<<

The Rocky Reality of Roadways and Timber ( horridus):

An Intersection of Spatial, Thermal, and Road Ecology

A thesis presented to

the faculty of

the College of Arts and Sciences of Ohio University

In partial fulfillment

of the requirements for the degree

Master of Science

Garrett P. Sisson

August 2017

© 2017 Garrett P. Sisson. All Rights Reserved.

2

This thesis titled

The Rocky Reality of Roadways and Timber Rattlesnakes (Crotalus horridus):

An Intersection of Spatial, Thermal, and Road Ecology

by

GARRETT P. SISSON

has been approved for

the Department of Biological Sciences

and the College of Arts and Sciences by

Willem M. Roosenburg

Professor of Biology

Shawn R. Kuchta

Associate Professor of Biology

Robert Frank

Dean, College of Arts and Sciences 3

ABSTRACT

SISSON, GARRETT P., M.S., August 2017, Biological Sciences

The Rocky Reality of Roadways and Timber Rattlesnakes (Crotalus horridus):

An Intersection of Spatial, Thermal, and Road Ecology

Directors of Thesis: Willem M. Roosenburg and Shawn R. Kuchta

A major challenge in conservation biology is balancing human transportation needs with biodiversity. I studied a remnant population of Ohio-endangered Timber

Rattlesnakes (Crotalus horridus) in a forested landscape recently fragmented by a four- lane highway that featured crossing structures and exclusion fencing. I evaluated the success of mitigation structures for while also quantifying the impacts of the road on spatial and thermal ecology using a combination of radio telemetry, mark- recapture, mortality surveys, camera traps, operative temperature modeling, and resource selection functions. Fencing was not successful in excluding reptiles from the right-of- way due to structural and design failings, and reptiles did not use the crossing structures to maintain connectivity across the road. Gravid rattlesnakes used habitats within the road corridor throughout gestation, while conspecifics avoided open canopy areas beyond the edge. The road corridor provided warmer temperatures for more hours of the day compared with the forest, but also exceeded voluntary maximum temperatures at the surface. My models indicated that most of the available gestation habitat was predicted to be concentrated within the road corridor. The placement of exclusion fencing should be adjusted to allow access to favorable habitats within the road corridor while also preventing road mortality. 4

DEDICATION

To Teal Richards-Dimitrie, who’s mentorship and contagious enthusiasm instilled in me the confidence I needed to pursue these curiosities.

And for ingraining “people, , data, and always in that order” which has helped to keep me grounded.

5

ACKNOWLEDGEMENTS

I thank my advisors Willem Roosenburg and Shawn Kuchta for their patience, support, knowledge, and guidance—steering me when needed, while allowing me the freedom to pursue outside endeavors. Their mentorship pushed me in unexpected directions that made me a more rounded and stronger biologist, and those outside endeavors provided me with some delusion of life balance. I thank my committee members, Viorel Popescu and Joe Johnson: Joe, for engaging in quality discussion at seminar (and the bar), and Viorel, my de facto third advisor, for providing generous analytical and professional advice. I thank my cohort and lab-mates for their support and friendship, especially: Maggie Hantak, Kaili Boarman, Tom Radomski, Melissa Liotta, and Alayna Tokash for putting up with my antics, and also Steve Krichbaum, Paul

Converse, Anthony Gilbert, Vinny Farallo, and Eric Gorscak for deepening my understanding of many topics. I thank my housemates Don Cerio, Almir de Paula, and

Buba. Their sleep and sanity have been repeatedly victimized by my musical noodling and obnoxiously loud guests. “Afugundo as magoas.” I thank my professors, Jim Dyer,

Don Miles, Molly Morris, Gaurav Sinha, and Kelly Williams, for their perspectives both inside and outside of the classroom. I thank the people who helped me (suffered for me) in the field: Christine Hanson, Adam Kabrick, Phil Miller, Seth Cones, Kellie Johnson,

Isabel Fisk Baruque, Mike McTernan, Tyler Stewart, Aspen Wilson, among others. I thank all that supported this work as a collaborator: Charlene Hopkins, Steve Porter, Gary

Conley, Matt Trainer, Rob Wiley, Eileen Wyza, and others at the Voinovich School of

Leadership and Public Affairs. I thank the Ohio Department of Transportation for 6 funding this research, and personally thank Matt Perlik, Mike Austin, Steve Williams, and Kelly Nye for their enthusiastic support. I thank the Wayne National Forest, and especially Lynda Andrews and her field technicians. I also thank the Ohio Department of

Natural Resources and Division of Wildlife: Ryan Harris, Kate Parsons, Mike Reynolds,

Allen Patton, Chris Dodge, Jared Abele, and all other wildlife officers and staff who supported this work through permits, enthusiasm, or by helping me extract rattlesnakes from private land. I thank my colleagues: Chris Howey, Bill Peterman, and Doug Wynn, who all provided critical advice or resources. I thank the Department of Biological

Sciences staff: Cindy Meyer, Wendy Kaaz, and Karen Keesey. I thank the Ohio

University IACUC for approving our protocol: 14-L-018. Finally, I thank Ryan

Friebertshauser, Mike Driscoll, Eli Chambers, Mike Varga, Emily Harger, and Merri

Collins, who brought much needed distraction, relief, laughter, debate, adventure, love, friendship, and growth.

7

TABLE OF CONTENTS

Page

Abstract ...... 3 Dedication ...... 4 Acknowledgements ...... 5 List of Tables ...... 10 List of Figures ...... 11 General Introduction ...... 12 Roads & Wildlife ...... 12 Impacts Beyond the Pavement ...... 13 Ectothermic Organisms in Focus ...... 15 Consequences for Mitigation ...... 16 The Nelsonville Bypass ...... 17

Chapter 1: Effectiveness of Rattlesnake Mitigation Structures at the Nelsonville Bypass ...... 19 Introduction ...... 19 Materials and Methods ...... 23 Study ...... 23 Study Site ...... 25 Mitigation Structures: Fence ...... 28 Mitigation Structures: Small Wildlife Ecopassages ...... 28 Evaluating the Effectiveness of the Rattlesnake Exclusion Fence ...... 31 Radio Telemetry ...... 31 Live Trapping and Measuring ...... 32 Marking ...... 33 Road Mortality Surveys ...... 33 Evaluating Space and Habitat Use ...... 34 Evaluating Road Avoidance ...... 35 Evaluating the Effectiveness of the Small Wildlife Ecopassages ...... 37 8

Results ...... 39 Radio Telemetry ...... 39 Fence Crossings by Capture-Mark-Recapture ...... 46 Road Mortality Data ...... 48 Correlated Random Walk ...... 54 Ecopassage Photo Data ...... 54 Discussion ...... 59 Mitigation Failure ...... 59 Rattlesnake Mitigation Elsewhere ...... 62 Road Mortality ...... 64 Road Avoidance ...... 68 Ecopassages ...... 69 Future of the Nelsonville Rattlesnake Population ...... 72

Chapter 2: Thermal Resources in a Road Fragmented Landscape: Biophysical Consequences of the Nelsonville Bypass ...... 76 Introduction ...... 76 Materials and Methods ...... 80 Study Species ...... 80 Study Site ...... 82 Thermal Resource Availability ...... 83 Radio Telemetry & Body Temperature ...... 84 Body Temperature & Thermal Habitats ...... 85 Macrohabitat Selection ...... 87 Availability of Gestation Habitat ...... 88 Results ...... 93 Body Temperature ...... 93 Thermal Resources ...... 93 Macrohabitat Selection ...... 98 Availability of Gestation Habitats ...... 99 Discussion ...... 101 9

Consequences for Management ...... 105 Conclusions ...... 108

Literature Cited ...... 110 Appendix 1: Operative Temperature Model Design ...... 129 Appendix 2: Thermal Modeling and RSF Surfaces ...... 136

10

LIST OF TABLES

Page

Table 1.1: List of captures at the Nelsonville Bypass...... 42 Table 1.2: Home ranges for radio-telemetered rattlesnakes...... 43 Table 1.3: Reptile captures at the Nelsonville Bypass...... 47 Table 1.4: Fence crossings as measured by capture-mark-recapture...... 47 Table 1.5: Summary of reptiles observed crossing or dead on the road...... 49 Table 1.6: Road survey observations of reptiles found crossing or dead on the road...... 52 Table 1.7: Results and summary data for the road avoidance simulations...... 54 Table 1.8: Photo observations in Small Wildlife Ecopassage 1 (with snake fencing). .... 57 Table 1.9: Photo observations in Small Wildlife Ecopassage 2 (without snake fencing). 58 Table 2.1: Predictor variables used in thermal landscape and RSF models...... 90 Table 2.2: Classification scheme of temperature ranges...... 95 Table 2.3: Average daily thermal conditions across macrohabitat type...... 98 Table 2.4: Model selection results of gravid female RSF...... 100

Table 2.5: Top RSF model (TeAVG + Edge) parameters estimates...... 100 Table 2.6: Predicted gestation habitat availability...... 101

11

LIST OF FIGURES

Page

Figure 1.1: Wildlife mitigation structures along the Nelsonville Bypass...... 27 Figure 1.2: The Nelsonville Bypass ROW and reptile mitigation structures...... 30 Figure 1.3: Example of correlated random walk analysis...... 37 Figure 1.4: 100% Minimum convex polygon home ranges...... 44 Figure 1.5: ROW habitats used by rattlesnakes...... 45 Figure 1.6: Reptile road crossing locations on the western Nelsonville Bypass...... 50 Figure 1.7: Reptile road crossing locations on the eastern Nelsonville Bypass...... 51 Figure 1.8: Ecopassage usage by predators and prey of rattlesnakes...... 56 Figure 1.9: Damage along the snake fence...... 60 Figure 2.1: Mean hourly Tb and Te profiles with 95% CI...... 94 Figure 2.2: Spatiotemporal availability of thermal resources across habitats...... 96 Figure 2.3: Availability of gestation Tb and voluntary temperature ranges...... 97 Figure 2.4: Manly selection ratios for rattlesnakes summer habitat use...... 99 Figure A1.1: Operative Temperature Model Design...... 131 Figure A1.2: August Te and Tb temperature profiles...... 134 Figure A2.1: Reference imagery for the NVBP study site...... 136 Figure A2.2: Percent canopy cover layers...... 137 Figure A2.3: Edgeness (canopy heterogeneity)...... 137 Figure A2.4: Insolation...... 138 Figure A2.5: TeAVG surface...... 139 Figure A2.6: TeMAX surface...... 140 Figure A2.7: Projected TeAVG + Edge RSF model...... 141

12

GENERAL INTRODUCTION

Roads & Wildlife

A major challenge in conservation biology is designing landscapes that balance human transportation needs with biodiversity conservation (Forman et al. 2003). It has been estimated that 20% of the land area is ecologically impacted by roads

(Forman 2000). However, roads are geographically widespread, and the global area of road-impacted zones will expand as human populations grow and the demand for transportation infrastructure follows. Globally, at least 25 million kilometers of new roads are anticipated by 2050, increasing the global extent of roads by 60% compared with 2010 (Laurance et al. 2014). Roads beget development, and development begets more roads (Forman & Alexander 1998, Beckmann et al. 2010), catalyzing feedback loops of anthropogenic landscape change, the broader conservation challenge of which road corridors are just one form (Fischer & Lindenmayer 2007).

Road development forecasted to occur in the 21st century is concerning because of the various ways roads disrupt ecological systems. Roads impact wildlife populations through four primary mechanisms: 1) habitat loss and modification, 2) population subdivision, 3) resource isolation, and 4) road mortality (Jaeger et al. 2005). These impacts can affect the spatial structure (McLellan & Shackleton 1988, Berger 2007,

Forman & Alexander 1998, Trombulak & Frissel 2000), demographics (Steen & Gibbs

2004, Aresco 2005a, Steen et al. 2006), densities (Fahrig et al. 1995), and genetics of populations (Holderegger & Di Giulio 2010), sometimes exerting enough pressure to trigger population crashes (Jones 2000) or elicit evolutionary responses over short 13 timescales (Brady & Richardson 2017). Other common road impacts include introducing and spreading invasive species (Hansen & Clevenger 2005, von der Lippe & Kowarik

2007, Hulme 2009, Mortensen et al. 2009) and various forms of pollution (nutrients, toxicants, sediments, noise, light: Karraker et al. 2008, McClure et al. 2013, Shannon et al. 2016, Longcore & Rich 2004). Roads, among other forms of habitat fragmentation, will likely pose additional challenges to biodiversity in the face of climate change, as wildlife will be pressed to track suitable climates or face extinction (Brodie et al. 2012).

Awareness of road effects on wildlife populations has inspired the development of mitigation structures to abate some of these impacts (Glista et al. 2009). Exclusion fencing and ecopassages (i.e. barrier fencing and wildlife crossing structures) are the most common forms of road mitigation, and are implemented to exclude wildlife from roads and facilitate dispersal across roadways (Beckman et al. 2010). European countries have led innovation in road mitigation efforts (Forman & Alexander 1998), but recent decades have seen an increasing number of ecopassages in and worldwide

(Clevenger & Ford 2010). Despite the growth of road ecology into a maturing field

(Forman et al. 2003, Beckman et al. 2010, van der Ree et al. 2015, Andrews et al. 2015), road mitigation often escapes quantitative evaluation, and there remain many unanswered questions regarding the factors that promote species interactions with roads and the effectiveness of mitigation structures (Rytwinski et al. 2015, van der Ree et al. 2015).

Impacts Beyond the Pavement

When roads are built, they change the habitat structure and alter the biophysical environment (Trombulak & Frissel 2000, Forman et al. 2003, Jackson et al. 2015). By 14 fragmenting landscapes, roads increase the edge to core habitat ratio, reduce habitat patch size, and generate a wide-range of structural, biotic, and abiotic changes, collectively referred to as edge-effects (Trombulak & Frissel 2000, Harper et al. 2005). Edge effects often include increased temperature, reduced humidity, increased wind, and changes in community composition, and penetrate remnant habitats to varying distances and magnitudes depending upon the habitat contrast and edge structure (Harper et al 2005).

The abiotic and ecological changes are the outcome of the ecotone created at the interface of open and closed canopy systems. While edges exist at the interface of natural habitat types, they have become far more pervasive as humans have carved the landscape with linear and reticulated transportation infrastructure (Trombulak & Frissel 2000).

Structural and biophysical impacts larger than edge-effects occur along roads where land-cover itself changes. In road, gas line, and electrical transmission corridors, these areas are often designated as the right-of-way (“ROW”), where vegetation is managed and other structural modifications are made. Transportation agencies often maintain ROWs by mowing, which suppresses woody plant growth, and promotes ruderal plants that are often exotic and invasive (Forman & Alexander 1998). The mowing regime and resulting habitat structure then influences the faunal communities inhabiting roadsides (Forman & Alexander 1998). In mountainous landscapes, road construction sometimes involves extensive manipulation of the local topography through earth removal and regrading, resulting in large exposed rock outcropping known as roadcuts. Because topographic features and vegetation removal also produce thermal heterogeneity on landscapes (Forman et al. 2003, Sears et al. 2011), roadcuts too alter the 15 biophysical conditions of roadside environments. Crevices in roadcut rock faces provide structural and thermal microhabtiats that can be used as retreat sites, roosts, and nesting cavities by a diversity of species. Other common modifications to roadside environments include the generation of roadside pools (e.g. drainage ditches and storm- retention ponds), and stone piles installed as drainage or erosion control structures. These resources and many others (e.g. carrion, other forage) often draw animals to roads and the adjacent purlieu to exploit these resources (Langen et al. 2015). Unfortunately, roadside habitats often function as ecological traps, because attracting animals to the roadway can result in increased mortality, or expose animals to pollution (Langen et al. 2015).

Ectothermic Organisms in Focus

The habitat heterogeneity generated by roads and other ROWs may be sought by ectotherms, especially reptiles, seeking thermoregulatory opportunities. Thermal resources can be provided by the road itself (Sullivan 1981), open habitats created during road construction (Sartorius et al. 1999), and forest edges (Blouin-Demers &

Weatherhead 2002). Many ectotherms rely on behavioral thermoregulation (Cowles &

Bogert 1944), adjusting habitat selection and activity times to maintain preferred body temperatures that optimize physiological performance (Huey 1982, Huey & Kingsolver

1989, Huey 1991, Angilleta 2009). Excluding endogenous factors (e.g. dispersal, physiological mechanisms), an ectotherm’s ability to thermoregulate is fundamentally limited by climate and habitat heterogeneity (Kearney et al. 2009, Sears et al. 2011), the latter of which can be a limited resource in both heavily shaded and predominantly open landscapes. In the case of heavily shaded environments, some reptiles have evinced 16 preference for disturbed habitats, including road ROWs (Sartorius et al. 1999,

Klingenböck et al. 2000). Reptiles have also been observed in roadcuts (Myers 1957,

Hertz 1979) and other rocky roadside features (Kovar et al 2014), but usually without formal evaluation of habitat preference. While it is generally supported that reptiles and other ectotherms can use edge habitat (Blouin-Demers & Weatherhead 2002) and rocky cover (Huey et al. 1989) to access these thermal resources, few researchers have (1) explicitly quantified the thermal opportunities provided by roadside habitats relative to unmodified habitats (Sartorius et al. 1999, Klingenböck 2000), (2) evaluated those resources in context of key life history stages, or (3) related those resources to the effective placement of road-mitigation structures.

Consequences for Mitigation

Roads are ubiquitous and expanding, and road mitigation strategies should seek to protect species drawn to roadside environments by either preventing the creation of ecological traps, or by converting ecological traps into compensatory habitat. While exclusion fencing can be a viable means to reduce wildlife-vehicle collisions, its functional placement is rarely evaluated. If road corridor habitats can provide important resources to focal species (e.g. locally endangered reptiles), it may be appropriate to position exclusion fencing in locations that reduce road mortality while also permitting access to those resources within the ROW. Restricting access to these habitats may compromise the effectiveness of mitigation structures by creating motivation to trespass, or fail to capitalize on newly generated habitats that may offset the impacts of habitat loss incurred during road creation. 17

The Nelsonville Bypass

The Nelsonville Bypass (NVBP) is a segment of U.S. Route 33 that was built through the Wayne National Forest, Ohio, and features wildlife mitigation structures to protect local fauna and motorists. The highway was constructed in an area harboring a remnant population of Ohio endangered Timber Rattlesnakes (Crotalus horridus) that was targeted for species-specific mitigation, including a snake-exclusion fence and a small wildlife ecopassage. These mitigation structures were warranted by the ’ state-endangered status, and evidence showing that snakes are commonly killed on roads

(Andrews et al. 2008), with Timber Rattlesnakes being particularly sensitive to road impacts due to a combination of life history traits, behaviors, and cultural prejudice. The highway threatened to increase mortality, and subdivide what appeared to be an already small population and its habitats, potentially isolating individuals from critical resources and exacerbating the risk of extirpation. While Timber Rattlesnakes were the focal species for the described mitigation, these structures could benefit a diversity of wildlife.

The Wayne National Forest (WNF) surrounding the NVBP is home to at least 16 reptile and 14 amphibian species that could be impacted by the highway, as well as terrestrial of concern including the bobcat (Lynx rufus) and gray fox (Urocyon cinereoargenteus).

The construction of the NVBP also resulted in the creation of a large, open canopy right-of-way containing roadcuts and stone piles that provided important thermal resources previously rare or unavailable to ectotherms in the surrounding landscape.

Problematically, these features are concentrated within the road corridor on the road-side 18 of exclusion fencing, and thus the configuration of fencing and habitats may have created an impetus for trespassing and an ecological trap, impacting rattlesnakes and other wildlife seeking basking habitat or other resources within the ROW. However, if rattlesnakes using roadside areas can avoid the road and additive mortality, they may benefit from access to habitats created in the ROW.

I monitored the Timber Rattlesnakes population in the NVBP ROW and surrounding WNF post-construction using VHF radio telemetry to evaluate the effectiveness of mitigation structures, identify habitat selection, and test for road avoidance. I also quantified thermal environments using operative temperature modelling to characterize the impact of the highway on thermal resources, and to determine if suitable gestation habitats were limited to or concentrated within the ROW area. Chapter

One focuses on the effectiveness of mitigation structures, while Chapter Two explores thermal resource availability and habitat selection. The findings of this study were used to provide resource managers (Ohio Department of Transportation) with recommendations for mitigation structures and ROW management that would best conserve the Timber

Rattlesnake population. However, the findings may have broader implications for imperiled reptiles in thermally challenged landscapes and important considerations for road mitigation in general.

19

CHAPTER 1: EFFECTIVENESS OF RATTLESNAKE MITIGATION STRUCTURES

AT THE NELSONVILLE BYPASS

Introduction

Anthropogenic habitat modification has rapidly altered the composition and spatial configuration of habitats and landscapes on earth, disrupting ecological processes through habitat loss, conversion, and fragmentation, and posing a global threat to biodiversity (Fischer & Lindenmayer 2007). Roads are a principal form of landscape change that often precede and catalyze further modification (e.g. urbanization), creating feedback loops of development (Beckmann et al. 2010). Roads themselves impact wildlife populations through a wide range of ecological effects (Forman & Alexander

1998, Trombulak & Frissell 2000, Fahrig & Rytwinski 2009, Holderegger & DiGuilio

2010, Andrews et al. 2015), primarily habitat loss and degradation, increased mortality resulting from wildlife-vehicle collisions, and reduced connectivity among habitats and populations (Jaeger et al. 2005).

Exclusion fencing and ecopassages (i.e. wildlife crossing structures) are becoming more common mitigation tools installed along both new roads and existing transportation infrastructure (Glista et al. 2009, Beckmann et al. 2010). Studies have demonstrated that the use of exclusion fencing and ecopassages can reduce road mortality and maintain connectivity among wildlife populations (Yanes et al. 1995, Clevenger & Waltho 2000,

Mata 2008, Clevenger et al. 2001, Sawaya et al. 2013, Sawaya et al. 2014), including reptiles and amphibians (Dodd et al. 2004, Aresco 2005b, Colley et al. 2017). Globally, a large number of reptiles and amphibians, collectively known as herpetofauna or herptiles, 20 are threatened with extinction (Stuart et al. 2004, McCallum 2007, Böhm et al. 2013), and these taxa are also the most negatively impacted by roads among the

(Rytwinski & Fahrig 2012). Despite this vulnerability, research evaluating road effects on herpetofauna has lagged behind studies on other vertebrates (Rhytinski & Fahrig 2012).

Perhaps unsurprisingly, studies that have evaluated road mitigation structures have often found that success is contingent on design details and maintenance of the structures (Feldhamer et al. 1986, Dodd et al. 2004, Jochimsen et al. 2004, Baxter-Gilbert et al. 2015). As these measures become more commonly integrated in transportation infrastructure, it becomes increasingly important to have best-management-practices

(BMPs) available to facilitate effective design. The installation and expense of ineffective measures wastes conservation resources and jeopardizes public support for future wildlife mitigation. As such, rigorous evaluation of wildlife mitigation treatments remains important to the development of BMPs for road mitigation, and allows for adaptive management if and when mitigation fails (Rytwinski et al. 2015, van der Ree et al. 2015).

Wildlife exclusion fencing is implemented to prevent animals from accessing the road to reduce or eliminate wildlife-vehicle collisions and road mortality. When fencing is paired with ecopassages, it can also be used to direct animals to those crossing structures. Thus, the effectiveness of fencing can be evaluated based on multiple biologically relevant criteria, including: the frequency or proportion of animals trespassing the fence, reduction in road mortality, and observations of animals being directed to ecopassages. Space and habitat use of target species are also important considerations for mitigation structure design and placement, specifically, to answer the 21 questions of “where” and “to what extent.” If home ranges exceed the linear extent of mitigation, then it would suggest that the extent of fencing is inadequate. In addition, if home ranges show considerable overlap with the road corridor, then fence placement may be excluding individuals from preferred habitats. In both scenarios, the effectiveness of the mitigation structures would likely be compromised, and the design may need to be modified to ensure success. From the perspective of design and maintenance, other important criteria include the physical integrity (material, durability) and upkeep that ensures functionality of fencing. However, some circumstances may negate the need for exclusion fencing. Some wildlife actively avoid crossing roads (Brehme et al. 2013), and exclusion fencing may exacerbate barrier effects when the impacts of isolation and subdivision outweigh the impacts of road mortality, such as the when traffic intensity is low and ecopassages are not installed or used (Jaeger & Fahrig 2004). Thus, the necessity and effectiveness of road mitigation can be specific to species and roadway.

When effective, crossing structures facilitate population and habitat connectivity across a roadway. Thus, ecopassage effectiveness can be evaluated based on the evidence of complete crossings from one entrance to the entrance on the opposite side of the road.

Another important criterion for ecopassage success is the integrity of the design, which may be evaluated by the suitability for target species, willingness of target species to use these features, effective placement in the landscape, and integration with exclusion fencing. Additionally, if predators use ecopassages to hunt prey, it creates the potential for prey trap formation (Little et al. 2002). Thus, evidence of frequent use by predators 22 and events within or near ecopassages should be considered when evaluating suitability for target species.

The Nelsonville Bypass (NVBP) is a segment of U.S. Route 33 that was built through the Wayne National Forest (WNF), Ohio, in an area that harbors a remnant population of Ohio endangered Timber Rattlesnakes (Crotalus horridus). Exclusion fencing (i.e. the snake fence) was installed to keep rattlesnakes off the NVBP, and small wildlife ecopassages (i.e. wildlife crossing structures) were installed under the roadway to allow snakes to safely cross to the other side. While timber rattlesnakes were the focal species for the described mitigation, these structures could benefit a diversity of reptiles and other wildlife. The WNF surrounding the NVBP is home to at least 16 reptile and 14 amphibian species that could be impacted by the highway, as well as terrestrial mammals of concern including bobcat (Lynx rufus) and gray fox (Urocyon cinereoargenteus).

I conducted a two-year field study evaluating the effectiveness of the rattlesnake mitigation structures. If successful, the rattlesnake road-mitigation structures should fulfill two basic criteria (adapted from Forman et al. 2003): 1) prevent wildlife from accessing the roadway and 2) allow dispersal across the roadway. To assess the effectiveness of exclusion fencing, I evaluated whether rattlesnakes and other reptiles trespassed the snake fence, whether snake fencing reduced reptile road mortality, and whether snake fencing diverted wildlife to ecopassages. I also quantified the size and placement of rattlesnake home ranges with respect to mitigation structures and their overlap with right-of-way habitats to evaluate whether the exclusion fence was sufficient in extent or isolating suitable habitats. In addition, I simulated rattlesnake movement 23 paths using correlated random walk analysis to evaluate road avoidance by rattlesnakes.

To assess the effectiveness of ecopassages, I evaluated whether the small wildlife ecopassages facilitated reptile dispersal across the roadway, while also quantifying general wildlife activity at crossing structures to evaluate evidence of prey-trap formation. Herein, I report on the results of these efforts and make recommendations for the improvement of mitigation infrastructure as relevant to Timber Rattlesnakes. This work is complementary to its sister paper (Chapter Two), which further explores habitat preference, thermal opportunities within the road corridor, and how those factors relate to management and mitigation.

Materials and Methods

Study Species

Timber Rattlesnakes (Crotalus horridus) are a large-bodied

(family: ) with a life history characterized by delayed maturity, infrequent reproduction, and high adult survival (Brown 1991, 1993, 2007, 2016, Martin 1993,

2001, Olson 2015). Crotalus horridus are sit-and-wait ambush predators that feed on a wide variety of endothermic prey, consisting of predominantly small mammals and sometimes (Reinert et al. 1984, 2011, Brown & Greenberg 1992, Ernst & Ernst

2003). Natural predators of C. horridus include birds of prey and mammalian mesocarnivores (Ernst & Ernst 2003), and other species of snake (e.g. Coluber constrictor) sometimes prey on juveniles. Their spatial ecology is characterized by intraspecific variation in home range size (< 5 ha to > 200 ha), movements, and habitat use, strongly correlated with sex and reproductive status (Reinert 1984a, Reinert & 24

Zappalorti 1998, Sealy 2002, Aldridge & Brown 1995, Coupe 2002, Waldron et al. 2006,

Anderson 2010). Generally, males and non-gravid females prefer closed-canopy throughout the activity season, while gravid females prefer open-canopy habitats with rocky substrate or downed woody debris (Reinert 1984a; Reinert & Zappalorti 1998).

Males often travel long distances in search of females throughout the mating season

(Aldridge & Brown 1995, Coupe 2002, Waldron et al. 2006), which occurs in the late summer to early autumn. Home ranges of all individuals are tethered to their winter hibernacula (den), for which they show strong site fidelity across years (Brown, 1993).

These sites are rarely abandoned even after significant disturbances to the forest ecosystem, such as logging (e.g. Reinert et al. 2011; MacGowan et al. 2017).

Crotalus horridus will use disturbed habitats, including roadsides, which has been observed primarily in gravid females (Brown 1993, Reinert & Zappalorti, 1998). Adult males are often the most frequently encountered on roads, especially during the mating season (Aldridge & Brown 1995). Vehicle strikes can be a substantial source of mortality for this species (Aldridge & Brown 1995, Sealy 2002, Adams 2005), and may result in female biased sex ratios (Sealy 2002). Crotalus horridus cross roads extremely slowly and often freeze in response to an approaching vehicle (Andrews & Gibbons, 2005).

Their laggard road crossing behavior combined with the persecution of snakes on roadways make it exceptionally difficult for C. horridus to safely cross unmitigated roads

(Langley et al. 1989; Ashley et al. 2007). It is thus not surprising that roads have also been associated with reduced population densities of C. horridus and reduced gene flow among populations, which ultimately erodes genetic diversity within populations 25

(Rudolph & Burgdorf 1997, Rudolph et al. 1999, Bushar et al. 2015, Clarke et al. 2010).

This combination of life history and behavioral traits render C. horridus vulnerable to road impacts and habitat fragmentation.

Study Site

The NVBP is a 13.6 km segment of U.S. 33 that passes through Athens and

Hocking Counties, Ohio (Figure 1). The NVBP is a high speed (112 km/h), high volume

(17,000 vehicles/day), four-lane divided highway. Construction of the NVBP began in

2007, and the road was opened to traffic in October, 2013. The NVBP bisects the Athens

Unit of Wayne National Forest (WNF), which includes more than 67,000 non-contiguous acres of forestland in five counties of southeastern Ohio.

The study site spanned habitats within the NVBP right-of-way (ROW) and the surrounding forest. The WNF is an eastern deciduous forest with a predominantly Oak-

Hickory (Quercus, Carya) overstory at the NVBP. The forest contains outcroppings of

Pennsylvanian sandstone that are shaded by canopy. Habitats within the ROW are predominantly open canopy, and include roadside fields dominated by grasses (Poaceae) and weedy vegetation (Asteraceae), roadcuts, stone piles (drainage control structures), early successional stands of sumac (Rhus sp.) and black locust (Robinia pseudoacacia), and small patches of remnant trees and forest edge. I define ROW habitat as the area within the bounds of the road corridor, excluding the road itself (Figure 1). This includes all disturbed habitats generated as a function of road construction or within ROW wildlife fencing (Figure 1). Areas outside the road corridor were predominantly forested, and these areas were defined as forest habitats (Figure 1). 26

While the surrounding landscape is largely forested, the NVBP travels less than 1 km outside the city of Nelsonville and crosses 6 two-lane roads, all intersecting portions of the forest. The patch of forest north of the bypass is mostly contiguous, but the south side of the bypass bears a large gap in forest cover at the site of a large landfill (The

Athens-Hocking Reclamation Center). Furthermore, the WNF near the NVBP contains approximately 75 miles of off-road vehicle (ORV/ATV) trails, a gas line right-of-way, and infrastructure for oil, gas, and timber extraction including well pads and access roads.

In the decade prior to the completion of the bypass, C. horridus had been observed in the Dorr Run area outside of Nelsonville, with confirmed sightings on the

WNF ORV trails, crossing OH-278, and around the Athens-Hocking Reclamation Center.

Pre-construction rattlesnake surveys were performed in WNF pursuant to an environmental impact assessment, but surveyors failed to detect rattlesnakes. However, during construction of the bypass, two rattlesnakes were encountered on separate occasions in the Dorr Run area, where the rattlesnake mitigation structures were ultimately installed.

27

Figure 1.1: Wildlife mitigation structures along the Nelsonville Bypass. 28

Mitigation Structures: Snake Fence

A 2.4 m tall ROW exclusion fence (“wildlife fence”) was constructed along the length of the bypass to prevent deer and other large wildlife from accessing the highway.

However, the mesh of the wildlife fence is permeable to smaller wildlife including rattlesnakes, and thus an additional reptile exclusion fence (“snake fence”) was installed using a high-density material (6.35 mm mesh galvanized hardware cloth). The snake fence stands 0.9 m to 1.2 m tall, with the base of the fence buried < 0.2 m into the ground. This fence extends along both sides of the highway for about 1.6 km west of the

Dorr Run interchange (Figure 1, 2), spanning the areas where rattlesnakes were observed during construction. The eastern half of the fence is attached to the base of the wildlife fence and tracks steep terrain across the ROW-forest boundary, while the western half is detached from the wildlife fence and travels closer to the roadway within the ROW. The snake fence was not built at a consistent distance from the roadway and ranges from ~7 m to 90+ m from the pavement.

Mitigation Structures: Small Wildlife Ecopassages

Five wildlife crossing structures (“ecopassages”) were installed along the U.S. 33 segment of the Nelsonville Bypass (Figure 1), and an additional two crossings were installed under the OH-78 access road. Two of these structures along the U.S. 33 segment were 1.2 m diameter corrugated steel culverts designated as herpetofauna crossings

(Figure 2), but can be used by any wildlife small enough to fit through the tunnels, and thus “small wildlife ecopassages” (SWE) may be a more appropriate designation. The ecopassages are 52 m long, and in the median, the circular culverts open into a 1.2 m 29 wide by 10 m long rectangular box culvert that spans the width of the median. The ceiling of the box culvert is an elevated metal grate (tall enough for a person to stand comfortably), and allows natural lighting at the middle of the passage. However, conditions throughout most of the SWEs remain quite dark (Figure 2) and would score low by metrics of openness (Yanes et al. 1995). One of these structures was installed in the Dorr Run area and is paired with the snake fence (SWE1, Figure 1, 2), while the other ecopassage was installed 3.2 km west and beyond the extent of snake fencing (SWE2,

Figure 1).

30

Figure 1.2: The Nelsonville Bypass ROW and reptile mitigation structures. A: The NVBP ROW. B: Snake-fence attached to the wildlife-fence. C: Snake-fence detached from wildlife-fence. D: Small-Wildlife Ecopassages 1 (“SWE1”) with snake- fencing. E: View inside SWE1.

31

Evaluating the Effectiveness of the Rattlesnake Exclusion Fence

I measured fence trespass rates using a combination of radio telemetry

(rattlesnakes) and capture-mark-recapture (CMR). Radio telemetry allowed reliable detection of rattlesnake fence crossings throughout the activity season, while capture- mark-recapture allowed for monitoring of fence crossings by rattlesnakes and other reptiles. Using both methods is useful for comparative purposes, as detectability, and by extension recapture rate, is often low for snakes (Steen 2010). I deployed bidirectional box-funnel traps (Burgdorf et al. 2005) and tin coverboards (Grant et al. 1992) along both sides of the snake fence to capture reptiles. Telemetry, trapping, and marking protocol are described below. I also quantified reptile road mortality on the NVBP in areas with and without snake-fencing to evaluate the effectiveness of the snake fence in reducing road mortality, and compared the number of mortality events between the NVBP section with snake fencing to the rest of the bypass without snake fencing while controlling for distance. Further, I evaluated the effectiveness of the snake fence for directing wildlife to ecopassages by comparing use rates as detected by cameras. I noted damage to the fencing throughout the field season and comprehensive surveys of its structural integrity were performed by collaborators.

Radio Telemetry

I surgically implanted radio transmitters into the body cavity of adult and large sub-adult rattlesnakes following Reinert and Cundall (1982). Snakes were implanted with an Advanced Telemetry Systems ® R1680 transmitter (3.6 grams; < 2% of body mass).

After surgery, snakes were provided with water and a warm enclosure in the lab where 32 they recovered for 2-7 days before being released at the location of capture, with the longer recovery periods being provided when weather conditions were not conducive to healing (e.g. cold, rainy). I tracked rattlesnakes 3 times per week between spring egress

(i.e. emergence; April – May) and autumn ingress (i.e. onset of ; September –

October) using a Communication Specialists Inc. R-1000 Telemetry Receiver and a 3-

Element Yagi Antenna, and relocations were recorded using a Garmin GPSMAP 64 (3 m accuracy). Locating rattlesnakes 3 times weekly was an acceptable tracking frequency because C. horridus can often remain sedentary for multiple days at a time: non-gravid snakes are sit-and-wait predators that often spend extended periods in ambush posture

(Reinert et al. 1984), while gravid females often have small home ranges and spend extended periods at gestation sites (Reinert & Zappalorti 1998, Sealy 2002, Anderson

2010). Rattlesnakes were commonly found at the same location across successive relocations throughout this study. I recorded the habitat characteristics and which side of the fence the snake was on at each relocation to ensure that GPS locations matched field observations.

Live Trapping and Measuring

I deployed 12 bidirectional box-funnel traps (Burgdorf et al. 2006) to capture animals moving along the snake fence. In 2015, six traps were deployed on each side of the highway. Traps were spaced 100 - 300 m apart, and targeted areas that were of suitable habitat and avoided trapping areas where the fence was visibly compromised. Six traps were deployed on both sides of the highway (12 total), with four traps placed on the forest side of the fence, and two traps placed on the ROW side. In all locations, traps 33 were positioned flush against the ground and the fence. I also deployed 32 tin cover board piles (Grant et al. 1992) adjacent to the snake fence; 16 board piles were distributed on each side of the highway, with 8 piles on each side of the fence. Each pile was constructed from 2-3 staggered sheets of tin. Individual tin sheets were approximately 2 x

0.5 m. Because I did not capture rattlesnakes on the south side of the bypass in 2015, I moved all box-traps to the north side of the highway in 2016, with six traps placed on each side of the fence, but cover boards were still monitored on the south side of the bypass.

Marking

Rattlesnakes and all other snake species (excluding leaf-litter snakes, e.g.

Carphophis amoenus, Diadophis punctatus) were marked with a Biomark ® MiniHPT8

PIT tag, and were marked with shell notches on the marginal scutes (Cagle 1939).

All equipment was sterilized prior to marking each : PIT tagging equipment was bathed in a chlorohexidine solution, and notching equipment was bathed in isopropyl alcohol and flamed. All animals were released the following day (except implanted rattlesnakes) at the location of capture.

Road Mortality Surveys

I drove the length of the NVBP in both directions 5-7 days/week throughout the rattlesnake activity season. While on the highway, we drove in the right lane at reduced speeds (88.5 km/h) and scouted for reptiles crossing or dead on the road (DOR). We documented all reptiles we observed on the NVBP, and recorded species, GPS location, vehicle lane, and when possible, sex and age class (adult versus juvenile). To test if the 34 snake fence was reducing road mortality, I compared mortality frequencies between fenced areas and unfenced areas using Fisher’s exact test, generating predicted frequencies proportional to the length of highway in each treatment. Though travel speeds were faster than ideal for road mortality surveys, this was the minimum speed allowed by the Ohio Department of Transportation for safety concerns.

Evaluating Space and Habitat Use

I used a combination of home range estimation techniques to quantify space and habitat use at the Nelsonville Bypass. To quantify the maximum extent of space use (i.e. home range length), I generated 100% minimum convex polygons (MCP) for all rattlesnakes for each activity season. If the maximum length of MCPs exceed the length of snake fencing or occur in areas that do not overlap snake fencing, then the snake fencing is not sufficient in extent. I used kernel density estimators (KDE, Worton 1989) to evaluate the overlap of general and core activity areas with the ROW. I generated 95%

(general home range) and 50% (core home range) KDEs for individual snakes each activity season. To address the criticisms raised by Row and Blouin-Demers (2006) that kernels often overestimate home range size, I used a conservative bandwidth operator

(PLUGIN method; comparisons provided in Millspaugh et al. 2012). The PLUGIN method consistently produced 95% KDEs that were smaller than 95% MCP for individuals with complete activity season data, meeting the standards suggested by Row and Blouin-Demers (2006). I did not produce KDEs for individuals with data deficient activity seasons, for example, when individuals we captured late in the activity season and the low number of relocations would have biased KDE-ROW overlap. 35

KDEs are probabilistic, and may still extrapolate activity areas to locations not used by an animal, and thus fail to detect hard boundaries in the environment. LoCoH

(local convex hull) methods are akin to a hybrid of MCP and KDE, where the distribution of points in space drives the formation of smaller polygon shapes using a nearest- neighbor algorithm (Getz et al. 2007). Thus, LoCoH can be used to generate utilization distributions that better identify hard boundaries. I generated 95% and 50% LoCoH home ranges to compare with KDE, and to allow for comparison with another study that used these same home range metrics (Andrews 2010). I used adaptive-LoCoH (a-LoCoH) because of its improved ability to model home ranges when the distribution of relocations contains both densely and sparsely populated activity areas (Getz et al. 2007). All home ranges were generated in R (R Core Development Team, 2016) using the AdehabitatHR package for MCPs (Calenge 2006), ks package for KDE (Duong 2007), and tlocoh package for LoCoH (Lyons et al. 2013).

Evaluating Road Avoidance

Exclusion fencing should prevent animals from entering the ROW, but if the snake fence fails to exclude rattlesnakes from the ROW and no evidence of rattlesnake road mortality is found, then rattlesnakes may be avoiding the roadway. I modified the methods described in Shepard et al. (2008a) to evaluate whether rattlesnakes were exhibiting road avoidance in the field based on movement data derived from radio telemetry. I used correlated random walk analysis (CRW, Kareiva & Shigesada 1983) to model rattlesnake movement paths, and quantified the frequency that simulated movements entered or crossed the highway. I then made statistical comparisons between 36 observed and expected crossings based on radio-telemetry observation and simulation results, respectively. CRW analysis was performed by simulating 1000 random walks

(i.e. movement paths) for each individual using the step length and turn angle distributions derived from each individual’s observed movement path (example provided in Figure 3). Each iteration was rooted to the individual’s den site, and the number of movements per random walk (steps) was determined by counting the number of movements > 0 m observed throughout the individual’s activity season. After movements paths were simulated, I calculated the number of times each simulated path crossed or entered the highway. Thus, for each individual I generated a distribution of road crossings that could be used to make statistical comparisons at the individual and population levels. At the individual level, I evaluated avoidance by calculating the probability of sampling a value from the individual’s simulated crossing distribution that was less than or equal to the observed number of crossings. At the population level, I evaluated road avoidance using a one-tailed paired t-test. Using individual-years as replicates, I compared the number of observed road crossings in an individual’s activity season to the median value from that individual’s simulated crossing distribution. I used individual-years, as opposed to individuals, to compensate for the small sample size of telemetered individuals. Two individuals that were tracked in consecutive years were used twice, but this is not a pseudo-replication problem because of large variation in movements and habitat use among years for these individuals (gravid vs. post-partum years for one snake, and the other snake was a juvenile with a shifting home range).

Movement path metrics and correlated random walks were generated in GME (Beyer 37

2015). Crossings were calculated in ArcMap (ESRI 2014) and statistical tests were performed in R (R Core Development Team 2016).

Figure 1.3: Example of correlated random walk analysis. Solid line = observed movement path; dashed lines = simulated random walks.

Evaluating the Effectiveness of the Small Wildlife Ecopassages

While radio telemetry provided a means of detecting use of the ecopassages by rattlesnakes, I also quantified use of the small animal ecopassages using game cameras, which allowed for simultaneous evaluation of prey trapping within the ecopassage. Each ecopassage was monitored by three Buckeye Game Cameras; a camera was deployed at both entrances, with an additional camera positioned in the middle of the tunnel. Cameras were triggered by movement (passive IR sensor), and photos were reviewed to quantify 38 crossings or other activity within the ecopassages. Photo events were scored as crossings when the same animal was detected at both entrances of the tunnel. Activity was scored as a potential crossing when an animal was detected at the middle of the tunnel and only one entrance, which made it impossible to tell where the animal had originally entered or ultimately exited the ecopassage; in such cases, animals may have failed to trigger cameras due to their speed or position in the tunnel relative to the sensor. Potential crossings also included events when the duration between observations at both entrances of the tunnel made it unclear as to whether it was the same individual. When animals approached or entered the tunnel but could not be further classified as a crossing or potential crossing, that event was scored as entrance activity.

To quantify the potential for prey trap formation, I tabulated the number of days within the rattlesnake activity season that predators and prey species were observed in the ecopassages. I considered only dates from May through September because activity at the tail ends of the season was usually restricted to areas near the dens (middle April, early

October). Rattlesnakes are underground in dens for the remainder of the year making predator activity irrelevant during those periods. For each field season, I calculated the number of days in which predator and prey species used the rattlesnake ecopassages.

Potential rattlesnake predators included all mammalian predators capable of killing a juvenile or adult rattlesnake (e.g. , fox, mink, , , , etc.).

Potential prey species included all potential mammalian prey items (e.g. mice, voles, , chipmunks, , ).

39

Results

Radio Telemetry

I captured 17 Timber rattlesnakes, including adults of both sexes (2♂, 2♀), one juvenile (1♀), and 12 neonates (5♂, 7♀; Table 1). I telemetered five rattlesnakes, which included four adults and one juvenile female, and generated 419 relocations over the course of two field seasons (163 relocations of four individuals in 2015, 256 relocations of four individuals in 2016). Of the radio-tracked snakes, four occurred within the range of rattlesnake mitigation structures, and one was captured east of the Dorr Run interchange beyond the snake fencing. Three of the adult rattlesnakes were initially captured within the ROW, including two individuals in the area with snake fencing. All five telemetered rattlesnakes crossed between the forest and ROW habitats despite four of the snakes occurring where the snake fence was present. In total, I observed 20 fence crossings by telemetered rattlesnakes (Table 2). Despite the snake fence being a permeable barrier, no telemetered rattlesnakes were killed on the road, nor did I observe road related injuries. I did not observe telemetered rattlesnakes cross or attempt to cross the road directly or by way of the ecopassages, suggesting that rattlesnake movements were bounded by the roadway.

MCP areas ranged from 3.6 to 25.7 hectares and maximum home range lengths ranged from 306 to 937 meters (Figure 5, Table 2). All MCP home ranges included both forest and ROW habitats, and were bounded by the highway (Figure 5). None of the rattlesnakes had individual home range lengths exceeding the length of the snake fence, but one individual dispersed beyond the extent of the snake fence, and another 40 individual’s home range was completely outside the extent of the snake fence (Figure 5).

95% KDEs (range: 1.2 – 22.6 ha, Table 2) predicted activity areas to include the road surface for 5 of 6 home ranges, but we did not observe rattlesnakes enter or cross the road. 95% LoCoH home ranges identified the NVBP as a boundary and ranged in size from 1.8 to 15.9 hectares. Five of six 50% KDE and four of six 50% LoCoH home ranges overlapped with ROW habitats, indicating that the ROW was a component the of core activity area. Kernels produced more biologically realistic core home ranges than

LoCoH. For example, a gravid female showed 90.7% overlap between her core KDE and the ROW, but only 35.6% overlap using LoCoH despite 71.6% of her relocations occurring within the ROW (Table 2). Due to the sensitivity of rattlesnakes to persecution and the accessibility of the study site, I have intentionally omitted figures specifying

KDE and LoCoH home ranges.

While male and non-gravid female rattlesnakes spent most of their time within the forest habitats (2.8 – 33.3% relocations within ROW), gravid females spent nearly all of their time within the ROW habitats until giving birth (85.4 – 95.2% relocations within the

ROW while gravid). Gravid females immediately returned to the forest after neonates dispersed, and remained in the forest through hibernation. Non-gravid rattlesnakes traveled to ROW intermittently to shed, bask after surgery, and ostensibly to forage or search for mates. Most of the activity within the ROW occurred near the forest edge, especially for non-gravid rattlesnakes that rarely traveled into the open canopy habitats more than 10 - 15 meters beyond the forest edge. Gravid females were more willing to travel further into the ROW, but they too were more active near the edge than away from 41 it. These findings suggest that important habitats were created within the ROW, often on the road-side of the snake exclusion fence, and likely motivated fence crossings.

In 2015, I also tracked four (2♂, 2♀) Black Racers (Coluber constrictor) and one

(♀) Copperhead (). Surgical implantation and tracking followed the same procedures described for C. horridus, but these snakes were tracked less frequently. Three of the four racers were captured on the south side of the bypass, and two of the races made long distance movements (1♂, 1♀) along the snake fence until eventually finding breaches, or reaching the end of the fence. These animals then spent extensive time foraging in roadside fields and basked in open habitats when they needed to shed. One individual was relocated in a bush growing along the fence, and the snake could be observed resting in an arboreal position intertwined with the wildlife fence. On another occasion, when chasing a C. constrictor along the snake fence, the individual attempted to evade capture by climbing vegetation to reach the top of the wildlife fence.

The copperhead we tracked was a gravid female that used a large stone pile in the ROW as a gestation site. In 2016, another copperhead was captured at a basking site on a road cut that was used by a gravid rattlesnake in the previous year.

42

Table 1.1: List of Timber Rattlesnake captures at the Nelsonville Bypass. Snake SVL Mass Capture Capture PIT ID Sex Captured Mitigation (cm) (g) Location Method Area

982000365990609 FS 5/27/2015 72 275 Y Forest Cover Board

982000365990652 FG 6/26/2015 103 964 Y ROW Incidental 982000365990672 M 8/5/2015 93 723 Y ROW Box Trap 982000365990621 M 8/23/2015 102 980 N ROW Incidental

982000365990587 MN 9/6/2015 33 27 Y ROW Box Trap

982000365990588 MN 9/19/2015 31 25 Y Forest Box Trap

982000365990601 FG 5/21/2016 105 1135 Y Forest Box Trap

982000365990524 FN 9/5/2016 32 25 Y ROW Incidental

982000365990554 MN 9/5/2016 30 22 Y ROW Incidental

982000365990541 FN 9/5/2016 31 26 Y ROW Fence Line

982000365990518 FN 9/5/2016 30 23 Y ROW Fence Line

982000365990561 MN 9/5/2016 31 23 Y ROW Cover Board

982000365990558 FN 9/6/2016 30 24 Y ROW Box Trap

982000365990542 FN 9/6/2016 31 24 Y ROW Cover Board

982000365990486 FN 9/6/2016 32 25 Y ROW Cover Board

982000365990526 MN 9/8/2016 32 27 Y ROW Box Trap

982000365990564 FN 9/8/2016 31 23 Y ROW Box Trap Sex: M, male; F, female; S, sub-adult; G = Gravid, N = Neonate. Capture location refers to the side of the wildlife exclusion fence where an individual was captured; Forest = outside the ROW, ROW = within the ROW. 43

Table 1.2: Home ranges for radio-telemetered rattlesnakes. Snake MCP 100% 50% 95% 50% 95% ID Year Sex Relocs. Fence Max MCP KDE KDE LoCoH LoCoH Crossings Length

652 2015 FG 55 3.6 0.2 1.2 0.6 1.9 5 306 (72.7) (30.4) (90.7) (46.9) (35.6) (24.6) 652 2016 F 71 17.1 3.5 12.9 1.9 8.5 0 682 (2.8) (1.0) (0.0) (6.9) (0.0) (0.3)

601 2016 FG 63 9.7 0.6 5.3 1.0 1.8 6 700 (65.1) (11.1) (64.9) (47.4) (5.8) (25.8)

609 2015 FS 71 5.7 0.8 3.2 0.7 1.9 4 378 (21.1) (7.9) (12.5) (17.3) (13.7) (16.3)

609 2016 FS 48 8.4 0.5 6.4 2.0 3.2 4 380 (33.3) (4.0) (5.4) (26.3) (11.1) (8.3) 672 2016 M 74 25.7 2.6 22.6 2.0 15.9 0 937 (17.6) (2.3) (12.4) (13.8) (0.0) (2.3) 672I 2015 M 20 10.4 NA NA NA NA 1 749 (20.0) (4.4) 621I 2015 M 17 6.2 NA NA NA NA NA 752 (17.6) (4.6) SVL is reported in cm. Relocs = the number of activity season relocations collected, with the percentage of relocations within the ROW reported in parentheses. Home ranges are reported in hectares with percent ROW-overlap in parentheses. MCP maximum length is reported in meters. I = Individual-year reflects incomplete activity season, and we did not generate KDE or LoCoH home ranges for these individual-years due to insufficient data. For purposes of comparison with home range diameters, the snake fence spans 1.6 km. 44

Figure 1.4: 100% Minimum convex polygon home ranges. All home ranges overlapped with the ROW, but were also bounded by the roadway; F = non-gravid female, Fg = gravid female, Fs = sub-adult female, M = male. 45

Figure 1.5: ROW habitats used by rattlesnakes. Top: Gravid rattlesnake basking on a man-made stone pile; Bottom: other ROW habitats used by rattlesnakes are pictured, including fields, early succession stands of locust, and rocky roadcuts

46

Fence Crossings by Capture-Mark-Recapture

We monitored box traps for 3696 traps nights, made 1824 coverboard observations, and spent approximately 2900 person hours in the field. We captured 223 reptiles (7 species of snakes, 1 species of turtles; Table 3), not counting leaf litter snakes

(42 Diadophis punctatus and 2 Carphophis amoenus) and Eastern Fence (55

Sceloporus undulatus), which were excluded from of our mark recapture-survey. We also captured 87 amphibians (9 species), 232 mammals (12 species), and 26 birds (6 species) as bycatch, and these animals were released upon capture. We observed 12 reptile and 7 amphibian species within the ROW. Sceloporus undulatus were captured commonly on both sides of the fence at the forest edge. Leaf litter snakes also were captured on both sides of the fence, but never far beyond the forest edge.

In total, we marked 174 individual reptiles across 8 species. Of those, we had 18 snake recaptures (15 individuals) of 5 species and 12 box recaptures of 8 individuals. We recaptured 2 rattlesnakes using traps over the course of the project, and detected only 1 fence crossing. Overall, mark recapture revealed 4 snake fence crossings to the forest side, and 3 crossings to the ROW side. Crossings were observed in

Thamnophis sirtalis (n=1), Lampropeltis Triangulum (n = 2), Coluber constrictor (n = 2), and C. horridus ( n = 1) (Table 4). We marked 45 individual Terrapene carolina and 35 were in the rattlesnake mitigation treatment area. Within the treatment area, we had 11 recaptures of 7 individual T. carolina, and documented 1 fence crossing to the ROW. We never documented a reptile cross the road.

47

Table 1.3: Reptile captures at the Nelsonville Bypass. Snakes Species Forest ROW Total Copperhead Agkistrodon contortrix 6 3 9 Northern Black Racer Coluber constrictor 16 13 29 Timber Rattlesnake Crotalus horridus 6 13 19 Eastern Hog-Nosed Snake Heterodon platirhinos 1 0 1 Eastern Milksnake Lampropeltis triangulum 13 11 24 Gray Ratsnake Pantherophis spiloides 8 4 12 Common Gartersnake Thamnophis sirtalis 57 8 65 Snakes Total 107 55 162 Turtles Eastern Box Turtle Terrapene carolina 49 12 61 Reptiles Total 156 67 223 Reptile captures from the forest and ROW sides of the wildlife exclusion fence. Data includes box trap, cover board, fence line, incidental captures, and recaptures.

Table 1.4: Fence crossings as measured by capture-mark-recapture. Cross to Cross to Total Recaptures Total Snakes Species Forest ROW (no. individuals) Marked Copperhead Agkistrodon contortrix 0 0 0 8 Northern Black Racer Coluber constrictor 1 1 3 (3) 25 Timber Rattlesnake Crotalus horridus 1 0 2 (2) 17 Eastern Hog-Nosed Snake Heterodon platirhinos 0 0 0 1 Lampropeltis Eastern Milksnake triangulum 1 1 6 (5) 15 Gray Ratsnake Pantherophis spiloides 0 0 1 12 Common Gartersnake Thamnophis sirtalis 1 0 6 (5) 51 Snakes Total 4 2 18 (15) 129 Turtles Eastern Box Turtle Terrapene carolina 0 1 12 (8) 45 Reptiles Total 4 3 30 (23) 174 Data includes box trap, cover board, fence line, incidental captures. Discrepancies Table 3 and 4 between total animals captured and marked are because Table 3 includes recaptures and some animals unfit for marking (i.e. too small, health issues).

48

Reptile Road Mortality Data

I did not observe rattlesnake road mortality on the bypass through telemetry or road surveys, but detected 26 reptiles of 8 species on the bypass (Table 5), 24 of which were found dead on the road (“DOR"). I captured and assisted two box turtles actively crossing the bypass. While these animals weren’t killed on the road, I included them in analyses because they identified crossing locations and I assumed a low probability of survival without human intervention due to high traffic intensity. Over the course of two field seasons, overall reptile mortality rates on the bypass were low (1.91 DOR/km).

Five crossing locations were within 1.6 km of snake fencing, and 21 crossing locations were within the 12 km of the NVBP without snake fencing (Table 5, 6, Figure 6, 7).

Reptile crossing locations were found in higher density in the snake-fenced area (3.125

DOR/km fenced versus 1.75 DOR/km unfenced areas), but sample sizes were small and these rates were not found to be significantly different (Fisher’s Exact Test, P = 0.703).

Two road mortalities were found within ~100 meters of the end of the snake-fence, and it is possible that these individuals were diverted along the fence before entering the roadway. Of the 26 reptiles found on the road, 14 were T. carolina (54%), and 13 occurred in areas without snake fencing (0.625 DOR/km in mitigated versus 1.083

DOR/km unmitigated areas). Of the 24 DOR reptiles, 75% were found in the right shoulder, suggesting most individuals were killed soon after entering the roadway.

49

Table 1.5: Summary of reptiles observed crossing or dead on the road. Within Snake Outside Snake Common Name Species Total Mitig. Area Mitig. Area Eastern Box Turtle Terrapene carolina 1 (0.625) 13 (1.083) 14 Painted Turtle Chrysemys picta 0 1 (0.083) 1 Snapping Turtle Chelydra serpentina 2 (1.25) 0 2 Turtles 3 (1.875) 14 (1.167) 17 Copperhead Agkistrodon contortrix 0 1 (0.083) 1 Northern Black Racer Coluber constrictor 0 1 (0.083) 1 Eastern Milksnake Lampropeltis triangulum 1 (0.625) 2 (0.167) 3 Gray Ratsnake Pantherophis spiloides 0 2 (0.167) 2 Common Gartersnake Thamnophis sirtalis 1 (0.625) 1 (0.083) 2 Snakes 2 (1.25) 7 (0.583) 9 Reptiles 5 (3.125) 21 (1.75) 26 Data displayed as raw count and distance corrected (reptiles/km) in parentheses. The distance correction was calculated as the raw count of reptiles divided by the length (km) of the treatment area (mitigated or unmitigated). Caution should be used when interpreting these data, as habitat patch size, type, and quality vary across the length of the highway.

50

Figure 1.6: Reptile road crossing locations on the western Nelsonville Bypass. Numbers correspond to key in numbers correspond to key in Table 6.

51

Figure 1.7: Reptile road crossing locations on the eastern Nelsonville Bypass. Numbers correspond to key in Table 6. In absence of snake fence, mortality occurred even in close proximity to a crossing structure.

52

Table 1.6: Road survey observations of reptiles found crossing or dead on the road. Map Highway Dead Date Common Name Species Sex Age Class Vehicle Lane Marker Direction /Alive 1 4/27/2015 Black Racer Coluber constrictor - Adult West Left Shoulder Dead 2 5/11/2015 Eastern Box Turtle Terrapene carolina M Adult East Right Lane Alive 3 5/25/2015 Eastern Box Turtle Terrapene carolina F Adult West Right Shoulder Dead 4 6/4/2015 Gray Ratsnake Pantherophis spiloides - Juvenile East Right Shoulder Dead 5* 6/6/2015 Eastern Box Turtle Terrapene carolina - Adult East Right Shoulder Dead 6 6/11/2015 Eastern Box Turtle Terrapene carolina M Adult West Right Shoulder Dead 7 6/24/2015 Eastern Box Turtle Terrapene carolina - Adult East U-Turn Dead 8* 6/27/2015 Snapping Turtle Chelydra serpentina - Juvenile West Right Shoulder Dead 9* 7/9/2015 Eastern Milksnake Lampropeltis triangulum - Juvenile East Berm Dead 10* 7/22/2015 Snapping Turtle Chelydra serpentina - Adult East Right Shoulder Dead 11 7/26/2015 Gray Ratsnake Pantherophis spiloides - Juvenile East Right Shoulder Dead 12* 8/9/2015 Common Gartersnake Thamnophis sirtalis - Adult East Right Shoulder Dead 13 8/16/2015 Copperhead Agkistrodon contortrix - Adult West Exit Lane Dead 14 9/5/2015 Eastern Box Turtle Terrapene carolina M Adult East Right Shoulder Dead 15 5/13/2016 Eastern Box Turtle Terrapene carolina F Adult East Right Shoulder Dead 16 5/27/2016 Eastern Box Turtle Terrapene carolina M Adult East Right Shoulder Dead 17 5/27/2016 Eastern Box Turtle Terrapene carolina M Adult East Right Shoulder Alive

18 6/4/2016 Eastern Box Turtle Terrapene carolina FG Adult West Exit Ramp Lane Dead 19 6/17/2016 Eastern Milksnake Lampropeltis triangulum - Juvenile West Right Shoulder Dead 20 6/23/2016 Eastern Milksnake Lampropeltis triangulum - Adult West Right Shoulder Dead 21 6/24/2016 Eastern Box Turtle Terrapene carolina - Adult West Right Shoulder Dead 22 7/29/2016 Eastern Box Turtle Terrapene carolina M Adult West Right Shoulder Dead 23 8/1/2016 Eastern Box Turtle Terrapene carolina M Adult West Right Shoulder Dead 24 8/1/2016 Eastern Box Turtle Terrapene carolina - Adult West Right Shoulder Dead 25 9/7/2016 Painted Turtle Chrysemys picta F Adult West Right Shoulder Dead

53

Table. 1.6: Continued. Exit Ramp Right 26 9/9/2016 Common Gartersnake Thamnophis sirtalis - Adult West Dead Shoulder

* = animal was found in the snake fenced area. - = sex unknown. F = female, M = male. FG = gravid female.

54

Correlated Random Walk

I performed CRW analyses for all individuals with sufficient activity season data

(n = 6, Table 7). I did not observe statistical avoidance at the individual level (Table 7) owed to the many potential movement paths that can avoid a single linear barrier, but did detect significant avoidance at the population level (T = 2.44, df = 5, P = 0.029). This suggests that a road crossing or road mortality event should have been observed if rattlesnake were moving randomly with respect to the road.

Table 1.7: Results and summary data for the road avoidance simulations. Predicted Crossings SVL Observed P ≤ ID Year Sex Steps Median Range (cm) Crossings observed

652 2015 FG 103 34 0 1 0 - 11 0.370 652 2016 F 103 54 0 3 0 - 18 0.200

601 2016 FG 105 31 0 2 0 - 14 0.240

609 2015 FS 72 33 0 0 0 - 12 0.576

609 2016 FS 77 35 0 0 0 - 9 0.539 672 2016 M 97 61 0 1 0 - 20 0.319 FG = gravid female; F = non-gravid female; FS = sub-adult female; M = male; SVL = snout-vent length; Steps = number of steps in random walk determined by the number of movements > 0 m within the activity season; P ≤ observed = the probability of sampling a value from the crossing distribution ≤ the number of observed crossings.

Ecopassage Photo Data

From 9 July 2015 through 21 December 2016, I observed 38 species of vertebrates at the small wildlife ecopassages (Table 7, 8), but did not observe reptiles or amphibians complete crossings. Activity and crossings were dominated by mammals, including deer mice ( spp.), eastern chipmunks ( striatus),

(Procyon lotor), Virginia (Didelphis virginiana), and eastern cottontails

55

(Sylvilagus floridanus). I also detected larger mammalian mesopredators including bobcat

(Lynx rufus), coyote (Canis latrans), red fox (Vulpis vulpis), gray fox (Urocyon cinereoargenteus), and American mink (Neovison vison).

I detected 2563 animal events (28 species) in SWE1 (with fencing) and 2647 events (28 species) in SWE2 (without fencing). While this initially suggests that the exclusion fence failed to redirect wildlife to crossings, it should be noted that 57% of the activity at SWE2 was by mice (Peromyscus spp.), compared to only 27% at SWE1.

When excluding Peromyscus, I observed 1853 events at SWE1, and 1229 events at

SWE2. I observed 667 crossings (17 species) through SWE1 and 276 crossings (13 species) through SWE2, suggesting that the snake-fence redirected animals to the ecopassage. This finding was consistent when including potential crossings (crossings + potential crossings; SWE1 = 1134, SWE2 = 813).

Ecopassages were monitored 235 days throughout the two rattlesnake activity seasons; SWE1 and SWE2 were used 80.0% and 54.0% of days by predators, and 91.5% and 72.3% of days by prey species, respectively (Figure 8). This indicated a high usage- rate for both predators and prey of rattlesnakes. Days of use were consistently higher in

SWE1 for both predators and prey. While no reptiles were preyed upon within the ecopassages, there were at least two recorded predation events by one raccoon and one weasel with ostensibly small mammalian prey items. A coyote was involved in an apparent scuffle with an unknown small animal on one occasion, and a mink was also observed carrying a carcass through an ecopassage, but it did not appear to be killed therein. Given these findings, there is the potential for prey trapping in the

56 ecopassages, but little evidence that it occurred. Despite the high volume of prey species using the ecopassages, there was no evidence of rattlesnakes foraging within or at the mouth of ecopassages.

Figure 1.8: Ecopassage usage by predators and prey of rattlesnakes. SWE1 is integrated with the snake fence, while SWE2 is only integrated with the wildlife fence. The ecopassage activity data used to generate this figure was restricted to the rattlesnake activity seasons.

57

Table 1.8: Photo observations in Small Wildlife Ecopassage 1 (with snake fencing). Potential Entrance Total Common Name Species Crossings Crossings Activity Activity REPTILES TOTAL 0 0 7 7 Northern Black Racer Coluber constrictor 0 0 1 1 Eastern Fence Sceloporus undulatus 0 0 1 1 Eastern Box Turtle Terrapene carolina 0 0 3 3 Common Gartersnake Thamnophis sirtalis 0 0 2 2 AMPHIBIANS TOTAL 0 0 2 2 American Toad Anaxyrus americanus 0 0 2 2 MAMMALS TOTAL 659 466 1408 2533 Coyote Canis latrans 1 0 3 4 Virginia Opossum Didelphis virginiana 180 28 79 287 Domestic Felis catus 10 6 6 22 Marmota monix 3 2 0 5 Striped Mephitis mephitis 2 0 12 14 Meadow Vole pennsylvanicus 5 0 71 76 Short Tailed Weasel Mustela erminea 8 4 9 21 American Mink Neovison vison 10 5 14 29 Mole Parascalops breweri 0 0 2 2 Peromyscus spp. 29 16 665 710 Raccoon Procyon lotor 297 114 133 544 Fox Scurius niger 5 7 139 151 Sorex / Blarina sp. 0 0 9 9 Sylvilagus floridanus 42 207 100 349 Southern Bog Synaptomys cooperi 1 0 1 2 Lemming Eastern Chipmunk Tamias striatus 59 68 158 285 Gray Fox Urocyon cinereoargenteus 1 1 1 3 Red Fox Vulpis vulpis 6 8 6 20 BIRDS TOTAL 8 1 12 21 Wood Duck Aix sponsa 8 0 0 8 House Sparrow Passer domesticus 0 0 4 4 Common Grackle Quiscalus quiscula 0 0 1 1 Thryothorus ludovicianus 0 0 6 6 American Robin Turdus migratorius 0 1 1 2 ANIMALS TOTAL 667 467 1429 2563

58

Table 1.9: Photo observations in Small Wildlife Ecopassage 2 (without snake fencing). Potential Entrance Total Common Name Species Crossings Crossings Activity Activity REPTILES TOTAL 0 0 4 4 Northern Black Racer Coluber constrictor 0 0 3 3 Ring-Necked Snake Diadophis punctatus 0 0 1 1 MAMMALS TOTAL 276 537 1788 2601 Coyote Canis latrans 0 1 0 1 Domestic Canis lupus familiaris 0 0 2 2 Beaver Castor canadensis 1 0 0 1 Virginia Opossum Didelphis virginiana 14 12 24 50 Domestic Cat Felis catus 2 0 4 6 Bobcat Lynx rufus 0 0 1 1 Groundhog Marmota monix 9 32 14 55 Striped Skunk Mephitis mephitis 1 4 26 31 Microtus Meadow Vole 0 0 15 15 pennsylvanicus Short Tailed Weasel Mustela erminea 1 3 13 17 American Mink Neovison vison 2 13 10 25 Hairy Tailed Mole Parascalops breweri 0 0 3 3 Deer Mouse Peromyscus sp. 112 166 1240 1518 Raccoon Procyon lotor 91 122 55 268 Fox Squirrel niger 7 21 65 93 Shrew Sorex / Blarina sp. 0 0 25 25 Eastern Cottontail Sylvilagus floridanus 15 141 191 347 Southern Bog Lemming Synaptomys cooperi 0 0 1 1 Eastern Chipmunk Tamias striatus 20 22 96 138 Red Fox Vulpis vulpis 1 0 3 4 BIRDS TOTAL 0 0 42 42 Eastern Wood Peewee Contopus virens 0 0 12 12 Dumetella Gray Catbird 0 0 3 3 carolinensis Chickadee Poecile sp. 0 0 1 1 Hooded Warbler Setophaga citrina 0 0 1 1 Thryothorus Carolina Wren 0 0 24 24 ludovicianus Brown Thrasher Toxostoma rufum 0 0 1 1 ANIMALS TOTAL 276 537 1834 2647

59

Discussion

Mitigation Failure

Rattlesnake mitigation structures at the NVBP failed to exclude reptiles from the

ROW, or maintain reptile population connectivity across the highway. These failures stemmed from factors related to mitigation design, placement, and extent. We discovered rattlesnakes distributed beyond the extent of the snake fencing, indicating that the fencing coverage was not sufficient for exclusion at the population level. In addition, radio telemetry and capture data both revealed that rattlesnakes and other reptiles cross the snake fence. The snake fencing was compromised by damage suffered from washouts on steep slopes, corrosion in areas with acidic soils, overgrowth, and tree falls that knocked down fence (Figure 9), which led to rattlesnakes and other reptiles exploiting gaps in the fence to access the ROW. It is also possible that rattlesnakes climbed over the fence via tree climbing, as evinced by their arboreal behavior (Coupe 2001, Rudolph et al. 2004, G.

Sisson pers. obs.). For these reasons, mitigation fencing should be built on level ground and away from the forest edge when possible. In areas with extremely acidic soils, corrosion resistant materials may be required. At the NVBP, a snake fence may be more effective and structurally resilient if built nearer to the road and away from the exterior boundary of the ROW. Assuming traffic accidents, road maintenance, and vandalism have minimal impact on exclusion fencing, it should be more cost effective to maintain fencing closer to roads where degradation by natural processes is reduced and access is easier.

60

Figure 1.9: Damage along the snake fence. Top: corrosion damage. Middle: tree-fall damage. Bottom: erosion damage.

61

The distribution of habitats on the landscape and the creation of suitable habitats in the ROW contributed to rattlesnake movement and mitigation failure. The road-cut on the north side of the bypass created large south-facing rock escarpments exposed to full sunlight and large riprap stone (Figure 4), creating conditions preferred by gravid females

(Reinert 1984a, Reinert & Zappalorti 1998), with thermal and structural properties conducive to both gestation (Chapter Two) and predator avoidance. Habitat use within the ROW was not restricted to rocky areas, as gravid and non-gravid snakes both used grassy fields and early successional habitats within the ROW (Figure 5). Gravid females used the ROW extensively throughout gestation, while non-gravid snakes used the ROW infrequently, but when they did, it was often associated with behaviors linked to thermoregulation such as shedding or healing. Use of the ROW habitats by reptiles highlights the importance of wildlife-landscape interactions when designing mitigation structures. When suitable habitats are created within the road corridor, animals will often be drawn to them (Andrews et al. 2008, Langen et al. 2015). The resulting impacts on those populations will depend on the increased additive mortality introduced through road kill or other mechanisms, determining whether the ROW becomes an ecological trap or beneficial habitat. Strategically placed mitigation could be used as a tool to capitalize on benefits by allowing access to road corridor habitats, yet simultaneously reducing road mortality. For focal taxa, it should be identified whether activity within the roadside, regardless of vehicle collisions, would be detrimental, and under those circumstances it may be appropriate to construct fencing further away from the road. More research is

62 needed to quantify the tradeoffs experienced by wildlife when using roadside habitats and should be a priority of road ecology.

Rattlesnake Mitigation Elsewhere

Rattlesnake road mitigation structures have been constructed elsewhere throughout North America. Jochimsen et al (2004) reported that a fence-culvert system

(0.8 m height cloth fence, 3 slotted culverts for lighting) was erected in Shawnee state forest to prevent mortality of migratory Timber Rattlesnakes that traversed a forest road as they commuted between limestone bluffs (dens) and wetland areas. Two culvert passages by rattlesnakes were recorded during monitoring efforts in the following two years (Jochimsen et al 2004), but road mortality data was not reported. Colley et al.

(2017) reported success in the Eastern Rattlesnake ( catenatus) road mitigation efforts in . The fence was a hardware cloth design similar to the

NVBP, but replaced with durable geofabric in locations where the fence rusted. Road mortality decreased significantly after fence construction. Four concrete box culverts with open grated ceilings were also installed (1.2 m width, ~0.5 m height). Despite nine being detected at culvert entrances, only one was observed successfully crossing. These crossings were 16% the length of the NVBP ecopassages, and allowed for ample light penetration. Thermal conditions within the ecopassages were evaluated and did not exceed thermal tolerances of the rattlesnakes.

Laidig & Golden (2004) studied a population of Timber Rattlesnakes adjacent to a residential development in New Jersey where a fence-culvert system much like the

NVBP system (6.35 mm stainless steel hardware cloth, 0.9 m height, buried 15.2 cm

63 beneath the surface, 2.7 km length) was installed to exclude rattlesnakes from the development and allow dispersal across a new roadway. Nine rattlesnakes were monitored using radio telemetry. Like the NVBP study, rattlesnakes traveled along the fence line until either finding a breach or reaching the end of the fence, and ultimately the fence did not prevent rattlesnakes from entering the development. The exclusion zone included core activity areas for several of the snakes, including gestation sites contributing to the snakes' motivation to trespass. After the conclusion of the study, one of the tracked snakes was found dead on a road within the development. The New Jersey site also featured five concrete box culverts at 90 m intervals (0.91 m width, 0.41 m height, 15 m long). At least two rattlesnakes traversed one of the culverts, but only after balking during a previous crossing attempt, reemerging on the side of entry, and waiting several days before finally passing through. The culverts used in New Jersey were 29% the length of the NVBP crossings.

People have been trying to exclude Timber Rattlesnakes with similar exclusion fences for more than 30 years. In 1986, two vacation camps in New York erected a snake fence (1 m height by 250 m length, 1.27 cm mesh galvanized hardware cloth) to prevent

Timber Rattlesnakes from traveling through camp grounds (Brown 1993). During the previous decades, rattlesnakes were killed as they passed through the campground during the mating season, which unfortunately coincides with vacation season. Brown (1993) reported that snakes were no longer observed within the campground for the following seven years. Another example occurred in the late 1980’s in the Blue Hills of

Massachusetts. Yankee Heights was a $300 million office-hotel-condominium complex

64 slated for construction in an area that harbored state endangered Timber Rattlesnakes. To mitigate the impacts on the rattlesnake population, an environmental agreement required that a snake-exclusion fence be constructed to envelop the clearing. When it was discovered in 1989 that the fence was constructed shoddily, with abundant gaps and covering only a fraction of the clearing (550 m length, chain-link fence with attached silt fence), the project was suspended and ultimately terminated (Palmer 2004).

Road Mortality

The driving speed for the road surveys compromised our ability to detect small carcasses, but likely did not affect our ability to detect rattlesnakes or other large reptiles that died on the road. Road surveys were helped by the light color of the road surface, which facilitated carcass detection. Turtle carcasses were often found in the right shoulder, and usually persisted on the road for days to weeks (sometimes months) following mortality. A study of a highway in (four-lane, 20,000+ vehicles/day) found that 95% of 343 DOR turtles were killed in the road shoulder and the remaining

5% were killed in the first two travels lanes (Aresco, 2005b). While carcass detection in travel lanes may be lower, as those carcasses are destroyed more rapidly by vehicles, our findings are consistent with the distribution of carcasses in Aresco (2005b), who surveyed roads 2-4 times daily. Carcass detection could have been reduced by animals that left the road after being struck (Adams, 2005; Row et al. 2007). Independent components of fieldwork required that I routinely walk sections of the NVBP berm, meaning that I also surveyed sections of the road on foot. This led to the detection of only two carcasses (L. triangulum and T. sirtalis). While these findings are consistent with

65 other studies that have demonstrated the superiority of pedestrian surveys for detecting small carcasses (e.g. Enge & Wood 2002), they do not suggest that substantial road mortality was missed.

While the number of DOR reptiles on the NVBP is low in the context of other studies (Ashley & Robinson 1996, Smith & Dodd 2003, Aresco et al. 2005b, Andrews et al. 2008), these numbers cannot be immediately dismissed as unimportant. We found that

Eastern Box Turtles experienced the most road mortality among reptiles at the NVBP. Of the 10 box turtles that could be sexed, only three were female, and only one was definitively gravid. These results did not provide clear evidence of the female-biased mortality that is common in turtles (Steen et al. 2006), but my sample was small. Shepard et al. (2008b) also reported weakly male-biased road mortality of box turtles in Illinois. I did not statistically evaluate seasonal patterns of reptile road mortality due to small sample sizes, but the data suggested that box turtles experienced more road mortality in the late spring to early summer, which was also consistent with Shepard et al. (2008b).

While box turtle road mortality rates on the NVBP were low, they should be considered in the context of population size, demographic structure, and life history. Terrapene carolina have a bet-hedging life history, making their populations sensitive to additive adult mortality (Congdon et al, 1994). Determining the proportion of the box turtle population being killed on the bypass annually will be necessary to understand these impacts and warrants further study (research currently in progress by M. Weigand & V.

Popescu).

66

Road mortality aggregations in space or time can be a function of traffic volume

(Seigel 1986, Bernardino & Dalrymple 1992), but often depend on additional ecological variables. Traffic volume varies at hourly and seasonal scales and by location, as do movement rates and dispersal routes as a function of natural history (e.g. nocturnal versus diurnal, migratory versus sedentary, sex or age class biased movement). Consequently, species specific and intraspecific road impacts are often observed (Bonnet et al. 1999,

Seigel & Pilgrim 2002, Steen et al. 2006, Shepard et al 2008b, Hartmann et al. 2011,

Rouse et al. 2011, Jochimsen et al. 2014). In snake populations, road mortality events are often correlated with migrations associated with egress and ingress from hibernacula (e.g.

Seigel & Pilgrim 2002, Gross 2013), and peak movements by males during the mating season (e.g. Aldridge & Brown 1995, Bonnet et al. 1999, Shepard et al. 2008b) or by females during the egg laying period (Bonnet et al. 1999). In some cases, juveniles experience higher mortality (e.g. Ciesiołkiewicz et al. 2006, Kovar et al. 2014) due to their higher relative abundance or dispersal tendencies (Bonnet et al. 1999). Turtle mortality is often female biased (Steen et al. 2006) and associated with the nesting season

(e.g. Chelydra serpentine, Haxton 2000), but may not be the case when males have larger home ranges (e.g. Terrapene carolina, Shepard et al. 2008b) or when movements occur among wetlands (e.g. Bodie & Semlitsch 2000, Beaudry et al. 2006). Many of these seasonal movements are triggered by weather (e.g. Ciesiołkiewicz et al. 2006, Shepard et al. 2008b), especially temperature and precipitation patterns that cue hibernation, emergence, mating, and nesting, and can be disrupted or exacerbated by extreme conditions such as flood or drought (e.g. Seigel & Pilgrim 2002).

67

The distribution of animal crossing locations can also be driven by factors that influence animal movements across landscapes, including the relative locations of habitats, site history, topography, and local grade of the road bank (upslope/downslope; e.g. Clevenger et al. 2003). Most studies reporting high reptile mortality are from sites adjacent to aquatic habitats (e.g. Bernardino & Dalrymple 1992; Ashley & Robinson

1996, Smith & Dodd 2003; Aresco 2005b), and typically occur when individuals make overland movements to access resources that have become spatially isolated (Roe et al.

2006). In comparison, the U.S. 33 portion of the NVBP is bordered by palustrine habitats and fields. More vagile snakes often cross roads more frequently (Bonnet et al. 1999, Roe et al. 2006), thus increasing their vulnerability to edge-effects and habitat fragmentation

(e.g. Dyrmachron couperi, Berringer et al. 2004). With these spatial considerations in mind, comparisons of road mortality between mitigation treatment areas with low sample sizes should be treated with appropriate caution, as landscape patterns could be playing a large role in animal movement. Even within landscapes, any given location has a suite of landform and habitat covariates that influence habitat use and movement. For this reason, reliable control treatments are difficult to obtain outside of BACI (“Before-After-Control-

Impact”) study designs. Unfortunately, BACI designs remain elusive in road ecology, which is often out of the control of researchers due to disparities in the timing of construction versus the timing of research funding, as was the case with this study.

Road mortality prior to the start of the study could have depressed local populations and thus led to the low rates of reptile road mortality observed on the bypass

(Eberhardt et al. 2013). Individuals with home ranges bisected by roads are likely the first

68 victims of road mortality. Even if NVBP reptile road mortality rates are naturally low

(low productivity system), it is still important to ensure that mitigation structures are functional, as several of the species impacted by road mortality are demographically sensitive to additive mortality (e.g. Congdon et al. 1994, Row et al. 2007), and because sites with good habitat and low road mortality could make suitable locations for population recovery (Eberhardt et al. 2013).

Road Avoidance

An potential explanation for low observed road mortality is that animals actively avoided the roadway. Rattlesnakes in our study did not cross the NVBP despite approaching within 5 m of the road edge, and analyses of road avoidance at the population level support that rattlesnakes avoided the roadway. It is generally supported that larger roads of higher traffic intensity generally impose larger road-effects (Forman

& Alexander 1998). Shine et al. (2004) observed Common Gartersnakes (Thamnophis sirtalis) avoid crossing a small gravel road, while Shepard et al. (2008a) demonstrated avoidance of roadways by Eastern Massasaugas Rattlesnakes (S. catenatus) and box turtles. Klingenböck et al. (2000) and Richardson et al. (2006) both observed road avoidance in reptiles (Egernia major and , respectively) that commonly used edge habitats near the roads. Brehme et al. (2013) examined the road avoidance of lizards and small mammals on three different road types including unpaved, secondary paved road, and a 2-lane highway. The authors found that the largest road examined, the 2-lane highway, was the most avoided by all species, and they speculated that traffic volume, road vibrations, and noise could all be deterrents against crossing. It

69 seems probable that road width and substrate are also important. By comparison, the

NVBP is twice the size of the largest road examined in Brehme et al. (2013), and it is likely that rattlesnakes and other animals perceive the highway as a threatening environment that is to be avoided.

Ecopassages

No rattlesnakes were observed using the ecopassages by either cameras or radio telemetry. Because rattlesnakes are ectotherms and their body temperatures are often similar to ambient conditions, it is plausible that rattlesnakes failed to trigger the passive

IR sensors (e.g. Pagnucco et al. 2011). For smaller ectotherms like amphibians, this is almost certainly true, and my amphibian crossing data are probably an underestimate. But for larger reptiles like rattlesnakes, this is unlikely. First, rattlesnakes’ large body size confers thermal inertia, causing their body temperature to lag behind environmental temperatures (see appendix). In addition, cameras were positioned and aimed within the ecopassages where it was constantly shaded, and thus incoming snakes would likely have been a substantively different temperature than the ambient conditions within the culverts, at least during daylight and early evening hours. Second, the cameras detected many of the common reptiles at the crossing entrances, including Eastern Box Turtles,

Northern Black Racers, Common Gartersnakes, Ring-Necked Snakes, an Eastern Fence

Lizard, and other small ectotherms, including a toad and multiple insects. Moreover, the cameras captured thousands of images of vegetation swaying at the entrance on windy days. While failing to register at one camera by a rattlesnake is certainly possible, any undetected animal that completed a crossing would have had to bypass three cameras.

70

Rattlesnakes may not have used the ecopassage for a combination of reasons.

First, because the population is small, it is unlikely that many individuals encountered the ecopassage during our two-year study. Without preconstruction data, it would be difficult or impossible to predict historic dispersal routes that would could have been hotspots for crossings or mortality (Patrick et al. 2010, 2012). Radio telemetry revealed that one rattlesnake passed near (< 3 m) the SWE1 entrance, but most individuals may remain unaware of the ecopassage. Alternatively, the habitat on the northern side of the highway may be meeting their ecological needs and crossing through the tunnel could pose an unnecessary risk. The ecopassages were used extensively by mammalian mesopredators, and throughout the activity season had some form of predator activity more days than not. Given the confined quarters within the ecopassage and the prevalence of predators, crossing these ecopassages would be a risky venture for rattlesnakes. Similar to other studies, there was evidence of predation within the ecopassages, but not with the regularity that would suggest the presence of an active prey trap (Little et al. 2002, Ford

& Clevenger 2010).

Design of the ecopassages at the NVBP could have deterred their use by reptiles.

The small wildlife ecopassages were not constructed to current BMP design specifications recommended for reptiles and amphibians (Clevenger & Huijser 2011,

Gunson et al. 2016), which became available after the design and construction of the

NVBP. First, current BMPs state that herpetofauna passages require exclusion fencing to direct these species to the structures (Clevenger & Huijser 2011, Cunnington et al. 2014,

Gunson et al. 2016). At the NVBP, the southern entrance of SWE1 protrudes from a

71 slope and sits away from the snake fence and some wildlife travelling along the fence can pass over the ecopassage rather than being led into it. The absence of snake fencing altogether at SWE2 fails to provide any mechanism to direct herpetofauna to the passage.

Clevenger and Huijser (2011) also recommend rectangular or square box culverts over circular culverts because vertical walls better facilitate the movement of amphibians and reptiles through a tunnel. They also suggest that steel is not a desirable material because of its high thermal conductivity and coldness during the spring migratory periods.

Most importantly, BMPs recommend that the tunnel width should scale with tunnel length. Herpetofauna tend to show preference for more open, lighted, and shorter crossings (Woltz et al. 2008). Clevenger recommend that tunnels 50-60 meters (i.e. 165-

200 ft) length should be no less than 2.4 meters (i.e. 8 ft) in diameter for circular culverts

(or 2.3 m width x 1.8 m height for rectangular culverts). However, the most recent BMP publications recommend nothing smaller than an underpass for highways of similar width to the NVBP (Gunson et al. 2016). Underpasses and wider tunnels would allow for more lighting and airflow compared to the 1.2 m diameter tunnels along the NVBP. Finally,

Clevenger recommends that the maximum distance between herpetofauna crossing structures should be 46 meters (but 60 m could be used if guiding walls or fencing are funnel-shaped to guide movements). In comparison at the NVBP, the nearest crossing structure to the SW1 is 2.3 km west (gas line ROW underpass, i.e. “butterfly bridge”). It is possible that rattlesnakes are distributed further west along the bypass, and that they and other reptiles utilize the gas line ROW and the bridge underpass. However, use of

72 this crossing may be unlikely for some time because much of the area surrounding the underpass remains unvegetated and thus exposed.

Unfortunately, the recommendations of Clevenger and Huijser (2011) and Gunson et al. (2016) were unavailable at the time of planning and construction of the NVBP. My findings reaffirm that crossing structures of this size are not adequate for herpetofauna passages mitigating large highways. By contrast, these structures were successful for many mammals. Four-lane highways (and larger) may require more open underpass designs with structural refuge to maintain herpetofauna connectivity. These findings should not discourage the use of road mitigation structures, as they have been demonstrated to be successful elsewhere when designed and maintained effectively (e.g.

Dodd et al. 2004, Aresco et al. 2005b).

Future of the Nelsonville Rattlesnake Population

Two field seasons have not provided us with enough data to statistically estimate the rattlesnake population size or make long term population projections. Given our sampling effort of nearly 3000 person hours and over 5000 trap nights (box traps and cover boards), capturing only five adult and sub-adult rattlesnakes (and recapturing two of them) indicates that the subpopulation adjacent to the Nelsonville Bypass is small.

However, our observations of rattlesnakes at seemingly disjunct locations along the bypass suggests that rattlesnakes may be spread diffusely across the Athens Unit of WNF and within proximity to the NVBP. We observed reproduction in both years of the study, which suggests that this population may have the capacity to persist in the long-term if adult mortality remains low. A 12-year study of rattlesnakes in eastern saw

73 the number of gravid females vary greatly between among years (Martin 2002), and our study may reflect years in which reproduction was relatively low. An alternative scenario is that the population could be in an extinction debt (Tilman et al. 1994). We documented the death of one juvenile female that was killed by a predator in the second year of tracking, and we have yet to recapture a marked neonate. Unfortunately, gathering rigorous population data on C. horridus in this region is challenged by their use of satellite dens with small population sizes. Ultimately, the fate of the population will likely depend upon metapopulation dynamics among satellite dens within WNF, maintaining low adult mortality, and habitat management that facilitates successful reproduction.

All rattlesnakes were found on the westbound (north) side of the bypass, but we cannot be sure if a subpopulation on the eastbound (south) side of the bypass was extirpated, or if their apparent absence was due to a lower encounter probability on that side of the highway. Rattlesnakes may have been more difficult to detect on the southern side of the NVBP because favorable habitats, south facing slopes and ridge tops, were found primarily on the south side of the patch, away from the bypass and closer to Old

U.S. 33 (Haydenville Road). Although we visually surveyed much of this forest, our trapping efforts were concentrated along the snake exclusion fence located on the northern side of the forest patch and away from those preferred habitats. Historically, there were at least two rattlesnake observations from this forest patch. A rattlesnake was tracked on the Athens-Hocking Reclamation Center years before the construction of the bypass (Doug Wynn, pers. comm.), and another rattlesnake was observed south of the

74 bypass during construction near the location where the rattlesnake ecopassage was installed (Mike Austin, pers. comm.). While it is possible that populations persist on the south side of the NVBP, it is unlikely that there will be gene flow among hibernacula separated by the NVBP if rattlesnakes are unwilling to use the ecopassages. It is possible that rattlesnakes may use the large underpass ecopassages constructed elsewhere along the bypass, but snakes were not identified in those areas during this study. Assuming there was dispersal across the landscape prior to construction, isolation would disrupt metapopulation processes and gene flow, render subpopulations as functionally independent units, and exacerbate vulnerability of these populations to the demographic and genetic consequences of small population size.

The NVBP rattlesnakes exhibit small to average home ranges sizes relative to other populations (Reinert & Zappalorti 1998, Reinert 1999, Sealy 2002, Adams 2005,

Waldron et al. 2006, Anderson 2010, Andrews 2010). Female rattlesnake home ranges were average, but the male rattlesnake had a home range starkly smaller compared with neighboring and regional populations (NVBP Male: 25.7 ha versus 50-100+ ha commonly reported in literature), and did not show large wandering movements towards in late summer associated with the mating season. Andrews (2010) found that Timber

Rattlesnakes had larger foraging ranges in parts of a landscape undergoing residential and recreational development compared to less disturbed areas, and suggested that snakes may have been traveling greater distances in pursuit of food resources. Following that logic, small home ranges at the NVBP suggest that the bypass did not compromise prey availability. However, there are competing explanations for these small home ranges at

75 the NVBP. Naturally occurring, large canopy gaps are uncommon in forest habitat at the study site, and the ROW may have created basking sites closer to dens than were previously available on the landscape, reducing the need to travel long distance for thermoregulation. The summer of 2015 was unusually cool and wet, and in that year, all four telemetered rattlesnakes moved to the forest edge and ROW habitats when shedding.

The thermal opportunities provided by the ROW are most important for gravid females that maintain higher body temperatures (Gardner-Santana & Beaupre 2009) and are encumbered by reduced locomotor performance throughout pregnancy (Seigel et al.

1987). Gravid females used habitats within the ROW extensively for gestation during both years of the study, and ROW habitats made up the majority of their 50% kernel home ranges. The NVBP appears to have increased the availability of high quality basking habitat. The proximity of these habitats to dens also reduces exposure to predators by providing both shorter commutes to these resources as well as structural refuge. Thus, the ROW may have modified habitat in some ways that benefit the local rattlesnake population in ways similar to management techniques used to improve habitats for reptiles by opening the canopy (e.g. Pike et al. 2011). However, these benefits are contingent on ROW habitats not increasing mortality rates. If rattlesnakes are attracted to the ROW and are subsequently killed by vehicles, humans, or other predators, then these features could function as an ecological trap.

76

CHAPTER 2: THERMAL RESOURCES IN A ROAD FRAGMENTED LANDSCAPE:

BIOPHYSICAL CONSEQUENCES OF THE NELSONVILLE BYPASS

Introduction

When roads are built, they modify habitats, biophysical environments, and the spatial arrangements of resources on landscapes (Trombulak & Frissel 2000, Forman et al. 2003, Jackson et al. 2015). One important resource is the thermal environment, as temperature affects many physiological processes, and can impact habitat selection, community assemblies, and species distributions (Angiletta 2009). Road corridors and other right-of-ways (e.g. gas lines, electrical transmission corridors, railroads, ski runs; here after “ROW”) modify thermal resource availability through three primary structural changes: 1) creation and maintenance of open canopy habitats exposed to increased incident solar radiation and wind, including edge-effects, 2) modification of topographic features through regrading, including roadcuts and microtopography, and 3) creation of microhabitat structures with distinct thermal properties, such as the road surface, rock escarpments, or stone piles with crevice retreats. These factors will alter thermal heterogeneity within a landscape depending on various properties, including elevation, slope, aspect, narrowness and orientation of the corridor, albedo, vegetation, and the structure of microhabitats sites, resulting in changes to the temporal biophysical regime

(Shine et al. 2002, Forman et al. 2003, Harper et al. 2005, Sears et al. 2011, Sullivan

1981, Huey et al. 1989).

The thermal habitat heterogeneity generated by road corridors and other ROWs can be important for ectotherms, particularly reptiles. Many reptiles rely on behavioral

77 thermoregulation (Cowles & Bogert 1944), adjusting habitat selection and activity times to maintain body temperatures that optimize physiological performance (e.g. Huey 1982,

Huey & Kingsolver 1989, Huey 1991). Excluding endogenous factors (e.g. dispersal capacity, physiological mechanisms), an ectotherm’s ability to thermoregulate in terrestrial environments is fundamentally limited by climate and habitat heterogeneity

(Kearney et al. 2009). Habitat heterogeneity can be a limited resource in both heavily shaded and predominantly open landscapes where the thermal conditions are uniform across space. In forestlands, some reptiles show preference for disturbed anthropogenic habitats including ROWs and forest edges, which researchers have often linked to thermal resource availability (Vitt et al. 1998, Sartorius et al. 1999, Klingenböck et al.

2000, Shine et al. 2002, Blouin-Demers & Weatherhead 2002). Reptiles have also been observed in roadcuts (Myers 1957, Hertz 1979, Anderson 2010) and other rocky roadside features (Kovar et al 2014), which can provide basking surfaces and retreats with buffered thermal environments (Huey et al. 1989). Because the thermal environment affects various metrics of ecological performance (Huey 1982, Huey & Kingsolver 1989,

Huey 1991) and population processes (Dunham et al. 1989), road corridors and other

ROWs can impact demographic vital rates, local abundances, and the distributional range limits of ectotherms (Sartorius et al. 1999, Shine et al. 2002). From the perspective of conservation and management, loss of thermal habitat heterogeneity through ecological succession has been implicated in local extinctions and population declines of multiple temperate reptiles (Hall 1994, Ballinger & Watts 1995, Jäggi & Baur 1999, Fitch 2006), and thus maintenances of open canopy patches may be important for population

78 persistence (Shoemaker & Gibbs 2010, Pike et al. 2011). In addition to thermal resources, other common modifications to roadside environments include the generation of roadside pools in the form of drainage ditches or storm-water retention ponds, and stone piles installed as drainage or erosion control structures. These resources often draw animals to roads and the adjacent purlieu, and problematically, often form as ecological traps because promotion of activity near the roadway can result in increased mortality through road kill, predation, abbreviated hydroperiod, , or pollution (Andrews et al.

2008, Langen et al. 2015).

Roads networks are ubiquitous and expanding (Laurance et al. 2014), and mitigation strategies should seek to protect species drawn to roadside environments by either preventing the creation of ecological traps, or by converting road corridors into compensatory habitat. While exclusion fencing can reduce wildlife-vehicle collisions

(Dodd et al. 2004, Aresco 2005b, Glista et al. 2009, Beckmann et al. 2010), its functional placement is rarely evaluated, and warrants input from both engineers and biologists.

Best management practices (BMPs) currently recommend fence placement as far from the road as possible to avoid conflicts with mowing, snow ploughing, and road maintenance (Gunson et al. 2016). However, if edge and ROW habitats can provide important resources to focal species, it may be appropriate to position exclusion fencing closer to the road to permit access to those resources while also preventing road mortality. Restricting access to ROW habitats may compromise the effectiveness of mitigation structures by creating motivation to trespass, or fail to capitalize on newly generated habitats that may offset the impacts of habitat loss incurred during road

79 creation.

I studied a small population of state-endangered Timber Rattlesnakes (Crotalus horridus) in southeast Ohio, where a major highway recently fragmented a forested landscape and generated a large, open canopy right-of-way. The highway featured exclusion fencing and crossing structures to mitigate road impacts on this state- endangered reptile. Radio telemetry revealed that rattlesnakes trespassed the fence to access roadside habitats, with gravid females spending nearly their entire gestation periods within the road corridor. Considering the importance of maternal thermoregulation (Packard et al. 1977, Shine & Harlow 1993, O’Donnel & Arnold 2005),

I hypothesized that ROW habitats were providing advantageous thermal resources. In this study, I quantified thermal habitats available to rattlesnakes within this landscape using operative temperature models, and related those resources to temperatures preferred by gravid females and voluntary tolerances. I evaluated macrohabitat selection from telemetry data to assess how gravid and non-gravid rattlesnakes were selecting among forest, edge, and ROW habitats. In addition, I evaluated whether the distribution of preferred gestation habitats was limited to or concentrated within the road corridor. This was achieved by generating spatially explicit models of gravid female resource selection based on thermal landscapes and structural variables, and then projecting the top model to quantify the availability of suitable sites within and outside the road corridor. I conclude by discussing management of forest dwelling ectotherms and the effective placement of road mitigation structures.

80

Materials and Methods

Study Species

Timber Rattlesnakes are a viviparous venomous snake (Viperidae) widely distributed across eastern North America (Ernst & Ernst 2003), but often occurring in fragmented, relict populations at its northern and western range due to habitat loss and persecution (Brown 1993). Their life history is characterized by delayed maturity, infrequent reproduction, and high adult survival (Brown 1991, 1993, 2016; Martin 1993,

2001; Olson, 2015), making their populations vulnerable to additive adult morality.

Crotalus horridus are sit-and-wait ambush predators that consume a wide variety of endothermic prey, but eat primarily small mammals (Reinert et al. 1984, 2011, Brown &

Greenberg 1992, Ernst & Ernst 2003). As ambush foragers, C. horridus spend significant periods of time thermoconforming. While foraging, individuals spend long periods of time sedentary at the surface, often multiple days in duration, suggesting that C. horridus must select foraging sites that remain within their thermal tolerances.

However, gravid females typically forgo foraging (Reinert 1984a; but see Brown

2016) like many other viviparous reptiles that exhibit gestation anorexia (Gregory et al.

1999). Gravid females instead spend their time thermoregulating throughout pregnancy to maintain elevated body temperatures (Santana-Gardener & Beaupre 2009). Maternal thermoregulation in viviparous reptiles is a key determinant of reproductive success, as temperatures experienced throughout gestation can affect aspects of embryonic development, including gestation length, developmental abnormalities, and embryo mortality (Packard et al. 1977, Shine & Harlow 1993, Lourdais et al. 2004, O’Donnel &

81

Arnold 2005). Gestation typically lasts 2.5 – 4 months, with parturition occurring from late August through October (Martin 1993). Postpartum females require multiple years to replenish fat deposits (capital breeders, e.g. Jönsson 1997, Bonnet et al. 1998), restricting

C. horridus to superannual reproductive cycles, with triennial being most common and some females going as many as seven years between births (Brown 1991, 2016, Martin

1993). Gravidity also impedes locomotor perform, inhibiting predator evasion (Seigel et al. 1987). Earlier parturition thus reduces time spent in a vulnerable condition and frees more time to forage before overwintering, thereby enhancing survival and potentially shortening the reproductive interval. Collectively, these factors place a high premium on effective and efficient maternal thermoregulation.

The tradeoffs between maternal thermoregulation and ambush foraging results in habitat separation between gravid and non-gravid C. horridus (Reinert 1984a, Reinert &

Zappalorti 1998), where gravid females prefer canopy gaps, and non-gravid snakes use forest habitats with greater canopy closure. Given that canopy gaps are likely limiting resources in forested landscapes, road corridors and ROWs may be important resources to gravid rattlesnakes, providing habitats with increased solar radiation for longer durations and facilitating the maintenance of elevated temperatures. While ROWs could provide high quality gestation sites for gravid females, the thermal conditions could also result in the loss of foraging habitat for conspecifics. Crotalus horridus have been known to use roadsides and other disturbed habitats, ostensibly for their thermal qualities

(Brown 1993, Reinert & Zappalorti 1998). Unfortunately, vehicle strikes can be a common source of mortality for C. horridus, and result in a range of demographic and

82 genetic impacts (e.g. Aldridge & Brown 1995, Rudolph et al. 1997, Sealy 2002, Adams,

2005, Clark et al. 2010, Bushar et al. 2015). Thus, ROWs may function as either an ecological opportunity or ecological trap, making C. horridus a useful model system for evaluating the trade-offs associated with roadside habitats and mitigation placement.

Study Site

The Nelsonville Bypass (NVBP) is a large, four-lane divided highway (13.6 km) that fragmented the Wayne National Forest (WNF) in Athens and Hocking Counties,

Ohio, in 2013. The WNF at this site is an Oak-Hickory (Quercus-Carya) deciduous forest with outcroppings of Pennsylvanian sandstone. In addition to fragmenting the landscape, the NVBP created a large ROW (250 hectares) of predominantly open canopy habitats, including road cuts of exposed sandstone, early successional stands of black locust

(Robinia pseudoacacia) and sumac (Rhus sp.), fields dominated by mixed grasses and weedy vegetation (Poaceae, Asteraceae), barren slopes with exposed soils, and large stone piles (i.e. riprap installed for erosion and drainage control). In the area where rattlesnakes were found, the ROW created south facing slopes and escarpments, providing large open areas for basking, with rocky subsurface retreats in concentrated areas. For the purposes of quantifying thermal resource limitations and habitat selection, I considered three habitat types: forest, edge, and ROW. Forest habitats were areas dominated by overstory trees beyond the limit of direct disturbance, outside the ROW.

Most of the rock outcroppings within the forest are heavily shaded by canopy, and basking opportunities are restricted to small tree fall gaps and other anthropogenic disturbances (e.g. ATV trails, gas line ROW). ROW habitats were areas within the limit

83 of direct disturbance, and characterized by the open canopy successional-roadcut mosaic.

Previous data demonstrated that rattlesnake activity within the ROW was concentrated along the forest edge (Chapter One). As other studies have found edge preference in snakes, I included edge habitats as a third habitat type. Edge habitat was defined as the region within 15 m of the overstory canopy boundary (Blouin-Demers & Weatherhead

2002). Thus, edge habitat was a 30 m wide band of habitat at the overstory canopy interface, which is a distance well within dispersal and movement capabilities of C. horridus. For thermal resource availability models, I define the road corridor as all areas within the limit of direct disturbance including the edge habitat at the canopy interface

(ROW + edge).

Thermal Resource Availability

I used operative temperature models (OTM; Bakken 1992) to quantify the spatiotemporal availability of ecologically relevant thermal resources across the landscape (Grant & Dunham 1988). OTMs model the physical properties of an animal that affect passive heat exchange with its environment (conduction, convection, and radiation) and provide an estimate of the operative environmental temperature (Te). Te is defined as the equilibrium body temperature that an animal would achieve under prevailing environmental conditions at a given microhabitat in the absence of metabolic heating and evaporative cooling (Bakken & Gates 1975, Bakken 1976, 1992). OTMs were constructed from 15.2 cm lengths of copper tubing (3.8 cm diameter, 0.15 cm thickness) and painted to approximate the reflectivity of C. horridus. OTM design and evaluation is described in detail in the appendix. Models were distributed at the surface of

84 each macrohabitat type (35-39 models per habitat, 110 total) at locations used by rattlesnakes and random sites to capture ecologically relevant variation. Because rattlesnakes forage at the surface where temperatures are going to be the most extreme, I prioritized surface locations over subsurface retreats. Radio telemetry data from the 2015 field season was used to select activity locations, and random sites were sampled as random walks from activity locations within the same habitat type by generating a random distance and bearing. Random distances were constrained to the interquartile range of movement step-lengths observed in telemetry data (~10-70 m). Models contained Thermochron ® iButton data loggers (DS1921G ± 0.5 °C resolution) that recorded temperature every 20 minutes.

Radio Telemetry & Body Temperature

I used VHF radio telemetry to track rattlesnakes in the 2015 and 2016 field seasons to evaluate habitat selection, estimate field preferred body temperatures

(hereafter Tb). Rattlesnakes large enough to bear a transmitter were surgically implanted with a temperature sensitive radio transmitter (ATS ® R1680 transmitter, 3.6 grams; <

2% of body mass) following Reinert and Cundall (1982). Prior to implantation, I generated a standard curve for each transmitter relating temperature to the inter-pulse period. This was done by placing each transmitter in a water bath of known temperature and recording the interpulse interval, repeating this process across a range of temperatures (10 – 40˚C), and fitting a power function (R2 > 0.99 for all transmitters).

Interpulse period measurements were then substituted into the transmitter specific regression equation to calculate Tb. I tracked rattlesnakes three times per week

85 throughout the active season using a Communication Specialist ® R1000 receiver and

Yagi 3-Element Antenna. Upon relocating each animal, I recorded the interpulse period,

GPS location (Garmin ® GPSMAP 64), behavior, and noted ambient weather conditions.

To quantify gestation Tb, I monitored a free ranging gravid female in the 2016 field season using a remote telemetry receiver that recorded the interpulse period (Lotek

SRX 800 ®, Ottowa, ; e.g. Beaupre 1995, Blouin-Demers & Weatherhead 2002).

Due to the rugged terrain, dispersal of snakes across the study site, and range of our equipment, we were not able to simultaneously monitor the body temperature of multiple snakes using this method. The telemetry receiver was positioned within 100 m of the gestation site, and sampled the interpulse period every 20 minutes. Tb measurements were collected during the warmest period of the summer from 26 July – 17 August, and thus field active body temperatures should approximate the thermal preferences of a gravid female given there was access to full and thermal refuge (i.e. rocky subsurface retreats, shade). I used observed body temperatures of all telemetered snakes (n = 5,

Chapter One) to identify field minimum and maximum voluntary Tb (hereafter VMIN and

VMAX; Cowles & Bogert 1944). In addition, I implanted a male rattlesnake with a

Therochron ® iButton data logger (DS1921G ± 0.5 °C resolution) that recorded Tb every 20 minutes for the month of August.

Body Temperature & Thermal Habitats

I calculated mean hourly body temperatures (Tb) of the gravid female and mean hourly Te for each habitat type. OTMs were treated as the unit of replication for Te. To identify how Te related to field preferred temperatures, I subtracted the field preferred Tb

86 of a gravid female from individual Te measurements, and calculating the average hourly differences (De) for each habitat. This method assigned positive values to Te above mean

Tb, and negative values to Te below Tb. I further compared thermal environments by assigning Te measurements to temperature classes related to mean gestation Tb and VMAX.

The classes included: (1) the range of mean hourly gestation Tb, (2) less than mean gestation Tb, (3) less than minimum voluntary gestation Tb, (4) greater than mean gestation Tb, (5) greater than maximum voluntary Tb, and (6) greater than overall VMAX.

Summer temperatures were not observed below overall minimum VMIN, and thus this category was excluded.

I quantified eight metrics of daily average thermal conditions across habitat types: mean Te, maximum Te, minimum Te, De (defined as Te – mean gestation Tb), hours above mean gestation temperature (hereafter HG), hours above VMAX (hours that would restrict surface activity, hereafter HRES), range, and standard deviation. I summarized these average daily indices across habitat types using both the mean and median. I tested for differences among habitats across all indices using linear or generalized linear mixed effects models (LMM and GLMM; Pinheiro & Bates 2000), which are functionally similar to repeated measures ANOVA blocking on individual OTMs as a random effect. I also used the ordinal date as a crossed random effect to account for heteroscedasticity produced by variation among days. GLMMs were used to specify the Poisson distribution when analyzing counts of HG and HRES. After running the LMMs and GLMMs, I assessed whether thermal metrics were significantly different among habitats using Tukey post hoc tests with Bonferroni corrections, setting a = 0.002. Statistical tests were performed

87 in R version 3.2 (R Core Team, 2016). LMMs and GLMMs were performed using the lme4 package (Bates et al. 2015), and multiple comparisons were performed using the multcomp package (Horton et al. 2008).

Macrohabitat Selection

I evaluated 3rd order macrohabitat selection (Johnson 1980) using Manly selection ratios (Manly et al. 2002). Manly selection ratios compare the frequency that specific habitats are used relative to their availability in the environment. Selection ratio (wi) were calculated for each defined habitat class, where wi greater than 1 indicates a habitat was selected, wi less than 1 indicates avoidance, and wi overlapping 1 suggests a habitat was used randomly with respect to its availability. I evaluated summer habitat selection (June

– August; the period during which females were gravid) for gravid and non-gravid rattlesnakes separately under both Type III (individual) and Type II (population) availability designs (Manly et al. 2002). Under the Type III availability design, a buffered

100% minimum convex polygon was created for each snake, bounding that individual’s relocations, and defining the extent of available habitat for that individual. The buffer distance was equal to the radius of the respective home range assuming a circular geometry. Under the Type II availability design, the individual buffered-home ranges were combined into a single shape. Forest, edge, and ROW habitats were classified remotely using aerial imagery and LiDAR canopy data in a GIS, where edge habitat was defined as areas within 15 m of the overstory canopy boundary (Blouin-Demers &

Weatherhead 2002). For both availability designs, the proportion of each respective habitat was calculated using ArcMap (ESRI, 2014), and available habitat was bounded by

88 the NVBP road edge because prior analyses have demonstrated that the rattlesnakes avoided the roadway (Chapter One). Selection ratios were calculated in R using the adehabitatHS package (Calenge 2006).

Availability of Gestation Habitat

Logistic regression-based resource selection functions (RSF) were used to predict the distribution of suitable gestation habitat at the landscape scale, and to evaluate whether gestation habitats were limited to or concentrated within the ROW. To include temperature, I generated four thermal landscape surfaces by predicting Te data collected by OTMs using air temperature and spatially referenced GIS layers, which were modeled using linear and generalized linear mixed effects models (Pinheiro & Bates 2000, methods adapted from Fridley 2009). Response variable data was calculated for each

OTM by day, and included daily maximum Te (TeMAX), daily mean Te (TeAVG), HG, or

HRES (Table 1). Fixed effects included a metric of air temperature (either daily maximum air temperature or daily average air temperature), canopy cover, and insolation (the amount of incident solar radiation reaching a given location as a function of latitude and the local geomorphology in units of kWh/m2; Table 1). Daily maximum and average air temperature data from a regional weather station (~20 km from the study site) was downloaded from the National Oceanic and Atmospheric Administration (NOAA), and both metrics were compared as predictors for each response variable using Akaike

Information Criteria (AIC; Burnham & Anderson 2002) to determine which provided better model fit. To obtain canopy cover, canopy heights were measured using LiDAR

(2.5 m resolution horizontal plane; data collected in 2014). Canopy heights were then

89 converted to a binary layer by assigning 1 to cells exceeding 2 m in canopy height.

Canopy cover was derived by calculating the mean value using a moving window approach across a range of sizes to obtain measures at multiple scales, that were evaluated to identify the optimal scale for each response variable (7.5 x 7.5 m [smallest possible square] to a 30 m radius circle). Focal statistics were performed in R using the raster package version 2.5-8 (Hijmans 2016). Solar insolation was calculated from June through August in ESRI ® ArcMap 10.3 (ESRI, 2014) using the Area Solar Radiation tool, and then divided by the number of days to obtain average daily kWh/m2; the digital elevation model (DEM) used to produce the insolation surface was also generated from the same LiDAR data at 2.5 m resolution. In preliminary analyses, interaction terms of fixed effects were included, but dropped because they did not improve model fit. In addition, other explanatory variables were considered in these models including elevation and topographic wetness index (TWI), but they contributed negligible explanatory power.

Elevation varies by only ~120 m across the study site, and permanent streams were largely absent. Canopy cover explained far more variation than insolation when modeling thermal landscapes.

90

Table 2.1: Predictor variables used in thermal landscape and RSF models. Program: ‘Package’ “Function” Predictor Variable Source (Key Parameters) Max Daily/Average Daily NOAA NA Air Temperature

Insolation LiDAR Digital Elevation Model ArcGIS: “Solar Radiation Area” R: ‘raster’ Canopy Cover LiDAR Binary Canopy Layer “Focal Statistics” (Mean) R: ‘raster’ Edgeness Canopy Cover “Focal Statistics” (SD) Average Daily Air Temp., T R: ‘nlme’ “lme” eAVG Canopy Cover, Insolation Max Daily Air Temp., Canopy T R: ‘nlme’ “lme” eMAX Cover, Insolation Max Daily Air Temp., Canopy R: ‘lme4’ “glmer” (Poisson H G Cover, Insolation Distribution) Max Daily Air Temp., Canopy R: ‘glmmADMB’ “glmmadmb” H RES Cover, Insolation (Zero-inflated Poisson Distribution)

All spatial data was sampled and modeled at 2.5 m resolution.

TeMAX and TeAVG models were fitted and evaluated in R using the nlme package version 3.1-128 (Pinheiro et al. 2016). For these models, OTM identity was a random effect and temporal autocorrelation was addressed by applying the corAR1 structure to ordinal date. The HG response variable was Poisson distributed, and thus fitted with generalized linear mixed models using the lme4 package (Bates et al. 2015). HRES was

Poisson distributed, but also zero inflated, and thus I used the glmmADMB package

(Fournier et al. 2012, Skaug et al. 2016), which allowed for the specification of a zero- inflated Poisson distribution. Autocorrelation terms could not be fit using the lme4 and glmmADMB packages, and ordinal date was thus listed as a crossed random effect. Six

OTMs were censored from this analysis because their residuals were consistently large, suggesting that canopy conditions had changed at these locations since LiDAR data was collected. After the best model was selected for each thermal surface, the raster layers

91 were plugged into the model equation to predict each thermal variable across the landscape. For maximum and average air temperature values, we chose the mean summer value for the duration that models were deployed (maximum air = 31˚C, average air =

24˚C).

Finally, logistic regression RSFs were used to evaluate gravid female (n=2 snakes, 89 relocations) habitat selection under a use-availability design, using Type III availability. I generated 300 random sample points in QGIS (Quantum GIS Development

Team, 2017) spaced a minimum of 20 m apart within each snake’s home range, and sampled values of each independent variable at snake relocations and random-availability locations using the raster package in R. This number of random points was chosen because it provided extensive and even coverage within the limits of available habitat. I generated a candidate set of models containing either thermal landscapes, or the constituent structural variables used to generate the thermal landscapes (canopy cover and solar radiation), with and without habitat “edgeness”. Edgeness was calculated by measuring the standard deviation of canopy cover using a moving window approach as described above, and can also be considered a metric of canopy heterogeneity. Predictor variables that showed a correlation greater than 0.4 were not included in the same model, and thus I did not test models that simultaneously included canopy cover and thermal landscapes. I did not consider geomorphological variables (e.g. slope, elevation) because parameter estimates would be confounded with basking habitat conditions in the ROW.

Most of the habitat in the ROW occurred on steep slopes due to the roadcut, making it difficult to disentangle the effects of slope and the thermal environment with this sample

92 size. RSFs were fitted using generalized linear mixed effects models, specifying the binomial distribution with the logit link function in R using the lme4 package (Bates et al.

2015). Individual rattlesnake identity was treated as a random effect. Models were evaluated using receiver operating characteristic (ROC) and 10-fold cross validation to measure the area under the curve (i.e. AUC, the classification ratio of correct to incorrect model predictions, where AUC = 1 indicates perfect classification, and AUC = 0.5 indicates no predictive ability). The top model, as measured by both AICc and AUC, was then used to predict the relative probability of suitable gestation habitat on the landscape, w(x), using the following equation (Boyce & McDonald 1999):

%#&(( + ( # + ( # + … + ( # ) ~" # = ) , , - - / / 1 + %#&(() + (,#, + (-#- + … + (/#/)

In the model, w(x) is estimated by k independent variables for x resource units expected to influence resource selection, and where b denotes parameter coefficients estimated by the model. For each value of k, a different raster surface is substituted into the equation and used to generate the resource selection probability surface. I used the quartile ranges of predicted values to bin the landscape into categories ranging from lowest potential (class 1) to highest potential (class 4) for suitable gestation habitat

(Hodder et al. 2014). I then tabulated the proportion of the most suitable areas within both the forest and ROW portions of rattlesnake habitat under the Type II availability area to determine whether the most suitable gestation habitats were limited to or concentrated within the ROW.

93

Results

Body Temperature

I recorded more than 460,000 Te readings (n = 110 OTMs) and more than 1,500

Tb readings (n = 2 snakes, 1 gravid female, 1 male). Of the Tb readings, 881 Tb were from the gravid female. The gravid female maintained a mean hourly Tb range between

26.8 - 30.2˚C (absolute range = 24.3 - 35.1˚C, overall mean hourly Tb = 28.6˚C,

SD=1.08), which varied less than hourly Te profiles of each habitat (Figure 1). The male exhibited lower mean hourly Tb with greater variation compared with the gravid female and generally conformed to the forest environment (male mean hourly Tb range = 21.38 -

28.43˚C male, overall mean hourly Tb = 24.51˚C, SD = 2.53; Appendix 1 Figure 2).

Though these time periods were not entirely overlapping, constraining the dates of male and gravid female Tb measurements to the overlapping period did not substantively change their values or relative relationships. The Te classification scheme (Table 2) was used to evaluate and compare spatiotemporal variation in thermal resource availability among the habitats (Figure 2), and to evaluate surface habitat availability for foraging and gestation (Figure 3)

Thermal Resources

ROW and edge habitats provided rattlesnakes significantly warmer habitats for longer durations of the day compared with forested habitats (Figures 1-3; Table 3). Mean hourly temperatures in the forest converged on preferred gestation temperatures during only the hottest hours of the day (Figure 1), while edge and ROW habitats exceeded those temperatures for 2 – 4 times as many hours (Table 3), warming earlier in the day (Figure

94

3), and remaining warmer later into the evening. However, ROW temperatures in the afternoon routinely exceeded VMAX at the surface throughout the summer (Figures 2-3).

While mean edge Te suggested that these habitats too exceeded VMAX, median values and surfaces plots revealed that these areas were generally tolerable (Table 3, Figure 2), with only about 25% of the edge surface habitats exceeding VMAX at the hottest moments of the day (Figure 3). While average temperatures at the ROW surface often exceeded VMAX throughout midday, 40-50% of surface habitat remained below VMAX, indicating that thermal refuge was available at the surface if snakes were willing to move

(Figure 2). While daily maximum Te was significantly different among habitat types, daily minimum Te was not (Table 3), and thus the range of Te within each habitat was a largely a function of maximum Te.

Figure 2.1: Mean hourly Tb and Te profiles with 95% CI. Left: mean hourly Tb of a gravid rattlesnake and mean hourly Te of each habitat type; dashed line = VMAX. Right: relative De (Te – Tb) across hours centered on mean hourly Tb. De > 0: Te exceeded Tb, De < 0: Te below Tb. Gravid snake Tb was monitored from 26 July – 17 August 2016, while Te was measured from 21 June – 29 August 2016. Restricting Te measurements to the period concurrent with Tb measurements did not substantively change the shape of the curves or magnitude of the differences shown.

95

Table 2.2: Classification scheme of temperature ranges. Classification Description Range (°C) Derived from

Field-preferred Range of mean hourly T recorded Mean Gestation T 26.8 – 30.2 b b gestation temperatures during gestation

T < Mean Gestation T & T ≥ Below mean hourly e b e < Mean Gestation T 24.3 – 26.7 absolute minimum observed gestation b gestation temperatures Tb Exceeding mean T > Mean Gestation T & T ≤ hourly gestation e b e > Mean Gestation T 30.2 – 35.1 absolute maximum observed gestation b temperatures T b

Exceeding maximum T > Absolute Max. Observed Gestation > Max Gestation Tb 35.1 – 39.8 gestation temperature Temp and T ≤ voluntary maximum Tb

Below minimum T < Absolute Min. Observed Gestation < Min Gestation Tb < 24.3 gestation temperature Tb

Exceeding maximum > Absolute maximum field observed T > V voluntary > 39.8 b MAX in a free ranging snake temperatures

96

Figure 2.2: Spatiotemporal availability of thermal resources across habitats. Top panels: average Te class available within each habitat for 20-minute intervals from June 21 - August 29. Middle panels: proportion of habitat (OTMs) available within each Te class across 20-minute intervals for each habitat. Bottom panel: overall proportion of time each experienced a given Te class.

97

Figure 2.3: Availability of gestation Tb and voluntary temperature ranges. Gestation availability shows the average proportion of OTMs in each habtiat between 26.8 – 30.2 °C Te throughout the day, and foraging availability (or general availability) shows the proportion of OTMs below VMAX, < 39.8 °C.

98

Table 2.3: Average daily thermal conditions across macrohabitat type.

Te Metric Forest Edge ROW P<0.002 Mean 23.3 (23.5) 25.7 (25.8) 29.1 (29.3) All SD 3.1 (2.9) 5.9 (4.2) 8.6 (9.5) All Maximum 31.0 (30.1) 40.2 (36.3) 47.6 (50.3) All Minimum 19.6 (20.0) 19.8 (20.5) 20.4 (21.3) NS Range 11.4 (9.9) 20.4 (16.3) 27.2 (29.5) All

De -5.3 (-5.1) -2.9 (-3.6) 0.5 (1.2) All

HG 2.1 (1.0) 4.9 (5.0) 8.5 (9.7) All

HRES < 0.1 (0.0) 1.5 (0.0) 3.9 (5.3) All Daily summary statistics were first calculated with OTMs as the unit of replication. Reported are mean values of each thermal habitat parameter with median values in parentheses. All units are ˚C except HG and HRES which are in hours. Significance of multiple comparisons among habitats are reported with a = 0.002. All = all pairwise comparison were significant. NS = not significant.

Macrohabitat Selection

Manly selection ratios indicated that gravid females (89 relocations of 2 individuals) selected edge habitat, used ROW habitats, and avoided forest habitats, while non-gravid rattlesnakes (291 relocations of 3 individuals; 1M, 2F) avoided ROW habitats, and used both forest and edge habitats (Figure 4). These findings were generally consistent between Type II and Type III availability designs, but gravid snake confidence intervals overlapped with 1 for edge habitat under Type III design, which appeared to be a consequence of one animal’s range spanning a large extent of forest edge (increasing its calculated availability), rather than reduced use of edge habitat. In summary, gravid females avoided interior forest habitats in favor of using the forest-edge and ROW, while non-gravid snakes avoided exposed habitats in the ROW.

99

Figure 2.4: Manly selection ratios for rattlesnakes summer habitat use. Selection ratios (Wi) ± 95% CI. Confidence intervals overlapping with 1 suggest a habitat was used randomly relative to its availaility, while Wi > 1 and Wi < 1 suggest a habitat was selected or avoided, respectively. To maintain a type I error rate of 0.05, a Bonferonni correction was applied to confidence intervals, dividing 2 by the number of habitats, resulting in an adjusted 2 = 0.0167. Top panels: gravid females (89 relocations of 2 individuals). Bottom panels: non-gravid rattlesnakes (males + non-gravid females; 291 relocations of 3 individuals). Left panels: design II availability (pooled population availability). Right panels: design III availability (indivdiaual based availability). Only summer reloaction data were used (June – August, 2015 and 2016).

Availability of Gestation Habitats

The top gravid female RSF model was positively related to average daily Te and habitat edge (Table 4, 5), and showed a high level of predictive accuracy (AUC = 0.90).

The top quartile range of fitted values (class 4 suitability, 75th percentile and above)

100 contained 83% of the gravid female relocations. The model predicted that more than 60% of class 4 habitats available to rattlesnakes were found within the road corridor, even though the road corridor only accounted for 10 - 20% of the total habitat depending on whether availability was defined as a buffered or unbuffered home range (Table 6).

Suitable gestation habitats were more concentrated within the road corridor (>50% of area, Table 6) compared with areas outside the road corridor (<10% of area, Table 6).

Table 2.4: Model selection results of gravid female RSF. Model K AICc ∆AICc AICc AUC (SE)

TeAVG + Edge 4 321.1 0.0 0.99 0.90 (0.02) Canopy 2 + Insolation + Edge 5 331.5 10.4 0.01 0.89 (0.02) Canopy 1 + Insolation + Edge 5 335.9 14.8 0.00 0.88 (0.02)

TeMAX + Edge 4 342.9 21.8 0.00 0.88 (0.02)

HG + Edge 4 343.2 22.1 0.00 0.88 (0.02)

HRES + Edge 4 363.2 42.2 0.00 0.85 (0.02) Edge 3 371.0 49.9 0.00 0.80 (0.03)

TeAVG 3 467.0 145.9 0.00 0.81 (0.02) Canopy 2 + Insolation 4 478.6 157.5 0.00 0.79 (0.02) Canopy 1 + Insolation 4 480.9 159.8 0.00 0.76 (0.01)

TeMAX 3 488.8 167.7 0.00 0.76 (0.01)

HG 3 509.6 188.6 0.00 0.79 (0.02)

HRES 3 529.4 208.4 0.00 0.76 (0.01) Null 2 530.2 209.1 0.00 0.50 (0.00) RSF models contained four thermal landscape metrics, edge (standard deviation of canopy cover), a solar insolation surface (kWH / m2*day), two metrics of canopy cover (5 m and 10 m radius, i.e. Canopy 1 and 2 respectively; larger windows were tested but found to erode predictive ability).

Table 2.5: Top RSF model (TeAVG + Edge) parameters estimates. Parameter Coefficient Estimate SE Z P Intercept -3.3228 0.2870 -11.58 <0.001

TeAVG 1.2880 0.1991 6.47 <0.001 Edge 1.3734 0.1347 10.20 <0.001 Coefficient estimates shown are scaled for interpretation, but habitat projections used unscaled estimates.

101

Table 2.6: Predicted gestation habitat availability. Gestation Habitat Availability Within Road Corridor Outside Road Corridor Buffered Home Range 7.6 ha (53.4%) 4.7 ha (6.8%) Unbuffered Home Range 1.8 ha (52.4%) 0.9 ha (2.6%) Area of gestation habitat is reported in hectares. The percentage each habitat's total area predicted to be suitable gestation habitat is shown in parentheses (the concentration).

Discussion

In this study, I evaluated how a road ROW altered the thermal landscape and influenced resource selection by gravid and non-gravid C. horridus. I was only able assess field preferred temperatures of one gravid C. horridus because this was a small, locally endangered population. However, overall mean hourly Tb this female was surprisingly equal to the value reported for gravid C. horridus in , and the minimum and maximum hourly means were within 1.6˚C (Arkansas: overall mean Tb =

28.6˚C, mean hourly Tb range = 25.2 - 31.7˚C, n = 5 gravid C. horridus; Santana-

Gardener & Beaupre 2009). My observed VMAX (39.8˚C) was only 2.4˚C higher than reported in Arkansas (cf. 37.4˚C, Wills & Beaupre 2000), and approaches values of

CTMAX common in many reptiles (Brattstrom 1965). Collectively, these findings suggest that the temperature classification used in this study and my evaluation of thermal parameters was reasonable, despite the low sample size of animals.

In both summers, I observed gravid females give birth on the side of an exposed rocky roadcut at the forest edge. Those gravid females used ROW and edge habitats through the duration that they were gravid, and returned to the forest habitats after parturition. Consistent with macrohabitat selection, RSF models identified that the road corridor was providing conditions sought by gravid females. While not limited to the road corridor, the RSF revealed that the most suitable conditions for gestation were

102 concentrated therein. All else being equal, gravid female RSF models that included average daily Te consistently outperformed models that included only canopy cover and solar insolation (Table 4). This suggests that temperature was an important driver of habitat selection, and that ecological process variables can be more informative than their structural surrogates (Peterman et al. 2014). However, habitat edgeness was the most important predictor of gravid female habitat selection, and suggests that structural habitat was an important component of habitat selection (Table 4). It is difficult to divorce the structural habitat from the thermal environment, as the former often generates heterogeneity in the latter. It’s possible that gravid females did not need to travel far beyond the forest edge to attain their desired Tb, and thus avoid excess irradiance and exposure to predators within the ROW. Rocky retreats within the road corridor provided protection from extreme temperatures, and allowed gravid females to maintain elevated body temperatures throughout the night (Figure 1). Consequently, gravid females should be able to thermoregulate more effectively within the road corridor, conferring benefits ranging from optimal embryonic development to reduced gestation length (Packard 1977,

Lourdais et al. 2004). In addition, the proximity of the ROW to hibernacula (~50 - 350 m) may have reduced travel distances to viable gestation sites, and thereby reduced energetic costs and predation risk.

In contrast, non-gravid snakes were most active within the forest, and sometimes used edge habitats. It’s worth noting that higher selection only indicated that a habitat was used more often relative to its availability, and in this case, forest was used more often than edge by non-gravid snakes despite edge having a higher selection ratio. Non-

103 gravid snakes were also active within the road corridor, but activity within the open canopy was restricted to densely vegetated areas close to the overstory boundary, and thus classified as edge. Activity in road corridor by non-gravid snakes was often associated with behaviors linked to elevated body temperatures (shedding, healing). The

ROW may be a poor foraging area for C. horridus. Higher average temperatures increase the metabolic costs of surface activity (e.g. Beaupre 1995). Moreover, maximum temperatures regularly exceeding VMAX would require commuting to and from retreat sites, further increasing energetic costs and risk of predation (Cowles & Bogert 1944,

Huey & Slatkin 1976). These conditions may be viable for active foraging snakes (e.g.

Coluber constrictor), but for an ambush predator that often spends days waiting for a meal, the ROW is a costly place to hunt.

Many studies have report antagonistic impacts of road corridors and other ROWs on reptile populations. Kovar et al. (2014) observed a Bohemian Aesculapian Snake

(Zamenis longissimus) population use a rocky road embankment for thermoregulation with little adult road mortality, but juveniles did not avoid the road and were often killed.

Blouin-Demers & Weatherhead (2002) found that snakes (Pantherophis spiloides

[ obsoleta]) preferentially used edge habitat as a thermal resource while gravid, but population viability analysis by Row et al. (2007) showed that even modest levels of road mortality could jeopardize population persistence. Hódar et al. (2000) found that areas of high road density and other anthropogenic habitats were providing habitats preferred by chameleons (Chameleo chameleon), but noted that this species commonly suffers from road mortality and illegal collection. Shine et al. (2002) found that a

104 powerline ROW alleviated thermal reproductive constraints on an oviparous lizard

(Bassiana duperreyi), but speculated that by concentrating lizards and nesting activity within these features, population connectivity would increase via erosion of the typical metapopulation structure imposed by dependence on forest gaps, and suggested that such changes could prevent local adaptation or increase disease transmission. Klingenböck et al. (2000) viewed road corridors as a feature that may enhance connectivity between patches of suitable habitat for the Australian lizard Egernia major, while Vitt et al. (1998) and Sartorius et al. (1999) viewed clearings and roads corridors as a means by which

Ameiva ameiva and other heliothermic lizards could disperse deeper into Neotropical forests, potentially disrupting prey populations and displacing competitors. It has been reported widely that road corridors often attract female turtles for nesting (Haxton 2000), which results in ecological trap formation and female biased mortality (Aresco 2005a,

Steen et al. 2006).

The common thread among these studies was that reptiles were drawn to disturbed habitats to access thermal resources, underscoring the critical importance of temperature in reptile habitat selection (Huey 1991, Reinert 1993). Road corridors and other ROWs create concentrated thermal resources that span large linear expanses, making them easier to locate and less ephemeral relative to small canopy gaps. The extreme thermal regimes found in ROW habitats increase the spatial and temporal available of warmer thermal environments, and provide opportunities for precise thermoregulation when structural heterogeneity and retreat sites are also available therein

(this study, Shine et al. 2002). Thus, ROWs and their edges may confer benefits over

105 natural canopy gaps, and ultimately become preferred habitat for some species (Sartorius et al. 1999). Simulations and experiments by Sears et al. (2016) predict that ectotherms should benefit more from landscapes with small and dispersed patches of thermal resources, rather than large and aggregated thermal resources, because they should increase the effectiveness and efficiency of thermoregulation by reducing energy expenditure and movement (Sears et al. 2016). However, our study and others have shown that large disturbances can create thermal conditions that are uncommon or absent in natural habitats (Vitt et al. 1998, Sartorius et al 1999). In thermally limited circumstances, reptiles may utilize whatever basking resources are available, even if they are not optimal.

Consequences for Management

Given that the landscape matrix remains forest, within which basking opportunities for gravid females are limited, it is likely that the benefits of new and improved basking habitat will outweigh the costs of reduced foraging territory. Similarly, other studies have found that C. horridus populations can tolerate modest logging operations (Reinert et al. 2011, MacGowan et al. 2017). However, scale is important: C. horridus is a forest dwelling snake whose reproductive ecology requires canopy gaps.

Thus, populations require a forest matrix interspersed with gaps, and not the opposite.

Though the ROW is substantial, snakes still have access to abundant forest habitat for foraging. Elsewhere in the Wayne National Forest, rattlesnakes could benefit from selective felling or canopy thinning, especially over south facing rock outcroppings, and it would be best to leave felled trees as microhabitat. By reducing thermal constraints,

106 demographic responses and population persistence will depend on connectivity, road avoidance, road mitigation, forest management, and other human dimensions including persecution and collection.

Connectivity among dens or populations has been challenging to quantify at the study site because rattlesnakes in Ohio often use small satellite dens scattered across the landscape (Coupe 1997). Camera surveys at the study site and reports of rattlesnakes crossing ATV trails have provided evidence that the population is spread diffusely across

WNF. While genetic diversity has not been evaluated, outward signs of inbreeding were not observed (e.g. monomorphic, aberrant colorations; Clark et al. 2011). On the other hand, inbreeding depression could develop over time if the highway severed historical routes of gene flow (Clark et al. 2010), which is a concern for this species because long- term isolation ostensibly rendered a C. horridus population in more vulnerable to disease (Clark et al. 2011).

The combined factors of small population size and long generation time suggest that the population will not be able to withstand much additive mortality, and thus road mortality, illegal collection, wanton killings, and increased predation could all cause rapid population decline and extirpation. If rattlesnakes were to attempt crossing the

NVBP, the probability of road mortality would be high because C. horridus cross roads slowly, often pausing at the sign of traffic, and the traffic intensity exceeds 17,000 vehicles per day (Andrews & Gibbons 2010). Crotalus horridus show some natural aversion to road crossing (Andrews & Gibbons 2005), and avoidance in other species scales with the road size and traffic intensity (e.g. Brehme et al. 2013). We have thus far

107 not observed road mortality at the NVBP by either radio telemetry or road mortality surveys, and prior analyses indicated that rattlesnake movements were consistent with road avoidance (Chapter One). However, my sample size was small. Given the high cost of additive mortality, it is recommended that mitigation fencing be reconstructed.

Best management practices recommend that amphibian and reptile exclusion fencing be placed as far as possible from the roadway (Gunson et al. 2016), but in this study I have arrived at the opposite conclusion. While road construction and habitat fragmentation should be avoided whenever possible, road ROWs can generate important habitat heterogeneity within forested landscapes, and exclusion fencing should not restrict access to those resources. My results suggest that reptiles will seek ways to trespass fencing under these conditions (Chapter One). Arguments for distant mitigation placement are often grounded in the logistics of transportation infrastructure, on the basis of avoiding interference with mowing, snow ploughing, and other road maintenance activities (Gunson et al. 2016). However, mitigation fencing placed 5-10 m from the road should not pose significant obstacles to those activities in most places, and instead, should facilitate the maintenance and vegetation control necessary for continued function

(e.g. Feldhamer et al. 1986, Dodd et al. 2004, Baxter-Gilbert et al. 2015). Placement closer to the road would also likely reduce damage by natural processes, as placement within forest habitats or at the forest edge increases vulnerability to tree falls. In addition, fence placement along the roadside where the earth has been graded and stabilized would help reduce erosion damage. Signs could also be posted encouraging the public to contact

108 transportation agencies when fencing is in disrepair, and may get the public more invested in road mitigation.

Conclusions

Given that sympatric species and conspecifics can have starkly different resource requirements and habitat preferences (Schoener 1974, e.g. Reinert 1984a, 1984b), the distinction between habitat restoration and habitat degradation can be complex.

Roadsides are often considered inherently degraded habitats because they favor weedy species, introduce edge-effects harmful to forest interior species (e.g. Chalfoun et al.

2002), and can become ecological traps (Langen et al. 2015). This sentiment extends to developed landscapes more broadly, where floral and faunal diversity has been homogenized in favor of urban adapted species (McKinney 2006). At global and regional scales, these trends are concerning. The impacts of urbanization and edge-effects have become ever more pervasive as humans have carved the globe with cities and a reticulated transportation infrastructure (McKinney 2006, Trombulak & Frissel 2000).

However, ecological responses can vary across scales (Wiens 1989, Levin 1992). At local scales, forest canopy thinning has been demonstrated as an effective restoration technique to increase diversity and abundance of some reptiles (Pike et al. 2011). Such habitat modifications may be especially important for improving habitat quality in fire- suppressed habitats (Webb et al. 2005). Similarly, many thermophilic reptiles use microhabitats characterized by rock outcroppings (e.g. Reinert 1984a) for their physical and thermal properties (Huey et al. 1989, Croak et al. 2008). These species can be impacted by the removal of rocky habitats (Shine et al. 1998) or their shading (Pike et al.

109

2011), and benefit from their restoration (Croak et al. 2010) and sunning (Pike et al.

2011). Many have also observed the negative consequences of unchecked forest succession on temperate reptile populations and communities, which has resulted in the declines and extirpations of numerous snakes and lizards (Hall 1994, Ballinger & Watts

1995, Fitch 2006). By creating or restoring open habitats, ROWs and roadcuts may benefit local reptile populations (Jäggi & Baur 1999, this study) under circumstances when additive mortality is low, which can be helped by effective placement of mitigation structures.

110

LITERATURE CITED

Adams, J. P. 2005. Home Range and Behavior of the Timber Rattlesnake (Crotalus horridus). Marshall University.

Aldridge, R., and W. Brown. 1995. Male Reproductive Cycle, Age at Maturity, and Cost of Reproduction in the Timber Rattlesnake (Crotalus horridus). Journal of Herpetology 29:399–407.

Anderson, C. D. 2010. Effects of Movement and Mating Patterns on Gene Flow among Overwintering Hibernacula of the Timber Rattlesnake (Crotalus horridus). Copeia 2010:54–61.

Andrews, K. M. 2010. Snakes in fragmenting landscapes: An investigation of linear barriers and landscape alterations. University of .

Andrews, K. M., J. W. Gibbons, and D. M. Jochimsen. 2008. Ecological Effects of Roads on Amphibians and Reptiles: A Literature Review. Pages 121–143in J. C. Mitchell, R. E. Jung Brown, and B. Bartholomew, editors. Urban Herpetology. Society for the Study of Amphibians & Reptiles, Salt Lake City, Utah.

Andrews, K. M., P. Nanjappa, and S. P. D. Riley. 2015. Roads and Ecological Infrastructure: Concepts and Applications for Small Animals. Johns Hopkins University Press, Baltimore, MD.

Andrews, K., and J. Gibbons. 2005. How Do Highways Influence Snake Movement? Behavioral Responses to Roads and Vehicles. Copeia 2005:772–782.

Angilleta Jr., M. J. 2009. Thermal Adaptation: A Theoretical and Empirical Synthesis. Oxford University Press, Oxford, New York.

Aresco, M. J. 2005a. The effect of sex-specific terrestrial movements and roads on the sex ratio of freshwater turtles. Biological Conservation 123:37–44.

Aresco, M. J. 2005b. Mitigation Measures to Reduce Highway Mortality of Turtles and Other Herpetofauna at a North Florida Lake. The Journal of Wildlife Management 69:549–560.

Ashley, E. P., and J. T. Robinson. 1996. Road mortality of amphibians, reptiles and other wildlife on the long point causeway, Lake Erie, Ontario. Canadian Field-Naturalist 110:403–412.

111

Ashley, P. E., A. Kosloski, and S. A. Petrie. 2007. Incidence of Intentional Vehicle– Reptile Collisions. Human Dimensions of Wildlife 12:137–143.

Bakken, G. S., and D. M. Gates. 1975. Heat-Transfer Analysis of Animals: Some Implications for Field Ecology, Physiology, and Evolution. Pages 255–290in D. M. Gates and R. B. Schmerl, editors. Perspectives of Biophysical Ecology. Springer Berlin Heidelberg, Berlin, Heidelberg.

Bakken, G. S., and M. J. A. Jr. 2014. How to avoid errors when quantifying thermal environments. Functional Ecology 28:96–107.

Bakken, G. S. 1976. An improved method for determining thermal conductance and equilibrium body temperature with cooling curve experiments. Journal of Thermal Biology 1:169–175.

Bakken, G. S. 1992. Measurement and Application of Operative and Standard Operative Temperatures in Ecology. American Zoologist 216:194–216.

Ballinger, R. E., and K. S. Watts. 1995. Path to Extinction: Impact of Vegetational Change on Lizard Populations on Arapaho in the Nebraska Sandhills. American Midland Naturalist 134:413–417.

Bates, D., M. Maechler, B. Bolker, and S. Walker. 2015. Fitting Linear Mixed-Effects Models Using lme4. Journal of Statistical Software 67:1–48.

Baxter-Gilbert, J. H., J. L. Riley, D. Lesbarreres, and J. D. Litzgus. 2015. Mitigating reptile road mortality: Fence failures compromise ecopassage effectiveness. PLoS ONE 10:1–15.

Beaupre, S. J. 1995. Effects of Geographically Variable Thermal Environment on Bioenergetics of Mottled Rock Rattlesnakes. Ecology 76:1655–1665.

Beckmann, J. P., A. P. Clevenger, M. P. Huijser, and J. A. Hilty. 2010. Safe Passages: Highways, Wildlife, and Habitat Connectivity. Island Press, Washington, D.C.

Berger, J. 2007. Fear, human shields and the redistribution of prey and predators in protected areas. Biology Letters 3:620–623.

Bernardino, F. S., and G. H. Dalrymple. 1992. Seasonal activity and road mortality of the snakes of the Pa-hay-okee wetlands of Everglades National Park, USA. Biological Conservation 62:71–75.

Beyer, H. L. 2015. Geospatial Modelling Environment.

112

Blouin-Demers, G., and P. J. Weatherhead. 2002. Habitat-specific behavioural thermoregulation by black rat snakes (Elaphe obsoleta obsoleta). Oikos 97:59–68.

Bodie, J. R., and R. D. Semlitsch. 2000. Spatial and Temporal Use of Floodplain Habitats by Lentic and Lotic Species of Aquatic Turtles. Oecologia 122:138–146.

Böhm, M., B. Collen, J. E. M. Baillie, P. Bowles, J. Chanson, N. Cox, G. Hammerson, M. Hoffmann, S. R. Livingstone, M. Ram, A. G. J. Rhodin, S. N. Stuart, P. P. van Dijk, B. E. Young, L. E. Afuang, A. Aghasyan, A. García, C. Aguilar, R. Ajtic, F. Akarsu, L. R. V Alencar, A. Allison, N. Ananjeva, S. Anderson, C. Andrén, D. Ariano-Sánchez, J. C. Arredondo, M. Auliya, C. C. Austin, A. Avci, P. J. Baker, A. F. Barreto-Lima, C. L. Barrio-Amorós, D. Basu, M. F. Bates, A. Batistella, A. Bauer, D. Bennett, W. Böhme, D. Broadley, R. Brown, J. Burgess, A. Captain, S. Carreira, M. del R. Castañeda, F. Castro, A. Catenazzi, J. R. Cedeño-Vázquez, D. G. Chapple, M. Cheylan, D. F. Cisneros-Heredia, D. Cogalniceanu, H. Cogger, C. Corti, G. C. Costa, P. J. Couper, T. Courtney, J. Crnobrnja-Isailovic, P. A. Crochet, B. Crother, F. Cruz, J. C. Daltry, R. J. R. Daniels, I. Das, A. de Silva, A. C. Diesmos, L. Dirksen, T. M. Doan, C. K. Dodd, J. S. Doody, M. E. Dorcas, J. Duarte de Barros Filho, V. T. Egan, E. H. El Mouden, D. Embert, R. E. Espinoza, A. Fallabrino, X. Feng, Z. J. Feng, L. Fitzgerald, O. Flores-Villela, F. G. R. França, D. Frost, H. Gadsden, T. Gamble, S. R. Ganesh, M. A. Garcia, J. E. García-Pérez, J. Gatus, M. Gaulke, P. Geniez, A. Georges, J. Gerlach, S. Goldberg, J. C. T. Gonzalez, D. J. Gower, T. Grant, E. Greenbaum, C. Grieco, P. Guo, A. M. Hamilton, K. Hare, S. B. Hedges, N. Heideman, C. Hilton-Taylor, R. Hitchmough, B. Hollingsworth, M. Hutchinson, I. Ineich, J. Iverson, F. M. Jaksic, R. Jenkins, U. Joger, R. Jose, Y. Kaska, U. Kaya, J. S. Keogh, G. Köhler, G. Kuchling, Y. Kumlutaş, A. Kwet, E. La Marca, W. Lamar, A. Lane, B. Lardner, C. Latta, G. Latta, M. Lau, P. Lavin, D. Lawson, M. LeBreton, E. Lehr, D. Limpus, N. Lipczynski, A. S. Lobo, M. A. López-Luna, L. Luiselli, V. Lukoschek, M. Lundberg, P. Lymberakis, R. Macey, W. E. Magnusson, D. L. Mahler, A. Malhotra, J. Mariaux, B. Maritz, O. A. V Marques, R. Márquez, M. Martins, G. Masterson, J. A. Mateo, R. Mathew, N. Mathews, G. Mayer, J. R. McCranie, G. J. Measey, F. Mendoza-Quijano, M. Menegon, S. Métrailler, D. A. Milton, C. Montgomery, S. A. A. Morato, T. Mott, A. Muñoz-Alonso, J. Murphy, T. Q. Nguyen, G. Nilson, C. Nogueira, H. Núñez, N. Orlov, H. Ota, J. Ottenwalder, T. Papenfuss, S. Pasachnik, P. Passos, O. S. G. Pauwels, N. Pérez-Buitrago, V. Pérez-Mellado, E. R. Pianka, J. Pleguezuelos, C. Pollock, P. Ponce-Campos, R. Powell, F. Pupin, G. E. Quintero Díaz, R. Radder, J. Ramer, A. R. Rasmussen, C. Raxworthy, R. Reynolds, N. Richman, E. L. Rico, E. Riservato, G. Rivas, P. L. B. da Rocha, M. O. Rödel, L. Rodríguez Schettino, W. M. Roosenburg, J. P. Ross, R. Sadek, K. Sanders, G. Santos-Barrera, H. H. Schleich, B. R. Schmidt, A. Schmitz, M. Sharifi, G. Shea, H. T. Shi, R. Shine, R. Sindaco, T. Slimani, R. Somaweera, S. Spawls, P. Stafford, R. Stuebing, S. Sweet, E. Sy, H. J. Temple, M. F. Tognelli, K. Tolley, P. J. Tolson, B. Tuniyev, S. Tuniyev, N. Üzüm, G. van Buurt, M. Van Sluys, A. Velasco, M.

113

Vences, M. Veselý, S. Vinke, T. Vinke, G. Vogel, M. Vogrin, R. C. Vogt, O. R. Wearn, Y. L. Werner, M. J. Whiting, T. Wiewandt, J. Wilkinson, B. Wilson, S. Wren, T. Zamin, K. Zhou, and G. Zug. 2013. The of the world’s reptiles. Biological Conservation 157:372–385.

Bonnet, X., D. Bradshaw, and R. Shine. 1998. Capital versus Income Breeding: An Ectothermic Perspective. Oikos 83:333–342.

Bonnet, X., G. Naulleau, and R. Shine. 1999. The dangers of leaving home: dispersal and mortality in snakes. Biological Conservation 89:39–50.

Boyce, M. S., and L. L. McDonald. 1999. Relating populations to habitats using resource selection functions. Trends in Ecology & Evolution 14:268–272.

Brady, S. P., and J. L. Richardson. 2017. Road ecology: shifting gears toward evolutionary perspectives. Frontiers in Ecology and the Environment 15:91–98.

Brattstrom, B. H. 1965. Body Temperatures of Reptiles. The American Midland Naturalist 73:376–422.

Brehme, C. S., J. A. Tracey, L. R. McClenaghan, and R. N. Fisher. 2013. Permeability of roads to movement of scrubland lizards and small mammals. Conservation Biology 27:710–720.

Breininger, D. R., M. L. Legare, and R. B. Smith. 2004. Eastern Indigo Snakes ( couperi) in Florida: Influence of Edge Effects on Population Viability. Pages 299–311in H. R. Akcakaya, M. A. Burgman, O. Kindvall, C. C. Wood, P. Sjogren-Gulve, J. S. Hatfield, and M. A. McCarthy, editors. Species Conservation and Management: Case Studies. Oxford University Press, Oxford New York.

Brodie, J. F., E. S. Post, and D. F. Doak. 2012. Wildlife Conservation in a Changing Climate. University of Chicago Press, Chicago, IL.

Brown, W. S., and D. B. Greenberg. 1992. Vertical-tree ambush posture in Crotalus horridus. Herpetological Review 23:78–82.

Brown, W. S. 1993. Biology, Status, and Management of the Timber Rattlesnake (Crotalus horridus): a Guide for Conservation. SSAR Publications No. 22:78pp.

Brown, W. S. 2016. Lifetime Reproduction in a Northern Metapopulation of Timber Rattlesnakes (Crotalus horridus). Herpetologica 72:331–342.

114

Brown, W., M. Kéry, and J. Hines. 2007. Survival of Timber Rattlesnakes (Crotalus horridus) Estimated by Capture-Recapture Models in Relation to Age, Sex, Color Morph, Time, and Birthplace. Copeia 2007:656–671.

Burgdorf, S., D. Rudolph, R. Conner, D. Saenz, and R. Schaefer. 2005. A Successful Trap Design for Capturing Large Terrestrial Snakes. Herpetological Review 36:421– 424.

Burnham, K. P., and D. R. Anderson. 2002. Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach. Second. Springer-Verlag, New York.

Bushar, L. M., N. Bhatt, M. C. Dunlop, C. Schocklin, M. A. Malloy, and H. K. Reinert. 2015. Population Isolation and Genetic Subdivision of Timber Rattlesnakes (Crotalus horridus) in the New Jersey Pine Barrens. Herpetologica 71:203–211.

Cagle, F. R. 1939. A System of Marking Turtles for Future Identification. Copeia 1939:170–173.

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.

Chalfoun, A. D., F. R. Thompson III, and M. J. Ratnaswamy. 2002. Nest Predators and Fragmentation: a Review and Meta Analysis. Conservation Biology 16:306–318.

Ciesiołkiewicz, J., G. Orłowski, and A. Elzanowski. 2006. High juvenile mortality of grass snakes Natrix natrix (L.) on a suburban road. Polish Journal of Ecology 54:465–472.

Clark, R. W., W. S. Brown, R. Stechert, and K. R. Zamudio. 2010. Roads, interrupted dispersal, and genetic diversity in timber rattlesnakes. Conservation Biology 24:1059–1069.

Clark, R. W., M. N. Marchand, B. J. Clifford, R. Stechert, and S. Stephens. 2011. Decline of an isolated timber rattlesnake (Crotalus horridus) population: Interactions between climate change, disease, and loss of genetic diversity. Biological Conservation 144:886–891.

Clevenger, A. P., B. Chruszcz, and K. E. Gunson. 2003. Spatial patterns and factors influencing small fauna road-kill aggregations. Biological Conservation 109:15–26.

Clevenger, A. P., and M. Huijser. 2011. Wildlife Crossing Structure Handbook Design and Evaluation in North America. Washington, D.C.

115

Clevenger, A. P., and N. Waltho. 2000. Factors Influencing the Effectiveness of Wildlife Underpasses in Banff National Park, , Canada. Conservation Biology 14:47– 56.

Clevenger, A. P., and A. T. Ford. 2010. Wildlife Crossing Structures, Fencing, and Other Highway Design Considerations. Pages 17–49in J. P. Beckmann, A. P. Clevenger, M. P. Huijser, and J. A. Hilty, editors. Safe Passages: Highways, Wildlife, and Habitat Connectivity. Island Press, Washington, D.C.

Colley, M., S. C. Lougheed, K. Otterbein, and J. D. Litzgus. 2017. Mitigation reduces road mortality of a threatened rattlesnake. Wildlife Research 44:48–59.

Congdon, J. D., A. Dunham, E. Sels, R. C. Van Loben, D. Congdon, and E. Dunham. 1994. Demographics of Common Snapping Turtles (Chelydra Serpentina): Implications for Conservation and Management of Long-Lived Organisms. American Zoologist 34:397–408.

Coupe, B. 2001. Arboreal Behavior in Timber Rattlesnakes (Crotalus horridus). Herpetological Review 32:83–85.

Coupe, B. 1997. Factors Affecting Movement of Radio-Tracked Timber Rattlesnakes (Crotalus horridus) in Southern Ohio. Ohio State University.

Cowles, R. B., and C. M. Bogert. 1944. A Preliminary Study of the Thermal Requirements of Lizards. Bulletin of the American Museum of Natural History 83:261–296.

Croak, B. M., D. A. Pike, J. K. Webb, and R. Shine. 2008. Three-dimensional crevice structure affects retreat site selection by reptiles. Animal Behaviour 76:1875–1884.

Croak, B. M., D. A. Pike, J. K. Webb, and R. Shine. 2010. Using artificial rocks to restore nonrenewable shelter sites in human-degraded systems: Colonization by fauna. Restoration Ecology 18:428–438.

Cunnington, G. M., E. Garrah, E. Eberhardt, and L. Fahrig. 2014. Culverts alone do not reduce road mortality in anurans. Écoscience 21:69–78.

Dunham, A. E., B. W. Grant, and K. L. Overall. 1989. Interfaces between Biophysical and Physiological Ecology and the Population Ecology of Terrestrial Vertebrate Ectotherms. Physiological Zoology 62:335–355.

Duong, T. 2007. ks: Kernel density estimation and kernel discriminant analysis for multivariate data in R. Journal Of Statistical Software 21:1–16.

116

Eberhardt, E., S. Mitchell, and L. Fahrig. 2013. Road kill hotspots do not effectively indicate mitigation locations when past road kill has depressed populations. Journal of Wildlife Management 77:1353–1359.

Enge, K. M., and K. N. Wood. 2002. A Pedestrian Road Survey of an Upland Snake Community in Florida. Southeastern Naturalist 1:365–380.

Ernst, C. H., and E. M. Ernst. 2003. Snakes of the United States and Canada. Smithsonian Press, Washington, D.C.

ESRI. 2014. ArcGIS Desktop. Environmental Systems Research Institute, Redlands, CA.

Fahrig, L., and T. Rytwinski. 2009. Effects of roads on animal abundance: an empirical review and synthesis. Ecology and Society 14:21–41.

Fahrig, L., J. H. Pedlar, S. E. Pope, P. D. Taylor, and J. F. Wegner. 1995. Effect of road traffic on amphibian density. Biological Conservation 73:177–182.

Feldhamer, G. A., J. E. Gates, D. M. Harman, A. J. Loranger, and K. R. Dixon. 1986. Effects of Interstate Highway Fencing on White-Tailed Deer Activity. The Journal of Wildlife Management 50:497–503.

Fischer, J., and D. B. Lindenmayer. 2007. Landscape modification and habitat fragmentation: a synthesis. Global Ecology and Biogeography 16:265–280.

Fitch, H. S. 2006. Ecological Succession on a Natural Area in Northeastern Kansas from 1948 to 2006. Herpetological Conservation and Biology 1:1–5.

Ford, A. T., and A. P. Clevenger. 2010. Validity of the Prey-Trap Hypothesis for Carnivore-Ungulate Interactions at Wildlife-Crossing Structures 24:1679–1685.

Forman, R. T. T., D. Sperling, J. a Bissonette, a P. Clevenger, C. D. Cutshall, V. H. Dale, L. Fahrig, R. France, C. R. Goldman, K. Heanue, J. a Jones, F. J. Swanson, T. Turrentine, and T. C. Winter. 2003. Road ecology: science and solutions. Island Press, Washington D.C.

Forman, R. T. T. 2000. Estimate of the Area Affected Ecologically by the Road System in the United States. Conservation Biology 14:31–35.

Forman, R. T. T., and L. E. Alexander. 1998. Roads and Their Major Ecological Effects. Annual Review of Ecology and Systematics 29:207–231.

Fournier, D. A., H. J. Skaug, J. Ancheta, J. Ianelli, A. Magnusson, M. N. Maunder, A. Nielsen, and J. Sibert. 2012. AD Model Builder: using automatic differentiation for

117

statistical inference of highly parameterized complex nonlinear models. Optimization Methods and Software 27:233–249.

Fridley, J. D. 2009. Downscaling climate over complex terrain: High finescale (<1000 m) spatial variation of near-ground temperatures in a montane forested landscape (). Journal of Applied Meteorology and Climatology 48:1033– 1049.

Gardner-Santana, L. C., and S. J. Beaupre. 2009. Timber Rattlesnakes (Crotalus horridus) Exhibit Elevated and Less Variable Body Temperatures during Pregnancy. Copeia 2009:363–368.

Getz, W. M., S. Fortmann-Roe, P. C. Cross, A. J. Lyons, S. J. Ryan, and C. C. Wilmers. 2007. LoCoH: Nonparameteric Kernel methods for constructing home ranges and utilization distributions. PLoS ONE 2.

Glista, D. J., T. L. DeVault, and J. A. DeWoody. 2009. A review of mitigation measures for reducing wildlife mortality on roadways. Landscape and Urban Planning 91:1–7.

Grant, B. W., A. D. Tucker, J. E. Lovich, A. M. Mills, P. M. Dixon, and J. W. Gibbons. 1992. The Use of Coverboards in Estimating Patterns of Reptile and Amphibian Biodiversity. Pages 379–403in D. R. McCullough and R. H. Barrett, editors. Wildlife 2001: Populations. London, England.

Grant, B. W., and A. E. Dunham. 1988. Thermally Imposed Time Constraints on the Activity of the Desert Lizard Sceloporus merriami. Ecology 69:167–176.

Gregory, P. T., L. H. Crampton, and K. M. Skebo. 1999. Conflicts and interactions among reproduction, thermoregulation and feeding in viviparous reptiles: are gravid snakes anorexic? Journal of Zoology 248:231–241.

Gross, I. P. 2013. Taking the road most travelled: Understanding patterns of snake (Colubridae; Storeria) movement and road mortality in a state park. Eastern Illinois University.

Gunson, K., D. Seburn, J. Kintsch, and J. Crowley. 2016. Best Management Practices for Mitigating the Effects of Roads on Amphibian and Reptile Species at Risk in Ontario.

Hall, R. J. 1994. Herpetofaunal diversity of the Four Holes Swamp, South Carolina. Page USDI/NBS Res Pub 198.

118

Hansen, M. J., and A. P. Clevenger. 2005. The influence of disturbance and habitat on the presence of non-native plant species along transport corridors. Biological Conservation 125:249–259.

Harper, K. A., E. Macdonald, P. J. Burton, J. Chen, K. D. Brosofske, S. C. Saunders, E. S. Euskirchen, D. Roberts, M. S. Jaiteh, and P.-A. Esseen. 2005. Edge Influence on Forest Structure and Composition in Fragmented Landscapes. Conservation Biology 78:356–782.

Hartmann, P. A., M. T. Hartmann, and M. Martins. 2011. Snake Road Mortality in a Protected Area in the Atlantic Forest of Southeastern Brazil. South American Journal of Herpetology 6:35–42.

Hertz, P. E. 1979. Comparative thermal biology of sympatric grass anoles (Anolis semilineatus and A. olssoni) in lowland Hispaniola (Reptilia, Lacertilia, Iguanidae). Journal of Herpetology 13:329–333.

Hijmans, R., van Etten Jacob, J. Cheng, M. Mattiuzzi, M. Sumner, J. A. Greenberg, O. P. Lamigueiro, A. Bevan, E. B. Racine, and A. Shortridge. 2016. raster: Geographic Data Analysis and Modeling.

Hódar, J. A., J. M. Pleguezuelos, and J. C. Poveda. 2000. Habitat selection of the common chameleon (Chamaeleo chamaeleon) (L.) in an area under development in southern Spain: implications for conservation. Biological Conservation 94:63–68.

Hodder, D. P., C. J. Johnson, R. V. Rea, and A. Zedrosser. 2014. Application of a Species Distribution Model to Identify and Manage Bear Den Habitat in Central , Canada. Wildlife Biology 20:238–245.

Holderegger, R., and M. Di Giulio. 2010. The genetic effects of roads: A review of empirical evidence. Basic and Applied Ecology 11:522–531.

Hothorn, T., F. Bretz, and P. Westfall. 2008. Simultaneous Inference in General Parametric Models. Biometrical Journal 50:346–363.

Huey, R. B. 1982. Temperature, Physiology, and the Ecology of Reptiles. Pages 25–91in C. Gans and F. H. Pough, editors. Biology of the Reptilia Volume 12: Physiology. Academic Press, New York.

Huey, R. B., C. R. Peterson, S. J. Arnold, and W. P. Porter. 1989. Hot rocks and not-so- hot rocks: retreat-site selection by garter snakes and its thermal consequences.

Huey, R. B., and M. Slatkin. 1976. Cost and Benefits of Lizard Thermoregulation. The Quarterly Review of Biology 51:363–384.

119

Huey, R. B. 1991. Physiological Consequences of Habitat Selection. The American Naturalist 137:S91–S115.

Huey, R. B., and J. G. Kingsolver. 1989. Evolution of thermal sensitivity of ectotherm performance. Trends in Ecology and Evolution 4:131–135.

Hulme, P. E. 2009. Trade, transport and trouble: Managing invasive species pathways in an era of globalization. Journal of Applied Ecology 46:10–18.

Jackson, S. D., T. A. Langen, D. M. Marsh, and K. M. Andrews. 2015. Natural History and Physiological Characteristics of Small Animals in Relation to Roads. Pages 21– 41in K. M. Andrews, P. Nanjappa, and S. P. D. Riley, editors. Roads and Ecological Infrastructure: Concepts and Applications for Small Animals. Johns Hopkins University Press, Baltimore, MD.

Jaeger, J. A. G., J. Bowman, J. Brennan, L. Fahrig, D. Bert, J. Bouchard, N. Charbonneau, K. Frank, B. Gruber, and K. T. Von Toschanowitz. 2005. Predicting when animal populations are at risk from roads: An interactive model of road avoidance behavior. Ecological Modelling 2005:329–348.

Jaeger, J. a G., and L. Fahrig. 2004. Effects of Road Fencing on Population Persistence. Conservation Biology 18:1651–1657.

Jäggi, C., and B. Baur. 1999. Overgrowing forest as a possible cause for the local extinction of Vipera aspis in the northern Swiss Jura mountains. Amphibia-Reptilia 20:25–34.

Jochimsen, D. M., C. R. Peterson, K. M. Andrews, and J. W. Gibbons. 2004. A Literature Review of the Effects of Roads on Amphibians and Reptiles and the Measures Used to Minimize Those Effects.

Jochimsen, D. M., C. R. Peterson, and L. J. Harmon. 2014. Influence of ecology and landscape on snake road mortality in a sagebrush-steppe ecosystem. Animal Conservation 17:583–592.

Johnson, D. H. 1980. The Comparison of Usage and Availability Measurements for Evaluating Resource Preference. Ecology 61:65–71.

Jones, M. E. 2000. Road upgrade, road mortality and remedial measures: impacts on a population of eastern quolls and Tasmanian devils. Wildlife Research 27:289–296.

Jönsson, K. I. 1997. Capital and Income Breeding as Alternative Tactics of Resource Use in Reproduction. Oikos 78:57–66.

120

Kareiva, P. M., and N. Shigesada. 1983. Analyzing insect movement as a correlated random walk. Oecologia 56:234–238.

Karraker, N. E., J. P. Gibbs, and J. R. Vonesh. 2008. Impacts of road deicing salt on the demography of vernal pool-breeding amphibians. Ecological Applications 18:724– 734.

Kearney, M., R. Shine, and W. P. Porter. 2009. The potential for behavioral thermoregulation to buffer “cold-blooded” animals against climate warming. Proceedings of the National Academy of Sciences of the United States of America 106:3835–3840.

Kenneth Dodd, C., W. J. Barichivich, and L. L. Smith. 2004. Effectiveness of a barrier wall and culverts in reducing wildlife mortality on a heavily traveled highway in Florida. Biological Conservation 118:619–631.

Klingenböck, A., K. Osterwalder, and R. Shine. 2000. Habitat use and thermal biology of the “Land Mullet” Egernia major, a large scincid lizard from remnant rain forest in southeastern Australia. Copeia 2000:931–939.

Kovar, R., M. Brabec, R. Vita, and R. Bocek. 2014. Mortality Rate and Activity Patterns of an Aesculapian Snake (Zamenis longissimus) Population Divided by a Busy Road. Journal of Herpetology 48:24–33.

Langen, T. A., K. M. Andrews, S. P. Brady, N. E. Karraker, and D. J. Smith. 2015. Road Effects on Habitat Quality for Small Animals. Pages 57–93in K. M. Andrews, P. Nanjappa, and S. P. D. Riley, editors. Roads & Ecological Infrastructure: Concepts and Applications for Small Animals. Johns Hopkins University Press, Baltimore, MD.

Langen, T. A., K. E. Gunson, C. A. Scheiner, and J. T. Boulerice. 2012. Road mortality in freshwater turtles: Identifying causes of spatial patterns to optimize road planning and mitigation. Biodiversity and Conservation 21:3017–3034.

Langley, W. M., H. W. Lipps, and J. F. Theis. 1989. Responses of Kansas Motorists to Snake Models on a Rural Highway. Transactions of the Kansas Academy of Science (1903-) 92:43.

Laurance, W. F., G. R. Clements, S. Sloan, C. S. O’Connell, N. D. Mueller, M. Goosem, O. Venter, D. P. Edwards, B. Phalan, A. Balmford, R. Van Der Ree, and I. B. Arrea. 2014. A global strategy for road building. Nature 513:229–232.

Levin, S. A. 1992. The Problem of Pattern and Scale in Ecology. Ecology 73:1943–1967.

121

Little, S. J., R. G. Harcourt, and A. P. Clevenger. 2002. Do wildlife passages act as prey- traps? Biological Conservation 107:135–145.

Longcore, T., and C. Rich. 2004. Ecological light pollution. Frontiers in Ecology and the Environment 2:191–198.

Lourdais, O., R. Shine, X. Bonnet, M. Guillon, and G. Naulleau. 2004. Climate affects embrionic development in a viviparous snake, Vipera aspis. OIKOS 104:551–560.

Lutterschmidt, W. I., and H. K. Reinert. 2012. Modeling body temperature and thermal inertia of large-bodied reptiles: Support for water-filled biophysical models in radiotelemetric studies. Journal of Thermal Biology 37:282–285.

Lyons, A. J., W. C. Turner, and W. M. Getz. 2013. Home range plus: a space-time characterization of movement over real landscapes. Movement ecology 1:1–14.

MacGowan, B. J., A. F. T. Currylow, and J. E. MacNeil. 2017. Short-term responses of Timber Rattlesnakes (Crotalus horridus) to even-aged timber harvests in . Forest Ecology and Management 387:30–36.

Manly, B. F. J., L. L. McDonald, D. L. Thomas, T. L. McDonald, and W. P. Erickson. 2002. Resource Selection by Animals: Statistical design and analysis for field studies. 2nd edition. Kluwer Academic Publishers, Dordrecht, Netherlands.

Martin, W. H. 1993. Reproduction of the Timber Rattlesnake (Crotalus horridus) in the . Journal of Herpetology 27:133–143.

Martin, W. H. 2002. Life history constraints on the timber rattlesnake (Crotalus horridus) at its climatic limits. Pages 285–306in G. W. Schuett, M. Hoggren, M. E. Douglas, and H. W. Greene, editors. Biology of the Vipers. Mountain Pub Lc, Eagle Mountain, UT.

Mata, C., I. Hervás, J. Herranz, F. Suárez, and J. E. Malo. 2008. Are motorway wildlife passages worth building? Vertebrate use of road-crossing structures on a Spanish motorway. Journal of Environmental Management 88:407–415.

McCallum, M. L. 2007. Amphibian Decline or Extinction? Current Declines Dwarf Background Extinction Rate. Journal of Herpetology 41:483–491.

McClure, C. J. W., H. E. Ware, J. Carlisle, G. Kaltenecker, J. R. Barber, and P. R. S. B. 2013. An experimental investigation into the effects of traffic noise on distributions of birds: avoiding the phantom road. Proceedings of the Royal Society B: Biological Sciences 280:20132290.

122

McKinney, M. L. 2006. Urbanization as a major cause of biotic homogenization. Biological Conservation 127:247–260.

McLellan, B. N., and D. M. Shackleton. 1988. Grizzly Bears and Resource-Extraction Industries: Effects of Roads on Behaviour, Habitat Use and Demography. Journal of Applied Ecology 25:451–460.

Millspaugh, J. J., D. C. Kesler, R. W. Kays, R. A. Gitzen, J. H. Schulz, C. T. Rota, C. M. Bodinof, J. L. Belant, and B. J. Keller. 2012. Analysis of Radiotelemetry Data. Pages 258–283in N. J. Silvy, editor. The wildlife techniques manual 1. Johns Hopkins University Press, Baltimore.

Mortensen, D. A., E. S. J. Rauschert, A. N. Nord, and B. P. Jones. 2009. Forest Roads Facilitate the Spread of Invasive Plants. Invasive Plant Science and Management 2:191–199.

Myers, C. W. 1957. Amphibians and reptiles of Washington State Park, Washington County, Missouri. Transactions of the Kansas Academy of Science 60:288–293.

O’Donnell, R. P., and S. J. Arnold. 2005. Evidence for Selection on Thermoregulation: Effects of Temperature on Embryo Mortality in the Thamnophis elegans. Copeia 2005:930–934.

Olson, Z. H., B. J. MacGowan, M. T. Hamilton, A. F. T. Currylow, and R. N. Williams. 2015. Survival of Timber Rattlesnakes (Crotalus horridus): Investigating Individual, Environmental, and Ecological Effects. Herpetologica 71:274–279.

Packard, G. C., C. R. Tracy, and J. J. Roth. 1977. The Physiological Ecology of Reptilian Eggs and Embryos, and the Evolution of Viviparity Within the Class Reptilia. Biological Reviews 52:71–105.

Pagnucco, K. S., C. A. Paszkowski, and G. J. Scrimgeour. 2011. Using Cameras to Monitor Tunnel Use by Long-Toed Salamanders (Ambystoma macrodactylum): An Informative, Cost-Efficient Technique. Herpetological Conservation and Biology 6:277–286.

Palmer, T. 2004. Landscape with Reptile. The Lyons Press, Guilford, Connecticut.

Patrick, D. A., C. M. Schalk, J. P. Gibbs, and H. W. Woltz. 2010. Effective Culvert Placement and Design to Facilitate Passage of Amphibians across Roads. Copeia 44:618–626.

123

Patrick, D. A., J. P. Gibbs, V. D. Popescu, and D. A. Nelson. 2012. Multi-scale habitat- resistance models for predicting road mortality “hotspots” for turtles and amphibians. Herpetological Conservation and Biology 7:407–426.

Peterman, W. E., G. M. Connette, R. D. Semlitsch, and L. S. Eggert. 2014. Ecological resistance surfaces predict fine-scale genetic differentiation in a terrestrial woodland salamander. Molecular Ecology 23:2402–2413.

Peterson, C. R., A. R. Gibson, and M. E. Dorcas. 1993. Snake Thermal Ecology: The Causes and Consequences of Body-Temperature Variation. Pages 241–314 in R. A. Seigel and J. T. Collins, editors. Snakes: Ecology and Behavior. The Blackburn Press, Caldwell, New Jersey.

Pike, D. A., J. K. Webb, and R. Shine. 2011. Removing forest canopy cover restores a reptile assemblage. Ecological Applications 21:274–280.

Pinheiro, J., D. Bates, S. DebRoy, D. Sarkar, and R. C. Team. 2016. nlme: Linear and Nonlinear Mixed Effects Models.

Pinheiro, J. C., and D. M. Bates. 2000. Mixed effects models in S and S-Plus. Springer Verlag, New York.

Quantum GIS Development Team. 2017. Quantum GIS Geographic Information System. Open Source Geospatial Foundation Project.

R Core Team. 2016. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Reinert, H. K., D. Cundall, and L. M. Bushar. 1984. Foraging Behavior of the Timber Rattlesnake, Crotalus horridus. Copeia 1984:976–981.

Reinert, H. K., G. A. MacGregor, M. Esch, L. M. Bushar, and R. T. Zappalorti. 2011. Foraging Ecology of Timber Rattlesnakes, Crotalus horridus. Copeia 2011:430– 442.

Reinert, H. K., W. F. Munroe, C. E. Brennan, M. N. Rach, S. Pelesky, and L. M. Bushar. 2011. Response of timber rattlesnakes to commercial logging operations. The Journal of Wildlife Management 75:19–29.

Reinert, H. K., and R. T. Zappalorti. 1998. Timber Rattlesnakes (Crotalus horridus) of the Pine Barrens: Their Movement Patterns and Habitat Preference. Copeia 1988:964–978.

124

Reinert, H. K. 1993. Habitat Selection in Snakes. Pages 201–240in R. A. Seigel and J. T. Collins, editors. Snakes: Ecology and Behavior. The Blackburn Press, Caldwell, New Jersey.

Reinert, H. K. 1984a. Habitat Variation Within Sympatric Snake Populations. Ecology 65:1673–1682.

Reinert, H. K. 1984b. Habitat Separation Between Sympatric Snake Populations. Ecology 65:478–486.

Reinert, H. K., and D. Cundall. 1982. An Improved Surgical Implantation Method for Radio-Tracking Snakes. Copeia 1982:702–705.

Richardson, M. L., P. J. Weatherhead, and J. D. Brawn. 2006. Habitat use and activity of prairie (Lampropeltis calligaster calligaster) in Illinois. Journal of Herpetology 40:423–428.

Roe, J. H., J. Gibson, and B. a. Kingsbury. 2006. Beyond the wetland border: Estimating the impact of roads for two species of water snakes. Biological Conservation 130:161–168.

Rouse, J. D., R. J. Willson, R. Black, and R. J. Brooks. 2011. Movement and Spatial Dispersion of Sistrurus catenatus and Heterodon platirhinos: Implications for Interactions with Roads. Copeia 2011:443–456.

Row, J. R., G. Blouin-Demers, and P. J. Weatherhead. 2007. Demographic effects of road mortality in black ratsnakes (Elaphe obsoleta). Biological Conservation 137:117– 124.

Row, J. R., and G. Blouin-Demers. 2006. Kernels Are Not Accurate Estimators of Home- range Size for Herpetofauna. Copeia 2006:797–802.

Rudolph, D. C., S. J. Burgdorf, R. N. Conner, and R. R. Schaefer. 1999. Preliminary evaluation of the impact of roads and associated vehicular traffic on snake populations in eastern Texas. Pages 129–136Proceedings of the International Conference on Wildlife Ecology and Transportation. Florida Department of Transportation, Tallahassee, Florida. Fl-ER-73-99.

Rudolph, D. C., R. R. Schaefer, D. Saenz, and R. N. Conner. 2004. Arboreal behavior in the Timber Rattlesnake, Crotalus horridus, in eastern Texas. Texas Journal of Science 56:395–404.

125

Rudolph, D. C., and S. J. Burgdorf. 1997. Timber Rattlesnakes and Pine Snakes of the West Gulf Coastal Plain: Hypotheses of Decline. The Texas Journal of Science 49:Supplement: 111-122.

Rytwinski, T., and L. Fahrig. 2012. Do species life history traits explain population responses to roads? A meta-analysis. Biological Conservation 147:87–98.

Rytwinski, T., R. van der Ree, G. M. Cunnington, L. Fahrig, C. S. Findlay, J. Houlahan, J. A. G. Jaeger, K. Soanes, and E. A. van der Grift. 2015. Experimental study designs to improve the evaluation of road mitigation measures for wildlife. Journal of Environmental Management 154:48–64.

Sartorius, S. S., L. J. Vitt, and G. R. Colli. 1999. Use of naturally and anthropogenically disturbed habitats in Amazonian rainforest by the teiid lizard Ameiva ameiva. Biological Conservation 90:91–101.

Sawaya, M. a, S. T. Kalinowski, and A. P. Clevenger. 2014. Genetic connectivity for two bear species at wildlife crossing structures in Banff National Park. Proceedings. Biological sciences / The Royal Society 281:20131705.

Sawaya, M. a., A. P. Clevenger, and S. T. Kalinowski. 2013. Demographic connectivity for ursid populations at wildlife crossing structures in Banff National Park. Conservation Biology 27:721–730.

Schoener, T. W. 1971. Resource Partitioning in Ecological Communities. Science 185:27–39.

Sealy, J. B. 2002. Ecology and behavior of the timber rattlesnake (Crotalus horridus) in the upper Piedmont of North Carolina: identified threats and conservation recommendations. Pages 561–578in G. W. Schuett, M. Hoggren, M. E. Douglas, and H. W. Greene, editors. Biology of the Vipers. Eagle Mountain Pub Lc, Eagle Mountain, UT.

Sears, M. W., M. J. Angilletta, M. S. Schuler, J. Borchert, K. F. Dilliplane, M. Stegman, T. W. Rusch, and W. A. Mitchell. 2016. Configuration of the thermal landscape determines thermoregulatory performance of ectotherms. Proceedings of the National Academy of Sciences of the United States of America 113:10595–10600.

Sears, M. W., E. Raskin, and M. J. Angilletta. 2011. The World Is not Flat: Defining Relevant Thermal Landscapes in the Context of Climate Change. Integrative and Comparative Biology 51:666–675.

Seigel, R. A., M. M. Huggins, and N. B. Ford. 1987. Reduction in locomotor ability as a cost of reproduction in gravid snakes. Oecologia 73:481–485.

126

Seigel, R. A., and M. A. Pilgrim. 2002. Long-term Changes in Movement Patterns of Massasaugas (Sistrurus catenatus). Pages 405–412in G. W. Schuett, M. Hoggren, M. E. Douglas, and H. W. Greene, editors. Biology of the Vipers. Eagle Mountain Pub Lc, Eagle Mountain, UT.

Shannon, G., M. F. McKenna, L. M. Angeloni, K. R. Crooks, K. M. Fristrup, E. Brown, K. A. Warner, M. D. Nelson, C. White, J. Briggs, S. McFarland, and G. Wittemyer. 2016. A synthesis of two decades of research documenting the effects of noise on wildlife. Biological Reviews 91:982–1005.

Shepard, D. B., A. R. Kuhns, M. J. Dreslik, and C. A. Phillips. 2008a. Roads as barriers to animal movement in fragmented landscapes. Animal Conservation 11:288–296.

Shepard, D. B., M. J. Dreslik, B. C. Jellen, and C. A. Phillips. 2008b. Reptile Road Mortality around an Oasis in the Illinois Corn Desert with Emphasis on the Endangered Eastern Massasauga. Copeia 2008:350–359.

Shine, R., and P. Harlow. 1993. Maternal Thermoregulation Influences Offspring Viability in a Viviparous Lizard. Oecologia 96:122–127.

Shine, R., E. G. Barrott, and M. J. Elphick. 2002. Some like It Hot: Effects of Forest Clearing on Nest Temperatures of Montane Reptiles. Ecology 83:2808–2815.

Shoemaker, K. T., and J. P. Gibbs. 2010. Evaluating Basking-Habitat Deficiency in the Threatened Eastern Massasauga Rattlesnake. Journal of Wildlife Management 74:504–513.

Skaug, H., D. Fournier, B. Bolker, M. A, and A. Nielsen. 2016. Generalized Linear Mixed Models using “AD Model Builder.”

Smith, L. L., and C. K. Dodd. 2003. Wildlife mortality on US highway 441 across Paynes Prairie, Alachua County, Florida. Florida Scientist 66:128–140.

Steen, D. A., and J. P. Gibbs. 2004. Effects of Roads on the Structure of Freshwater Turtle Populations. Conservation Biology 18:1143–1148.

Steen, D. A. 2010. Snakes in the grass: Secretive natural histories defy both conventional and progressive statistics. Herpetological Conservation and Biology 5:183–188.

Steen, D. A., M. J. Aresco, S. G. Beilke, B. W. Compton, E. P. Condon, C. K. Dodd, H. Forrester, J. W. Gibbons, J. L. Greene, G. Johnson, T. A. Langen, M. J. Oldham, D. N. Oxier, R. A. Saumure, F. W. Schueler, J. M. Sleeman, L. L. Smith, J. K. Tucker, and J. P. Gibbs. 2006. Relative vulnerability of female turtles to road mortality. Animal Conservation 9:269–273.

127

Stuart, S. N., J. S. Chanson, N. A. Cox, B. E. Young, A. S. L. Rodrigues, D. L. Fischman, and R. W. Waller. 2004. Status and trends of amphibian declines and extinctions worldwide. Science 306:1783–1786.

Sullivan, B. K. 1981. Observed Differences in Body Temperature and Associated Behavior of Four Snake Species. Journal of Herpetology 15:245–246.

Tilman, D., R. M. May, C. L. Lehman, and M. A. Nowak. 1994. and the extinction debt. Nature 371:65–66.

Trombulak, S. C., and C. A. Frissell. 2000. Review of Ecological Effects of Roads on Terrestrial and Aquatic Communities. Conservation Biology 14:18–30. van der Ree, R., J. A. G. Jaeger, T. Rytwinski, and E. A. Van Der Grift. 2015. Good Science and Experimentation are Needed in Road Ecology. Pages 71–81in R. van der Ree, D. J. Smith, and C. Grilo, editors. Handbook of Road Ecology. Wiley- Blackwell, Oxford, UK. van der Ree, R., D. J. Smith, and C. Grilo. 2015. Handbook of Road Ecology. Wiley- Blackwell, Oxford, UK.

Vitt, L. J., T. C. S. Avila-Pires, J. P. Caldwell, and R. L. Veronica. 1998. The Impact of Individual Tree Harvesting on Thermal Environments of Lizards in Amazonian Rain Forest. Conservation Biology 12:654–664. von der Lippe, M., and I. Kowarik. 2007. Long-distance dispersal of plants by vehicles as a driver of plant invasions. Conservation Biology 21:986–996.

Waldron, J. L., J. D. Lanham, and S. H. Bennett. 2006. Using Behaviorally-Based Seasons to Investigate Canebreak Rattlesnake (Crotalus horridus) Movement Patterns and Habitat Selection. Herpetologica 62:389–398.

Webb, J. K., R. Shine, and R. M. Pringle. 2005. Canopy Removal Restores Habitat Quality for an Endangered Snake in a Fire Suppressed Landscape. Copeia 2005:894–900.

Wiens, J. A. 1989. Spatial scaling in ecology. Functional Ecology 3:385–397.

Wills, C. A., and S. J. Beaupre. 2000. An Application of Randomization for Detecting Evidence of Thermoregulation in Timber Rattlesnakes (Crotalus horridus) from Northwest Arkansas. Physiological and Biochemical Zoology 73:325–334.

128

Woltz, H. W., J. P. Gibbs, and P. K. Ducey. 2008. Road crossing structures for amphibians and reptiles: Informing design through behavioral analysis. Biological Conservation 141:2745–2750.

Worton, B. 1989. Kernel Methods for Estimating the Utilization Distribution in Home- Range. Ecology 70:164–168.

Yanes, M., J. M. Velasco, and F. Suarez. 1995. Permeability of roads and railways to vertebrates: The importance of culverts. Biological Conservation 71:217–222.

129

APPENDIX 1: OPERATIVE TEMPERATURE MODEL DESIGN

Models were made of copper tubing (Figure A1; length 15.2 cm, diameter 3.8 cm, thickness 0.15 cm; similar to Wills & Beaupre 2000), which is a standard material for operative temperature models (OTMs) because of its high conductivity and affordability.

We measured the diameter of snake tubes used to handle gravid females to estimate an appropriate diameter for OTMs. Previous work has shown that increasing operative model length beyond 15 cm has diminishing returns in accuracy for small snake models

(Peterson et al. 1993), ostensibly because a snake’s cylindrical morphology and high surface–to-volume scaling provides a large area for rapid and even heat exchange.

Models contained a Thermochron ® iButton (DS1921G ± 0.5 °C resolution) temperature data logger suspended within the model using a frame of folded hardware cloth. Models were sealed on both ends using rubber-stoppers.

Models were painted to approximate the reflectivity of C. horridus. Estimating the

“correct” reflectivity of C. horridus is challenged by the fact that the species varies greatly in dorsal ground coloration among individuals within a population, displaying banding patterns, color polymorphism (black phase and yellow phase individuals), and considerable color variation within each morph. Consequently, there is no single

“correct” color to paint OTMs for this species. To paint models appropriate colors, I measured the reflectivity of C. horridus dorsal coloration using an Ocean Optics Jaz

Spectrometer ®. I used a live snake from our population (light yellow phase), as well as three dead C. horridus that were donated by a colleague (two yellow phase ranging from yellow to brown; one black morph that was a dark gray; all from Ohio populations). Dead

130 snakes were preserved by freezing, and thus coloration was not faded. Two of these snakes died of natural causes, and the third was a road mortality; snakes were not harmed for the purposes of this study. After thawing, I sampled reflectivity curves at ten points within both the ground color and band locations for each specimen. I took the dead specimens to a local hardware store where they were scanned to produce matching paint samples. I then sampled the reflectivity of the paints using the spectrometer, and selected the paints that best approximated ground and band reflectivity curves. I selected a representative ground color for each yellow and black morph, and a common color for cross bands. Models were hand-painted to approximate the banding pattern of C. horridus

(Figure A1).

131

Figure A1.1: Operative Temperature Model Design. Models were produced for both yellow and black color morphs. The models are pictured with a shed skin of an adult C. horridus, an iButton, a hardware-cloth frame, and meter tape.

Models recorded temperature once every 20 mins (approximately once every two time constants as determined by laboratory heating and cooling trials, methods adapted from Wills & Beaupre 2000). I deployed 55 OTMs of each color morph (N = 110). Sites were selected based on rattlesnake relocations within each habitat type, along with paired models that were deployed using random walks from those relocations so that ecologically relevant variation was sampled within the habtiats. Randon model locations were referenced to known activity locations, and displaced a random distance sampled

132 from the interquartile range of movement distances (~10 - 64 m) and at a random bearing.

Thus OTM locations were generally cronstrained to a population level minimum convex polygon. Sampling effort among habitats was roughly equal (edge = 35 models, forest =

36 models, ROW = 39 models). All models were deployed at surface activity locations regardless of vegetative cover. Retreat crevices used as gestation sites were difficult to access (often on cliff faces), and we did not wish to deter use of these sites by plugging them with models. We also do not know exactly how far snakes venture into subsurface retreats, making it difficult to accurately capture the range of conditions snakes can experience in refuges. Additionaly, the thermal buffering qualities of subsurface retreats has been well studied (Huey et al. 1989). My primary goal was also to compare thermal limitations, and thus it seemed most appropriate to characterize surface temperatures where the most extreme thermal environments would be experienced.

We evaluated model accuracy in the field by placing models next to telemetered snakes after measuring Tb, and recording the temperature after two time constants (n = 26 calibration measurements). Major-axis regression demonstrated a strong relationship

2 between model Te and Tb of free ranging snakes (R = .89), and did not reveal a significant deviation from a slope of 1 (95% CI: 0.90-1.21) and an intercept of 0 (95%

CI: -7.03 - 0.54), suggesting that models accurately modeled field body temperatures; however, the intercept was trending towards significance. The same analysis performed on model surface temperatures versus snake surface temperatures revealed a higher level of fit (R2 = 0.93, intercept 95% CI: -0.87 - 0.99, slope 95% CI: 0.88 - 1.12), indicating that OTMs were converging on similar surface temperatures as snakes, and observed

133 differences in internal temperatures were likely the result of higher heat capacity in live snakes, which is expected when using OTMs. This was also reflected when comparing Te and Tb data of a male rattlesnake inhabiting the forest. We implanted a male rattlesnake with an iButton during transmitter replacement surgery, and after recovering the iButton in Spring 2017, we plotted average hourly Tb against average hourly Te for the month of

August. While models accurately approximated average temperatures experience by the rattlesnake, the snake had greater thermal inertia, leading to heating and cooling curves being offset by 1-2 hours (Figure A2). Otherwise, Te models accurately captured the range of observed Tb (average difference between mean daily minimum and maximum temperatures achieved were < 1 ˚C, Figure A2). This should not be viewed as problematic. In theory, operative temperature models should respond to the environment faster than an actual organism, providing an instantaneous measure of conditions that would be experienced by the organism (Angilletta 2009). Bakken (1992) also argued that models should respond instantaneously, as animals are often capable of peripheral thermoreception and likely respond to environmental conditions more rapidly than changes in core body temperatures might predict. Surprisingly, neither mean Te nor average daily maximum Te differed significantly between model color morphs over the course of our study (Mean Te: t = -0.48, df = 108, P = 0.63; Max Te: t = 0.14665, df =

108, P = 0.8837), and did not change when controlling for habitat type (Mean Te: t =

0.99, df = 106, P = 0.32; Max Te: t = -0.097, df = 106, P = 0.923); thus, models of both morphs were pooled in analyses.

134

Figure A1.2: August Te and Tb temperature profiles. The male rattlesnake was implanted with a temperature logging iButton in late July during transmitter replacement. This individual spent most of the August within the forest habitat. Average daily minimum and maximum body temperatures were within 1˚C of those observed measured by forest Te models. Heating and cooling rates appeared to be approximately equal between the snake and Te models, though offset by 1 – 2 hours indicating that snakes had greater thermal inertia when shifting between heating and cooling phases.

Mass is an important property that affects Tb of large bodied organisms that have internal temperature gradients and thermal inertia. Models can be filled with fluids or other substances to account for this inertia (Lutterschmidt & Reinert 2012), effectively adding thermal capacitance to models. The primary advantage of thermal inertia models is to achieve closer estimates of actual Tb that would be experienced in real time by accounting for the animal’s thermal history. However, doing so comes with tradeoffs.

135

Fluid-filled models significantly complicate model design and field implementation, especially when models are widely distributed and contain electronic data loggers

(Bakken & Angilletta 2014). If models leak, heating and cooling rates will deviate over time, incurring irreparable errors. Equally as concerning, data from thermal inertia models are not comparable to studies reporting Te from standard OTMs, and have been shown to be ineffective for capturing extreme environmental conditions experienced by an organism (Bakken & Angilletta 2014). In contrast, OTMs quantify the thermal environment as it would be experienced by the organism, and reflect the Tb that would eventually be achieved under prevailing conditions. This will generally differ from instantaneously realized Tb because diel patterns of heating and cooling ensure that thermal environments are in a constant state of flux, and all else being equal, virtually any difference in heat capacitance between the organism and the OTM will yield differences between instantaneous Te and Tb. For these reasons, it seemed appropriate to use standard OTMs and avoid thermal inertia models.

136

APPENDIX 2: THERMAL MODELING AND RSF SURFACES

Various spatial data layers were used to model thermal landscapes and resource selection functions at the NVBP (Figure A3). These included canopy cover and edgeness

(canopy heterogeneity), which were evaluated at multiple spatial scales (Figures A4, A5).

Solar insolation was derived for the summer 2016 field season (Figure A6). Canopy cover, insolation, and air temperature data were used to predict OTM data, and derive thermal landscape surfaces (Figures A7, A8). Thermal landscapes, constituent layers, and habitat edgeness were then used to model gravid female RSF models (Figure A9).

Figure A2.1: Reference imagery for the NVBP study site.

137

Figure A2.2: Percent canopy cover layers. Left: fine scale canopy cover derived by taking the mean value of 3x3 window of 2.5 m cells. Right: coarser scale canopy cover derived from a 10 m radius circle window.

Figure A2.3: Edgeness (canopy heterogeneity). This layer was derived by taking the standard deviation of a fine scale canopy cover layer using a 10 m radius focal window.

138

Figure A2.4: Insolation. Insolation was derived using the Solar Radiation Area tool in ArcGIS for the summer of 2016, which was then divided by the total number of days and converted to kilowatts to show insolation in units of kWh/day.

139

Figure A2.5: TeAVG surface. TeAVG was derived from average air temperature, a 25 m radius canopy cover layer, and insolation using a linear mixed effects model. The model was project with air temperature set to 24˚C, which was the mean daily average air temperature throughout the summer. The model had a moderately high level of fit: marginal R2 = 0.71.

140

Figure A2.6: TeMAX surface. TeMAX was derived from maximum air temperature, a 3x3 cell window canopy cover layer, and insolation using a linear mixed effects model. The model was projected with air temperature set to 31˚C, which was the mean daily maximum air temperature. The model had a moderate level of fit: marginal R2 = 0.61.

141

Figure A2.7: Projected TeAVG + Edge RSF model. Top: The RSF model was binned into 20th percentile ranges of fitted values, where 1 = lowest suitability and ranges up to the 0th to 20th percentile, and 5 = highest suitability and is the 80th percentile and greater. These ranges differ from the 25th percentile ranges reported in Chapter Two but illustrate the same methods. Bottom: count of gravid female relocations within each suitability class (73% in suitability class 5).

! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !

! ! Thesis and Dissertation Services