattlesnake conservation, detectability, habitat use, and demographics in southern Michigan . These objectives were : 1) Validate the Bailey (2010) massasauga Habitat Suitability Index (HSI) model across a range of habitat conditions i n southern Michigan and explore other habitat model structures (Chapter 1) . 2) Determine factors affecting massasauga detectability and recommend a standardized survey methodology that is effective and efficient (Chapter 2) . 3) Estimate probability of occupancy of massasaugas for habitat s classified as low, (Chapter 1) . 4) Estimate population demographic rates (survival, fecundity) for adult, juvenile, and neonate massasaugas for a range of habitat condit ions in southern Michigan (Chapter 3) . 5) ( Appendix B ) . ix TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ........................ x i LIST OF FIGURES ................................ ................................ ................................ ..................... xiv CHAPTER 1: ASSESSING HABITAT SUITABILITY AND PREDICTING OCCUPANCY IN POTENTIAL EASTERN MA SSASAUGA RATTLESNAKE HABITAT ................................ ... 1 ABSTRACT ................................ ................................ ................................ ................................ . 1 INTRODUCTION ................................ ................................ ................................ ....................... 2 STUDY AREA ................................ ................................ ................................ ............................ 6 METHODS ................................ ................................ ................................ ................................ .. 7 Initial Site Assessment ................................ ................................ ................................ .............. 7 Capture and Marking ................................ ................................ ................................ ................ 8 Habitat Suitability Assessment ................................ ................................ ................................ . 9 Vegetation Sampling ................................ ................................ ................................ ............ 12 Habitat Suitability Index ................................ ................................ ................................ ...... 13 Resource S election Probability Function ................................ ................................ ................ 13 Predicting Occupancy ................................ ................................ ................................ ............. 14 RESULTS ................................ ................................ ................................ ................................ .. 15 Habitat Suitability ................................ ................................ ................................ ................... 15 Kernel Density Estimation ................................ ................................ ................................ ...... 23 Resource Selection Probability Function ................................ ................................ ................ 27 Predicted Occupancy ................................ ................................ ................................ .............. 35 DISCUSSION ................................ ................................ ................................ ............................ 36 MANAGEMENT IMPLICATIONS ................................ ................................ ......................... 40 ACKNOWLEDGMENTS ................................ ................................ ................................ ......... 41 CHAPTER 2: SURVEY METHODOLOGY FOR EASTERN MASSASAUGA RATTLESNAKES AND FACTORS INFLUENCING DETECTION ................................ ........ 42 ABSTRACT ................................ ................................ ................................ ............................... 42 INTRODUCTION ................................ ................................ ................................ ..................... 43 STUDY AREA ................................ ................................ ................................ .......................... 44 METHODS ................................ ................................ ................................ ................................ 45 Initial Site Assessment ................................ ................................ ................................ ............ 45 Capture and Marking ................................ ................................ ................................ .............. 46 Detection Surveys ................................ ................................ ................................ ................... 47 Detection Factors ................................ ................................ ................................ .................... 50 Analysis ................................ ................................ ................................ ................................ ... 54 RESULTS ................................ ................................ ................................ ................................ .. 54 DISCUSSION ................................ ................................ ................................ ............................ 62 MANAGEMENT IMPLICATIONS ................................ ................................ ......................... 65 ACKNOWLEDGMENTS ................................ ................................ ................................ ......... 65 x CHAPTER 3: EASTERN MASSASAUGA RATTLESNAKE SURVIVORSHIP, MOVEMENTS, AND REPRODUCTION ................................ ................................ ................... 67 ABSTRACT ................................ ................................ ................................ ............................... 67 INTRODUCTION ................................ ................................ ................................ ..................... 68 STUDY AREA ................................ ................................ ................................ .......................... 70 METHODS ................................ ................................ ................................ ................................ 71 Initial Site Assessment ................................ ................................ ................................ ............ 71 Capture and Marking ................................ ................................ ................................ .............. 71 Survivorship Estimates ................................ ................................ ................................ ........... 73 RESULTS ................................ ................................ ................................ ................................ .. 75 Estimated Daily and Period Survivorship ................................ ................................ ............... 75 Embryo Counts, Litter Size, and Dates of Parturition ................................ ............................ 77 Post - parturition Radiotelemetry of Neonates ................................ ................................ .......... 82 DISCUSSION ................................ ................................ ................................ ............................ 85 ACKNOWLEDGMENTS ................................ ................................ ................................ ......... 89 CHAPTER 4: CONCLUSIO NS AND M ANAGEMENT RECOMMENDA TIONS .................. 90 DISSERTATION OVERVIEW ................................ ................................ ................................ . 90 RECOMMENDATIONS FOR HABITAT MANAGEMENT AND POPULATION ASSESSMENT ................................ ................................ ................................ .......................... 92 Identifying Potential Massasauga Sites ................................ ................................ .................. 92 Assessing Habitat Suitability ................................ ................................ ................................ .. 93 Detecting Massasaugas ................................ ................................ ................................ ........... 93 Habitat Management ................................ ................................ ................................ ............... 94 FUTURE RESEARCH ................................ ................................ ................................ .............. 96 APPENDICES ................................ ................................ ................................ .............................. 98 APPENDIX A: Dissemination of Findings and Related Information ................................ ....... 99 APPENDIX B: Snake Fungal Disease Sampling Results ................................ ......................... 102 APPENDIX C: Study Site Information and Maps ................................ ................................ ... 111 APPENDIX D: Documentation of All Encountered Eastern Massasaug a Rattlesnakes .......... 115 APPENDIX E: Snake Fungal Disease Disinfection Protocol ................................ .................. 124 APPENDIX F: Study Site and Detection Survey Subsite Naming Description ....................... 127 APPENDIX G: R Code ................................ ................................ ................................ ............. 130 APPENDIX H: Documentation of Three Mortality Events ................................ ..................... 158 APPENDIX I: Data Sheets ................................ ................................ ................................ ....... 160 LITERATURE CITED ................................ ................................ ................................ ................ 165 xi LIST OF TABLES Table 1.1. Habitat suitability index (HSI) scores and the associated suitability index (SI) scores for measured habitat variables c at each study site by sampling date across southern Michigan. Scores based on Bailey (2010) HSI model for eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ). Sites where massasaugas were located during 2015 or 2016 field seasons are marked wi th an asterisk (*) ................................ ................................ ................................ ..... 17 Table 1.2. Habitat suitability index (HSI) and associated suitability index (SI) scores for habitat variables c measured at eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) locations by site throughout southern Michigan, 2015 - 2016 ................................ ................................ ....... 20 Table 1.3. Welch two sample t - test results comparing vegetation for early (May - June) and late seasons (July - August) in southern Michigan. Vegetation variables correspond to habitat components for eastern massasauga rattlesnakes ( Sistrurus ca tenatus catenatus ). Confidence intervals (LCI = lower 95% confidence interval, UCI = upper 95% confidence interval) portray range of values bounding the difference of early - and late - sample means. Degrees of freedom (DF) based on number of vegetation sampl es collected for each site. An asterisk (*) indicates significance ( ................................ ................................ ................................ .................. 22 Table 1.4. Consistent AIC (CAIC) tables ranking the resource selection probability functions for each % kernel home range at 4 study sites in southern Michigan, 2015 and 2016 . Only models that converged are included ................................ ................................ ................................ .......... 25 Table 1.5. Number of snake locations (n s,total ) and vegetation sampling plots ( n v,total ) in each 20 ha study site, and corresponding kernel home range with the number of snake locations ( n s,hr ) and vegetation s ampling plots ( n v,hr ) within, southern Michigan, 2015 - 2016 ................................ ..... 26 Table 1.6. Comparative summary of vegetation structure (mean, minimum, maximum observed within the 5 x 20 m vegetation sampling plot) in used versus unused (i.e., available) areas based on the top - ranking resource selection probability function home range kernels (Table 1.4) for eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) at 4 20 - ha sites throughout southern Michigan in 2015 and 2016. Sample size ( n ) indicates the number of v egetation samples that the values are based on within use or available areas ................................ .............. 30 Table 1.7. Generalized linear model p arameter estimates and associated standard errors (SE), z scores, and p values for relationships between occupancy probabilities and easte rn massasauga rattlesnake ( Sistrurus catenatus catenatus ) habitat suitability index (HSI) scores (Bissell 2006, Bailey 2010). Includes full model, thermal variables (LHC, DHC, STDS, and BAT), and the landscape - level variables (AEDU and AEDW) ................................ ................................ ............ 36 xii Table 2.1. Administrative information, habitat quality, and number of telemetered eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) by stage (adult, juvenile) and sex ( m ale, f emale) for 20 ha sites in southern Michigan, USA, for the sites where detecti on surveys were carried out in 2016 ................................ ................................ ................................ ........................ 49 Table 2.2. All detection covariates (environmental, habitat, searcher, and survey) considered and measured during 2015 and 2016, quantified to evaluate their importance when detecting eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) in their habitats throughout southern Michigan ................................ ................................ ................................ ................................ ....... 51 Table 2.3. The 27 20 - ha study sites where detection surveys were carried out for detection of the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) in 2015 and 2016 throughout southern Michigan ................................ ................................ ................................ ........................ 55 Table 2.4. Survey count and massasauga detection for occupied subsites (2 ha) in the 4 sites (20 ha) where the final detection surveys were conducted for eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ), southern Michigan, USA, in 2016 ................................ ............. 56 Table 2.5. AIC table for models found to have the greatest influence on detectability of eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) at 2 ha survey subsites in southern Michigan, USA. All models are statistically supported ( p ................................ ............... 57 Table 2.6. AICc table for all 39 covariates considered (plus the null model) to have a potential influence on detection probability for eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ), southern Michiga n, USA, 2016 ................................ ................................ .................. 58 Table 3.1. Number of adult, juvenile, and neonate eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) used to estimate survivorship throughout southern Michigan in 2015 and 2016. The habitat suitability index (HSI) represents a suitability score assigned to each site based on the vegetation composition and structure within the site (Chapter 1) ................................ ..... 75 Table 3.2. Daily survivorship estimates for the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) at 6 20 - ha study sites throughout southern Michigan in 2015 and 2016 .... 76 Table 3.3. Period (137 days) survivorship estimates for the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) at 6 20 - ha study sites throughout southern Michigan in 2015 and 2016 ................................ ................................ ................................ ................................ ............... 76 Table 3.4 . Number of embryos, the method by which the count was obtained, and observed neonates following parturition for all gravid female eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) found to be gravid during the 2015 and 2016 field seasons at our study sites throughout southern Michigan. This table does not include litter counts for instances where we were unable to obtain an embryo count for the female and does not include litters found with no associated adult female as this limited our abili ty to make comparisons between pre - and post - parturition litter size estimates ................................ ................................ ................................ ...... 78 xiii Table 3.5. Radiotracking information for all telemetered neonate eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) obtained during the 2016 field season in sou thern Michigan. All neonates were fitted with external transmitters weighing less than 5% of their body weight (Lentini et al. 2011). Exact date of parturition is unknown for all neonates. Horizontal lines within the table delineate three different litters based on proximity and date of location (i.e., one litter at NE_L1, and two litters at SW_M2) ................................ ................................ .................. 83 Table B.1. All eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) sampled in 2015 and 2016 throughout southern Michigan for snake fungal disease ( Ophidiomyces ophiodiicola ) and outcome of testing for each (positive [+] , negative [ - ] ). Multiple samples were collected when an individual exhibited clinical signs of infection by the fungal pathogen ........................ 106 Table C.1. The 27 20 - ha study sites , locations, and property names that were included in the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) research throughout southern Michigan in 2015 and 2016 ................................ ................................ ................................ ......... 111 Table C.2. Site information for the 27 20 - ha study sites where field work was carried out during the 2015 and 2016 field seasons for the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) research throughout southern Michigan ................................ ................................ ..... 112 Table D.1. E astern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) encountered and marked during this study ( 2015 and 2016 ) throughout southern Michigan, and fate of the individual at final date of the study ................................ ................................ .............................. 115 Table D.2. Number of eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) encounters of either unidentified or unknow n indi viduals observed during 2015 and 2016 field seasons throughout southern Michigan ................................ ................................ ................................ ..... 119 Table D.3. Weights (g) and lengths (snout - to - rattlesnakes ( Sistrurus catenatus catenatus ) captured during the 2015 and 2016 field s easons in southern Michigan. Additional metadata associated with each massasauga are provided in Appendix D: Table D.1 ................................ ................................ ................................ ................ 120 Table F.1 : Site code names for the 20 - ha study sites included in the 2015 and 2016 eastern massasauga rattlesnake ( Sistrurus catenat us catenatus ) field work in southern Michigan ......... 128 Table F.2: Site and subsite code names (original and final) for 20 - ha study sites and respective 2 - ha detection survey subsites included in the 2015 and 2016 eastern massasauga rattlesnake ( Sistrurus caten atus catenatus ) field work throughout southern Michigan ................................ . 129 xiv LIST OF FIGURES Figure 1.1. Production functions representing suitability indices for habitat components comprising the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) habitat su itability index model as defined by Bissell (2006), as amended by Bailey (2010), for southern Michigan ................................ ................................ ................................ ................................ ....................... 11 Figure 1.2. Habitat suitability production functions for components of the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) habitat sui tability index model (Bissell 2006, Bailey 2010) for southern Michigan. The blue line overlaying each production function represents range of suitabilities observed across 27 field sites during the active snake season throughout southern Michigan, 2015 - 201 6. Points marked with x and represent median and mean values, respectively (for 2c, observed mean = 1179.38 stems/ha and median = 930 stems/ha) ............... 16 Figure 1.3. Eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) % home range kernels (50 - 95 %), snake locations (red dots), and vegetation sampling locations (green dots) for study sites NE_L1 (1.5a), SE_H3 (1.5b), SW_H2 (1.5c), and SW_M2 (1.5d), in southern Michigan in 2015 and 2016 ................................ ................................ ................................ .......... 24 Figure 1.4. Marginal effects plots showing the relationship between availability of live herbaceous cover (0 - 100% transect coverage out of 0 - 10 meters as displayed in the x - axis , covering the range of values observed within the site ) and absolute probability of use by eastern massasauga rattlesnake s ( Sistrurus catenatus catenatus ) for the top - ranking home range kernels for sites NE_L1 (60% kernel), SE_H3 (85% kernel), SW_H2 (90% kernel), and SW_M2 (80% kernel) in southern Michigan (Table 1.4) ................................ ................................ ..................... 28 Figure 1.5. Marginal effects plots showing the relationship betwee n availability of dead herbaceous cover (0 - 100% transect coverage out of 0 - 10 meters as displayed in the x - axis , covering the range of values observed within the site ) and absolute probability of use by eastern massasauga rattlesnake s ( Sistrurus catenatus catenatus ) for the top - ranking home range kernels for sites NE_L1 (60% kernel), SE_H3 (85% kernel), SW_H2 (90% kernel), and SW_M2 (80% kernel) in southern Michigan (Table 1.4) ................................ ................................ ..................... 29 Figure 1.6: Marginal effects plots showing the relationship between the numb er of woody stems (covering the range of values observed within the site) and absolute probability of use by eastern massasauga rattlesnake s ( Sistrurus catenatus catenatus ) for the top - ranking home range kernels for sites NE_L1 (60% kernel), SE_H3 (85% kernel), SW_H2 (90% kernel), and SW_M2 (80% kernel) in southern Michigan (Table 1.4) .... 32 Figure 1.7: Marginal effects plots showing the relationship between the average diameter at breast height (cm) of trees and shrubs m in height within the 20 x 5 m plot (covering the range of values observed within the site) and absolute probability of use by eastern massasauga rattlesnake s ( Sistrurus catenatus catenatus ) for the top - ranking home range kernels for sites NE_L1 (60% kernel), SE_H3 (85% kernel), SW_H2 (90% kernel), and SW_M2 (80% kernel) in southern Michigan (Table 1.4) ................................ ................................ ................................ ...... 34 xv Figure 1.8: Predicted occupancy probabilities for eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) habitat suitability index (HSI) s cores (Bissell 2006, Bailey 2010). Includes full model, thermal variables (LHC, DHC, STDS, and BAT), and the landscape - level variables (AEDU and AEDW) ................................ ................................ ................................ ..................... 35 Figure 2.1. Schematic of eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) detection su rvey methodology conducted at 2 ha subsites (n=4) in southern Michigan. Filled - in numbered points indicate the pre - determined GPS locations along the subsite boundary used for navigation. Searchers simultaneously started at opposite ends ( i.e., searcher 1: points 1 to 14; searcher 2: points 14 to 1 ) and loosely traversed each transect searching for massasaugas ......... 48 Figure 2.2. Predicted detection probabilities for eastern massasauga rattlesnakes by minutes spent actively searching (i.e., total length of search time in minutes for a given survey), at 2 ha survey sites, southern Michigan, USA, during the active season 2016. Detection probability is indicated by the solid line; 95% confidence intervals are indicated by the dotted lines .............. 60 Figure 2.3. Predic ted detection probabilities for massasauga rattlesnakes across a range of minimum air temperatures (C) at 2 ha survey sites, southern Michigan, USA, during the active season 2016. Detection probability is indicated by the solid line; 95% confidence interva ls are indicated by the dotted lines ................................ ................................ ................................ ......... 61 Figure 3.1. Film radiograph image (photographed) taken on 30 Jun 2015 showing 16 embryos (R. Colburn, personal communication) in female eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) from site NE_L1 (PIT tag # 836560350 ; Appendix D ) in southern Michigan during the 201 5 field season . Parturition for this female occurred between 28 August and 2 September 2015. On 2 September 2015, 15 neonates were observed in close proximity to this female. Lateral view . Image credits: John Ball Zoo, Grand Rapids, MI; R. Colburn .................. 79 Figure 3.2. Digital radiographs taken on 29 July 2016 showing 18 embryos (R. Colburn, personal communication, 2016) in female eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) fr om site NE_L1 ( lateral view [2a], dorsal view [2b]; PIT tag # 840543607; Appendix D ) in southern Michigan during the 2016 field season . Parturition for this female occurred between 7 and 15 August 2016. On 15 August 2016, 15 neonates were observed in close proximity to this female. Image credits: John Ball Zoo, Grand Rapids, MI; R. Colburn, B. Flanagan ................... 80 Figure C.1. The 27 20 - ha study sites, represented by yellow dots, throughout southern Michigan where field work was conducted during 2015 and 2016 field seasons. Counties containing field sites included Barry, Calhoun, Jackson, Lenawee, Oakland, Washtenaw, and Livingston. These sites were selected based on historical presence of eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ; MNFI 2014) ................................ ................................ ............................... 114 xvi Figure D.1. P ossible hypo - melanism (J. Harding, personal communication, 2018 ) observed in two neonate eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) in Calhoun County in southern Michigan in 2016. Figure D. 1a shows a close - up of head. Fi gure D.1b shows a neonate exhibiting possible hypo - melanism beside a normal - colored neonate, presumed to be a litter mate, for comparison ................................ ................................ ................................ ........... 122 Figure D.2. Apparent deformity observed in one neonate eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) in Calhoun County in southern Michigan in 2016. Figure D. 2a: i ndividual did not move normally; a rolling motion was used for locomotion, the neck appeared permanently . Figure D.2b shows the individual in a clear plastic bag, which were used for weighing small juveniles and neonates ................................ ...................... 123 Figure E.1. D isinfection checklist developed for this project used before leaving each field site to prevent the spread of snake fungal disease ( Ophidiomyces ophiodiicola ) between sites i n southern Michigan during the 2015 and 2015 field seasons (see Appendix B) ........................... 126 Figure I.1. Data sheet used to record all pertinent information for captured eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) during the 2015 and 2016 field sea sons throughout southern Michigan ................................ ................................ ................................ ....................... 161 Figure I.2. Data sheet used to record all vegetation characteristics measured during the 2015 and 2016 field seasons for assessing eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) habitat throughout sout hern Michigan based on the Bailey (2010) habitat suitability index model ................................ ................................ ................................ ................................ ...................... 162 Figure I.3. Detection survey data sheet used during eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) detection surveys (see Chapter 2) conducted throughout southern Michigan. developed during the 2015 and 2015 field seasons ................................ ................................ ..... 163 Figure I.4. Postoperative monitoring data sheet (Bailey 2010) used following radiotransmitter implantation surgeries for eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) captur ed throughout southern Michigan during the 2015 and 2016 field seasons ...................... 164 1 CHAPTER 1: ASSESSING HABITAT SUITABILITY AND PREDICTING OCCU PANCY IN POTENTIAL EASTERN MA SSASAUGA RATTLESNAKE HABITAT Stephanie A. Shaffer, Henry Campa, III, Gary Roloff, Ryan Colburn, Daniel Kennedy ABSTRACT Eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) are a federally Threatened species, due in part to degradation and fragmentation of habitats. Where concerns for habitat exist, the ability to ide ntify and rank habitat suitability will inform management. Our goal was to evaluate a habitat suitability index (HSI) model for massasaugas in southern Michigan across varying habitat qualities. We a priori applied the HSI model to 27 20 - ha sites where mas sasauga presence was confirmed within the past 20 years (9 were re - confirmed during this study). We used a stratified random sample to collect vegetation attributes at each site and used these attributes to calculate HSI scores. Habitat suitability index s cores ranged from 0.21 (low quality) to 0.95 (high quality), and we found massasaugas at sites that represented medium to high quality scores, suggesting that the relationship between HSI score and massasauga occupancy was unreliable. However, HSI scores based on vegetation data collected at massasauga locations performed better, with HSIs ranging from 0.60 1.00 HSI, suggesting that the HSI model is best suited for localized scales (i.e., the 20 x 5 m sampling plot). On a subset of sites (n = 4) we mappe d kernel use areas from massasauga locations collected through visual encounter and radio telemetry and used a resource selection probability function (RSPF) to verify shape of the suitability index relationships. The RSPF corroborated the importance of th e habitat variables defined by the habitat suitability index model for massasaugas, illustrating positive relationships between probability of use and % live and dead herbaceous cover, and negative relationships between probability of use and DBH and numbe height. Finally, based on HSI scores and our records of confirmed occupancy by site, we 2 estimated occupancy for sites of varying suitability. For sites approaching maximum suitability (HSI = 1), predicted occupancy by massasaugas e xceeded 0.5 (i.e., we can expect over 50% of sites at this level of suitability to be occupied). Alternately, for minimally suitable sites (HSI scores approaching 0), predicted occupancy was < 0.2. These results link HSI scores to a population state, in th is case occupancy probability, and illustrate that sites with a relatively low habitat suitability index can be occupied by massasaugas. Greatest habitat management benefits include improving these low suitability sites to facilitate population expansion . H abitat assessments using the variables described here may be used by managers as a guide to determine current and potential habitats and as a means of determining the amount and type of habitat management needed to improve habitats for conserving the mass asauga. INTRODUCTION Populations of eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ; hereafter, massasauga) are declining range wide (e.g., Szymanski 1998, Shoemaker and Gibbs 2010). In 2016, the U.S. Fish and Wildlife Service listed massasa ugas as federally Threatened , partly because of habitat degradation and fragmentation (USFWS 2016). The range of massasaugas extends from central New York and southern Ontario in the east to Iowa, Minnesota, and Missouri in the west, and includes southern Wisconsin, Michigan, Illinois, Indiana, Ohio, and Pennsylvania (USFWS 2016). Considering the federal Threatened status, decline and degradation of massasauga habitat rangewide, and importance of identifying and conserving remaining habitats for massasaugas , it is imperative to understand relationships between habitats and population states. The ability to quantify quality, or suitability, of habitats is an invaluable tool for managers as they assess and manage sites occupied by massasaugas. Massasauga habi tat is generally characterized by structural attributes that include limited canopy cover and early - 3 Kodrich 1982:170) in upland and wetland vegetation types rangewide (e.g., Johnson 199 5, Moore and Gillingham 2006, Durbian et al. 2008, McCluskey et al. 2018). Much available literature describes vegetation types where massasaugas occur as lowland bogs, peatlands, and fens (e.g., Johnson 1995, Marshall et al. 2006), and upland open, dry, h erbaceous grasslands, and early successional deciduous vegetation types (e.g., Durbian et al. 2008, McCluskey et al. 2018). Although highly varied, predominant plant species include sedges ( Carex spp.), dogwoods Cornus spp.), rushes ( Juncus spp.), and catt ails ( Typha spp.; Marshall et al. 2006, Moore and Gillingham 2006, Durbian et al. 2008). Few studies quantitatively described specific structural aspects of habitat (e.g., Bissell 2006, Moore and Gillingham 2006, Bailey 2010). For a massasauga population i n southeastern Michigan, Moore and Gillingham (2006:750) reported height and primarily used areas with an open canopy and high amounts of groundcover vegetation rangi ng from 0.5 1.5 m in height. Similarly, for a southwestern Michigan population, Bissell (2006) identified optimally suitable ranges of specific habitat components for the massasauga including live herbaceous cover (optimal values ranging from 60 - 100% cov er), dead herbaceous cover (51.5 - - 60 - 0.116 m 2 /ha) . Such vegetation structures, characteristic of open herbaceous and early success ional vegetation types , are consistent with massasauga needs for thermoregulation via basking or hiding during the active season, and are important aspects of reproduction and digestion (e.g., Foster et al. 2009, Harvey and Weatherhead 2010). Habitat suita bility index (HSI) models define and evaluate suitability of a given area for species of interest to guide habitat management. These models are a tool based on quantification 4 of habitat structure and composition (USFWS 1981). Numerous HSI models exist for multiple taxa for game and non - game species, including barred owl ( Strix varia ; Allen 1987), eastern wild turkey ( Meleagris gallopavo sylvestris ; Schroeder 1985), American alligator ( Alligator mississippiensis ; Newsom 1987), and sea cucumber ( Apostichopus japonicus ; Zhang et al. 2017). Habitat suitability index models have been used to guide habitat management, determine habitat use by species (e.g., Dussault et al. 2006), and can be implemented without observing the species of interest within the habitat an especially important characteristic for cryptic species such as massasaugas. To confirm multi - site applicability of these models for a range of habitat conditions, model validation is an important step (Roloff and Kernohan 2006). For example, Soniat an d Brody (1988) validated an HSI model for American oyster ( Crassostrea virginica ; Cake 1983) by correlating oyster density with HSI values across multiple sites. Based on field studies and observation of environmental conditions relevant to the model (i.e. , the model variables), Soniat and Brody (1988) simplified the Cake (1983) model by including negative population effects on the species. Soniat and Brody (1988) noted that while HSI models are intended for multi - site use, location - specific modifications t o the model may be necessary. Further, accounting for uncertainty arising from multi - site application of HSI models will aid in interpretation and management (e.g., Zajac et al. 2015). Previous work in southwestern Michigan quantified habitat suitability requirements for massasaugas (Bissell 2006, Bailey 2010), and resource selection in managed landscapes (Bailey et al. 2012). Bissell (2006) developed a massasauga HSI model for southwestern Michigan based on vegetation samples collected in 2004 and 2005 i n areas used by telemetered massasaugas. This model was later modified by Bailey (2010) based on vegetation sampling and quantification of massasauga habitat selection patterns and movements for the same population in 2008 and 5 2009. This model relies on ve getation composition and structural attributes (e.g., availability of live and dead herbaceous cover, woody stem density, and basal area) at multiple spatial scales (e.g., the structure of microhabitat selected by snakes, landscape - level variables includin g the percent of the site made up of early successional deciduous upland or lowland vegetation types). These habitat elements related to several population fitness measures of massasaugas (e.g., survivorship, reproductive rates, home range size; Bailey 201 0). The revised model was built from observations in relatively high quality habitat, so to appropriately validate this model a range of habitat conditions across southern Michigan should be assessed (e.g., Roloff and Kernohan 1999). Having a validated HSI model would aid natural resource managers in planning and implementing more effective habitat management practices to maintain or enhance massasauga habitat. Resource selection can be used to determine habitat attributes selected by massasaugas to illustr ate whether specific habitat components are used disproportionately (i.e., visited more frequently than others) among what is available within a given area (Lele and Keim 2006). Validation of HSI models involves correlating measures of population states (e.g., occupancy, abundance, use, survival) with model outputs (Roloff and Kernohan 2006). Occupancy probability is defined as the likelihood a site is occupied by a species (MacKenzie et al. 2006:26), has been used to monitor populations, and is useful f or cryptic or rare species, such as massasaugas, when capture rates can be low (Durso et al. 2011). Durso et al. (2011) determined occupancy probability of 7 aquatic snake species within wetland sites based on site characteristics (e.g., wetland permanence , distance from floodplain) and availability of prey. The ability to predict occupancy based on vegetation composition and structure can allow managers to identify important sites for conservation by estimating an occupancy likelihood for a given 6 area. Hab itat use by a species may also be used as a metric for validation of HSI models. For example, Thomasma et al. (1991) f ound that fishers ( Martes pennanti ) used high - suitability areas more frequently than available lower - suitability areas and stated that tho ugh improvements on the model were needed, a positive relationship was documented between fisher habitat 1991 :295) . Using home range kernel estimation and RSPF (Lele et al. 2006), we can delineate areas used by m assasaugas can compare vegetation structures among use and available areas. Our objective was to validate the Bissell (2006; as amended by Bailey 2010) massasauga HSI model using occupancy and use. We measured vegetation to populate the HSI model and surveyed for massasaugas, comparing occupancy to HSI score. On a subset of sites, we estimated massasauga use probabilities and correlated these values to HSI vegetation measures. Collectively, these methods allowed us to quantify the suitability of a giv en area for massasaugas based on specific habitat components, to identify and define use by massasaugas of those components, and to relate HSI score to occupancy probability at the site level. STUDY AREA Southern Michigan is a temperate region with moderat e spring - summer (i.e., May Aug) temperatures that ranged from approximately 10 32° C during our study (2015 to 2016 in Jackson, MI; NOAA - NWS 2017). Physiography of our research counties (Barry, Calhoun, Jackson, Lenawee, Livingston, Oakland, and Washtenaw) consists of glacially deposited outwash plains, moraines, till plains, and lacustrine plains (Striker and Harmon 1961, Engberg and Austin 1974, Engel 1977, McLeese 1981, Feenstra 1982, Thoen 1990, Tardy 1997). Soils were well drained or poorly drained and loamy with interspersed sandy - loam, loamy sand, or mucky soil types (USDA - NRCS 2017). 7 We identified 27 sites based on confirmed reports of massasaugas within the last 25 years (Michigan Natural Heritage Database [MNFI] 2014). These sites represented a ran ge of vegetation types and HSI qualities for massasaugas on private and public lands (Appendix C: Table C. 1). Of the 27 sites, 11 occurred on private lands owned by citizens, non - profit conservation groups, or corporations , while the remaining 16 sites occ urred on public lands (Appendix C). Sites were 20 ha in size (the maximum 95% fixed kernel home range for a massasauga in southern Michigan; Bissell 2006) and (the maximum distance moved by an individual massasauga in a single season in southe rn Michigan; Bissell 2006). METHODS To assess massasauga habitat suitability, we used the HSI model (Bissell 2006; as modified by Bailey 2010) at our 27 20 - ha study sites. We also predicted massasauga use of habitat structures defined by the HSI model usi ng a resource selection probability function. We used a use versus available design ( Lele and Keim 2006 ), where documented massasauga locations represented used areas and points outside the 95% fixed kernel use area represented available areas. Lastly, usi ng the occupancy status and HSI score of each of our sites, we determined how probability of occupancy varied by HSI score. Initial Site Assessment We coarsely classified habitat quality a priori into low, medium, and high based on the proportion of suit able and unsuitable vegetation types for massasaugas within each 20 - ha site. The purpose of this classification was to ensure inclusion of sites with varying habitat structure and composition throughout southern Michigan. We used the 2006 National Land Cov er Database (hereafter NLDC; MRLC 2015; Appendix C: Table C. 2) and considered grassland/herbaceous openings, woody wetlands, and emergent herbaceous wetlands as suitable types for massasaugas. We considered open water, development, deciduous forest, evergr een 8 forest, mixed forest, shrub/scrub, pasture/hay, and cultivated crops as non - suitable. We based Management Analysis Program (IFMAP; MDNR 2001) land cover data. We de signated sites with <40% in suitable vegetation types as low site quality, between 40 60% suitable vegetation types as medium, and >60% suitable vegetation types as high quality (Appendix C: Table C. 1). Capture and Marking From May through August 2015 and 2016, we located and captured massasaugas via random encounter surveys within and around study sites. We checked all captured massasaugas for a passive integrated transponder (PIT) tag (AVID; Norco, CA), and newly captured individuals were injected with a PIT tag subcutaneously into the dorsal region approximately 4 6 cm caudal to the cloaca (e.g., Bissell 2006). We transported massasaugas weighing >100 g to a veterinary clinic for surgical implantation of a radio transmitter. While in captivity under vete rinary observation, massasaugas were held individually in locked glass - front reptile aquariums (model S24T; Neodesha Plastics Inc., Neodesha, KS), and provided a place to hide, a dish of water, and sheets of paper towel for additional hiding cover and to a bsorb liquids (e.g., spilled water, defecation). Veterinary staff surgically implanted massasaugas with one of three radio transmitters of varying weights, depending on body mass (models R1515, 7 g, and R1680, 3.1 g, Advanced Telemetry Systems [ATS], Inc. , Isanti, MN; model SB - 2, 5.3 g, Holohil Systems Ltd., Ontario, Canada). A licensed veterinarian from the John Ball Zoological Park veterinary clinic in Grand Rapids, Michigan, followed transmitter implantation procedures described by Bailey (2010) and Ba iley et al. (2011). The combined weight of the implanted radio transmitter and the PIT tag did not exceed 5% of body weight to minimize marking effects (Lentini et al. 2011). Following d with a transmitter each 9 year (further, no more than one of these three was a gravid female). Within 7 days of capture (in most cases 3 4 days), implanted massasaugas were examined and approved for release by the veterinarian and subsequently released at their original capture location (Appendix D). We used external transmitters attached with cyanoacrylate - for massasaugas > 100 g lo cated after the final day of surgery (i.e., after July 31 of both years to ensure sufficient healing of the surgical site prior to hibernation). We also used external transmitters for neonate and juvenile massasaugas weighing < 100 g. These transmitters in cluded ATS model R1635 (0.75 g), and ATS model A2414 (0.30 g), and were <5% of the body weight in combination with the PIT tag (Lentini et al. 2011). Snake capturing, handling and housing methods were approved by Michigan State nimal Care and Use Committee (IACUC), protocol 05/15 - 087 - 00, and by the Michigan Department of Natural Resources (MDNR) Fisheries Division (Scientific Collectors Permit #PR8114) . Public land access was approved by the Michigan Department of Natural Resourc es Permit to Use State Land #PR1136 - 1. Private land access was arranged directly with landowners prior to the start of any research activity at that property. In accordance with the protocol and requirements outlined in our IACUC approval and MDNR Scientif ic Collectors Permit, boots and all gear used in the field were disinfected using a 10% bleach solution following each day in the field and before moving among field sites to reduce potential spread of pathogens (Appendix E). Habitat Suitability Assessment The Bailey (2010) HSI model is based on percent live herbaceous cover (LHC; Figure 1. 1a), percent dead herbaceous cover (DHC; Figure 1. 1b), stem density of trees and shrubs > 3 m in height (SDTS; Figure 1. 1c), and basal area of trees >3 m in height (BAT; Figure 1. 1d). Each 10 of these variables was measured in the field following vegetation sampling techniques detailed by Bailey (2010). Percent area in early successional deciduous uplands (AEDU; Figure 1. 1e), and percent area in early successional deciduous w etlands (AEDW; Figure 1. 1f) , two additional variables, were quantified using IFMAP land cover data ( MDNR 2001) in Arc GIS 10.2 for each site (Bissell 2006, Bailey 2010). The 6 variables were quantified for each 20 ha study site and used to determine a suit ability index ranging from 0 to 1, with 1 being most suitable (Figure 1. 1). We note that our use of variable BAT differs from variable ADT in Bailey (2010) reported an incorrect range of optimal suitability for ADT, and the error was not acc ounted for in either Bissell (2006) or Bailey (2010) but was communicated to us via Bailey (personal communication, 11 1. 1a 1. 1c 1. 1e 1. 1b 1. 1d 1. 1f Figure 1. 1. Production functions representing suitability indices for habitat components comprising the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) habitat suitability index model as d efined by Bissell (2006), as amended by Bailey (2010), for southern Michigan. 0.00 0.20 0.40 0.60 0.80 1.00 0 20 40 60 80 100 Suitability Index Percent Live Herbaceous Cover 0.00 0.20 0.40 0.60 0.80 1.00 0 200 400 600 800 Suitability Index Stem Density of Trees and Shrubs >3 m (stems/ha) 0.00 0.20 0.40 0.60 0.80 1.00 0 50 100 Suitability Index Percent area in early successional deciduous upland 0.00 0.20 0.40 0.60 0.80 1.00 0 20 40 60 80 100 Suitability Index Percent Dead Herbaceous Cover 0.00 0.20 0.40 0.60 0.80 1.00 0 20 40 Suitability Index Basal Area of trees >3 m (m 2 /ha) 0.00 0.20 0.40 0.60 0.80 1.00 0 50 100 Suitability Index Percent area in early successional deciduous wetland 12 Vegetation Sampling During the 2015 and 2016 field seasons, we selected 10 - 12 stratified random points for vegetation sampling at each site. We stratified sample locations based on amount of each vegetation type (i.e., NLCD; MRLC 2015) to ensure proportional representation of vegetation types within a study site. For example, if 3 vegetation types occurred in equal amounts within the site, 4 vegetation samples were randomly selected within each type. If 3 vegetation types represented 0.2, 0.2, and 0.6 of the area of a site, 2, 2, and 6 vegetation sampling points were randomly selected within each type, respectively. At each point we used a north - south 20 m line - intercept to estimate suitability indices LHC and DHC (Bailey 2010). We placed a 5 x 20 m plot eastward along each l ine - intercept (the line intercept being the west boundary of the plot) and counted and measured the diameter at breast height (DBH) of woody stems >3m in height. Diameter at breast height was measured for every tree within the plot up to 9 stems; if > 9 st ems were present the three stems closest to the 0, 10, and 20 m marks on the line - intercept were measured. These stem measurements were used to calculate SDTS and BAT. In addition to randomly selected sampling points throughout each study site, we conducte d vegetation sampling at locations of newly found massasaugas and at telemetered massasauga locations when they moved into a new vegetation type. At each site, we collected early season (i.e., May, June) and late season samples (i.e., July, August) to desc ribe vegetation variability throughout the active season, and different random points were selected for each year. We only measured LHC and DHC during the late season sample, as we assumed that variables SDTS and BAT would remain unchanged within a span of 2 - 3 months (absent habitat management). To compare early season (May - June, 2015 and 2016) and late season (July - August, 2015 and 2016) samples for LHC and DHC, we used a Welch two sample t - test ( = 0.05). 13 We assessed AEDU and AEDW for each site using Arc GIS (10.2) and IFMAP (MDNR 2001). Using IFMAP vegetation classes, we considered early successional deciduous upland (i.e., AEDU) as forage crop, herbaceous openland, row crop, and upland shrub, whereas early successional deciduous wetland (i.e., AEDW) incl uded emergent wetland, floating aquatic, lowland shrub, and mixed non - forest wetland (MDNR 2001). Habitat Suitability Index We calculated SI scores for each habitat component (LHC, DHC, SDTS, BAT, AEDU, and AEDW) and a composite HSI for each site (Bissell 2006, Bailey (2010): We calculated HSI for both early and late seasons in 2015 and 2016, and averaged scores across both years and seasons to obtain a single overall HSI for each site. Sites were subsequently ranked based on overall HSI scores as: 0.00 to 0.25 = poor suitability; 0.26 to 0.50 = marginal suitability; 0.51 - 0.75 = good suitability; 0.76 - 1.00 = optimal suitability (Bailey 2010). Our use of the HSI model. Resource Selection Probability Function To assess massasauga use of vegetation structural characteristics identified as habitat components by the HSI model, we quantified use ve rsus availability using a resource selection 2006, Lele et al. 2017) in R Studio. We defined use areas as a home range kernel from snake locations, compared to availab le samples outside the kernel. We generated home range kernels locations recorded in 2015 and 2016 within and around our study sites. We only used sites with ations for this analysis ( Appendix F) . We delineated 25, 50, 60, 70, 80, 85, 90, and 14 assumes the utilization distribution is bivariate normal around each location (Ca lenge 2015, 2018). Because our study sites varied in the relative proportions of suitable versus unsuitable vegetation types, the range of kernel sizes allowed us to examine how these variations influenced patterns of use at sites of varying quality (i.e., what is used among what is available at these varying levels), and how vegetation structure varied among these kernels. The smallest kernels (i.e., 25, 50%) allowed us to examine core - use areas. We clipped these kernels to the extent of a 20 - ha study site and overlaid the late season vegetation sampling points that represented vegetation at its structural peak during the active snake season (a majority [>77%] of massasauga locations were collected during July through August or later). The resource selectio n probability quantify use versus availability (Lele and Keim 2006). Habitat variables included live herbaceous cover, dead herbaceous cover, number of woody stems > 3 m in height, and DBH of woody stems >3 m in height (mean value at each sampling point). Study sites included in this analysis were those for which we obtained >30 massasauga locations throughout 2015 and 2016. These sites included: SW_H2 (Otis Audubon Sa nctuary), SW_M2 (Baker Audubon Sanctuary North), SE_H3 (Ives Road Fen North), and NE_L1 (Seven Lakes State Park East; Appendix F). Predicting Occupancy from HSI Components To examine the relationship between components of the HSI model and massasauga occup ancy, we used logistic regression in R Studio using the glm() function. We previously demonstrated that time spent searching (>1.5 hours for a 2 ha site) was the most important determinant of detection probability for massasaugas (Chapter 2) and, given our time spent at each site searching for massasaugas and conducting vegetation sampling, we were reasonably confident in our assessment of occupancy. We evaluated relationships between site occupancy 15 probability and three derivations of the HSI model includ ing the full model scores, thermal subset scores (i.e., suitability indices LHC, DHC, SDTS, and BAT), and landscape - level subset scores (i.e., suitability indices AEDU and AEDW). We assessed these alternatives to the full HSI model because management conte xt differs for each. Landscape variables can be quantified via remote imagery and require no field - based vegetation sampling, and management focuses broadly on patch types and locations. In contrast, thermal variables represent within patch vegetation stru cture and composition managed through prescriptions (e.g., herbicide or burn to reduce encroachment of woody stems). RESULTS Habitat Suitability Our measured values for each suitability index across all sites represented nearly the entire range of s uitability from poor to optimal (Figure 1. 2). Average conditions across sites were optimal - or good - (i.e., SI = 0.00; Figure 1. 2c), with mean counts of 1,179 stems/ha. This result indicates that woody stem encroachment currently limits average site quality for massasaugas across southern Michigan, specifically on sites that supported historical populati ons that are currently unoccupied. 16 1. 2a 1. 2c 1. 2e 1. 2b 1. 2d 1. 2f Figure 1. 2. Habitat suitability production functions for components of the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) habitat suitability index model (Bissell 2006, Bailey 2010) for southern Michigan. The blue line overlaying each production function represents range of suitabilities observed across 27 field sites during the active snake season throughout southern Michi gan, 2015 - 2016. Points marked with x and represent median and mean values, respectively (for 2c, observed mean = 1179.38 stems/ha and median = 930 stems/ha). 0.00 0.20 0.40 0.60 0.80 1.00 0 50 100 Suitability Index Percent Live Herbaceous Cover 0.00 0.20 0.40 0.60 0.80 1.00 0 200 400 600 800 Suitability Index Stem Density of Trees and Shrubs >3 m (stems/ha) 0.00 0.20 0.40 0.60 0.80 1.00 0 20 40 60 80 100 Suitability Index Percent area in early successional deciduous upland 0.00 0.20 0.40 0.60 0.80 1.00 0 20 40 60 80 100 Suitability Index Percent Dead Herbaceous Cover 0.00 0.20 0.40 0.60 0.80 1.00 0 10 20 30 40 Suitability Index Basal Area of trees >3 m (m2/ha) 0.00 0.20 0.40 0.60 0.80 1.00 0 20 40 60 80 100 Suitability Index Percent area in early successional deciduous wetland 17 Habitat suitability index scores (averaged across all samples within a site) ranged from 0. 21 to 0.95 (mean = 0.64; Table 1.1 ). For sites where no massasaugas were observed during this study ( n = 18), HSI scores ranged from 0.21 - ranged from 0.28 to 0.89, indicating that massasaugas occurred at sites o f marginal ( n = 2; HSI range 0.26 - 0.50), good ( n = 5; 0.51 - 0.75), and optimal ( n = 2; 0.76 - 1.00) suitability (no occupied massasauga occupancy, suggesting that averaging vegetation conditions at random points across a site was not a useful application of the HSI model. Table 1.1 . Habitat suitability index (HSI) scores and the associated suitability index (SI) scores for measured habitat variables c at each st udy site by sampling date across southern Michigan. Scores based on Bailey (2010) HSI model for eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ). Sites where massasaugas were located during 2015 or 2016 field seasons are marked with an aster isk (*). Study Site a Date b n c Suitability Indices d HSI e Overall HSI f LHC (SI1) DHC (SI2) STDS (SI3) BAT (SI4) AEDU (SI5) AEDW (SI6) NE_H1 06 - 06 - 15 12 1.00 0.80 0.00 0.67 1.00 1.00 0.62 0.56 17 - 07 - 15 12 1.00 1.00 0.00 0.67 1.00 1.00 0.67 29 - 06 - 16 12 1.00 0.94 0.00 0.00 1.00 1.00 0.49 08 - 08 - 16 12 1.00 0.96 0.00 0.00 1.00 1.00 0.49 NE_H2 04 - 08 - 15 10 1.00 0.85 0.00 0.05 1.00 1.00 0.48 0.62 18 - 08 - 16 10 0.98 0.99 0.00 0.76 1.00 1.00 0.68 02 - 06 - 16 10 1.00 0.99 0.00 0.76 1.00 1.00 0.69 NE_H3 04 - 08 - 15 11 1.00 1.00 0.00 1.00 1.00 1.00 0.75 0.75 02 - 06 - 16 11 1.00 1.00 0.00 1.00 1.00 1.00 0.75 25 - 08 - 16 11 1.00 1.00 0.00 1.00 1.00 1.00 0.75 NE_L1 * 25 - 05 - 15 11 0.79 0.42 0.00 0.29 1.00 0.34 0.25 0.28 09 - 07 - 15 11 0.84 0.43 0.00 0.29 1.00 0.34 0.26 19 - 05 - 16 11 0.51 0.53 0.00 0.73 1.00 0.34 0.30 21 - 07 - 16 11 0.66 0.59 0.00 0.73 1.00 0.34 0.33 NE_L2 19 - 05 - 15 11 0.31 0.31 0.00 0.36 1.00 0.10 0.13 0.21 16 - 07 - 15 11 0.97 0.15 0.00 0.36 1.00 0.10 0.20 31 - 05 - 16 11 0.81 0.23 0.00 0.65 1.00 0.10 0.23 15 - 08 - 16 11 0.82 0.40 0.00 0.65 1.00 0.10 0.26 18 Study Site a Date b n c Suitability Indices d HSI e Overall HSI f LHC (SI1) DHC (SI2) STDS (SI3) BAT (SI4) AEDU (SI5) AEDW (SI6) NE_L3 19 - 05 - 15 12 0.47 0.29 0.00 0.00 1.00 0.56 0.15 0.21 16 - 07 - 15 12 0.91 0.33 0.00 0.00 1.00 0.56 0.24 31 - 05 - 16 12 0.86 0.19 0.00 0.02 1.00 0.56 0.21 12 - 08 - 16 12 0.75 0.39 0.00 0.02 1.00 0.56 0.23 NE_M1 * 25 - 05 - 15 11 0.91 0.40 0.00 0.85 1.00 0.88 0.51 0.58 20 - 07 - 15 11 1.00 1.00 0.00 0.85 1.00 0.88 0.67 19 - 05 - 16 10 0.69 0.68 0.00 0.96 1.00 0.88 0.55 21 - 07 - 16 10 1.00 0.57 0.00 0.96 1.00 0.88 0.59 NE_M2 20 - 05 - 15 10 1.00 0.75 0.00 0.00 1.00 1.00 0.44 0.58 16 - 07 - 15 10 1.00 0.96 0.00 0.00 1.00 1.00 0.49 31 - 05 - 16 10 1.00 0.89 0.00 0.87 1.00 1.00 0.69 15 - 08 - 16 10 1.00 1.00 0.00 0.87 1.00 1.00 0.72 NE_M3 17 - 07 - 15 10 1.00 1.00 0.74 1.00 0.84 1.00 0.86 0.87 29 - 06 - 16 10 1.00 1.00 0.80 1.00 0.84 1.00 0.87 08 - 08 - 16 10 1.00 1.00 0.80 1.00 0.84 1.00 0.87 SE_H1 * 28 - 05 - 15 11 0.91 0.69 0.00 0.94 1.00 1.00 0.63 0.72 10 - 07 - 15 11 1.00 0.96 0.00 0.94 1.00 1.00 0.72 18 - 06 - 16 11 1.00 0.77 0.31 0.90 1.00 1.00 0.74 25 - 07 - 16 11 1.00 0.86 0.31 0.90 1.00 1.00 0.77 SE_H2 19 - 06 - 15 11 1.00 0.86 0.00 0.74 1.00 1.00 0.65 0.71 15 - 07 - 15 11 1.00 1.00 0.00 0.74 1.00 1.00 0.68 14 - 06 - 16 11 1.00 0.94 0.00 1.00 1.00 1.00 0.74 17 - 08 - 16 11 1.00 1.00 0.00 1.00 1.00 1.00 0.75 SE_H3 * 19 - 05 - 15 11 1.00 1.00 0.39 0.94 1.00 0.63 0.68 0.66 07 - 07 - 15 11 1.00 1.00 0.39 0.94 1.00 0.63 0.68 20 - 05 - 16 11 1.00 0.85 0.26 0.92 1.00 0.63 0.62 22 - 07 - 16 11 1.00 1.00 0.26 0.92 1.00 0.63 0.65 SE_H4 20 - 05 - 15 11 0.84 1.00 0.65 1.00 0.43 0.24 0.29 0.31 07 - 07 - 15 11 1.00 1.00 0.65 1.00 0.43 0.24 0.31 22 - 07 - 16 11 0.85 1.00 0.81 1.00 0.43 0.24 0.31 13 - 05 - 16 11 1.00 1.00 0.81 1.00 0.43 0.24 0.32 SE_L1 14 - 05 - 15 24 0.69 0.40 0.00 0.91 1.00 0.39 0.35 0.42 15 - 07 - 15 24 1.00 0.97 0.00 0.91 1.00 0.39 0.50 17 - 05 - 16 11 1.00 0.54 0.00 0.73 1.00 0.39 0.39 23 - 08 - 16 11 0.95 0.89 0.00 0.73 1.00 0.39 0.45 SE_L2 * 18 - 05 - 15 11 0.92 0.58 0.23 1.00 1.00 0.35 0.46 0.57 08 - 07 - 15 10 1.00 1.00 0.23 1.00 1.00 0.35 0.54 13 - 05 - 16 11 1.00 1.00 0.75 1.00 1.00 0.35 0.63 22 - 07 - 16 11 1.00 1.00 0.75 1.00 1.00 0.35 0.63 SE_L3 20 - 07 - 15 11 1.00 0.83 0.00 0.99 1.00 0.31 0.46 0.43 11 - 08 - 16 11 0.88 0.47 0.18 0.90 1.00 0.31 0.40 16 - 05 - 16 11 1.00 0.62 0.18 0.90 1.00 0.31 0.44 SE_L4 05 - 08 - 15 10 0.97 0.66 0.00 0.87 1.00 0.75 0.55 n/a SE_M1 * 05 - 08 - 15 12 1.00 0.67 0.00 0.18 1.00 0.89 0.44 0.40 26 - 07 - 16 12 1.00 0.57 0.00 0.00 1.00 0.89 0.37 12 - 05 - 16 12 1.00 0.67 0.00 0.00 1.00 0.89 0.39 19 Study Site a Date b n c Suitability Indices d HSI e Overall HSI f LHC (SI1) DHC (SI2) STDS (SI3) BAT (SI4) AEDU (SI5) AEDW (SI6) SW_H1 31 - 05 - 15 10 1.00 1.00 0.00 0.94 1.00 0.97 0.72 0.73 22 - 07 - 15 10 1.00 1.00 0.00 0.94 1.00 0.97 0.72 23 - 05 - 16 10 1.00 1.00 0.00 1.00 1.00 0.97 0.74 02 - 08 - 16 10 1.00 1.00 0.00 1.00 1.00 0.97 0.74 SW_H2 * 30 - 05 - 15 10 1.00 1.00 0.00 1.00 1.00 1.00 0.50 0.80 23 - 07 - 15 10 1.00 1.00 0.00 1.00 1.00 1.00 0.75 01 - 06 - 16 10 1.00 1.00 0.93 1.00 1.00 1.00 0.98 02 - 08 - 16 10 1.00 1.00 0.93 1.00 1.00 1.00 0.98 SW_H3 * 02 - 06 - 15 11 1.00 0.69 0.00 0.67 1.00 1.00 0.59 0.65 23 - 06 - 15 11 1.00 0.96 0.00 0.67 1.00 1.00 0.66 25 - 05 - 16 11 0.89 0.81 0.04 0.80 1.00 1.00 0.64 03 - 08 - 16 11 1.00 1.00 0.04 0.80 1.00 1.00 0.71 SW_L1 25 - 07 - 15 10 0.85 0.71 0.00 0.75 1.00 0.11 0.32 0.35 25 - 06 - 16 10 0.58 0.59 0.35 1.00 1.00 0.11 0.35 03 - 08 - 16 10 0.73 0.59 0.35 1.00 1.00 0.11 0.37 SW_L2 27 - 07 - 15 11 0.84 0.69 0.00 0.00 1.00 0.95 0.37 0.37 27 - 05 - 16 11 0.78 0.51 0.00 0.00 1.00 0.95 0.32 04 - 08 - 16 11 1.00 0.75 0.00 0.00 1.00 0.95 0.43 SW_L3 25 - 07 - 15 10 1.00 0.27 0.00 0.17 1.00 0.10 0.20 0.26 25 - 06 - 16 10 0.97 0.33 0.00 0.70 1.00 0.10 0.27 04 - 08 - 16 10 0.94 0.68 0.00 0.70 1.00 0.10 0.32 SW_M1 30 - 07 - 15 10 1.00 1.00 0.96 1.00 1.00 0.90 0.94 0.95 18 - 05 - 16 10 1.00 1.00 1.00 1.00 1.00 0.90 0.95 01 - 08 - 16 10 1.00 1.00 1.00 1.00 1.00 0.90 0.95 SW_M2 * 28 - 07 - 15 11 1.00 1.00 0.37 1.00 1.00 1.00 0.84 0.89 18 - 05 - 16 11 1.00 1.00 0.64 1.00 1.00 1.00 0.91 01 - 08 - 16 11 1.00 1.00 0.64 1.00 1.00 1.00 0.91 SW_M3 24 - 07 - 15 10 1.00 0.90 0.00 0.17 1.00 1.00 0.52 0.65 24 - 05 - 16 10 1.00 0.87 0.00 0.94 1.00 1.00 0.70 02 - 08 - 16 10 1.00 0.92 0.00 0.94 1.00 1.00 0.72 a Study site represents the area (20 ha study site) within which the HSI score was assessed. b Date is the date (or final date) of sampling. Samples collected during the months of May and June indicate an c n = number of vegetation plots for that site. d Suitability indices represent the suitability (0 - 1) for thermal and landscape - level habitat characteristics measured within each site and are defined as follows: LHC (SI1) = live herbaceous cover, DHC (SI2) = dead herbaceous cover, STDS (SI3) = stem density of trees and shrubs >3m (stems/ha), BAT (SI4) = basal area of trees >3m (m 2 /ha), AEDU (SI5) = area in early successional deciduous upland, AEDW (SI6) = area in early successional deciduous wetland. e HSI r epresents the HSI score for each respective site by date. f Overall HSI represents the mean of the individual HSI scores for that site. 20 We subsequently evaluated HSI scores based on vegetation data collected at snake ample at massasauga locations at 8 occupied sites ( Table 1.2 massasaugas associated with vegetation structures at finer resolutions than the site. For example, the overall si te HSI for NE_L1 based on randomly located vegetation measures was 0.28 (i.e., our lowest - ranking occupied site), yet HSI score based on vegetation data collected at massasauga locations was 0.60. This result indicates that the HSI model is best suite d and biologically significant at the individual snake level of use (i.e., perhaps the size of the sampling plot, 20 x 5 m). Table 1.2 . Habitat suitability index (HSI) and associated suitability index (SI) scores for habitat variables c measured at eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) locations by site throughout southern Michigan, 2015 - 2016. Study Site a n b Suitability Indices c HSI d LHC (SI1) LHC (SI2) STDS (SI3) BAT (SI4) AEDU (SI5) AEDW (SI6) NE_L1 35 1.00 1.00 0.59 1.00 1.00 0.34 0.60 NE_M1 1 1.00 0.62 1.00 1.00 1.00 0.88 0.85 SE_H1 9 1.00 1.00 0.78 1.00 1.00 1.00 0.94 SE_H3 33 1.00 1.00 1.00 1.00 1.00 0.63 0.82 SE_M1 1 1.00 1.00 0.54 1.00 1.00 0.89 0.84 SW_H3 1 1.00 1.00 1.00 1.00 1.00 1.00 1.00 SW_M2 17 1.00 1.00 1.00 1.00 1.00 1.00 1.00 SW_H2 13 1.00 1.00 0.75 1.00 1.00 1.00 0.94 a Study site represents the area (within 20 ha study sites) corresponding to snake locations. b n = number of massasauga locations at which vegetation sampling was conducted. c Suitability indices represent average suitability (0 - 1) for snake locations within each site and include: LHC (SI1) = live herbaceous cover, DHC (SI2) = dead herbaceous cover, STDS (SI3) = stem density of trees and shrubs >3m (stems/ha), BAT (SI4) = basal area of trees >3m (m 2 /ha), AEDU (SI5) = area in early successional deciduous upland, AEDW (SI6) = area in early successional deciduous wetland. d HSI is average from all snake locations within a site. 21 We found that early - and late - season vegetation sa mples differed (Welch two sample t - test, < 0.05) for LHC for 7 of 26 study sites and DHC for 4 of 26 study sites ( Table 1.3 ). We were unable to test early and late season samples at one study site (SE_L4) because we collected a single late season sample in 2015 due to landowner availability. Hence, for most of our study sites changes in structural components of massasauga habitat were generally not significant throughout the active season. Furthermore, changes in vegetation between early and late - season s amples at a site rarely shifted HSI score into a new quality category ( Table 1.1 ). For example, at site NE_H1 in 2015, the early - and late - season HSI scores was 0.62 and 0.67, respectively, both 22 Table 1.3 . We lch two sample t - test results comparing vegetation for early (May - June) and late seasons (July - August) in southern Michigan. Vegetation variables correspond to habitat components for eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ). Confiden ce intervals (LCI = lower 95% confidence interval, UCI = upper 95% confidence interval) portray range of values bounding the difference of early - and late - sample means. Degrees of freedom (DF) based on number of vegetation samples collected for each site. An asterisk (*) indicates significance ( Study Site % Live Herbaceous Cover (LHC) % Dead Herbaceous Cover (DHC) Early Late t p DF LCI UCI Early Late t p DF LCI UCI NE_H1 8.333 8.333 0.000 1.000 46 - 1.183 1.183 4.417 5.292 - 0.915 0.365 4 6 - 2.800 1.050 NE_H2 7.300 6.476 1.068 0.294 2 9 - 0.753 2.401 5.100 4.952 0.121 0.905 20 - 2.394 2.690 NE_H3 8.091 7.545 0.666 0.511 28 - 1.132 2.223 6.364 6.955 - 0.490 0.630 1 9 - 3.118 1.936 NE_L1 4.045 4.636 - 0.655 0.516 38 - 2.417 1.235 2.136 2.364 - 0.226 0.822 4 2 - 2.254 1.799 NE_L2 3.455 5.545 - 2.837 0.007 * 4 2 - 3.578 - 0.604 0.955 1.000 - 0.098 0.922 36 - 0.984 0.893 NE_L3 4.125 5.167 - 1.236 0.223 4 6 - 2.738 0.655 0.792 1.500 - 1.387 0.174 3 7 - 1.743 0.327 NE_M1 5.000 6.524 - 1.536 0.133 3 9 - 3.531 0.483 2.476 4.095 - 1.549 0.130 36 - 3.739 0.501 NE_M2 6.300 7.350 - 1.585 0.121 3 8 - 2.391 0.291 4.100 5.200 - 0.947 0.350 3 8 - 3.453 1.253 NE_M3 9.400 9.700 - 0.906 0.382 12 - 1.020 0.420 8.000 9.100 - 1.469 0.166 1 3 - 2.721 0.521 SE_H1 6.955 8.409 - 2.342 0.024 * 40 - 2.710 - 0.199 3.591 4.636 - 1.140 0.261 41 - 2.897 0.806 SE_H2 8.545 8.773 - 0.481 0.634 3 5 - 1.188 0.733 4.591 6.864 - 2.474 0.018 * 40 - 4.129 - 0.416 SE_H3 7.955 9.045 - 2.278 0.028 * 41 - 2.058 - 0.124 4.727 5.955 - 1.218 0.230 4 2 - 3.261 0.807 SE_H4 6.000 6.682 - 0.854 0.398 4 2 - 2.294 0.930 6.727 8.455 - 2.937 0.005 * 41 - 2.915 - 0.539 SE_L1 5.143 7.514 - 3.382 0.001 * 6 8 - 3.771 - 0.972 1.971 4.829 - 4.281 0.000 * 61 - 4.191 - 1.523 SE_L2 6.682 8.810 - 3.672 0.001 * 3 9 - 3.300 - 0.955 4.227 6.190 - 2.662 0.011 * 38 - 3.456 - 0.471 SE_L3 7.000 6.864 0.163 0.872 2 2 - 1.596 1.869 3.000 3.136 - 0.121 0.904 26 - 2.448 2.175 SE_M1 6.583 6.917 - 0.476 0.638 3 1 - 1.763 1.097 3.250 2.958 0.317 0.754 22 - 1.616 2.199 SW_H1 8.600 9.050 - 0.822 0.416 37 - 1.558 0.658 6.750 7.550 - 1.014 0.318 35 - 2.402 0.802 SW_H2 7.300 9.400 - 3.921 0.000 * 3 3 - 3.190 - 1.010 8.300 8.400 - 0.161 0.873 3 8 - 1.359 1.159 SW_H3 6.591 6.955 - 0.347 0.731 41 - 2.482 1.754 3.727 5.045 - 1.198 0.238 4 2 - 3.539 0.903 SW_L1 3.600 4.900 - 0.959 0.348 2 2 - 4.113 1.513 2.800 3.150 - 0.221 0.827 1 8 - 3.670 2.970 SW_L2 4.818 5.682 - 0.700 0.492 2 1 - 3.432 1.705 2.364 3.545 - 1.205 0.239 26 - 3.196 0.833 SW_L3 6.000 6.400 - 0.484 0.636 14 - 2.170 1.370 1.300 2.150 - 1.235 0.227 2 8 - 2.260 0.560 SW_M1 7.800 9.050 - 1.508 0.158 1 2 - 3.059 0.559 6.500 7.400 - 0.869 0.397 17 - 3.082 1.282 SW_M2 7.000 9.318 - 3.455 0.005 * 1 2 - 3.786 - 0.850 5.909 6.864 - 0.919 0.370 1 9 - 3.131 1.221 SW_M3 7.700 6.900 0.846 0.407 2 3 - 1.158 2.758 4.400 4.650 - 0.199 0.845 15 - 2.924 2.424 23 Kernel Density Estimation Figure 1.3 ), and at least one RSPF converged for these sites ( Table 1.4 ). Model convergence depended on: 1) spatial extent of the utilization distribution (e.g., 95% kernel for SW_H2 covered 98% of the site [ Table 1.5 ] , limiting the number of available plots), and 2) amount of variation in vegetation measurements within kerne ls (e.g., available LHC within 50% kernel for SW_M2 only ranged from 80 - 100%, used LHC from 70 - 90%, thereby limiting differentiation space for the RSPF). Number of snake locations ranged from 49 - 262, and number of vegetation plots ranged from 30 - 56 within a site ( Table 1.5 ). The 95% kernels covered 30 - 98% of a site, whereas 25% core use areas covered 2 - 12% ( Table 1.5 ). 24 1. 3 a 1. 3 c 1. 3 b 1. 3 d Figure 1.3 . Eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) % home range kernels (50 - 95%), snake locations (red dots), and vegetation sampling locations (green dots) for study sites NE_L1 (1.5a), SE_H3 (1.5b), SW_H2 (1.5c), and SW_M2 (1.5d), in southern Michigan in 2015 and 2016. 25 Table 1.4 . Consistent AIC (CAIC) tables rankin g the resource selection probability functions for each % kernel home range at 4 study sites in southern Michigan, 2015 and 2016. Only models that converged are included. Site % Kernel CAIC a dCAIC b wCAIC c NE _ L1 60 81.007 0.000 0.410 95 82.094 1.087 0.238 90 82.094 1.087 0.238 50 83.621 2.614 0.111 70 92.352 11.345 0.001 25 92.487 11.480 0.001 85 108.551 27.544 0.000 SE _ H3 85 29.122 0.000 1.000 95 54.500 25.378 0.000 70 116.060 86.939 0.000 SW _ H2 90 13.331 0.000 1.000 60 54.108 40.778 0.000 SW _ M2 80 99.466 0.000 1.000 a C AIC = the consistent Akaike information criterion . b dCAIC = change in C AIC between univariate models . c wCAIC = the CAIC weights. 26 Table 1.5 . Number of snake locations (n s,total ) and vegetation sampling plots ( n v,total ) in each 20 ha study site, and corresponding kernel home range with the number of snake locations ( n s,hr ) and vegetation sampling plots ( n v,hr ) within, southern Michigan, 2015 - 2016. Study Site (20 ha) n s,total n v,total % Kernel Home Range Area (ha) % Area n s,hr n v,hr NE_L1 262 56 95 7.45 37% 262 41 90 6.25 31% 259 41 85 5.42 27% 255 37 80 4.7 24% 246 32 70 3.42 17% 241 31 60 2.53 13% 220 28 50 1.87 9% 208 27 25 0.57 3% 115 14 SE_H3 98 50 95 6.02 30% 97 37 90 4.65 23% 97 36 85 3.57 18% 93 33 80 2.96 15% 90 29 70 2.01 10% 88 26 60 1.43 7% 79 25 50 1.01 5% 70 16 25 0.32 2% 56 16 SW_H2 49 30 95 19.53 98% 49 30 90 17.32 87% 49 28 85 15.01 75% 49 24 80 13.11 66% 49 23 70 10.13 51% 46 21 60 7.97 40% 42 19 50 6.05 30% 34 16 25 2.47 12% 26 14 SW_M2 133 38 95 9.71 49% 132 28 90 7.15 36% 129 27 85 5.61 28% 123 20 80 4.52 23% 119 18 70 3.15 16% 115 15 60 2.67 13% 106 12 50 1.64 8% 98 10 25 0.62 3% 57 4 27 Resource Selection Probability Function For RSPFs that converged, kernel contours of top - ranking models varied by site. Massasauga use was best described by site vegetation at the 60% kernel scale (CAIC wt = 0.41) for site NE_L1, and at the >80% kernel scale for the other sites ( Table 1.4 ). Site NE_L1 was low - marginal quality (HSI = 0.28; Table 1.1 ) with massasauga habitat patchily distributed (HSI at snake locations was 0.60 ; Table 1.2 ), where as the other sites were medium and high quality with more uniform distribution of suitable habitat, particularly within the different kernel bands. Relationships between fitted values (i.e., the absolute probability of use) and habitat variables correspon ded to relationships portrayed in the HSI model for LHC and DHC . Our top - ranking RSPF functions (Table 1.4) showed increases in snake use when LHC and DHC >50% for site NE_L1 and when DHC >30% at site SW_M2 (Fig ures 1. 4 , 1. 5 ) . This relationship was not well illustrated at sites SE_H3 and SW_H2 for LHC and DHC , and at SW_M2 for LHC, likely due to the homogenous distribution of highly suitable habitat at these sites and within all home range kernels. For example, observed LHC ranged from 9 0 - 10 0% at SW _H W and 8 0 - 1 0 0 % at SW_M2 ( Figure 1.4 ) , while DHC ranged from 8 0 - 10 0% at SE_H3 and 7 0 - 10 0% at SW_H2 ( Figure 1.5 ) , and due to the homogeneity of these variables throughout these sites, there was little difference in structure between used and available (yet unused) areas (Table 1.6) ; thus, the consistency of high - quality habitat limited our ability to illustrate any difference in massasauga use based on these habitat variables at these sites . 28 Figure 1.4 . Marginal effects plots showing the relationship betw een availability of live herbaceous cover (0 - 100% transect coverage out of 0 - 10 meters as displayed in the x - axis , covering the range of values observed within the site ) and absolute probability of use by eastern massasauga rattlesnake s ( Sistrurus catenatus catenatus ) for the top - ranking home range kernels for sites NE_L1 (60% kernel), SE_H3 (85% kernel), SW_H2 (90% kernel), and SW_M2 (80% kernel) in southern Michigan (Table 1.4) . 29 Figure 1.5 . Marginal effects plots showing the relationship between availability of dead herbaceous cover (0 - 100% transect coverage out of 0 - 10 meters as displayed in the x - axis , covering the range of values observed within the site ) and absolute probability of use by eastern massasauga rattlesnake s ( Sistrurus cate natus catenatus ) for the top - ranking home range kernels for sites NE_L1 (60% kernel), SE_H3 (85% kernel), SW_H2 (90% kernel), and SW_M2 (80% kernel) in southern Michigan (Table 1.4) . 30 Table 1.6 . Comparative summary of vegetation structure (mean, minimum, maximum observed within the 5 x 20 m vegetation sampling plot) in used versus unused (i.e., available) areas based on the top - ranking resource selection probability function home range kernels (Table 1.4) for eastern massasauga rattlesnakes ( Sistrurus cate natus catenatus ) at 4 20 - ha sites throughout southern Michigan in 2015 and 2016. Sample size ( n ) indicates the number of vegetation samples that the values are based on within use or available areas. Vegetation Structure Site Kernel Use Available Mean Min Max n Mean Min Max n Percent Live Herbaceous Cover a NE _ L1 60% 87.5 0.0 100.0 28 50.4 0.0 100.0 28 SE _ H3 85% 84.9 30.0 100.0 33 88.2 50.0 100.0 17 SW _ H2 90% 93.9 50.0 100.0 28 95.0 90.0 100.0 2 SW _ M2 80% 88.3 70.0 100.0 18 94.5 80.0 100.0 20 Percent Dead Herbaceous Cover b NE _ L1 60% 81.4 0.0 100.0 28 24.6 0.0 100.0 28 SE_H 3 85% 95.2 80.0 100.0 33 50.6 10.0 100.0 17 SW _ H2 90% 90.4 40.0 100.0 28 65.0 60.0 70.0 2 SW _ M2 80% 83.3 30.0 100.0 18 68.5 0.0 100.0 20 No. Woody Stems c NE _ L1 60% 2.7 0.0 19.0 28 13.9 2.0 34.0 28 SE _ H3 85% 0.0 0.0 0.0 33 7.2 0.0 11.0 17 SW _ H2 90% 2.0 0.0 33.0 28 62.5 45.0 80.0 2 SW _ M2 80% 0.1 0.0 1.0 18 5.0 0.0 48.0 20 Average DBH (cm) d NE _ L1 60% 6.4 0.0 29.0 28 12.2 5.1 48.5 28 SE _ H3 85% 0.0 0.0 0.0 33 13.4 0.0 25.1 17 SW _ H2 90% 3.1 0.0 56.0 28 6.3 3.5 9.0 2 SW _ M2 80% 0.9 0.0 16.8 18 5.8 0.0 37.4 20 a Percentage of the 20 m transect covered by live herbaceous vegetation. b Percentage of the 20 m transect covered by dead herbaceous vegetation. c d 31 S nake use declined with woody stems >500/ha (5 stems/100m 2 ; Fig. 1.5 ) at site NE_L1, suggesting that massasaugas were more tolerant of higher woody stem densities than predicted by the HSI model at this site . S imilarly , at site SW_M2, massasauga use decreased with w oody stems >100/ha ( Figure 1.6 ) . At all sites, areas used by massasaugas contained lower stem densities than measured in the available areas, though at sites SE_H3 and SW_M2 these differences were less pronounced (Table 1.6) due to the highly suitable vege tation structure throughout the site . As with LHC and DHC, structural homogeneity in use areas likely limited our ability to illustrate any relationship between probability of use and stem density, particularly for site SE_H3 ( Figure 1.6 ), where 0 woody st ems occurred within areas of use (Table 1.6). 32 Figure 1.6 . Marginal effects plots showing the relationship between the number of woody stems (covering the range of values observed within the site) and absolute probability of use by eastern massasauga rattlesnake s ( Sistrurus catenatus catenatus ) for the top - ranking home range kernels for sites NE_L1 (60% kernel), SE_H3 (85% kernel), SW_H2 (90% kernel), and SW_M2 (80% kernel) in southern Michigan ( Table 1 .4 ) . 33 At both sites NE_L1 and SE_H3, probability of use by massasaugas declined to 0 when average DBH exceeded approximately 5 cm ( Figure 1.7 ) indicating massasauga use of areas with fewer overall trees , though this result should be interpreted cautiously as the relationship between average DBH and use does not account for the number of stems contributing to the DBH estimation. This fact likely contributed to our inability to detect a consistent pattern between the RSPF analysis for this variable ( Figure 1 .7 ) and the comparison of average DBH measurements among use versus available areas (Table 1.6). 34 Figure 1.7 . Marginal effects plots showing the relationship between the average diameter at breast height (cm) m in height within the 20 x 5 m plot (covering the range of values observed within the site) and absolute probability of use by eastern massasauga rattlesnake s ( Sistrurus catenatus catenatus ) for the top - ranking home range kernels for sites NE_L1 (60% kernel), SE_H3 ( 85% kernel), SW_H2 (90% kernel), and SW_M2 (80% kernel) in southern Michigan ( Table 1.4 ) . 35 Predicted Occupancy T he HSI score from the full model as a predictor variable resulted in the greatest range of occupancy probabilities, ranging from 0. 18 to 0. 53 fo r sites with HSI scores ranging from 0.2 1 to 0.95 ( Figure 1.8 ) , and had a greater rate of change (i.e., slope of 2.161; Table 1.7) than the thermal or landscape - level models (slopes of 1.716 and 2.136, respectively) . Alternately, when u sing the thermal variables ( LHC, DHC, SDTS, BAT) predicted occupancy ranged from 0.20 to 0.47 (HSI scores ranging from 0.27 to 1.00), and using the landscape - level variables (AEDU, AEDW) predicted occupancy ranged from 0.14 to 0.40 (HSI scores ranging from 0.34 to 1.00). T his finding indicates that use of the full model (i.e., including all 6 habitat variables) offers more biologically meaningful occupancy prediction resulting in a greater range of occupancy values. Figure 1.8 . Predicted occupancy probabilities for easte rn massasauga rattlesnake ( Sistrurus catenatus catenatus ) habitat suitability index (HSI) scores (Bissell 2006, Bailey 2010). Includes full model, thermal variables (LHC, DHC, STDS, and BAT), and the landscape - level variables (AEDU and AEDW). 0.00 0.20 0.40 0.60 0.80 1.00 0.00 0.20 0.40 0.60 0.80 1.00 Predicted Occupancy HSI Score HSI - Thermal HSI - Landscape HSI - Full Model 36 Table 1.7 . G eneralized linear model p arameter estimates and associated standard errors (SE), z scores, and p values for relationships between occupancy probabilities and eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) habitat suitability index (HSI) sco res (Bissell 2006, Bailey 2010). Includes full model, thermal variables (LHC, DHC, STDS, and BAT), and the landscape - level variables (AEDU and AEDW). Estimate SE z value Pr(>|z|) Intercept - 1.831 1.593 - 1.150 0.250 HSI thermal 1.716 2.287 0.750 0.453 Intercept - 2.539 2.521 - 1.128 0.259 HSI landscape - level 2.136 2.520 0.848 0.397 Intercept - 1.939 1.280 - 1.514 0.130 HSI full model 2.161 2.046 1.056 0.291 DISCUSSION We found that eastern massasauga rattlesnakes occupied sites (20 ha) with a wide range of suitability scores, ranging from marginal to high quality (i.e., HSI = 0.28 - 0.89). Generally, massasaugas were located within patches of greater suitability (i.e., HSI = 0.60 - 1.00) within our sites, consistent with the non - uniform distribution of higher quality habitats. Generally, RSPF indicated that massasaugas used vegetation structures consistent with relationships portrayed in the HSI model (Bailey 2010) for LHC and DHC but deviated for density of woody stems. Our results indic ated that massasaugas in southern Michigan were using habitats with higher woody stem densities than proposed by the HSI model (e.g., approximately 360 stems/ha, SI = 0.59, at site NE_L1) . Massasaugas in southeastern Michigan tend to avoid overstory tree c over and used herbaceous vegetation 0.5 - 1.5 m in height ( e.g., Moore and Gillingham 2006, Bailey et al. 2010, Johnson et al. 2016). Our observations of massasaugas using areas with higher woody stem densities than observed by others (e.g., Bissell 2006, Ba iley 2010) could be explained by lack of effective management at occupied sites (resulting in woody stem encroachment over time), or 37 greater plasticity in habitat use by massasaugas than originally thought. Our results lend support for using the HSI model to portray areas likely used by massasaugas across a range of habitat conditions in southern Michigan, but we caution that the relationship between use and woody stem density warrants further evaluation, and spatial extent is highly relevant. At lower qual ity sites occupied by massasaugas in southern Michigan, use of patchy habitats is best quantified at localized scales (e.g., 60% kernel). For these types of sites (e.g., NE_L1 in our study), small high - quality patches of habitat were interspersed in a matr ix of unsuitable vegetation types or structures. In other situations, sites contained larger, distinct regions of high quality used and unsuitable vegetation types (e.g., sites SE_H3 and SW_M2), and use was best portrayed at larger scales (i.e., >80% kern el). Both SE_H3 and SW_M2 have been actively managed (e.g., invasive species control via herbicide, burning, or mowing) to maintain early successional herbaceous vegetation, resulting in large areas of consistently high - quality massasauga habitat. Lastly, massasauga use at sites with homogenous high quality vegetation (e.g., SW_H2) was also best described at a large kernel (90%) because snakes were located throughout the site. The power of RSPF relies on used versus available (yet unused) locations and asso ciated habitat variables (i.e., Lele and Keim 2006), hence models for sites with homogenous vegetation (e.g., few unused plots) will struggle to converge. The relationship between probability of use and average DBH was likely confounded by our inability to account for number of stems contributing to the DBH variable for this RSPF. For example, within an area used by massasaugas a single small - diameter tree would result in a low average DBH and a low stem density estimate (i.e., optimal suitability), while n umerous small - diameter trees densely distributed throughout a n unused site would still indicate low average DBH yet would result in high stem density estimates (i.e., poor suitability), thus confounding any relationship between 38 massasauga use and average D BH. As such, we do not recommend using average DBH with RSPF as a predictor of massasauga use. Our findings highlight the importance of including heterogeneity when using RSPF to understand animal habitat use (i.e., the Modifiable Areal Unit Problem; Jelin ski and Wu 1996). In our analysis, 20 ha sites (corresponding to massasauga home range size) were too small to rigorously apply RSPF, especially when locations for individual snakes are pooled to represent a population - level response. We recommend doubling this size for future RSPF analyses in southern Michigan habitats. Spatial extent of the modeling area can significantly affect results and interpretations. For example, one of our sites (NE_L1) was scored on the low end of marginal (HSI = 0.28). Yet, th is site had relatively high massasauga abundance (Appendix D), and we documented successful parturitions and litters in 2015 and 2016 (Chapter 3). Although the HSI scores for this site suggested marginal suitability, patches ranging in size from 0.12 0.3 8 ha of highly suitable habitat occurred within. These low HSI score areas may hold the greatest potential for management in that the population seems poised to occupy new habitats as they become available. Habitat degradation due to succession is a threa t to massasauga populations woody vegetation encroachment are a conservation priority. The HSI model offers a means to assess and define current massasauga habi tats, can be used in lieu of species distribution models for species such as the massasauga where presence data are limited (e.g., Zajac et al. 2015), and can indicate amount of management effort necessary to improve lower - quality habitat. For example, mas sasauga occupancy probability for study site NE_L1 (HSI = 0.28) was estimated to be 0.21. According to the HSI model, this site was limited by woody stem densities (i.e., average stem density of 1573 stems per ha, resulting 39 in SI = 0). If we decrease stem density by 75% (to 393 stems per ha), the SI value for woody stem densit y becomes 0.55 and HSI for the site increases to 0.48 with predicted massasauga occupancy increases to 0.29. If we further reduce stem density by 90% (i.e., 157 stems per ha), the resu lting SI score is 0. 87 , resulting in HSI score of 0.56 and predicted massasauga occupancy of 0.33. This example illustrates how HSI model relationships, when supported by validation data, can portray localized effects on massasauga occupancy. Furthermore, manipulation of stem density can directly affect basal area SI, and indirectly affect live and dead herbaceous cover, resulting from decrease in canopy cover. and posit ively influence individual fitness (Johnson et al. 2016) if implemented appropriately (e.g., properly timing prescribed burns or mowing; Durbian 2006, Bailey et al. 2012). Woody stem density (SI3) was the most limiting factor among all our sites. Thus, in many situations, management of woody stem density alone via tree and shrub removal will likely improve HSI score and occupancy probability. For a New York population of massasaugas, Johnson et al. (2016) observed increased use of basking areas that had bee n managed (i.e., shrub removal) compared to unmanaged areas. They suggested that shrubs be cut to <0.25 m in height to limit regrowth and recommended shrub removal in habitats threatened by succession (Johnson et al. 2016). The full HSI model (i.e., consi dering all habitat attributes described by Bailey 2010) predicted the widest range of occupancy probabilities, demonstrating the importance of considering thermal (LHC, DHC, STDS, BAT) and landscape - level variables (AEDU, AEDW). Still, only considering a s ubset of HSI variables may be useful where time or resources for habitat assessments may be limited. For example, for interests in assessing occupancy probability 40 of new sites, landscape - level variables (AEDU and AEDW) are quantifiable using GIS software a nd a landcover database that do not require field work (save for confirmation of vegetation types). Likewise, where GIS and landcover database information is not available (e.g., some private lands), thermal variables may be used for habitat assessment and occupancy prediction. Such partial assessments may not be ideal in that they do not illustrate each aspect of massasauga habitat as defined by the HSI model but can be a useful starting point when identifying potentially occupied areas. MANAGEMENT IMPLIC ATIONS Based on the application of the Bailey (2010) HSI model at our study sites throughout southern Michigan, we determined that stem density is most often the primary habitat variable resulting in reduced habitat suitability for massasaugas. As such, w e recommend managers consider tree and shrub removal as a first step in improving massasauga habitats. Furthermore, with the manipulation of stem density, it follows that basal area and the herbaceous variables (LHC, DHC) will improve as well. Our validati on of the Bailey (2010) HSI model, along with results of our RSPF analysis, indicate that the HSI model can be used when determining areas of conservation value both for massasauga populations that are already identified and in need of management (e.g., si te NE_L1), and for delineating new areas that may potentially be occupied. These methods can be implemented within areas where massasauga occupancy is unknown using straightforward field methods such as the vegetation sampling techniques described here, an d in turn, can offer predicted occupancies for managers as they focus on sites that have potential for supporting massasaugas. Additionally , implementing our detection survey methods (Chapter 2) can then be used within these areas to further inform researc hers of habitat use by massasaugas. 41 ACKNOWLEDGMENTS We acknowledge the support of Michigan State University Department of Fisheries and Wildlife, and S. Winterstein and J. Tsao for their contributions to this work. Funding for this project was provided through the Michigan Department of Natural Resources State Wildlife Grant # F15AP00096 in cooperation with the U.S Fish and Wildlife Service, Wildlife and Sport Fish Restoration Program. We thank K. Bissell, A. Derosier, M. Donnelly, R. Fahlsing, C. Hanabu Michigan Department of Natural Resources. We thank J. Dingledine and S. Hicks from the U.S. Fish and Wildlife Service. We thank the employees and veterinary staff of John Ba ll Zoo, Grand Rapids, Michigan, including R. Colburn, B. Flanagan, and H. Teater. We also thank T. Meyers Harrison for veterinary assistance. We thank S. and C. Weaver and J. Buck; S. Leavitt, C. May, and R. Villegas from The Nature Conservancy; T. Funke a nd R. Roake from Michigan Audubon Society; Y. Lee from the Michigan Natural Features Inventory. We thank R. Bailey for assistance with HSI model application. We thank field technicians T. Brockman, B. Brodowski, C. Burden, G. Payter, and H. Reynolds (Britz ). 42 CHAPTER 2: SURVEY METHODOLOGY FOR EASTERN MASSASAUGA RATTLESNAKES AND FACTORS INFLUENC ING DETECTION Stephanie A. Shaffer, Henry Campa, III, Gary Roloff ABSTRACT The eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) was listed as Threatened in 2016 by the U.S. Fish and Wildlife Service under the Endangered Species Act. To aid researchers and managers in successfully detecting massasaugas in occupied habitat for conservation efforts we developed a visual encounter survey method in s outhern Michigan to quantify the detection probability for massasaugas and to determine which factors are most important in influencing our ability to detect them (i.e., environmental conditions, surveyor conditions). In 2015, we developed and implemented the detection survey methodology at sites throughout southern Michigan as a pilot study. In 2016, using the finalized detection survey methods, we conducted 54 surveys (paired, independent searchers) in 2 ha areas on 4 sites that were occupied by telemeter ed massasaugas. To parameterize the detection process, we collected data on environmental conditions (air temperature, humidity, solar radiation, surface temperature, precipitation) and included searcher characteristics (e.g., experience, time spent search ing) . We detected massasaugas on 11 of the 54 surveys. Detection probability from the null model was 0.31, but we found that time spent searching and minimum air temperature were important correlates of detection probability. Detection probability approach ed 1.00 as searcher time exceeded 90 minutes and approached 0.80 on cooler mornings (12.8°C) . Our findings offer a means to understand the reliability of visual encounter surveys for massasaugas. 43 INTRODUCTION The range of eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ; hereafter, massasaugas) extends from central New York and southern Ontario in the east to Iowa, Minnesota, and Missouri in the west, and includes southern Wisconsin, Michigan, Illinois, Indiana, Ohio, and Pennsylvania (USFWS 2016). Massasaugas were designated as federally Threatened throughout this range as of October 31, 2016, because of threats that included habitat degradation and fragmentation, land use change, human persecution, and road mortality (USFWS 2016). Considerin g that all projects with the potential to impact massasaugas are now subjected to federal review, the ability to implement standardized field survey techniques that reliably denote occupancy status is imperative for conservation efforts. Detection probabi lity is an important component of occupancy surveys and is defined as the likelihood of detecting an organism given that it occurs in a location (MacKenzie et al 2006). For elusive, cryptic species such as massasaugas, knowing the factors affecting detecti on probability is crucial for surveying (Harvey 2005), and ultimately for informing population and habitat management or monitoring (Christy et al. 2010). Estimating detection probability for snakes is challenging (Mazerolle et al. 2007, Durso et al. 2011, Willson et al. 2011), often because space use is poorly understood, capture and observation can be difficult, and movement of individuals can vary across landscapes (Steen 2010). Detectability of snakes can be influenced by numerous factors including hab itat composition and structure, availability of micro - habitat features (e.g., burrows, basking areas), and weather conditions such as cloud cover, temperature, and wind speed (Mazerolle et al. 2007, Moreno - Rueda and Pleguezuelos 2007, Foster et al. 2009, C hristy et al. 2010). Characteristics of individual snakes may also influence detection probability (Christy et al. 2010), particularly because thermoregulatory needs of individuals vary by age, sex, and reproductive status (Foster 44 et al. 2009). Indeed, the effects of thermal environments on space use have been documented for numerous species (Elmore et al. 2017). For snakes, thermoregulatory behaviors frequently determine whether the individual is readily observable for detection by researchers (e.g., openl y basking), or hidden. Additionally, detection probability may vary by observer (e.g., Harvey 2005). The goal of this part of our study was to quantify factors affecting massasauga detection probability during the active season (i.e., following emergence from hibernation and settling in summer habitats; Siegel 1986). Habitats used by massasaugas from June through August (approximately) are where snakes bask, forage, find mates, and in the case of reproductive females, have their young and brood (e.g., Rei nert 1981, Johnson 2000), and can differ from habitats used in early spring and fall around hibernacula (Reinert and Kodrich 1982). Given that habitat management projects such as mowing, invasive plant species treatment, and prescribed fire may occur durin g the massasauga active season (e.g., Durbian 2006), and that prevalence of construction projects (e.g., road work, urban developments) increases during summer, insights into active season surveys for massasaugas are critical to species conservation. Our a im was to quantify detection probability with respect to environmental and searcher variables for sites known to be occupied by massasaugas, and to offer a standardized survey protocol supported by empirical estimates of uncertainty . STUDY AREA Southern Mi chigan is a temperate region with moderate spring - summer (i.e., May Aug) temperatures that ranged from approximately 10 32° C during our study (201 5 to 201 6 in Jackson, MI; NOAA - NWS 2017) . Physiography of the counties within which this research was conduct ed (Barry, Calhoun, Jackson, Lenawee, Livingston, Oakland , and Washtenaw) consists of glacially deposited outwash plains, moraines, till plains, and lacustrine plains (Striker and 45 Harmon 1961, Engberg and Austin 1974, Engel 1977, McLeese 1981, Feenstra 1982, Thoen 1990, Tardy 1997). Soils within our study areas were well drained or poorly drained and loamy with interspersed sandy - loam , loamy sand, or mucky soil types (USDA - NRCS 2017) . We identified 27 study sites based on confirmed reports of ma ssasaugas within the last 25 years (Michigan Natural Heritage Database [MNFI] 2014) . These 27 sites represented a range of vegetation types and habitat qualit ies for massasaugas on private and public lands ( Appendix C: Table C. 1 ). Of the 27 sites, 11 occur red on private lands owned by citizens, non - profit conservation groups, or corporations, while the remaining 16 sites occurred on public lands (Appendix C). Sites were 20 ha in size (the maximum home range for a massasauga in southern Michigan; Bissell 200 6) a nd (the maximum distance moved by an individual massasauga in a single season in southern Michigan ; Bissell 2006). METHODS Initial Site Assessment When selecting sites, we coarsely classified habitat quality a priori into low, medium, and high based on proportion of suitable and unsuitable vegetation types for massasaugas within each 20 - ha site . The purpose of this classification was to ensure inclusion of sites of varying habitat structure and composition throughout s outhern Michigan. We used the 2006 National Land Cover Database (hereafter NLDC; MRLC 2015; Appendix C: Table C. 2 ) to identify vegetation types w ithin our study sites. The NLCD includes: open water, developed (open space), developed (low intensity), develo ped (medium intensity), barren land (rock/sand/clay), deciduous forest, evergreen forest, mixed forest, shrub/scrub, grassland/herbaceous, pasture/hay, cultivated crops, woody wetlands, and emergent herbaceous wetlands (MRLC 2015). We considered grassland/ herbaceous openings, woody wetlands, and emergent herbaceous wetlands as suitable types for massasaugas. We considered open water, development, deciduous forest, 46 evergreen forest, mixed forest, shrub/scrub, pasture/hay, and cultivated crops as non - suitable . We based these suitability groups from the Integrated Forest Management Analysis Program (IFMAP ; MDNR 2001 ) land cover data . We designated sites with <40% in suitable vegetation types as low site quality, between 40 60% suitable vegetation types as medium, and >60% suitable vegetation types as high quality ( Appendix C: Table C. 1 ). Capture and Marking Radio telemetering massasaugas helped us determine the location of massasaugas within our study sites and to quantify the vegetation structures used by individual massasaugas . From May through August 2015 and 2016, we located and captured massasaugas via random encounter surveys within and around study sites. We checked all captured massasaugas for a passive integrated transponder (PIT) tag ( 12 mm in length; AVID; Norco, CA ), and newly captured individuals were injected with a PIT tag subcutaneously into the dorsal region approximately 4 6 cm caudal to the cloaca (e.g., Bissell 2006). We transported m assasaugas weighing >100 g to a veterinary clinic for surgical implantation of a radio transmitter. While in captivity under veterinary observation, massasaugas were held individually in locked glass - front reptile aquariums ( model S24T; Neodesha Plastics I nc., Neodesha, KS), and provided a place to hide, a dish of water, and sheets of paper towel for additional hiding cover and to absorb liquids (e.g., spilled water, defecation). Veterinary staff surgically implanted massasaugas with one of three radio tra nsmitters of varying weights, depending on body mass (models R1515, 7 g, and R1680, 3.1 g, Advanced Telemetry Systems [ATS] , Inc., Isanti, MN; model SB - 2, 5.3 g, Holohil Systems Ltd., Ontario, Canada). A licensed veterinarian from the John Ball Zoological Park veterinary clinic in Grand Rapids, Michigan, followed transmitter implantation procedures described by Bailey (2010) and 47 Bailey et al. (2011) . The combined weight of the implanted radio transmitter and the PIT tag did not exceed 5% of body weight to m inimize marking effects (Lentini et al. 2011). Following protocol with the John Ball Zoo, individuals per site were implanted with a transmitter each year (further, no more than one of these three was a gravid female). Within 7 days of capture (in most cases 3 4 days), implanted massasaugas were examined and approved for release by the veterinarian and subsequently released at their original capture location (Appendix D) . Detection Surveys During the 2015 pilot season, detection survey methods were dev eloped and carried out at 12 2 - ha survey subsites within 11 of the 27 study sites . In 2016, using the finalized detection survey methods, surveys were carried out at 2 2 - ha survey subsites within or adjacent to each of 4 study sites. Survey subsites were n on - overlapping. The survey subsite size of 2 ha represents the minimum annual home range size for massasaugas in southern Michigan (Bissell 2006). In 2016, prior to and after each detection survey we visually confirmed that the respective subsite was occup ied by a telemetered massasauga by locating the snake with a Yagi antenna and receiver (model R - 1000, Communications Specialists, Inc., Orange, CA). Searchers had no knowledge of where the massasauga(s) was located. Although searchers crossed paths in the approximate center of the subsites, they were not allowed to discuss their results during the survey. For the detection surveys, to produce the survey replication needed to estimate detection probability, we paired two searchers that simultaneously started at opposite ends of the same subsite ( Figure 2. International Inc., Olathe, KS), searchers walked in general north - south and south - north transects simultaneously, resulting in 7 transect s with 20 m separation ( Figure 2. 1). We did not require searchers to walk in straight lines along transects. Rather, they meandered along transects 48 offering the opportunity to more thoroughly search the habitat. Searchers recorded the location of detected massasaugas, but to reduce their influence on snake detectability (i.e., for the other searcher as they walked through the site), they were not allowed to scan for a PIT tag or to determine if it was implanted with a radio transmitter. We conducted surveys between 0800 and 1800. Because of the limited number of telemetered snakes (Table 2. 1), we conducted surveys at the same subsites over multiple days. These surveys were at least 24 hours apart and we alternated searchers, searcher pairs, and start locatio ns to reduce searcher bias. Figure 2. 1. Schematic of eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) detection survey methodology conducted at 2 ha subsites (n=4) in southern Michigan. Filled - in numbered points indicate the pre - determined GPS locations along the subsite boundary used for navigation. Searchers simultaneously started at opposite ends ( i.e., searcher 1: points 1 to 14; searcher 2: points 14 to 1 ) and loosely traversed each transect searching for massasaugas. 49 Table 2. 1. Administrative information, habitat quality, and number of telemetered eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) by stage (adult, juvenile) and sex ( m ale, f emale) for 20 ha sites in southern Michigan, USA, for the sites where detection surveys were carried out in 2016. Site Code Ownership County Quality Ranking a Habitat Amount (% 20 ha) b Telemetered Massasaugas Age, Sex Suitable Unsuitable NE_L1 SE_H1 SE_H3 SW_M2 State park Conservation group Conservation group Conservation group Oakland Jackson Lenawee Calhoun Low High High Medium 18% 64% 64% 47% 82% 36% 36% 53% 2 2 1 2 Adult , adult Adult , juvenile Adult Adult , adult a We designated sites with <40% suitable vegetation types as low site quality, between 40 60% suitable vegetation types as medium, and >60% suitable vegetation types as high quality. b From the 2006 National Land Cover Data layer (MRLC 2015), suitable vege tation types included : grassland/herbaceous openings, woody wetlands, and emergent herbaceous wetlands; non - suitable included : open water, development, deciduous forest, evergreen forest, mixed forest, shrub/scrub, pasture/hay, and cultivated crops. 50 Detection Factors We a priori identified weather factors thought to be important to massasauga detection (Table 2. 2) that included variables related to air temperature, ground surface temperature, solar radiation, precipitation, and humidity (Mazerolle et al. 2007, Moreno - Rueda and Pleguezuelos 2007, Foster et al. 2009, Christy et al. 2010). We measured these variables at the start and end of each survey, and every 10 minutes during each survey as searchers conducted their walks through the su bsites. Air temperature, ground surface temperature, and humidity were measured using an Extech® Hygro - Thermometer with Infrared Thermometer (model RH101, Extech Instruments, Waltham, MA). Solar radiation was measured using an Extech® Foot Candle/Lux Meter (model 401025). We examined the influence of minimum, mean, maximum, and change in these variables during the survey on massasauga detection probability. We also included same day and previous day minimum air temperature (data collected a posteriori using the Climate Data Online Data Tools from the National Oceanic and Atmospheric Administration National Centers for Environmental Information; NOAA - NCEI 2017). Other potential massasauga detection factors that we analyzed included start and end time of each survey, Julian date, and the proportion of canopy cover of woody vegetation and herbaceous vegetation types in the 2 - ha subsites, estimated using aerial imagery in ArcGIS (Environmental Systems Research Institute, Inc., Redlands, CA, USA). The proportion o f woody and herbaceous vegetation types was explored as this would be a simple and repeatable way for researchers to assess sites potentially supporting massasauga populations. Searcher covariates included an individual identifier, searcher skill (a relati ve index among all searchers based on the amount of previous field experience with the species ranging from 0 3; 0 = no previous work with massasaugas, 1 = 1 year [spring - summer field season] previous work with massasaugas, 2 = 51 Table 2. 2. All detection c ovariates (environmental, habitat, searcher, and survey) considered and measured during 2015 and 2016, quantified to evaluate their importance when detecting eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) in their habitats throughout south ern Michigan . Category Covariate Name Comments Environmental AirAvg Air temperature average during survey AirDif Air temperature change during survey AirMax Air temperature maximum during survey AirMin Air temperature minimum during survey difNOAAminAirmax SameDayMinNOAA minus AirMax difNOAAminAirmin SameDayMinNOAA minus AirMin difNOAAminSurfmin SameDayMinNOAA minus SurfMin HumAvg Humidity average during survey HumMax Humidity maximum during survey HumMin Humidity minimum during survey Prec Precipitation during survey (0 for no rain, 1 for rain) PrevDayMinNOAA Previous day air temperature minimum (NOAA) SameDayMinNOAA Same day air temperature minimum (NOAA) SoilAvg Soil temperature average during survey a SoilMax Soil temperature maximum during survey a SoilMin Soil temperature minimum during survey a SolRadAvg Solar radiation average during survey SolRadMax Solar radiation maximum during survey SolRadMin Solar radiation minimum during survey SolRadSD Solar radiation standard deviation during survey SolRadSE Solar radiation standard error during survey SurfAvg Soil surface temperature average during survey b SurfMax Soil surface temperature maximum during survey b SurfMin Soil surface temperature minimum during survey b Habitat herbtowood Herbaceous:woody vegetation type PropHerb Proportion herbaceous vegetation type PropWood Proportion woody vegetation type SqmHerb Square meters of herbaceous vegetation type SqmWood Square meters of herbaceous vegetation type Searcher MMTs Number of times searcher stopped to take measurements OverallRank Rank of surveyor skill c RankSnake Index of surveyor success d RankSnakePerSite Proportion of successful surveys for each surveyor SearchMins Time spent actively searching e SrchrCode Individual searcher identification code (1 6) TotalMins Total survey time f Survey DaysSince Days since last survey EndMins End time of survey Julian Julian date of survey SiteOccu Subsite occupancy status g 52 Category Covariate Name Comments StartMins Start time of survey SubsiteCode Individual identification for each survey subsite SurveyOrder Number of times subsite had been surveyed a Soil temperature was measured in 2015 but not in 2016 . Soil temperature covariates were excluded from the final analysis. b Soil surface temperature was measured in 2016 but not in 2015. c Rank of searcher skill from 1 (Low) to 6 (High), partially informe d by the index of searcher success covariate. d Index of searcher success based on number of massasaugas located during field season (0 3). e Does not include time spent measuring detection covariates during the survey. f Includes time spent searching for massasaugas and measuring detection covariates. g Covariate SiteOccu excluded from final analysis which included only 2016 data since the occupancy = 1 for all survey subsites in 2016. 53 2 years previous work with massasaugas, 3 = >2 years previous work with massasaugas), proportion of successful detections by each searcher (i.e., the proportion of total surveys which resulted in a positive detection), and total minutes spent actively sear ching during a survey (i.e., not including time spent measuring environmental variables; Table 2. 2) . because the methods discussed here were developed as a potential tool for researchers and managers to use, and such vegetation information would not be readily available for any given site. The general habitat quality ranking that was assessed as a detection variable for this analysis (i.e., the NLCD habitat score used to qualify sites as low, medium, or high quality) is a broader representation of quality and is both accessible and easily applied without requiring on - site vegetation sampling efforts. All snake capturing, handling and housing methods were approved by Mi chigan State - 087 - 00, and by the Michigan Department of Natural Resources (MDNR) Fisheries Division (Scientific Collectors Permit #PR8114) . Public land access was approved by t he Michigan Department of Natural Resources Permit to Use State Land #PR1136 - 1 . Private land access was arranged directly with landowners prior to the start of any research activity at that property. In accordance with the protocol and requirements outline d in our IACUC approval and MDNR Scientific Collectors Permit, boots and all gear used in the field were disinfected using a 10% bleach solution following each day in the field and before moving among field sites to reduce the potential spread of pathogens ( Appendix E ). 54 Analysis We standardized model covariates using the scale() function in R. Given the large set of a priori detection covariates that measured similar things (e.g., 4 measures of air temperature; Table 2. 2), we built univariate logistic regression models (predicting detection probability) to help identify those with statistical support for inclusion in the final model set (Table 2. 2). We subsequently used AIC c to rank all logistic models (Table 2. 2) and removed redundant variables by selecting only the variable among those with the highest AIC c score. Single season occupancy models (with occupancy fixed at 1.00), based on various combinations of these covariates, were subsequently ranked using AIC using t he R program Unmarked (Fiske et al. 2011, Fiske and Chandler 2011, 2015). We calculated pairwise Pearson r correlation coefficients for the final 6 RESULTS In 2015, 24 pilot detection survey pairs were conducted at 11 sites (5 occupied, 6 unoccupied) to develop our survey methods (Table 2. 3) ; 5 total massasaugas were observed at 3 of sites . Data from the 2015 detection surveys were omitted from further analys is due to sparsity of data and the fact that these were pilot surveys to develop and test our methods. Final survey methods were carried out in 2016 . In 2016, we captured 7 massasaugas at 4 of our 27 study sites from 12 May to 14 October in 2016 that were PIT tagged and telemetered (Table 2. 1, Table 2. 3 ; Appendix D ). From 8 June to 17 August, we surveyed 8 occupied 2 - ha subsites at the 4 sites (2 subsites per site) from 5 - 9 times, omitting one subsite (subsite SE_H3 _s2) because we never had any telemetered massasaugas within that subsite, resulting in 54 survey pairs (Table 2. 4). We detected at least one massasauga on 11 of those surveys (Table 2. 4). Our null model estimated massasauga detection probability at 0.31 ( SE = 0.167). All analyses and results are based on the data collected during the 2016 detections surveys . 55 Table 2. 3 . The 27 20 - ha study sites where detection surveys were carried out for detection of the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) in 2015 and 2016 throughout southern Michigan. Year Site Code a Property Name Survey Subsite Code b # Individual Surveys c Massasaugas Detected d Site Occupancy e 2015 f NE_H1 Indian Springs Metropark NE_H1_s1 4 0 0 NE_L1 Seven Lakes State Park NE_L1_s1 4 0 1 NE_M1 Seven Lakes State Park NE_M1_s1 4 0 1 NE_M3 Indian Springs Metropark NE_M3_s1 4 0 0 SE_H1 Liberty Fen SE_H1_s1 4 0 1 SE_L4 Fay Lake SE_L4_s1 4 0 0 SE_H3 Ives Road Fen SE_H3_s1 4 3 1 SE_H3_s2 4 1 1 SE_L3 Sharonville State Game Area SE_L3_s1 4 0 0 SW_H2 Otis Audubon Sanctuary SW_H2_s1 4 0 1 SW_M1 Baker Audubon Sanctuary SW_M1_s1 4 0 0 SW_M2 Baker Audubon Sanctuary SW_M2_s1 4 1 1 2016 NE_L1 Seven Lakes State Park NE_L1_s1 18 3 1 NE_L1_s2 18 3 1 SE_H1 Liberty Fen SE_H1_s1 12 0 1 SE_H1_s2 12 1 1 SE_H3 Ives Road Fen SE_H3_s1 20 8 1 SE_H3_s2 g 20 0 1 SW_M2 Baker Audubon Sanctuary SW_M2_s1 10 2 1 SW_M2_s2 10 1 1 a Site Code indicates focal region (NE = North - East, SE = South - East, SW = South - West), site suitability (L = Low, M = Medium, and H = High, based on our preliminary analysis of the proportions of available vegetation types using the NLCD land cover data [MR LC 2015]) . b Survey Subsite Code indicates the Site Code of the 20 - ha site and the number of surveys carried out within that site (e.g., s1, s2). c # Individual Surveys indicates the number of individual surveys carried out within that survey subsite (e.g ., 4 surveys equates to 2 pairs of surveyors as discussed in the detection survey methods). d Massasaugas Detected indicates the total number of massasaugas observed by surveyors within that survey subsite. e Site Occupancy indicates whether the survey subsite was known to be occupied at any point in time during the study via radio telemetry or observation (1 = occupied, 0 = unoccupied). f Year 2015 pilot detection data were omitted from our final analysis due to sparsity of data, unconfirmed occ upancy status for many of the subsites, and because of the limited amount of time we had for surveys resulting from the time it took to develop the final survey methodology. g This subsite was omitted from our final analysis because no massasauga was dete cted within this subsite at any time during the extent of this research. 56 Table 2. 4. Survey count and massasauga detection for occupied subsites (2 ha) in the 4 sites (20 ha) where the final detection surveys were conducted for eastern massasa uga rattlesnake ( Sistrurus catenatus catenatus ), southern Michigan, USA, in 2016. Site (20 ha) Subsite (2 ha) Survey Pairs Paired Surveys with Detected Massasaugas NE_L1 SE_H1 SE_H3 SW_M2 1 2 1 1 2 1 2 9 9 6 10 10 5 5 3 1 1 3 0 2 1 The top - ranking univariate models estimating detection probability included time spent searching ( w i = 0.94), followed by minimum air temperature (AIC w i = 0.04) and the number of times a subsite was surveyed (AIC w i = 0.01; Table 2. 5; Table 2. 6 shows AIC results for all 39 covariates evaluated ; Appendix G ). Parameter coefficients from the top - ranking model indicated that massasauga detection probability increased nearly to 1 as search minutes approached 95 minutes ( = 2.46, SE = 0.767, p = 0.001 ; Figure 2 . 2 ) and decreased to below 0.25 as minimum air temperature increased to 21.1 °C ( = 1.13, SE = 0.384, p = 0.003 ; Figure 2. 3 ). As subsites were repeatedly surveyed (i.e., number of times a subsite was surveyed) detection probability increased ( = 1.11, S aic = 9.39; Burnham and Anderson 2002) . Our results indicated that the likelihood of detecting massasaugas was high (i.e. detection probability >0.9) if searchers spent >90 min in our 2 - ha subsites , and almost undetectable if they spent <60 min ( Figure 2. 2). We also found that detection probability was highest when minimum air temperatures ranged between 12.8 and 21.1° C (resulting in average detection probabilities >0.20), but those results were hi ghly variable. ( Figure 2. 3). On hot days (i.e., >25° C) massasaugas were almost undetectable on average ( Figure 2. 3). 57 Table 2. 5. AIC table for models found to have the greatest influence on detectability of eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) at 2 ha survey subsites in southern Michigan, USA. All models are statistically supported ( p Univariate Model K a AIC b i c w i d Minutes spent searching Minimum air temperature Number of times subsite was surveyed Same day minimum temperature (NOAA) Julian date Minimum surface temperature Null model a 3 3 3 3 3 3 3 67.01 73.14 76.40 77.95 78.67 79.06 81.50 0.00 6.12 9.39 10.93 11.66 12.05 14.49 0.94 0.04 0.01 0.00 0.00 0.00 0.00 a K = the number of estimated parameters for each model. b AIC = the Akaike information criterion . c . d wi = the Akaike weights. e The null model, which includes no observation covariates. 58 Table 2. 6 . AICc table for all 39 covariates considered (plus the null model) to have a potential influence on detection probability for eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ), southern Michigan, USA, 2016 . Covariate Covariate Name K a AIC b i c w i d Time spent actively searching e SearchMins 3 67.01 0.00 0.64 Total survey time f TotalMins 3 69.30 2.29 0.20 Number of times surveyor stopped to take measurements MMTs 3 70.69 3.68 0.10 Air temperature - minimum during survey AirMin 3 73.14 6.12 0.03 Number of times subsite had been surveyed SurveyOrder 3 76.40 9.39 0.01 Air temperature - average during survey AirAvg 3 77.61 10.60 0.00 Same day air temperature minimum (NOAA) SameDayMinNOAA 3 77.95 10.93 0.00 Julian date of survey Julian 3 78.67 11.66 0.00 Soil surface temperature minimum during survey SurfMin 3 79.06 12.05 0.00 Start time of survey StertMins 3 79.84 12.83 0.00 Air temperature - maximum during survey AirMax 3 80.01 13.00 0.00 Air temperature - change during survey AirDif 3 80.01 13.00 0.00 Solar radiation standard error during survey SolRadSE 3 81.47 14.46 0.00 Null model EMRfm1 2 81.50 14.49 0.00 End time of survey EndMins 3 81.75 14.73 0.00 Soil surface temperature maximum during survey SurfMax 3 82.25 15.23 0.00 Humidity - average during survey HumAvg 3 82.35 15.34 0.00 Solar radiation standard deviation during survey SolRadSD 3 82.39 15.38 0.00 Herbaceous:woody vegetation type herbtowood 3 82.47 15.45 0.00 Humidity - minimum during survey HumMin 3 82.57 15.55 0.00 Humidity - maximum during survey HumMax 3 82.57 15.55 0.00 Solar radiation average during survey SolRadAvg 3 82.59 15.57 0.00 Proportion of successful surveys for each surveyor RankSnakePerSite 3 82.81 15.80 0.00 SameDayMinNOAA minus AirMax difNOAAminAirmax 3 82.89 15.87 0.00 Solar radiation maximum during survey SolRadMax 3 83.04 16.03 0.00 Precipitation during survey (0 for no rain, 1 for rain) Prec 3 83.04 16.03 0.00 Same day minimum (NOAA) minus surface temperature minimum difNOAAminSurfmin 3 83.14 16.13 0.00 Index of surveyor success g RankSnake 3 83.23 16.22 0.00 Solar radiation minimum during survey SolRadMin 3 83.28 16.27 0.00 Individual identification for each survey subsite SubsiteCode 3 83.36 16.35 0.00 59 Table 2.6. Covariate Covariate Name K a AIC b i c w i d SameDayMinNOAA minus AirMin difNOAAminAirmin 3 83.41 16.39 0.00 Individual searcher identification code (1 - 6) SrchrCode 3 83.41 16.40 0.00 Rank of surveyor skill h OverallRank 3 83.43 16.42 0.00 Previous day air temperature minimum (NOAA) PrevDayMinNOAA 3 83.43 16.42 0.00 Proportion woody vegetation type PropWood 3 83.44 16.43 0.00 Square meters of woody vegetation type SqmHerb 3 83.44 16.43 0.00 Square meters of herbaceous vegetation type SqmWood 3 83.44 16.43 0.00 Proportion herbaceous vegetation type PropHerb 3 83.44 16.43 0.00 Soil surface temperature average during survey SurfAvg 3 83.49 16.47 0.00 Days since last survey DaysSince 3 83.50 16.49 0.00 a K = the number of estimated parameters for each model. b AIC = the Akaike information criterion. c i = change in AIC between univariate models. d w i = the Akaike weights. a Does not include time spent measuring detection covariates during the survey. b Includes time spent searching for massasaugas and measuring detection covariates. c Index of searcher success based on number of massasaugas located during field season (0 3). d Rank of searcher skill from 1 (Low) to 6 (High), partially informed by the in dex of searcher success covariate. 60 Figure 2. 2. Predicted detection probabilities for eastern massasauga rattlesnakes by minutes spent actively searching (i.e., total length of search time in minutes for a given survey), at 2 ha survey sites, southern Michigan, USA, during the active season 2016. Detection probability is indicated by the solid line; 95% confidence intervals are indicated by the dotted lines. 61 Figure 2. 3. Predicted detection probabilities for massasauga rattlesnakes across a range of minimum air temperatures (C) at 2 ha survey sites, southern Michigan, USA, during the active season 2016. Detection probability is indicated by the solid line; 95% confidence i ntervals are indicated by the dotted lines. 62 DISCUSSION We identified factors that can influence detection probability of massasauga rattlesnakes during visual encounter surveys on 2 ha subsites during the active season. Time spent actively searching and minimum air temperature were most influential; incorporating these factors into survey planning can improve the probability of detecting massasaugas in occupied habitat of varying quality. Detect ion probability can be relatively low for cryptic species, particularly snakes (e.g., Durso et al. 2011), but our work illustrated that detection probability can be substantially improved during optimal survey conditions. For example, we estimated that mas sasauga detection probability (i.e., detecting at least 1 massasauga within the 2 ha detection survey area) approached 1.0 with the proper combination of search time and minimum temperature (e.g., 90 min search time at ~13°C). Hence, our results indicated that massasaugas can be detected on occupied sites with a relatively easy to implement visual encounter survey protocol, provided that search time and temperatures are considered . Other studies reporting on factors influencing the ability to find massasau gas also found that temperature was important (Casper et al. 2001, Shoemaker and Gibbs 2010). For example, in a consensus of expert opinion from massasauga researchers throughout the species range, Casper et al. (2001) suggested that surveys for massasauga (Casper et al., 2001:3) ; C asper et al. (2001) state that their survey methods may not be applicable south and west of the Missouri River . For a northeastern New York population, Shoemaker and Gibbs (2010:508) found that most massasauga observations occurred during times of day when temperatures were increasing . Collectively, these results confirm that thermoregulatory behaviors are important considerations when surveying for massasaugas. Thermoregulatory behaviors 63 influence how massasaugas use vegetation and other micro - site feature s (e.g., basking sites). For example, Harvey and Weatherhead (2010: 414) found that massasaugas at the edge of their northern range in Ontario, Canada, thermoregulated primarily by microhabitat selection based on favorable thermal ranges (as opposed to macr ohabitat selection), becoming more visible as forest cover declined, resulting in more efficient thermoregulatory behaviors . Such behavior was exhibited by massasaugas at our study sites, generally basking either partially or completely out in the open in herbaceous vegetation types with little to no overstory cover. For ectotherms like snakes, the relationship between individual body temperature and ambient temperatures within a given site is an important determinant of behavior and space use (e.g., Elmor e et al. 2017). Studies to date have focused on snake body temperature and microsite (in close proximity to the snake) ambient temperature (e.g., Harvey and Weatherhead 2010), recognizing that microsite temperatures are confounded by thermoregulatory behav iors of individual animals. Indeed, massasauga basking sites tend to be warmer than temperatures observed at random locations (Shoemaker and Gibbs 2010). During 20 min time - limited detection trials within 0.25 ha for massasaugas using temperature - logging i mplanted radio transmitters, Harvey (2005) found that detection probability of massasaugas on the Bruce Peninsula in Ontario, Canada, was greatest when body temperatures ranged from 20 30° C. Furthermore, Shoemaker and Gibbs (2010 :508) reported average mas sasauga body and average environmental temperatures of 29.3° C and 29.1° C, respectively, at time of capture for a population near Syracuse, New York. In their description of suitable conditions for effective massasauga surveys, Casper et al. (2001:3) reco mmended surveying between 10 and 27° C. Our results are generally consistent with these previous observations, indicating that detection probability is highest during cooler days (e.g., approaching 0.80 at 12.8°C ) when massasaugas 64 apparently seek microsite s that expose them to direct sunlight. Hence, training crews to look for these microsites (e.g., exposed mats of dead vegetation, grass tussocks) during cooler mornings or afternoons improves detection probability. Our results also indicated that as temper atures exceed 24° C massasauga detection probability was <0.1, suggesting that surveys during warmer days will be considerably less effective, or that searchers should aim to begin surveys earlier in the day during ideal temperature conditions. For researc hers locating massasaugas in other parts of the species range (i.e., northern or southern limits) optimal temperatures for detection may need to be adjusted. For example, for southern populations , massasauga basking habits may differ from massasaugas in th e north because of differences in daily and seasonal temperature variability across the species range. Given optimal search time and temperature conditions, our results indicated that massasauga surveys could be conducted throughout the day (i.e., survey s tart time was not an important covariate). We also failed to find a searcher or habitat quality effect (similar to Harvey 2005), indicating that basic training of field staff in snake and microsite identification, field navigation (e.g., use of GPS, compas s), and training on the survey protocol prior to data collection, can result in reliable surveys for massasaugas during the active season throughout a range of habitat conditions. Finally, though survey fatigue was not a factor accounted for in this study, anecdotally we found that two surveys per searcher per day was optimal for maintaining searcher alertness and interest. Further, to maintain optimal searcher performance, we recommend taking a break between surveys. Three surveys were attempted on multipl e occasions and in these instances the third survey always proved to not be worth the effort for the searchers involved due to fatigue and decreased interest and alertness. 65 Conservation concerns impacting massasaugas (e.g., federally Threatened status, hab itat degradation and fragmentation), combined with a cryptic life history in dense vegetation types, highlights the need for reliable , standardized survey methodologies as described here. A standardized method would ensure that researchers are following re producible methods with comparable effort when detection the species rangewide. Our results can be used to increase the effectiveness of conducting eastern massasauga rattlesnake surveys and to quantify survey reliability. We encourage biologists to use st andardized survey methodologies to detect massasaugas throughout their range and across a range of habitat conditions, providing additional avenues for conservation planning. MANAGEMENT IMPLICATI ONS We offer a standardized survey technique for determining occupancy status of sites potentially supporting massasaugas. During the active season, we recommend that 2 - ha areas be surveyed for 90 min using coordinated transect walks during days or at a time of day with cooler temperatures (~13° C). These temperatur es tend to occur early in the morning during spring or summer in southern Michigan, though this pattern can vary daily and annually; e.g., NOAA - NWS 2017). For larger areas we recommend distributing 2 - ha survey subsites throughout as opposed to conducting a long, single survey to avoid searcher fatigue. The 2 - ha area of our survey subsites represented minimum annual home range size for massasaugas in southern Michigan (Bissell 2006), so we recommend that researchers use subsite sizes that align with region - s pecific massasauga home ranges. ACKNOWLEDGMENTS We acknowledge the support of Michigan State University Department of Fisheries and Wildlife, and S. Winterstein and J. Tsao for their contributions to this work. Funding for this project was provided throug h the Michigan Department of Natural Resources State Wildlife 66 Grant # F15AP00096 in cooperation with the U.S Fish and Wildlife Service, Wildlife and Sport Fish Restoration Program. We thank K. Bissell, A. Derosier, M. Donnelly, R. Fahlsing, C. Hanaburgh, A Michigan Department of Natural Resources. We thank J. Dingledine and S. Hicks from the U . S . Fish and Wildlife Service . We thank the employees and veterinary staff of John Ball Zoo , Grand Rapids, Michigan, including R. Colburn, B. Flanagan, and H. Teater. We also thank T. Meyers Harrison for veterinary assistance. We thank S. and C. Weaver and J. Buck; S. Leavitt, C. May, and R. Villegas from The Nature Conservancy; T. Funke and R. Roake from Michigan Audubon Society; Y. Lee from the Michigan Natural Features Inventory; and M. Dreslik at the University of Illinois for their persistent support of our project. We thank field technicians T. Brockman, B. Brodowski, C. Burden, G. Payter, H. Reynolds (Britz), V. Romanek, and A. Tarnowski. 67 CHAPTER 3: EASTERN MASSASAUGA R ATTLESNAKE SURVIVORSHIP , MOVEMENTS, AND REPRODUCTION Stephanie A. Shaffer, Henry Campa, III, Ryan Colburn, Scott Winterstein ABSTRACT The eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) was listed as Threatened in 2016 by the U.S. Fish and Wildlife Service under the Endangered Species Act in part due to a documented population decline rangewide . While some life history and demographic data are available for large , well - studied populations occurring in high quality habitats, little is known about massasauga populations occurring outside of these areas . To quantify survivorship and obtain life history data for massasaugas throughout southern Michigan, we survey ed for massasaugas at 27 20 - ha study sites in areas known to have been historically occupied by massasaugas . From May through September (i.e., the active season) in 2015 and 2016, t elemetry was used to relocate snakes . S ix of the 27 study sites resulted in sufficient massasauga locations to estimate survivorship rates across the study region . For each 137 - day study period in the 2015 and 2016 active season , adult massasauga survivorship using a modified version of the Mayfield method was 0. 767 (SE = 0.016 ) . For juveniles and neonates ( n = 10) , apparent survivorship was 0.65 . Using radiograph imagery or palpation, embryo counts for gravid females ( n = 17) across all sites ranged from 5 to 18, and litter counts ( n = 6) in the field ranged fro m 1 to 15 . Neonate massasaugas were radio tracked from 2 to 26 days, moving up to 551.2 m from the gestation site within this period . Our active season survivorship rates fall within the range of estimates from other massasauga populations and add to the c urrent literature which is relatively lacking in life history data for neonate and juvenile massasaugas. Because demographic rates can be used to infer population viability, such information may be used by 68 managers to assess small and fragmented populations in lower - quality habitats throughout the species range. INTRODUCTION The eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ; hereafter, massasauga) was designated as federally Threatened rangewide as of October 31, 2016, due to threa ts that included habitat degradation and fragmentation, land use change, human persecution, and road mortality (USFWS 2016). Massasauga range extends from central New York and southern Ontario in the east to Iowa, Minnesota, and Missouri in the west, and i ncludes southern Wisconsin, Michigan, Illinois, Indiana, Ohio, and Pennsylvania ( USFWS 2016). Research has suggested that demographic rates for massasaugas can vary greatly among populations throughout this range ( e.g., adult survivorship; Bissell 2006, Ba iley 2010, Bailey et al. 2011, Jones et al. 2012), yet few studies estimate neonate or juvenile survivorship rates (e.g., Hileman et al. 2018), or recruitment rates, fecundity, or movements (e.g., Jellen and Kowalski 2007). Further, t hough demographic info rmation is available for specific populations, little is understood regarding demographic rates for populations occurring throughout a range of habitat conditions . This, along with population fragmentation (i.e., USFWS 2016 ) , creates difficulty for managers interested in monitoring and evaluating trends and shifts in local populations and modeling population viability. Because massasauga populations in Michigan are considered among the most viable ( e.g., Szymanski 1998 ), Mich igan offers an opportunity to research life history traits pertaining to all life stages. It is imperative that demographic rates are estimated across a range of habitat conditions so that natural resources managers can predict population viability and imp lement appropriate habitat management to help conserve populations . S uch information can be used as performance measures to indicate fitness of the population (e.g., 69 Haufler et al. 2002). A dding information on these vitality rates to currently available li terature would assist managers attempting to assess population viabilit y for a range of habitat qualities . Further, little information is available on dispersal and movements of snake neonates following parturition . Such information is relevant for managers as they assess habitat use by massasaugas ; however, it is difficult to obtain due to difficulties associated with locating and successfully radio tracking snakes of small body mas s . S till, some researchers have successfully collected such data via radio telemetry for neonate rattlesnakes (e.g., Cobb et al. 2005, Jellen and Kowalski 2007). Movement of animals throughout their habitat is related to the availability of habitat components and quality (i.e., suitability ) . Examining movements of indi vidual massasaugas combined with habitat assessment tools (e.g., see Chapter 2) , can indicate habitat suitability and areas of high use within an area. T his information can be used to inform habitat management. When assessing demographic parameters, radio telemetry has been successfully used across multiple taxa including snakes and other relatively small - bodied animals (e.g., timber rattlesnake neonates [ Crotalus horridus ; Cobb et al. 2005 , Howze et al. 2012 ], various bat ) . Telemetry studies involving snakes have mainly resulted in estimates of adult demographic data (e.g., Plummer and Mills 2000, Bailey et al. 201 1 ) , though Jellen and Kowalski (2007) successfully telemetered neonate massasaugas in Pennsylvania by attaching t ® Vetbond Tissue Adhesive. Similarly, k; Cobb et al. (2005) reported that following external attachment of transmitters using a similar adhesive to neonate timber rattlesnakes ( Crotalus horridus ), 4 individuals were s uccessfully tracked from 39 - 42 days in Tennessee. 70 The purpose of our work was to estimate survival rates for massasaugas in southern Michigan . Further, in the interest of supplementing available literature on life history of neonate massasaugas, we report on individual movements of radio telemetered neonate massasaugas and reproductive information ( embryo counts of gravid females, litter sizes ) . STUDY AREA Southern Michigan is a temperate region with moderate spring - summer (i.e., May Aug) tempera tures that ranged from approximately 10 32° C during our study (201 5 to 201 6 in Jackson, MI; NOAA - NWS 2017) . Physiography of the counties within which this research was conducted (Barry, Calhoun, Jackson, Lenawee, Livingston, Oakland , and Washtenaw) consis ts of glacially deposited outwash plains, moraines, till plains, and lacustrine plains (Striker and Harmon 1961, Engberg and Austin 1974, Engel 1977, McLeese 1981, Feenstra 1982, Thoen 1990, Tardy 1997). Soils within our study areas were well drained or po orly drained and loamy with interspersed sandy - loam , loamy sand, or mucky soil types (USDA - NRCS 2017) . We identified 27 study sites based on confirmed reports of massasaugas within the last 25 years (Michigan Natural Heritage Database [MNFI] 2014) . These 2 7 sites represented a range of vegetation types and habitat qualit ies for massasaugas on private and public lands ( Appendix C: Table C. 1 ). Of the 27 sites, 11 occurred on private lands owned by citizens, non - profit conservation groups, or corporations, while the remaining 16 sites occurred on public lands (Appendix C). Sites were 20 ha in size (the maximum home range for a massasauga in south ern Michigan; Bissell 2006) a nd (the maximum distance moved by an individual massasauga in a single season in southern Michigan ; Bissell 2006). 71 METHODS Initial Site Assessment When selecting sites, we coarsely classified habitat quality a priori into low, medium, and high based on proportion of suitable and unsuitable vegetation types for massasaugas within each 20 - ha site . The purpose of this classification was to ensure inclusion of sites of varying habitat structure and composition throu ghout southern Michigan. We used the 2006 National Land Cover Database (hereafter NLDC; MRLC 2015; Appendix C: Table C. 2 ) to identify vegetation types w ithin our study sites. The NLCD includes: open water, developed (open space), developed (low intensity), developed (medium intensity), barren land (rock/sand/clay), deciduous forest, evergreen forest, mixed forest, shrub/scrub, grassland/herbaceous, pasture/hay, cultivated crops, woody wetlands, and emergent herbaceous wetlands (MRLC 2015) . We considered grassland/herbaceous openings, woody wetlands, and emergent herbaceous wetlands as suitable types for massasaugas. We considered open water, development, deciduous forest, evergreen forest, mixed forest, shrub/scrub, pasture/hay, and cultivated crops as non - suitable . We based these suitability groups from the Integrated Forest Management Analysis Program (IFMAP ; MDNR 2001 ) land cover data . W e designated sites with <40% in suitable vegetation types as low site quality, between 40 60% suitable vegetation types as medium, and >60% suitable vegetation types as high quality ( Appendix C: Table C. 2 ). Capture and Marking From May through August 2015 and 2016, we located and captured massasaugas via random encounter surveys within and around study sites. We checked all captured massasaugas for a passive integrated transponder (PIT) tag ( 12 mm in length; AVID; Norco, CA), and newly captured individuals were injected with a PIT tag subcutaneously into the dorsal region 72 approximately 4 6 cm caudal to the cloaca (e.g., Bissell 2006). We transported massasaugas weighing >100 g to a veterinary clinic for surgical implantation of a radio transmitter. While in captivity under veterinary observation, massasaugas were held individually in locked glass - front reptile aquariums ( model S24T; Neodesha Plastics Inc., Neodesha, KS), and provided a place to hide, a dish of water, and sheets of paper towel for additional hiding cover and to absorb liquids (e.g., spilled water, defecation). Veterinary staff surgically implanted massasaugas with one of three radio transmitters of varying weights, depending on body mass (models R1515, 7 g, and R1680, 3.1 g, Advanced Telemetr y Systems [ATS] , Inc., Isanti, MN; model SB - 2, 5.3 g, Holohil Systems Ltd., Ontario, Canada). A licensed veterinarian from the John Ball Zoological Park veterinary clinic in Grand Rapids, Michigan, followed t ransmitter implantation procedures described by Bailey (2010) and Bailey et al. (2011). The combined weight of the implanted radio transmitter and the PIT tag did not exceed 5% of body weight to minimize marking effects (Lentini et al. 2011). Following protocol with the John Ball Zoo, individuals pe r site were implanted with a transmitter each year (further, no more than one of these three was a gravid female). Within 7 days of capture (in most cases 3 4 days), implanted massasaugas were examined and approved for release by the veterinarian and subse quently released at their original capture location (Appendix D) . While in captivity, all female snakes were examined for presence of embryos via palpation and ultrasound or radiograph, and embryo counts were recorded. To determine dates (or a range of dates) of parturition, we noted when the adult female was last observed while gravid and related it to the first observation of a neonate in close proximity to her. We also conducted t elemetry external transmitters attache d with cyanoacrylate - based 73 massasaugas > 100 g that were locate d after the final day of surgery (i.e., after July 31 of both years to ensure sufficient healing of the surgical site prior to hibernation), and for neonate and juvenile individuals weighing < 100 g . These transmitters included ATS model R1635 (0.75 g), an d ATS model A2414 (0.30 g), and again did not exceed 5% of the body weight in combination with the PIT tag (Lentini et al. 2011). All snake capturing, handling and housing methods were approved by Michigan State se Committee (IACUC), protocol 05/15 - 087 - 00, and by the Michigan Department of Natural Resources (MDNR) Fisheries Division (Scientific Collectors Permit #PR8114) . Public land access was approved by the Michigan Department of Natural Resources Permit to Use State Land #PR1136 - 1 . Private land access was arranged directly with landowners prior to the start of any research activity at that property. In accordance with the protocol and requirements outlined in our IACUC approval and MDNR Scientific Collectors Pe rmit, boots and all gear were disinfected using a 10% bleach solution following each day in the field and before moving among field sites to reduce the potential spread of pathogens ( Appendix E ). Survivorship Estimates To estimate survival of massasaugas, we used the Mayfield method (Mayfield 1961) as modified by Bart and Robson (1982). The Mayfield method has been used for massasauga survival estimates (e.g., Bailey et al. 2011), and is recommended for smaller sample sizes , and it allows for staggered entr y assuming constant survival throughout the study period (Winterstein et al. 2001) . Due to the large number of field sites and logistical limitations, visits to sites for tracking massasaugas were not regular (e.g., intervals between visits ranged from 1 t o 57 days ), T he modification to the Mayfield estimator introduced by Bart and Robson (1982) allows for 74 irregular intervals between visits . Bart and Robson ( 1982 ) do not explicitly account for censored individuals and therefore we modified the calculation for censoring (e.g., Bunck and Pollock 1993, Winterstein et al. 2001) and calculated a daily survival rate ( ): where is the number of mortalities observed during the study, L is the maximum interval b etween visits, l is the interval length, is the number of intervals of length l during which a mortality did not occur, h is a fraction of the days during which mortality did not occur ( h = 0.4 for long intervals, as recommended by Bart and Robson [ 1982 ] and Miller and Johnson [ 1978 ], representing the presumed number of days after which death occurred between the second - to - final and final locations ), and is the number of intervals of length l during which a mortality did occur (Bart and Robson 1982 ). Variable , the number of intervals of length l during which an individual was censored, is our modification to the original formula presented by Bart and Robson (1982). A massasauga was cens ored when we could no longer locate the individual , the radio transmitter, or both, and the fate of the individual was unknown . A mortality event was documented when a carcass was located and identified via telemetry or PIT tag, or when we located a transm itter that had been surgically implanted with no sign of the snake (presuming the snake had been depredated and the transmitter discarded) . Using this method to estimate survival , we included massasaugas where 2 locations were collected following release from initial capture or surgery. We pooled data across all field sites and years due to small sample sizes, treating any massasaugas observed in both years independently. 75 RESULTS Estimated Daily and Period Survivorship Massasaugas were located at 9 of th e 27 study sites, 3 of which only resulted in a single massasauga sighting (study sites SE_L2, SE_M1, and SW_H3) . These massasaugas were omitted from further analysis . For the remaining 6 study sites (NE_L1, NE_M1, SW_M2, SE_H1, SE_H3, and SW_H2) at least one massasauga with 2 or more locations was found (Appendix D) . Site NE_L1 was a marginal quality site (HSI = 0.2 8 , Chapter 1) and was where we found a majority of massasaugas included in this analysis ( Table 3. 1 ). Table 3. 1 . Number of adult, juvenile, and neonate eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) used to estimate survivorship throughout southern Michigan in 2015 and 2016 . The habitat suitability index (HSI) represents a suitability score assigned to each site based on the v egetation composition and structure within the site (Chapter 1). Site HSI Adults Juveniles Neonates NE_M1 0.28 10 a 1 3 SW_M2 0.89 5 0 3 SE_H1 0.72 3 2 a 0 SE_H3 0.66 4 2 0 SW_H2 0.8 2 0 0 a 1 individual tracked in both years, treated independently for this analysis Three adult massasauga mortalities occurred in 2015 and 2016 ( Appendix H ) . Years and massasaugas were treated independently and the study period was 137 days in length (26 May to 10 Oct for each year ) . Among all massasaugas including adults, juveniles, and neonates both telemetered and untelemetered ( n = 36; 19 censored, 3 mortalities) , daily survivorship was 0.998 (SE = 0.001, UCI = 1.000, LCI = 0.996 ; Table 3. 2 ) , and period survivorship (i.e., 137 days ) was 76 0.7 85 (SE = 0.01 4 , UCI = 0 . 811 , LCI = 0.7 55 ; Table 3. 3 ) . Results differed slightly when only adult massasaugas ( including both telemetered and untelemetered ; n = 25; 9 censored, 3 mortalities ) were included . D aily survivorship was 0.998 (SE = 0.001, UCI = 1.000, LCI = 0.995; Table 3. 2 ) and period survivorship was 0.767 (SE = 0.016, UCI = 0.796, LCI = 0.732; Table 3. 3 ) . Table 3. 2 . Daily survivorship estimates for the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) at 6 20 - ha study sites throughout southern Michigan in 2015 and 2016. Groups Included a Survival b LCI c UCI c n Censored Mortalities d All Adult, Juvenile, Neonate 0.998 (0.001) 0.996 1.000 36 19 3 All Adult 0.998 (0.001) 0.995 1.000 25 9 3 Telemetered Adult, Juvenile, Neonate 0.998 (0.001) 0.995 1.000 30 13 3 Telemetered Adult 0.998 (0.001) 0.995 1.000 20 4 3 a Group Included indicates massasaugas of which stage were included in the analysis and whether telemetered and b Surviv c LCI, lower 95% confidence interval, and UCI, upper 95% confidence interval. d Observed mortalities. Table 3. 3 . Period (137 days) survivorship estimates for the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) at 6 20 - ha study sites throughout southern Michigan in 2015 and 2016. Group s Included a Survival b LCI c UCI c n Censored Mortalities e All Adult, Juvenile, Neonate 0.785 (0.014) 0.755 0.811 36 19 3 All Adult 0.767 (0.016) 0.732 0.796 25 9 3 All Juvenile, Neonate 0.650 d n/a n/a 11 n/a 0 Telemetered Adult, Juvenile, Neonate 0.761 (0.015) 0.728 0.789 30 13 3 Telemetered Adult 0.739 (0.017) 0.701 0.770 20 4 3 Telemetered Juvenile, Neonate 0.650 d n/a n/a 10 n/a 0 a Group Included indicates massasaugas of which stage were included in the analysis and whether telemetered and b Surviv c LCI, lower 95% confidence interval, and UCI, upper 95% confidence interval. d Estimated using apparent survival due to no observed mortalities among the neonate and juvenile stages. e Total number of o bserved mortalities. 77 W e observed no mortalities among juvenile or neonate stages of the massasauga populations we encountered, likely due to their small size and because radio transmitters could only be attached externally. Therefore, survival could not be estimated using the Mayfield method . Instead, we estimated survival for these groups using apparent survival (i.e., length of the period during which neonates were observed [ 89 days; 5 Jun to 2 Sept ] divided by the total length of the study [137 days; 26 May to 10 Oct]); juveniles and neonates were pooled due to small sample size ( n = 11). Apparent survivorship for the 137 - day p eriod was 0.65 for all neonates and juveniles for 2015 and 2016 ( Table 3. 3 ) and represents a minimum survival rate within this timeframe. Embryo Counts , Litter Size , and Dates of Parturition Embryo counts for gravid female massasaugas ranged from 5 to 18 ( mean = 8.5; Table 3. 4 ) . Observed litters in the field ( mean = 7.5; Table 3. 4 ) . In 2015, we attempted to locate litters from known gravid females at 4 study sites . That year, a litter was first observed at site SW_M2 on 13 August and the telemetered female associated with this litter was last observed the day before on 12 August with no neonates nearby . A t site NE_L1 the first litter was observed on 2 September and the telemetered female (PIT tag # 836560 350; Figure 3. 1) was last observed with no neonates on 28 August . N o neonates were observed at site NE_M1 although the telemetered female was observed periodically into October . Likewise, no litters were observed at site SE_H3 , though we were unable to clo sely monitor any females due to mortality of the sole telemetered gravid female at that site ( Table 3. 4 ; Appendix H ) . 78 Table 3. 4 . Number of embryos, the method by which the count was obtained, and observed neonates following parturition for all gravid female eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) found to be gravid during the 2015 and 2016 field seasons at our study sites throughout southern Michigan. This table does not include litter counts for instances where we were unable to obtain an embryo count for the female and does not include litters found with no associated adult female as this limited our ability to make comparisons between pre - and post - partur ition litter size estimates. Year Site Code a PIT I.D. b # of Embryos Method Date of Observation c Post - Parturition Comments Parturition Date # Neonates Observed d 2015 NE_L1 836560332 8 palpate July 9 Female observed Unknown 0 836560350 radiograph June 30 Female observed with litter 28 Aug 2 Sept 836568071 9 palpate July 9 n/a Unknown n/a 836580078 8 radiograph June 30 Female observed with litter Prior to 4 Sept 4 NE_M1 836550298 ~5 radiograph July 14 Female observed Unknown 0 SW_M2 836561814 e radiograph July 29 Female observed with litter 12 13 Aug 6 - 7 SE_H3 75578601 8 - 9 radiograph May 22 n/a Unknown n/a 100075839 e palpate July 8 n/a Unknown n/a 836546351 8 radiograph May 22 mortality n/a n/a 836550585 e palpate July 12 n/a Unknown n/a 836559770 6 radiograph June 8 n/a Unknown n/a 836567798 ~5 palpate July 8 n/a Unknown n/a 836570859 7 palpate July 8 n/a Unknown n/a 2016 NE_L1 840533265 5 palpate June 9 n/a Unknown n/a 840543607 18 radiograph July 29 Female observed with litter 7 15 Aug SW_M2 840525562 e palpate June 15 n/a Unknown n/a 840533081 8 - 9 palpate June 21 Female observed with litter 9 11 Aug 1 SE_H3 75567567 8 palpate June 1 n/a Unknown n/a 99874823 11 palpate June 1 n/a Unknown n/a 840520549 6 palpate June 1 n/a Unknown n/a 840527780 8 - 10 palpate June 1 Female observed Unknown 0 840544608 e palpate June 30 n/a Unknown n/a SE_M1 836551889 e palpate May 12 n/a Unknown n/a SW_H2 840515348 e radiograph June 24 Female observed with litter 16 24 Aug 4 a Study site code. b PIT (passive integrated transponder) tag identification. c Date that embryo count was obtained. d Number of neonates per litter observed following parturition. e No embryo count obtaine d. 79 Figure 3. 1. Film radiograph image (photographed) taken on 30 Jun 2015 showing 16 embryos (R. Colburn, personal communication) in female eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) from site NE_L1 (PIT tag # 836560350 ; Ap pendix D ) in southern Michigan during the 201 5 field season . Parturition for this female occurred between 28 August and 2 September 2015. On 2 September 2015, 15 neonates were observed in close proximity to this female. Lateral view. Image credits: John Ba ll Zoo, Grand Rapids, MI; R. Colburn. 80 3. 2a Figure 3. 2. Digital radiographs taken on 29 July 2016 showing 18 embryos (R. Colburn, personal communication, 2016) in female eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) from site NE_L1 ( lateral view [2a], dorsal view [2b]; PIT tag # 840543607; Appendix D ) in southern Michigan during the 2016 field season . Parturition for this female occurred between 7 and 15 August 2016. On 15 August 2016, 15 neonates were observed in close proximi ty to this female. Image credits: John Ball Zoo, Grand Rapids, MI; R. Colburn, B. Flanagan. 81 Figure 3. 3. 2b 82 In 2016, we attempted to locate litters from known gravid females at 5 study sites ( Table 3. 4 ) . At site NE_L1, the first neonate was observed on 10 August (the telemetered female thought to be associated with this neonate due to close proximity was last observed with no neonates on 7 August and was again observed with 7 neonates on 15 August; Figure 3. 2). At site SW_M2, the first litter was observed on 12 August, though no adult female was located in close proximity; 3 neonates observed on 16 August at this site exhibited unusual physical characteristics including 2 possible cases of hypo - melanism (J. Harding, personal communication, 2018), and one case of apparent congenital deformity (see Appendix D : Figures D. 1 and D. 2). At site SW_H2 the first litter was observed in close proximity to the associated telemetered female on 24 August (the female, PIT tag # 840515348, was last observed on 16 August and was still gravid at that time). At site SE_H3 no gravid females were implanted with transmitters, limiting our ability to monitor and detect any parturition events , and none of the females we had PIT tagg were observed on 9 August; the female attending this litter was unmarked . Likewise, at site SE_M1, no female massasaugas were telemetered and no neonates were observed. Post - partu rition Radiotelemetry of Neonate s In 2016, we tracked neonate massasaugas from litters located via telemetered adult females . We successfully attached transmitters externally to 4 neonates at site NE_L1 and to 6 neonates at site SW_M2 . All neonates at site NE_L1 were from the same litter while 3 neonates each from two litters were telemetered at site SW_M2 . The number of locations collected among all neonates ranged from 2 to 7 over a span of 2 to 26 days ( Table 3. 5 ) . For 4 of the 10 neonates, at some unknown time following the initial transmitter attachment and release, the transmitter fell off prior to our next return to the site and the second location is based on the location of the transmitter . 83 Table 3. 5 . Radiotracking information for all telemetered neonate eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) obtained during the 2016 field season in southern Michigan. All neonates were fitted with external transmitters weighing less than 5% of their body weight (Lentini et al. 2011). Exact date of parturition is unknown for all neonates. Horizontal lines within the table delineate three different litters based on proximity and date of location (i.e., one litter at NE_L1, and two litters at SW_M 2). Site Code a Snake ID Dates Track ed b # of Locations c Parturition Distance (m) d Total Distance (m) e Comments NE_L1 N_NEL1_2016_10 15 17 Aug f 2 4.5 4.5 n/a N_NEL1_2016_11 15 Aug 10 Sep t g 7 170.2 368.4 Last observed with transmitter on 2 Sept. N_NEL1_2016_12 15 25 Aug 5 445.9 551.2 No visual observation at final location but thought to be moving while being tracked. N_NEL1_2016_13 17 25 Aug 4 259.8 275.1 Observed with transmitter at final location. SW_M2 N_SWM2_2016_1 12 16 Aug f 2 1 1 n/a N_SWM2_2016_2 12 18 Aug f 3 95.6 96.2 No visual observation following initial release, thought to be burrowed. N_SWM2_2016_3 12 16 Aug f 2 93.2 93.2 n/a N_SWM2_2016_4 16 18 Aug f 2 11.2 11.2 n/a N_SWM2_2016_6 16 24 Aug 3 41.1 62.7 Observed with transmitter at final location. N_SWM2_2016_7 16 24 Aug g 3 158.7 165.8 Last observed with transmitter on 18 August. a Study site code. b Range of dates that the neonate was tracked. First date indicates the d ate of capture, transmitter attachment, and release . Second date indicates the date that the final location was collected, based on the location of a dropped transmitter or the telemetered neonate. c Number of locations collected, including the initial capture location and the final location, whether the n eonate was observed or not . d Total linear distance from gestation site to the final telemetry location . e Total linear distance between successive telemetry locations, representing the minimum distance moved by the individual durin g the timeframe within w hich it was tracked. f O nly transmitter found at final location (i.e., had fallen off neonate); neonate not located. g No neonate or transmitter observed at final location (i.e., transmitter signal was heard but underground and unrecoverable) 84 At site NE_L1 on 15 August, a litter of 15 neonates was located in close proximity to a telemetered female ( PIT tag # 840543607 ; Table 3. 4 ; Figure 3. 2; Appendix D ); parturition had likely occurred at some point since 7 August . 18 embryo s (R. Colburn, personal communication, 2016) , the greatest number of embryos counted during this study. Three neonates from this litter were fitted with external radio transmitters on 15 August ; on e transmitter fell off and was recovered within two days, and on 17 August another neonate from this litter was fitted with a transmitter . For one neonate from this litter, we obtained 7 total locations spanning a linear traveled distance of 368.4 m over 2 6 days (the last telemetry location of this neonate was 170.2 meters from the gestation site ; Table 3. 5 ) but the fate of this neonate is unknown as it was not observed at the final location and the transmitter was not recovere d. The transmitter remained at tached to this individual for at least 18 - 26 days and was the longest span of time during which a neonate was telemetered in our study . Another telemetered neonate from this litter traveled a linear distance of 551.2 m within 10 days of tracking, the furth est distance moved following parturition recorded among all neonates telemetered in our study. The directionality of movement differed among these three neonates; while neonate N_NEL1_2016_11 generally trailed after the mother (i.e., southeastward of the gestation site), neonates N_NEL1_2016_1 2 and N_NEL1_2016_1 3 moved eastward . At study site SW_M2, no adult female massasauga was located in cl ose proximity to the unmarked female had been located, PIT tagged, and fitted with an external radiotransmitter on 12 Aug within 2 meters of that location (PIT ta g # 840543057; Appendix D). On 16 Aug, the transmitter that had been attached to the female was found in close proximity to the neonates, but no adult massasauga was found. Interestingly, neonates N_SWM2_2016_2 and N_SWM2_2016_3 moved east and slightly nor th of their gestation site and were relocated 85 among the neonates from the second litter at this site. For the second litter, n eonate N_SWM2_2016_7 moved northward and slightly east at a considerable distance compared to its littermates, while neonate N_SWM 2_2016_6 moved southeast from the gestation site. DISCUSSION Our estimate s of adult period survival (0.767) and minimum apparent survival for the neonate and juvenile group (0.65) are comparable to estimates for other massasauga populations . Hileman et al . (2018) report ed that for a massasauga population in southwestern Michigan, rate of survivorship increase d with age, which corroborates our findings and is reported for other reptile species (e.g., Congdon et al. 1994 ) . Hileman et al. (2018) estimated annual adult survivorship at 0.66 for males and 0.71 for females . Apparent survival for ages 0, 1, and 2 was 0.38, 0.65, and 0.67, respectively (Hileman et al. 2018) . Likewise, Jones et al. (2012) report ed that active season (comparable to our study period of May through October) survivorship was 0.77 rangewide . Our estimates of period adult survival were considerably lower than adult active season survival estimate of 0.947 reported by Bailey et al. (2011) for massasaugas in southern Michigan using the May field method . Bailey et al. (2011) suggested t hat their relatively high estimate may be because the study area is managed specifically for massasauga s , due to a lack of human development in the area , or potentially a result of low sample size . Further, our estimate was obtained at sites ranging from marginal to optimal suitability (Chapter 1) . The 6 study sites from which we obtained survivorship estimates ranged from marginal to optimal suitability (1 marginal, 3 good, 2 optimal) based on the Bailey (2010) habitat suitability index model for the massasauga in southern Michigan (Chapter 1) , yet suitability ranking for these 20 - ha study sites do es not necessarily correlate with population viability . Thus, we note the importance of considering scale when relat ing habitat suitability to population fittness (e.g., Wiens 1989, Gaillard et al. 2010). For example, w hile one site, NE_L1, was ranked as 86 marginal ly suitable for the massasauga, patches of optimally suitable habitat occurred within a matrix of unsuitable vegetation types and this site is where we observed the two largest litter sizes documented during this study. This finding suggests that massasaugas are more strongly influenced by micro - habitat quality (i.e., vegetation structure and composition) as oppo sed to macro - habitat attributes (e.g., predominant vegetation type within a defined area) . Habitat quality has been identified as a driving factor for population viability including increased survival (e.g., Bailey et al. 2011) , though Johnson et al. (2016 ) found no change in massasauga reproductive rates or survival when basking site quality was improved for a New York population . Embryo counts obtained during our study ranged from 5 18 . Bailey (2010 , unpublished data), observed an average embryo count of 6.6 (ranging from 1 - 10) for a massasauga population in southwestern Michigan using counts from radiograph a nd palpation . Our largest counts (based on radiograph imagery) were 16 (Figure 3. 1 ) and 18 (F igure 3. 2 ) , both in massasaugas from the same study site (NE_L1) in 2015 and 2016 respectively . F ollowing parturition for both females, we documented 15 neonates for each litter . For massasauga parturitions monitored during our study, litter sizes ranged from 1 to 15 (mean = 7.5) neonates based on field observations . Our findings are consistent with previously reported litter sizes, for example, Bissell (2006) reported a litter size range of 4 - 12 (mean = 9) for a southern Michigan population, though thes e counts included both the number of observed neonates at a gestation site and the number of viable embryos counted via ultrasound s . Additional reports range from 2 - 16 (mean = 7.6, captive - born litters in Michigan ; Hileman et al. 2018), 5 7 for a Wisconsin population (Reinert 1981), and 4 10 for a Missouri population (mean = 6.35; Seigel 1986) . T he se authors did not report whether litter counts (both pre - and post - parturition) included non - viable offspring which may be due to the difficulty associated with identifying non - viable individuals using any method of counting, though Baker (2016) 87 observed total litter sizes for captive massasaugas ranging from 2 - 15 (mean = 7.74) and viable offspring ranging from 1 15 (mean = 6.67) . We did not obs erve any non - viable offspring in the field and were unable to confirm whether all neonates were accounted for when we located litters in the field . Our two observations of 15 viable neonates is on the upper end of litter sizes that have been reported (e.g. , Reinert 1981, Bissell 2006, Hileman 2018) and considering the number of embryos counted for these females (i.e., 16 and 18), is similar to overall embryo viability rate of 85% reported by Baker (2016) . 15% of embryos were found t o be nonviable across all massasaugas examined in the study . It is important to note that litter counts can vary depending on the method with which they were obtained . For example , in a study comparing effectiveness of ultrasound, radiography, palpation, and post - parturition litter observation , Bissell (2006, unpublished data) found that for massasauga s , radiography, palpation, and post - parturition litter counts consistently underestimated litter size compared to ultrasound which was found to be the most a ccurate . Likewise , in a study by Donini et al. (2017), accuracy of egg or follicle counts for Diamondback Terrapins ( Malaclemys terrapin ) was greater with ultrasound than with radiography . In the current study, we did not use ultrasound for litter counts a nd always relied on palpation in the field or on film radiograph or digital radiograph imagery when massasaugas were transported to the veterinary clinic for surgical implantation of transmitters . Thus, it is likely that for female massasaugas that were no t selected for transmitter implantation, palpation counts resulted in underestimated litter size ( Table 3. 4 ); Bissell et al. ( 2006, unpublished data) reported that the mean difference between litter counts obtained via palpation and ultrasound was 2.5 . In the current study all radiographs were produced using film - based radiographic techniques in 2015; after 2016, our collaborators upgraded to a digital radiography system (e.g., Figure 3. 1 , Figure 88 3. 2 ) . In a medical study , researchers found no significant di fferences between diagnostic readings of film versus digital radiograph imagery (e.g., Franzblau et al. 2018 ). Parturitions for massasauga s monitored in our study occurred during August, with one instance possibly occurring the first two days of September . Likewise, Jellen and Kowalski (2007) report that for a western Pennsylvania massasauga population, parturition occurred between 5 and 27 August 2015 . Foster et al. (2009) found that the mean date of parturition in 200 4 and 2005 was 17 August for a southern Michigan population . Further, parturition date may vary based upon environmental conditions (e.g., Lourdais et al. 2004 ). Jellen and Kowalski (2007) found that for telemetered massasauga neonates in western Pennsylvania , average total distance mov ed within 42 days was 257.3 m (± 88.6 m) , and at the end of the period average distance from gestation site was 83.3 m (± 44.5 m ) . Our distances traveled by neonate massasaugas exceed these numbers (i.e., site NE_L1 in 2016), but such movements may be infl uenced by environmental factors such as temperature (e.g., Eskew and Todd 2017) and habitat quality or vegetation structure and its distribution throughout a site ( e.g., Kernohan et al. 2001 , Durbian et al. 2008 ) . In 2015 and 2016, site NE_L1 was classifie d as a poor - marginal quality site for massasaugas ( see Chapter 1) comprised of a patchy matrix of suitable and unsuitable vegetation types , and this is where we observed the greatest neonate movements observed during this study (i.e., 551.2 m moved within 10 days) . Conversely, at site SW_M2, a good - optimal suitability site containing homogenous suitable habitat, neonate movements never exceeded 165.8 m within 8 days. Our results offer demographic rates and life history information for massasaugas at multip le sites ranging in habitat quality from marginal to optimal throughout southern Michigan. W e offer this analysis as a contribution to the knowledge base for massasaugas within this part of the range, for populations within areas identified as poor - margina l quality for the species (e.g., 89 site NE_L1) , and to add information for previously unstudied and undocumented populations . The reproductive success of massasauga populations occurring in geographically small and threatened areas (e.g., due to late success ional encroachment) highlights the importance of identifying and conserving such seemingly insignificant populations . The application of appropriate habitat management (e.g., the creation of corridors, reversing succession) can promote genetic exchange amo ng such populations , thereby improving population viability and ensuring the long - term success of the species throughout the state. ACKNOWLEDGMENTS We acknowledge the support of Michigan State University Department of Fisheries and Wildlife, and G. Roloff and J. Tsao for their contributions to this work. Funding for this project was provided through the Michigan Department of Natural Resources State Wildlife Grant # F15AP00096 in cooperation with the U.S Fish and Wildlife Service, Wildlife and Sport Fish Restoration Program. We thank K. Bissell, A. Derosier, M. D onnelly, R. Fahlsing, C. Michigan Department of Natural Resources. We thank J. Dingledine and S. Hicks from the U . S . Fish and Wildlife Service . We thank the employees and veterinary staff of John Ball Zoo, Grand Rapids, Michigan, including R. Colburn, B. Flanagan, and H. Teater. We also thank T. Meyers Harrison for veterinary assistance. We thank S. and C. Weaver and J. Buck; S. Leavitt, C. May, and R. Villegas from The Nature Conservancy; T. Funke and R. Roake from Michigan Audubon Society; Y. Lee from the Michigan Natural Features Inventory; and M. Dreslik at the University of Illinois for their persistent support of our project. We thank field technicians T. Brockman, B. Brodowski, C. Burden, G. Payter, and H. Reynolds (Britz) . 90 CHAPTER 4: CONCLUSIO NS AND MANAGEMENT RE COMMENDATIONS DISSERTATION OVERVIE W In Chapter 1 of this dissertation, I presented a validation of the Bailey (2010) habitat suitability index (HSI) model for the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) at 27 20 - ha study sites of varying quality throughout southern Michigan on private and public lands. Disproportionate use by massasaugas of the habitat elements defined as importa nt by Bissell (2006) and Bailey (2010) was illustrated via resource selection probability function analysis, showing a positive relationship between probability of use and the amount of live and dead herbaceous cover and a negative relationship between pro bability of use and DBH and number of woody stems 3 m in height. Further, I reported predicted occupancy rates based where suitability was most optimal (i.e., HS I = 1), suggesting that I can expect 50% of the sites at that level of suitability to be occupied by massasaugas within southern Michigan. This work describes methods that may be used by managers in habitat assessments for the massasauga, and a way to iden tify habitats with the greatest conservation value (i.e., habitats where HSI scores indicate a high probability of occupancy). Chapter 2 identifies environmental and searcher variables that should be considered when attempting to detect massasaugas in the ir habitats; specifically, minimum temperature should be considered (detection probability approaches 0.8 when minimum temperatures reach 12.8 °C ), and time spent searching should exceed an hour (detection probability approaches 1 at occupied sites when 1. 5 hours are spent searching within a 2 - ha area ). Further, I offer a standardized survey methodology that can be applied at sites where massasaugas may exist 91 throughout a range of habitat suitabilities. This work offers a method for detecting massasaugas wi thin their habitats and can aid managers attempting to maximize their detection probability. Chapter 3 describes my findings on the demographic and life history characteristics for massasaugas throughout my study sites in southern Michigan in sites ranging in suitability from marginal to optimal based on the HSI modeling in Chapter 1. Survivorship estimates are available for adult massasaugas rangewide (e.g., Jones et al. 2010) yet little is available in the literature regarding neonate or juvenile survivor ship or movements . Further, little is understood regarding demographic rates for populations occurring throughout a range of habitat conditions . The aim of this work was to address some of these data gaps at sites of varying suitability . I reported daily a dult survivorship (0.998), active - season adult survivorship (0.767), and minimum active season survivorship of juveniles and neonates ( 0.65). I reported average litter sizes (mean = 7.5, ranging from 1 - 15), embryo counts (mean = 8.5; ranging from 5 - 18), a nd summarized neonate movements following parturition based on radio tracking. Ten neonates were tracked for 2 - 26 days, moving distances up to over 551 m. As a federally Threatened species (USFWS 2016), conservation of massasauga populations and both known and potential habitats are imperative for the future of the species. Knowledge of habitat use, habitat composition and structure, and life history informa tion including movements, demographic rates, and reproduction in habitats of varying quality is necessary for understanding and defining massasauga populations . Based on the findings of this work and my evaluation of massasauga population and habitat ecolo gy, I offer management recommendations for conservation of the massasauga and its habitat throughout southern Michigan on public - and privately - owned lands . 92 RECOMMENDATIONS FOR HABITAT MANAGEMENT A ND POPULATION ASSESSMENT Identifying Potential Massasauga Sites For natural resource managers interested in locating potential massasauga sites throughout southern Michigan, I recommend first referring to recent and historical reports of observations of the species within and around the region of interest. When i dentifying specific locations for further investigation, I recommend remotely using a landcover database (e.g., Integrated F orest M anagement A nalysis P rogram in suitable vegetation or cover types for massasaugas (e.g., early successional deciduous or wetland vegetation types containing a high proportion of open, herbaceous vegetation with limited overstory cover; Bailey [2010]). For example, using the NLCD data base at my study sites in southern Michigan, I identified areas containing suitable massasauga cover types which included grassland/herbaceous openings, woody wetlands, and emergent herbaceous wetlands (MRLC 2015). Further, using the IFMAP database which p rovides more detailed vegetation classification that NLCD, I identifi ed the specific vegetation types that were available within my study sites. For identifying massasauga habitat using IFMAP, suitable early successional upland vegetation types include for age crop, herbaceous openland, row crop, and upland shrub; suitable early successional wetland vegetation types include emergent wetland, floating aquatic, lowland shrub, and mixed non - forest wetland (MDNR 2001). Following a remote assessment of the avail able vegetation types, I recommend following up with ground truthing efforts to confirm that vegetation types in the database are representative of what currently exists at the site. This validation may involve a simple walk - through of the site to confirm that specific vegetation types of interest are still present at the location. Recent 93 management, development, or successional changes may result in stark differences between the database and current vegetation composition and structure. While in the field, microhabitats conducive to massasauga use can be located (e.g., high percentages of live or dead herbaceous cover, limited overstory; Chapter 1, Bailey 2010). Assessing Habitat Suitability For managers interested in assessing suitability of massasauga ha bitats to guide management, the Bailey (2010) HSI model can be used in its full form (i.e., accounting for all vegetation composition and structural variables; Chapter 1), using only the landscape level variables (% area in early successional deciduous upl and or wetland areas), or using only the thermal stand - level variables (% live and dead herbaceous cover, basal area of trees, and stem density). Where resources are limited, an assessment using only the landscape - level variables can be conducted using GIS software and a landcover database and do es not require the user to be present in the field. Likewise, the thermal variables may be used when landcover data is unavailable, which may be the case for some private lands, or for areas that have rece ntly undergone extensive management. As illustrated in Chapter 1, the resulting HSI score can then be used to determine the probability of occupancy for a given area. This method can be used in conjunction with my recommendations for identifying potential massasauga sites (see previous section), further indicating the likelihood of a site being occupied or suitable for massasaugas. Detecting Massasaugas I recommend that detection surveys, as described in Chapter 2, be implemented in areas where massasaugas are thought to exist (e.g., based on sighting reports or the availability of suitable vegetation types, composition, and structure, as defined in Chapter 1). Surveys should be carried out during the active period ( i.e. , spring and summer) when there is suf ficient time to 94 spend at least 1 - 1.5 hours searching (e.g., to maximize detection probability; Chapter 2) per area of interest, and I recommend keeping the survey area to 2 ha (i.e., the smallest annual home range for massasaugas in southern Michigan; Biss ell 2006) so the area can be sufficiently searched within that timeframe. I do not recommend conducting a survey at more than 2 2 - ha areas per day, per person, as this resulted in considerable fatigue for searchers during the pilot work in 2015. Finally, i t is imperative that suitable temperatures be considered when deciding on a day and/or time for a survey. In southern Michigan, detection probabil ity is maximized (e.g., approaching 0.8) at 12.8 °C , is approximately 0.5 at 17 °C , and approach es 0 .0 when te mperatures exceed 24 °C . Considering the greater detection probability at cooler temperatures, I recommend searching earlier in the season (May, June) throughout the day, or in the morning during the hotter months (July, August) when temperatures are coole r, as a general rule. Habitat Management Though I did not assess impacts of habitat management or management practices on massasauga populations as a part of this study, I summarize findings in the literature to make recommendations for management practice s and future research . Habitat management is an integral component of maintaining and restoring suitable massasauga habitat (e.g., Johnson et al. 2000, Cross et al. 2015), especially considering the threat of habitat degradation (i.e., driven by succession ) impacting massasauga populations (USFWS 2016). Important methods for maintaining early successional vegetation types include mowing, targeted shrub removal, and burning (e.g., Johnson et al. 2000) , but a s a relatively sedentary, slow - moving species, mass asaugas may be at an increased mortality risk from such management activities (e.g., mowing, burning; e.g., Durbian 2006, Cross et al. 2015) . As such, managers may be interested in reducing incidental mortality of massasaugas resulting from management acti vities. 95 Cross et al. (2015) recommended that burns occur prior to emergence from hibernation or following movement into hibernation sites , that burns not be conducted around known hibernacula during ingress or egress, and that burns be conducted in the wi nter months when possible to avoid risk to massasaugas. For management occurring outside of this timeframe (i.e., during the active season), the authors state that burns be patchy to provide suitable areas for massasaugas to relocate to during the burn, th at burn areas not be in close proximity to one another, and that refugia such as brush piles be removed from burn areas to reduce massasauga mortality. For mowing, I recommend that mowing decks be set high to prevent massasauga mortalities ; s pecifically, Johnson et al. (2000:24 - 25) recommend mowing at least 10.16 - 15.24 4 - 6 inches above the ground to avoid massasaugas that may be present . Johnson et al. (2000) also recommend mowing - 3 p.m , stating that this time is when most massasaugas are under cover. Further, the authors recommend that mowing take place prior to emergence in the spring or following hibernation in the fall (Johnson et al. 2000:24). High s tem density (stems/ha of trees and shrubs >3 m in height) was often the greatest contributing factor limiting habitat quality for massasaugas when I assessed habitat at the 27 20 - ha study sites throughout southern Michigan in 2015 and 2016 (i.e., Chapter 1: Table 1. 1 ) . Specifically, f or most of these sites, stem density estimates for all sites but one exceeded the optimal range (0 - 58 stems / ha of trees and shrubs >3 m in height ) , and 16 sites had stem density estimations 800 stems per ha ( i.e., at which suitability = 0 ; Bailey 2010). Likewise, Shoemaker and Gibbs (2010) reported that decreased canopy cover (i.e., via cutting shrubs to 0.5 - 0.25 m in height) improved basking opportunities for massasaugas for a New York population. As such, I suggest r educing stem density at sites to within optimal (0 - 58 stems/ha) range of suitability via 96 shrub and tree removal at many sites as a primary step in improving massasauga habitat . As discussed in Chapter 1, reduction of this variable alone can result in improvement of the other h abitat variables (e.g., increasing the amount of live and dead herbaceous cover) over time, further improving the suitability of the habitat. Shoemaker and Gibbs (2010) recommend that shrubs be cut to 0.25 - 0.5 m in height to decrease canopy cover and to al low for improved basking opportunities, while Johnson et al. (2016) recommend cutting shrubs to 0.25 m in height. Where habitat management is warranted for the massasauga, I recommend tree and shrub removal prior to implementing other management activitie s since this factor will likely positively influence other habitat attributes. This method is relatively simple to implement (i.e., removal of trees and shrubs) compared to the use of fire, and when possible, stems may be left on - site to create brush piles , offering refugia for massasaugas. FUTURE RESEARCH Future research for the massasauga throughout its range should seek to include not only well - studied or heavily - documented populations occurring in high quality habitats, but also smaller populations in lower - quality habitats that may otherwise remain undocumented ( e.g., such as our study site NE_L1; Chapter 1). While heavily studied populations offer invaluable information and insight into the ecology, habitat, and dynamics of the species, smaller popula tions in poor quality habitats similar to our study site NE_L1 are equally important potential metapopulation dynamics . H abitat management to improve these areas can increase connectivity throughout the landscape by creating additional suitable habitat , ultimately decreasing the opportunity for population isolation and extirpation . Further, population dynamics of massasaugas within lower - quality sites should be assessed (i.e., survivorship, fecundity) as indicator s of population fitness. Many reports of massasauga survivorship come from well - studied populations in high quality habitats (e.g., Jones et al. 2012), 97 range. Massasauga movements should continue to be examined to determine individual and population home ranges and to further identify dynamics between movement and habitat quality. The movement of massasaugas through poor quality habitats may differ from movement s occurring in high quality habitats (i.e., differing spatial distribution of resources and habitat components within poor - versus high - quality habitats) and can inform researchers of how massasaugas are using specific habitat components throughout the spe the habitat suitability index model described in detail in Chapter 1 and by Bissell (2006) and Bailey (2010) is applicable throughout southern Michigan, but may be limited in its usefulness in other parts of the species range (e .g., at the northernmost or western limits) due to differences in geomorphology, hydrology, and land cover or vegetation types, and may require modification for certain populations to address these limitations. As such, the continued study of massasauga mo vements and habitat use rangewide is imperative. 98 APPENDICES 99 APPENDIX A Dissemination of Findings and Related Information The following documents presentations, publication, and outreach where findings or information related to this project were disseminated to the scientific community or the public. Peer - Reviewed Publication Shaffer, S.A., G. Roloff, H. Campa, III. In Progress . Survey methodology for detecting eastern massasauga rattlesnakes in southern Michigan. Wildlife Society Bulletin . Presentations: Professional Meetings - The Wildlife Society 24th Annual Conference o September 26, 2017; Albuquerque, New Mexico o o Authors: Stephanie A. Shaffer, Henry Campa, III, Gary Roloff, Daniel Kennedy - Michigan Partners in Amphibian and Reptile Conservation o March 22, 2017; Belle Isle Nature Zoo; Detroit, Michigan o o Authors: Stephanie A. Shaffer, Henry Campa, III, Gary Roloff, Daniel Kennedy - 12th Annual Department of Fisheries and Wildlife Graduate Student Research Symposium o February 24, 2017; Michigan State University; East Lansing, Michigan o o Authors: Stephanie A. Shaffer, Henry Campa, III, Gary Roloff, Daniel Kennedy 100 - Midwest Fish and Wildlife 77 th Annual Conference o February 7, 2017; Lincoln, Nebraska o o Authors: Stephanie A. Shaffer, Henry Campa, III, Gary Roloff, Daniel Kennedy - Midwest Fish and Wildlife 76 th Annual Conference o January 26, 2016; Grand Rapids, Michigan o uthern o Authors: Stephanie A. Shaffer, Henry Campa, III, Gary Roloff, Daniel Kennedy - 10 th Annual Department of Fisheries and Wildlife Graduate Student Research Symposium o February 27, 2015; Michigan State University; East Lansing, Michig an o Assessing the Status, Habitat Quality, and Management of Eastern Massasauga Rattlesnake Populations in Michigan o Authors: Stephanie A. Shaffer, Henry Campa, III Presentations: Public Engagement and Outreach - Sierra Club Michigan Chapter; Cro ssroads Group Meeting o September 20, 2017; Brighton District Library; Brighton, Michigan o There are Rattlesnakes in Michigan? o Authors: Stephanie A. Shaffer, Henry Campa, III - Huron Valley Audubon Society Meeting o October 10, 2016; Kensington Nature Center; Milford, Michigan o Conservation of the Eastern Massasauga Rattlesnake in Michigan o Authors: Stephanie A. Shaffer, Henry Campa, III 101 - Wildlife Weekend o August 8, 2015; Indian Springs Metropark; White Lake, Michiga n o The Eastern Massasauga Rattlesnake in Michigan o Authors: Stephanie A. Shaffer, Henry Campa, III Media Engagement - News Coverage: "MSU Researchers Work to Protect Rattlesnakes" o July 2016; WILX News 10; Lansing, Michigan o - Video Feature: "Sustaining a Ssssssspecies" o 2016; MSU Today; Michigan State University; East Lansing, Michigan o 102 APPENDIX B Snake F ungal D isease S ampling R esults Stephanie A. Shaffer, Henry Campa, III Introduction Snake fungal disease, a fungal pathogen ( Ophidiomyces ophiodiicola , formerly Chrysosporium sp SFD ) is a recently emerging potential threat to snake populations (Allender et al. 2015). The spread of such a pathogen holds important implications, especially considering that many massasauga populations remain fragmented and isolated (USFWS 2016). Still, little is known about SFD (Allender et al. 2011 , 2015 ) though the number of reports on the pathogen have increased in recent years (e.g., Dolinski et al. 2014 , Allender et al. 201 5 ) . Snake fungal disease has been detected in multiple snake species across N orth America, Europe, Asia, and Australasia (Sigler et al. 2013) . In the United States, reported instances of infection include the plains garter snake in Illinois ( Thamnophis radix ; Dolinski et al. 2014), black rat snake in Georgia ( Elaphe obsoleta obsole ta ; Rajeev et al. 2009), and eastern massasauga ( Sistrurus catenatus catenatus ) in Illinois (Allender et al. 2011) . Across multiple snake taxa, potential instances of infection include symptoms such as lumps or bumps under the skin (i.e., granuloma), lesi ons or ulcers on the skin, inflammation, - up under scales or scutes (Rajeev et al. 2009 , Allender et al. 2011, 2016, Dolinski et al. 2014) . It is also important to note that the pathogen may also be present upon the skin of an individual showing no physical signs of infection ( Allender et al. 2016 ) . Snake fungal disease has been confirmed in Michigan ( e.g., Barry, Cass, Crawford, and Kalkaska counties; Allender et al. 2016 ) . I n massasaugas specifically , this disease has been 103 reported to affect individuals around the nasolabial region (i.e., heat sensing pits) which can result in facial swelling and ulcers, and a distortion of mouth and eyes in more severe cases (Allender et al. 2011 , 2016 ) . Considering the conservatio n status of massasaugas (i.e., USFWS 2016), f urther assessment of SFD prevalence is warranted . Study Area Southern Michigan is a temperate region with moderate spring - summer (i.e., May Aug) temperatures that ranged from approximately 10 32° C during our st udy (201 5 to 201 6 in Jackson, MI; NOAA - NWS 2017) . Physiography of the counties within which this research was conducted (Barry, Calhoun, Jackson, Lenawee, Livingston, Oakland , and Washtenaw) consists of glacially deposited outwash plains, moraines, till pl ains, and lacustrine plains (Striker and Harmon 1961, Engberg and Austin 1974, Engel 1977, McLeese 1981, Feenstra 1982, Thoen 1990, Tardy 1997). Soils within our study areas were well drained or poorly drained and loamy with interspersed sandy - loam , loamy sand, or mucky soil types (USDA - NRCS 2017) . We identified 27 study sites based on confirmed reports of massasaugas within the last 25 years (Michigan Natural Heritage Database [MNFI] 2014) . These 27 sites represented a range of vegetation types and habitat qualit ies for massasaugas on private and public lands ( Appendix C: Table C. 1 ). Of the 27 sites, 11 occurred on private lands owned by citizens, non - profit conservation groups, or corporations, while the remaining 16 sites occurred on public lands (Appendi x C). Sites were 20 ha in size (the maximum home range for a massasauga in southern Michigan; Bissell 2006) a nd (the maximum distance moved by an individual massasauga in a single season in southern Michigan ; Bissell 2006). 104 Methods Capture and Marking From May through August 2015 and 2016, we located and captured massasaugas via random encounter surveys within and around study sites. We checked all captured massasaugas for a passive integrated transponder (PIT) tag ( 12 mm in length; AVID; N orco, CA ), and newly captured individuals were injected with a PIT tag subcutaneously into the dorsal region approximately 4 6 cm caudal to the cloaca (e.g., Bissell 2006). Sample Collection Captured massasaugas were immediately examined for clinical sign s of SFD . Swab samples were collected from all massasaugas regardless of clinical signs (at least one sample in 2015, and at least two samples one upon initial capture, and one at a later date - in 2016 when possible; e.g., Hileman et al. 2018). If clini cal signs of SFD were present, a photograph was taken, and information provided to Dr. Matt Allender (including swab samples and photographs) as part of the data reporting process. Massasaugas were sampled at first, upon recommendation by Dr. Allender, by swabbing the nasolabial pits . This method proved to be impractical (and unsuccessful in many instances due to the difficulty of restraining the snake in a tube while attempting to swab its face ), and so nasolabial pit swabs were only taken if there were clinical signs of SFD on the face of the massasauga in question . Otherwise, swab samples were taken from along the back and belly of an individual showing no clinical signs and directly from any locations on the body where clinical signs were present (i.e. , lesion, lump, scab, or scar; each location sampled individually) . In all instances, massasaugas were sampled while in the field, prior to transportation for radio transmitter implantation. All snake capturing, handling and housing methods were approved by Michigan State - 087 - 00, 105 and by the Michigan Department of Natural Resources (MDNR) Fisheries Division (Scientific Collectors Permit #PR8114) . Public land access was approved by the Michigan Department of Natural Resources Permit to Use State Land #PR1136 - 1 . Private land access was arranged directly with landowners prior to the start of any research activity at that property. In accordance with the protocol and requirements ou tlined in our IACUC approval and MDNR Scientific Collectors Permit, boots and all gear used in the field were disinfected using a 10% bleach solution following each day in the field and before moving among field sites to reduce the potential spread of path ogens ( Appendix E ). Results In 2015, 24 swab samples were collected and a single massasauga tested positive for SFD (Table B . 1). This male was captured and sampled at site SE_H1 on 17 June 2015 and surgically implanted with a radiotransmitter but found dea d (presumed depredation; Appendix H) on 31 July 2015. In 2016, 46 swab samples were collected and two massasaugas tested positive for SFD, a juvenile male and an adult nongravid female (Table B . 1). The juvenile male was captured and sampled at site SE_H1 on 5 July 2016 and was tracked to hibernation on 11 October 2016 (in 2015 this individual had been fitted with an external radiotransmitter and sampled for SFD, testing negative in that year). Th e adult female for which there was a positive SFD result in 2016 (based on a single sample collected) was captured at site SW_M2 and fitted with an external radiotransmitter on 15 Jun 2016 (the transmitter was found unattached the following day and the mas sasauga was never located again); a swab sample collected from this individual in 2015 tested negative for SFD. 106 Table B . 1. All eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) sampled in 2015 and 2016 throughout southern Michigan for snake fungal disease ( Ophidiomyces ophiodiicola ) and outcome of testing for each (positive [+] , negative [ - ] ). Multiple samples were collected when an individual exhibited clinical signs of infection by the fungal pathogen. Year Site Code Snake Identification Date of Sample Sample Record # Number of Samples Sample Vial Identification Test Results 2015 NE_L1 836519789 25 - May - 15 4 1 NEL1_01 - 836560332 9 - Jul - 15 16 1 NEL1_05 - 836560350 29 - Jun - 15 10 1 NEL102 - 836568071 9 - Jul - 15 15 1 NEL1_04 - 836580078 29 - Jun - 15 11 1 NEL103 - 836571332 840382378 13 - Jul - 15 19 1 NEL1_06 - NE_M1 836550298 13 - Jul - 15 18 1 NEM1_01 - SW_M2 836560014 29 - Jul - 15 21 1 SWM2_02 - 836561814 28 - Jul - 15 20 1 SWM2_01 - 836564778 29 - Jul - 15 22 2 SWM3_03_A, SWM3_03_B - SE_H1 836551570 17 - Jun - 15 9 1 SEH1_02 - 836574862 17 - Jun - 15 8 1 SEH1_01 + SE_H3 75578601 21 - May - 15 1 1 SEH3_01 - 99873268 22 - May - 15 3 1 SEH3_03 - 100075839 8 - Jul - 15 14 1 SEH3_08 - 836546351 21 - May - 15 2 1 SEH3_02 - 836549811 7 - Aug - 15 23 1 SEH3_10 - 836550585 12 - Jul - 15 17 1 SEH3_09 - 836559770 5 - Jun - 15 6 1 SEH304 - 836567798 8 - Jul - 15 12 1 SEH3_06 - 836570859 8 - Jul - 15 13 1 SEH3_07 - J_SEH3_2015_1 5 - Jun - 15 7 1 SEH305 - SW_H2 836547069 1 - Jun - 15 5 1 SWH0201 - 2016 NE_L1 836519789 17 - Aug - 16 136 1 NEL1_78 - 840380005 19 - May - 16 104 1 NEL1_09 - 4 - Aug - 16 126 1 NEL1_68 - 840521571 4 - Aug - 16 127 1 NEL1_69 - 840533265 9 - Jun - 16 113 1 NEL1_11 - 1 - Aug - 16 125 1 NEL1_67 - 840543607 28 - Jul - 16 124 1 NEL1_66 - 836571332 840382378 19 - May - 16 105 1 NEL1_10 - 107 Table B Year Site Code Snake Identification Date of Sample Sample Record # Number of Samples Sample Vial Identification Test Results 17 - Aug - 16 137 1 NEL1_79 - J_NEL1_2016_1 19 - May - 16 103 1 NEL1_08 - N_NEL1_2016_1 15 - Aug - 16 135 1 NEL1_72 - N_NEL1_2016_2 15 - Aug - 16 135 1 NEL1_73 - N_NEL1_2016_3 15 - Aug - 16 135 1 NEL1_74 - N_NEL1_2016_4 15 - Aug - 16 135 1 NEL1_75 - N_NEL1_2016_5 15 - Aug - 16 135 1 NEL1_76 - N_NEL1_2016_6 15 - Aug - 16 135 1 NEL1_77 - SW_M2 836560014 15 - Jun - 16 115 3 SWM2_07a, SWM2_07b, SWM2_07c + 836571116 18 - May - 16 102 1 SWM2_04 - 840517848 6 - Jun - 16 112 2 SWM2_05a, SWM2_05b - 5 - Aug - 16 129 1 SWM2_10 - 840525562 15 - Jun - 16 114 1 SWM2_06 - 840533081 21 - Jun - 16 116 1 SWM2_08 - 5 - Aug - 16 128 1 SWM2_09 - 840533890 9 - Aug - 16 130 1 SWM2_11 - 840543057 12 - Aug - 16 131 1 SWM2_12 - N_SWM2_2016_1 12 - Aug - 16 132 1 SWM2_13 - N_SWM2_2016_2 12 - Aug - 16 133 1 SWM2_14 - N_SWM2_2016_3 12 - Aug - 16 134 1 SWM2_15 - SE_H1 836551570 5 - Jul - 16 119 2 SEH1_03a, SEH1_03b + 840525094 20 - Jul - 16 122 1 SEH1_04 - 840532564 25 - Jul - 16 123 1 SEH1_05 - SE_H3 75567567 1 - Jun - 16 107 1 SEH3_11 - 99874823 1 - Jun - 16 109 1 SEH3_13 - 840382344 18 - Jul - 16 121 1 SEH3_18 - 840516095 6 - Jun - 16 111 1 SEH3_15 - 840520549 1 - Jun - 16 108 1 SEH3_12 - 840524297 14 - Jul - 16 120 1 SEH3_17 - 840527780 1 - Jun - 16 110 1 SEH3_14 - 840544608 30 - Jun - 16 118 1 SEH3_16 a n/a a SE_M1 836551889 12 - May - 16 101 1 SEM1_01 - SW_H2 840515348 23 - Jun - 16 117 1 SWH2_03 - 108 Table B Year Site Code Snake Identification Date of Sample Sample Record # Number of Samples Sample Vial Identification Test Results 840518007 24 - May - 16 106 1 SWH2_01 - a This sample was collected and stored following our snake fungal disease sampling protocol, but we were unable to locate the sample at the end of the field day. 109 Discussion Our results indicate positive incidences of SFD occur in areas where the pathogen had not yet been located (Calhoun and Jackson counties, sites SW_M2 and SE_H1, respectively), though it is important to note that, to our knowledge, these massasauga populations have not previously been formally studied and it is possible that the pathogen existed within these populations prior to our study. The occurrence of SFD at these sites poses a potential long - term threat to these populations (e.g., Allender et al. 2015), yet additional study is necessary to evaluate these impacts. T here is a possibility of false negatives with the analysis used to test for SFD, particularly for individuals that show no clinical signs of the disease . S pecifically, a negative relationship was illustrated between number of samples and rate of false positive s for SFD (Hileman et al. 2018), which was the reasoning behind the decision to take a second sample later in the season in 2016. As such, it is possible that our sampling results underestimate the presence of SFD across our study sites ; more samples (i.e. , multiple samples per snake, per season; Hileman et al. 2018) will be necessary to assess whether the pathogen is present at sites where it was not detected . Researchers have assessed the potential increase susceptibility of telemetered massasaugas to SFD ; Hileman et al. (2018) found no relationship between increased rates of SFD and individuals implanted with radiotransmitters . Acknowledgments We acknowledge the support of Michigan State University Department of Fisheries and Wildlife, and G. Roloff and S. Winterstein for their contributions to this work. Funding for this project was provided through the Michigan Department of Natural Resources State Wildlife Grant # F15AP00096 in cooperation with the U.S Fish and Wildlife Service, Wildlife and Sport Fish Restoration Program. We thank K. Bissell, A. Derosier, M. Donnelly, R. Fahlsing, C. 110 Michigan Department of Natural Resources. We thank J. Dingledine and S. Hicks from the U . S . Fish and Wildlife Service . We thank M. Allender and staff at the University of Illinois Wildlife Epidemiology Lab for sample analysis, support with this work, and background information. We thank the employees and veterinary staff of John Ball Zo o, Grand Rapids, Michigan, including R. Colburn, B. Flanagan, and H. Teater. We also thank T. Meyers Harrison for veterinary assistance. We thank S. and C. Weaver and J. Buck; S. Leavitt, C. May, and R. Villegas from The Nature Conservancy; T. Funke and R. Roake from Michigan Audubon Society; Y. Lee from the Michigan Natural Features Inventory; and M. Dreslik at the University of Illinois for their persistent support of our project. We thank field technicians T. Brockman, B. Brodowski, C. Burden, G. Payter, and H. Reynolds (Britz) . 111 APPENDIX C Study S ite I nformation and M aps Table C. 1. The 27 20 - ha study sites , locations, and property names that were included in the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) research throughout southern Michigan in 2015 and 2016 . Focal Area a County Site Code b Property Name c North - East Livingston NE_L2 Brighton State Recreation Area NE_L3 Brighton State Recreation Area NE_M2 Brighton State Recreation Area Oakland NE_H1 Indian Springs Metropark NE_H2 Independence Oaks County Park NE_H3 Independence Oaks County Park NE_L1 Seven Lakes State Park NE_M1 Seven Lakes State Park NE_M3 Indian Springs Metropark South - East Jackson SE_H1 Liberty Fen (private) SE_H2 Grand River Fen (The Nature Conservancy ; private ) SE_L4 Fay Lake (private) Lenawee SE_H3 Ives Road Fen (The Nature Conservancy ; private ) SE_L1 Wacker Chemical Company (private) SE_L2 Ives Road Fen (The Nature Conservancy ; private ) SE_H4 Ives Road Fen (The Nature Conservancy ; private ) Washtenaw SE_L3 Sharonville State Game Area SE_M1 Nan Weston (The Nature Conservancy ; private ) South - West Barry SW_H1 Barry State Game Area SW_H2 Otis Audubon Sanctuary (private) SW_H3 Yankee Springs Recreation Area SW_L1 Yankee Springs Recreation Area SW_L2 Yankee Springs Recreation Area SW_L3 Yankee Springs Recreation Area SW_M3 Barry State Game Area Calhoun SW_M1 Baker Audubon Sanctuary (private) SW_M2 Baker Audubon Sanctuary (private) a Focal area indicates the geographic region within which 9 sites were distributed; counties included in each focal area are listed. b Site code indicates focal region (NE = n orth e ast, SE = s outh e ast, SW = s outh w est), site suitability (L = Low , M = Medium, and H = High, based on our preliminary analysis of the proportions of available cover types using the NLCD land cover data [MRLC 2015]), and number of sites within that focal area at that level of suitability (ranging from 1 - 4). c Due to federal listing status of massasaugas , detailed locations of study sites are not provided . Locations for study sites are available via communication with Daniel Kennedy (Michigan Department of Natural Resources), Henry ( Rique ) Campa III (Michigan State Uni versity), or Stephanie Shaffer. 112 Table C. 2. Site information for the 27 20 - ha study sites where field work was carried out during the 2015 and 2016 field seasons for the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) research throu ghout southern Michigan. Site Code Suitability (NLCD) a Total EMR Observed b EMR Telemetered c Historical Observations d Last sighting e Primary Vegetation Class (IFMAP classification) f Suitability (HSI) g Recent Management h NE_H1 High 0 0 Y 2006 early - mid - late deciduous wetland, early - mid upland deciduous marginal - good NE_H2 High 0 0 Y (within 10m) 2009 early - mid - late deciduous wetland marginal - good NE_H3 High 0 0 Y 2009 early - mid - late deciduous wetland good NE_L1 Low 58 (+ 20 sightings) 12 i Y 2014 late upland deciduous, early - mid upland deciduous poor - marginal shrub/brush removal (for this research) NE_L2 Low 0 0 Y (on property) 2013 late upland deciduous poor NE_L3 Low 0 0 Y 2013 late upland deciduous poor NE_M1 Medium 1 1 Y 2014 late deciduous wetland, early - mid deciduous upland marginal - good NE_M2 Medium 0 0 Y (on property) 2013 wetland deciduous early - mid, late upland deciduous marginal - good NE_M3 Medium 0 0 Y 2006 early - mid upland deciduous, early - mid deciduous wetland optimal SE_H1 High 4 (+ 1 sighting) 4 a Y 2013 deciduous wetland early - mid - late good - optimal SE_H2 High 0 0 other n/a deciduous wetland early - mid - late good SE_H3 High 22 (+ 13 sightings) 4 Y 2013 late deciduous wetland, upland deciduous early - mid - late good ongoing SE_H4 High 0 0 other n/a upland deciduous early - mid marginal ongoing SE_L1 Low 0 0 other n/a upland deciduous early - mid - late marginal SE_L2 Low 1 0 Y 2013 upland deciduous early - mid - late, late deciduous wetland marginal - good ongoing SE_L3 Low 0 0 other n/a upland deciduous early - mid - late marginal Tree/shrub removal (for this research) SE_L4 Low 0 0 Y (within 600m) 2010 late upland deciduous, late deciduous wetland good SE_M1 Medium 1 0 Y (on property) 2004 late wetland deciduous, upland deciduous early - mid marginal 113 Site Code Suitability (NLCD) a Total EMR Observed b EMR Telemetered c Historical Observations d Last sighting e Primary Vegetation Class (IFMAP classification) f Suitability (HSI) g Recent Management h SW_H1 High 0 0 other n/a early - mid deciduous wetland good SW_H2 High 9 (+ 2 sightings) 3 Y 2013 deciduous wetland marginal - good - optimal ongoing (herbicide) SW_H3 High 1 0 Y 2009 early - mid deciduous wetland, late upland deciduous good SW_L1 Low 0 0 Y unknown early - mid deciduous wetland, late upland deciduous marginal SW_L2 Low 0 0 Y 2009 early - mid deciduous wetland, late upland coniferous/deciduous marginal SW_L3 Low 0 0 Y 2009 late upland deciduous/coniferous poor - marginal SW_M1 Medium 0 0 other n/a upland deciduous early - mid, wetland deciduous early - mid - late optimal ongoing SW_M2 Medium 28 (+ 17 sightings) 12 other n/a early - mid deciduous wetland, upland deciduous early - mid - late optimal ongoing (burn/mow) SW_M3 Medium 0 0 other n/a early - mid deciduous wetland, late upland deciduous good a Suitability (NLCD) represents the initial suitability assessment for each site using the NLCD land cover dataset ( MRLC 2015 ; Chapters 1, 2, 3 ). b Total EMR observed is a count of all individual massasauga sightings . c EMR Telemetered is the number of individual massasaugas that were telemetered in 2015, 2016, or in both years, either by surg ical implantation or external attachment methods . d Histo other ) indicates past known presence of massasaugas based on the Michigan Natural Heritage Database of massasauga sightings ( MNFI 2014) . he MNFI (2014) database, but recent reports from collaborating property owners or agencies were provided. e Last Sighting indicates the year of last observation recorded in the Michigan Natural Heritage Database of massasauga sightings ( MNFI 2014) . f Prima ry Vegetation Class was determined by examining landcover within each site using the IFMAP landcover dataset (MDNR 2001) . g Suitability (HSI) based on assessment of each site using Bailey (2010) habitat suitability index model for massasauga s . h Recent M anagement represents habitat management activity at the site during or prior to this project. i One individual at this site was telemetered in both 2015 and 2016. 114 Figure C. 1. The 27 20 - ha study sites, represented by yellow dots, throughout southern Michigan where field work was conducted during 2015 and 2016 field seasons . Counties containing field sites include d Barry, Calhoun, Jackson, Lenawee, Oakland, Washtenaw, and Livingston . T hese sites were selected based on historical presence of eastern massasauga rattlesnake s ( Sistrurus catenatus catenatus ; MNFI 2014). 115 APPENDIX D Documentation of A ll E ncountered E astern M assasauga R attlesnakes Table D. 1. E astern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) encountered and marked during this study ( 2015 and 2016 ) throughout southern Michigan , and fate of the individual at final date of the study . Identification Mark a Site Code Telemetry Frequency b Stage, Sex c # Obs. d Capture Date e Date of Final Obs. f Fate 075566848 g, h PIT SE_H3 U 1 06 - Jul - 15 06 - Jul - 15 Unknown - single observation only 075567567 h PIT SE_H3 A, F 1 01 - Jun - 16 01 - Jun - 16 Unknown - single observation only 075578601 g, h PIT SE_H3 A, F 3 21 - May - 15 08 - Jul - 15 Unknown 099873268 g, h PIT SE_H3 A, U 1 22 - May - 15 22 - May - 15 Unknown - single observation only 099874823 h PIT SE_H3 A, F 2 01 - Jun - 16 06 - Jun - 16 Unknown - final location was a release - had been taken in for surgery, but surgery canceled 100075839 g, i PIT SE_H3 A, F 1 08 - Jul - 15 08 - Jul - 15 Unknown - single observation only 836519789 PIT NE_L1 A, F 1 25 - May - 15 25 - May - 15 Unknown - single observation only 149.264 (ext) A, F 5 17 - Aug - 16 02 - Sep - 16 Unknown - telemetry signal followed, but no observation of snake or transmitter at final location 836546351 h PIT SE_H3 152.142 (int) A, F 28 21 - May - 15 12 - Aug - 15 Mortality (Appendix H) 836547069 PIT SW_H2 150.901 (ext) J, M 3 01 - Jun - 15 04 - Jun - 15 Transmitter recovered - no observation of snake at final location 836549811 PIT SE_H3 A, M 1 07 - Aug - 15 07 - Aug - 15 Unknown - single observation only 836550298 PIT NE_M1 152.002 (int) A, F 31 13 - Jul - 15 13 - Dec - 15 Hibernating - no observation of snake or transmitter; tracked to a potential hibernation site 836550585 PIT SE_H3 A, F 1 12 - Jul - 15 12 - Jul - 15 Unknown - single observation only 836551570 PIT SE_H1 150.692 (ext) J, M 7 17 - Jun - 15 06 - Jul - 15 Unknown - lost signal, could not locate snake or transmitter at final attempt 151.83 (int) J, M 19 05 - Jul - 16 11 - Oct - 16 Hibernating - no observation of snake or transmitter; tracked to a potential hibernation site 836551889 PIT SE_M1 A, F 1 12 - May - 16 12 - May - 16 Unknown - single observation only 836559770 PIT SE_H3 A, F 2 05 - Jun - 15 08 - Jun - 15 Unknown - final location was a release - had been taken in for surgery, but surgery canceled 836560014 PIT SW_M2 A, F 1 29 - Jul - 15 29 - Jul - 15 Unknown - single observation only 151.62 (ext) A, F 2 15 - Jun - 16 16 - Jun - 16 Transmitter recovered - no observation of snake at final location 836560332 PIT NE_L1 A, F 5 09 - Jul - 15 25 - Aug - 15 Unknown - observed basking at final location 836560350 PIT NE_L1 152.061 (int) A, F 37 29 - Jun - 15 13 - Dec - 15 Hibernating - no observation of snake or transmitter; tracked to a potential hibernation site 116 Identification Mark a Site Code Telemetry Frequency b Stage, Sex c # Obs. d Capture Date e Date of Final Obs. f Fate 836561814 PIT SW_M2 152.042 (int) A, F 20 28 - Jul - 15 01 - Nov - 15 Hibernating - no observation of snake or transmitter; tracked to a potential hibernation site 836564778 PIT SW_M2 A, F 2 29 - Jul - 15 06 - Aug - 15 Unknown 836567798 PIT SE_H3 A, F 1 08 - Jul - 15 08 - Jul - 15 Unknown - single observation only 836568071 PIT NE_L1 A, F 1 09 - Jul - 15 09 - Jul - 15 Unknown - single observation only 836570859 h PIT SE_H3 A, F 2 08 - Jul - 15 07 - Aug - 15 Unknown 836571116 h PIT SW_M2 J, F 1 18 - May - 16 18 - May - 16 Unknown - single observation only 836574862 PIT SE_H1 152.162 (int) A, M 13 17 - Jun - 15 31 - Jul - 15 Mortality (Appendix H) 836580078 h PIT NE_L1 A, F 4 29 - Jun - 15 04 - Sep - 15 Unknown - last observed in close proximity to 4 neonates 840380005 PIT NE_L1 152.181 (int) A, F 34 19 - May - 16 02 - Sep - 16 Unknown - telemetry signal followed, but no observation of snake or transmitter at final location 840382344 h PIT SE_H3 A, F 1 18 - Jul - 16 18 - Jul - 16 Unknown - single observation only 840515348 PIT SW_H2 152.342 (int) A, F 18 23 - Jun - 16 07 - Oct - 16 Unknown - observed basking in woodpile at last location 840516095 h PIT SE_H3 A, F 1 06 - Jun - 16 06 - Jun - 16 Unknown - single observation only 840517848 PIT SW_M2 152.202 (int) A, M 29 06 - Jun - 16 10 - Oct - 16 Unknown - telemetry signal followed, but no observation of snake or transmitter at final location 840518007 PIT SW_H2 152.102 (int) A, M 24 24 - May - 16 07 - Oct - 16 Unknown - telemetry signal followed, but no observation of snake or transmitter at final location 840520549 h PIT SE_H3 A, F 1 01 - Jun - 16 01 - Jun - 16 Unknown - single observation only 840521571 PIT NE_L1 151.981 (ext) A, F 6 04 - Aug - 16 17 - Aug - 16 Transmitter recovered - no observation of snake at final location 840524297 PIT SE_H3 150.872 (ext) J, F 3 14 - Jul - 16 18 - Jul - 16 Transmitter recovered - no observation of snake at final location 840525094 PIT SE_H1 151.62 (ext) A, F 2 20 - Jul - 16 25 - Jul - 16 Mortality (Appendix H) 840525562 PIT SW_M2 151.77 (ext) A, F 3 15 - Jun - 16 16 - Jun - 16 Transmitter recovered - no observation of snake at final location 840527780 PIT SE_H3 152.02 (int) A, F 29 01 - Jun - 16 10 - Oct - 16 Unknown - observed basking at last location 840532564 PIT SE_H1 151.8 (int) A, M 13 25 - Jul - 16 11 - Oct - 16 Unknown - telemetry signal followed, but no observation of snake or transmitter at final location 840533081 PIT SW_M2 150.581 (int) A, F 20 21 - Jun - 16 10 - Oct - 16 Unknown - telemetry signal followed, but no observation of snake or transmitter at final location 840533265 PIT NE_L1 151.65 (ext) A, F 6 09 - Jun - 16 10 - Aug - 16 Transmitter recovered - no observation of snake at final location 840533890 PIT SW_M2 A, M 1 09 - Aug - 16 09 - Aug - 16 Unknown - single observation only 117 Identification Mark a Site Code Telemetry Frequency b Stage, Sex c # Obs. d Capture Date e Date of Final Obs. f Fate 840543057 PIT SW_M2 151.711 (ext) A, F 2 12 - Aug - 16 16 - Aug - 16 Transmitter recovered - no observation of snake at final location 840543607 PIT NE_L1 151.92 (ext) A, F 13 28 - Jul - 16 10 - Sep - 16 Unknown - telemetry signal followed, but no observation of snake or transmitter at final location 840544608 PIT SE_H3 A, F 1 30 - Jun - 16 30 - Jun - 16 Unknown - single observation only 836571332, 840382378 j PIT NE_L1 150.46 (int) A, M 33 13 - Jul - 15 13 - Dec - 15 Unknown - found basking (on Dec 13) at final location; near what was presumed to be a hibernation site 150.44 (int) A, M 38 19 - May - 16 14 - Oct - 16 Hibernating - no observation of snake or transmitter; tracked to a potential hibernation site J_NE_L1_2016_1 none NE_L1 150.933 (ext) J, U 9 19 - May - 16 09 - Jun - 16 Unknown - lost signal, could not locate snake or transmitter at final attempt J_SE_H3_2015_1 whiteout SE_H3 151.47 (ext) J, U 10 05 - Jun - 15 01 - Jul - 15 Unknown - lost signal, could not locate snake or transmitter at final attempt N_NE_L1_2016_1 whiteout NE_L1 N, U 1 15 - Aug - 16 15 - Aug - 16 Unknown - single observation only N_NE_L1_2016_10 whiteout NE_L1 149.264 (ext) N, U 2 15 - Aug - 16 17 - Aug - 16 Transmitter recovered - no observation of snake at final location N_NE_L1_2016_11 whiteout NE_L1 149.123 (ext) N, U 7 15 - Aug - 16 10 - Sep - 16 Unknown - telemetry signal followed, but no observation of snake or transmitter at final location N_NE_L1_2016_12 whiteout NE_L1 149.143 (ext) N, U 5 15 - Aug - 16 25 - Aug - 16 Unknown - telemetry signal followed, but no observation of snake or transmitter at final location N_NE_L1_2016_13 whiteout NE_L1 149.284 (ext) N, U 4 17 - Aug - 16 25 - Aug - 16 Unknown - observed moving through a meadow at last location N_NE_L1_2016_2 whiteout NE_L1 N, U 1 15 - Aug - 16 15 - Aug - 16 Unknown - single observation only N_NE_L1_2016_3 whiteout NE_L1 N, U 1 15 - Aug - 16 15 - Aug - 16 Unknown - single observation only N_NE_L1_2016_4 whiteout NE_L1 N, U 1 15 - Aug - 16 15 - Aug - 16 Unknown - single observation only N_NE_L1_2016_5 whiteout NE_L1 N, U 1 15 - Aug - 16 15 - Aug - 16 Unknown - single observation only N_NE_L1_2016_6 whiteout NE_L1 N, U 1 15 - Aug - 16 15 - Aug - 16 Unknown - single observation only N_NE_L1_2016_7 whiteout NE_L1 N, U 1 15 - Aug - 16 15 - Aug - 16 Unknown - single observation only N_NE_L1_2016_8 whiteout NE_L1 N, U 1 15 - Aug - 16 15 - Aug - 16 Unknown - single observation only N_NE_L1_2016_9 whiteout NE_L1 N, U 1 15 - Aug - 16 15 - Aug - 16 Unknown - single observation only N_SW_M2_2015_1 whiteout SW_M2 N, U 2 20 - Aug - 15 21 - Aug - 15 Unknown - at final observation was found approximately 2.5 meters from location of initial observation N_SW_M2_2015_2 whiteout SW_M2 N, U 1 20 - Aug - 15 20 - Aug - 15 Unknown - single observation only N_SW_M2_2015_3 whiteout SW_M2 N, U 1 21 - Aug - 15 21 - Aug - 15 Unknown - single observation only N_SW_M2_2016_1 whiteout SW_M2 149.323 (ext) N, U 2 12 - Aug - 16 16 - Aug - 16 Transmitter recovered - no observation of snake at final location 118 Identification Mark a Site Code Telemetry Frequency b Stage, Sex c # Obs. d Capture Date e Date of Final Obs. f Fate N_SW_M2_2016_2 k whiteout SW_M2 149.165 (ext) N, U 3 12 - Aug - 16 18 - Aug - 16 Transmitter recovered - no observation of snake at final location N_SW_M2_2016_3 whiteout SW_M2 149.204 (ext) N, U 2 12 - Aug - 16 16 - Aug - 16 Transmitter recovered - no observation of snake at final location N_SW_M2_2016_4 k whiteout SW_M2 149.244 (ext) N, U 2 16 - Aug - 16 18 - Aug - 16 Transmitter recovered - no observation of snake at final location N_SW_M2_2016_5 l whiteout SW_M2 N, U 1 16 - Aug - 16 16 - Aug - 16 Unknown - single observation only N_SW_M2_2016_6 whiteout SW_M2 149.183 (ext) N, U 3 16 - Aug - 16 24 - Aug - 16 Unknown - observed basking at last location N_SW_M2_2016_7 whiteout SW_M2 149.204 (ext) N, U 3 16 - Aug - 16 24 - Aug - 16 Unknown - telemetry signal followed, but no observation of snake or transmitter at final location a Massasaugas were marked for identification , or nail enamel painted on rattle. b The radiotelemetry frequency associated with the transmitter attached to the massasauga; nt mitter, indicates an externally attached transmitter . A blank indicates that no transmitter was attached. c Stage (A = adult, J = juvenile, N = neonate, U = unknown) and sex (F = female, M = male, U = unknown) . d Number of times this massasauga was o bserved in the field either via radiotelemetry or by random encounter . e Date of initial capture. f Last date the individual was observed or tracked. g PIT tag from previous study . h Melanistic coloration . i Blue nail enamel on rattle from previous study . j Original PIT tag lost at some point during summer 2016, was replaced . k Exhibited possible hypo - melanism ( J. Harding, personal communication, 2018; Figure D. 1 a, D. 1b ) . l Deformity - neck appears "kinked", body coiled, makes rolling movements (Figure D. 2 a , D. 2b ) . 119 Table D. 2. Number of eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) encounters of either unidentified or unknow n individuals observed during 2015 and 2016 field seasons throughout southern Michigan . Site Code Year Stage Number Observed a Beginning Date End Date NE_L1 2015 J uvenile 1 31 - Jul - 15 n/a N eonate 30 2 - Sep - 15 19 - Sep - 15 U nknown 3 6 - Jul - 15 13 - Aug - 15 2016 Adult 5 26 - May - 16 28 - Jul - 16 N eonate 16 10 - Aug - 16 2 5 - Aug - 16 U nknown 3 1 - Jul - 16 1 - Aug - 16 6 b 8 - Jun - 16 9 - Jun - 16 SE_H1 2016 U nknown 1 29 - Jul - 16 n/a 1 b 19 - Jul - 16 n/a SE_H3 2015 Adult 1 12 - Jun - 15 n/a J uvenile 1 21 - May - 15 n/a U nknown 6 20 - Jun - 15 11 - Aug - 15 2016 Adult 2 9 - Aug - 16 11 - Aug - 16 2 b 29 - Jun - 16 29 - Jun - 16 J uvenile 1 29 - Jun - 16 n/a N eonate 3 9 - Aug - 16 9 - Aug - 16 U nknown 4 2 - Jul - 16 18 - Jul - 16 4 b 29 - Jun - 16 2 - Jul - 16 SE_L2 2016 U nknown 1 22 - Jul - 16 n/a SW_H2 2015 J uvenile 1 29 - May - 15 n/a 2016 Adult 3 16 - Jun - 16 11 - Aug - 16 N eonate 4 24 - Aug - 16 24 - Aug - 16 SW_H3 2016 U nknown 1 3 - Aug - 16 n/a SW_M2 2015 N eonate 22 13 - Aug - 15 24 - Aug - 15 2016 Adult 2 11 - Aug - 16 24 - Aug - 16 1 b 17 - Jun - 16 n/a N eonate 1 10 - Oct - 16 n/a U nknown 1 9 - Aug - 16 n/a 2 b 16 - Jun - 16 n/a a T hese data do not include counts of massasaugas that were marked for telemetry or recapture or individually identifiable in any way . b Observed during a detection survey (Chapter 2) . 120 Table D. 3. Weights (g) and lengths (snout - to - of eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) captured during the 2015 and 2016 field seasons in southern Michigan . Additional metadata associated with each massasauga are provided in Appendix D : Table D. 1 . Stage Sex PIT tag I . D . Year Gravid a SVL (mm) Weight (g) b Adult Female 75567567 2016 Y 555 189 75578601 2015 Y 529 198 (211) 99874823 2016 Y 610 289 100075839 2015 Y 544 205 836519789 2015 unk 550 206 2016 unk 574 225 836546351 2015 Y 527 163 836550298 2015 Y 548 277 836550585 2015 Y 545 244 836551889 2016 unk 555 308 836559770 2015 Y 526 165 836560014 2015 unk 653 332 2016 N 654 267 836560332 2015 Y 530 211 836560350 2015 Y 605 395 (181*) 836561814 2015 Y 569 265 (162*) 836564778 2015 unk 657 357 836567798 2015 Y 528 250 836568071 2015 Y 535 254 836570859 2015 Y 530 222 (199) 836580078 2015 Y 520 206 840380005 2016 unk 650 268 (282) 840382344 2016 unk 531 224 840515348 2016 Y 540 274 840516095 2016 unk 520 165 840520549 2016 Y 493 149 840521571 2016 unk 517 164 840525094 2016 unk 569 307 840525562 2016 Y 560 226 840527780 2016 Y 607 259 840533081 2016 Y 568 230 (214) 840533265 2016 Y 534 193 (205, 208) 840543057 2016 unk 534 131 840543607 2016 Y 658 566 (234*) 840544608 2016 Y 559 204 Male 836549811 2015 n/a 599 150 836574862 2015 n/a 458 157 840517848 2016 n/a 655 405 (411) 840518007 2016 n/a 645 305 840532564 2016 n/a 444 140 840533890 2016 n/a 640 192 121 Stage Sex PIT tag I . D . Year Gravid a SVL (mm) Weight (g) b 836571332 840382378 2015 n/a 480 188 2016 n/a 535 170 (183, 194) Unknown 99873268 2015 n/a 607 209 Juvenile Female 836571116 2016 n/a 464 106 840524297 2016 n/a 358 52 Male 836547069 2015 n/a 246 28 836551570 2015 n/a 298 44 2016 n/a 434 103 Unknown J_NEL1_2016_1 2016 n/a 204 14 J_SEH3_2015_1 2015 n/a 263 16 Neonate Unknown N_NEL1_2016_1 2016 n/a 8 N_NEL1_2016_10 2016 n/a 12 N_NEL1_2016_11 2016 n/a 13 N_NEL1_2016_12 2016 n/a 8 N_NEL1_2016_13 2016 n/a 12 N_NEL1_2016_2 2016 n/a 12 N_NEL1_2016_3 2016 n/a 12 N_NEL1_2016_4 2016 n/a 14 N_NEL1_2016_5 2016 n/a 12 N_NEL1_2016_6 2016 n/a 13 N_NEL1_2016_7 2016 n/a 14 N_NEL1_2016_8 2016 n/a 14 N_NEL1_2016_9 2016 n/a 12 N_SWM2_2015_2 2015 n/a 185 9 N_SWM2_2015_3 2015 n/a 172 10 N_SWM2_2016_1 2016 n/a 9 N_SWM2_2016_2 2016 n/a 12 N_SWM2_2016_3 2016 n/a 13 N_SWM2_2016_4 2016 n/a 13 N_SWM2_2016_5 2016 n/a 12.5 N_SWM2_2016_6 2016 n/a 11 N_SWM2_2016_7 2016 n/a 13 a Reproductive status for female massasaugas ( b Weights in parentheses indicate additional weight measurements taken at a later date; weights marked with an asterisk (*) indicate a measurement taken following parturition. 122 D. 1a D. 1b Figure D. 1. P ossible hypo - melanism (J. Harding, personal communication, 2018 ) observed in two neonate eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) in Calhoun County in southern Michigan in 2016. Figure D. 1a shows a close - up of head. Figure D. 1b shows a neonate exhibiting possible hypo - melanism beside a normal - colored neonate, presumed to be a litter mate, for comparison. 123 D. 2a D. 2b Figure D. 2. Apparent deformity observed in one neonate eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) in Calhoun County in southern Michigan in 2016. Figure D. 2a: i ndividual did not move normally; a rolling motion was used for locomotion, the neck app eared permanently . Figure D. 2b shows the individual in a clear plastic bag, which were used for weighing small juveniles and neonates. 124 APPENDIX E Snake F ungal D isease D isinfection P rotocol Field b iosecurity p rotocol used by all personnel involved in this research while in the field to reduce potential sp r ead of snake fungal disease ( Ophidiomyces ophiodiicola ) between sites . This protocol was developed by M. Allender (University of Illinois, Wildlife Epidemiology Laboratory) , discussed with all massasauga researchers, and was required by the Michigan issued for this research . 1) Field Biosecurity Protocols a) Aim: Keep wildlife populations healthy b) Prevent the introduction of wildlife diseases c) Prevent the spread of wildlife diseases between populations d) Prevent transmission of diseases between animals and field workers e) Minimize costs (both financial and environmental) of disease outbreaks 2) Personal hygiene a) Wear latex or nitril e examination gloves b) Change gloves between each individual c) Wash hands frequently d) Wash clothes daily or wear a washable scrub top/ lab coat e) Wear rubber boots when possible and rinse with 10% bleach between sites. f) If rubber boots are not used, still rinse a nd spray with dilute bleach between sites g) The use of disposable latex gloves when handling cameras, phones, backpacks, etc. These items are rarely disinfected, and are frequently used after handling snakes. 125 3) Sampling Gear Any gear that physically touches animals, including snake tongs, hooks, holding bags, tubes, buckets, gender probes, PIT tag injectors, etc. a) Wash sampling gear in between sites (site=interbreeding population of the species being handled). b) Wash all sampling gear in a disinfecting solutio n. A disinfecting solution such as diluted bleach [nine parts water to one part bleach] or a commercially prepared solution, such as Virkon is recommended. Chlorhexidine and Novalsan are effective with bacteria, but do not appear to kill Ophidiomyces ophio diicola in culture, and are not recommended. c) Soaking gear in a solution is more effective than spraying. Be sure to rinse all solution from gear. d) Soaking of tubes is recommended for 1 hour and then overnight e) When not feasible, then all equipment is wiped with Chlorox wipe f) Contact time is critical, the longer the disinfection solution is applied, the more effective it will be. Also, mechanical action (i.e. scrubbing) will also increase effectiveness. g) Washing should be done between individuals in the populat ion. h) Machine launder snake holding bags using detergent and bleach with hot water. It may be helpful to use different bags for different sites. i) Using disposable (one time use) sterilized PIT tag injectors are recommended. j) If this is not possible, the use of an alcohol soak can be used to clean injectors, as well as other items, such as gender probes, and should be done between individuals, even if within the same site as these tools are invasive. k) Be sure there is no remaining alcohol within the tip of the PIT tag injector. Protocol credit: M. Allender , University of Illinois Wildlife Epidemiology Laboratory 126 Figure E. 1. D isinfection checklist developed for this project used before leaving each f ield site to prevent the spread of snake fungal disease ( Ophidiomyces ophiodiicola ) between sites in southern Michigan during the 2015 and 2015 field seasons (see Appendix B ) . 127 APPENDIX F Study Site and Detection Survey Subsite Naming Description All F. 1 and F. 2) upon site selection and estimation of general suitability for massasaugas (see Site Selection in Chapters 1, 2, or 3) and qualified as low, medium, or high quality based on the proportion of available vegetation types considered suitable for the massasauga based on the Bailey (2010) habitat suitability index model. Upon visiting the site this initial suitability assessment was either confirmed or contradicted. For 2 of the 27 stu dy sites, the initial assessment was found to be based on vegetation types that were available prior to extensive habitat manipulations at the site F. 1 and F. 2). All sites codes were formatted to reflect the difference between the original code and the final code and are shown in Tables F. 1 and F. 2 for all study sites and subsites included in this research. 128 Table F. 1. Site code names for the 20 - ha study site s included in the 2015 and 2016 eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) field work in southern Michigan. Original Site Code a , b Final Site Code b, c Property Name NEH1 NE_H1 Indian Springs Metropark NEH2 NE_H2 Independence Oaks C ounty Park NEH3 NE_H3 Independence Oaks C ounty Park NEL1 NE_L1 Seven Lakes State Park NEL2 NE_L2 Brighton State Recreation Area NEL3 NE_L3 Brighton State Recreation Area NEM1 NE_M1 Seven Lakes State Park NEM2 NE_M2 Brighton State Recreation Area NEM3 NE_M3 Indian Springs Metropark SEH1 SE_H1 Liberty Fen (private) SEH2 SE_H2 Grand River Fen SEH3_M SE_H3 b Ives Road Fen SEH4 SE_H4 Ives Road Fen SEL1 SE_L1 Wacker Chemical Company SEL2 SE_L2 Ives Road Fen SEL3 SE_L3 Sharonville State Game Area SEL5 SE_L4 b Fay Lake (private) SEM1 SE_M1 Nan Weston SWH1 SW_H1 Barry State Game Area SWH2 SW_H2 Otis Audubon Sanctuary SWH3 SW_H3 Yankee Springs Recreation Area SWL1 SW_L1 Yankee Springs Recreation Area SWL2 SW_L2 Yankee Springs Recreation Area SWL3 SW_L3 Yankee Springs Recreation Area SWM1 SW_M1 Baker Audubon Sanctuary SWM2 SW_M2 Baker Audubon Sanctuary SWM3 SW_M3 Barry State Game Area a Site code u sed in initial data collection, data sheets, raw data files, and analyses (e.g., R code). b Within the site codes, NE = northeast focal area, SE = southeast focal area, and SW = southwest focal area (Appendix C); H = high suitability, M = medium suitability, L = low suitability. c Site code u sed t hroughout this dissertation and in any related pu blications; altered from the original for clarity and consistency among sites . 129 Table F. 2. Site and subsite code names (original and final) for 20 - ha study sites and respective 2 - ha detection survey subsites included in the 2015 and 2016 eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) field work throughout southern Michigan . Year Property Name Original Site Code a , b Final Site Code b, c Original Subsite Code a Final Subsite Code b 2015 Baker Audubon Sanctuary SWM1 SW_M1 5y SW_M1_s1 2015 Baker Audubon Sanctuary SWM2 SW_M2 7y SW_M2_s1 2015 Fay Lake (private) SEL5 SE_L4 16 SE_L4_s1 2015 Indian Springs Metropark NEH1 NE_H1 18 NE_H1_s1 2015 Indian Springs Metropark NEM3 NE_M3 17y NE_M3_s1 2015 Ives Road Fen SEH3_M SE_H3 8 SE_H3_s1 32 SE_H3_s2 2015 Liberty Fen (private) SEH1 SE_H1 0 SE_H1_s1 2015 Otis Audubon Sanctuary SWH2 SW_H2 18 SW_H2_s1 2015 Seven Lakes State Park NEL1 NE_L1 19y NE_L1_s1 2015 Seven Lakes State Park NEM1 NE_M1 27y NE_M1_s1 2015 Sharonville State Game Area SEL3 SE_L3 16 SE_L3_s1 2016 Baker Audubon Sanctuary SWM2 SW_M2 7 SW_M2_s1 17 - 21 SW_M2_s2 2016 Ives Road Fen SEH3_M SE_H3 8 SE_H3_s1 32 SE_H3_s2 2016 Liberty Fen (private) SEH1 SE_H1 25 SE_H1_s1 111 SE_H1_s2 2016 Seven Lakes State Park NEL1 NE_L1 15 NE_L1_s1 19 - 24 NE_L1_s2 a Site or subsite code u sed in initial data collection, data sheets, raw data files, and analyses (e.g., R code) . Original subsite code was based on vegetation sampling points and named thusly . Final subsite code reflects the name of the larger study site for clarity. b Within the site codes, NE = northeast focal area, SE = southeast focal area, and SW = southwest focal area (Appendix C); H = high suitability, M = medium suitability, L = low suitability. c Site or subsite code u sed t hroughout this dissertation and in any related publications; altered from the original for clarity and consistency among sites . 130 APPENDIX G R C ode Chapter 1 : Kernel Home Range s and Resource Selection Probability Function Analysis ############################################################################ # SEVEN LAKES ############################################################################ NEL1_emrlocs < - read.table("NEL1_emrlocs_TableToExcel.txt", header=TRUE, sep="") #reads in the raw data NEL1_emr < - NEL1_emrlocs #17T NEL1_emr < - subset(NEL1_emr, select=c("UTMeasting", "UTMnorthin", "Code")) #extract the columns of data I need coordinates(NEL1_emr) < - c("UTMeasting", "UTMnorthin") #turn snake locations into a "spatial points data frame" proj4string(NEL1_emr) < - CRS("+init=epsg:32617") # add CRS information... NEL1s < - spTransform(NEL1_emr, CRS("+init=epsg:32617")) # transform # Utilization distribution. NEL1_kud < - kernelUD(NEL1s, h="href") ## Reference smooth ing parameter. # Estimating a home range from a utilization distribution. NEL1_homerange95 < - getverticeshr(NEL1_kud, percent=95) NEL1_homerange90 < - getverticeshr(NEL1_kud, percent=90) NEL1_homerange85 < - getverticeshr(NEL1_kud, percent=85) NEL1_homer ange80 < - getverticeshr(NEL1_kud, percent=80) NEL1_homerange70 < - getverticeshr(NEL1_kud, percent=70) NEL1_homerange60 < - getverticeshr(NEL1_kud, percent=60) NEL1_homerange50 < - getverticeshr(NEL1_kud, percent=50) NEL1_homerange25 < - getverticeshr(NEL1 _kud, percent=25) # Clip homeranges to the extent of the study site , export/import shapefile for ArcGIS and save to ArcGIS folder library(GISTools) library(rgdal) # Create directory in ArcGIS folder to save from R #dir.create ("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR") 131 # save polygon to Arc folder for clipping and/or determining use/availability of veg sampling sites in Arc writeOGR(NEL1_homerange95, dsn="C:/Users/Stephy/Documents/ArcGIS /MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="NEL1_HR_95", driver="ESRI Shapefile") writeOGR(NEL1_homerange90, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="NEL1_HR_90", driver ="ESRI Shapefile") writeOGR(NEL1_homerange85, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="NEL1_HR_85", driver="ESRI Shapefile") writeOGR(NEL1_homerange80, dsn="C:/Users/Stephy/Documents/ArcGIS/M SU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="NEL1_HR_80", driver="ESRI Shapefile") writeOGR(NEL1_homerange70, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="NEL1_HR_70", driver=" ESRI Shapefile") writeOGR(NEL1_homerange60, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="NEL1_HR_60", driver="ESRI Shapefile") writeOGR(NEL1_homerange50, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU /habitat_selection/RSPF/HR_shapefiles_fromR", layer="NEL1_HR_50", driver="ESRI Shapefile") writeOGR(NEL1_homerange25, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="NEL1_HR_25", driver="ES RI Shapefile") # Import clipped homerange polygon (clipped to the extent of the site) NEL1_HR_95_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","NEL1_HR_95_proj_Clip") NEL1_HR_95_clip < - spTransform(NEL1_HR_95_clip, CRS("+i nit=epsg:32617")) #transform NEL1_HR_90_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","NEL1_HR_90_proj_Clip") NEL1_HR_90_clip < - spTransform(NEL1_HR_90_clip, CRS("+init=epsg:32617")) #transform NEL1_HR_85_clip < - readOGR("C: /Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","NEL1_HR_85_proj_Clip") NEL1_HR_85_clip < - spTransform(NEL1_HR_85_clip, CRS("+init=epsg:32617")) #transform NEL1_HR_80_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","NEL1_HR_80_proj_Clip") NEL1_HR_80_clip < - spTransform(NEL1_HR_80_clip, CRS("+init=epsg:32617")) #transform NEL1_HR_70_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selec tion/RSPF","NEL1_HR_70_proj_Clip") NEL1_HR_70_clip < - spTransform(NEL1_HR_70_clip, CRS("+init=epsg:32617")) #transform NEL1_HR_60 < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","NEL1_HR_60_proj") NEL1_HR_60 < - spTransform(NEL1_HR_ 60, CRS("+init=epsg:32617")) #transform NEL1_HR_50 < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","NEL1_HR_50_proj") NEL1_HR_50 < - spTransform(NEL1_HR_50, CRS("+init=epsg:32617")) #transform NEL1_HR_25 < - readOGR("C:/Users/Stephy/ Documents/ArcGIS/MSU/habitat_selection/RSPF","NEL1_HR_25_proj") NEL1_HR_25 < - spTransform(NEL1_HR_25, CRS("+init=epsg:32617")) #transform # resource selection probability function: library(ResourceSelection) 132 NEL1_veg_use_avail < - read.table("NEL1_vegsam ples_TableToExcel.txt", header=TRUE, sep="") head(NEL1_veg_use_avail) #REMOVE EARLY SAMPLES AND UNNECESSARY DATA BEFORE BRINING INTO R NEL1_RSPF_95 < - rspf(status95 ~ LHC + DHC + woody3m + avgDBH, NEL1_veg_use_avail, m=0, B = 99) NEL1_RSPF_90 < - rspf(stat us90 ~ LHC + DHC + woody3m + avgDBH, NEL1_veg_use_avail, m=0, B = 99) NEL1_RSPF_85 < - rspf(status85 ~ LHC + DHC + woody3m + avgDBH, NEL1_veg_use_avail, m=0, B = 99) NEL1_RSPF_80 < - rspf(status80 ~ LHC + DHC + woody3m + avgDBH, NEL1_veg_use_avail, m=0, B = 99) NEL1_RSPF_70 < - rspf(status70 ~ LHC + DHC + woody3m + avgDBH, NEL1_veg_use_avail, m=0, B = 99) NEL1_RSPF_60 < - rspf(status60 ~ LHC + DHC + woody3m + avgDBH, NEL1_veg_use_avail, m=0, B = 99) NEL1_RSPF_50 < - rspf(status50 ~ LHC + DHC + woody3m + avgDBH, NEL1_veg_use_avail, m=0, B = 99) NEL1_RSPF_25 < - rspf(status25 ~ LHC + DHC + woody3m + avgDBH, NEL1_veg_use_avail, m=0, B = 99) ## Visualize the relationships par(mfrow=c(2,2), ask=FALSE) plot(NEL1_RSPF_95) # FIT LINE WITH CI par(mfrow=c(2,3)) mep( NEL1_RSPF_95) summary(NEL1_RSPF_95) par(mfrow=c(2,2), ask=FALSE) plot(NEL1_RSPF_90) par(mfrow=c(1,1)) par(mfrow=c(2,3), ask=FALSE) mep(NEL1_RSPF_90) summary(NEL1_RSPF_90) par(mfrow=c(2,2), ask=FALSE) plot(NEL1_RSPF_85) par(mfrow=c(2,3)) mep(NEL1_RSPF_8 5) summary(NEL1_RSPF_85) par(mfrow=c(2,2), ask=FALSE) plot(NEL1_RSPF_80) par(mfrow=c(2,3)) mep(NEL1_RSPF_80) summary(NEL1_RSPF_80) # did not converge par(mfrow=c(2,2), ask=FALSE) 133 plot(NEL1_RSPF_70) par(mfrow=c(2,3)) mep(NEL1_RSPF_70) summary( NEL1_RSPF_70) par(mfrow=c(2,2), ask=FALSE) plot(NEL1_RSPF_60) par(mfrow=c(2,3)) mep(NEL1_RSPF_60) summary(NEL1_RSPF_60) # TOP MODEL par(mfrow=c(2,2), ask=FALSE) plot(NEL1_RSPF_50) par(mfrow=c(2,3)) mep(NEL1_RSPF_50) summary(NEL1_RSPF_50) par(mfrow=c( 2,2), ask=FALSE) plot(NEL1_RSPF_25) par(mfrow=c(2,3)) mep(NEL1_RSPF_25) summary(NEL1_RSPF_25) # Rank with consistent AIC - all models that converged caic_NEL1 < - CAICtable(NEL1_RSPF_95,NEL1_RSPF_90,NEL1_RSPF_85,NEL1_RSPF_70,NEL1_RSPF_60,NEL1_RSPF_50,NE L1_RSPF_25) write.table(caic_NEL1, file = "caic_NEL1.txt", sep = " \ t") ############################################################################ # IVES NORTH ############################################################################ SEH3_emrlocs < - read.table("SEH3_emrlocs_TableToExcel.txt", header=TRUE, sep="") #reads in the raw data SEH3_emr < - SEH3_emrlocs #17T SEH3_emr < - subset(SEH3_emr, select=c("UTMeasting", "UTMnorthin", "Code")) #extract the columns of data I need coordinates(SEH3_emr) < - c( "UTMeasting", "UTMnorthin") #turn snake locations into a "spatial points data frame" proj4string(SEH3_emr) < - CRS("+init=epsg:32617") # add CRS information... SEH3s < - spTransform(SEH3_emr, CRS("+init=epsg:32617")) # transform # Utilization distribution. SEH3_kud < - kernelUD(SEH3s, h="href") ## Reference smoothing parameter. 134 # Estimating a home range from a utilization distribution. SEH3_homerange95 < - getverticeshr(SEH3_kud, percent=95) SEH3_homerange90 < - getverticeshr(SEH3_kud, percent=90) SEH3_homer ange85 < - getverticeshr(SEH3_kud, percent=85) SEH3_homerange80 < - getverticeshr(SEH3_kud, percent=80) SEH3_homerange70 < - getverticeshr(SEH3_kud, percent=70) SEH3_homerange60 < - getverticeshr(SEH3_kud, percent=60) SEH3_homerange50 < - getverticeshr(SEH3 _kud, percent=50) SEH3_homerange25 < - getverticeshr(SEH3_kud, percent=25) # Clip homeranges to the extent of the study site , export/import shapefile for ArcGIS and save to ArcGIS folder # save polygon to Arc folder for clipping and/or determining use/av ailability of veg sampling sites in Arc writeOGR(SEH3_homerange95, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SEH3_HR_95", driver="ESRI Shapefile") writeOGR(SEH3_homerange90, dsn="C:/Users/Steph y/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SEH3_HR_90", driver="ESRI Shapefile") writeOGR(SEH3_homerange85, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SEH3_HR_85", driver="ESRI Shapefile") writeOGR(SEH3_homerange80, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SEH3_HR_80", driver="ESRI Shapefile") writeOGR(SEH3_homerange70, dsn= "C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SEH3_HR_70", driver="ESRI Shapefile") writeOGR(SEH3_homerange60, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SEH3_HR_60", driver="ESRI Shapefile") writeOGR(SEH3_homerange50, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SEH3_HR_50", driver="ESRI Shapefile") writeOGR(SEH3_homerange25, dsn="C :/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SEH3_HR_25", driver="ESRI Shapefile") # Import clipped homerange polygon (clipped to the extent of the site) SEH3_HR_95_clip < - readOGR("C:/Users/Stephy/Doc uments/ArcGIS/MSU/habitat_selection/RSPF","SEH3_HR_95_proj_Clip") SEH3_HR_95_clip < - spTransform(SEH3_HR_95_clip, CRS("+init=epsg:32617")) #transform SEH3_HR_90_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SEH3_HR_90_proj_ Clip") SEH3_HR_90_clip < - spTransform(SEH3_HR_90_clip, CRS("+init=epsg:32617")) #transform SEH3_HR_85_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SEH3_HR_85_proj_Clip") SEH3_HR_85_clip < - spTransform(SEH3_HR_85_clip, CRS( "+init=epsg:32617")) #transform SEH3_HR_80_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SEH3_HR_80_proj_Clip") SEH3_HR_80_clip < - spTransform(SEH3_HR_80_clip, CRS("+init=epsg:32617")) #transform 135 SEH3_HR_70 < - readOGR("C:/U sers/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SEH3_HR_70_proj") SEH3_HR_70 < - spTransform(SEH3_HR_70, CRS("+init=epsg:32617")) #transform SEH3_HR_60 < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SEH3_HR_60_proj") SEH 3_HR_60 < - spTransform(SEH3_HR_60, CRS("+init=epsg:32617")) #transform SEH3_HR_50 < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SEH3_HR_50_proj") SEH3_HR_50 < - spTransform(SEH3_HR_50, CRS("+init=epsg:32617")) #transform SEH3_HR_25 < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SEH3_HR_25_proj") SEH3_HR_25 < - spTransform(SEH3_HR_25, CRS("+init=epsg:32617")) #transform # resource selection probability function: SEH3_veg_use_avail < - read.table(" SEH3_vegsamples_TableToExcel.txt", header=TRUE, sep="") head(SEH3_veg_use_avail) #REMOVE EARLY SAMPLES AND UNNECESSARY DATA BEFORE BRINING INTO R SEH3_RSPF_95 < - rspf(status95 ~ LHC + DHC + woody3m + avgDBH, SEH3_veg_use_avail, m=0, B = 99) SEH3_RSPF_90 < - rspf(status90 ~ LHC + DHC + woody3m + avgDBH, SEH3_veg_use_avail, m=0, B = 99) SEH3_RSPF_85 < - rspf(status85 ~ LHC + DHC + woody3m + avgDBH, SEH3_veg_use_avail, m=0, B = 99) SEH3_RSPF_80 < - rspf(status80 ~ LHC + DHC + woody3m + avgDBH, SEH3_veg_use_avail , m=0, B = 99) SEH3_RSPF_70 < - rspf(status70 ~ LHC + DHC + woody3m + avgDBH, SEH3_veg_use_avail, m=0, B = 99) SEH3_RSPF_60 < - rspf(status60 ~ LHC + DHC + woody3m + avgDBH, SEH3_veg_use_avail, m=0, B = 99) SEH3_RSPF_50 < - rspf(status50 ~ LHC + DHC + woody3m + avgDBH, SEH3_veg_use_avail, m=0, B = 99) SEH3_RSPF_25 < - rspf(status25 ~ LHC + DHC + woody3m + avgDBH, SEH3_veg_use_avail, m=0, B = 99) ## Visualize the relationships par(mfrow=c(2,2), ask=FALSE) plot(SEH3_RSPF_95) # FIT LINE WITH CI par(mfrow=c(2,3)) mep(SEH3_RSPF_95) summary(SEH3_RSPF_95) par(mfrow=c(2,2), ask=FALSE) plot(SEH3_RSPF_90) par(mfrow=c(2,3)) mep(SEH3_RSPF_90) summary(SEH3_RSPF_90) #did not converge par(mfrow=c(2,2), ask=FALSE) plot(SEH3_RSPF_85) par(mfrow=c(2,3)) mep(SEH3_RSPF_85) 136 summary(SEH3_RSPF_85) # top model par(mfrow=c(2,2), ask=FALSE) plot(SEH3_RSPF_80) par(mfrow=c(2,3)) mep(SEH3_RSPF_80) summary(SEH3_RSPF_80) # did not converge par(mfrow=c(2,2), ask=FALSE) plot(SEH3_RSPF_70) par(mfrow=c(2,3)) mep(SEH3_RSPF_70) summary( SEH3_RSPF_70) par(mfrow=c(2,2), ask=FALSE) plot(SEH3_RSPF_60) par(mfrow=c(2,3)) mep(SEH3_RSPF_60) summary(SEH3_RSPF_60) # DID NOT CONVERGE par(mfrow=c(2,2), ask=FALSE) plot(SEH3_RSPF_50) par(mfrow=c(2,3)) mep(SEH3_RSPF_50) summary(SEH3_RSPF_50) # did not converge par(mfrow=c(2,2), ask=FALSE) plot(SEH3_RSPF_25) par(mfrow=c(2,3)) mep(SEH3_RSPF_25) summary(SEH3_RSPF_25) # did not converge # RANK WITH consistent AIC - all models that converged caic_SEH3 < - CAICtable(SEH3_RSPF_95,SEH3_RSPF_85,SEH3_R SPF_70) write.table(caic_SEH3, file = "caic_SEH3.txt", sep = " \ t") ############################################################################ # OTIS ############################################################################ 137 SWH2_emrlocs < - read.table( "SWH2_emrlocs_TableToExcel.txt", header=TRUE, sep="") #reads in the raw data SWH2_emr < - SWH2_emrlocs #17T SWH2_emr < - subset(SWH2_emr, select=c("UTMeasting", "UTMnorthin", "Code")) #extract the columns of data I need coordinates(SWH2_emr) < - c("UTMeasting ", "UTMnorthin") #turn snake locations into a "spatial points data frame" proj4string(SWH2_emr) < - CRS("+init=epsg:32616") # add CRS information... SWH2s < - spTransform(SWH2_emr, CRS("+init=epsg:32616")) # transform # Utilization distribution. SWH2_kud < - kernelUD(SWH2s, h="href") ## Reference smoothing parameter. # Estimating a home range from a utilization distribution. # Create a __% polygon home range (note: these are not clipped to the study site yet) SWH2_homerange95 < - getverticeshr( SWH2_kud, percent=95) SWH2_homerange90 < - getverticeshr(SWH2_kud, percent=90) SWH2_homerange85 < - getverticeshr(SWH2_kud, percent=85) SWH2_homerange80 < - getverticeshr(SWH2_kud, percent=80) SWH2_homerange70 < - getverticeshr(SWH2_kud, percent=70) SWH2_ homerange60 < - getverticeshr(SWH2_kud, percent=60) SWH2_homerange50 < - getverticeshr(SWH2_kud, percent=50) SWH2_homerange25 < - getverticeshr(SWH2_kud, percent=25) # Clip homeranges to the extent of the study site , export/import shapefile for ArcGIS and save to ArcGIS folder # save polygon to Arc folder for clipping and/or determining use/availability of veg sampling sites in Arc writeOGR(SWH2_homerange95, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", la yer="SWH2_HR_95", driver="ESRI Shapefile") writeOGR(SWH2_homerange90, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SWH2_HR_90", driver="ESRI Shapefile") writeOGR(SWH2_homerange85, dsn="C:/Users/S tephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SWH2_HR_85", driver="ESRI Shapefile") writeOGR(SWH2_homerange80, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", laye r="SWH2_HR_80", driver="ESRI Shapefile") writeOGR(SWH2_homerange70, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SWH2_HR_70", driver="ESRI Shapefile") writeOGR(SWH2_homerange60, dsn="C:/Users/Ste phy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SWH2_HR_60", driver="ESRI Shapefile") writeOGR(SWH2_homerange50, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer= "SWH2_HR_50", driver="ESRI Shapefile") writeOGR(SWH2_homerange25, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SWH2_HR_25", driver="ESRI Shapefile") 138 # Import clipped homerange polygon (clipped to the extent of the site) SWH2_HR_95_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWH2_HR_95_proj_Clip") SWH2_HR_95_clip < - spTransform(SWH2_HR_95_clip, CRS("+init=epsg:32616")) #transform SWH2_HR_90_clip < - readOGR("C:/U sers/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWH2_HR_90_proj_Clip") SWH2_HR_90_clip < - spTransform(SWH2_HR_90_clip, CRS("+init=epsg:32616")) #transform SWH2_HR_85_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","S WH2_HR_85_proj_Clip") SWH2_HR_85_clip < - spTransform(SWH2_HR_85_clip, CRS("+init=epsg:32616")) #transform SWH2_HR_80_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWH2_HR_80_proj_Clip") SWH2_HR_80_clip < - spTransform(SWH2_H R_80_clip, CRS("+init=epsg:32616")) #transform SWH2_HR_70_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWH2_HR_70_proj_Clip") SWH2_HR_70_clip < - spTransform(SWH2_HR_70_clip, CRS("+init=epsg:32616")) #transform SWH2_HR_60_c lip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWH2_HR_60_proj_Clip") SWH2_HR_60_clip < - spTransform(SWH2_HR_60_clip, CRS("+init=epsg:32616")) #transform SWH2_HR_50_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habita t_selection/RSPF","SWH2_HR_50_proj_Clip") SWH2_HR_50_clip < - spTransform(SWH2_HR_50_clip, CRS("+init=epsg:32616")) #transform SWH2_HR_25 < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWH2_HR_25_proj") SWH2_HR_25 < - spTransform(S WH2_HR_25, CRS("+init=epsg:32616")) #transform # resource selection probability function: SWH2_veg_use_avail < - read.table("SWH2_vegsamples_TableToExcel.txt", header=TRUE, sep="") head(SWH2_veg_use_avail) #REMOVE EARLY SAMPLES AND UNNECESSARY DATA BEFORE BRINING INTO R #SWH2_RSPF_95 < - rspf(status95 ~ LHC + DHC + woody3m + avgDBH, SWH2_veg_use_avail, m=0, B = 99) #does not work (all veg points fall within the 95% home range) SWH2_RSPF_90 < - rspf(status90 ~ LHC + DHC + woody3m + avgDBH, SWH2_veg_use_avail , m=0, B = 99) SWH2_RSPF_85 < - rspf(status85 ~ LHC + DHC + woody3m + avgDBH, SWH2_veg_use_avail, m=0, B = 99) SWH2_RSPF_80 < - rspf(status80 ~ LHC + DHC + woody3m + avgDBH, SWH2_veg_use_avail, m=0, B = 99) SWH2_RSPF_70 < - rspf(status70 ~ LHC + DHC + woody3m + avgDBH, SWH2_veg_use_avail, m=0, B = 99) SWH2_RSPF_60 < - rspf(status60 ~ LHC + DHC + woody3m + avgDBH, SWH2_veg_use_avail, m=0, B = 99) SWH2_RSPF_50 < - rspf(status50 ~ LHC + DHC + woody3m + avgDBH, SWH2_veg_use_avail, m=0, B = 99) SWH2_RSPF_25 < - rspf(s tatus25 ~ LHC + DHC + woody3m + avgDBH, SWH2_veg_use_avail, m=0, B = 99) ## Visualize the relationships #par(mfrow=c(2,2), ask=FALSE) #plot(SWH2_RSPF_95) # FIT LINE WITH CI #par(mfrow=c(2,3)) #mep(SWH2_RSPF_95) #summary(SWH2_RSPF_95) 139 par(mfrow=c(2,2), a sk=FALSE) plot(SWH2_RSPF_90) par(mfrow=c(2,3)) mep(SWH2_RSPF_90) summary(SWH2_RSPF_90) par(mfrow=c(2,2), ask=FALSE) plot(SWH2_RSPF_85) par(mfrow=c(2,3)) mep(SWH2_RSPF_85) summary(SWH2_RSPF_85) # does not converge par(mfrow=c(2,2), ask=FALSE) plot(SWH 2_RSPF_80) par(mfrow=c(2,3)) mep(SWH2_RSPF_80) summary(SWH2_RSPF_80) # DOES NOT CONVERGE par(mfrow=c(2,2), ask=FALSE) plot(SWH2_RSPF_70) par(mfrow=c(2,3)) mep(SWH2_RSPF_70) summary(SWH2_RSPF_70) # DOES NOT CONVERGE par(mfrow=c(2,2), ask=FALSE) plot( SWH2_RSPF_60) par(mfrow=c(2,3)) mep(SWH2_RSPF_60) summary(SWH2_RSPF_60) par(mfrow=c(2,2), ask=FALSE) plot(SWH2_RSPF_50) par(mfrow=c(2,3)) mep(SWH2_RSPF_50) summary(SWH2_RSPF_50) # DOES NOT CONVERGE par(mfrow=c(2,2), ask=FALSE) plot(SWH2_RSPF_25) 140 par(mf row=c(2,3)) mep(SWH2_RSPF_25) summary(SWH2_RSPF_25) # did not converge # RANK WITH consistent AIC - all models that converged caic_SWH2 < - CAICtable(SWH2_RSPF_90,SWH2_RSPF_60) write.table(caic_SWH2, file = "caic_SWH2.txt", sep = " \ t") ############################################################################ # BAKER NORTH ############################################################################ SWM2_emrlocs < - read.table("SWM2_emrlocs_TableToExcel.txt", header=TRUE, sep="") #reads in the raw data SWM2_emr < - SWM2_emrlocs #17T SWM2_emr < - subset(SWM2_emr, select=c("UTMeasting", "UTMnorthin", "Code")) #extract the columns of data I need coordinates(SWM2_emr) < - c("UTMeasting", "UTMnorthin") #turn snake locations into a "spatial points data frame" proj4string(SWM2_emr) < - CRS("+init=epsg:32616") # add CRS information... SWM2s < - spTransform(SWM2_emr, CRS("+init=epsg:32616")) # transform # Utilization distribution. SWM2_kud < - kernelUD(SWM2s, h="href") ## Reference smoothing parameter. # Estimating a home range from a utilization distribution. SWM2_homerange95 < - getverticeshr(SWM2_kud, percent=95) SWM2_homerange90 < - getverticeshr(SWM2_kud, percent=90) SWM2_homerange85 < - getverticeshr(SWM2_kud, percent=85) SWM2_homerange80 < - getve rticeshr(SWM2_kud, percent=80) SWM2_homerange70 < - getverticeshr(SWM2_kud, percent=70) SWM2_homerange60 < - getverticeshr(SWM2_kud, percent=60) SWM2_homerange50 < - getverticeshr(SWM2_kud, percent=50) SWM2_homerange25 < - getverticeshr(SWM2_kud, percent=2 5) # Clip homeranges to the extent of the study site , export/import shapefile for ArcGIS and save to ArcGIS folder # save polygon to Arc folder for clipping and/or determining use/availability of veg sampling sites in Arc writeOGR( SWM2_homerange95, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SWM2_HR_95", driver="ESRI Shapefile") writeOGR(SWM2_homerange90, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_ shapefiles_fromR", layer="SWM2_HR_90", driver="ESRI Shapefile") writeOGR(SWM2_homerange85, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SWM2_HR_85", driver="ESRI Shapefile") 141 writeOGR(SW M2_homerange80, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SWM2_HR_80", driver="ESRI Shapefile") writeOGR(SWM2_homerange70, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_sh apefiles_fromR", layer="SWM2_HR_70", driver="ESRI Shapefile") writeOGR(SWM2_homerange60, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SWM2_HR_60", driver="ESRI Shapefile") writeOGR(SWM2 _homerange50, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shapefiles_fromR", layer="SWM2_HR_50", driver="ESRI Shapefile") writeOGR(SWM2_homerange25, dsn="C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF/HR_shap efiles_fromR", layer="SWM2_HR_25", driver="ESRI Shapefile") # Import clipped homerange polygon (clipped to the extent of the site) SWM2_HR_95_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWM2_HR_95_proj_Clip") SWM2_HR_95_clip < - spTransform(SWM2_HR_95_clip, CRS("+init=epsg:32616")) #transform SWM2_HR_90_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWM2_HR_90_proj_Clip") SWM2_HR_90_clip < - spTransform(SWM2_HR_90_clip, CRS("+init= epsg:32616")) #transform SWM2_HR_85_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWM2_HR_85_proj_Clip") SWM2_HR_85_clip < - spTransform(SWM2_HR_85_clip, CRS("+init=epsg:32616")) #transform SWM2_HR_80_clip < - readOGR("C:/Use rs/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWM2_HR_80_proj_Clip") SWM2_HR_80_clip < - spTransform(SWM2_HR_80_clip, CRS("+init=epsg:32616")) #transform SWM2_HR_70_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWM 2_HR_70_proj_Clip") SWM2_HR_70_clip < - spTransform(SWM2_HR_70_clip, CRS("+init=epsg:32616")) #transform SWM2_HR_60_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWM2_HR_60_proj_Clip") SWM2_HR_60_clip < - spTransform( SWM2_HR_60_clip, CRS("+init=epsg:32616")) #transform SWM2_HR_50_clip < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWM2_HR_50_proj_Clip") SWM2_HR_50_clip < - spTransform(SWM2_HR_50_clip, CRS("+init=epsg:32616")) #transform SWM2_H R_25 < - readOGR("C:/Users/Stephy/Documents/ArcGIS/MSU/habitat_selection/RSPF","SWM2_HR_25_proj") SWM2_HR_25 < - spTransform(SWM2_HR_25, CRS("+init=epsg:32616")) #transform # resource selection probability function: SWM2_veg_use_avail < - read.table("SWM2_v egsamples_TableToExcel.txt", header=TRUE, sep="") head(SWM2_veg_use_avail) #REMOVE EARLY SAMPLES AND UNNECESSARY DATA BEFORE BRINING INTO R SWM2_RSPF_95 < - rspf(status95 ~ LHC + DHC + woody3m + avgDBH, SWM2_veg_use_avail, m=0, B = 99) SWM2_RSPF_90 < - rspf (status90 ~ LHC + DHC + woody3m + avgDBH, SWM2_veg_use_avail, m=0, B = 99) SWM2_RSPF_85 < - rspf(status85 ~ LHC + DHC + woody3m + avgDBH, SWM2_veg_use_avail, m=0, B = 99) SWM2_RSPF_80 < - rspf(status80 ~ LHC + DHC + woody3m + avgDBH, SWM2_veg_use_avail, m=0, B = 99) SWM2_RSPF_70 < - rspf(status70 ~ LHC + DHC + woody3m + avgDBH, SWM2_veg_use_avail, m=0, B = 99) SWM2_RSPF_60 < - rspf(status60 ~ LHC + DHC + woody3m + avgDBH, SWM2_veg_use_avail, m=0, B = 99) 142 SWM2_RSPF_50 < - rspf(status50 ~ LHC + DHC + woody3m + avg DBH, SWM2_veg_use_avail, m=0, B = 99) SWM2_RSPF_25 < - rspf(status25 ~ LHC + DHC + woody3m + avgDBH, SWM2_veg_use_avail, m=0, B = 99) ## Visualize the relationships par(mfrow=c(2,2), ask=FALSE) plot(SWM2_RSPF_95) # FIT LINE WITH CI par(mfrow=c(2,3)) mep(SW M2_RSPF_95) summary(SWM2_RSPF_95) # did not converge par(mfrow=c(2,2), ask=FALSE) plot(SWM2_RSPF_90) par(mfrow=c(2,3)) mep(SWM2_RSPF_90) summary(SWM2_RSPF_90) # DID NOT CONVERGE par(mfrow=c(2,2), ask=FALSE) plot(SWM2_RSPF_85) par(mfrow=c(2,3)) mep( SWM2_RSPF_85) summary(SWM2_RSPF_85) # DID NOT CONVERGE par(mfrow=c(2,2), ask=FALSE) plot(SWM2_RSPF_80) par(mfrow=c(2,3)) mep(SWM2_RSPF_80) summary(SWM2_RSPF_80) par(mfrow=c(2,2), ask=FALSE) plot(SWM2_RSPF_70) par(mfrow=c(2,3)) mep(SWM2_RSPF_70) su mmary(SWM2_RSPF_70) # did not converge par(mfrow=c(2,2), ask=FALSE) plot(SWM2_RSPF_60) par(mfrow=c(2,3)) mep(SWM2_RSPF_60) summary(SWM2_RSPF_60) # did not converge 143 par(mfrow=c(2,2), ask=FALSE) plot(SWM2_RSPF_50) par(mfrow=c(2,3)) mep(SWM2_RSPF_50) summary(SWM2_RSPF_50) # did not converge par(mfrow=c(2,2), ask=FALSE) plot(SWM2_RSPF_25) par(mfrow=c(2,3)) mep(SWM2_RSPF_25) summary(SWM2_RSPF_25) # DID NOT CONVERGE # RANK WITH consistent AIC - all models that converged caic_SWM2 < - CAICtable(SWM2_RS PF_80) write.table(caic_SWM2, file = "caic_SWM2.txt", sep = " \ t") 144 Chapter 2 : Detection Factors and Predicted Occupancy Estimates . ######################################################################################################## # Detection using Unmarked ######################################################################################################## library(unmarked) # For estimating detection library(AICcmodavg) # For AIC tables library(MuMIn) # For model averaging library(ggplo t2) # For plotting predicted detection probabilites based on top covariates #################### # Read in Data #################### AllData < - read.table("DetectionSurveyData.txt", header=TRUE, sep=" \ t") #properly formatted AllData < - AllData[! AllData$Year.1 == "2015", ] #Remove 2015 data (only want to analyze 2016) AllData < - AllData[! AllData$SurveyID == "GRLF.SEH1.25.1", ] #Remove unoccupied subsite (the site that we #never found an EMR at, SEH1.25) AllData < - AllData[! AllData$SurveyID == "GRLF.SEH1.25.2", ] AllData < - AllData[! AllData$SurveyID == "GRLF.SEH1.25.3", ] AllData < - AllData[! AllData$SurveyID == "GRLF.SEH1.25.4", ] AllData < - AllData[! AllData$SurveyID == "GRLF.SEH1.25.5", ] AllData < - AllData[! AllData$SurveyID == "GRLF.SEH1.2 5.6", ] ########################################################################################### # COVARIATES ########################################################################################### # Detection history: EMRDet < - AllData[,73:74] # d etection history column numbers from the data table # Site covariate: EMRsitecovs < - AllData[,c("QualityRank")] # site covariates, i.e., site quality. 1=low, 2=medium, 3=high # Observation covariates: EMRobscovs < - list(SubsiteCode=AllData[,c("SubsiteCod e.1","SubsiteCode.2")], Prec=AllData[,c("Prec.1","Prec.2")], SiteOccu=AllData[,c("SiteOccu.1","SiteOccu.2")], SrchrCode=AllData[,c("SrchrCode.1","SrchrCode.2")], RankSnake=AllDa ta[,c("RankSnake.1","RankSnake.2")], RankSnakePerSite=AllData[,c("RankSnakePerSite.1","RankSnakePerSite.2")], 145 OverallRank=AllData[,c("OverallRank.1","OverallRank.2")], DaysSince=AllData[,c ("DaysSince.1","DaysSince.2")], #continuous StartMins=AllData[,c("StartMins.1","StartMins.2")], #continuous EndMins=AllData[,c("EndMins.1","EndMins.2")], #continuous SqmHerb=AllData[,c("SqmHerb.1" ,"SqmHerb.2")], #continuous SqmWood=AllData[,c("SqmWood.1","SqmWood.2")], #continuous herbtowood=AllData[,c("herbtowood.1","herbtowood.2")], #continuous AirDif=AllData[,c("AirDif.1","AirDif.2")], #continuous PrevDayMinNOAA=AllData[,c("PrevDayMinNOAA.1","PrevDayMinNOAA.2")], #continuous SurfAvg=AllData[,c("SurfAvg.1","SurfAvg.2")], #continuous SurfMax=AllData[,c("SurfMax.1","SurfMax.2")], #co ntinuous SoilAvg=AllData[,c("SoilAvg.1","SoilAvg.2")], #continuous SoilMin=AllData[,c("SoilMin.1","SoilMin.2")], #continuous SoilMax=AllData[,c("SoilMax.1","SoilMax.2")], #continu ous SolRadAvg=AllData[,c("SolRadAvg.1","SolRadAvg.2")], #continuous SolRadMin=AllData[,c("SolRadMin.1","SolRadMin.2")], #continuous SolRadMax=AllData[,c("SolRadMax.1","SolRadMax.2")], #continuous SolRadSD=AllData[,c("SolRadSD.1","SolRadSD.2")], #continuous SolRadSE=AllData[,c("SolRadSE.1","SolRadSE.2")], #continuous HumAvg=AllData[,c("HumAvg.1","HumAvg.2")], #continuous HumMin=AllData[,c("HumMin.1","HumMin.2")], #continuous HumMax=AllData[,c("HumMax.1","HumMax.2")], #continuous difNOAAminSurfmin=AllData[,c("difNOAAminSurfmin.1","difNOAAminSurfmin.2")],#continuous difNOAAminAirmax=AllData[,c("difNOAAminAirmax.1","difNOAAminAirmax.2")], #continuous difNOAAminAirmin=AllData[,c("difNOAAminAirmin.1","difNOAAminAirmin.2")], #continuous PropHerb=AllData[,c("PropHerb.1","PropHerb.2 ")], #continuous PropWood=AllData[,c("PropWood.1","PropWood.2")], #continuous Julian=AllData[,c("Julian.1","Julian.2")], #continuous SurveyOrder=AllData[,c ("SurveyOrder.1","SurveyOrder.2")], #continuous TotalMins=AllData[,c("TotalMins.1","TotalMins.2")], #continuous MMTs=AllData[,c("MMTs.1","MMTs.2")], #continuous SearchMins=AllDat a[,c("SearchMins.1","SearchMins.2")], #continuous AirMin=AllData[,c("AirMin.1","AirMin.2")], #continuous SameDayMinNOAA=AllData[,c("SameDayMinNOAA.1","SameDayMinNOAA.2")], #continuous AirAvg=AllData[,c("AirAvg.1","AirAvg.2")], #continuous AirMax=AllData[,c("AirMax.1","AirMax.2")], #continuous SurfMin=AllData[,c("SurfMin.1","SurfMin.2")] #continuous ) 146 EMRsitecovsdf < - data.frame(EMRsitecovs) # Observation covariates as dataframe EMR.data < - unmarkedFrameOccu(y=EMRDet, siteCovs=EMRsitecovsdf, obsCovs=EMRobscovs) # Standardize continuous covariates, which is what we want. ocforscale < - obsCovs(EMR.data) # remove the observation covariates from the dataframe ocforscale[,8:43] < - scale(ocforscale[,8:43]) # scale the continuous variables obsCovs(EMR.data) < - ocforscale # put them back into the dataframe summary(EMR.data) ############################################## ############################################################ # Correlation among covariates ########################################################################################################## #using the ocforscale dataframe (data is scaled here) # These are all covariates significant at 5%, plus PropHerb and PropWood (which we did not ultimately end # up using because of their non - significance) correlation_5percent < - cor(ocforscale[32:43], method="pearson") correlation_5percent pairs(correlation_5percent) CorrTable < - as.table(correlation_5percent) ########################################################################################################### # Fitting models: ######################################################### ################################################## # Null Model: #~~~~~~~~~~~~~~~~~~~~~~~~~~ EMRfm1 < - occu(~1 ~1, EMR.data) backTransform(EMRfm1, 'det') # backtransform to get actual detection estimate, for no - covariate models backTransform(EMRfm1, 'stat e') # backtransform to get actual occupancy estimate, for no - covariate models ############################## # Univariate Models: ############################## # significant at alpha = 0.05 #~~~~~~~~~~~~~~~~~~~~~~~~~~ TotalMins < - occu(~ TotalMins ~ 1, E MR.data) #sig backTransform(linearComb(TotalMins, coefficients = c(1,0), type = 'det')) MMTs < - occu(~ MMTs ~ 1, EMR.data) #sig 147 backTransform(linearComb(MMTs, coefficients = c(1,0), type = 'det')) SearchMins < - occu(~ SearchMins ~ 1, EMR.data) #sig backT ransform(linearComb(SearchMins, coefficients = c(1,0), type = 'det')) AirMin < - occu(~ AirMin ~ 1, EMR.data) #sig backTransform(linearComb(AirMin, coefficients = c(1,0), type = 'det')) SameDayMinNOAA < - occu(~ SameDayMinNOAA ~ 1, EMR.data) #sig . A s min temp gets colder, detection prob increases backTransform(linearComb(SameDayMinNOAA, coefficients = c(1,0), type = 'det')) AirAvg < - occu(~ AirAvg ~ 1, EMR.data) #sig backTransform(linearComb(AirAvg, coefficients = c(1,0), type = 'det')) AirMax < - occu(~ AirMax ~ 1, EMR.data) #sig backTransform(linearComb(AirMax, coefficients = c(1,0), type = 'det')) SurfMin < - occu(~ SurfMin ~ 1, EMR.data) #sig backTransform(linearComb(SurfMin, coefficients = c(1,0), type = 'det')) SurveyOrder < - occu(~ SurveyOrder ~ 1, EMR.data) #sig backTransform(linearComb(SurveyOrder, coefficients = c(1,0), type = 'det')) Julian < - occu(~ Julian ~ 1, EMR.data) #sig backTransform(linearComb(Julian, coefficients = c(1,0), type = 'det')) #~~~~~~~~~~~~~~~~~~~~~~~~~~ # significant at al pha = 0.10 #~~~~~~~~~~~~~~~~~~~~~~~~~~ AirDif < - occu(~ AirDif ~ 1, EMR.data) # close to significant! backTransform(linearComb(AirDif, coefficients = c(1,0), type = 'det')) #~~~~~~~~~~~~~~~~~~~~~~~~~~ # non - significant #~~~~~~~~~~~~~~~~~~~~~~~~~~ StartMins < - occu(~ StartMins ~ 1, EMR.data) backTransform(linearComb(StartMins, coefficients = c(1,0), type = 'det')) SolRadSE < - occu(~ SolRadSE ~ 1, EMR.data) 148 backTransform(linearComb(SolRadSE, coefficients = c(1,0), type = 'det')) SubsiteCode < - occ u(~ SubsiteCode ~ 1, EMR.data) backTransform(linearComb(SubsiteCode, coefficients = c(1,0), type = 'det')) Prec < - occu(~ Prec ~ 1, EMR.data) backTransform(linearComb(Prec, coefficients = c(1,0), type = 'det')) SrchrCode < - occu(~ SrchrCode ~ 1, EMR.da ta) backTransform(linearComb(SrchrCode, coefficients = c(1,0), type = 'det')) RankSnake < - occu(~ RankSnake ~ 1, EMR.data) backTransform(linearComb(RankSnake, coefficients = c(1,0), type = 'det')) RankSnakePerSite < - occu(~ RankSnakePerSite ~ 1, EMR.da ta) backTransform(linearComb(RankSnakePerSite, coefficients = c(1,0), type = 'det')) OverallRank < - occu(~ OverallRank ~ 1, EMR.data) backTransform(linearComb(OverallRank, coefficients = c(1,0), type = 'det')) DaysSince < - occu(~ DaysSince ~ 1, EMR.dat a) backTransform(linearComb(DaysSince, coefficients = c(1,0), type = 'det')) EndMins < - occu(~ EndMins ~ 1, EMR.data) backTransform(linearComb(EndMins, coefficients = c(1,0), type = 'det')) PrevDayMinNOAA < - occu(~ PrevDayMinNOAA ~ 1, EMR.data) backTransform(linearComb(PrevDayMinNOAA, coefficients = c(1,0), type = 'det')) SurfAvg < - occu(~ SurfAvg ~ 1, EMR.data) backTransform(linearComb(SurfAvg, coefficients = c(1,0), type = 'det')) SurfMax < - occu(~ SurfMax ~ 1, EMR.data) backTransform(linea rComb(SurfMax, coefficients = c(1,0), type = 'det')) SolRadAvg < - occu(~ SolRadAvg ~ 1, EMR.data) backTransform(linearComb(SolRadAvg, coefficients = c(1,0), type = 'det')) SolRadMin < - occu(~ SolRadMin ~ 1, EMR.data) 149 backTransform(linearComb(SolRadMin, coefficients = c(1,0), type = 'det')) SolRadMax < - occu(~ SolRadMax ~ 1, EMR.data) backTransform(linearComb(SolRadMax, coefficients = c(1,0), type = 'det')) SolRadSD < - occu(~ SolRadSD ~ 1, EMR.data) backTransform(linearComb(SolRadSD, coefficients = c (1,0), type = 'det')) HumAvg < - occu(~ HumAvg ~ 1, EMR.data) backTransform(linearComb(HumAvg, coefficients = c(1,0), type = 'det')) HumMin < - occu(~ HumMin ~ 1, EMR.data) backTransform(linearComb(HumMin, coefficients = c(1,0), type = 'det')) HumMax < - occu(~ HumMax ~ 1, EMR.data) backTransform(linearComb(HumMax, coefficients = c(1,0), type = 'det')) difNOAAminSurfmin < - occu(~ difNOAAminSurfmin ~ 1, EMR.data) backTransform(linearComb(difNOAAminSurfmin, coefficients = c(1,0), type = 'det')) difNOAAminAirmax < - occu(~ difNOAAminAirmax ~ 1, EMR.data) backTransform(linearComb(difNOAAminAirmax, coefficients = c(1,0), type = 'det')) difNOAAminAirmin < - occu(~ difNOAAminAirmin ~ 1, EMR.data) backTransform(linearComb(difNOAAminAirmin, coefficien ts = c(1,0), type = 'det')) PropHerb < - occu(~ PropHerb ~ 1, EMR.data) backTransform(linearComb(PropHerb, coefficients = c(1,0), type = 'det')) PropWood < - occu(~ PropWood ~ 1, EMR.data) backTransform(linearComb(PropWood, coefficients = c(1,0), type = 'det')) SqmHerb < - occu(~ SqmHerb ~ 1, EMR.data) backTransform(linearComb(SqmHerb, coefficients = c(1,0), type = 'det')) SqmWood < - occu(~ SqmWood ~ 1, EMR.data) backTransform(linearComb(SqmWood, coefficients = c(1,0), type = 'det')) herbtowood < - occ u(~ herbtowood ~ 1, EMR.data) 150 backTransform(linearComb(herbtowood, coefficients = c(1,0), type = 'det')) #SiteOccu < - occu(~ SiteOccu ~ 1, EMR.data) # Does not converge . Omitted from analysis. #SoilMax < - occu(~ SoilMax ~ 1, EMR.data) # Does not converge . Omitted from analysis. #SoilAvg < - occu(~ SoilAvg ~ 1, EMR.data) # Does not converge . Omitted from analysis. #SoilMin < - occu(~ SoilMin ~ 1, EMR.data) # sig nificant , but does not converge . Omitted from analysis. ############################## # AIC table - Univariates only ############################## # AIC table for all univariate models (except models that did not converge, SiteOccu, SoilMax, SoilAvg, and SoilMin) univar.set.all < - list(EMRfm1, SubsiteCode,Prec,SrchrCode,RankSnake,RankSnakeP erSite,OverallRank, SurveyOrder,DaysSince,Julian,StartMins,EndMins,TotalMins,MMTs,SearchMins,AirAvg, AirMin,AirMax,AirDif,PrevDayMinNOAA,SameDayMinNOAA,SurfAvg,SurfMin,SurfMax,SolRadAvg, SolRadMin,SolRadMax,SolRadSD,SolRadSE,HumAvg,HumMin,HumMax,difNOAAminSurfmin, difNOAAminAirmax,difNOAAminAirmin,PropHerb,PropWood,SqmHerb,SqmWood,herbtowood) allmodelsAIC < - aictab(univar.set.all, modnames=c("EMRfm1","SubsiteCode","P rec","SrchrCode","RankSnake", "RankSnakePerSite","OverallRank","SurveyOrder", "DaysSince","Julian","StartMins","EndMins","TotalMins", "MMTs","SearchMins","AirAvg","AirMin","AirMax","AirDif", "PrevDayMinNOAA","SameDayMinNOAA","SurfAvg","SurfMin", "SurfMax","SolRa dAvg","SolRadMin","SolRadMax","SolRadSD", "SolRadSE","HumAvg","HumMin","HumMax","difNOAAminSurfmin", "difNOAAminAirmax","difNOAAminAirmin","PropHerb", "PropWood","SqmHerb","SqmWood","herbtowood"), second.ord=FALSE, nobs=NULL, sort=TRUE) ############################## # Multivariate Models ############################## # All significant at alpha = 0.05 #~~~~~~~~~~~~~~~~~~~~~~~~~~ multivar.sig_0.05 < - occu(~ TotalMins + MMTs + SearchMins + AirMin + SameDayMinNOAA + AirAvg + AirMax + SurfMin + SurveyOrder + Julian ~ 1, EMR.data) backTransform(linearCo mb(multivar.sig_0.05, coefficients = c(1,0,0,0,0,0,0,0,0,0,0), type = 'det')) #AIC table multivar.sig_0.05_AIC < - list(TotalMins,MMTs,SearchMins,AirMin,SameDayMinNOAA,AirAvg,AirMax,SurfMin, 151 SurveyOrder,Julian,EMRfm1) aictab(m ultivar.sig_0.05_AIC, modnames=c("TotalMins","MMTs","SearchMins","AirMin","SameDayMinNOAA", "AirAvg","AirMax","SurfMin","SurveyOrder","Julian","EMRfm1"), second.ord=FALSE, nobs=NULL, sort=TRUE) # In "multiva r.sig_0.05", the following covariates are correlated with one another: # TotalMins corr w/ MMTs # TotalMins corr w/ SearchMins # MMTs corr w/ SearchMins # AirMin corr w/ AirAvg # AirMin corr w/ AirM ax # AirMin corr w/ SurfMin # SameDayMinNOAA corr w/ SurfMin # SameDayMinNOAA corr w/ Julian # AirAvg corr w/ AirMax # AirAvg corr w/ SurfMin # AirMax corr w/ SurfMin #~~~~~~~~~~~~~~~~~~~~~~~~~~ # Final Set #~~~~~~~~~~~~~~~~~~~~~~~~~~ # From all covariates significant at 0.05: # Omitted TotalMins and MMTs because they are quite redundant with SearchMins (and not relevant to people # implementing the surveys) # Omitted AirAvg and AirMax beca use they are redundant with AirMin and rank lower with AIC than AirMin # SearchMins, AirMin, SameDayMinNOAA, SurveyOrder, Julian, SurfMin FinalSet < - occu(~ SearchMins + AirMin + SameDayMinNOAA + SurfMin + SurveyOrder + Julian ~ 1, EMR.data) backTransform (linearComb(FinalSet, coefficients = c(1,0,0,0,0,0,0), type = 'det')) #AIC table # second.ord=TRUE for AICc, second.ord=FALSE for AIC . FinalSet_AIC < - list(SearchMins,AirMin,SameDayMinNOAA,SurveyOrder,Julian,SurfMin,EMRfm1) aictab(FinalSet_AIC, modnames= c("SearchMins","AirMin","SameDayMinNOAA","SurveyOrder","Julian","SurfMin", "EMRfm1"),second.ord=FALSE, nobs=NULL, sort=TRUE) # In "FinalSet", the following covariates are correlated with one another: # SurfMin corr w/ AirMin # SurfMin corr w/ SameDayMinNOAA 152 # Julian corr w/ SameDayMinNOAA #~~~~~~~~~~~~~~~~~~~~~~~~~~ # Non - correlated Final Models #~~~~~~~~~~~~~~~~~~~~~~~~~~ # All combinations of our FinalSet covariates # No combination contains correlated covariates; all are significant at alpha = 0.05 SearchMins_SurfMin_SurveyOrder_Julian < - occu(~ SearchMins + SurfMin + SurveyOrder + Julian ~ 1, EMR.data) backTransform(linearComb(SearchMins_SurfMin_SurveyOrder_Julian, coefficients = c(1,0,0,0, 0), type = 'det')) SearchMins_SurfMin_SurveyOrder < - occu(~ SearchMins + SurfMin + SurveyOrder ~ 1, EMR.data) backTransform(linearComb(SearchMins_SurfMin_SurveyOrder, coefficients = c(1,0,0,0), type = 'det')) SearchMins_SurfMin_Julian < - occu(~ SearchMin s + SurfMin + Julian ~ 1, EMR.data) backTransform(linearComb(SearchMins_SurfMin_Julian, coefficients = c(1,0,0,0), type = 'det')) SearchMins_SurveyOrder_Julian < - occu(~ SearchMins + SurveyOrder + Julian ~ 1, EMR.data) backTransform(linearComb(SearchMins_ SurveyOrder_Julian, coefficients = c(1,0,0,0), type = 'det')) SurfMin_SurveyOrder_Julian < - occu(~ SurfMin + SurveyOrder + Julian ~ 1, EMR.data) backTransform(linearComb(SurfMin_SurveyOrder_Julian, coefficients = c(1,0,0,0), type = 'det')) SearchMins_Sur fMin < - occu(~ SearchMins + SurfMin ~ 1, EMR.data) backTransform(linearComb(SearchMins_SurfMin, coefficients = c(1,0,0), type = 'det')) SearchMins_SurveyOrder < - occu(~ SearchMins + SurveyOrder ~ 1, EMR.data) backTransform(linearComb(SearchMins_SurveyOrde r, coefficients = c(1,0,0), type = 'det')) SearchMins_Julian < - occu(~ SearchMins + Julian ~ 1, EMR.data) backTransform(linearComb(SearchMins_Julian, coefficients = c(1,0,0), type = 'det')) SurfMin_SurveyOrder < - occu(~ SurfMin + SurveyOrder ~ 1, EMR.dat a) backTransform(linearComb(SurfMin_SurveyOrder, coefficients = c(1,0,0), type = 'det')) SurfMin_Julian < - occu(~ SurfMin + Julian ~ 1, EMR.data) backTransform(linearComb(SurfMin_Julian, coefficients = c(1,0,0), type = 'det')) SurveyOrder_Julian < - occu( ~ SurveyOrder + Julian ~ 1, EMR.data) 153 backTransform(linearComb(SurveyOrder_Julian, coefficients = c(1,0,0), type = 'det')) SearchMins_AirMin_SurveyOrder_Julian < - occu(~ SearchMins + AirMin + SurveyOrder + Julian ~ 1, EMR.data) backTransform(linearComb(Se archMins_AirMin_SurveyOrder_Julian, coefficients = c(1,0,0,0,0), type = 'det')) SearchMins_AirMin_SurveyOrder < - occu(~ SearchMins + AirMin + SurveyOrder ~ 1, EMR.data) backTransform(linearComb(SearchMins_AirMin_SurveyOrder, coefficients = c(1,0,0,0), typ e = 'det')) SearchMins_AirMin_Julian < - occu(~ SearchMins + AirMin + Julian ~ 1, EMR.data) backTransform(linearComb(SearchMins_AirMin_Julian, coefficients = c(1,0,0,0), type = 'det')) AirMin_SurveyOrder_Julian < - occu(~ AirMin + SurveyOrder + Julian ~ 1, EMR.data) backTransform(linearComb(AirMin_SurveyOrder_Julian, coefficients = c(1,0,0,0), type = 'det')) SearchMins_AirMin < - occu(~ SearchMins + AirMin ~ 1, EMR.data) backTransform(linearComb(SearchMins_AirMin, coefficients = c(1,0,0), type = 'det')) Ai rMin_SurveyOrder < - occu(~ AirMin + SurveyOrder ~ 1, EMR.data) backTransform(linearComb(AirMin_SurveyOrder, coefficients = c(1,0,0), type = 'det')) AirMin_Julian < - occu(~ AirMin + Julian ~ 1, EMR.data) backTransform(linearComb(AirMin_Julian, coefficients = c(1,0,0), type = 'det')) SearchMins_AirMin_SurveyOrder_SameDayMinNOAA < - occu(~ SearchMins + AirMin + SurveyOrder + SameDayMinNOAA ~ 1, EMR.data) backTransform( linearComb(SearchMins_AirMin_SurveyOrder_SameDayMinNOAA, coefficients = c(1,0,0,0,0), type = 'det')) SearchMins_AirMin_SameDayMinNOAA < - occu(~ SearchMins + AirMin + SameDayMinNOAA ~ 1, EMR.data) backTransform(linearComb(SearchMins_AirMin_SameDayMinNOAA, coefficients = c(1,0,0,0), type = 'det')) SearchMins_SurveyOrder_SameDayMinNOAA < - occu(~ SearchMins + SurveyOrder + SameDayMinNOAA ~ 1, EMR.data) backTransform(linearComb(SearchMins_SurveyOrder_SameDayMinNOAA, coefficients = c(1,0,0,0), type = 'det')) A irMin_SurveyOrder_SameDayMinNOAA < - occu(~ AirMin + SurveyOrder + SameDayMinNOAA ~ 1, EMR.data) backTransform(linearComb(AirMin_SurveyOrder_SameDayMinNOAA, coefficients = c(1,0,0,0), type = 'det')) SearchMins_SameDayMinNOAA < - occu(~ SearchMins + SameDayM inNOAA ~ 1, EMR.data) backTransform(linearComb(SearchMins_SameDayMinNOAA, coefficients = c(1,0,0), type = 'det')) 154 AirMin_SameDayMinNOAA < - occu(~ AirMin + SameDayMinNOAA ~ 1, EMR.data) backTransform(linearComb(AirMin_SameDayMinNOAA, coefficients = c(1,0,0 ), type = 'det')) SurveyOrder_SameDayMinNOAA < - occu(~ SurveyOrder + SameDayMinNOAA ~ 1, EMR.data) backTransform(linearComb(SurveyOrder_SameDayMinNOAA, coefficients = c(1,0,0), type = 'det')) #~~~~~~~~~~~~~~~~~~~~~~~~~~ # Univariate models, and the null model #~~~~~~~~~~~~~~~~~~~~~~~~~~ #AIC table # All combinations of the covariates included in FinalSet . # Each combination contains non - correlated covariates, and all covariates are significant at alpha = 0.05 . # 32 SINGLE SEASON OCCUPANCY MODELS USING THE 6 FINAL COVARIATES: (there are 32 models) FinalAll_list < - list(AirMin,AirMin_Julian,AirMin_SameDayMinNOAA,AirMin_SurveyOrder, AirMin_SurveyOrder_Julian,AirMin_SurveyOrder_SameDayMinNOAA,EMRfm1,Julian, SameD ayMinNOAA,SearchMins,SearchMins_AirMin,SearchMins_AirMin_Julian, SearchMins_AirMin_SameDayMinNOAA,SearchMins_AirMin_SurveyOrder, SearchMins_AirMin_SurveyOrder_Julian,SearchMins_AirMin_SurveyOrder_SameDayMinNOAA, SearchMins_Julian,SearchMins_SameDayMinNOAA,SearchMins_SurfMin, SearchMins_SurfMin_Julian,SearchMins_SurfMin_SurveyOrder, SearchMins_SurfMin_SurveyOrder_Julian,SearchMins_SurveyOrder, SearchMins_SurveyOrder_Julian,SearchMins_SurveyOrder_SameDayMinNOAA,SurfMin, SurfMin_Julian,SurfMin_SurveyOrder,SurfMin_SurveyOrder_Julian,SurveyOrder, SurveyOrder_Julian,SurveyOrder_SameDayMinNOA A) FinalAll_AICc_table < - aictab(FinalAll_list, modnames=c("AirMin","AirMin_Julian","AirMin_SameDayMinNOAA", "AirMin_SurveyOrder","AirMin_SurveyOrder_Julian", "AirMin_SurveyOrder_SameDayMinNOAA","EMRfm1", "Julian","SameDayMinNOAA","SearchMins", "SearchMins_AirMin","SearchMins_AirMin_Juli an", "SearchMins_AirMin_SameDayMinNOAA", "SearchMins_AirMin_SurveyOrder", "SearchMins_Ai rMin_SurveyOrder_Julian", "SearchMins_AirMin_SurveyOrder_SameDayMinNOAA", "SearchMins_Julian","SearchMins_SameDayMinNOAA", "SearchMins_SurfMin","SearchMins_SurfMin_Julian", "SearchMins_SurfMin_SurveyOrder", "SearchMins_SurfMin_Surve yOrder_Julian", 155 "SearchMins_SurveyOrder", "SearchMins_SurveyOrder_Julian", "SearchMins_SurveyOrder_SameDayMinNOAA","SurfMin", "SurfMin_Julian","SurfMin_SurveyOrder", "SurfMin_SurveyOrder_Julian","SurveyOrder", "SurveyOrder_Julian","SurveyOrder_SameDayMinNOAA"), second.ord=TRUE, nobs=NULL, sort=TRUE) # note: second.ord=TRUE for AICc , =FALSE for AIC ################################ # Model Averaging (using package MuMIn) ################################ # All models included: FinalAll_AICc_modavg < - model.avg(FinalAll_list) FinalAll_AICc_modavg_summary < - summary(model.avg(FinalAll_l ist)) # model average everything where delta < 7 (Burnham and Anderson 2002) # create a new model list to subset only models where delta < 7 based on the model.avg of the full model FinalAll_subset_list_delta7 < - list(SearchMins,SearchMins_AirMin,SearchMi ns_AirMin_Julian, SearchMins_AirMin_SameDayMinNOAA,SearchMins_AirMin_SurveyOrder, SearchMins_AirMin_SurveyOrder_Julian, SearchMins_AirMin_SurveyOrde r_SameDayMinNOAA,SearchMins_Julian, SearchMins_SameDayMinNOAA,SearchMins_SurfMin,SearchMins_SurfMin_Julian, SearchMins_SurfMin_SurveyOrder,SearchMins_SurfMin_SurveyOrder_Julian, SearchMins_SurveyOrder,SearchMins_SurveyOrder_Julian, SearchMins_SurveyOrder_SameDayMinNOAA) # Only models where delta < 7 included: FinalAll_subset_AICc_delta7_modavg < - model.avg(FinalAll _subset_list_delta7) FinalAll_subset_AICc_delta7_modavg_summary < - summary(model.avg(FinalAll_subset_list_delta7)) confint(FinalAll_subset_AICc_delta7_modavg_summary, full=TRUE) # full=TRUE gives the CI for the full averaged results; # full=FALSE is default and gives the CI for the conditional results (the subset) ############################## # PREDICTING ############################## # All scaled and unscaled code values for the following predict and ggplot code are from using 2016 # data only , and only including only the 7 occupied subsites. 156 # This includes our top univariate models based on the AIC table for FinalSet # SearchMins, AirMin #~~~~~~~~~~~~~~~~~~~~~~~~~~ # SearchMins_AirMin #~~~~~~~~~~~~~~~~~~~~~~~~~~ prediction_test_searchmins_ airmin < - data.frame(SearchMins = - 2.5:5.5, AirMin = - 2.5:5.5) #these are the standardized values # the ranges set here need to cover the range of BOTH covariates for each of the # covariates to be able to include all rows for each covariate . round( predict(SearchMins_AirMin, type = 'det', newdata = prediction_test_searchmins_airmin, appendData=TRUE), 3) #~~~~~~~~~~~~~~~~~~~~~~~~~~ # SearchMins #~~~~~~~~~~~~~~~~~~~~~~~~~~ predict_searchmins < - data.frame (SearchMins = - 3:6) #these bound the standardized values, see predict_searchmins_a predict_searchmins_a < - round(predict(SearchMins, type = 'det', newdata = predict_searchmins, appendData=TRUE), 3) predict_searchmins_df < - as.data.frame(predict_searchmin s_a) scatterplot(Predicted ~ SearchMins, data=predict_searchmins_df) #ggplot: searchmins_ggplot < - ggplot(predict_searchmins_df, aes(SearchMins, Predicted)) + geom_line() searchmins_ggplot < - searchmins_ggplot + geom_ribbon(data=predict_searchmins_df,aes( ymin=lower,ymax=upper) ,alpha=0.4) #alpha is the shade of the CI #The following line relabels the standardized numbers (the decimals, e.g., - 2.81247) with the # appropriate unstandardized associated "a ctual value" #(e.g., " - 2.81247,"="30") determined using the Excel spreadsheet from G. Roloff ("Plot formulas (from G Roloff)"): searchmins_ggplot < - searchmins_ggplot + scale_x_continuous(name="Minutes Spent Actively Searching", breaks=c( - 2.81247, - 2.05594, - 1.29942, - 0.54290, 0.21363,0.97015,1.72668,2.48320, 3.239 72,3.99625,4.75277,5.50930), limits=c( - 3,6), labels=c(" - 2.81247,"="30"," - 2.05594"="40", " - 1.29942"="50"," - 0.54290"="60", "0.21363"="70","0.97015"="80", "1.72668"="90","2.48320"="100", "3.23972"="110","3.99625"="120", "4.75277"="130","5.50930"="140")) 157 searchmins_ggplot < - searchmins_ggplot + scale_y_co ntinuous(name="Predicted Detection Probability", breaks=c(0.0,0.2,0.4,0.6,0.8,1.0)) #variance - covariance matrix vcov(SearchMins) #~~~~~~~~~~~~~~~~~~~~~~~~~~ # AirMin: #~~~~~~~~~~~~~~~~~~~~~~~~~~ predict_airmin < - data.frame(AirMin = - 2.5:2.5) #these bound the standardized values, #see in #predict_airmin_a predict_airmin_a < - round(predict(AirMin, type = 'det', newdata = p redict_airmin, appendData=TRUE), 3) predict_airmin_df < - as.data.frame(predict_airmin_a) scatterplot(Predicted ~ AirMin, data=predict_airmin_df) #ggplot: airmin_ggplot < - ggplot(predict_airmin_df, aes(AirMin, Predicted)) + geom_line() airmin_ggplot < - air min_ggplot + geom_ribbon(data=predict_airmin_df,aes(ymin=lower,ymax=upper),alpha=0.4) #alpha is the shade of the CI #The following line relabels the standardized numbers (the decimals, e.g., - 3.13278) with the #appropriate unstandardized associated "actu al value" (e.g., " - 3.13278"="55") determined using the Excel #spreadsheet from G. Roloff ("Plot formulas (from G Roloff)"): airmin_ggplot < - airmin_ggplot + scale_x_continuous(name="Minimum Air Temperature During Survey (Deg F)", breaks=c( - 3.13278, - 2.37675, - 1.62071, - 0.86468, - 0.10864, 0.64739,1.40343,2.15946,2.91550), limits=c( - 2.5, 2.5), labels=c(" - 3.13278"="55"," - 2.37675"="60", " - 1.62071"="65"," - 0.86468"="70", " - 0.10864"="75 ","0.64739"="80", "1.40343"="85","2.15946"="90", "2.91550"="95")) airmin_ggplot < - airmin_ggplot + scale_y_continuous(name="Predicted Detection Probability", breaks=c(0.0,0.2,0.4,0.6,0.8,1.0)) #variance - covariance matrix vcov(AirMin) 158 APPENDIX H Documentation of T hree M ortality E vents 2015 : Site SE_H3 - Adult Female Adult female (gravid , PIT tag # 836546351 ; Appendix D ) massasauga was first captured on 21 May 2015 . Brought in for surgical implantation of radiotransmitter (7g ATS) on 22 May, released back at location of capture on 26 May 201 5 . Continued to collect loc ations and closely monitored this individual as we expected parturition to occur in early - mid August . Following a final visual observation on 4 Aug where behavior appeared normal , the massasauga moved to a different location and remained unseen for the nex t two telemetry locations (6 Aug and 7 Aug), then moved to a final location and remained unseen for one telemetry location (10 Aug) . Upon our second visit to this final location (12 Aug), we noticed the ground had been slightly dug up (apparently by a scav enging animal as claw marks were seen) where we had assumed the female was burrowed for parturition . Upon further inspection, we discovered the snake carcass and the transmitter was recovered (PIT tag scanner confirmed the identity of this massasauga) . Car cass was too decomposed for necropsy to determine cause of mortality . Prior to this finding, w e saw no indication that parturition had occurred at any of the final locations (i.e., no signs of neonates or neonatal ecdysis) . 2015: Site SE_H1 - Adult Male Adult male (PIT tag # 836574862 ; Appendix D ) captured and transported for surgical implantation of radiotransmitter on 17 Jun 2015 . Following surgery, released at location of capture on 20 Jun . Located via telemetry approximately twice per week until last observed alive on 24 July 2015 where the massasauga was found basking in the garden of the landowner whose property we were allowed access to for this research . Radiotracked again on 28 Jul and found 159 with no visual observation of the massasauga . On 31 Jul, radiotracking again behind the house revealed the radiotransmitter laying at the base of a tree with the anchoring suture used by the veterinarian still intac t around the transmitter with a massasauga rib attached . Bite marks found on the transmitter may indicate that this individual was depredated, and the landowner informed us that he had seen many raccoons around his house recently . Additionally, t his individual tested positive for snake fungal disease ( Appendix B ) but it is unknown if this influenced the mortality of the individual . 2016 : Site SE_H1 - Adult Female Adult female (nongravid , PIT tag # 840525094 ; Appendix D ) captured and fitted with an external radiotransmitter on 20 Jul 2016 . During captu re and handling, we noted sluggish behavior and an emaciated appearance . Released at location of capture the same day . Relocating the individual via radiotelemetry on 25 Jul led us to the carcass of this massasauga where a PIT tag scan identified the indiv idual and the radiotransmitter was recovered . Carcass too decomposed for necropsy. 160 APPENDIX I Data S heets The following pages (Figures I. 1 - I. 4) illustrate and describe the data sheets used during the 2015 and 2016 field seasons for the eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) study throughout southern Michigan in 2015 and 2016 . 161 Figure I. 1. Data sheet used to record all pertinent information for captured eastern massasauga rattlesnakes ( Sistruru s catenatus catenatus ) during the 2015 and 2016 field seasons throughout southern Michigan. 162 Figure I. 2. Data sheet used to record all vegetation characteristics measured during the 2015 and 2016 field seasons for assessing eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) habitat throughout southern Michigan based on the Bailey (2010) habitat suitability index model. 163 Figure I. 3. Detection survey data sheet used during eastern massasauga rattlesnake ( Sistrurus catenatus catenatus ) detection surveys (see Chapter 2) conducted throughout southern Michigan. developed during the 2015 and 2015 field seasons. 164 Figure I. 4. Postoperative monitoring data sheet (Bailey 2010) used following radiotransmitter implantation surgeries f or eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) captured throughout southern Michigan during the 2015 and 2016 field seasons . 165 LITERATURE CITED 166 LITERATURE CITED Allen, A.W. 1987. Habitat suitability index models: barred owl. U.S. Fish Wild. Serv. Biol. Rep. 82(10.143). 17 pp. Allender, M.C., D.B. Raudabaugh, F.H. Gleason, and A.N. Miller. 2015. The natural history, ecology, and epidemiology of Ophidiomyces ophiodiicola and its potential impact on free - ranging sna ke populations. Fungal Ecology 17:187 - 196. Allender, M.C., M. Dreslik, S. Wylie, C. Phillips, D.B. Wylie, C. Maddox, M.A. Delaney, and M.J. Kinsel. 2011. Chrysosporium sp. Infection in eastern massasauga rattlesnakes. Emerging Infectious Diseases 17(12):2 383 - 2384. Allender. M.C., E.T. Hileman, J. Moore, and S. Tetzlaff. 2016. Detection of Ophidiomyces, the causative agent of snake fungal disease, in the eastern massasauga ( Sistrurus catenatus ) in Michigan, USA, 2014. Journal of Wildlife Diseases 52(3):694 - 698. Bailey, R.L, H. Campa, III, T.M. Harrison, and K. Bissell. 2011. Survival of eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) in Michigan. Herpetologica 67:167 173. Bailey, R.L. 2010. Modeling habitat suitability and population demog raphics of the eastern massasauga rattlesnake in managed lands in southwestern Michigan. Thesis, Michigan State University, East Lansing, USA. Bailey, R.L., H. Campa, III, T.M. Harrison, and K. Bissell. 2011. Survival of eastern massasauga rattlesnakes ( S istrurus catenatus catenatus ) in Michigan. Herpetologica 67(2):167 - 173. Bailey, R.L., H. Campa, III, T.M. Harrison, and K. Bissell. 2012. Resource selection by the eastern massasauga rattlesnake on managed land in southwestern Michigan. Journal of Wildlif e Management 76:414 - 421. Baker, S.J. 2016. Life and death in a corn desert oasis: reproduction, mortality, genetic diversity, University of Illinois at Urbana - Champai gn, Urbana, USA. Bart, J., and D.S. Robson. 1982. Estimating survivorship when the subjects are visited periodically. Ecology 63(4):1078 - 1090. Bissell, K. 2006. Modeling habitat ecology and population viability of the eastern massasauga rattlesnake in so uthwestern lower Michigan. Thesis, Michigan State University, East Lansing, USA. Bunck, C.M., and K.H. Pollock. 1993. Estimating survival of radio - tagged birds. Pages 51 63 in J.C. Lebreton and P.M. North, editors. Marked individuals in the study of bird populations. Birkhäuser Verlag, Basel, Switzerland. 167 Burnham, K.P., and D.R. Anderson. 2002. Model Selection and Multimodel Inference: A Practical Information - Theoretical Approach. Second Edition. Springer - Verlag, New York, New York, USA. Buskirk, S.W., and J.J. Millspaugh. 2006. Metrics for studies of resource selection. Journal of Wildlife Management 70(2):358 - 366. Cake, E.W., Jr. 1983. Habitat suitability index models: Gulf of Mexico American oyster. United States Department of the Interior Fish and Wildlife Service FWS/OBS - 82/10.57. 37 p. Calenge, C. 2006. The package adehabitat for the R software: a tool for the analysis of space and habitat use by animals. Ecological Modelling 197:516 - 519 Calenge, C. 2015. Home range estimation in R: the adehabitatHR package. In CRAN. Casper, G.S., T.G. Anton, R.W. Hay, A.T. Holycross, R.S. King, B.A. Kingsbury, D. Mauger, C. Parent, C.A. Phillips, A. Resetar, R.A. Seigel, T.P. Wilson. 2001. Recommended st andard survey protocol for the eastern massasauga, Sistrurus catenatus catenatus . Report for the Milwaukee Public Museum, Milwaukee, Wisconsin, USA. Christy, M.T., A.A. Yackel Adams, G.H. Rodda, J.A. Savidge, and C.L. Tyrrell. 2010. Modelling detection pr obabilities to evaluate management and control tools for an invasive species. Journal of Applied Ecology 47:106 113. Cobb, V.A., J.J. Green, T. Worrall, J. Pruett, and B. Glorioso. 2005. Initial den location behavior in a litter of neonate Crotalus horrid us (timber rattlesnakes). Southeastern Naturalist 4(4):723 - 730. Congdon, J.D., A.E. Dunham, R.C. van Loben Sels. 1994. Demographics of common snapping turtles ( Chelydra serpentina ): implications for conservation and management of long - lived organisms. Ame rican Zoologist 34(3):397 - 408. Cross, M.D., K.V. Root, C.J. Mehne, J. McGowan - Stinski, and D. Pearsall. 2015. Multi - scale responses of eastern massasauga rattlesnakes ( Sistrurus catenatus ) to prescribed fire. American Midland Naturalist 173:346 - 362. Dol inski, A.C., M.C. Allender, V. Hsiao, C.W. Maddox. 2014. Systemic Ophidiomyces ophiodiicola infection in a free - ranging plains garter snake ( Thamnophis radix ). Journal of Herpetological Medicine and Surgery 24(1 - 2):7 - 10. Donini, J.T., W. Selman, and R. A. Valverde. 2017. A comparison of reproductive assessment techniques to determine the reproductive status of female diamondback terrapins ( Malaclemys terrapin ). Herpetological Review 48(4):763 - 766. 168 Conservation potential of prescribed fire for maintaining habitats and populations of an endangered rattlesnake Sistrurus c. catenatus . Endangered Species Research 22:51 - 60. Durbian, F.E. 2006. Effects of mowing and summer burning on the massasauga ( Sist rurus catenatus ). American Midland Naturalist 155(2):329 - 334. Durbian, F.E., R.S. King, T. Crabill, H. Lambert - Doherty, and R.A. Seigel. 2008. Massasauga home range patterns in the Midwest. Journal of Wildlife Management 72(3):754 - 759. Durso, A.M., J.D. Willson, and C.T. Winne. 2011. Needles in haystacks: Estimating detection probability and occupancy of rare and cryptic snakes. Biological Conservation 144(2011):1508 - 1515. Dussault, C., R. Courtois, and J.P. Ouellet. 2006. A habitat suitability index mod el to assess moose habitat selection at multiple spatial scales. Canadian Journal of Forest Research 36:1097 - 1107. Elmore, R.D., J.M. Carroll, E.P. Tanner, T.J. Hovick, B.A. Grisham, S.D. Fuhlendorf, and S.K. Windels. 2017. Implications of the thermal env ironment for terrestrial wildlife management. Wildlife Society Bulletin 41(2):183 193. Engberg, C.A., and F.R. Austin. 1974. Soil Survey of Livingston County, Michigan. U.S. Department of Agriculture, Soil Conservation Service, Washington, D.C., USA. Eng el, R.J. 1977. Soil Survey of Washtenaw County, Michigan. US Department of Agriculture, Soil Conservation Service, Washington, D.C., USA. Eskew, E.A., and B.D. Todd. 2017. Too cold, too wet, too bright, or just right? Environmental predictors of snake mov ement and activity. Copeia 105(3):584 - 591. Feenstra, J.E. 1982. Soil Survey of Oakland County, Michigan. US Department of Agriculture, Soil Conservation Service, Washington, D.C., USA. Fiske, I., and R. Chandler. 2011. Unmarked: an R package for fitting hierarchical models of wildlife occurrence and abundance. Journal of Statistical Software 43(10):1 23. Fiske, I., and R. Chandler. 2015. Overview of unmarked: an R package for the analysis of data from unmarked animals. In CRAN. Fiske, I., R. Chandler, J.A. Royle, and M. Kery. 2011. unmarked: Models for Data from Unmarked Animals. R package version 0.9 0. Foster, M.A., K.M. Bissell, H. Campa, III, and T. Myers Harrison. 2009. The influence of reproductive status on thermal ecology and vegetation use of female eastern massasauga 169 rattlesnakes ( Sistrurus catenatus catenatus ) in southwestern Michigan. Herpetological Conservation and Biology 4(1):48 - 54. Lockey, and R.I. Ehrlich. 2018. Comparison of digital and film chest radiography for detection and medical surveillance of silicosis in a setting with a high burden of tuberculosis. American Journal of Industrial Medicine 61:229 - 238. Gaillard, J.M., M. Hebblewhite, A . Loison, M. Fuller, R. Powell, M. Basille, and B. Van Moorter. 2010. Habitat - performance relationships: finding the right metric at a given spatial scale. Philosophical Transactions of the Royal Society B 365:2255 - 2265. Harvey, D.S. 2005. Detectability of a large - bodied snake ( Sistrurus c. catenatus ) by time - constrained searching. Herpetological Review 36(4):413 415. Harvey, D.S., and P.D. Weatherhead. 2010. Habitat selection as the mechanism for thermoregulation in a northern population of massasauga r attlesnakes ( Sistrurus catenatus ). Écoscience 17(4):411 419. 2002. Performance measures for ecosystem management and ecological sustainability. Technical Review 02 - 1. The Wildlife Society, Bethesda, Maryland, USA. Hileman, E.R., M.C. Allender, D.R. Bradke, L.J. Faust, J.A. Moore, M.J. Ravesi, and S.J. Tetzlaff. 2018. Estimation of Ophidiomyces prevalence to evaluate snake fungal disease risk. Journal of Wildlif e Management 82(1):173 - 181. Hileman, E.T., R.B. King, L.J. Faust. 2018. Eastern massasauga demography and extinction risk under prescribed - fire scenarios. Journal of Wildlife Management DOI: 10.1002/jwmg.21457 Howze, J.M., K.M. Stohlgren, E.M. Schlimm, a nd L.L. Smith. 2012. Dispersal of neonate timber rattlesnakes ( Crotalus horridus ) in the southeastern coastal plain. Journal of Herpetology 46(3):417 - 422. Jellen, B.C., and M.J. Kowalski. 2007. Movement and growth of neonate eastern massasaugas ( Sistrurus catenatus ). Copeia 2007(4):994 - 1000. Johnson, B.D., J.P. Gibbs, K.T. Shoemaker, and J.B. Cohen. 2016. Demography of a small and isolated population of eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ) threatened by vegetative succession. Journal of Herpetology 50(4):534 - 540. Johnson, B.D., J.P. Gibbs, T.A. Bell, Jr., K.T. Shoemaker. 2016. Manipulation of basking sites for endangered eastern massasauga rattlesnakes. Journal of Wildlife Management 80(5):803 - 811. 170 Johnson, G. 1995. Spatial e cology, habitat preference, and habitat management of the eastern massasauga, Sistrurus c. catenatus in a New York weakly - minerotrophic peatland. PhD Dissertation. State University of New York, Syracuse, New York. Johnson, G. 2000. Spatial ecology of the eastern massasauga ( Sistrurus c. catenatus ) in a New York peatland. Journal of Herpetology 34(2):186 192. Johnson, G., B. Kingsbury, R. King, C. Parent, R. Seigel, and J. Szymanski. 2000. The eastern massasauga rattlesnake: a handbook for land managers. U .S. Fish and Wildlife Service, Fort Snelling, Minnesota, USA. Jones, P.C., R.B. King, R.L. Bailey, N.D. Bieser, K. Bissell, H. Campa, T. Crabill, M.D. Cross, B.A. Degregorio, M.J. Dreslik, F.E. Durbian, D.S. Harvey, S.E. Hecht, B.C. Jellen, G. Johnson, B. A. Kingsbury, M.J. Kowalski, J. Lee, J.V. Manning, J.A. Moore, J. Oakes, C.A. Phillips, K.A. Prior, J.M. Refsnider, J.D. Rouse, J.R. Sage, R.A. Seigel, D.B. Shepard, C.S. Smith, T.J. Vandewalle, P.J. Weatherhead, and A. Yagi. 2012. Range - wide analysis of e astern massasauga survivorship. Journal of Wildlife Management 76:1576 1586. Kernohan, B. J., R. A. Gitzen, and J. J. Millspaugh. 2001. Analysis of animal space use and movements. Pages 125 166 in J. J. Millspaugh and J. H. Marzluff, editors. Radio tracki ng and animal populations. Academic Press, San Diego, California, USA. Lele, S.R, J.L. Keim, and P. Solymos. 2017. ResourceSelection: Resource Selection (Probability) Functions for Use - Availability Data. R package version 0.3 - 2. Lele, S.R., and J.L. Keim. 2006. Weighted distributions and estimation of resource selection probability functions. Ecology 87(12):3021 - 3028. Lentini, A. M., G.J. Crawshaw, L.E. Licht, and D.J. McLelland. 2011. Pathologic and hematologic responses to surgically implanted tran smitters in eastern massasauga rattlesnakes ( Sistrurus catenatus catenatus ). Journal of Wildlife Diseases 47(1):107 125. Lourdais, O., R. Shine, X. Bonnet, M. Guillon, and G. Naulleau. 2004. Climate affects embryonic development in a viviparous snake, Vip era aspis . Oikos 104:551 - 560. MacKenzie, D.I., J.D. Nichols, J.A. Royle, K.H. Pollock, L.L. Bailey, and J.E. Hines. 2006. Occupancy Estimation and Modeling: Inferring Patterns and Dynamics of Species Occurrence. First edition. Academic Press, San Diego, C alifornia, USA. Marshall, J.C., Jr., J.V. Manning, and B.A. Kingsbury. 2006. Movement and macrohabitat selection of the eastern massasauga in a fen habitat. Herpetologica 62(2):141 - 150. Mayfield, H. 1961. Nesting success calculated from exposure. The Wil son Bulletin 73(3):255 261. 171 Mazerolle, M.J., L.L. Bailey, W.L. Kendall, J.A. Royle, S.J. Converse, and J.D. Nichols. 2007. Making great leaps forward: accounting for detectability in herpetological field studies. Journal of Herpetology 41(4):672 689. McC luskey, E.M., S.N. Matthews, I.Y. Ligocki, M.L. Holding, G.J. Lipps, Jr., and T.E. Hetherington. 2018. The importance of historical land use in the maintenance of early successional habitat for a threatened rattlesnake. Global Ecology and Conservation 13:1 - 11. McLeese, R.L. 1981. Soil Survey of Jackson County, Michigan. US Department of Agriculture, Soil Conservation Service, Washington, D.C., USA. MDNR. 2001. 2001 Integrated forest management analysis program/Gap analysis program land cover. Forest, Mine ral, and Fire Management Department, Lansing, Michigan, USA. Miller, H.W., and D.H. Johnson. 1978. Interpreting the results of nesting studies. The Journal of Wildlife Management 42(3):471 - 476. MNFI. 2014. Michigan Natural Heritage Database; Michigan Nat ural Features Inventory. Lansing, Michigan, USA. Moore, J.A., and J.C. Gillingham. 2006. Spatial ecology and multi - scale habitat selection by a threatened rattlesnake: the eastern massasauga ( Sistrurus catenatus catenatus ). Copeia 2006(4):742 - 751. Moreno - Rueda, G., and J.M. Pleguezuelos. 2007. Long - term and short - term effects of temperature on snake detectability in the wild: a case study with Malpolon monspessulanus . Herpetological Journal 17:204 207. MRLC. 2015. National Land Cover Database 2006; Multi - Resolution Land Characteristics Consortium. Available at: http://www.mrlc.gov/index.php. Accessed 8 - Mar - 2015. Newsom, J.D., T. Joanen, and R.J. Howard. 1987. Habitat suitability index modes: American alligator. U.S. Fish Wildl. Serv. Biol. Rep. 82( 10.136). 14 pp. NOAA - NCEI. 2017. Climate Data Online Data Tools; National Oceanic and Atmospheric Administration National Centers for Environmental Information. Available at: https://www.ncdc.noaa.gov/cdo - web/datatools. Accessed 26 - May - 2017 NOAA - NWS. 201 7. Climate Data Annual Data for Jackson, MI 2011 - 2017; National Oceanic and Atmospheric Administration National Weather Service. Available at: http://www.weather.gov/grr/climate/plots. Accessed 30 - October - 2017 2014. 50 years of bat tracking: device attachment and future directions. Methods in Ecology and Evolution 5:311 - 319. 172 Plummer, M.V., and N.E. Mills. 2000. Spatial ecology and survivorship of resident and translocated hognose snakes ( Heterodon platirhinos ). Journal of Herpetology 34(4):565 - 575. Rajeev, S., D.A. Sutton, B.L. Wickes, D.L. Miller, D. Giri, M. Van Meter, E.H. Thompson, M.G. Rinaldi, A.M. Romanelli, J.F. Cano, and J. Guarro. 2009. Isolation and characterization of a new fungal species, Chrysospo rium ophiodiicola , from a mycotic granuloma of a black rat snake ( Elaphe obsoleta obsoleta ) Journal of Clinical Microbiology. 47(4):1264 - 1268. Reinert, H.K. 1981. Reproduction by the massasauga ( Sistrurus catenatus catenatus ). American Midland Naturalist 105(2):393 - 395. Reinert, H.K., and W.R. Kodrich. 1982. Movements and habitat utilization by the massasauga, Sistrurus catenatus catenatus . Journal of Herpetology 16(2):162 - 171. Roloff, G.J., and B.J. Kernohan. 1999. Evaluating reliability of habitat suit ability index models. Wildlife Society Bulletin 27(4):973 - 985. Schroeder, R.L. 1985. Habitat suitability index models: Eastern wild turke y . U.S. Fish Wildl. Serv. Biol. Rep. 82(10.106). 33 pp. Seigel, R.A. 1986. Ecology and conservation of an endangered rattlesnake, Sistrurus catenatus , in Missouri, USA. Biological Conservation 35(1986):333 - 346. Shoemaker, K.T., and J.P. Gibbs. 2010. Evaluating basking - habitat deficiency in the threatened eastern massasauga rattlesnake. Journal of Wildlife Management 74(3):504 513. Siegel, R.A. 1986. Ecology and conservation of an endangered rattlesnake, Sistrurus catenatus , in Missouri, USA. Biological Conservation 35(1986):333 - 346. Sigler, L., S. Hambleton, J.A. Paré. 2013. Molecular characterization of reptile pathogens currently known as members of the Chrysosporium anamorph of Nannizziopsis vriesii complex and relationship with some human - associated isolates. Journal of Clinical M icrobiology 51(10):3338 - 3357. Soniat, T.M., and M.S. Brody. 1988. Field validation of a habitat suitability index model for the American oyster. Estuaries 11(2):87 - 95. Steen, D.A. 2010. Snakes in the grass: secretive natural histories defy both conventio nal and progressive statistics. Herpetological Conservation and Biology 5(2):183 188. Steen, D.A., C.J.W. McClure, J.C. Brock, D.C. Rudolph, J.B. Pierce, J.R. Lee, W.J. Humphries, B.B. Gregory, W.B. Sutton, L.L. Smith, D.L. Baxley, D.J. Stevenson, and G. Guyer. 2012. Landscape - level influences of terrestrial snake occupancy within the southeastern United States. Ecological Applications 22)4):1084 - 1097. 173 Striker, M.M., and L.I. Harmon. 1961. Soil Survey of Lenawee County, Michigan. US Department of Agricultu re, Soil Conservation Service, Washington, D.C., USA. Szymanski, J. 1998. Status assessment for the eastern massasauga ( Sistrurus c. catenatus ). U.S. Fish and Wildlife Service, Ft. Snelling, Minnesota, USA. Tardy, S.W. 1997. Soil Survey of Calhoun County , Michigan. US Department of Agriculture, Soil Conservation Service, Washington, D.C., USA. Thoen, G.F. 1990. Soil Survey of Barry County, Michigan. US Department of Agriculture, Soil Conservation Service, Washington, D.C., USA. Thomasma, L.E., T.D. Drum mer, and R.O. Peterson. 1991. Testing the habitat suitability index model for the fisher. Wildlife Society Bulletin 19(3):291 - 297. United States Fish and Wildlife Service [USFWS]. 1981. Standards for the development of habitat suitability index models. Un ited States Fish and Wildlife Service, Release Number 1 - 81, 103 ESM. United States Fish and Wildlife Service [USFWS]. 2016. Endangered and Threatened Wildlife and Plants; Threatened Species Status for the Eastern Massasauga Rattlesnake. Federal Register 8 1(190):67193 - 67214. September 30, 2016. USDA - NRCS. 2017. Web Soil Survey; United States Department of Agriculture Natural Resources Conservation Service. Available at: https://websoilsurvey.nrcs.usda.gov/app/ Wiens, J.A. 1989. Spatial scaling in ecology. Functional Ecology 3(4):385 - 397. Willson, J.D., C.T. Winne, and B.D. Todd. 2011. Ecological and methodological factors affecting detectability and population estimation in elusive species. Journal of Wildlife Management 75(1):36 45. Winterstein, S.R., K .H. Pollock, and C.M. Bunck. 2001. Analysis of survival data from radiotelemetry studies. Pp. 351 - 380. In J.J. Millspaugh and J.H. Marzluff (eds.), Radio Tracking and Animal Populations. Academic Press, San Diego, California, USA. Zajac, Z., Stith, B., A. C. Bowling, C.A. Langtimm, and E.D.Swain. 2015. Evolution of habitat suitability index models by global sensitivity and uncertainty analyses: a case study for submerged aquatic vegetation. Ecology and Evolution 5(13):2503 - 2517. Zhang, Z., J. Zhou, J. Song , Q. Wang, H. Liu, and X. Tang. 2017. Habitat suitability index model of the sea cucumber Apostichopus japonicus (Selenka): a case study of Shandong Peninsula, China. Marine Pollution Bulletin 122:65 - 76.