Ecology and Conservation Biology of the North American Wood Turtle (Glyptemys

insculpta) in the Central Appalachians

A dissertation presented to

the faculty of

the College of Arts and Sciences of Ohio University

In partial fulfillment

of the requirements for the degree

Doctor of Philosophy

Steven P. Krichbaum

May 2018

© 2018 Steven P. Krichbaum. All Rights Reserved. 2 This dissertation titled

Ecology and Conservation Biology of the North American Wood Turtle (Glyptemys

insculpta) in the Central Appalachians

by

STEVEN P. KRICHBAUM

has been approved for

the Department of Biological Sciences

and the College of Arts and Sciences by

Willem Roosenburg

Professor of Biological Sciences

Robert Frank

Dean, College of Arts and Sciences 3 Abstract

KRICHBAUM, STEVEN P., Ph.D., May 2018, Biological Sciences

Ecology and Conservation Biology of the North American Wood Turtle (Glyptemys insculpta) in the Central Appalachians

Director of Dissertation: Willem Roosenburg

My study presents information on summer use of terrestrial habitat by IUCN

“endangered” North American Wood Turtles (Glyptemys insculpta), sampled over four years at two forested montane sites on the southern periphery of the species’ range in the central Appalachians of Virginia (VA) and West Virginia (WV) USA.

The two sites differ in topography, stream size, elevation, and forest composition and structure. I obtained location points for individual turtles during the summer, the period of their most extensive terrestrial roaming. Structural, compositional, and topographical habitat features were measured, counted, or characterized on the ground (e.g., number of canopy trees and identification of herbaceous taxa present) at Wood Turtle locations as well as at paired random points located 23-300m away from each particular turtle location.

First, I report and discuss basic morphometric and activity area data of the

VA and WV turtles. Chapter two uses a nine-year dataset of adult WV Wood

Turtles to estimate population size, population growth rate (lambda), and survivorship with open population Cormack-Jolly-Seber and Pradel models in program MARK. My third chapter assess Wood Turtle thermal ecology by examining three data sets of environmental and turtle temperatures: 1) temperatures 4 in three different microhabitat types (unshaded by ground cover [exposed], under vegetation [UV], under litter [UL]) recorded by iButtons at arrays throughout the two study sites; 2) ground temperatures at the locations of radio-tracked individuals and their paired random points measured within 300 meters and 30 minutes of each other; 3) body temperatures estimated with iButtons attached to the shell bridges of adult Wood Turtles. In the fourth chapter, I examine highly localized conditions resulting from short-term weather patterns and fine-scale microhabitat characteristics by comparing ground-level relative humidity at the locations of radio-tracked Wood Turtles to those at paired random points. I use the GIS-based water balance model developed by Dr. James Dyer to examine landscape conditions (such as water deficit [“DEF”] and actual evapotranspiration [“AET”]) resulting from long-term climate patterns and broad-scale habitat conditions (e.g., topographical aspect and soil types).

The final two chapters are the heart of my dissertation. Vegetation was identified, measured, counted, or characterized in plots at 640 locations (394 in

VA, 246 in WV), evenly distributed between adult turtle and random points.

Importance values for overstory trees ≥ 10cm dbh were calculated in 400m2 plots; herbaceous plant taxa were identified in 400m2 and 1m2 plots; woody seedling taxa were identified in 1 m2 plots; forest types were specified at the 400m2 plot and stand (5-20ha) scales. I used the R program “indicspecies”, paired logistic regression, and classification and regression trees (CART) to analyse these data.

Over thirty herbaceous and woody seedling taxa were indicators for Wood Turtle 5 presence at the 400m2 and/or 1m2 scales at the VA and WV study sites. I used a series of conditional logistic regressions to quantify habitat use of Wood Turtles at multiple scales across a range of different forest types. At each of the turtle and random points proportions of ground cover were visually estimated within 1m2 plots to assess microhabitat use; structural, compositional, and topographical habitat features were measured in 400m2 circular plots to capture meso-scale ecological data; and stand scale (5-20ha) designations of forest type and seral stage were used to assess macro-scale habitat use. I found that Wood Turtles showed a preference for specific environmental conditions: older forest sites with relatively more herbaceous ground cover, large woody debris, canopy openness, and turtle- level obscurity, and with gentler slopes and warmer aspects.

6 Dedication

For my family, especially my parents, Donald William and Mary Mihailoff

Krichbaum, and my friends, especially Sherman Bamford, Shay and Kim Clanton,

Lloyd Clayton, Jacques and Ulysses Desportes, Nancy Eckel-Dickenson, Bob Fener,

Joe and Jackie Glisson, Joe and Laura Hazelbaker, Allan Hench, Henry C.

Familiarus, Jack Hutchinson, Mike Jones, Linda Lee and Andy Mahler, Ernie and Sue

Reed, Jody Schaub, Andrew Sterrett, Dwight Worker, and Christina Wulf.

.

7 Acknowledgments

This dissertation was possible thanks to many people.

The usual suspects: Dr. Willem Roosenburg of Ohio University for consultation and patience; Dr. Thomas Akre of Smithsonian Conservation Center for equipment and inspiration; Dr. James Dyer of Ohio University for GIS assistance and chocolate; Dr. Viorel Popescue of Ohio University for advice and statistical assistance; my committee – Dr. Scott Moody, Dr. Matt White, and Dr.

Don Miles.

The OU Biological Sciences Department faculty, students, and staff for support, especially Cindy Meyer and Karen Keesey.

Dr. Robert Hunzicker (Naturalist Extraordinaire), Dr. Robert Mueller

(Geologist Extraordinaire), Kyle Elza, Preston Sheaks, and Heroic Henry the Best

Boy in the World for field assistance.

The Theodore Roosevelt Memorial Grant from the American Museum of

Natural History for funding. Yoyd Clayton, Dwight Worker, Heartwood, and Wild

Virginia for additional funding and entertainment.

US Forest Service (Ken Landgraf), Virginia Department of Game and Inland

Fisheries (Shirl Dressler, J.D. Kloepfer), and WV Department of Natural Resources

(Barbara Sargent, Kieran O’Malley) for permits.

8 TABLE OF CONTENTS

Page

Abstract ...... 3 Dedication ...... 6 Acknowledgments ...... 7 List of Tables ...... 13 List of Figures [figures are in Supplement document] ...... 17 Chapter 1: Aspects of Spatial Ecology and Morphometrics of Wood Turtles (Glyptemys insculpta) at Their Southern Range Periphery ...... 21 Introduction ...... 21 Focal Species ...... 23 Methods ...... 24 Study Area ...... 24 Field Procedures ...... 25 Analytic Procedures ...... 27 Results ...... 28 Spatial Ecology ...... 28 Morphology ...... 30 Discussion ...... 31 Spatial Ecology ...... 31 Morphology ...... 38 Chapter 2: Demographic Estimates and Conservation Implications for a Southern Population of Wood Turtles (Glyptemys insculpta) ...... 47 Introduction ...... 47 Focal Species ...... 49 Methods ...... 50 Study Area ...... 50 Field Procedures ...... 51 Analytic Procedures ...... 52 Results ...... 55 Field Census ...... 55 MARK Analyses ...... 56 9 Discussion ...... 59 Chapter 3: Thermal Patterns of Wood Turtles (Glyptemys insculpta) and Their Associated Microhabitats in Montane Forests of Virginia and West Virginia ...... 75 Introduction ...... 75 Focal Species ...... 78 Methods ...... 80 Study Area ...... 80 Field Procedures ...... 81 Microhabitat Temperatures ...... 81 Turtle Temperatures ...... 83 Ground Temperatures at Turtle and Random Locations ...... 84 Analytic Procedures ...... 85 Analysis Rationale ...... 86 Results ...... 88 Microhabitat Temperatures ...... 88 Diurnal Microhabitat Temperatures ...... 88 Spatial Patterns of Diurnal Microhabitat Temperatures ...... 91 Nocturnal Microhabitat Temperatures ...... 92 Correlations of Diurnal Microhabitat Temperatures and Environmental Attributes ...... 93 Surface Temperatures at Turtle and Random Points ...... 94 VA 2011-2014 ...... 96 WV 2011-2014 ...... 96 VA – WV 2011-2014 ...... 97 Turtle Shell Temperatures ...... 97 Diurnal ...... 97 Nocturnal ...... 99 Discussion ...... 100 Environmental Temperatures ...... 100 Turtle Temperatures ...... 105 Use of Habitat ...... 112 Conservation Considerations ...... 121 10 Chapter 4: Using a GIS-based Water Balance Approach to Investigate Chelonian Habitat Use in Central Appalachian Forests ...... 138 Introduction ...... 138 Focal Species ...... 141 Methods ...... 142 Study Area ...... 142 Water Balance Model Rationale ...... 142 GIS Procedures ...... 143 Field Procedures ...... 145 Analytic Procedures ...... 146 Relative Humidity ...... 146 Water Balance ...... 146

Ground Level Relative Humidity and Temperature, and AET/DEF25 Values at Turtle and Random Points in 2011-2012 ...... 149 Results ...... 151 Ground Level Relative Humidity at Turtle and Random Points ...... 151 Water Balance at Turtle and Random Points ...... 154 Correlations Between Ground Level Relative Humidity and Water Balance Values at Turtle and Random Points ...... 154 Conditional Logistic Regression with AET Values for 2011-2012 ...... 155

Conditional Logistic Regression with DEF25 Values for 2011-2012 ...... 157 Discussion ...... 159 Chapter 5: Vegetative Indicators for Wood Turtle (Glyptemys insculpta) Habitat Use in Central Appalachian Forests ...... 186 Introduction ...... 186 Focal Species ...... 188 Methods ...... 190 Study Area ...... 190 Field Procedures ...... 190 Analytic Procedures ...... 192 Herbaceous and Woody Seedling Taxa ...... 192 Tree Taxa, Forest Types, and Seral Stages ...... 195 Results ...... 200 11 Forest Types and Seral Stages of Stands and Plots ...... 200 Overstory Trees ...... 203 Paired Logistic Regression ...... 204 CART ...... 206 Herbaceous Flora ...... 209 400m2 plots ...... 209 1m2 plots ...... 213 Woody Seedlings ...... 216 Virginia 1m2 ...... 217 West Virginia 1m2 ...... 217 Herbaceous Richness, Forest Types, Seral Stages, Herbaceous Cover ...... 218 Discussion ...... 220 Herbaceous and Seedling Taxa ...... 221 Overstory Tree Composition and Structure ...... 224 Forest Types and Seral Stages – Patch Scale ...... 227 Synthesis ...... 229 Conservation Recommendations ...... 235 Chapter 6: Multi-scale Habitat Preferences of Wood Turtles (Glyptemys insculpta) in Central Appalachian Forests ...... 261 Introduction ...... 261 Focal Species ...... 263 Methods ...... 264 Study Area ...... 264 Field Procedures ...... 264 Analytic Procedures ...... 265 Microhabitat Preference ...... 267 Mesohabitat Preference ...... 268 Macrohabitat Preference ...... 269 Study Rationale ...... 270 Results ...... 274 Preference for Microhabitat Features ...... 274 Preference for Mesohabitat Features ...... 276 12 Virginia ...... 277 West Virginia ...... 279 Pooled VA and WV Turtles ...... 279 Preference for Mesohabitat Features Using Random Effect Models ..... 282 Discussion ...... 284 Habitat Variables ...... 286 Canopy Openness ...... 286 Turtle- and Eye-level Obscurity ...... 289 Slope Inclination ...... 290 Aspect ...... 290 Woody Debris ...... 292 Snags ...... 294 Herbaceous Richness and Cover ...... 295 Shrub Taxa ...... 296 Tree Taxa ...... 296 Forest Type of Stand and Plot ...... 297 Stand Age and Seral Stage ...... 299 General Issues ...... 300 Conservation Recommendations ...... 303 References ...... 335 Appendix 1. Flora at Study Area ...... 385 Appendix 2. Annual Capture History (2006-2014) Used in MARK Analyses ...... 386 Appendix 3. Details of Microhabitat Temperature Patterns ...... 387 Appendix 4. Water Balance Conditional Logistic Regression Models ...... 394 Appendix 5. Importance Value Conditional Logistic Regression Models ...... 396 Appendix 6. Literature Review and Natural History of the Wood Turtle (Glyptemys insculpta) ...... 398 13 List of Tables

Page

Table 1.1 Spatial metrics for Wood Turtles radio-tracked in VA and WV ...... 41 Table 1.2 Morphometrics ANOVA results ...... 42 Table 1.3 Correlations of morphometrics and spatial metrics for adult Wood Turtles ...... 43 Table 1.4 Morphometrics for adult and juvenile Wood Turtles in VA and WV ..... 44 Table 1.5 Results of two-way ANOVAs comparing morphometrics ...... 45 Table 1.6 Results of paired t-tests comparing BMi to BMe ...... 46 Table 2.1 Annual sampling of WV Wood Turtles ...... 69 Table 2.2 Morphometrics for WV Wood Turtles ...... 70 Table 2.3 Summary of well-supported Cormack-Jolly-Seber models ...... 71 Table 2.4 Apparent survival (ϕ) and capture probability (ρ) estimates ...... 71 Table 2.5 Well-supported CJS models for estimated Wood Turtle abundance ...... 72 Table 2.6 Well-supported Pradel models for estimated population growth rate (λ) ...... 73 Table 2.7 Estimated population growth rate (λ) and seniority (γ) values ...... 73 Table 2.8 Reported demographic information for Wood Turtles ...... 74 Table 3.1 Environmental attributes at microhabitat iButton array sites ...... 126 Table 3.2 Diurnal microhabitat temperatures (C°) recorded by iButtons ...... 127 Table 3.3 Proportions of diurnal microhabitat temperatures recorded by iButtons ...... 128 Table 3.4 Nocturnal microhabitat temperatures (C°) recorded by iButtons ...... 129 Table 3.5 Metrics and array sites for diurnal microhabitat temperatures (C°) recorded by iButtons ...... 130 Table 3.6 Proportions of nocturnal microhabitat temperatures recorded by iButtons ...... 131 Table 3.7 Surface temperatures (C°) at turtle and random points in VA and WV 132 Table 3.8 Results of paired t-tests comparing surface temeratures at turtle and random points ...... 133 Table 3.9 Proportions of ground temperatures (“TempT”) immediately adjacent to turtles ...... 134 14 Table 3.10 Proportions of ground temperatures recorded in shade (“TempGr”) at turtle and paired random points ...... 135 Table 3.11 Temperatures (C°) recorded by iButtons attached to adult Wood Turtles ...... 136 Table 3.12 Proportions of temperatures recorded by iButtons attached to adult Wood Turtles ...... 137 Table 4.1 Values for ground surface-level relative humidity at turtle and paired random points ...... 170 Table 4.2 Results of paired t-tests comparing humidity at turtle and random points ...... 171 Table 4.3 Ground level relative humidity at pooled turtle and paired random points ...... 172 Table 4.4 ANOVA results for ground level relative humidity variables ...... 173 Table 4.5 Values for environmental variables at turtle and paired random points in 2011-2012 used in logistic regression models ...... 174 Table 4.6 Water balance values (AET, DEF, PET) at 400m2 plots ...... 175 Table 4.7 Wilcoxon test results for water balance at turtle and paired random points ...... 176 Table 4.8 Wilcoxon test results for water balance at turtle points and GIS- generated random points ...... 177 Table 4.9 Best conditional logistic regression model variables – using AET and months pooled ...... 178 Table 4.10 Well-supported conditional logistic regression models – using AET and months pooled ...... 179 Table 4.11 Best conditional logistic regression model variables – using monthly AET ...... 180 Table 4.12 Well-supported conditional logistic regression models – using monthly AET ...... 181

Table 4.13 Best conditional logistic regression model variables – using DEF25 and months pooled ...... 182

Table 4.14 Well-supported conditional logistic regression models – using DEF25 and months pooled ...... 183 Table 4.15 Best conditional logistic regression model variables – using monthly

DEF25 ...... 184 Table 4.16 Well-supported conditional logistic regression models – using monthly

DEF25 ...... 185 15 Table 5.1 Importance values of tree taxa in 400m2 plots at turtle and paired random points in VA and WV ...... 240 Table 5.2 Importance values of tree taxa in 400m2 plots at pooled turtle and paired random points in VA and WV ...... 241 Table 5.3 Forest type groups of 400m2 plots at turtle and paired random points 242 Table 5.4 Forest types of stands at turtle and paired random points ...... 243 Table 5.5 Seral stage of stands at turtle and paired random points ...... 244 Table 5.6 Floristic richness and cover in plots at turtle and paired random points ...... 245 Table 5.7 Paired t-test results comparing floristic variables at turtle and random points ...... 246 Table 5.8 Well-supported conditional logistic regression models – using importance values (IVs) of trees in 400m2 plots ...... 247 Table 5.9 Best conditional logistic regression model variables – using IVs ...... 248 Table 5.10 Synopsis of best conditional logistic regression model variables – using IVs for turtle groups in VA and WV ...... 249 Table 5.11 Well-supported conditional logistic regression models – using IVs and pooled turtle groups ...... 250 Table 5.12 Best conditional logistic regression model variables – using IVs and pooled turtle groups ...... 251 Table 5.13 Herbaceous taxa used in indicator species analyses – 400m2 plots in VA and WV ...... 252 Table 5.14 Herbaceous taxa with significant indicator values – 400m2 plots ..... 253 Table 5.15 Herbaceous taxa used in association analyses – 400m2 plots ...... 254 Table 5.16 Herbaceous taxa with significant association values – 400m2 plots .. 255 Table 5.17 Herbaceous taxa used in indicator species analyses – 1m2 plots ...... 256 Table 5.18 Herbaceous taxa used in association analyses – 1m2 plots ...... 257 Table 5.19 Woody seedling taxa used in indicator species analyses – 1m2 plots 258 Table 5.20 Woody seedling taxa used in association analyses – 1m2 plots ...... 259 Table 5.21 Herbaceous taxa used in association analyses – 400m2 plots and forest type groups ...... 260 Table 6.1 Ground cover variables measured in 1m2 plots ...... 311 Table 6.2 Environmental variables measured in 400m2 plots ...... 312 Table 6.3 Conditional logistic regression models used – 1m2 plots ...... 313 Table 6.4 Conditional logistic regression models used – 400m2 plots ...... 314 16 Table 6.5a Values of ground cover variables measured in center 1m2 plots ...... 318 Table 6.5b Values of ground cover variables measured in four peripheral 1m2 plots ...... 319 Table 6.6a Values of environmental variables measured in 400m2 plots ...... 320 Table 6.6b Values of environmental variables in 400m2 plots – pooled data ...... 321 Table 6.7 Well-supported conditional logistic regression models – using cover variables in 1m2 plots ...... 322 Table 6.8 Best conditional logistic regression model variables – cover variables in 1m2 plots ...... 323 Table 6.9 Well-supported conditional logistic regression models – using variables in 400m2 plots 2011-2014 ...... 324 Table 6.10 Best conditional logistic regression model variables for 400m2 plots 2011-2014 ...... 325 Table 6.11 Well-supported conditional logistic regression models – using variables in 400m2 plots, including number of herbaceous taxa 2011-2013 ...... 326 Table 6.12 Best conditional logistic regression model variables for 400m2 plots 2011-2013, including number of herbaceous taxa ...... 327 Table 6.13 Well-supported conditional logistic regression models – using variables in 400m2 plots and pooled data 2011-2014 ...... 328 Table 6.14 Best conditional logistic regression model variables for 400m2 plots, using pooled data 2011-2014 ...... 329 Table 6.15 Well-supported conditional logistic regression models – using variables in 400m2 plots, including number of herbaceous taxa and IVs, pooled data 2011- 2013...... 330 Table 6.16 Best conditional logistic regression model variables for 400m2 plots, including number of herbaceous taxa and IVs, using pooled data 2011-2013 .... 331 Table 6.17 Synopsis of best conditional logistic regression model variables in 400m2 plots for various turtle groups in VA and WV ...... 332 Table 6.18 Well-supported mixed conditional logistic regression models – using variables in 400m2 plots, including stand and plot forest type and seral stage, pooled data 2011-2013 ...... 333 Table 6.19 Best mixed conditional logistic regression model variables for 400m2 plots, using stand and plot forest type and seral stage, pooled data 2011-2013 ... 334

17 List of Figures

Page

Figure 1.1. Map of study area ...... 6 Figure 1.2. Map with activity areas (MCPs) of Wood Turtles in 2014...... 7 Figure 2.1. Map of WV study site with 2011-2014 locations of Wood Turtles ...... 8 Figure 2.2. Photos of plastrons of two adult Wood Turtles in different years...... 9 Figure 3.1. Map of locations of 2013 microhabitat array sites ...... 10 Figure 3.2. Means of daily nocturnal microhabitat temperatures recorded by iButtons...... 11 Figure 3.3. Means of daily diurnal microhabitat temperatures recorded by iButtons...... 12 Figure 3.4. Means, mean maxima, and mean minima of all diurnal microhabitat temperatures...... 13 Figure 3.5. Means of diurnal microhabitat temperatures recorded every 30 minutes...... 14 Figure 3.6. Proportions of diurnal temperatures recorded by iButtons in VA...... 15 Figure 3.7. Diurnal microhabitat temperatures recorded every 30 minutes on a single day at a single array site...... 16 Figure 3.8. Means of nocturnal microhabitat temperatures recorded every 30 minutes...... 17 Figure 3.9. Ground level temperatures immediately adjacent to turtles (“TempT”) and in nearby shade (“TempGr”)...... 18 Figure 3.10. Mean temperatures recorded by iButtons attached to Wood Turtles. . 19 Figure 3.11. Diurnal temperatures on a single day recorded by iButtons attached to three Wood Turtles ...... 20 Figure 3.12. Daily mean diurnal temperatures recorded by iButtons attached to Wood Turtles...... 21 Figure 3.13. Mean diurnal temperatures recorded every 15 minutes by iButtons attached to Wood Turtles...... 22 Figure 3.14. Means, mean maxima, and mean minima of diurnal temperatures recorded by iButtons attached to Wood Turtles...... 23 Figure 3.15. Hourly means of summer diurnal temperatures (UV and WT) recorded by iButtons in Virginia and West Virginia ...... 24 Figure 3.16. Proportions of diurnal temperatures recorded by iButtons attached to adult Wood Turtles...... 25 18 Figure 3.17. Daily mean diurnal minimum and maximum temperatures recorded by iButtons attached to Wood Turtles...... 27 Figure 3.18. Diurnal temperatures recorded by iButtons attached to two adult Wood Turtles (female and male)...... 28 Figure 3.19. Diurnal temperatures recorded by an iButton attached to an adult male Wood Turtle in Virginia on August 5-August 15, 2014...... 29 Figure 3.20. Proportions of nocturnal temperatures recorded by iButtons attached to Wood Turtles...... 30 Figure 4.1. Ground level relative humidity measured in shade at time of Wood Turtle capture/recapture (“HumGr”) ...... 31 Figure 4.2. Ground level relative humidity immediately adjacent to Wood Turtles (“HumT”) and in closeby shade (“HumGr”)...... 32 Figure 4.3. Mean monthly actual evapotranspiration (AET) at Virginia and West Virginia study sites calculated with Dyer water balance model...... 33

Figure 4.4. Mean monthly water deficits (DEF25) at Virginia and West Virginia study sites calculated with Dyer water balance model...... 34

Figure 4.5. Map of distribution of July DEF25 (segregated by median value) at WV study site...... 35 Figure 4.6. Map depicting distribution of NRCS soil types at WV study site...... 36 Figure 5.1. Proportions of turtle and random points in different forest type groups in VA and WV based on importance values calculated for each 400m2 plot ...... 37 Figure 5.2. Counts of turtle and random points in different forest type groups in VA and WV based on IVs in 400m2 plots...... 38 Figure 5.3. Counts of turtle and random points in different forest type groups in VA and WV based on USFS stand inventory data...... 39 Figure 5.4. Map of turtle and random points in delineated stands of different forest type groups in VA...... 40 Figure 5.5. Map of turtle and random points in delineated stands of different forest type groups in WV...... 41 Figure 5.6. Comparison of counts of pooled turtle and random points in forest type groups characterized at 400m2 plot scale and stand scale...... 42 Figure 5.7. Counts of points in stands of different seral stages in Virginia...... 43 Figure 5.8. Map of points in stands of different seral stages in Virginia...... 44 Figure 5.9. Virginia CART diagram of importance values of tree taxa in 400m2 plots...... 45 Figure 5.10. Virginia CART diagram of importance values of tree taxa in 400m2 plots with Chestnut Oak, models without CO and SO ...... 46 19 Figure 5.11. Virginia CART diagram of importance values of tree taxa in 400m2 plots with Chestnut Oak, models without CO ...... 47 Figure 5.12. West Virginia CART diagram of importance values of tree taxa in 400m2 plots ...... 48 Figure 5.13. West Virginia CART diagram of importance values of tree taxa in 400m2 plots without Sycamore ...... 49 Figure 5.14. Mean numbers of herbaceous taxa in 400m2 plots in VA and WV. ... 50 Figure 5.15. Mean numbers of herbaceous taxa in 1m2 plots in VA and WV...... 51 Figure 5.16. Number of herbaceous taxa in 400m2 plots at random points in different seral stages in Virginia...... 52 Figure 5.17. Number of herbaceous taxa in 400m2 plots at random points in different forest types in VA and WV...... 53 Figure 5.18. Regression of number of herbaceous taxa in 400m2 plots in VA on IVs of Scarlet Oak and White Ash...... 54 Figure 5.19. Regression of number of herbaceous taxa in 400m2 plots in WV on IVs of Virginia Pine and Sugar Maple ...... 55 Figure 5.20. Mean amounts (%) of herbaceous coverage (combined forb and grass cover) in 1m2 plots in VA and WV ...... 56 Figure 5.21. Photo showing abundant grass cover at a site with high importance value for Virginia Pine in WV ...... 57 Figure 6.1. Proportions of ground cover types estimated in 1m2 plots at turtle and paired random points in VA and WV ...... 58 Figure 6.2. Logistic regression coefficient values for cover variables at 1m2 plots .. 59 Figure 6.3. Mean amount (%) of herbaceous coverage (combined forb and grass cover) in central and peripheral 1m2 plots...... 60 Figure 6.4. Mean values for some environmental variables measured in 400m2 plots at turtle and paired random points...... 61 Figure 6.5. Conditional logistic regression coefficient values for habitat variables in 400m2 plots, 2011-2014...... 62 Figure 6.6. Conditional logistic regression coefficient values for habitat variables in 400m2 plots, including the number of herbaceous taxa, 2011-2013...... 63 Figure 6.7. Conditional logistic regression coefficient values for habitat variables in 400m2 plots, using pooled VA and WV turtles...... 64 Figure 6.8. Conditional logistic regression coefficient values for tree IVs in 400m2 plots, using pooled VA and WV turtles ...... 65 Figure 6.9. Photo of male Wood Turtle basking beside LWD...... 66 20 Figure 6.10. Map of Wood Turtle location points in relation to recently logged sites (esh) in Virginia...... 67

21 CHAPTER 1: ASPECTS OF SPATIAL ECOLOGY AND MORPHOMETRICS OF

WOOD TURTLES (GLYPTEMYS INSCULPTA) AT THEIR SOUTHERN RANGE

PERIPHERY

Introduction

Home range size and shape is a fundamental and frequently studied ecological parameter for many species (Hayne 1949, Nilsen et al. 2005). As with other taxa, spatial activity patterns of North American Wood Turtles (Glyptemys insculpta) are influenced by intra- and inter-specific factors and variation in dietary preference, predator avoidance, thermo- and osmo-regulation, mating opportunities, competition, morphology, and sex (Strang 1983, Foscarini 1994, Lindeman 2000,

Compton et al. 2002, Penick et al. 2002, Converse and Savidge 2003, Plummer

2003, Arvisais et al. 2004, Akre and Ernst 2006, Rossell et al. 2006, Mueller and

Fagan 2008, Millar and Blouin-Demers 2011, Jaeger and Cobb 2012). Organisms face tradeoffs in resource allocation to various compartments of their time and energy budgets and their choices can have significant implications for their use of space and ultimately their survival and reproductive output (Congdon 1989, Doak et al. 1992,

Kearney 2006, Morris et al. 2008). These necessities and choices may increase or decrease the use of available habitats and thereby determine an individual’s home range or activity area (Huey 1991, Downes 2001, Halstead et al. 2009).

When a Wood Turtle is located repeatedly at the same site it is not known whether this is due to frequent return to the site or to prolonged stay. Either way, stay periods at a particular area suggest the favorability of that habitat within the context 22 of the allocation tradeoffs (Owen-Smith et al. 2010). Abundant, spatially concentrated, and predictable resources tend to promote small home ranges, while productivity decline or more patchily distributed or spatially unpredictable resources can result in home range expansion (i.e., lowering the population density in the landscape) (Kapfer et al. 2010). However, larger home ranges may not be a response to low resource availability. Instead, turtles and other fauna perhaps range farther where suitable habitat is abundant (e.g., sites with a high availability of humid microclimates or food items) (Ross and Anderson 1990, Hamernick 2000, Jaeger and

Cobb 2012) because the opportunities for expansive roaming increase fitness.

In addition to differential habitat use between males and females (Tingley et al. 2010), sexual dimorphism has been observed in Wood Turtles (Lovich et al. 1990,

Stephens and Wiens 2009). Turtle size dimorphism has been interpreted in terms of sexual selection: Males are larger than females as an adaptation to increase success in male combat or to enable forcible insemination of females (Berry and Shine 1980).

Male Wood Turtles exhibit aggressive behavior such as chasing, biting, and lunging during male-male combat (Barzilay 1980, Kaufmann 1992b); Walde et al. (2003) documented bleeding and injuries to post-combat males and to female Wood Turtles resulting from aggressive mating. Wood Turtles maintain a hierarchal social structure based on size, sex, and maturity, with male rank affecting reproductive success

(Kaufmann 1992b, Galbraith 1993).

I present information on Wood Turtles sampled over the course of four years at relatively undisturbed forested montane sites on the southern periphery of the 23 species’ range in Virginia and West Virginia USA. Herein, as radio-telemetry was used only during the summer, I calculated seasonal “activity areas”. The objective was to reveal basic spatial ecology and morphological parameters, such as activity area size and body size, with the goal of informing in situ conservation.

Focal Species

The Wood Turtle is a small (ca. 1 kg when adult) long-lived amphibious omnivore. It is seasonally biphasic, living terrestrially during warm periods and being more aquatic when temperatures are lower, with underwater hibernation throughout the winter. Wood Turtles generally prefer deciduous, coniferous, and mixed forests around clear, low gradient 1st to 5th order streams with moderate currents and hard bottoms (Ernst 2001b). The Wood Turtle has a broad geographic range in southeastern Canada and northeastern USA and is one of the more northern ranging Nearctic reptiles (Buhlman et al. 2009, TTWG 2017). Populations in the Virginias are at the southern periphery of their range.

The International Union for the Conservation of Nature (“IUCN”) classified the Wood Turtle as an “endangered” species on its Red List to reflect range- contraction and increased threats and impacts from human encroachment (Van Dijk and Harding 2011). A petition advocating the species’ listing under the federal

Endangered Species Act is currently before the US Secretary of the Interior (CBD

2012, USDI FWS 2015). The Wood Turtle is state listed as “threatened” in Virginia.

West Virginia considers it a "Priority Group 1" species in the state’s wildlife conservation strategy (“species of greatest conservation need” in WVDNR 2005). 24 See Appendix 6 for further information on the species.

Methods

Study Area

Fieldwork took place at two forested montane sites close to the species southern range limit on the George Washington National Forest (“GWNF”) of

Virginia (VA) and West Virginia (WV). Both sites are within the Ridge and Valley physiographic province (Hunt 1974, McNab et al. 2007) and are upper-elevation headwaters of the Potomac River drainage basin of the Chesapeake Bay watershed.

Elevations at Wood Turtle locations range from ca. 260-400m in WV and 425-610m in VA, although surrounding ridges within the watersheds reach altitudes of ca. 865m above sea level. The study sites lie within two orographic rainshadows, from the Blue

Ridge Mountains to the east and the Allegheny Mountains and Plateau to the west.

As a result, this is one of the drier locales in the species’ range, with average annual precipitation measuring only ca. 95 cm, of which ca. 27cm falls in June through

August; the mean temperature during this time period is 22.3°C (NCDC 2013).

The portion of the main stream at the WV study site on the National Forest is ca. 3.5km in length, while that at VA is ca. 12km in length. Both main streams are low gradient 1st – 3rd order streams with mostly pebble-cobble-boulder substrates and numerous riffle-run-pool habitats replicated along the length of each stream. Summer water depths at both streams range from ca. 3-100 cm, with channel widths of ca. 1-

7 m. Associated with the VA stream are broad gently sloping (1-7° inclination) flats and benches up to ca. 300m wide. In contrast, the stream channel in WV is generally 25 more narrowly entrenched and sharply incised, with closeby steep slopes and some flats up to ca. 125m wide (Fig. 1.1).

Forests here are within the Oak – Chestnut Region of Braun (1950) and the

Appalachian Oak Section of the Mesophytic Region identified by Dyer (2006).

Though many herbaceous and woody species are held in common, the forests found at the two study sites are also noticeably different (see Appendix 1 for common plant taxa at the two sites). The VA site is predominantly composed of mixed oak and other deciduous forest types, while the WV site is comprised of a greater proportion of relatively more-xeric pine and mixed pine-deciduous forest types (Tables 5.3 and

5.4).

In addition to differences in topography, stream size, elevation, and forest composition between the two sites, there are some structural differences as well.

Commercial even-age logging fabricated stands of early-successional (1-35 years old) and mid-successional forest (36-75 years old) in VA that do not occur at the WV site.

A gravel road runs the length of the VA site, but not the WV site.

Field Procedures

I tracked movement patterns of individual G. insculpta using radio telemetry.

At the two study sites, individual Wood Turtles were searched for visually and hand- captured on land in June, July, and August of 2011-2014. I affixed radio-transmitters

(Advanced Telemetry Systems R2030; ca. 25gm each, ca. 40gm total with adhesive) with plumber’s putty or epoxy glue to the rear marginal and pleural scutes of the

Turtles; these transmitters were removed at the end of each field season. A portable 26 telemetry receiver (Telonics, Inc. TR-5 164-168 MHz radio receiver) and hand-held antennae were used to locate the turtles. In 2011 eleven Wood Turtles were outfitted with transmitters: 5 adults (4 males) at WV and 6 adults (4 females) at VA; in 2012 nineteen Wood Turtles: 8 adults (2 males) and 1 juvenile at WV, 9 adults (2 males) and 1 juvenile at VA; in 2013 nineteen Turtles: 8 adults (2 males) and 1 juvenile at

WV, 9 adults (2 males) and 1 juvenile at VA; in 2014 nineteen Turtles: 8 adults (2 males) and 1 juvenile at WV, 9 adults (2 males) and 1 juvenile at VA. In total, 29 adult turtles (19F:10M) and 3 juveniles were radio-tracked in WV and 33 adults

(25F:8M) and 3 juveniles in VA. Individuals were considered adults if carapace length was > 160mm; sex was determined by observation of male secondary sexual characteristics (viz., a plastral concavity and longer thicker pre-cloacal tail length).

Turtles were released at the point of initial capture within two hours. Each individual was located on average ca. every 10 days during June-August in a single season.

Geographic coordinates for all turtle locations and their paired random points were recorded with a Garmin ETrex GPS unit. In addition to the 68 radio-tracked turtles, morphometric data were also collected for non-transmittered Wood Turtles serendipitously encountered in the field during the four field seasons; hence, I report morphometrics obtained from 81 adult individuals (40 in VA, 41 in WV). Unless otherwise noted, juvenile plots were not used in the reported analyses due to low sample size.

The following abbreviations are used in this dissertation: VA female Wood

Turtles = VFT, VA male Wood Turtles = VMT; WV female Wood Turtles = WFT, WV 27 male Wood Turtles = WMT, WV juvenile Wood Turtles = WJT; for both states combined, female Wood Turtles = FWT, male Wood Turtles = MWT; all turtles pooled = WT.

Analytic Procedures

Seasonal activity areas (minimum convex polygons – MCPs) and maximum straight-line distances across these areas were calculated in ArcGIS (vers. 9.3; ESRI

2010) for individuals with at least three location points. The time intervals (number of days) between the two points with the maximum separation were used to calculate the straight-line distance traveled per day for each individual. I divided the activity area (m2) by the maximum distance to derive an average width in meters for the activity areas. For each activity area I divided the maximum distance (i.e., length) by the mean width to derive a shape metric; the greater this number then the more elongated the shape of the activity area. I also calculated sizes of transformed activity areas by using a formula derived by J.M. Turner (Barrett 1990) that corrects activity areas for sample size bias by dividing the calculated areas by 0.257(ln(n)−0.31), where n is the number of animal observations used in each area calculation. Activity area data were normalized with a natural log transformation and analysed with t- tests, ANOVAs, post hoc Tukey HSD tests, and Pearson’s correlation tests in R (R

Core Team 2015). Because of the small sample size (2), comparisons involving juveniles were not tested.

When initially encountered in the field Wood Turtles were measured and weighed using standard protocols (Table 1.2). I used Haglof calipers to obtain 28 straight-line shell dimensions (±1mm) and Pesola spring scales (±5 g), to obtain body

mass. The morphometrics included: CL = carapace length (using nuchal notch), CLmax

= maximum carapace length (using marginal outer edges), CW = carapace width at widest axis, SD = shell depth (bottom of plastron to top of vertebrals) at thickest point,

PL = plastron length (using the inside of gular and anal notches), PLmax = plastron length using outer margins of gular and anal scutes), and BMi = body mass when first weighed at beginning of annual field season. Metrics were obtained from 49 female and 32 male adult turtles, along with nine juveniles. For a smaller subset of turtles that were recaptured, I also obtained an end-of-field-season body mass measurement

(BMe). I calculated change in mass as dM = BMi – BMe (including 4 females and 5 males of 42 total turtles that declined in mass), and dM% = change in mass as a proportion of initial mass. I calculated means and standard errors for all turtle morphometric data and analysed with Pearson correlation tests, ANOVAs, Tukey

HSD tests, and students t-tests.

Results

Spatial Ecology

There were no differences in activity area size, corrected activity area size, maximum distance across activity area, distance moved per day, maximum distance from the main streams, and activity area shape, ANOVA results indicated no significant differences between states, sexes, or state:sex interaction (Tables 1.1 &

1.2). Maximum distance from the main streams differed between males and females pooled for both states. Within state comparisons between males and females for this 29 metric did not differ (Tukey HSD p-values = 0.23 (VA) and 0.70 (WV)). Mean distance to the main streams was 93.2m ± 6.1m for all terrestrial turtle points, with females locating farther away from the main streams than did males (Tables 1.2 and Table

6.6). In VA the distance from the main stream containing 95% of female terrestrial location points was 375m and 256m for males, while in WV the distance was 406m for females and 243m for males.

For pooled turtles (sites and sexes combined) there was an inverse correlation between activity area size and end of season body mass (BMe; r = -0.347, t = -2.22, p = 0.033) and a marginally significant negative correlation between activity area size and initial body mass (BMi; r = -0.270, t = -1.98, p = 0.053). These correlations were driven by male turtles, results for female turtles were not significant for either

BMe or BMi (Table 1.3). There was no significant correlation between activity area

size and maximum carapace length (CLmax; r = -0.073, t = -0.48, p = 0.634), or carapace width (CW; r = -0.242, t = -1.88, p = 0.065) for all turtles combined. For female turtles and all turtles combined there were no significant correlations between maximum distances from the main streams and either BMe or BMi. However, for male turtles there were significant negative correlations between maximum distances from the main streams and both BMe and BMi (Table 1.3).

Biomass of adult Wood Turtles per unit area at the WV site was ca. 372 gms/ha. This was derived using an estimated adult population size of 76 (Chapter 2), a mean turtle mass of 1072gms (mean of BMi and BMe at Table 1.4), and a total site 30 area of 219 ha (based on a 290m buffer zone around the WV mainstream, within which were 95% of turtle location points).

Morphology

Male and female Wood Turtles differed in CL (carapace length), CLmax, CW, and BMi, but not BMe, with males having greater mean carapace lengths and widths as well as greater initial body masses than females (Tables 1.4 & 1.5). Shell depths and plastron lengths were equivalent between the sexes. Females had slightly

rounder carapaces than males, with a mean CW to CLmax ratio of 0.768 (se = 0.004) vs. 0.739 (se = 0.005) for males. Turtles from the two states were roughly of the same mean dimensions, particularly the males (e.g., several Tukey HSD pairwise comparisons had p-values >0.997). The only differences between VA and WV Wood

Turtles were CLmax and PLmax (maximum plastron length) (Table 1.5). In pairwise comparisons of the four groups (VFT/VMT/WFT/WMT), West Virginia females

repeatedly differed from Virginia males (BMi, CD, CL, CLmax) whereas other comparisons differed less frequently. BMi and BMe differed for VA females, but no other turtles (Table 1.6). For both sexes the strongest correlate with BMi was CW;

CL and PL were also strongly correlated (Table 1.3). CW, CL, and PL also were strongly correlated with BMe.

Female turtles tended to gain proportionately greater mass than did males

(2.6% of BMi vs. 1.1%). Juveniles (n = 2) far out-performed adults, with a mean gain in mass of 10.8% (Table 2). 31 Discussion

Spatial Ecology

Concordant with findings from elsewhere, Wood Turtles at my sites exhibited high site fidelity and short-distance movements during the summer. Wood Turtle home range sizes vary geographically, being generally larger at higher latitudes

(Quinn and Tate 1991, Brooks et al. 1992, Arvisais et al. 2002), and are usually anchored on a creek, stream, or river and elongate in shape as a result (Strang 1983,

Kaufman 1995, Daigle 1997, Saumure 2004, Curtis and Vila 2015, McCoard et al.

2016a). Home ranges tend to be small, but are highly variable among individuals

(Quinn and Tate 1991, Daigle 1997, Arvisais et al. 2002, Saumure 2004, McCoard et al. 2016a), both within a population and among populations inhabiting different landscape mosaics (Akre and Ernst 2006). In this respect, Wood Turtles display an individual variability in home range size, activity, and habitat use similar to that observed in other forest-dwelling testudines (cf., Moskovits and Kiester 1987

[Geochelone], Lue and Chen 1999 [Cuora], Lawson 2006 [Kinixys], Seibert and

Belzer 2015 [Terrapene]).

The activity areas of the radio-tracked individuals spanned almost two orders of magnitude, from 0.2ha to 13.3ha. The shapes of many of the activity areas delineated for this study were indeed elongate (see, e.g., Fig. 1.2), though statistical tests indicated the shapes of the males’ activity areas were not more elongate than those of the females. In addition, the size of activity areas did not differ between males and females. Within other Wood Turtle populations, home range sizes did not 32 differ between males and females, nor were home range sizes correlated to turtle size or dominance (Ross et al. 1991, Kaufmann 1995, Arvisais et al. 2002, Tuttle and

Carroll 2003, Remsberg et al. 2006, Akre and Ernst 2006, Curtis and Vila 2015,

McCoard et al. 2016). Though in Massachusetts, Jones (2009) found that older females had larger home ranges than younger females. In the mountains of North

Carolina, Smith and Cherry (2016) observed similar patterns in the Wood Turtle’s congener G. muhlenbergii (Bog Turtle), with individuals’ home ranges spanning an order of magnitude and no difference detected between sexes. For male Wood

Turtles, my study found both the size of activity areas and maximum distance from the main streams to be inversely correlated with body mass (Table 1.3). For larger males, this pattern may potentially increase mating opportunities, which are positively correlated with size (Kaufmann 1992b). The smaller activity area size may also be indicative of high habitat quality and not having to cover as much ground to successfully forage.

Though for this study the shapes and sizes of activity areas did not differ between sexes, males tended to locate closer than females to the main streams (Table

1.1). Tingley and colleagues (2009) found a similar pattern for Wood Turtles in Nova

Scotia, Canada (95% of female locations were within 235m of water, while 95% of male locations were within 43m), as did Tuttle and Carroll (2003) in New Hampshire

(95% of female locations were within 188m of water, while 95% of male locations were within 61m), Parren (2013) in Vermont (276m mean maximum distance for females and 107m for males), and McCoard et al. (2016) in West Virginia (mean 33 maximum distance from the river for females was 140 ± 26m, while that for males was 86 ± 20m). Thus, male summer activity areas may include more stream length than do females (Jones 2009). Though many expansive female movements pertain to nesting behavior, post-nesting females also may locate distant from water (Tuttle and

Carroll 2003, Akre and Ernst 2006, Jones 2009, Flanagan et al. 2013, Parren 2013).

Although often found near the main streams, in this study Wood Turtles of both sexes were observed at dry drainages, slopes, and ridges far from water (e.g., >500m away).

These movements are presumably an infrequent but critical component of Wood

Turtle behavior and demography and somehow confer fitness benefits.

A typical explanation for differences in home ranges between sexes is based upon differences in life history strategy, with female patterns based on metabolic needs (eggs are energetically expensive compared to sperm) and male patterns structured by efforts to increase reproductive success (mate with more females;

Schoener and Schoener 1982). My results support this explanation. Although the

Wood Turtle is seasonally biphasic, living terrestrially during warm periods and being more aquatic when temperatures are lower, most individuals temporarily return to the main streams occasionally during the summer. Tuttle and Carroll (2003) and Akre and Ernst (2006) found, in New Hampshire and Virginia populations, respectively, that females tended to be more terrestrial than males during the active season. Likewise, in this study male Wood Turtles were found more often in aquatic habitats than were females. In WV, 38.2% of 144 male locations were aquatic, while

21.7% of 152 female locations were. In VA, 17.6% of 102 male locations were 34 aquatic, while 9.7% of 247 female locations were. Over the course of the 2011-2014 field seasons, 100% of telemetered WV males and 41.2% of WV females were radio- located aquatically at least once. In Virginia 57.1% of radio-telemetered males and

53.6% of females were located aquatically at some time. The great majority of courtship mountings take place aquatically (Ernst and Lovich 2009); ca. 95% I observed were in water (Krichbaum pers. obs.). By locating activity areas closer to the streams and being more aquatic, male Wood Turtles increase their likelihood of intercepting females, thereby increasing mating opportunities.

If the size of home ranges scales to energetic constraints and metabolic rates

(McNab 1963), then the size of female turtles’ areas could be expected to be larger than those of males. I found no difference, however, between the size of activity areas of males and females, consistent with the results of the meta-analysis of turtle home ranges conducted by Slavenko and colleagues (2016). They also found that predation mode (viz., herbivory, omnivory, or carnivory) had a weak effect on home range size and that turtles generally have small home ranges even for terrestrial ectotherms (compared to similarly sized frogs or snakes). The minimal energetic requirements of turtles can result in other extrinsic factors, such as predator avoidance or topography, having greater effect on home range size than does energy acquisition. The apparent high habitat quality at my study sites (suggested by the observed philopatry and high survival over the years – Krichbaum pers. obs. and

Chapter 2) in concert with the low Wood Turtle densities (e.g., an estimated 0.70 adults/ha in WV – Chapter 2) may serve to provide suitable habitat without any 35 limiting resource across broad areas, thus obviating any overt discrepancy between the areal extent of habitat use by the different sexes.

That space to live is not a constraint here for either sex is indicated by the low estimated biomass of the WV turtles (372 gms/ha). To place this figure in perspective,

Wood Turtle biomass here is far less than the 8,730gms/ha of Eastern Box Turtles

(Terrapene carolina) at a northern Virginia site (T.P. Boucher 1999 in Ernst and Lovich

2009), the 4,300-8,700gms/ha estimated for Pennsylvania Spotted Turtles (Clemmys guttata; Iverson 1982), or the 1,927gms/ha of Eastern Box Turtles at a Maryland site

(estimated by me using an average turtle mass of 367gms; Quinlan et al. 2003). It is also less than the approximate 1998gms/ha for adult Wood Turtles in New Jersey

(estimated by me using the turtle population numbers and 61.3ha study area in Farrell and Graham 1991 and 850gms per turtle). The biomass estimate for the WV turtles, however, does align with those of other Wood Turtle populations, based on consistently calculated densities (see Table 2.8 in Chapter 2). Estimation of density and biomass are sensitive to the amount of area used in the calculations; it is unknown how the preceding cited researchers defined their areas of computation.

In general, animals should be trying to leave poor-quality habitats and trying to stay in higher-quality habitats (Garshelis 2002). As a consequence of their philopatry and long life spans (Kaufmann 1995, Ernst 2001a&b, Arvisais et al. 2002,

Tuttle and Carroll 2003, Akre and Ernst 2006, Willoughby et al. 2008, Parren 2013,

Chapter 2), adult Wood Turtles are presumably familiar with their activity area and the location of resources and habitat patches therein, so their location at any time 36 can be reasonably presumed to represent habitat selection (McLellan 1986). It is difficult to envision how the low population densities of Wood Turtles at the study sites could appreciably impact available foraging, cover, or thermoregulatory resources; therefore, activity area or home range selection is probably not influenced by intraspecific competition for these resources. However, in Ontario Smith (2002) found Wood Turtle density to be negatively correlated with the size of home ranges.

No data exist for how and to what extent interspecific competitors may be affecting

Wood Turtle spatial ecology. Perhaps sympatric insectivores, omnivores, or herbivores such as White-tailed Deer (Odocoileus virginianus), mice, voles, shrews,

Wild Turkeys (Melleagris gallopavo), or Box Turtles are influencing habitat use here.

Though areal extents vary depending on the individual and local habitat conditions, as well as from year to year, Wood Turtle home ranges are generally limited in size: from 1-80 hectares, with an average of perhaps 2-20, though the time periods and the number of location points used to derive areal estimates varied among studies (Quinn and Tate 1991, Foscarini 1994, Kaufmann 1995, Arvisais et al. 2002, Smith 2002, Tuttle and Carroll 2003, Akre and Ernst 2006, Remsberg et al.

2006, Castellano 2008, Greaves 2008, Jones 2009, Curtis and Vila 2015, McCoard et al. 2016, this study). Comparisons are confounded by inconsistencies in the methodologies used to derive home range estimates (e.g., variation in the time periods used to calculate home range, the number of individual animal locations, and the range estimators employed, such as the use of minimum convex polygon vs. adaptive kernel models; detailed in Saumure 2004 and Jones 2009). 37 The size of Wood Turtle home ranges here were smaller than the mean values for seven species of semi-aquatic turtles (8.51ha, se = 5.28) and 28 taxa of chelonian omnivores (26.1ha, se = 7.9) (Slavenko et al. 2016), which could be inferred to indicate the sites’ high habitat quality. The time period of my study, however, was only the summer post-nesting season. Since vernal, autumnal, and nesting movements were excluded, it is to be expected that the estimated sizes of activity areas were relatively reduced. In addition, the number of location points used to delineate each individual’s activity area was very limited (generally only ca. 6-7) and varied between individuals; a greater number of points can be expected to result in larger estimated areas (Jaeger and Cobb 2012).

It is noteworthy that the size of activity areas did not significantly differ between Turtles in VA and WV, though different environmental conditions exist

(Chapters 3-6). Because home ranges of animals can be expected to be efficient with respect to spatially distributed resources (Mitchell and Powell 2007), this result suggests similar resource availability even though the forests differ between the two states, VA being much more dominated by deciduous taxa. The five taxa of overstory and midstory trees with the highest importance values in VA were Quercus montana,

Q. alba, Q. concinna (Chestnut, White, and Scarlet Oaks), Acer rubra (Red Maple), and Liriodendron tulipifera (Tulip Tree); while in WV the most prevalent taxa were

Pinus strobus and virginiana (White and Virginia Pines), Q. alba and montana, and

Carya spp. (Hickories) (Table 5.2). In addition, relief at the WV site is much more pronounced (see Fig. 1.1) with significantly steeper slopes. Mean slope inclination 38 of 123 random points in WV was 20.1° (se = 0.79), while for 197 random points in

VA it was only 9.8° (se = 0.36) (Table 6.6). Not only could steeper slopes be more metabolically expensive to traverse, there could be a physical limit to a turtle’s ability to use steep slopes; Box Turtles (T. carolina and T. ornata) had trouble maintaining their position on slopes greater than 40º (Muegel and Claussen 1994, Claussen et al.

2002). Though Wood Turtles are larger, have stronger limbs (Abdala et al. 2008), and are more adept at climbing than Box Turtles (Pope 1939, Krichbaum pers. obs.), there could still be metabolic, physical, and behavioral constraints upon their use of steep slopes. Being lower in altitude, temperatures were also generally higher in WV than

VA (Chapter 3). Moreover, the main stream in VA is wider and deeper with a clearly greater volume of flowing water in the summer.

Morphology

The body dimensions of the Wood Turtles in this study generally accord with those of populations elsewhere in the species’ range (Lovich et al. 1990, Table 4 in

Greaves and Litzgus 2009). Maximum size varies geographically, with the largest animals (up to 238-242 mm straight-line carapace length) occurring in the northern part of the range (Saumure 1992, Walde et al. 2003, Greaves and Litzgus 2009).

Generally, intermediate-sized turtles are found in the southern part of the taxon’s range, with the smallest turtles found in the center of the range (Greaves and Litzgus

2009). There is some support for the hypothesis that Turtle size is negatively correlated with the number of frost-free days (Walde et al. 2003). However, large individuals (ca. 215-220 mm in carapace length) occur at the southern limits of the 39 range as well (Akre 2002, Greaves and Litzgus 2009, this study). Of note is that though the latitude of this study is at the southern range periphery, the altitudes of these locales are some of the highest recorded in the literature, particularly at the

Virginia site where some turtle location points exceeded 600m in elevation. Because of the altitude (as well as climatic patterns) some habitat conditions are similar to those at considerably higher latitudes in the Turtle’s range; e.g., the number of frost- free days is similar to some places in Maine and New York (cite).

Of note is that Virginia females started out (BMi) smaller than males, but by the end of the field season (BMe) had caught up in size. This study occurred during the post-nesting period. Many of the females are known to have nested before the study commenced (Krichbaum unpub. data), and many of the others can be presumed to have nested prior to initial weighing (Akre 2002). Much of their energy allocation had gone into egg production, so it is not surprising that post-nesting BMi differed between the sexes. Females made up for this by gaining mass at over twice the rate of males (see ∆M% at Table 1.4). This is important because females with better body condition can have greater reproductive output over the long-term

(Litzgus et al. 2008).

Previous studies found males to reach a larger maximum size than females, have greater mean carapace lengths, and have larger wider heads (Lovich et al. 1990,

Farrell and Graham 1991, Tuttle 1996, Akre 2002, Greaves and Litzgus 2009, Curtis and Vila 2015). Large animals may approach 1.5 kg in mass (Akre 2002, Arvisais et al. 2002, Walde et al. 2003, Castellano et al. 2008, Greaves and Litzgus 2009). Tuttle 40 (1996) found no statistically significant difference between the mass of males (mean

= 750.3 ± 134.0 g) and females (mean = 711.4 ± 85.9 g). The carapace of males is proportionately narrower than females’ (Harding 2002, Akre 2002, Greaves and

Litzgus 2009, Jones and Willey 2015). The results of this study are congruent with the above findings.

41 Table 1.1

Spatial metrics for Wood Turtles radio-tracked in VA and WV Spatial metrics for Wood Turtles radio-tracked in Virginia and West Virginia during June-August 2011-2014. Reported are means, standard errors, and ranges; linear measurements in meters, areas in hectares. Seasonal activity areas were calculated from MCPs formed from single-year location points of individuals. Corrected activity areas were corrected for sample size (number of observation points) bias. Maximum distance = straight-line distance between the two MCP points farthest apart. Distance per day = maximum distance divided by the number of days between the two location points used to define the maximum distance. Number of points = number of terrestrial radio-locations used to calculate activity area MCPs. N = number of individual turtles. VFT = Virginia female Wood Turtles, VMT = VA male Wood Turtles, WFT = West Virginia female Wood Turtles, WMT = WV male Wood Turtles, WJT = WV juvenile Wood Turtles, FWT = female Wood Turtles (VA & WV together), MWT = male Wood Turtles (VA & WV together), WT = adult Wood Turtles (sexes and states aggregated).

VFT VMT WFT WMT WJT FWT MWT WT Activity area size 1.90 2.19 2.35 2.92 7.17 2.07 2.57 2.25 SE 0.40 1.09 0.78 0.94 5.21 0.38 0.81 0.38 Range 0.2-7.5 0.2-13.3 0.2-11.5 0.3-13.1 2.0-12.4 0.2-11.5 0.2-13.3 0.2-13.3 Corrected act. area 4.41 5.09 5.59 5.19 24.4 4.85 5.14 4.95 SE 0.99 2.57 1.46 1.13 20.4 0.82 1.36 0.71 Range 0.4-19.6 0.6-27.4 0.9-22.5 0.8-10.8 4.0-44.8 0.4-22.5 0.6-27.4 0.4-27.4 Maximum distance 305.8 377.8 391.4 402.6 408.5 336.4 390.7 355.7 across activity area SE 36.6 45.5 74.9 48.2 158.5 36.3 37.8 27.0 Range 93-881 116-636 95-1111 178-674 250-567 93-1111 116-674 93-1111 Distance /day 29.6 50.0 21.7 19.5 9.2 26.7 31.1 28.3 SE 8.5 17.0 8.7 5.3 6.2 4.9 9.9 4.7 Range 3-96 6-212 2-126 9-44 3-15 2-126 6-212 2-212 Maximum distance to 200.9 122.7 163.3 85.2 140.1 187.0 103.9 166.7 main stream SE 17.3 35.5 41.1 12.5 39.3 22.9 18.5 18.8 Range 32-545 39-340 13-525 27-152 32-476 13-545 27-340 13-545 Number of points 7.63 6.00 6.21 5.91 6.50 7.11 5.95 6.69 SE 0.38 0.47 0.58 0.73 2.50 0.35 0.43 0.28 Range 5-12 4-9 3-11 3-11 4-9 3-12 3-11 3-12 N 24 10 14 11 2 38 21 59 42 Table 1.2

Morphometrics ANOVA results Results of two-way ANOVAs comparing spatial metrics of adult Wood Turtles in Virginia and West Virginia using data obtained during June-August 2011-2014. Reported below are p values, F statistics, and degrees of freedom. State Sex State:Sex

Activity area p 0.205 0.936 0.523 size F 1.645 0.006 0.414 df 1, 55 1, 55 1, 55

Transformed p 0.119 0.915 0.923 activity area F 2.510 0.011 0.009 size df 1, 54 1, 54 1, 54

Max. distance p 0.261 0.204 0.747 across act. area F 1.290 1.656 0.105 df 1, 55 1, 55 1, 55

Distance moved p 0.125 0.246 0.564 per day F 2.431 1.375 0.338 df 1, 55 1, 55 1, 55

Shape of p 0.667 0.287 0.867 activity area F 0.187 1.157 0.029 df 1, 52 1, 52 1, 52

Max. distance p 0.870 0.0365 0.624 to main stream F 3.038 4.597 0.243 df 1, 54 1, 54 1, 54

43 Table 1.3

Correlations of morphometrics and spatial metrics for adult Wood Turtles Correlations of morphometrics and spatial metrics for adult Wood Turtles captured in Virginia and West Virginia during June-August 2011-2014. Reported are Pearson correlation coefficients with p-values below for ln transformed data; coefficients with significant p-values are in bold. Linear measurements in millimeters obtained with Haglof calipers, mass in grams obtained with Pesola scales. Number of individuals measured varied for different metrics; females n = 23-31, males n = 7- 20. CL = carapace length (using nuchal notch), CW = carapace width at widest axis, SD = shell depth (bottom of plastron to top of vertebrals) at thickest point, PL

= plastron length (using gular and anal notches), BMi = body mass when first weighed at beginning of annual field season, BMe = body mass when weighed at end of annual field season, ∆M = change in end and initial masses (data included turtles that declined in mass, 4 females and 5 males – 21% of 42 total turtles), ActA = activity areas (ha) calculated from MCPs formed from single-year location points of individuals, Dstr = straight-line maximum distance (m) of turtles to main streams, Dday = distance per day calculated from the maximum distance across an activity area (m) divided by the number of days between the two location points used to define the maximum distance. Coefficients in bold were ≥ 0.5 and had significant p-values (≤ 0.05). Females CL CW SD PL BMi BMe ∆M ActA Dstr Dday CL 0.76 0.52 0.79 0.70 0.75 -0.08 -0.13 0.23 -0.19 0.00 0.00 0.00 0.00 0.00 0.71 0.47 0.22 0.30

CW 0.78 0.64 0.83 0.85 0.82 -0.06 -0.19 0.23 -0.43 0.00 0.00 0.00 0.00 0.00 0.79 0.32 0.22 0.02

SD 0.55 0.67 0.53 0.66 0.64 0.09 0.13 0.17 -0.42 0.01 0.00 0.00 0.00 0.00 0.69 0.48 0.35 0.02

PL 0.85 0.77 0.65 0.84 0.85 -0.11 -0.17 0.08 -0.45 0.00 0.00 0.00 0.00 0.00 0.62 0.37 0.67 0.01

BMi 0.84 0.85 0.59 0.82 0.96 -0.14 -0.11 0.14 -0.33 0.00 0.00 0.01 0.00 0.00 0.51 0.56 0.47 0.08 Males

BMe 0.86 0.81 0.61 0.83 0.93 0.03 -0.25 0.05 -0.39 0.00 0.00 0.05 0.00 0.00 0.88 0.21 0.82 0.05

∆M 0.29 -0.24 -0.18 0.42 0.06 0.26 -0.41 -0.50 -0.22 0.52 0.60 0.70 0.35 0.90 0.58 0.06 0.02 0.31

ActA -0.47 -0.47 -0.39 -0.54 -0.56 -0.75 -0.32 0.52 0.44 0.03 0.04 0.09 0.02 0.01 0.01 0.49 0.00 0.01

Dstr -0.44 -0.36 -0.44 -0.45 -0.50 -0.64 -0.29 0.70 0.18 0.05 0.12 0.05 0.06 0.02 0.03 0.53 0.00 0.32

Dday -0.12 -0.16 -0.26 -0.16 -0.22 -0.29 0.16 0.49 0.46 0.63 0.51 0.27 0.53 0.35 0.38 0.73 0.03 0.04 44 Table 1.4

Morphometrics for adult and juvenile Wood Turtles in VA and WV Morphometrics for adult and juvenile Wood Turtles captured in Virginia and West Virginia during June-August 2011-2014. Reported are means and standard errors; linear measurements in millimeters obtained with Haglof calipers, mass in grams obtained with Pesola scales; number of individuals measured varied for different metrics. VFT = Virginia female Wood Turtles (n = 11-25), VMT = Virginia male Wood Turtles (n = 5-15), WFT = West Virginia female Wood Turtles (n = 13-26), WMT = West Virginia male Wood Turtles (n = 9-18), FWT = female Wood Turtles (VA & WV together, n = 28-51), MWT = male Wood Turtles (VA & WV together, n = 14-33), JWT = juvenile Wood Turtles (VA & WV together, n = 6-9), WT = adult Wood Turtles (sexes and states aggregated, n = 42-84). CL = carapace length (using nuchal notch), CLmax = maximum carapace length (using outer edge of marginals), CW = carapace width at widest axis, SD = shell depth (bottom of plastron to top of vertebrals at deepest axis), PL = plastron length (using gular and anal notches),

PLmax = plastron length using margins of gular and anal scutes, BMi = body mass when first weighed at beginning of annual field season, BMe = body mass when weighed at end of annual field season, ∆M = change in end and initial masses (calculated means and errors included turtles that declined in mass, 4 females and 5 males – 21% of 42 total turtles), ∆M% = change in mass as a proportion of BMi.

CL CLmax CW SD PL PLmax BMi BMe ∆M ∆M% VFT 185.8 192.8 146.3 72.8 172.4 185.3 1071.9 1119.1 28.7 2.9 SE 1.3 2.3 1.3 0.7 1.4 2.3 27.1 28.5 10.3 1.5

WFT 181.5 185.3 142.4 72.7 167.7 178.7 969.4 1011.2 22.3 2.2 SE 1.5 1.5 1.0 0.9 1.4 1.4 23.3 38.7 14.8 1.4

VMT 198.1 202.8 148.3 72.8 172.4 181.9 1122.9 1153.0 14.0 1.2 SE 2.1 2.5 1.6 1.0 1.8 2.2 28.6 46.2 14.1 1.3

WMT 195.5 198.3 147.4 72.9 168.3 177.8 1102.8 1118.3 8.9 1.0 SE 3.4 2.5 1.3 1.3 2.8 2.7 37.3 42.4 16.3 1.4

FWT 183.6 187.6 144.3 72.8 170.0 180.6 1015.2 1072.3 25.7 2.6 SE 1.03 1.34 0.85 0.55 1.05 1.31 19.06 24.9 8.7 0.9

MWT 196.9 200.0 147.8 72.9 170.6 179.8 1112.2 1130.7 10.7 1.1 SE 1.92 1.86 0.99 0.79 1.61 1.79 23.64 31.1 11.3 1.0

JWT 127.2 135.3 105.9 51.4 118.6 128.6 354.4 482.5 47.5 10.8 SE 8.1 6.1 4.2 2.8 8.0 6.7 40.4 32.5 12.5 2.4

WT 188.3 193.0 145.7 72.8 170.2 180.3 1053.0 1090.9 20.7 2.1 SE 1.2 1.3 0.7 0.4 0.9 1.0 15.7 19.9 6.9 0.7 45 Table 1.5

Results of two-way ANOVAs comparing morphometrics Results of two-way ANOVAs comparing morphometrics of adult Wood Turtles in Virginia and West Virginia using data obtained during June-August 2011-2014. Reported are p values, F statistics, and degrees of freedom. See Table 1.4 for definitions of variables. State Sex State:Sex

BMi p 0.066 0.00038 0.107 F 3.490 13.953 2.669 df 1,70 1,70 1,70

BMe p 0.127 0.122 0.544 F 2.426 2.501 0.374 df 1,39 1,39 1,39

CL p 0.094 <0.0001 0.418 F 2.874 55.568 0.664 df 1,72 1,72 1,72

CLmax p 0.007 <0.0001 0.449 F 7.930 34.91 0.580 df 1,59 1,59 1,59

CW p 0.149 0.011 0.318 F 2.128 6.830 1.008 df 1,77 1,77 1,77

PL p 0.055 0.764 0.758 F 3.815 0.091 0.096 df 1,71 1,71 1,71

PLmax p 0.0497 0.419 0.454 F 4.032 0.665 0.570 df 1,53 1,53 1,53

SD p 0.253 0.890 0.859 F 1.328 0.019 0.032 df 1,68 1,68 1,68

46 Table 1.6

Results of paired t-tests comparing BMi to BMe Results of paired t-tests comparing BMi to BMe for Wood Turtles measured in Virginia (VA) and West Virginia (WV) during June-August 2011-2014; reported in descending order are p values, t statistic values, and degrees of freedom. BMi = body mass when first weighed at beginning of annual field season, BMe = body mass when weighed at end of annual field season. FWT = female Wood Turtle points, MWT = male Wood Turtle points. Comparisons with significant results are in bold. VA WV FWT MWT FWT MWT BMi –— BMe p 0.0149 0.377 0.166 0.601 t -2.774 -0.994 -1.484 -0.544 df 14 4 11 8

47 CHAPTER 2: DEMOGRAPHIC ESTIMATES AND CONSERVATION IMPLICATIONS

FOR A SOUTHERN POPULATION OF WOOD TURTLES (GLYPTEMYS

INSCULPTA)

Introduction

Births, deaths, and demography directly affect both population dynamics

(Lande and Arnold 1983, Schoener 2011) and evolutionary processes (Futuyma

2010, Harts et al. 2014). Persistence in small stable populations is highly influenced by the evolutionary constraints that have formed the current life history

(Lande 1993, Kinniston and Hairston 2007) and by ecological factors that affect reproductive rates (Dunham et al. 1989). The self-sustainability of populations in the long-term is a function of the current population size (Willi and Hoffmann

2009, Reed and McCoy 2014) and the quality of the habitat relative to the species ecological requirements. Thus, the diminishment, isolation, or fragmentation of animal populations are of high conservation concern (Fahrig 2003, Rivera-Ortíz et al. 2015).

Typically, a large effective population (Ne) is needed to maintain genetic variation (Frankham 2003, Reed 2005); for example, Fridgen et al. (2013) reported lowered genetic diversity in reduced populations of Wood Turtles (Glyptemys insculpta) in southern Ontario. The lower genetic variation present in small populations may diminish a species’ ability to persist through future environmental challenges (Frankham 2003, Reed and Frankham 2003, Traill et al. 2010). This may be of particular concern for taxa such as Testudines, which, due to their generally 48 low genetic variability and slow microevolutionary rate (Avise et al. 1992,

Lourenco et al. 2012, Shaffer et al. 2013), may have limited ability to adapt to the accelerated anthropogenic changes to their environment. In this sense, expansive areas may be serving as population “sinks” or “ecological traps” wherein human modifications of the habitat in which populations evolved occur at a rate faster than the populations can adaptively respond (Quintero and Wiens 2013, Robertson et al. 2013).

Small, isolated, or declining populations are particularly at risk due to three forms of stochastic influences: genetic, demographic, and environmental (Soulé

1987, Lande 1993, Primack 2010). A mutual reinforcement of these biotic and abiotic processes serves to deteriorate population dynamics and collectively drive a population to extinction (Fagan and Holmes 2006). In long-lived organisms such as many turtle species, pervasive deterministic (such as predation or habitat loss) or demographic factors are likely to be greater threats to population viability than are genetic factors (Kuo and Janzen 2004, Pittman et al. 2011). Particularly in long- lived species, currently expressed genetic signals and status may be decoupled from contemporary demographic status (Marsack and Swanson 2009, Fridgen et al.

2013). Hence, genetic data must be coupled with basic habitat and demographic information, such as population size and trends, for effective conservation and management of threatened species (Avise 1995, O’Grady et al. 2004).

The Wood Turtle possesses the life history traits of slow growth, late maturity, high natural mortality of eggs and hatchlings, and low reproductive 49 potential (Lovich et al. 1990, Gibbs and Amato 2000, Ernst and Lovich 2009) that make their populations especially vulnerable and sensitive to increased human- caused loss and mortality. Due to the energetic and demographic implications of these traits, turtle populations may not be able to sustain even modest additive adult take or mortality (Congdon et al. 1993 & 1994, Garber and Burger 1995,

Enneson and Litzgus 2008). High adult survivorship and extreme iteroparity are generally necessary to maintain population viability (Doroff and Keith 1990,

Heppell 1998, Heppell et al. 2000, Mitro 2003, Reed and Gibbons 2003).

Nonetheless, given habitat stability and unchanged vital rates, small (<50) populations of long-lived turtles can persist for long periods of time (Shoemaker et al. 2013).

All of these concerns and impacts underscore the importance of maintaining the ecological integrity and connectivity of undeveloped sites and their intact populations. This study presents demographic information on Wood Turtles sampled over the course of nine years at a relatively undisturbed forested site on the southern periphery of the species’ range in West Virginia USA. The objective was to reveal basic demographic parameters such as population size and growth rate, with the goal of informing in situ conservation.

Focal Species

See “Focal Species” in Chapter 1 for information on Wood Turtles. 50 Methods

Study Area

I studied a Wood turtle population in the George Washington National

Forest (GWNF) of West Virginia. The site is part of the Central Appalachian

Broadleaf Forest – Coniferous Forest – Meadow Province (M221) (ecoregional nomenclature and enumeration sensu McNab et al. 2007). Forests here are within the Oak – Chestnut Region of Braun (1950) and the Appalachian Oak Section of the Mesophytic Region identified by Dyer (2006). This location lies within two orographic rainshadows, from the Blue Ridge Mountains to the east and the

Allegheny Mountains and Plateau to the west. As a result, this is one of the drier locales in the species’ range, with average annual precipitation measuring only ca.

95 cm (NCDC 2013). Elevations in the proximity of the turtle occurrence points are ca. 260-400m asl, although surrounding ridges within the watershed reach altitudes of ca. 850m.

The numerous ephemeral or intermittent drainages feeding the small main stream at this montane site are upper-elevation headwaters of the Potomac River drainage basin of the Chesapeake Bay watershed. The portion of the main stream within the study site on the National Forest is ca. 3.5km in length and is a low gradient 1st – 3rd order stream with pebble-cobble-boulder substrates and riffle-run- pool habitats. Summer water depths range from 3-80 cm, with channel widths of 1-

5 m. The stream channel is generally narrowly entrenched and sharply incised, with close by steep slopes and some flats up to ca. 125m wide (Fig. 2.1). 51 The forest at the study site is relatively undisturbed, with forest stands >100 years of age predominant. One old unpaved logging road enters the site. The most recent logging on the National Forest in the general vicinity took place around fifty years ago. Agricultural uses occur on privately owned lands up- and downstream

(with the closest about 400 meters from the study site). Many stands at the study site are comprised of sub-mesic pine and mixed pine-deciduous forest types: FT10

= White Pine/Upland Hardwoods, 33 = Virginia Pine, 42 = Upland

Hardwoods/White Pine, 45 = Chestnut Oak – Scarlet Oak – Yellow Pine, 52 =

Chestnut Oak, 53 = White Oak – Northern Red Oak – Hickory (stand designations, nomenclature, and enumeration as per USDA Forest Service inventory data).

Field Procedures

I searched for Wood Turtles at the study site at least once per annum from

2006-2014. I searched on foot following a meandering route that criss-crossed perpendicular to the stream along its length. Both aquatic and terrestrial habitats were searched, seldom more than 200 meters from the stream’s banks. Searches occurred when time permitted and when weather conditions promoted turtle activity throughout the day.

I searched for Wood Turtles on 3 dates in 2006 (September 28, October 21,

November 11), 6 dates in 2007 (March 15, April 21, June 12, August 11,

September 24, December 10 [0 turtles identified]), 4 dates in 2008 (April 25, June

28 [0 turtles observed], September 17, November 2), 3 dates in 2009 (April 13,

June 18, August 18), 1 date in 2010 (July 29), 5 dates in 2011 (June 19, July 11-14), 52 3 dates in 2012 (March 23, June 15, June 22), 2 dates in 2013 (July 5-6), and 3 dates in 2014 (May 20, May 31, June 1). I visited the site on numerous additional days in June-August 2011-2014 conducting other fieldwork and spent little time searching for turtles. During these occasions some additional turtles were observed, photographed, identified, and recorded.

Carapace length and width, plastron length, and shell height were measured with a straightedge ruler or, in later years, with Haglof™ bracket calipers accurate to 1 mm. Turtles were weighed with Pesola™ 41000 or 42500 spring scales accurate within 10 grams. Genders were determined by external morphological features (e.g., presence or absence of plastron concavity, length and thickness of tail, head width). Using magnified digital photographs, I estimated age by counting annuli (growth rings) on three costal scutes (assumption: 1 annulus/year); some older individuals (>20 years of age) were too worn for accurate counts. Turtles were considered to be adults when presenting at least 14 annuli or a straight-line carapace length of at least 160mm (Akre 2002, Akre and Ernst 2006).

Visual examinations of unique plastral blotch patterns and scute shapes on digital photographs were used to distinguish and identify individuals (Fig. 2.2).

After processing (~15 minutes), all captured turtles were immediately released unharmed at the point of capture.

Analytic Procedures

Observations of individual turtles were compiled using year as the sampling period in analyses. Because no juveniles were recaptured, only adult data were 53 used in the analyses (see, e.g., Daigle and Jutras 2005). The adult turtle capture- recapture data were split by sex. The sex ratio of adults was compared to a 50:50 sex ratio with a Chi-square goodness-of-fit test. I used the program MARK vers. 7.1 to model/estimate survival and recapture parameters, population size, and the geometric rate of population growth (λ).

I analyzed the nine-year adult mark–recapture histories by year in MARK by using a hierarchical model testing approach to examine sex-specific apparent survival (ϕ) and capture probability (ρ) of marked adult animals in open population

Cormack-Jolly-Seber models (allowing births/deaths and immigration/emigration).

The parameter ϕt is the probability that an individual alive at time t will also be alive and in the population at time t+1 (apparent survival). Emigration off the study site results in apparent survival being “true survival” times the probability that the animal remains on site (fidelity probability). Apparent capture (or encounter) probability (ρ) is the product of the probability an individual is available for encounter (has not temporarily emigrated) and the true probability of detection

(Cooch and White 2013). As the t throughout this analysis refers to time in years, ϕ is an annual apparent probability of survival and ρ is an annual apparent probability of capture. The size of the study area was constant and equal catchability of marked and unmarked animals was assumed (and validated by

RELEASE Test 2).

I examined models that included different combinations of fixed (i.e., constant or non-time-dependent, signified by a “.”) or time-dependent (signified 54 with a “t”) probabilities of capture (ρ) and survival (ϕ), sometimes including variation between sexes (denoted by “sex” in parentheses). The best-fit model was

determined by comparing AICc values of all the models, the model with the lowest

AICc value being ranked as best. I considered models with AICc or QAICc scores within two of the top score to be well supported.

Results from the above analysis were then used to estimate population size by running similar models in the POPAN routine in MARK. Model parameters included capture probability (ρ), apparent survival (ϕ), and probability of entrance

(“pent”) into the population. Constant or time-dependent survival and capture parameters were used, with and without sex differentiation. In all models “pent” was variable with time.

I used Pradel survival and lambda models in MARK to estimate seniority (γ) and realized population growth rate (λ). Pradel models assume that an animal can enter the study on any occasion (Cooch and White 2013). Seniority (gamma) is a form of hindcasting in which transitions among capture occasions are examined backwards in time; Pradel estimations of γ are equivalent to analyzing reverse

capture histories with CJS models (Mitro 2003). The parameter γt is the probability that an individual alive at time t also was alive and in the population at time t-1. In

Pradel models, realized λ is the growth rate only of the age class represented in the encounter history database, in this case, adults exclusively. Therefore, this realized lambda might not be equal to the growth rate of the population. Realized λ is equal

to the ratio of abundances in successive time steps (i.e., Nt+1/Nt); however, it is also 55 equal to the ratio of apparent survival and seniority, as well as to the sum of apparent survival and recruitment (f):

$%&' ∅% ! = = = ∅* + ,* $% )%&'

The Pradel models were parameterized with various combinations of constant or time dependent or sex specific survival, capture probability, and gamma.

The MARK module RELEASE was used to examine model goodness-of-fit, dispersion, and variance inflation factors (ĉ) used for adjusting the likelihood term, yielding the quasi-likelihood adjusted QAICc. Test 2.C addresses capture heterogeneity; it tests the CJS assumption that all marked animals should be equally

‘detectable’ at occasion i+1 independent of whether or not they were captured at occasion i. For estimates of abundance in open populations, both marked and unmarked animals must have the same probability of capture (Cooch and White

2013). Test 3 deals with differential survival; it tests the assumption that all marked animals alive at time i have the same probability of surviving to i+1.

Results

Field Census

More than thirty field searches in the period 2006-2014 resulted in the identification of 72 individual Wood Turtles at this study site: comprising

32F:27M:13J, or 44.4% female, 37.5% male, and 18.1% juvenile (Table 2.1).

Morphometrics for a subsample of these Turtles are reported at Table 2.2. The 72 56 individuals were captured a total of 138 times. The fifty-nine adults were captured

125 times, with thirty-two individuals recaptured at least once; of the sixty-six total recaptures, all were adults. Of the 59 adults, 54.2% were female and 45.8% male.

Nineteen of the thirty-two Turtles recaptured were female (59.4%) and 13 were male (40.6%). However, a higher proportion of the individual recapture incidents were for males than females: 57.6% (38 of 66) of the total recaptures were males, with 42.4% (28 of 66) being female. Of the 27 adult males, 48.1% (thirteen) were recaptured at least once. Of the 32 adult females, 59.4% (nineteen) were recaptured at least once.

Three males were observed six different years and one male was observed five years (see capture history data used for MARK analysis at Appendix 2). Only one female turtle was found in in five different years, with four others being found in three years or more. Genders differed in their overall capture ratios: 27 males were captured a total of 65 times (2.41 captures/male), while 32 females were captured a total of 60 times (1.88 captures/female); the difference was not statistically significant (χ2 = 1.96, 1 df, α = 0.05, P > 0.05). Nor was the observed

32F:27M adult sex ratio significantly different from 50:50 (χ2 = 0.210, 1 df, α =

0.05, P > 0.05).

MARK Analyses

The mark-recapture data best supported CJS models in which both apparent survival and capture probabilities were constant over time (Table 2.3). In the two top models capture probability varied between sexes; the two models differed in 57 that one had male and female survival as equal (the top model), while the other did not. In the three well-supported models capture probability estimates for females ranged from 0.3022-0.4083 (95% confidence interval = 0.1931-0.5078), while those for males ranged from 0.4083-0.5221 (95% CI = 0.3158-0.6598) (Table 2.4).

In the top three models apparent survival estimates for females ranged from

0.8885-0.9423 (95% CI = 0.7232-0.9893) and those for males ranged from 0.8539-

0.8916 (95% CI =0.7386-0.9415) (Table 2.4).

Program RELEASE indicated that CJS assumptions were not violated: there was no significant difference in apparent survival or capture rates between turtles captured previously and those not. RELEASE tests also indicated the capture data were underdispersed (i.e., less variation than expected by chance). Estimates of ĉ derived from χ2 /df calculations were less than 1; the goodness of fit results by group (TEST 2 + TEST 3) yielded a chi-square value of 10.87 with 25 df. This may be a result of high survival probabilities, high capture rates for some individuals, and the somewhat small sample size. It is not clear how to deal with underdispersed data in MARK (“there is lack of unanimity on how to handle ĉ < 1 .

. . set ĉ =1, and ‘hold your nose’”, Cooch and White 2013, pg. 5 - 6). The RELEASE output also indicated there was insufficient data in some test procedures to be able to trust the result.

Estimation of population size by running CJS models in POPAN yielded three well-supported models (Table 2.5). In these three models population sizes for males ranged from 31.44-34.16 (95% CIs = 26.32-43.49), while those for females 58 ranged from 40.75-47.50 (95% CIs = 31.10-72.56). Arithmetic means of these three models yielded a population estimate of ≈ 75 adult Wood Turtles at this site (32.20 males and 42.96 females, or 42.84% male and 57.16% female). Weighted means of twelve models for male and female populations were 32.69 and 43.76 respectively. The estimated adult sex ratio was not significantly different from 50:50

(χ2 = 1.58, 1 df, α = 0.05, P > 0.05).

Pradel models in MARK (Table 2.6), using parameterization of ϕ and ρ similar to the CJS models discussed above, yielded slightly different values for survival and catchability (Table 2.7). In the top four models apparent survival estimates for females ranged from 0.8493-0.9479 (95% CI = 0.7621-0.9899) and those for males ranged from 0.8493-0.8897 (95% CI = 0.7479-0.9391). In the top four models (c-hat=1.0) (using means of individual time-dependent estimates for the time-dependent model) apparent capture probabilities for females ranged from

0.2903-0.3268 (95% CI = 0.1898-0.4490) and those for males ranged from 0.5035-

0.5253 (95% CI = 0.3742-0.6577).

The data best-supported four Pradel models with both constant survivorship and seniority (γ) (Table 2.6). The two top models differed in that one had sex- differentiated survival and the other did not. One of the four models had time dependent capture; this (the 3rd-best model) yielded a suspect estimate of female gamma (i.e., a value of 1.00). For the other three models, seniority values for females ranged from 0.7677-0.8142 (95% CIs = 0.6651-0.8745), while those for males ranged from 0.8074-0.8414 (95% CIs = 0.7288-0.9121) in the four models 59 (Table 2.7). The weighted averages of twelve models for female gamma values were 0.8430-0.8461 (95% CIs = 0.7087-0.9222). The weighted averages of twelve models for male gamma values were 0.8430-0.8461 (95% CIs = 0.7087-0.9222).

From the estimated survival and seniority parameters, the Pradel models also derived the realized growth rate (λ) (Table 2.7). In the top four models, lambda values for females ranged from 0.8493-1.1643 (95% CIs = 0.7770-1.2760), while those for males ranged from 1.0102-1.1020 (95% CIs = 0.9048-1.1772). The geometric mean for female λ of the top four models is 1.0596. The geometric mean for male λ of the top four models is 1.0526. The weighted averages of twelve unadjusted models for female λ were 1.0724-1.0867 (95% CIs = 0.7433-1.4301).

The weighted averages of twelve unadjusted models for male λ were 1.0380-

1.0524 (95% CIs = 0.7796-1.3251).

Discussion

My population estimate of ≈ 76 (95% confidence interval of 58-95; Table

2.5) is low compared to size estimates generated for some other sites in the species’ range (Table 2.8). The mean estimated population size at twelve other locales was

≈ 189. However, some of those estimates included juveniles; mean adult population size at the seven studies that did not include juveniles was ≈ 118. The estimated population size of 76 is close to the median value (85) for the twelve cited Wood Turtle studies (Table 2.8). That different methods were used to generate estimates, such as Lincoln-Peterson or Schumacher & Eschmeyer, must be considered when comparing values and likely contribute to the variation in 60 estimates among populations. In addition, the presumed “populations” were estimated along different lengths of stream as well as along streams or rivers of different orders or size. The annual survival estimates for Wood Turtles in this population were lower than estimates in the literature for some emydids, but higher than those for others (Table 2.4). Importantly, most of the population growth rates

(λ) generated with Pradel models were greater than 1 for both males and females

(Table 2.7).

The morphometrics of the adult Wood Turtles at this site were congruent with previous studies. Intermediate-sized turtles are found in the southern part of the taxon’s range, with the smallest turtles found in the center and the largest in the north (Greaves and Litzgus 2009).

Using the adult field census numbers of 27 males and 32 females and the

formula Ne= 4NmNf/(Nm + Nf), the effective population size (Ne) for this site is 58.6.

Using MARK estimated population numbers of 32.69 males and 43.76 females, the

Ne is 74.85. Given the somewhat low estimated size of this Wood Turtle population, loss of genetic diversity through drift is a legitimate concern. However,

the true Ne may be underestimated and gene flow into this local population may be occurring; in 2015 a radio-tracked adult male Wood Turtle from ca. 10km away in

Virginia relocated to this study site. Even a small amount of gene flow can maintain substantial genetic variation (Jamieson and Allendorf 2012). That Wood Turtles express a social hierarchy, which may result in non-random mating and 61 reproductive success (Kaufmann 1992), further complicates genetic diversity within this population.

The minimum density of Wood Turtles necessary to maintain population viability at a site is unknown. Here, estimated density is 0.65 turtles/ha using the total census population size of 72 and a site area of 110ha (area calculated with a

200m bufferzone on both sides of the stream); using an estimated adult population size of 76 and 110ha, density is 0.69 adults/ha or 1 adult/1.45ha. The estimated density of Wood Turtles here is higher than eight of the thirteen other study sites

(Table 2.8). Populations of terrestrial turtles generally have low densities (<15 individuals/ha) (Luiselli 2006). Nonetheless, population density can still be a factor of concern for chelonians (Belzer 2000) due to potential component Allee effects, viz., difficulty in finding a mate (Berec et al. 2007). In addition, because of their popularity in the pet trade anthropogenic Allee effects could also impact Wood

Turtles; depending upon individuals’ market value, even large populations can be in danger of extinction (Holden and McDonald-Madden 2017). Late-maturing chelonians such as Wood Turtles appear to lack a density-dependent demographic compensation, meaning increased reproductive output in response to a decreased population density (Brooks et al. 1991, Galbraith et al. 1997). So, even after the cause of decline is removed, sparse populations can continue to decline (Strayer et al. 2004); in other words, there is an “extinction debt” from past harms, degradations, and diminishments to populations and habitat (Tilman et al. 1994,

Vellend et al. 2006). As to how Wood Turtles find mates, it is uncertain whether 62 visual, olfactory, auditory, or a combination of cues are used. Because Wood

Turtles’ courtship generally occurs in the spring and fall when they are concentrated in and around the aquatic habitats of relatively small area (Akre 2002,

Krichbaum pers. obs.), finding or encountering mates is more likely at these times than when dispersed on land in the summer.

The ratio of adult female turtles to adult males was not significantly different from 1:1 here. Similar findings were reported for other populations in Michigan

(Harding and Bloomer 1979), New Hampshire (Tuttle 1996), New Jersey (Farrell and Graham 1991), Vermont (Parren 2013), West Virginia (Niederberger and Seidel

1999), and Quebec (Walde et al. 2003). Female biased populations were documented in Michigan (Remsberg et al. 2006) and two sites in Ontario (Quinn and Tate 1991, Brooks et al. 1992).

At this site 18.1% (13 of 72) of censused Wood Turtles were juveniles.

Similarly, Spradling et al. (2010) reported 16% of turtles to be juveniles/subadults at their WV site and 0% in Iowa. At a WV river site juveniles composed 21% of captured turtles (McCoard et al. 2016), while at another WV site juveniles composed 46% of captured turtles (Niederberger and Seidel 1999). Juveniles comprised 12.8, 14.1, and 29.2% of censused turtles at three nearby Virginia sites

(the second site being most similar in habitat to the site reported here and only ca.

5km away) (Akre and Ernst 2006). In Canada, 17.3% of turtles were juveniles at a site in southern Quebec in 1995 and 8.3% in 2002 (Daigle and Jutras 2005). At another Quebec site juveniles represented 26.6% of captures (Walde et al. 2003), 63 while in Ontario, juveniles composed 35% of the census population (Greaves and

Litzgus 2009).

Adult Wood Turtles are strongly philopatric over extended time periods, with an individual often occupying somewhat the same home range over multiple years (Quinn and Tate 1991, Kaufmann 1995, Ernst 2001a, Arvisais et al. 2002 &

2004, Tuttle and Carroll 2003, Akre and Ernst 2006, Jones 2009, Parren 2013,

Krichbaum unpub. data). This high degree of site fidelity militates against there being a high number of adult “transients” in the sampling counts at this GWNF study site.

Just as indicated by the raw field census numbers and ratios, the capture history data best supported MARK CJS and Pradel models with capture probabilities different between the sexes (Tables 3 & 6). Estimates of capture probabilities for males were consistently higher than those for females (Tables 4 & 7). This discrepant detectability may be due to differences in habitat use, with males generally being more spatially concentrated in and around streams (Tingley et al.

2010) and therefore perhaps more likely to be detected. In June-August of 2011-

2014 at this WV site the mean distance from the main stream of female locations was 93.3m ± 14.5, while that for males was 54.0m ± 13.3 (Table 6.6). As different amounts of time were spent searching for turtles in different years, capture probability (p) might also vary over time; some of the well-supported models reflect this.

Annual survivorship estimates for Wood Turtles in this study (0.8187-0.8916 64 for males and 0.8475-0.9479 for females) are comparable to estimates reported for other non-marine chelonians. Wood Turtles are long-lived organisms and my models with constant adult survivorship are well supported (Tables 3 & 6).

Estimates of annual survivorship for females were consistently higher than those for males (Tables 4 & 7). Congruent with other studies of freshwater and terrestrial turtles, the turtles here of both sexes display high annual survivorship (> 0.8). Still, the survival probabilities generated by the CJS and Pradel models may be somewhat low, particularly those for males. Survivorships for Wood Turtles

(including juveniles and not differentiated by sex) at three different sites in Virginia were estimated to be 0.921, 0.916, and 0.808 (the second site being most similar to this GWNF site and ca. 5km away) (Akre and Ernst 2006).

For the Diamond-backed Terrapin (Malaclemys terrapin) in Rhode Island apparent survival of breeding females declined over a ten-year period from 0.959 to 0.944 (Mitro 2003). Calculated annual survival of western Nebraska Ornate Box

Turtles (Terrapene ornata) over nineteen years varied from 0.723-0.944 for adult males and 0.810-0.965 for adult females (Converse et al. 2005). For Florida Box

Turtles (T.carolina bauri) at the southernmost limit of their range, apparent survival estimates were similar for females (0.56) and males (0.67) (Verdon and Donnelly

2005). For northern California populations of the Western Pond Turtle (Actinemys marmorata), Ashton et al. (2011) used MARK to calculate annual survival of adult males as 0.956-0.968 and adult females as 0.964-0.973. From other published studies, Heppell (1998) derived mean annual adult survival rates from pre-breeding 65 age-based Leslie matrices (only females are modeled in life tables). The values ranged from 0.76 for Chrysemys picta and 0.814 for Trachemys scripta to 0.96 for

Emydoidea blandingii and 0.966 for Chelydra serpentina. Annual survival estimates

(not differentiated by sex) ranged from 0.61 for Psammobates geometricus and 0.81 for Trachemys scripta to 0.97 for Chelydra serpentina and 0.98 for Chelodina longicollis (see references cited in Shine and Iverson 1996; the methodologies used for these estimates are not clear).

Assuming the calculated survival estimates accurately reflect this Wood

Turtle population’s status and trend, it is not clear whether this somewhat low annual survivorship is problematic for population stability or persistence. For an apparently declining North Carolina population of the congener Bog Turtle, G. muhlenbergii, Pittman and colleagues (2011) estimated adult annual survival at ca.

0.89 (SE = 0.018, 95% CI = 0.853-0.924). For a Canadian population of Spotted

Turtles, Clemmys guttata, Enneson and Litzgus (2008) concluded that annual adult survival of less than 0.934 resulted in population decline. It is notable that the populations reported by Heppell (1998) of Chelydra (high ϕ) and Trachemys (low ϕ) were said to be declining (i.e., λ < 1), while those of Chrysemys (low ϕ) and

Emydoidea (high ϕ) were increasing (i.e., λ slightly > 1). The survival estimates for

Wood Turtles here, in combination with the lambda estimates, suggest that this population is stable.

Though the annual survival estimates for this Wood Turtle population may be somewhat low, most of the multiplicative population growth rates (λ) generated 66 from the same data are greater than 1 for both males and females (Table 2.7).

Nonetheless, though these rates suggest viability, this could be a “ghost population”. A ghost population is one in which adults are surviving from year to year, but there is insufficient reproduction and/or recruitment to maintain population viability (Vellend 2004). Combining the estimates of the top models for males and females (geometric means of four values) returns λ values of 1.0526 and

1.0596 respectively, indicating population growth. In addition, the presence of thirteen juvenile turtles of varying ages and sizes (Table 2.2) that were not included in the estimates of demographic metrics suggests recruitment may be taking place.

So, this may be a growing, albeit small, population of Wood Turtles. However, even with an overall λ estimated to be greater than 1, populations may still decrease in size over time due to variance in the actual yearly rates (Converse et al.

2005). In addition, patterns of abundance can be a misleading indicator of habitat quality (van Horn 1983).

Many populations/colonies of Wood Turtles are already small (Table 2.8), which means their persistence is already at risk (O’Grady et al. 2004). Such populations may not at present be robust enough to be considered self-sustaining over the next 50-100 years, at least by the standard of the so-called 50-500 rule

(Traill et al. 2010). Fagan and Holmes (2006) examined declining populations of ten different vertebrate species, including two populations of Wood Turtles. They found that the times to extinction for these Turtle populations were less than 20 years. The populations began their final decline when composed of 31 and 58 67 individuals. Very small populations of long-lived organisms may nonetheless be considered to be viable populations for shorter time periods. Demographic modeling by Shoemaker et al. (2013) on New York Bog Turtles indicated that colonies with as few as 15 breeding females had a >90% probability of persisting for >100 years if environmental conditions and survival remained relatively constant. Additionally, heterozygosity may not vary with population size (see, e.g.,

Davy 2013 with regards to Clemmys guttata). Clearly, even somewhat small populations such as reported herein could be valuable conservation reservoirs and restoration nuclei, providing for demographic and genetic exchange as well as facilitating range shifts in response to climate change (Shoemaker et al. 2013 &

2014).

For Wood Turtles, there is valid concern about continuing widespread threats to population persistence (Jones and Willey 2015). The species’ natural range, the northeastern United States and southeastern Canada, is a region experiencing intense human population and development pressures (Fulton et al.

2001). At a region-wide scale much Wood Turtle habitat has been taken over by agricultural, industrial, commercial, or residential development (Riitters et al. 2002,

Sanderson et al. 2002). As habitat loss and modification will likely continue as long as the human population and economy grows (Trauger et al. 2003), threats to the

Wood Turtle will not just go away and must be directly addressed if the species is to persist in our landscape. Additionally, perceptions and surveys of the distribution and health of present-day populations can be particularly misleading and deceptive 68 for long-lived species, as they may reflect the historical landscape configuration or condition rather than the present one, e.g., a time of less habitat fragmentation

(Honnay et al. 2005, Vellend et al. 2006, Marsack and Swanson 2009, Willoughby et al. 2013). Most of the land in the Turtle’s range is in private hands where typically the focus is not on chelonian conservation. Consequently, Turtle habitat on National Forests and other public lands is especially significant due to the highly degraded and fragmented condition of habitat elsewhere in their range

(Jones et al. 1997, Riitters et al. 2002, Harper et al. 2005). Since humans have appropriated so much of the landscape, peripheral places that formerly may have been relatively suboptimal habitat may now be the best that are available or be of intensified conservation value (Channell and Lomolino 2000). Moreover, peripheral lineages may harbor unique genotypes that confer greater resistance to climate-change-induced stress (Hampe and Petit 2005). Unlike many other Wood

Turtle locations, this particular site has not had recent logging and is devoid of open roads; older forests and natural disturbances prevail here. The philopatry, survival, and population growth revealed by this study indicate that these habitat conditions should be maintained. With pressures on the species mounting, sites on relatively undeveloped public lands grow increasingly crucial as refugia for the

Turtle. Preserving Wood Turtle populations and habitat in our National Forests and other public lands appears critical for ensuring their long-term survival.

69 Table 2.1.

Annual sampling of WV Wood Turtles Annual sampling and the numbers of Wood Turtles captured. Counts with asterisks include juveniles (total of thirteen with no recaptures). The + indicates that during 2011-2014 turtles were found on additional days in the field not wholly dedicated to searching for turtles. Sampling Total Total new Total New Total New trips captures individual captures adult captures adult turtles of adult males of adult females year males females 2006 3 10* 10* 6 6 1 1 2007 6 16* 13* 8 5 6 6 2008 4 6* 2* 4 0 1 1 2009 3 12 9 8 5 4 4 2010 1 5 1 3 1 2 0 2011 5 18* 10* 9 3 8 6 2012 3 22* 12* 8 3 12 7 2013 2 20 6 8 0 12 6 2014 3 29* 9* 11 4 14 1

totals 30+ 138* 72* 65 27 60 32

70 Table 2.2

Morphometrics for WV Wood Turtles Morphometrics for Wood Turtles at the WV study site; only individuals measured consistently (i.e., with Hagloff calipers and Pesola scales) are included. Lengths, widths, and heights are maximums. Reported are means, standard errors (in parentheses), and ranges. Metrics are in millimeters and grams. Carapace Carapace Plastron Shell Mass length width length height Males (n=17) 198.9 147.3 177.4 72.1 1098 (2.56) (1.38) (2.81) (1.38) (38.10) 176-220 138-160 168-191 62-78 720-1280 Females (n=26; 185.4 142.4 178.6 72.5 969 all nongravid) (1.48) (5.09) (1.44) (0.89) (23.41) 172-202 130-151 165-190 62-78 625-1100 Juveniles (n=7) 131.4 103.6 124.4 49.4 329 (6.99) (4.94) (8.28) (3.44) (48.98) 98-150 80-118 94-144 36-56 120-455

71 Table 2.3

Well-supported Cormack-Jolly-Seber models for apparent survival (ϕ) estimates Summary of model selection for the estimation of apparent survival (ϕ) and capture probability (ρ) for Wood Turtles using Cormack-Jolly-Seber models in MARK. c- hat=1.

AICc ∆AICc AICc Model Number of weight likelihood parameters Model ϕ(.)ρ(sex(.)) 270.339 0.00 0.35955 1.0000 3 ϕ(sex(.))ρ(sex(.)) 270.793 0.45 0.28655 0.7970 4 ϕ(.)ρ(.) 271.647 1.31 0.18702 0.5201 2

Table 2.4

Apparent survival (ϕ) and capture (ρ) probability estimates Summary of apparent survival (ϕ) and capture (ρ) probability estimates (with 95% CI) for Wood Turtles using Cormack-Jolly-Seber models in MARK. ϕ for ϕ for Ρ for Ρ for males Number of females males females parameters Model ϕ(.)ρ(sex(.)) 0.8885 0.8885 0.3340 0.5024 3 (0.81-0.94) (0.81-0.94) (0.23-0.46) (0.36-0.64) ϕ(sex(.))ρ(sex(.)) 0.9423 0.8539 0.3022 0.5221 4 (0.74-0.99) (0.74-0.92) (0.19-0.44) (0.38-0.66) ϕ(.)ρ(.) 0.8916 0.8916 0.4083 0.4083 2 (0.81-0.94) (0.81-0.94) (0.32-0.51) (0.32-0.51) 72 Table 2.5

Well-supported CJS models for estimated Wood Turtle abundance Summary of model selection for the estimation of abundance of Wood Turtles using apparent survival (ϕ) and capture (ρ) probabilities from CJS models run in the POPAN module in MARK; standard errors in parentheses. All models were parameterized with time-dependent probabilities of entry (“pent”). c-hat=1.

Number Number AICc ∆AICc AICc Model Number of of females of weight likelihood parameters Model males ϕ(t)ρ(sex(.))pent(t) 44.31 31.44 351.31 0.00 0.391 1.000 13 (5.91) (2.85) ϕ(t)ρ(sex(.))pent(sex(t)) 42.23 32.92 352.88 1.58 0.178 0.455 18 (5.15) (3.61) ϕ(sex(.))ρ(sex(.))pent(sex(t)) 42.34 32.24 353.12 1.82 0.157 0.403 15 (5.17) (3.35) Weighted average of 43.76 32.69 twelve models (5.83) (3.28) with 95% CI 31.8-55.7 25.9-39.5 73 Table 2.6

Well-supported Pradel models for estimated population growth rate (λ) Summary of model selection for the estimation of realized population growth rate (λ) and seniority (γ) for Wood Turtles using Pradel models in MARK. c-hat = 1.

AICc ∆AICc AICc weight Model Number of likelihood parameters Model ϕ(sex(.))ρ(sex(.))γ(.) 528.72 0.00 0.2886 1.000 5 ϕ(.)ρ(sex(.))γ(.) 530.12 1.40 0.1431 0.496 4 ϕ(.)ρ(sex(t))γ(sex(.)) 530.33 1.61 0.1287 0.446 21 ϕ(.)ρ(sex(.))γ(sex(.)) 530.43 1.71 0.1227 0.425 5

Table 2.7

Estimated population growth rate (λ) and seniority (γ) values Summary of values (with 95% CIs) for realized population growth rate (λ), seniority (γ), apparent survival (ϕ), and capture probabilities (ρ) for Wood Turtles using Pradel models in MARK. λ for λ for γ for γ for ϕ for ϕ for p for p for females males females males females males females males Model ϕ(sex(.))ρ(sex(.))γ(.) 1.1643 1.0448 0.8142 0.8142 0.9479 0.8507 0.2903 0.5253 (1.06- (0.95- (0.73- (0.73- (0.77- (0.75- (0.19- (0.39- 1.26) 1.14) 0.87) 0.87) 0.99) 0.92) 0.42) 0.66) ϕ(.)ρ(sex(.))γ(.) 1.1020 1.1020 0.8074 0.8074 0.8897 0.8897 0.3169 0.5111 (1.03- (1.03- (0.73- (0.73- (0.81- (0.81- (0.22- (0.37- 1.18) 1.18) 0.87) 0.87) 0.94) 0.94) 0.44) 0.65) ϕ(.)ρ(sex(t))γ(sex(.)) 0.8493 1.0102 1.0000 0.8408 0.8493 0.8493 0.0182- 0.2939- (0.78- (0.90- (0.99- (0.73- (0.76- (0.76- 0.4754 0.6619 0.92) 1.12) 1.00) 0.91) 0.91) 0.91) (0.00- (0.07- 0.69) 0.91) ϕ(.)ρ(sex(.))γ(sex(.)) 1.1570 1.0556 0.7677 0.8414 0.8882 0.8882 0.3268 0.5035 (1.04- (0.96- (0.67- (0.74- (0.81- (0.81- (0.22- (0.37- 1.28) 1.15) 0.85) 0.91) 0.94) 0.94) 0.45) 0.64) 74 Table 2.8

Reported demographic information for Wood Turtles Demographic information reported for Wood Turtle populations. Population estimates with a “j” include juvenile turtles. Densities denoted with an * were estimated by this author using published population sizes and using a 200 meter zone on both sides of the stream (i.e., 400m total width) to calculate the area of available habitat. C-J-S = Cormack-Jolly- Seber, J-S = Jolly-Seber, L-P = Lincoln-Peterson. Census Sampling Estimated Method of Density Stream Researcher numbers duration population estimation (turtles/ha) length Location (F:M:J) (years) size (km) WV 32:27:13 9 76 C-J-S 0.70* 3.5 Krichbaum (this study) VA 44:38:12 3 73 J-S 1.14* 1.6 Akre & Ernst 2006 VA 32:23:9 8 44 J-S 0.27* 4.1 Akre & Ernst 2006 VA 42:43:35 3 88 J-S 0.85* 2.6 Akre & Ernst 2006 WV Not 3 331 L-P 0.33* 25 Spradling et al. reported 2010 WV 49:52:86 2 287 j Schnabel 4.22* 1.7 Niederberger & (~2.28 Seidel 1999 adults) WV 88:137:59 2 793 j J-S 1.45* 13.7 McCoard et al. 2016 PA 39:28:21 6 159 Schumacher & 0.66 NA Ernst 2001a Eschmeyer IA Not 6 77 L-P 0.16* 12 Spradling et al. reported 2010 WI 15:8:1 4 Not na 0.19* 3.2 Ross et al. reported 1991 NH 29:17:36 1 82 j capture 2.6 na Tuttle 1996 Ontario 21:15:19 2 56 j L-P 0.21* 4.5 Greaves & Litzgus 2009 Quebec 188 2 238 j L-P 0.44 7.8 Walde et al. 2003 Quebec 23:20:9-> 2 52.4 -> 25.6 J-S 0.23 -> 5.7 Daigle & Jutras 11:11:2 0.11* 2005 75 CHAPTER 3: THERMAL PATTERNS OF WOOD TURTLES (GLYPTEMYS

INSCULPTA) AND THEIR ASSOCIATED MICROHABITATS IN MONTANE

FORESTS OF VIRGINIA AND WEST VIRGINIA

Introduction

Organisms can either contend with or exploit their environment to optimize their physiological states and in turn enhance their fitness (Huey and Stevenson

1979, Tracy and Christian 1986, Blouin-Demers and Weatherhead 2008). For example, reptiles often do not passively respond to environmental temperatures

(thermoconformity), but instead, when confronted with a variety of thermal

microclimates, actively regulate their body temperatures (Tb) by utilizing extrinsic heat or cold sources to exploit rates of heat loss or gain (Bogert 1959, Hutchison

1979, Sturbaum 1981, Seebacher and Grigg 1997). Body temperatures can be

selected that optimize various functions (Topt), such as digestion, locomotion, minimization of metabolic costs, or procurement of mates (Huey and Stevenson

1979, Grigg and Kirshner 2015, Gunderson and Leal 2016). Prediction of Topts can

be based upon critical thermal maxima (CTmax), preferred temperatures (Tp) in

laboratory thermal gradients, or estimated Tb of active individuals in the field (Huey et al. 2012).

Defining ambient or ‘environmental temperature’ (Te) in the field can be problematic (Bakken and Angilletta Jr. 2014). To capture the complexity of ambient

temperatures, operative temperature (To) is a measure of the microclimatic heat available to an organism from the interplay of conduction, convection,

76 evaporation, and radiation resulting from such environmental factors as topography, wind speed, cloud cover, air temperature, substrate, vegetation, and time of day (Crawford et al. 1983, Huey et al. 2012, Bakken and Angilletta Jr.

2014). Crawford et al. (1983) found substrate temperatures to be the best single

predictor of Tos available to ectotherms. For ectotherms that inhabit heterogeneous macrohabitats, such as this study area, an individual’s morphology, physiology, and behavior (e.g., selection of microhabitats) serve as filters to transfer the environmental thermal regime into its actual body temperature (Huey 1991).

Along with alteration of activity time or body posture, site selection is a well-known method of behavioral thermoregulation and for ectotherms may be a primary driver of habitat selection (Meek and Avery 1988, Seigel and Collins 1993,

Row and Blouin-Demers 2006, Rowe and Dalgarn 2009, Kapfer et al. 2010, Huey et al. 2012). Body temperatures higher or lower than air or surface temperatures are often attained by shuttling, i.e., moving in and out of sunlit and shaded areas (see, e.g., the Bowsprit Tortoise (Chersina angulata) in Branch 1984). Different microhabitats available within a home range can have widely divergent microclimatic regimes (Adolph 1990, Huey 1991, Converse and Savidge 2003,

Farrallo and Miles 2016). The vegetational structure of microhabitats can be a primary driver of thermoregulatory conditions, and hence activity or habitat selection patterns (Reagan 1974). For example, the attributes (surface temperature, relative humidity, and understory plant cover) most important for defining the microhabitat of Eastern Box Turtles (Terrapene carolina) were related to

77 thermoregulation and minimizing water loss (Penick et al. 2002, Rossell et al.

2006).

This study investigates summer use of terrestrial habitat by North American

Wood Turtles (Glyptemys insculpta) at two forested montane sites in the Ridge and

Valley physiographic province of Virginia and West Virginia. To assess relevant thermal factors, I examined three types of environmental and turtle temperatures.

First, to examine available thermal diversity, I placed iButtons in three different microhabitat types at arrays throughout two study sites. Data were obtained in the middle of summer during the 2013 post-nesting period, the time when the potential for thermal stress can be expected to be greatest. I hypothesized that significant differences would occur between microhabitats and predicted that under litter temperatures would be significantly lower than exposed surface temperatures.

Second, over the course of four years of fieldwork (June-August 2011-2014), I recorded ground temperatures at the locations of radio-tracked individuals and their associated random points within 30 minutes of each other. I hypothesized that ground-level shaded surface temperatures would not significantly differ between turtle and random sites. And third, I roughly estimated body temperatures with iButtons attached to the bridges of adult male and female Wood Turtles during the summer of 2014. As I had previously observed some differences in behavior and habitat use, I hypothesized that mean temperatures would differ between males and females. Due to the different energetic demands related to reproduction, I predicted females would have higher mean body temperatures than males.

78 Focal Species

See “Focal Species” in Chapter 1 for general background on the Wood

Turtle. Some Wood Turtle populations are near the northern limits of the geographical distribution of Nearctic turtles (Buhlman et al. 2009, TTWG 2017), reaching at least 47° N latitude in Quebec and 46° N latitude in Michigan and

Ontario (Iverson 1992, Arvisais et al. 2002, Walde et al. 2003, Greaves and Litzgus

2009), Hence, G. insculpta is considered a cold-adapted species (Ernst and Lovich

2009, Stephens and Weins 2009). Populations in the Virginias are at the southern periphery of their global range. For these reasons, iButton temperature data were collected in the middle of summer as this presumably would be the period of greatest potential heat stress.

Wood Turtles can be active over a wide range of body temperatures, 3.4-

37°C (Ernst 1986, Farrell and Graham 1991, Ross et al. 1991, Profitt and Chance

2003, Dubois et al. 2009). The lowest water and air temperatures at which

Pennsylvania Wood Turtles were observed to feed when aquatic and terrestrial

were 17.2 and 23°C respectively (Ernst 1986). A mean CTmax of 41.3°C (39.6°-

42.5°) was recorded during laboratory studies (Hutchison et al. 1966). Hence,

Wood Turtles have a “thermal tolerance range” (CTmax – CTmin) of ca. 37°C.

Ernst (1986) measured cloacal temperatures of 205 Pennsylvania Wood

Turtles, mostly in March through June; active turtles made up 50.2% (103) of these records, and had cloacal temperatures ranging from 7.5-30.0°C (mean = 21.0°, sd

= 5.72). Farrell and Graham (1991) reported the Tb of active New Jersey WTs to be

79 3.4°-31.0°C, with a mean of 16.2°C. The 46 cloacal temperatures of 15 Wisconsin

WTs (measured May-August) ranged from 11.9-31.5°C (mean = 23.6°, sd = 4.55)

(Ross et al. 1991). For 368 measurements (April-Dec) of New Hampshire WTs, cloacal temperatures ranged from 3.8°C for a turtle found dormant in water to

32.0°C for one found dormant on land; the mean cloacal temperature for all

activities and months was 23.4°C ± 5.7°C (Tuttle 1996). Summer Tbs for Quebec

WTs ranged from 9.8 to 34.5°C (Dubois et al. 2009). In Iowa, the highest cloacal temperature at which a Wood Turtle was observed as active was 34.3˚C, and at an air temperature of 35.7˚C (Berg 2014).

Studies of Wood Turtles have found generally close agreement between preferred temperatures determined in the laboratory and those observed in field

studies (Ernst 1986, Graham and Hutchison 1979). Tset is the preferred temperature

around which Tb is regulated in a laboratory thermal gradient (Hertz et al. 1993).

Nutting and Graham (1993) found the Tset of Massachusetts WTs to be 27.5°C

(range 21.3-31.9°, sd = ca. 3°). Using Quebec WTs, Dubois and colleagues (2008)

defined Tset as 30°C based on the upper selected Tb (95% percentile) of fed animals in laboratory settings; temperatures around 29-30°C were considered optimal for

Wood Turtle digestive performance and energy acquisition. The efficacy of metabolism depends on the cellular environment that in turn is influenced by the external environment. Changes in temperature can affect biochemical reaction rates, so most organisms function optimally at some relatively limited range of body

temperatures (Seebacher and Franklin 2012). For turtles in general, the Topt for

80 digestive efficiency is 29–30˚C and the Topt for developmental and growth rates is

30–31˚C (Parmenter 1980, Holt 2000). For example, for captive Yellow-margined

Box Turtles (Cuora flavomarginata), an amphibious species that lives in hilly east

Asian temperate forests, Chen and Lue (2008) found optimal temperatures for feeding efficiency and growth rate to be 26-31°C.

Except when nesting or engaged in courtship activities, Wood Turtles are almost entirely diurnal (Ernst 1986, Ernst and Lovich 2009). Wood Turtles that I occasionally retained overnight were never active after dark, nor have I ever observed, aside from nesting females, a free-ranging individual terrestrially active nocturnally (Krichbaum pers. obs.). Though female G. insculpta choose exposed unshaded nesting sites, they avoid heat stress by nesting in the evening, night, or during rainy or overcast conditions (Krichbaum pers. obs.). In New Brunswick,

Canada earliest activity occurred at 6:30 and no movement was recorded after

20:30, with average onset of movement at 11:30 (sd = 61 minutes) (Flanagan et al.

2013).

Methods

Study Area

For detailed description of the area see “Study Area” in Chapter 1, Fig. 1.1,

Tables 5.3-5.5, and Appendix 1.

81 Field Procedures

Microhabitat Temperatures

Surface temperatures were collected by Thermochron iButton data loggers

(Maxim Dallas Semiconductor DS1921G, Sunnyvale, California, USA, accuracy ±

0.5°C). Temperatures were recorded every 30 minutes from July 12, 2013 at 0:01 to August 16, 2013 at 3:01. iButtons were inserted in tied opaque vinyl sleeves and emplaced with flag wire staking through the ends of the vinyl with the flags to the north of the buttons. At each “array site” an iButton was placed in three different microhabitats. One was placed on the surface in an exposed area of soil or litter, unshaded by ground cover (“open” or “exposed”). Another was placed on an area of soil or litter that was shaded by ground cover, either herbaceous, grass, or shrub

(“under veg” or “UV”). The third was placed under leaf litter or coarse woody debris (“under litter” or “UL”). The greatest distance separating iButtons at a single array site was ca. 3 meters. IButtons were only situated in places that the Wood

Turtles could actually occupy, i.e., the microsites were not inaccessible, such as under rocks or in vegetation above the ground.

Aspects of an organism’s environment can be perceived as a resource continuum along which gradients or partitioning may occur (Tracy and Christian

1986). For this study I therefore preliminarily subdivided the organism’s environment at both the Virginia and West Virginia sites into categories optimal

(A), suboptimal (B), and inadequate (C). These categories were based upon my searches for and observations of G. insculpta at the study area over the previous

82 eight years. The “A” sites were locations where I had found Wood Turtles multiple times; the “B” sites were locations where I had only occasionally found Wood

Turtles, perhaps only a single individual; the “C” sites were locations where I had never found any Wood Turtles.

It is possible for iButton temperature data collected in the field to overestimate thermal opportunities because some locations are actually inaccessible because of biotic or abiotic factors (Bakken and Angilletta Jr. 2014).

Although Wood Turtles may not have used some A/B/C locations, all were capable of being accessed. They were available to the Turtles in that all were nearby known occurrence points of various individuals; even the “C” sites were within several hundred meters of known turtle points. There were no topographic or other barriers to prevent access to any array site, so healthy animals were physically capable of reaching them. As they were within 300 meters of the main streams, the array locations were in the general zone used by the turtles.

I measured environmental attributes at each array location (Table 3.1-A/B).

Overstory canopy openness was measured with a spherical densitometer held at breast height, slope with a clinometer, aspect with a compass (followed by a Beers transform), and distance from permanent running water by pacing-off in the field and GIS. Values for the forest type, site index (an indication of site productivity), age (in years), and seral stage (based upon the age of specific forest types) of the stands for each array location were obtained from the attribute tables of US Forest

Service GIS layers opened in ArcGIS v. 9.3 (ESRI, Inc.).

83 Some iButtons were unrecovered, presumably displaced or ingested by wildlife, such as voles (Microtus spp.), Chipmunks (Tamias striatus), or Turkeys

(Meleagris gallopavo). I initially put out 69 iButtons at 23 array sites and obtained usable data from 43 iButtons (23 in WV, 20 in VA) at 20 array sites (9 in WV, 11 in

VA) (Fig. 3.1).

Turtle Temperatures

Data loggers on the shell surface can be used to reliably estimate turtle Tb in the field (Grayson and Dorcas 2004, Chen and Lue 2008). In 2014 I glued iButtons to the bridges (the lateral part of the shell which connects the carapace and plastron) immediately below the marginal scutes of twelve adult Wood Turtles. The iButtons were situated approximately midway between both the anterior-posterior axis and the dorsal-ventral axis. I reasoned that this position would have the closest congruence with core body temperature and be least susceptible to extreme temperatures caused by direct exposure of the iButtons to the sun when turtles basked. Deep body temperatures of basking turtles can be substantially different from those on the carapace surface (McGinnis and Voigt 1971). Depending upon the angle of the sun, the aspect and inclination of the slope at a site, and the posture of the individual turtle, it was possible at times that part of an iButton could be in direct sun. Due to adhesive or transmitter failure, I only recovered 9 usable iButtons from seven males (4 in WV, 3 in VA) and two females (both in VA). The temperatures from VA and WV males were pooled for statistical tests.

84 Ground Temperatures at Turtle and Random Locations

See Chapter 1 “Field Procedures” for general information on radio-telemetry of turtles and data collection.

To capture substrate temperatures, the proxy indicator for operative temperatures, when a Wood Turtle was terrestrially encountered in the field during the summers of 2011-2014m I measured ground surface temperatures with a digital thermometer at a point immediately adjacent to the position occupied by the turtle

(“TempT”), regardless of whether the individual was in the sun, shade, or under litter. When a turtle was not on the surface in the shade I also measured the ground temperature in the shade nearby (“TempGr”), usually within 1-2m of the turtle.

Likewise, I measured TempGr at paired random points (at a random distance and compass heading 23-300m from each individual turtle point).

In 2011-2012 I located the random points immediately after I located the turtles; these random point TempGr were taken within 30 minutes of those taken at the turtle point (mean 29.0 minutes, sd = 15.1). Ground temperatures in the shade were also taken subsequently (“TempVeg”) when I returned to measure habitat characteristics at the turtle and random points (mean 7.5 days after locating the turtle, sd = 5.0). In 2013-2014 I located the random points when I returned to the turtle points to acquire habitat data (mean of 15.0 days later, sd = 8.8). In these latter two years TempVeg at the random point was typically taken within 20 minutes of that taken at the turtle point (mean 15.8 minutes, sd =11.6).

85 For each state (VA and WV) I pooled data from 2011-2012 (TempGr), 2013-

2014 (TempVeg), and 2011-2014 (TempT-TempGr) and tested for differences in temperatures between turtle vs. random points. I also compared temperatures at female turtle points to those at male turtle points, and made comparisons between states.

Analytic Procedures

I used ANOVAs with individual turtle as the random effect, paired t-tests,

Wilcoxon rank sum tests, and Pearson and Spearman correlation tests to analyse data. Natural log and square root transformations were used to normalize some data prior to analysis. A Bonferonni correction was used to adjust p-values when multiple comparisons were used on the same data. Assumptions of normality were tested with the Shapiro method and Levene’s Test was used to test for homogeneity of variances. Statistical tests were accomplished in R (R Development Core Team

2015).

The following acronyms are used herein: VA female Wood Turtles = VFT,

VA female random points = VFR, VA male Wood Turtles = VMT, VA male random points = VMR; WV female Wood Turtles = WFT, WV female random points =

WFR, WV male Wood Turtles = WMT, WV male random points = WMR; for both states combined, female Wood Turtles = FWT, female random points = FRP, male

Wood Turtles = MWT, male random points = MRP.

86 Analysis Rationale

Huey (1982), Seebacher and Grigg (1997), and Gunderson and Leal (2016) provided conceptual paradigms for investigating thermal constraints on ectotherm activity. The “thermal activity window” (TAW) (sensu Gunderson & Leal 2016) is the range of temperatures at which activity occurs, i.e., when the animal is not taking refuge somewhere. This window is narrower than the thermal tolerance range. Different “modes” of activity within the TAW have different thresholds than the thresholds for overall activity; e.g., courtship may occur at different temperatures than feeding. Similarly, “vigour”, how much time an ectothermic animal engages in a certain mode of activity or the rate at which it is performed, is likely linked to metabolic expenditure and therefore is also temperature dependent.

Performance breadth (sensu Huey 1982) reaches a maximum at some region of

optimal temperature (Topt). Estimations of Topt can be based upon the mean Tb of active individuals in the field (Huey et al. 2012).

I posited bounds for different thermal categories with which to calculate proportions of various temperatures available at the different microhabitats as well as proportions of turtle temperatures. The temperature ranges for temperature- dependent activity are not precisely known for the Wood Turtle populations that are the subject of this study. It may be that the broadest behavioral repertoire (the most modes available) occurs under intermediate thermal conditions (Gunderson and Leal 2016). Hence, I used the mean of the diurnal shell temperatures measured on turtles plus/minus one standard deviation to designate a “preferred body

87 temperature range” (PBTR) (sensu Seebacher & Grigg 1997): 17-27°C. During the day turtles are often inactive, quiescent in forms or under LWD or leaf litter; hence, for the TAW I posited a warmer range of temperatures than the PBTR as a basis for analysis. From published information (Ernst 1986, Farrell and Graham 1991, Ross et al. 1991, Tuttle 1996, Dubois et al. 2009), I set the TAW for Wood Turtles at 21-

32°C. Also based on published information (Nutting and Graham 1993, Dubois et

al. 2008), I used 27.5-30.0°C as an approximation for Topt.

At night average temperatures are lower and spatial and temporal thermal heterogeneity decline as well (Figs. 3.2 & 3.3). For Wood Turtles the nocturnal thermal environment can be considered as “unavoidable suboptimal” in that much

of the time an animal cannot obtain Topt anywhere in its activity area (Tracy and

Christian 1986); e.g., in WV Topt was never available nocturnally and TAW temperatures were available only 16% of the time then, but 57% of the time diurnally. Nocturnally Wood Turtles can take refuge in forms under leaf litter or in water that during some nights may have temperatures slightly higher than terrestrial habitats.

I calculated the proportions of temperature readings that were within and

outside the TAW, PBTR, Topt, and CTmax separately for diurnal and nocturnal time periods. Because Wood Turtles are chiefly diurnal, the bulk of my reportage and discussion pertains to temperature data obtained between the hours of 08:00 and

20:15.

88 Results

Microhabitat Temperatures

Diurnal Microhabitat Temperatures

Threeway ANOVAs (state, microhabitat type, and array category as fixed effects) with individual iButton as the random effect found temperatures did not

differ between array category (A, B, C) (F(2,37) = 0.836, p = 0.441), but did differ

among microhabitats (open, UV, UL) (F(2,37) = 15.604, p = <0.0001), with a

marginally significant difference between states (F(1,37) = 3.678, p = 0.063). Because of these results, along with the clear differences in environmental conditions between the two states (e.g., forest types and topography), I subsequently performed separate analyses for each state and the data from the A, B, and C array types were pooled for each microhabitat.

ANOVAs with individual iButton as the random effect found temperatures

differed between microhabitats in both WV (F(2,20) = 13.165, p = 0.0002) and VA

(F(2,17) = 5.180, p = 0.0175). In WV pairwise ANOVAs showed temperatures

differed between UV and UL microhabitats (F(1,13) = 14.396, p = 0.002) and open

and UL (F(1,12) = 36.605, p < 0.00001), but not between open and UV microhabitats

(F(1,15) = 1.134, p = 0.304). Pairwise ANOVAs for VA showed temperatures differed

between open and UL (F(1,1o) = 7.358, p = 0.0218), but not between open and UV

(F(1,12) = 4.354, p = 0.059) or UV and UL microhabitats (F(1,12) = 2.506, p = 0.139).

With the exception of the under litter temperatures, mean temperatures in

VA microhabitats were lower than those in WV, not unexpected given the

89 adiabatic cooling resulting from elevations at array sites in VA being on average

185m higher than those in WV (Tables 3.1 & 3.2). The mean of all diurnal microhabitat temperatures was 0.9°C lower in VA (17,281 readings from 20 iButtons) than in WV (19,549 readings from 23 iButtons) (Table 3.2, Fig. 3.4).

Omitting the UL microhabitats, the mean of diurnal surface temperatures in VA was

22.35°C (sd = 4.95; 12,031 readings from 14 iButtons), and in WV 23.50°C (sd =

5.25; 14,491 readings from 17 iButtons); using the means of the open and UV temperatures (since the microhabitats had different numbers of iButtons) the averages were 22.46°C for VA and 23.52°C for WV.

Although temperatures differed between the two states, some general patterns were similar (Table 3.2). In both VA and WV the open microhabitats had the highest mean temperatures as well as the highest maximum temperatures.

These habitats also exhibited the greatest temperature variance, with the broadest ranges as well as the largest standard deviations. In contrast, the UL habitats exhibited the least variation, with the narrowest temperature range and smallest standard deviations. Temperatures for the iButtons shaded by ground vegetation had means and variation between the above two extremes (Table 3.2 and Figs. 3.3

& 3.5). One of the more salient differences between the two states were temperatures less than 21°C (the lower bound of the posited TAW) in the UV microhabitats; in VA 48.2% of the temperatures were cooler than the TAW, while only 29.7% in WV were (Table 3.3).

90 In both states, open sites had the highest proportion of their temperatures in

the Topt range as well as the highest proportion of temperatures in the >30° C range

(Table 3.3 and Fig. 3.8). Temperatures in WV under litter never reached the posited

Topt range of 27.5-30.0°C, while in VA less than 1% were in this range. In contrast,

4.0% of the under veg temperatures were in the Topt range in VA, while 7.2% were in WV. In WV the under veg microhabitats had the greatest proportion of daily diurnal temperatures in the TAW, while in VA the open microhabitats had the highest proportion. In contrast, in both states the open sites had the lowest proportion of their temperatures in the PBTR and the UL microhabitats had the highest proportion. All these under litter temperatures were toward the lower bound of the PBTR.

Synchronous broad temperature differentials were obtained for different microhabitats at the same array site, e.g., a difference of over 27°C. The maximum single day temperature differential at a single microhabitat was 61°C. The temperature range across all microhabitats during a single diurnal period was usually greater than 15ºC, often more than 20 (see, e.g., Figs. 3.4 & 3.7). The diurnal temperature range under litter was usually less than 5 degrees, often less than 3 (Fig. 3.5). Under litter sites were cooler diurnally and warmer nocturnally than open locations (Tables 2 & 4, Figs. 3.5 & 3.8). During the entire study period the diurnal temperatures recorded by the iButtons spanned 63°C in Virginia and

69°C in West Virginia, with both the lowest and highest temperatures occurring in the open microhabitats (Table 3.2). From 10:30 to 12:30 marked the greatest shift

91 in mean temperatures over any two hour diurnal time period, with rises of 7.1°C

(21.4° to 28.5°C) in VA and 7.3°C (22.4° to 29.7°C) in WV (Fig. 3.5). In both states, the rates of temperature change were far less in the UV and UL microhabitats.

For details on iButton temperatures see Appendix 3.

The great majority of the recorded microhabitat temperatures were below the

species’ CTmax of 39.6°C. Because this study’s sampling took place for 35 days only during July-August, all the recorded temperatures were well above the critical thermal minimum (presumably ca. 3°C). In both states most of the temperatures ≥

CTmax were recorded in the open microhabitats. This circumstance only occurred for <3% of the total diurnal open habitat readings, an average of ca. 20 minutes per day. For the UV microhabitats less than 0.4% of the temperature readings were ≥

CTmax, an average of under 3 minutes per day. With one exception, temperatures ≥

CTmax never occurred diurnally in UL microhabitats or nocturnally in any

microhabitat. It is important to note that even when temperatures ≥ CTmax were

recorded in an open or UV microhabitat, temperatures well below the CTmax were available only a meter or two away in the UL or UV microhabitats (Fig. 3.7). For

details regarding CTmax see Appendix 3.

Spatial Patterns of Diurnal Microhabitat Temperatures

Some consistent patterns emerged in both states regarding diurnal daily and hourly microhabitat means and extremes at array sites (Tables 1, 2, 5). See

Appendix 3 for details of these patterns with respect to aspect, slope inclination, canopy openness, and forest type.

92 Nocturnal Microhabitat Temperatures

Overall mean nocturnal temperatures averaged 3.8°C less than diurnal temperatures (Tables 3.2 & 3.4). The temperature range was also far less nocturnally than diurnally, 16.7°C versus 69.0°C over the entire study period. The daily nocturnal temperature range was usually less than five degrees, often less than three, particularly for under litter microhabitats (Fig. 3.8).

There was much less variance in temperatures between microhabitats nocturnally than diurnally (Tables 3.2 & 3.4, Figs. 3.2 & 3.3). Nocturnally, the under litter mean temperatures were 0.8°C warmer than the open microhabitats in

VA and 0.9°C warmer in WV (Fig. 3.8); in contrast, diurnally the under litter mean temperatures were generally at least 3-4°C cooler than the open microhabitats

(Tables 3.4 & 3.2). Under litter microhabitats had far fewer nocturnal temps ≤

15°C, an order of magnitude difference from the other two microhabitats; temperatures <17°C occurred under litter less than half as frequently as they did in the under veg or open microhabitats (Table 3.6). Just as diurnally, in both states the

UL microhabitats had the greatest proportion of daily nocturnal temperatures in the posited PBTR. In WV the UV microhabitats had the highest proportion of daily nocturnal temperatures in the TAW, while in VA the UL microhabitats had the

highest (Table 3.6). Nocturnal temperatures were never in the Topt range or warmer in any microhabitat in either VA or WV.

93 Correlations of Diurnal Microhabitat Temperatures and Environmental Attributes

Using Pearson or Spearman tests, for each of the microhabitat types I examined the correlation of diurnal mean temperatures to each of six environmental attribute metrics measured at the array sites (canopy openness, transformed aspect, slope inclination, distance to main stream, elevation, forest stand age, and distance to road [only in VA]). With a Bonferonni correction for multiple tests, for each type of microhabitat adjusted significance level for WV was alpha = 0.0083, for VA 0.0071. Using these corrected significance levels, there were no correlations of temperature with environmental attributes in either VA or

WV. The results reported below are those tests with p-values ca. 0.05 or under.

For VA open microhabitats, the only attribute significantly correlated with temperature was amount of canopy openness (Pearson test: t = 2.99, df = 4, p =

0.040, r = 0.831). For VA under veg and under litter microhabitats, no environmental attribute was significantly correlated with temperature; there was a marginally significant correlation with canopy openness for UV microhabitats

(Spearman test: S = 26, p = 0.069, rho = 0.690).

For WV open microhabitats, three attributes were significantly correlated with temperature: canopy openness (t = 2.76, df = 6, p = 0.033, r = 0.748), elevation (S = 13.6, p = 0.009, rho = 0.838), and inversely with stand age (S = 149, p = 0.023, rho = -0.779). None of these three attributes was significantly correlated with the others. For WV under veg microhabitats, there were marginally significant correlations of temperature with transformed aspect (t = -2.19, df = 7, p = 0.065, r

94 = -0.637) and canopy openness (t = 2.17, df = 7, p = 0.066, r = 0.635). Canopy openness and aspect were not significantly correlated. For WV under litter microhabitats, there were significant correlations of temperature with canopy openness (S = 2, p = 0.017, rho = 0.943), distance from the main stream (t = 2.76, df = 4, p = 0.051, r = 0.810), and inversely with stand age (t = -3.46, df = 4, p =

0.026, r = -0.866). There were significant inverse correlations of stand age with distance from the main stream (t = -2.98, df = 4, p = 0.041, r = -0.830) and with canopy openness (S = 65.8, p = 0.021, rho = -0.880). Otherwise, canopy openness, distance from the main stream, and stand age were not significantly correlated.

Surface Temperatures at Turtle and Random Points

Overall, only 10.1% of 337 terrestrial locations of 86 adult Wood Turtles encountered during this four-year study were in the direct sun, while 18.4% were in dappled conditions, 54.9% in shade, and 16.6% were under litter. These proportions, along with the turtle shell iButton data and the lack of difference between concurrent TempTs and TempGrs, are evidence of the turtles’ general affinity here for shaded conditions.

The similarity between temperatures at turtle points and random points

(Table 3.7) corroborates what was indicated by the microhabitat iB data (Fig. 3.15), viz. the broad-scale availability throughout the study sites of temperatures suitable for the turtles. In 2011-2012 (when I located the random points immediately after I located the turtles), surface temperatures in the shade (TempGr) did not differ between turtle and random points for either males or females in WV or males in

95 VA, but did for females in VA (Tables 3.7 & 3.8). In 2013-2014 (when random points were located immediately after the turtle points during vegetation work), surface temperatures in the shade (TempVeg) did not differ between turtle and random points for females in WV or males in VA, but did for females in VA and males in WV (Tables 3.7 & 3.8).

In 2011-2014 TempT did not differ from concurrently measured TempGr for males or females in VA or WV (Tables 3.7 & 3.8, Fig. 3.9). With one exception,

TempTs of turtles sitting in direct sun were below 39.6°C. The mean of twenty- seven TempTs in direct sun was 31.5°C (sd = 3.86, se=0.74). For turtles who were situated under leaf litter or woody debris, TempTs were always well below 39.6°C; mean of forty-seven TempTs under litter was 24.5°C (sd=2.74, se=0.40).

Temperatures under litter differed from concurrently measured surface temperatures in the shade in both VA (paired t-test: t= 5.739, df = 39, p = 0.001) and WV (paired t-test: t= 5.212, df = 19, p < 0.0001); mean temperature under litter = 25.9 ± 0.56, RPs = 25.4 ± 1.00). Temperatures under litter did not differ at turtle points and concurrently measured random points in 2011-2012 (paired t-test: t= 0.701, df = 11, p = 0.498; mean temperature at WT points = 25.9 ± 0.56, RPs =

25.4 ± 1.00) (for temperature under litter tests α = 0.0167).

Of 300 TempTs recorded in 2011-2014, only one was less than 17°C and only one was outside the taxon’s thermal tolerance range, i.e., above 39.6°C.

Temperatures at turtle locations in WV were generally warmer than VA. Only 7.0% of TempTs were >30°C in VA, but 19.3% were in WV (Table 3.9). Of VA TempTs

96 (mean 25.5°C, sd=3.19, se = 0.23), over 78% were within the PBTR, while in WV less than half of TempTs were (mean 27.9°C, sd=3.58, se = 0.34). In WV 32.5% of

TempTs were in the Topt range, while only 13.5% were in VA.

The majority of TempGrs at turtle points were within the PBTR, 43.6% in

WV (mean 26.9°C, sd = 2.85, se = 0.28) and 75.0% in VA (mean 25.3°C, sd =

2.75, se = 0.20). In 2011-2012 only 3.9% of TempGrs were >30°C in VA, but

17.9% were in WV (Table 3.10). All of the TempGrs at both turtle points and

random points were within the taxon’s thermal tolerance range, i.e., below CTmax.

The highest such shade temperatures at turtle location points were 33.5°C on

August 8, 2011 and 35.4°C on July 17, 2013 in WV, and 31.4°C on June 29, 2012 and 32.7°C on July 16, 2013 in VA.

VA 2011-2014

There was no difference between male and female Wood Turtles for TempT

(F(1,42) = 1.687, p = 0.201) or TempGr (F(1,42) = 1.011, p = 0.320). Canopy openness

(Wilcoxon rank sum test: W = 395, p = 0.0367) and gap area (Wilcoxon rank sum test: W = 410, p = 0.0202) were both lower in early successional plots (<35 years old) relative to plots in older (>75 years old) seral stages.

WV 2011-2014

TempT did not differ between adult male and female turtles (F(1,43) = 0.103, p

= 0.750) or TempGr (F(1,40) = 0.094, p = 0.761). TempT also did not differ between

juveniles (n = 12 points), females, and males (F(2,45) = 1.938, p = 0.156), nor did

TempGr (F(2,40) = 2.138, p = 0.131). At turtle points there was no difference between

97 concurrently measured TempT and TempGr for juveniles (paired t-test: t = 1.332, df

= 15, p = 0.203).

VA – WV 2011-2014

Temperatures did not differ between sexes, but did between states. A two- way ANOVA with sex and state as fixed effects and individual turtle as the random

effect found no difference between the TempTs of males and females (F(1,86) =

3.528, p = 0.0637), but did detect a difference between VA and WV turtles (F(1,86) =

36.672, p < 0.0001) (α = 0.0125). A similar two-way ANOVA found no difference

between TempGr for males and females (F(1,82) = 1.734, p = 0.192), but did between

VA and WV turtles (F(1,82) = 17.949, p < 0.0001) (α = 0.0125).

Elevation differed between random points in VA (mean = 528m ± 2.5) and

WV (mean = 343m ± 1.8) (Wilcoxon rank sum test: W = 22560, p < 0.00001). For pooled VA-WV random points, elevation had weak inverse correlations with both

TempGr (Spearman’s correlation test: S = 342348, p-value = 0.00055, rho = –

0.316066) and TempVeg (S = 4702835, p = 0.0132, rho = –0.145). TempGr at VA-

WV pooled random points was weakly correlated with slope inclination (mean slope = 13.9° ± 0.48) (S = 200675, p = 0.014, rho = 0.229).

Turtle Shell Temperatures

Diurnal

During diurnal periods the temperature range was typically no more than

20°C (ca. 15°-35°), whereas nocturnally it was usually less than 8°C (ca. 15°-23°)

(Table 3.11, Figs. 3.10 & 3.11). The mean diurnal temperature of female Wood

98 Turtles was 1.5°C higher than that of males, with a mean of 22.8°C (sd = 4.4, se =

0.06, median = 22.2°C; 5,082 readings) vs. 21.3°C (sd = 4.0, se = 0.03, median =

20.2°; 15,589 readings) (Figs. 3.12 & 3.13). During the middle of the diurnal time period (11:00-17:00) the mean temperature of females was 2°C higher than that of males, 24.6 vs. 22.6°C (Fig. 3.14). A two-way ANOVA with sex and state as fixed effects and individual turtle as the random effect showed no significant difference

between the temperatures of VA and WV turtles (F(1,6) = 0.118, p = 0.743) or

between males and females (F(1,6) = 3.10, p = 0.129). The mean of all diurnal Wood

Turtle shell temperatures (21.7°C, sd = 4.16, se = 0.03; 20,671 readings) corresponded closely to the mean of all diurnal microhabitat temperatures as measured the previous summer by iButtons (22.2°C, sd = 4.72, se = 0.02; 36,830 readings) (Tables 3.2 & 3.10). The daily rise and fall of shell temperatures aligned with the daily rise and fall of UV microhabitat temperatures (Fig. 3.15).

Male turtles had a greater proportion of shell temperatures in the 17-27°C

PBTR than did females, whereas females had a greater proportion of temperatures in the 21-32°C TAW than did males (Table 3.12). Females had greater proportions

of their temperatures in the Topt (27.5-30°) and >30°C ranges than did males, while males had greater proportions of their temperatures in the <21° and <17°C ranges than did females (Fig. 3.16). Perhaps the most salient differences between the two sexes involved temperatures <21° and ≥27.5°C: 17.8% of females’ temperatures were ≥ 27.5°, while only 10.8% of males’ were, and 59.1% of male temperatures were <21°C, while only 41.7% of females’ were. Two other notable patterns

99 involve daily temperature maxima and minima. Minimum daily temperatures of males and females tracked each other very closely, whereas the mean daily maxima for females were consistently higher than those of males (Fig. 3.17). Most individuals of both sexes had cyclical spikes in temperature maxima, with prominently higher temperatures occurring almost daily (Fig. 3.18), however some individuals exhibited peaks only every 3-4 days and some spent long periods at consistently low temperatures (Fig. 3.19).

Nocturnal

Temperatures varied much less nocturnally, with little difference between the sexes; the mean nocturnal temperature of female Wood Turtles was only 0.5°C higher than that of males, 19.1°C vs. 18.6°C (Table 3.11, Fig. 3.10). For both sexes greater proportions of shell temperatures were in the PBTR during nocturnal periods than during diurnal periods, though most of these temperatures were <

21°C (Table 3.12, Fig. 3.20). Nocturnally, greater proportions of temperatures were less than 17°C and 21°C compared to diurnally and there were also no nocturnal temperatures ≥ 27°C (Table 3.11, Fig. 3.20). Male and female turtles had similar proportions of shell temperatures in the PBTR, whereas females, just as diurnally, had a greater proportion of temperatures in the TAW than did males (Table 3.12).

The mean of all nocturnal Wood Turtle shell temperatures (18.7° C, sd = 1.87, se =

0.04; 20,412 readings) corresponds closely to the mean of all nocturnal microhabitat temperatures measured the previous summer (18.4° C, sd = 2.30, se =

0.15; 33,626 readings).

100 Discussion

This study indicates that Wood Turtles at these Central Appalachian forested sites are not as thermophilic as some other chelonian species and live in an environment where it is difficult to exceed their critical thermal maximum. The environment is thermally heterogeneous with substantial variation at fine spatial scales, and thermally suitable sites are not a limiting resource for the turtles in the summer. The results support my three hypotheses/predictions as 1) the under litter microhabitat temperatures were significantly lower than exposed surface temperatures, 2) the paired ground-level shade temperatures measured during the same restricted time period (viz., within ca. 30 minutes of each other) did not differ between turtle and random sites, and 3) females tended to higher mean body temperatures than males, but the variance over a 24 hour period overwhelmed the ability to detect this difference with statistical tests. When examined categorically, females spent more time in warmer temperatures (e.g., 27.5-30°C and >30°C) than did males (Fig. 3.16). The results of this study are generally consistent with those obtained for other North American terrestrial and semi-terrestrial emydids, such as

Eastern Box (Terrapene carolina), Blanding’s (Emydoidea blandingii), and Spotted

(Clemmys guttata) Turtles.

Environmental Temperatures

In stark contrast to a heavily forested tropical area where Huey et al. (2012) found thermal heterogeneity beneath the canopy was limited at all times, thermal heterogeneity was not narrowly limited in these VA-WV forests (see Figs. 3.5 &

101 3.7). Certain microhabitats available to Wood Turtles offered a more stable thermal regime: the ranges and standard deviations indicate thermal conditions under litter were more constant than the more exposed microsites. Conversely, daily temperature spikes often occurred at the open microsites (Fig. 3.7). The spectrum of thermal regimes available in these forests suggests that microsites with usable temperatures are not a limiting resource for Wood Turtles here. Availability of thermal heterogeneity is augmented for an amphibious species that can access streams with water temperatures that remain consistently ca. 18° C during the summer; over the 2011-2014 field seasons, 137 out of 679 total turtle location points (20.2%) were aquatic. It is noteworthy that in WV the UV microhabitats had the greatest proportion of daily diurnal temperatures in the posited TAW of 21–

32°C, while in the higher altitude VA site the open microhabitats had the highest proportion. This suggests Wood Turtle spatial ecology might be constrained at higher elevations – and at higher latitudes – by the availability of more open microhabitats.

The microhabitat iButton data indicate that diurnal temperatures in the Topt range (27.5-30°C) are available in exposed microsites as well as those shaded by ground floor vegetation, while cooler temperatures (almost always under 30°C) are available under nearby litter and woody debris. Thus, any necessary thermoregulatory shuttling can be accomplished at a fine spatial scale, broad-scale movements would seldom if ever be necessary unless a turtle was exposed to site conditions that were drastically altered across large extents, e.g., removal of a high

102 proportion of overstory or understory vegetation across an area as large as or larger than its activity area, such as by logging or burning. These anthropogenic disturbances can substantially change a thermal landscape (Currylow et al. 2012,

Howey 2014). Wood Turtles can use forms under leaf litter not just for cool retreats diurnally, but for warmer refugia at night as well. During cold snaps they could also maximize heat conservation by moving to aquatic locations. They can maximize heat absorption during the day by positioning in direct sunlight in exposed microhabitats (indicated by temperature spikes in the open microhabitats and in the turtle temperatures). The oscillations in diurnal turtle shell temperatures indicate shuttling between warmer and cooler microclimates (Figs. 3.7, 3.11, 3.18).

High temperature is a “stress” that may “strain” the stressed system in response (Bakken and Angilletta Jr. 2014). At the latitude of the present study, perhaps the main diurnal thermal constraint for Wood Turtles is not basking opportunities, but the avoidance of excessive heat gain. The risk of high heat loads can force a restriction of a reptile’s activity time (Grant and Dunham 1988, Kearney

2013). In New Brunswick, Canada when temperatures exceeded 25°C after 14:00

Wood Turtles typically moved to shelter (Flanagan et al. 2013). Avoidance of overheating presumes the existence of vegetation, LWD, or litter that provide the

cover necessary for equable temperatures below CTmax. Ground temperatures are buffered from extremes by plant canopies (Oke 1997), both overstory trees

(Cunnington et al. 2008) and understory grass, herbs, and seedlings (see “under veg” at Table 3.2). Wood Turtles here were only intermittently exposed to

103 temperatures ≥ CTmax during 10:30-15:30, and this was only for a very small amount of time in a few microhabitats. Hence, unlike for some tropical or desert reptiles (Grant and Dunham 1988, Moulherat et al. 2014), at these forested sites thermal constraint upon Wood Turtle activity time and habitat use was minor (Fig.

3.4).

Both natural and anthropogenic disturbances that alter vegetative conditions can influence thermal conditions on the ground (Saunders et al. 1998), though generally natural disturbances may have less of an impact than anthropogenic ones

(Lewis 1998, Saunders et al. 1998). Typically, thinning or removal of the forest canopy results in reduced relative humidity and moisture at the ground surface and increased mean temperature, temperature fluctuations, and solar radiation (Collins et al. 1985). A road, roadside, or newly logged site may not provide buffering plant cover conditions. For example, in oak-hickory forests in southern Indiana Currylow and colleagues (2012) found ground temperatures in exposed recently logged sites to be significantly warmer (as much as 13°C) than forested control sites. They concluded that the summer temperature extremes in the logged sites (0.15 - 4.4ha in size) reduced their suitability for T. carolina and other herpetofauna. Similarly,

the highest proportions of temperatures above CTmax here in my study were at array sites of anthropogenic disturbance, a roadside and a fabricated opening, while the

only VA site that recorded no temperatures ≥ CTmax was in a thickly regenerated 30- years-old clearcut.

104 For an amphibious species with somewhat low preferred body temperatures inhabiting locations such as these Central Appalachian sites that are consistently warm during the summer, individuals can often access uninterrupted warmth during their terrestrial diurnal activities, with little need for basking. The lack of difference between turtle points and random points for any of the ground-level temperatures (TempGr and TempVeg) or between temperatures of the same microhabitats at the different A-B-C array sites indicate that at this time of year sites under forest cover are thermally benign habitats for Wood Turtles, with suitable temperature regimes available throughout the overall study area. Hence, temperature is not a driver of broad-scale spatial ecology here. That VA and WV microhabitat temperatures were in the posited PBTR for high proportions of the time both diurnally (79-96%) and nocturnally (70-88%) (Tables 3.3 & 3.6) is cogent evidence of the study area’s thermal benignity. Wills and Beaupre (2000) formed similar conclusions for the Eastern Timber Rattlesnake (Crotalus horridus) in

Arkansas deciduous and mixed forests. Depending upon the taxa and the spatial scale, nutritional requirements and avoidance of predators may be greater drivers of habitat selection than thermoregulation (Willems and Hill 2009). Though decreases in environmental temperatures can decrease ingestion and digestion rates in turtles (Gianopulos and Rowe 1999), changes in energy requirements wrought by cooler temperatures may also diminish nutrient quality and quantity requirements (Seebacher and Franklin 2012). Therefore, the use of lower temperature microhabitats, such as those under some sort of cover (e.g., litter or

105 woody debris), could result in advantageous lower resting energy expenditure

(Beaupre 1995, Polo-Cavia et al. 2012). As Wood Turtles may have access to foods such as berries, insects, and mushrooms only seasonally or intermittently, perhaps they utilize an energy minimizing strategy whereby cool microhabitats are selected to reduce overall metabolic costs when food is less abundant (such as in mid- to late summer) and energy costs are elevated by high ambient temperatures (see

Penick et al. 2002 regarding Box Turtles).

Turtle Temperatures

The proportions of diurnal shell temperatures in the TAW were consistently less than those in the cooler and slightly more narrowly defined PBTR. Because the turtles here had somewhat low mean shell temperatures (viz., 21.7°C), they exhibited a consistent preference for temperatures lower than the TAW minimum

of 21°C. As a result, the PBTR combined with the Topt more accurately reflect Wood

Turtle thermal behavior at this site than does the TAW (accounting for 89% vs.

44% of diurnal shell temperatures) (Table 3.12). Nonetheless, as the Topt and TAW were postulated with the use of data obtained from Wood Turtles located in

different parts of their range (viz., more northerly turtles), the posited Topt and TAW, as well as the PBTR, may simply not be accurate for this ecosystem and hence of limited utility here. The mean shell temperatures of the turtles in this study, however, did generally align with the mean cloacal temperatures of Wood Turtles from elsewhere: Pennsylvania (Ernst 1986), Wisconsin (Ross et al. 1991), and New

Hampshire (Tuttle 1996).

106 The mean temperatures of male and female Wood Turtles here were far cooler than those of some other reptiles (Figs. 3.13-3.14); for example, a tropical lizard, the Komodo Monitor (Varanus komodoensis), had mean daily diurnal body temperatures of 32.5-33.5°C (Harlow et al. 2010). Though mean female shell temperatures were not significantly higher than those of males, females maintained higher proportions of their temperatures in the upper ranges (>27° and >30°C); e.g.,

a higher proportion of their temperatures was in the posited Topt range (13% vs.

8%). This could be due to the greater energetic needs of adult females, i.e., the increased allocation of energy to reproduction. During the summer periods for which I have body mass data (mean periods of 60 days in Virginia and 72 days in

West Virginia), most females put on mass, more so than did males (2.6% gain vs.

1.1%; Chapter 1). While 80% of VA and 92% of WV females increased in body mass, only 60% of VA and 67% of WV males did.

Mean surface temperatures (TempT and TempGr) at female locations were not higher than those of males in either state, but there were some ostensible inconsistencies between the behavior of turtles located in the field and the temperature patterns revealed by the turtle iButtons. Of 229 female terrestrial locations, 21 (9.2%) were in direct sun, while 13 of 108 (12.0%) male locations were. Moreover, 46 of 229 (20.1%) female terrestrial locations were under litter, while only 10 of 108 (9.3%) male locations were. These two intersexual patterns could lead one to suspect females’ temperatures would be lower than males’, not higher as the iButtons showed.

107 Usually, at this study area the only way for a turtle to access temperatures above 32°C is to be in the direct sun (Table 3.3). That some shell temperatures were above 32°C indicates that Turtles commonly use habitats exposed to the sun; of course, sometimes even temperatures in the direct sun are lower than 32°C and sometimes temperatures in the shade are over 32°C. Though most of the time the turtles were apparently in the shade, the iButton temperatures indicate that for limited time periods they were in direct sun (Figs. 3.17 & 3.18); 3% of male turtle iB temperatures were >30°C, while 5% of female temperatures were. Only 0.7% of male turtle shell temperatures were >32°C, while 1.8% of female temperatures were. So, direct field observations together with the iB data suggest females were using terrestrial thermal extremes (i.e., under litter and in direct sun) more than were males. Overall, only 10.1% of 337 terrestrial locations of 86 adult Wood

Turtles encountered during this four-year study were in the direct sun, while 18.4% were in dappled conditions, 54.9% in shade, and 16.6% were under litter where the mean temperature was 24.5°C. These proportions, plus the lack of difference between concurrent TempTs and TempGrs, as well as the turtle and under veg iB data (Fig. 3.15), are evidence of the turtles’ general affinity here for shaded conditions. Similarly, Ernst (1986) remarked that most of the time Wood Turtles at his forested Appalachian study site in Pennsylvania were either in forms (turtle-size depressions in leaf litter) or foraging in the shade. Eastern Box Turtles during July in western Virginia also tracked temperatures characteristic of shaded forest conditions (Fredericksen 2014).

108 Though female Wood Turtles may bask more frequently prior to nesting

(Ernst 1986), my study took place post-nesting. Female turtles, however, might bask

more and maintain higher Tbs throughout the active season to enhance follicle production (Congdon and Tinkle 1982, Carriere et al. 2008). Conversely, males of

some species may need higher Tb to facilitate increased energy acquisition/processing to accommodate extensive movements searching for mates

(Spencer et al. 1998). Male Wood Turtles may avoid or reduce this constraint by spending more time closer to and in the streams; at some point in time all females will move closer to and enter aquatic habitat. Males were more often aquatic than were females and this could account for their lower mean shell temperatures

(discussed below in “Use of habitat” subsection).

Shell temperatures of both sexes generally exhibited a midday peak. Male temperatures peaked around an hour earlier than did females’, at 14:00 vs. 15:00

(Fig. 3.13). Elevated mid-day body temperatures have been observed in other chelonian species, such as Chrysemys picta (Rowe and Dalgarn 2009) and

(Terrapene ornata) (Plummer 2003). It may be that the intermittent temperature spikes detected here by the carapacial iButtons (Fig. 3.18) were to facilitate some non-continuous function, such as digestion (Regal 1966). Unlike many other ectotherms, such as lepidosaurs, the primary purpose of elevated temperatures or basking in turtles is probably to increase digestive efficiency and energy intake, not locomotor performance (Parmenter 1981, Polo-Cavia et al. 2012). During the summer the temperature a turtle happens to be when it encounters a predator

109 probably has little effect on its relative fitness, unlike a squamate reliant on sprint speed for survival (Christian and Tracy 1981). Hammond et al. (1988) found fed

Pseudemys scripta to bask much more than nonfed ones and fed females to bask much longer than fed males. Temperature, as well as both the quantity and quality of food consumed, can influence the rate of digestion (related to digestive efficiency and the rate of passage along the digestive tract) (Zimmerman and Tracy

1989). Various digestive processes are temperature sensitive; for example, higher temperatures may be needed by herbivorous taxa with gut endosymbionts that digest cell wall components (Zimmerman and Tracy 1989). Congdon (1989) predicted that a turtle that has fed will maintain a body temperature close to the preferred range to attain an optimal rate of digestion. For this and other reasons, some reptiles prefer temperatures in the upper parts of their thermal range, perhaps only a few degrees below their critical thermal maxima (Cloudsley-Thompson

1971).

This was not the case for Wood Turtles here, shell temperatures were

consistently well below the CTmax (39.6-42.5°C), a finding consistent with studies on the taxon from elsewhere (Ernst 1986, Ross et al. 1991, Nutting and Graham

1993, Dubois et al. 2009). The paucity of heliotropic basking I observed or inferred from iButton data may be due to predator avoidance or to the availability throughout these study sites of relatively warm diurnal temperatures at this time of year. Unlike Hammond et al. (1988), Boyer (1965) found nutritional status to have little effect on the amount of time turtles (unidentified aquatic taxa) spent basking;

110 perhaps Wood Turtles are similarly inclined. Impedance of locomotor ability due to a full stomach (Brown and Brooks 1991) is likewise an unlikely explanation for the lack of a significant thermophilic response in Wood Turtles. Though the rate of digestion (based on elapsed time between ingestion and fecal excretion) increases with increased temperature (Zimmerman and Tracy 1989), Wood Turtles may not require elevated temperatures post-feeding. For Snapping Turtles (C. serpentina) there may be shifts between metabolic rates and temperature such that enhanced digestion may occur at low temperatures (Brown and Brooks 1991).

Diel variation in the Tb of ectotherms such as turtles results from basking

(aquatic or aerial), occupying shaded/cooler sites, or thermoconforming to microhabitat temperatures (Rowe and Dalgarn 2009). In two senses the results of this study suggest that Wood Turtles in the forested habitats at this location may operate as thermal conformers during June-August: one, a broad range of thermoregulatory options was available at a fine spatial scale to which they could conform (Figs. 3.5 & 3.7), and two, their shell temperatures were generally congruent with the shaded under vegetation temperatures (Fig. 3.15). As did lizards in habitats of high thermal quality (Blouin-Demers and Nadeau 2005), Spotted

Turtles (C. guttata) in Ontario thermoconformed in the summer (Yagi and Litzgus

2013).

Although the Wood Turtles’ mean shell temperatures generally conformed with the UV microhabitat temperatures, the temperature spikes and dips as well as the proportions of temperatures in the different thermal categories suggest

111 thermoregulation and selection of different microclimates (Tables 3.3 & 3.11, Figs.

3.16 & 3.18). Variation in Tb resulted from time of day, sex, and microhabitat use for the emydid C. picta (Rowe and Dalgarn 2009). For Minnesota Blanding’s Turtles

(E. blandingii), Sajwaj and Lang (2000) found thermal patterns to vary with weather conditions in the summer: a thermoregulating pattern was observed on sunny days, with large differences between minimum and maximum body temperatures (range

> 11°C), a thermoconforming pattern was observed when solar radiation was limited or absent (range < 6°C), and an intermediate pattern occurred on partially cloudy days. Due to similar contingencies the Wood Turtles here may have manifested the types of patterns and variation observed in the preceding three turtle taxa.

Thermal specialists (stenotherms) have narrow thermal-niche widths (Huey and Slatkin 1976). Wood Turtles, with their observed ability to occupy a variety of habitats and operate over a broad range of temperatures, perhaps are thermal generalists (eurythermic). I say ‘perhaps’ because though found in a wide range of thermal environments (e.g., 17° water or a 45°C terrestrial sun splotch), most of the time they appear to maintain a somewhat narrow range of body temperatures (at least in the summer); for example, ca. 77% of the diurnal shell temperatures of the turtles in this study were in the 17-27°C range, while ca. 85% of the nocturnal temperatures were.

The temperature patterns observed here in VA and WV are similar to those observed for Wood Turtles in the field at a similar latitude on the coastal plain of

112 Maryland. Turtles there achieved a peak temperature of 37°C before noon (based on TidbiT temperature sensors mounted on their carapaces), then maintained a temperature of 20-25°C for the rest of the diurnal hours, with temperatures of 15-

20°C at night (Profitt and Chance 2003). The Wood Turtles here in VA and WV maintained generally cooler temperatures than did those in Iowa. Tamplin (2005) reported that turtles there closely regulated their mean body temperatures near

25°C in June and July. At two sites in Iowa, Berg (2014) found postnesting mean cloacal temperatures were similar for females (26.4˚C ± 0.3 SE; n = 267, n refers to number of temperature readings) and males (26.8˚C ±0.4 SE; n = 114) at the first site, and with females (25.5˚C ±0.3 SE; n = 164) and males (24.6˚C ±0.4 SE; n =

104) at the second site. These IA Wood Turtles basked much more those in VA-

WV; e.g., at one site females were basking on land during 44% of the encounters.

Use of Habitat

Clearly, daily temporal heterogeneity of the thermal environment exists at this study area. Except for a need to sometimes avoid (if only for brief periods)

direct sun in open microhabitats with temperatures ≥ CTmax, what is not so clear are the constraints imposed by the costs and benefits of fine scale thermal patchiness upon Wood Turtle summer spatial ecology. According to activity budget theory and the allocations of time and energy (Dunham et al. 1989), perhaps the most efficient use of time and space would involve a form of multi-tasking (Fortin et al.

2004), meaning that while one was engaged in thermoregulation, one would also be engaged with foraging, avoiding predation, and/or the seeking of mates or other

113 reproductive activities. Under this scenario, higher quality thermoregulatory microhabitats would also have food and cover available (or at least immediately adjacent), thus diminishing biophysical constraints (such as resource availability or harvesting and processing rates) on energy acquisition (Congdon 1989).

Small canopy gaps in mature forests, and the litter, woody debris, and herbaceous vegetation found therein, may afford Wood turtles just such an opportunity to implement synchronous thermal, foraging, and predator avoidance strategies, i.e., provide for more modes and vigour (sensu Gunderson and Leal

2016) at a single site. As turtles of the opposite sex may also be attracted to such sites, they could additionally facilitate mating opportunities. Microhabitats open to the sun could provide access to high temperatures, while the under litter/LWD and under vegetation microhabitats could provide cooler and more stable (less variable) temperatures. This propinquitous multimodality can be conceived as provision of consilient opportunities to thermoconform to different within-site microclimates.

The possible advantages of such multitasking quasi-thermoconformity are the reduction in exposure to predation, reduction in search time for seeking out sunny spots, shady shelter, or foraging patches, and reduction in the energy expenditures consequent of locomotion to such locations.

As with chelonians living elsewhere in habitats with a high degree of forest cover (e.g., the Serrated Hinge-back Tortoise (Kinixys erosa) of West Africa –

Luiselli 2005), there is not much risk of overheating for Wood Turtles in the relatively intact forest at these study sites (Table 3.3). Mean canopy openness as

114 measured with a spherical densitometer at 197 paired random points at the VA study site was ca. 13.1% (se = 0.40), for 123 random points at the WV study site it was 16.7% (se= 0.97); for turtle locations it was even higher, 19.6% (se = 0.97) for

123 WV turtle points and 20.1% (se = 0.96) for 197 VA turtle points (Chapter 6).

These proportions should not be interpreted as implying that usable basking sites are significantly limited to Wood Turtles here. Canopy gaps and broken canopies created by natural disturbances in mature forests furnish numerous openings in a variety of sizes that provide access to solar radiation for Wood Turtles and other small chelonians ca. 20cm length and 1kg in mass (Donaldson and Echternacht

2005, Luiselli 2005, Remsburg et al. 2006, Krichbaum this study). Mean amount of ground area beneath canopy gaps ≥ 9m2 estimated in the 400m2 plots at 197 adult turtle points at the VA study site was 41.4m2 (se = 4.41), for 123 adult turtle points at the WV study site it was 33.1m2 (se = 4.11) (Chapter 6); these numbers do not include all those smaller areas under 9m2 created by broken canopies that can supply sun splotches for basking as well (such areas are incorporated into the canopy openness measurements). The behavioral exploitation of within-habitat patchiness through basking has been observed in other reptiles (Carrascal et al.

1992).

Canopy gaps provide opportunities for basking, while the LWD and other ground cover there provide thermal refugia that allow escape from high midday temperatures. The microhabitat array sites all had varying degrees of canopy openness and amounts of canopy gaps. It is important to remember that greater

115 canopy openness in the overstory tree stratum does not necessarily translate to more open conditions or higher temperatures on the ground (i.e., at the scale/level that a Wood Turtle lives). In fact, the density of ground vegetation could actually increase under conditions of greater overstory canopy openness (Krichbaum pers. obs.). In this way, by providing both shaded and exposed microhabitats that allow for a range of humidity and temperature conditions, broken canopies and gaps facilitate efficient osmo- and thermo-regulatory shuttling. In addition, by providing a greater range of forest floor light levels and temperature regimes, gaps can allow for more floristic richness and/or abundance, i.e., increased foraging opportunities.

Just as the location or quality of optimal thermal patches may change seasonally or daily, so too may the availability or quality of food patches. Net energy gain (and hence growth and reproduction), and even survival, depend not only upon food availability but on other considerations as well; consequently, patches may be selected that are “suboptimal” for either food or thermal resources

(Huey 1991). Turtles face tradeoffs in resource allocation to various compartments of their energy budget, choices that ultimately have significant implications for reproductive output (Congdon 1989, Penick et al. 2002). At these VA and WV sites the period of year when Wood Turtles may feed (e.g., when water and air temperatures are at least 17.2° and 23°C, respectively (Ernst 1986)) could be 7 months or more. In the years 2006-2015 I observed Wood Turtles foraging at various VA and WV sites in all months from March to October. This means that in the spring females may feed before nesting. By not having to rely on stored body

116 lipids from the previous year, energy and time allocations here (e.g., parental investment), and thus habitat use, may differ from those at more northern populations (Rollinson et al. 2012).

If attainment of optimal temperatures involved considerable evaporative water loss, Wood Turtles, being amphibious, could deal with osmotic challenges as well as obtain cooling conditions by accessing substantially cooler water in the nearby streams at any time during the summer. Turtles often remain near or in waterways during the spring and autumn (when water temperatures may be warmer than air temperatures), but move much farther afield during summer – up to 300-600m from waterways (Kaufmann 1992, Arvisais et al. 2002 & 2004,

Flanagan et al. 2013, Parren 2013, this study). Wood Turtles can move at a speed of at least 0.32km/hr (Woods 1945) or 1.1km/hr over short distances (Swanson

1940). In 2010 prior to this study I radio-tracked an adult female in VA who traversed at least 1km in ca. 26 hours (Wood Turtles are typically diurnal so it is doubtful she was walking at night); I also observed another VA female walk ca.

90m in 85 minutes. The greatest known distance moved in a 24-hr period by an individual during this study was 511m (straight-line distance) by an adult male.

Clearly, during the summer period of greatest terrestrial dispersal a turtle is typically at most a few hours walk from cool water. Some of the Wood Turtles involved in this study evidently stayed for consecutive days in aquatic stream habitat (with water temperatures ca. 17-18°C) (Fig. 3.19).

117 For both females and males, the proportions of turtle iB temperatures >30°C

and 27.5-30°C (Topt) were less than the proportions of such categories recorded for

TempTs, while the proportions <17°C were greater than those for the TempTs

(Tables 3.7 & 3.10). This is explained by the fact that most of the ground temperatures for radio-tracked turtles were collected in the afternoon, typically the warmest part of the day, while the diurnal iButton temperatures include earlier and later times of day that are cooler. In addition, the iB temperatures include those recorded when turtles were aquatic, while the ground temperatures recorded in the field were exclusively terrestrial. Male mean temperatures may have been slightly lower than females because they were using aquatic habitats more frequently.

Twenty percent of 645 total adult turtle locations were aquatic, with male Wood

Turtles found more often in aquatic habitats than were females. In WV, 38.2% of male locations (out of a total n = 144) were aquatic, while 21.7% of female locations were (total n = 152). In VA, 17.6% of male locations (total n = 102) were aquatic, while 9.7% of female locations were (total n = 247). The higher average temperatures at the lower elevation WV site may have contributed to these different behavioral patterns. In WV, 100% of radio-telemetered males were located aquatically at some time over the course of the 2011-2014 field seasons, while only 41.2% of females were. In VA, 57.1% of radio-telemetered males were located aquatically at some time and 53.6% of females were.

Although the sexually dichotomous trend is similar, the proportionate use of aquatic habitat here was less than that displayed by Iowa Wood Turtles (Berg

118 2014); at two sites there, 23.2-34.6% of post-nesting period (i.e., summer) locations of radio-telemetered females were aquatic, while 47.7-49.5% of male locations were. Arvisais et al. (2004) reported that a Canadian population used aquatic habitats more frequently (59.1%) than terrestrial habitats (40.9%) during the active season. The use of aquatic habitats by turtles in VA and WV is similar, however, to that of another Canadian population in Algonquin Park that utilized aquatic habitats only 14.0% of the time during the spring and summer (Quinn and Tate

1991) and to Wood Turtles in Maine where Compton et al. (2002) recorded 16.0% of all turtle observations from May – September (excluding females during the nesting season) in lotic water sources. At two sites in northern WV 30% of summer

Wood Turtle locations were aquatic (Curtis and Vila 2015).

Whether it be nutritional, thermal, or other resources, an individual’s summer activity area must supply its needs. The mean average size of summer activity areas (MCPs) calculated for 59 Wood Turtles radio-tracked during this study was 2.25ha (=22,500 m2) (Chapter 1). A sitting/standing adult Turtle takes up an area of ca. 0.045 m2. Ergo, at any given time an individual Wood Turtle occupies an extremely small portion of its activity area, around 1/500,000 on average. This fact, coupled with their apparently low population densities (e.g., a density estimate for adults of 0.7/ha at the WV site; Chapter 2), make it unlikely that competition for space (habitat restriction hypothesis) is a factor constraining a

Wood Turtle from attempting to attain physiologically optimal body temperatures.

119 Instead, attainment of such states results from the selection of different habitats

(habitat exploitation hypothesis) (Tracy and Christian 1986).

Tracy and Christian (1986) outlined three situations concerning use of thermal patches and tradeoffs involved: 1) when there is no intraspecific interference or competition – if optimal patches are limited or difficult to exploit, then the costs involved with using them may outweigh the benefits; 2) when there is intraspecific interference or competition – accessibility to optimal patches may be reduced by conspecifics; 3) when there is interspecific interference or competition – some niche separation may occur. As Wood Turtles are not territorial

(Kaufmann 1992b), occur at low population densities (Chapter 2), have limited mobility (with a concomitant inability to patrol their activity areas and chase away competitors), and occupy such small portions of their activity areas, situation #2 does not appear to be operant. Aside from the Eastern Box Turtle, there are few if any omnivores of comparable size sharing the Wood Turtle’s terrestrial trophic niche. Both of these turtle taxa normally have low population densities and low metabolic requirements (Luiselli 2006). As did Strang (1983), I observed non- agonistic individuals of both species within a meter or less of each other. Nor am I aware of other competing herbivores or carnivores that would physically limit accessibility to optimal thermal patches; however, behavioral avoidance of predators by turtles may in effect produce such inaccessibility. Thus, situation #3 does not clearly appear to be operant here either. When diet is similar among species, habitat can be partitioned with regard to foraging tactics (Williams and

120 Christiansen 1981) or with seasonal and annual changes in resource availability

(James 1991). Wood Turtles feed in aquatic habitats where Box Turtles do not and occupy aquatic habitats for much longer time periods than does T. carolina, including hibernating underwater during the winter.

Situation #1 appears to be the most likely pattern operant at these study sites. The similarities among temperatures at sites with different environmental conditions (Table 3.1) revealed by the microhabitat iButton data indicate that at the macro-scale there is a broad distribution of salubrious thermal environments. These data along with the turtle iButton and location data indicate that individuals can attain similar body temperatures in different macro-environments, i.e., forest stands of different composition and structure. For example, Wood Turtles here used stands of ten different forest types, including Chestnut Oak – Scarlet Oak, Virginia Pine,

Upland Hardwoods – White Pine, and White Oak – Northern Red Oak – Hickory

(Table 5.4). The costs involved with using patches may involve not competition or limited availability, but perceived predation risk. Nonetheless, it is possible that unrecognized interspecific competitors do exist that serve to constrain Wood Turtle home ranges to some areal zone with widespread thermal optima.

With reduced risk of overheating at these forested montane sites, the thermal safety margin could be greater, so that body temperatures might be expected to be higher, i.e., closer to the upper bound of the TAW. But instead, at these southern

study sites Wood Turtles generally maintained a low Tb relative to ambient substrate

or air temperatures (Ta). Perhaps there are latitudinal and/or elevational differences

121 in thermoregulation (Weatherhead et al. 2012), with Wood Turtle Tb often being >

Ta in the north, Tb = Ta in the middle of the range, and Tb < Ta in the south. Southern turtles may be less thermally constrained by herbivory since they have lengthier diel and seasonal periods when high temperatures conducive to higher rates of digestion are available.

Conservation Considerations

The microhabitat and turtle data (Tables 3.1, 3.5, 3.8, 3.9, 3.11) can be used to identify thermal risk zones for implementation of better-informed management decisions regarding Wood Turtles. Habitat use restriction from high temperatures appears to be the primary potential thermal concern at these southern latitude sites.

In the VA deciduous forest with anthropogenic alterations the array locations with

the highest proportions of temperatures ≥ CTmax for all three microhabitats were a roadside and a scrubby anthropogenic opening with high degrees of canopy openness; these sites also had the highest minimum, maximum, and mean temperatures, as well as the greatest variance. Conversely, the only site with no

recorded temperatures ≥ CTmax was a high stem density early successional habitat

(clearcut 30 years before); for the open microhabitats this site also had the lowest minimum, maximum, and mean temperatures, as well as the least variance. For the under veg and under litter microhabitats a mature eastern aspect mixed oak site had the lowest minimum, maximum, and mean temperatures, as well as the least variance.

122 Identification of potential risk zones in the mixed pine-deciduous WV forest where all the sites were older and unroaded is not so apparent. There, sites with below average degree of canopy openness had some of the highest proportions of

temperatures ≥ CTmax for open microhabitats. The highest average daily maximum and mean temperatures and greatest variance were at a canopy gap in mature forest with a NE aspect and slightly higher than average degree of canopy openness; this was a relatively flat site so the influence of aspect was negligible.

The only site that recorded no open microhabitat temperatures ≥ CTmax was in mature forest with a northern aspect and below average degree of canopy openness; this site had the lowest average daily maximum and mean temperatures and greatest variance. For the under veg microhabitats, the pattern was clearer, as

temperatures ≥ CTmax only occurred in mature tracts with SW aspects, steep slopes, and high degree of canopy openness, all of which are typical correlates of warmer ground-level temperature regimes (Cantlon 1953, Pringle et al. 2003, Fridley 2009).

The lowest average daily maximum and mean temperatures, and lowest SD for WV under veg microhabitats occurred at an eastern aspect mature forest site with below average degree of canopy openness. For the WV under litter microhabitats, the highest average maximum and highest average mean temperatures, and highest

SDs occurred at a location with a northern aspect (but with little inclination) and the highest degree of canopy openness of any array site. The lowest under litter temperatures and the least variance occurred at a NW aspect site with below average degree of canopy openness.

123 All these outcomes, as well as the correlation test results for temperature and environmental attributes at array sites, indicate amount of canopy openness as a primary driver of fine-scale thermal patterns in these Central Appalachian forested locales, with some influence from aspect as well. Sites with high degree of canopy openness, such as roadsides or recently logged sites, have the highest temperatures and greatest temperature variance (Collins et al. 1985, Currylow et al. 2012), while sites with very closed canopies, such as a regenerating clearcut with high stem density, offer the least thermal diversity. Though sites that are too open can be

potentially problematic regarding CTmax for Wood Turtles in the summer, such risk changes with the season, geographic location, and behavior. Habitat of limited suitability for use by adults in the summer due to excessive temperatures, such as roadsides or anthropogenic openings, may nevertheless provide valuable nesting sites earlier in the year (Krichbaum pers. obs.). In addition, sites detrimental in the south or summer may be more suitable in the north or spring and vice-versa.

Furthermore, due to differences in life history characteristics, habitat preferences of animals at the periphery of their geographic ranges may differ from those at the core (Kapfer et al. 2008).

During the post-nesting period the Wood Turtles at these sites tended to avoid areas of anthropogenic early successional habitat (i.e., recent tracts of even- age logging) as well as roadside edge habitat. The Turtles would go right to the edge of the recent cut units, but no further. With one exception, whenever I did find them at the edge of a cut unit it was at the site of a large mature leave-tree (the

124 trees left standing at “modified shelterwood” cuts, which are ca. 90% of a clearcut) around which the shading understory (e.g., blueberry bushes (Vaccinium spp.)) was undisturbed. The one exception was a female turtle located ca. 30m into a 5-year old regenerating modified shelterwood cut who was eating blackberries (Rubus) growing there. I never found Wood Turtles in the regenerating clearcuts 5-35 years old with a high stem density of tree saplings. Part of the reason for this may be that the closed-over canopy at such logging sites, resulting from both the high stem density sapling regeneration as well as the lack of canopy gaps, makes for constantly shaded conditions that preclude spatially efficient shuttling in addition to manifesting a scantily developed ground floor herbaceous layer (Roberts 2004) that decreases foraging opportunities and exacerbates exposure to predators. The roadsides present the opposite problem, with consistently high temperatures outside of the PBTR. Except for nesting females, I never observed a Wood Turtle using roadside open or edge habitat. In short, all these lines of evidence indicate that from a thermal perspective, at least in some parts of their range, it is a good idea to avoid fabricating large openings in Wood Turtle habitat.

Mature forests are often characterized as “closed canopy” habitats. This characterization is not precisely accurate; a mature forest can be “intact” or

“contiguous” yet have numerous small canopy openings due to a variety of natural disturbances (McCarthy 2001). Mature forests are of the age that a mosaic of habitats is gaining expression due to the operant disturbance regime (Franklin et al.

2002). Broad-scale habitat alterations, such as intensive logging of areas 10ha in

125 size or road building, could fabricate conditions that Wood Turtles might avoid from a thermoregulatory perspective. Conversely, small canopy gaps, the result of the natural disturbance regime in NE forests (Runkle 1985 & 1990 & 1991b, Rentch

2006, Glasgow and Matlack 2007), create diverse conditions at a fine spatial scale that allow for shuttling and thermoregulatory ease as well as multi-tasking. Mature forests with such naturally provided structural complexity and heterogeneity, and concomitant thermal diversity, should be encouraged in Wood Turtle habitat. At some places, fabricated small canopy gaps, such as some group selection cuts of ca. 6 overstory trees per hectare at points with desired species advanced regeneration, could possibly improve thermal site conditions for Wood Turtles.

126 Table 3.1

Environmental attributes at iButton array sites; VA at 1a, WV at 1b 3.1a. Attributes of sites where iButton arrays were successfully deployed in Virginia during July-August 2013. Forest types: 39 = Table Mountain Pine, 41 = Cove Hardwoods – White Pine, 53 = White Oak – Northern Red Oak – Hickory, 54 = White Oak, 60 = Chestnut Oak – Scarlet Oak. Site index = expected tree height at end of 50 years of age (an indicator of site relative productivity). ESH = early successional habitat (0-35 years old), Mid-suc. = mid-successional forest (35-75 years old), Mature = forest at least 75 years old, OG = old growth (minimum years of age [90-140] depends upon forest type). Canopy Aspect Aspect Slope Distance Elevation Distance Stand Stand Stand Site openness (degrees) (Beers (degrees) from (m) from forest site age seral VA IB (%) transform) main road (m) type index (yrs.) stage Array stream Sites (m) PRA1 17.25 322 1.12 2 44 495 153 53 70 119 Mature PRA2 47.5 90 1.71 2 37 544 58 54 70 141 Shrubby opening PRA3 12 350 1.57 2 80 546 199 53 70 45 Mid-suc. CRA2 12.25 270 0.29 3 26 514 81 41 80 108 Mature seep PRB1 13.25 88 1.73 10 117 472 35 53 60 94 Mature PRB2 14.5 170 0.43 17 360 579 238 39 40 101 OG PRB3 21.5 158 0.61 10 92 594 96 54 70 144 OG PRC1 13 110 1.42 7 157 510 86 53 80 43 Mid-suc. PRC2 43 292 0.61 7 68 498 174 53 70 5 ESH PRC3 13.75 294 0.64 7 124 507 68 60 60 30 ESH PRC4 27.75 120 1.26 7 111 528 5 53 60 44 Roadside means 21.4 205.8 1.0 6.7 110.6 526.1 108.5 NA 66.4 79.5 NA se 3.8 30.3 0.2 1.4 27.7 11.1 21.9 3.4 14.4

3.1b. Attributes of sites where iButton arrays were successfully deployed in West Virginia during July-August 2013. Forest types: 10 = White Pine – Upland Hardwoods, 33 = Virginia Pine, 42 = Upland Hardwoods – White Pine, 53 = White Oak – Northern Red Oak – Hickory. Site index = expected tree height at end of 50 years of age (an indicator of relative site productivity). ESH = early successional habitat (0-35 years old), mid-suc. = mid-successional forest (35-75 years old), Mature = forest at least 75 years old, OG = old growth (minimum years of age [90-140] depends upon forest type). Canopy Aspect Transformed Slope Distance Elevation Stand Stand Stand Site openness (°) aspect (°) from (m) forest site age seral WV IB (%) (2 = cool, main type index (yrs) stage Array 0 = warm) stream Sites (m) SRA1 24.75 55 1.98 2 10 438 10 70 112 Mature SRA2 18.5 315 1.00 3 40 320 53 80 122 Mature SRA3 12 346 1.52 4 51 326 53 80 122 Mature SRA4 14 308 0.88 3 16 325 53 80 112 Mature SRB1 18.75 110 1.42 5 16 338 10 70 112 Mature SRB2 32 236 0.02 28 29 329 42 60 97 Mature SRB3 57.5 14 1.87 6 105 345 33 50 92 OG SRB4 17.5 265 0.23 3 12 344 33 60 75 Mature SRC1 15.75 298 0.71 35 27 344 10 70 112 Mature SRC2 21.75 156 0.64 25 150 327 42 60 90 Mature SRC3 46 212 0.03 18 78 346 33 60 116 OG SRC4 13.5 20 1.91 11 53 366 53 80 122 Mature means 24.3 194.6 1.0 11.9 48.9 345.7 NA 68.3 107.0 NA se 4.1 34.7 0.2 3.4 12.4 9.2 3.0 4.3

127 Table 3.2

Diurnal microhabitat temperatures (C°) recorded by iButtons Diurnal temperatures (C°) recorded by iButtons placed in three different microhabitats in Virginia and West Virginia July 12-August 15, 2013. Diurnal = 8:00- 20:00. Aggregated = all recorded temperatures in both VA and WV. Number of temperatures recorded by each iButton = ca. 857; temperatures recorded every 30 minutes. Diurnal temperatures Microhabitats Mean SD Minimum Maximum Virginia Open (n=6) 23.39 6.20 11.5 74.5 Under veg (n=8) 21.54 3.50 11.5 44.0 Under litter (n=6) 20.34 2.78 12.5 40.5 West Virginia Open (n=8) 23.91 6.08 10.5 79.5 Under veg (n=9) 23.13 4.36 11.0 59.0 Under litter (n=6) 20.19 2.11 14.0 26.5 VA and WV Open (n=14) 23.68 6.14 10.5 79.5 Under veg (n=17) 22.39 4.06 11.0 59.0 Under litter 20.20 2.48 12.5 40.5 (n=12) VA (n=20) 21.74 4.50 11.5 74.5 WV (n=23) 22.64 4.87 10.5 79.5 Aggregated 22.22 4.72 10.5 79.5

128 Table 3.3

Proportions of diurnal microhabitat temperatures recorded by iButtons Proportions of diurnal temperatures recorded by iButtons placed in three different microhabitats in Virginia and West Virginia July 12-August 15, 2013. Diurnal = 8:00- 20:00. PBTR = preferred body temperature range for Wood Turtles posited for this study, TAW = thermal activity window for Wood Turtles posited for this study, Topt = optimal body temperature for Wood Turtles posited for this study. Number of temperatures recorded by each iButton = ca. 857; temperatures recorded every 30 minutes. Temperature categories

(PBTR) (Topt) (TAW) 17-27 >30 <17 27.5-30 >27 21-32 >32 <21 Virginia Open .803 .091 .033 .073 .164 .613 .059 .328 Under veg .912 .008 .040 .040 .048 .518 .000 .482 Under litter .970 .001 .020 .009 .010 .377 .000 .623 West Virginia Open .777 .101 .032 .089 .190 .670 .065 .264 Under veg .858 .040 .029 .072 .112 .694 .010 .297 Under litter .945 .000 .055 .000 .000 .337 .000 .663 VA and WV Open .790 .096 .033 .081 .177 .642 .062 .296 Under veg .885 .024 .035 .056 .080 .606 .005 .390 Under litter .958 .000 .037 .005 .005 .357 .000 .643

129 Table 3.4

Nocturnal microhabitat temperatures (C°) recorded by iButtons Nocturnal temperatures (C°) recorded by iButtons placed in three different microhabitats in Virginia and West Virginia July 12-August 15, 2013. Nocturnal = 20:30-7:30. Aggregated = all measurements in both VA and WV. Number of temperatures recorded by each iButton = ca. 782; temperatures recorded every 30 minutes. Temperatures

Mean SD Minimum Maximum Virginia Open (n=6) 17.80 2.39 10.00 25.00 Under veg (n=8) 17.93 2.21 11.50 25.00 Under litter (n=6) 18.75 1.58 13.75 24.25 West Virginia Open (n=8) 18.29 2.71 10.25 26.75 Under veg (n=9) 18.56 2.54 10.60 26.50 Under litter (n=6) 19.02 1.82 14.00 24.25 VA & WV Open (n=14) 17.95 2.56 10.00 25.50 Under veg (n=17) 18.24 2.38 10.60 26.50 Under litter (n=12) 18.88 1.70 13.75 24.25 VA (n=20) 18.16 2.15 10.00 25.00 WV (n=23) 18.60 2.40 10.25 26.75 Aggregated(n=43) 18.39 2.29 10.00 26.75

130 Table 3.5

Statistical values and array sites of diurnal microhabitat temperatures (C°) recorded by iButtons Average diurnal temperatures (C°) recorded by iButtons placed in three different microhabitats in Virginia and West Virginia July 12-August 15, 2013. Reported are the values of the individual iButton that attained the highest and lowest mean value for each type of metric, with its array site location acronym below (see Table 3.1 for the environmental attributes found at these locations). SD = standard deviation, Min = minimum temperature, Max = maximum temperature, Mean = mean temperature, subscript “d” denotes average temperatures calculated on a daily basis across time periods, subscript “h” denotes average temperatures calculated on a half-hourly time period basis across days. Open = exposed microhabitats unshaded by ground floor vegetation, Under veg = microhabitats shaded by ground floor vegetation, Under litter = microhabitats under leaf litter and coarse woody debris. Diurnal = 8:00-20:00. Number of temperatures recorded by each iButton = ca. 857; temperatures recorded every 30 minutes. VA WV Open Under veg Under litter Open Under veg Under litter high low high low high low high low high low high low

SDd 7.0 2.0 3.2 0.9 2.4 0.7 8.3 2.4 5.7 1.8 1.2 0.6 PRA2 PRC3 PRC4 PRB1 PRC4 CRA2 SRA1 SRA3 SRB2 SRB1 SRB3 SRA4

Mind 19.0 17.6 18.7 17.6 19.4 17.2 19.2 17.6 19.2 17.7 19.6 17.6 PRC4 PRC3 PRC4 CRA2 PRC4 PRC3 SRC2 SRA1 SRC1 SRC2 SRC2 SRA4

Maxd 44.8 25.1 30.6 21.2 28.7 20.0 50.8 27.4 42.3 24.6 22.5 19.4 PRA2 PRC3 PRC4 PRB1 PRC4 CRA2 SRA1 SRA3 SRB2 SRB1 SRB3 SRA4

Meand 26.3 21.4 23.5 19.9 23.7 18.9 26.1 22.9 25.9 21.5 21.3 18.8 PRC4 PRA3 PRC4 PRB1 PRC4 CRA2 SRA1 SRA3 SRB2 SRB1 SRB3 SRA4

SDh 5.3 2.9 4.2 2.0 2.9 1.5 5.2 3.0 4.3 2.3 2.0 1.6 PRA2 PRA3 PRC4 PRB1 PRC4 PRB3 SRA1 SRB1 SRC3 SRB1 SRB3 SRA4

Minh 18.3 15.9 17.2 15.0 18.8 15.5 19.0 17.2 18.9 16.7 18.4 15.8 PRA2 PRC3 PRA2 CRA2 PRC4 CRA2 SRB4 SRC1 SRC3 SRC1 SRC2 SRA4

Maxh 38.5 28.0 33.1 24.2 31.1 22.7 36.4 29.7 35.0 26.1 25.4 21.6 PRA2 PRA3 PRC4 PRB1 PRC4 PRC3 SRA1 SRB1 SRC3 SRB1 SRB3 SRA4

Meanh 26.3 21.4 23.5 19.9 23.7 18.9 26.1 22.9 25.9 21.5 21.3 18.8 PRC4 PRA3 PRC4 PRB1 PRC4 CRA2 SRA1 SRA3 SRB2 SRB1 SRB3 SRA4

131 Table 3.6

Proportions of nocturnal microhabitat temperatures (C°) recorded by iButtons Proportions of nocturnal temperatures (C°) recorded by iButtons placed in three different microhabitats in Virginia and West Virginia July 12-August 15, 2013. Nocturnal = 20:30-7:30. PBTR = preferred body temperature range for Wood Turtles posited for this study, TAW = thermal activity window for Wood Turtles posited for this study, Topt = optimal body temperature for Wood Turtles posited for this study. Number of temperatures recorded by each iButton = ca. 782; temperatures recorded every 30 minutes. Nocturnal temperature categories

(PBTR) (Topt) (TAW) 17-27 >30 <17 27.5-30 >27 21-32 >32 <21 Virginia Open .702 .000 .298 .000 .000 .076 .000 .924 Under veg .707 .000 .293 .000 .000 .072 .000 .927 Under litter .876 .000 .124 .000 .000 .092 .000 .908 West Virginia Open .707 .000 .293 .000 .000 .156 .000 .844 Under veg .753 .000 .247 .000 .000 .181 .000 .819 Under litter .880 .000 .120 .000 .000 .163 .000 .837 VA and WV Open .705 .000 .295 .000 .000 .122 .000 .878 Under veg .731 .000 .269 .000 .000 .130 .000 .870 Under litter .878 .000 .122 .000 .000 .127 .000 .873

132 Table 3.7

Ground temperatures (C°) at turtle and random points in VA and WV Values for ground temperature variables (C°) obtained at turtle points and random points in Virginia and West Virginia during June-August 2011-2014; reported in descending order are means, standard deviations, and ranges. TempT = temperature immediately adjacent to turtles when they were located, TempGr = temperature in shade when/where turtles were located, TempVeg = temperature in shade when habitat data were obtained at plots. Point types: FWT = female Wood Turtle points, FRP = female random points, MWT = male Wood Turtle points, MRP = male random points. Virginia West Virginia Point Types Temperature variable FWT FRP MWT MRP FWT FRP MWT MRP TempT 25.3 NA 25.9 NA 27.8 NA 28.0 NA 2011-2014 3.24 NA 3.03 NA 3.81 NA 3.20 NA 15.6-39.1 NA 18.1-34.1 NA 21.0-40.0 NA 21.8-35.7 NA

TempGr 25.1 NA 25.6 NA 26.8 NA 27.0 NA 2011-2014 2.78 NA 2.67 NA 2.86 NA 2.86 NA 16.2-32.7 NA 18.1-30.1 NA 20.6-33.5 NA 21.2-35.4 NA

TempGr 25.8 26.8 26.4 27.4 27.9 29.3 28.1 28.7 2011-2012 2.54 3.17 2.88 2.28 2.55 2.98 2.06 2.54 19.4-31.4 19.5-34.6 18.1-30.1 22.2-32.2 23.4-33.5 25.4-36.1 24.3-31.7 24.5-35.4

TempVeg 24.3 24.9 24.6 24.9 25.3 25.3 24.3 25.2 2013-2014 2.49 2.45 2.33 2.15 3.52 2.39 2.22 2.38 18.7-29.7 20.4-30.8 20.1-29.4 20.2-29.2 17.5-34.1 20.5-32.8 20.9-29.3 21.2-29.7

133 Table 3.8

Surface temperature paired t-tests results Results of paired t-tests for surface temperature variables obtained at turtle points and paired random points in Virginia (VA) and West Virginia (WV) during June-August 2011-2014; reported in descending order are p values, t statistic values, and degrees of freedom. TempT = temperature immediately adjacent to turtles when they were located, TempGr = temperature in shade when/where turtles were located, TempVeg = temperature in shade when habitat data were obtained at plots. Point types: FWT = female Wood Turtle points, FRP = female random points, MWT = male Wood Turtle points, MRP = male random points. Comparisons with significant results are in bold. TempVeg test for VA females used square root transformed data. TempT-TempGr comparison involved turtle points only, not random points. Tests involving TempGr 2011-2012 had α = 0.05, TempGr 2011-2014 α = 0.0167 in VA and 0.0125 in WV, TempVeg 2013-2014 α = 0.05. VA WV FWT-FRP MWT-MRP FWT-FRP MWT-MRP TempGr 2011-2012 p 0.00012 0.520 0.0732 0.805 t -4.146 -0.658 1.891 -0.251 df 52 17 20 15

TempVeg 2013-2014 p <0.0001 0.172 0.629 0.0223 t -4.387 1.397 -0.486 2.451 df 79 31 48 23

TempT-TempGr p 0.271 0.213 0.119 0.5321 2011-2014 t 1.106 1.262 1.580 0.631 df 132 51 61 33

134 Table 3.9

Proportions of ground temperatures (“TempT”) immediately adjacent to turtles Proportions of diurnal ground temperatures recorded immediately adjacent to Wood Turtles (“TempT”) in Virginia and West Virginia during June-August of 2011-2014; individuals were in sun, shade, or under litter. FT = female Wood Turtle points (VA n=134, WV n=70), MT = male Wood Turtle points (VA n=52, WV n=44), WT = Wood Turtle points (both genders together). Diurnal = 9:00-19:30. PBTR = preferred body temperature range for Wood Turtles posited for this study, TAW = thermal activity window for Wood Turtles posited for this study, Topt = optimal body temperature for Wood Turtles posited for this study.

Temperature categories

(PBTR) (Topt) (TAW) 17-27 >30 <17 27.5-30 >27 21-32 >32 <21 VA FT .813 .075 .007 .104 .179 .895 .030 .075 MT .731 .058 .000 .212 .269 .923 .019 .058 WT .785 .070 .005 .140 .210 .903 .027 .070 WV FT .471 .200 .000 .329 .529 .914 .086 .000 MT .500 .182 .000 .318 .500 .886 .114 .000 WT .482 .193 .000 .325 .518 .904 .096 .000

135 Table 3.10

Proportions of ground temperatures recorded in shade (“TempGr”) at turtle and paired random points Proportions of diurnal ground temperatures recorded in shade (“TempGr”) at Virginia and West Virginia Turtle and paired random points during June-August of 2011 and 2012. FT = female Wood Turtles (VA n=56 points, WV n=24), FR = female paired random points, MT = male Wood Turtles (VA n=20 points, WV n=23), MR = male paired random points, WT = Wood Turtles (both genders together), RP = paired random points (both genders together). Diurnal = 9:00-19:30. PBTR = preferred body temperature range for Wood Turtles posited for this study, TAW = thermal activity window for Wood Turtles posited for this study, Topt = optimal body temperature for Wood Turtles posited for this study.

Temperature categories

(PBTR) (Topt) (TAW) 17-27 >30 <17 27.5-30 >27 21-32 >32 <21 VA FT .786 .054 .000 .161 .214 .964 .000 .036 FR .566 .113 .000 .315 .434 .905 .057 .038 MT .650 .000 .000 .350 .350 .950 .000 .050 MR .500 .111 .000 .389 .500 .944 .056 .000 WT .750 .039 .000 .211 .250 .961 .000 .039 RP .549 .113 .000 .338 .451 .916 .056 .028 WV FT .455 .227 .000 .318 .545 .955 .045 .000 FR .391 .391 .000 .217 .609 .826 .174 .000 MT .412 .118 .000 .471 .588 1.00 .000 .000 MR .273 .272 .000 .455 .727 .955 .045 .000 WT .436 .179 .000 .385 .564 .974 .026 .000 RP .333 .333 .000 .333 .667 .889 .111 .000

136 Table 3.11

Temperatures (C°) recorded by iButtons attached to adult Wood Turtles Temperatures (C°) recorded by iButtons attached to shell bridges of adult Wood Turtles in Virginia and West Virginia July 13-August 31, 2014. FT = female Wood Turtles (n=2), MT = male Wood Turtles (n=7), WT = Wood Turtles (both genders together) (n=9). Diurnal = 8:00-20:15, nocturnal = 20:30-7:45. Number of temperatures recorded by each iButton = ca. 2298 (diurnal), ca. 2268 (nocturnal); temperatures recorded every 15 minutes. Temperatures

Mean SD Min Max Diurnal FT 22.77 4.41 12.2 36.2 MT 21.30 4.00 12.5 36.2 WT 21.66 4.16 12.2 36.2 Nocturnal FT 19.08 2.14 12.4 25.4 MT 18.56 1.79 13.3 24.2 WT 18.68 1.87 12.4 24.4

137 Table 3.12

Proportions of temperatures (C°) recorded by iButtons attached to adult Wood Turtles Proportions of temperatures (C°) recorded by iButtons glued to shell bridges of adult Wood Turtles in Virginia and West Virginia July 13-August 31, 2014. FT = female Wood Turtles (n=2), MT = male Wood Turtles (n=7), WT = Wood Turtles (both genders together) (n=9). Diurnal = 8:00-20:15, Nocturnal = 20:30-7:45. PBTR = preferred body temperature range for Wood Turtles posited for this study, TAW = thermal activity window for Wood Turtles posited for this study, Topt = optimal body temperature for Wood Turtles posited for this study. Number of temperatures recorded by each iButton = ca. 2298 (diurnal), ca. 2268 (nocturnal); temperatures recorded every 15 minutes.

(PBTR) (Topt) (TAW) 17-27 >30 <17 27.5-30 >27 21-32 >32 <21 Diurnal FT .753 .049 .069 .129 .178 .565 .018 .417 MT .803 .026 .088 .082 .108 .402 .007 .591 WT .792 .031 .084 .093 .123 .438 .009 .553 Nocturnal FT .861 .000 .139 .000 .000 .203 .000 .797 MT .844 .000 .156 .000 .000 .105 .000 .895 WT .848 .000 .152 .000 .000 .127 .000 .873

138 CHAPTER 4: USING A GIS-BASED WATER BALANCE APPROACH TO

INVESTIGATE CHELONIAN HABITAT USE IN CENTRAL APPALACHIAN FORESTS

Introduction

The heterogeneous mosaic of patches composing a forested landscape can be defined at multiple spatial and temporal scales (Kotliar and Wiens 1990, Levin

1992, Edwards et al. 2003) and represented by a multitude of variables (Pickett and

Cadenasso 1995, Pickett and Rogers 1997). For example, patches can be defined by the types or the ages of the dominant floristic species in the canopy or understory (Fleming and Coulling 2001, Storch 2003, Bollmann et al. 2005), by soil type or slope aspect, or by components of water balance (Dyer 2009, Anning et al.

2014). Though multiple interwoven conditions, stressors, and disturbances affect the structural and compositional heterogeneity of landscapes (Fletcher, Jr. 2005,

McEwan et al. 2010), temperature and moisture certainly constrain vegetation patterns, including the distribution of different forest types (Whittaker 1956,

DeMars and Runkle 1992, Stephenson 2011). Since landscape patches provide animals with food, cover, and other environmental conditions where their activities take place, patches pattern animal spatial ecology in sundry ways (MacArthur and

Pianka 1966, Doak et al. 1992, Grover 2000, Baxley and Qualls 2009).

The patches composing a forest mosaic are heterogeneous in space and time due to the interactions of three pattern-forming templates: disturbance regimes, biotic interactions, and site-specific physical conditions (Swanson et al. 1988,

Pickett and Rogers 1997, Angelstam 2003). The physical environment includes

139 such factors as edaphic conditions, elevation, inclination, aspect, temperature, and precipitation. These factors influence fine-scale microclimatic patches and gradients that affect patterns of vegetation composition and structure (Jackson and

Newman 1967, Chen et al. 1999, Dyer 2009, Dobrowski 2010, Anning et al.

2014). For instance, at montane sites in western Virginia, differences in soil moisture and depth, aspect, and topography explained differences in vegetation on upper and lower slopes (Stephenson and Mills 1999). Faunal microhabitat selection is affected in turn due to foraging preferences and since behavioral thermo- and osmo-regulation involves the selection of optimal microclimates (Grover 2000,

Converse and Savidge 2003, Dubois et al. 2008 & 2009). For example, microhabitat selection by Eastern Box Turtles (Terrapene carolina) was related to thermoregulation and minimizing water loss (Penick et al. 2002, Rossell et al.

2006).

Terrestrial turtles may experience high rates of evaporative water loss

(“EWL”) (Weathers and White 1971). Evaporation of water serves to reduce the rate of heat gain during warming and increase the rate of heat loss during cooling. This mechanism thereby serves to prevent body temperatures from reaching critical thermal maxima, a benefit that must be balanced with the availability of and costs associated with sources of rehydration. Evaporative cooling can be expected to be less significant in situations of high relative humidity; in this way high humidity can serve to stabilize body temperatures. Vapor density difference (“VDD”), the difference in concentration of water vapor within an animal and that in the free air

140 beyond the animal’s body’s boundary layer, is the driving force of EWL and is directly related to relative humidity (Foley and Spotila 1978). All else being equal, as relative humidity increases VDD decreases, hence EWL decreases as well

(Finkler 2001). Water loss problems may be addressed through habitat selection

(e.g., moving into relatively more humid microhabitats) (Finkler et al. 2000) or diet

(such as foraging on succulent herbaceous material) (Strang 1983, Ernst 1986).

Since dehydration tolerance depends in part upon the severity of environmental conditions, survival can be directly related to relative humidity (Finkler 2001).

My study investigates the amphibious North American Wood Turtle’s

(Glyptemys insculpta) use of terrestrial habitat at forested montane sites in Virginia and West Virginia. Over the course of four years of fieldwork (June-August 2011-

2014) investigating summer habitat preferences in the post-nesting period, measurements of ground-level relative humidity were taken at the locations of radio-tracked Wood Turtles and associated random points. These data reflect highly localized conditions resulting from short-term weather patterns and fine- scale microhabitat characteristics. I hypothesized that potential EWL may be a driver of Wood Turtle spatial ecology and predicted that ground-level relative humidity would be significantly higher at turtle points than random points.

In addition, I used a water balance approach to examine variations in microclimate/microhabitat and their potential affect upon Wood Turtle spatial ecology during summer months of low water supply and high demand. With this method moisture availability and demand can be assessed across topographically

141 and edaphically variable sites (Dyer 2009). Use of readily available high resolution climate, soils, and elevation coverages in GIS allows soil moisture values at different sites to be estimated in absolute terms. These data represent landscape- level conditions resulting from long-term (ten-year monthly averages) climate patterns and more-broad-scale habitat conditions (e.g., topographical aspect and soil types). The monthly moisture demand at turtle locations was compared to that at random locations. The objective was to ascertain if GIS based ecological data and models are useful for revealing patterns of habitat selection by a small forest tetrapod. Due to the findings of previous studies (e.g., Ernst 1968, Strang 1983,

Dubois et al. 2008 & 2009), I hypothesized that Wood Turtles may prefer more mesic habitats and predicted water balance conditions (such as water deficit and actual evapotranspiration) would differ between turtle sites and random sites.

Focal Species

“Focal Species” in Chapter 1 provides a general description of Wood

Turtles. Terrestrially, Wood Turtles are often characterized as being associated with more humid microclimates and mesic habitats (Mitchell 1994, Akre and Ernst

2006, Ernst and Lovich 2009). Juvenile Wood Turtles were found to be more vulnerable to evaporative water loss than similarly sized adults of the sympatric

Box Turtle (Terrapene carolina; Ernst 1968). Wood Turtles forage on herbaceous plant leaves (e.g., Viola), mushrooms (e.g., Boletus and Amanita), earthworms, fruits (e.g., Rubus), slugs, beetles, and millipedes (see, e.g., Strang 1983, Kaufmann

1992, Niederberger and Seidel 1999, Ernst 2001, Compton et al. 2002, Walde et

142 al. 2003, Jones 2009, Krichbaum pers. obs.). Turtles may respond to fine-scale abundances of litter arthropods, fungi, and herbs that are affected by microhabitat/microclimate related to moisture differences in the forest (Meier et al.

1995, Hutchinson et al. 1999, Gilliam 2007).

Methods

Study Area

My study took place in montane forests of Virginia and West Virginia at the southern periphery of the species’ range. See “Study Area” in Chapter 1, Fig. 1.1,

Tables 5.3-5.5, and Appendix 1 for detailed description of the area.

Water Balance Model Rationale

Topography influences forest composition through its influence upon radiation load and soil moisture (Stephenson 1998). Topographically and edaphically controlled variations in microclimate and microhabitat affect ecological characteristics such as species richness and site productivity (Lookingbill and Urban 2005). By synthesizing a monthly time-step with readily available high- resolution data on climate, elevation, topography, and soil, the Dyer (2009) GIS- based water balance model directly assesses and quantifies moisture demand and availability at a site in absolute terms (e.g., millimeters of deficit). Water balance

(W) at a site is a function of elevation, slope, aspect, soil depth, soil texture, solar radiation, temperature, and precipitation. The basic components of a water balance approach are potential evapotranspiration (“PET”), actual evapotranspiration

(“AET”), and deficit (“DEF”). PET is a measure of moisture demand that represents

143 the amount of water that could be evapotranspired from a vegetated surface if water was not a limiting factor, while AET reflects given water availability. DEF is the difference between PET and AET, or evaporative demand not met by available water. Demand is met in part by drawing upon moisture stored in the soil, i.e., a site’s available water capacity (“AWC”), which is dependent upon the site-specific depth and texture of the soil. For further details see the Dyer (2009) paper as well as the guides and posters available for download from https://people.ohio.edu/dyer/water_balance.html.

GIS Procedures

I used the “Water Balance” toolbox and “Spatial Analyst” extension in

ArcGIS software (vers. 9.3; ESRI 2010) to develop a water balance model. Digital elevation map (“DEM”), soil AWC, and monthly solar radiation, temperature, and precipitation layers built a single model 10km long X 8km wide encompassing both study areas. I downloaded a fine-scale DEM (3 X 3m, with a minor extent at 9 X

9m resolution; USGS “National Map” site) and digitized soil surveys with estimates of AWC (NRCS). All cells in the study area were assigned the same monthly temperature and precipitation (2010 monthly 30-year normal for Winchester VA weather station ca. 40km to the NE; NCDC). Solar radiation (R) grids for the study area were computed monthly (Dyer user manual) from the closest solar radiation collection station (Dulles Airport in northern Virginia ca. 80km to the ENE).

Monthly radiation grids, DEM, soil AWC, and temperature – precipitation grids, were used as initial inputs for the Water Balance toolbox calculation of monthly

144 PET using the Turc (1961) method. PET was used to derive subsequent monthly calculations for: P-PE (precipitation minus potential evapotranspiration, i.e., supply minus demand), storage, surplus, AET, and DEF.

I calculated DEF using two soil depths for the AWC, 25cm and 100cm

(“DEF25” and “DEF100”). In deciduous forests such as occupy these study sites, 95%

of roots occur within the top 100cm of soil (Jackson et al. 1996). Thus, DEF100 reflects water availability for overstory tree growth and affects forest type, while

DEF25 aligns with the water available to herbaceous and gramineous ground floor vegetation in the understory layer. When limiting the AWC to only the top 25cm of soil, monthly deficits are greater than those calculated using 100cm of depth.

Wood Turtles may be spatially responding directly or indirectly to either or both

strata of vegetation cover. AET, which synthesizes the PET and DEF100 model outputs into a single metric, has a strong positive correlation with net primary productivity (Rosenzweig 1968, Webb et al. 1978). AET is also correlated with temperature, not just through effects of shade or albedo, but also in that, given equal net radiation conditions, increased transpiration from a vegetated area results in a lowering of maximum surface temperatures (Mahmood et al. 2014). Because

my sites are heavily forested AET was calculated with DEF100 values. And because

Wood Turtles are confined to the surface and are known to eat a variety of ground-

floor herbaceous vegetation, I used the DEF25 metric as it may be more congruent with their habitat preferences and spatial ecology. For the above reasons, my analyses and discussion focus on these two metrics, AET and DEF.

145 Field Procedures

See Chapter 1 “Field Procedures” for general information on radio-telemetry of turtles and data collection.

All turtle and random points were expanded in the field by an 11.3m radius to generate 400m2 plots with which to collect physiographic and biotic habitat data. The 11.3m radius also compensated for the error associated with the GPS unit that defines the precise points in the GIS. This technique assumes the turtles were selecting the conditions in the overall plot area rather than a precise pixel identified by the GPS. When a Wood Turtle was located I measured ground surface relative humidity (“HumT”) and temperature (“TempT”) immediately adjacent to the turtle with a digital thermometer-hygrometer, regardless of whether the individual was in the shade, sun, or under litter. When a turtle was not on the surface in the shade, I also measured the ground surface humidity (“HumGr”) and temperature

(“TempGr”) in the closest shade, usually within 1-2m of the turtle.

In 2011-2012 I located the random points immediately after I located the turtles. Random point HumGr and TempGr were typically taken within 30 minutes of the turtle point (mean 29.0 minutes, se = 2.2). Later (mean of 15.0 days after an individual was located, se = 1.0) at the turtle points and random points I measured habitat characteristics, including ground surface humidity (“HumVeg”) and temperature (“TempVeg”) in the shade. In 2013-2014 I measured HumVeg and

TempVeg at the random point and turtle point within ca. 16 minutes of each other

(mean 15.8 minutes, se = 1.3).

146 Analytic Procedures

Relative Humidity

For each state (VA and WV) I pooled data from 2011-2012, 2013-2014, and

2011-2014 and tested for differences in relative humidity between turtle vs. random points. I also compared humidities at female turtle points to those at male turtle points, as well as made comparisons between states. I used ANOVAs (one way and two way with individual turtle as a random effect), paired t-tests,

Wilcoxon tests (rank sum and paired signed rank), and correlation tests (Pearson and Spearman) to analyse data. Natural log, square root, or reciprocal transformations were used to normalize some data prior to analysis. A Bonferonni correction was used to adjust p-values when multiple comparisons were made on the same data. Assumptions of normality were tested with the Shapiro method and

Levene’s Test for homogeneity of variances. Statistical tests were accomplished in R

(R Development Core Team 2011).

The following acronyms are used herein: VA female Wood Turtles = VFT,

VA female paired random points = VFR, VA male Wood Turtles = VMT, VA male paired random points = VMR; WV female Wood Turtles = WFT, WV female paired random points = WFR, WV male Wood Turtles = WMT, WV male paired random points = WMR.

Water Balance

For statistical analyses I used the monthly mean values of PET, AET, and DEF calculated across all pixels within each 400m2 plot (41 pixels) (ArcGIS “zonal stats”

147 tool – ESRI 2010). To increase sample sizes the sexes were pooled within each state for statistical tests. WT = pooled male and female Wood Turtle points in VA or WV,

TR = pooled male and female paired random points in VA or WV, RP = GIS-

generated random points in VA or WV. In each state I compared AET and DEF25 in turtle points to paired random points (WT-TR) and to GIS-generated random points

(WT-RP) for each month (June, July, and August) (Table 4.3). Due to non-normality of data, I used Wilcoxon paired signed rank tests for WT-TR, while Wilcoxon rank sum tests were used for WT-RP comparisons.

For the matched pair analyses (viz., turtle plots vs. paired random plots)

(Table 4.4), the dataset for VA comprised 63 WT/TR plots in June (37F/26M),

70WT/TR plots in July (59F/11M), and 38WT/TR plots in August (32F/6M), totaling

171 WT plots and 171 TR plots = 342. The matched pair dataset for WV comprised

34WT/TR plots in June (21F/13M), 48WT/TR plots in July (30F/18M), and 29WT/TR plots in August (17F/12M), totaling 111 WT plots and 111 TR plots = 222.

In ArcGIS I generated 760 random points in the WV buffer zone (delimiting ca. 219ha) and 2087 random points in the VA buffer zone (delimiting ca. 586ha).

These buffer zones, extending out 290m (WV) and 295m (VA) from both sides of each main stream, contained approximately 95% of the Wood Turtle location points. The paired random points designated in the field (TR) represent sites available within an individual’s home range (corresponding to Johnson’s [1980] third-order habitat selection), whereas the GIS generated random points (RP) represent sites available to all individuals in each population and are in the area

148 within which home ranges are located (Johnson’s second-order selection). The density of generated random points per unit area approximates the density of the random points located in the field (i.e., the density of paired random points [TR] calculated as a function of the size of activity areas of individual turtles; see

Chapter 1). The random plots from points generated in the GIS (expanded by an

11.3m radius) are designated herein as VRP, WRP, or RP (VRP = Virginia generated random plots, WRP = West Virginia generated random plots, RP = generated random plots for VA or WV). Just as with the paired turtle/random plots, monthly

PET, AET, and DEF25/100 values were calculated for each individual generated random plot.

For the comparisons of turtle points to these generated random points (Table

4.5), I used an expanded dataset that included turtle points for which I had not located paired random points in the field. These modeling data for VA included 78

WT plots in June, 103 WT in July, and 75 WT plots in August (totaling 256 WT plots). The expanded dataset for WV included 51 WT plots in June, 74 WT in July, and 51 WT plots in August (totaling 176 WT plots).

Assumptions of normality were tested with the Shapiro method, and

Levene’s Test was used to test for homogeneity of variances. Statistical calculations were accomplished with the program R (R Development Core Team 2011).

149

Ground Level Relative Humidity and Temperature, and AET/DEF25 Values at Turtle and Random Points in 2011-2012

To evaluate the relative importance on Wood Turtle habitat use from broad- scale macroclimatic – topographic factors (water balance) as opposed to fine-scale microclimatic factors (ground humidity and temperature), using each state’s 2011-

2012 data I performed separate conditional logistic regressions for June, July, and

August as well as for the three summer months combined. In this method variables at each individual turtle point are compared with those at the paired random point.

I ran models with different combinations of variables and used an information- theoretic approach to compare models and model averaged coefficients.

The variables used in the modeling were AET, HumGr, TempGr, amount of canopy openness at each point (estimated with a spherical densitometer;

“Canopy”), and amount of ground area (m2) in the 400m2 plot that was under canopy gaps ³ 9m2 (visually estimated; “Gapsize”). The same variables and

modeling process were also used with DEF25 in place of AET. I included the canopy openness and gap size variables in the regression modeling because the water balance model, though it uses a range of environmental conditions (e.g., slope aspect, slope inclination, and soil type), does not include these two relevant site factors, either of which could conceivably affect ground surface temperature or moisture. Thinning or removal of the forest canopy typically results in reduced relative humidity and moisture at the ground surface and increased mean temperature, temperature fluctuations, and solar radiation (Collins et al. 1985).

150 Variables used in model construction were chosen based on their lack of correlation with each other; only the least correlated variables (viz., those with

Spearman correlation coefficient (rho) values less than 0.50) were retained for multivariate candidate models. With this criterion in mind, various combinations of variables had to be dropped from consideration, so candidate models were formulated with minor changes between VA and WV, different months, and AET

and DEF25 (Appendix 4).

Akaike’s information criterion (AICc), modified for small sample sizes

(relative to estimated parameters), was used to rank models, select the most

parsimonious models, and for averaging of habitat variable coefficients. AICc values

were rescaled to ∆i values for ease of interpretation and ranking, and Akaike

weights (wi) were calculated to give the approximate probability of each model

being the best model in the set. The model with the lowest AICc value was

considered to be the top model, and all models with AICc values within two of the minimum were considered to be well supported (Burnham and Anderson 2002). Of the well-supported models, that with the smallest number of variables was considered to be the best model (Arnold 2010).

Analyses were conducted with the R statistical program (R Development

Core Team 2015), specifically, the “Survival” package was used for the conditional

logistic regressions and the “Aicmodavg” package for AICc values and model averaged coefficients.

151 Results

When conducting multiple analyses on the same dependent variable, a

Bonferroni correction is conducted to protect from the increased likelihood of Type

I errors. Much of the results reported below consists of exploratory data analysis. I only had a few hypotheses to confirm, and those that I had involved the differences

between turtle points and random points, such as for DEF25/100 or surface humidity. I knew a priori that in some cases I would be testing the sexes and states separately and tests would involve subsets of a dependent variable, such as for a certain month or time period (e.g., 2011-2012 when HumGr for turtle points and random points was concurrently measured or July for AET). These tests are the main outcome of my research, whereas the other exploratory analyses can be considered as secondary outcomes or ancillary hypotheses. For this reason, the Bonferroni adjustments I applied to alpha values were different although the dependent variable was in a sense the same. For instance, I performed two tests of the HumGr

2011-2012 data (confirmatory), and used an adjusted alpha value of 0.025. I performed eight other tests that involved the HumGr 2011-2014 data (exploratory), and for these used an adjusted alpha value of 0.00625. Some of the exploratory correlation or comparison results may be incentive for future confirmatory research to determine their validity.

Ground Level Relative Humidity at Turtle and Random Points

In 2011-2012 (when I located the random points immediately after I located the turtles), HumGr differed between turtle and random points for both males and

152 females in VA, but only females in WV (Tables 4.1 & 4.2, Fig. 4.1). In 2013-2014

(when random points were located immediately after I located the turtle points during vegetation work), HumVeg differed between turtle and random points for males and females in both VA (marginally for males) and WV (Tables 4.1 & 4.2). In

2011-2014 HumT did not differ from concurrently measured HumGr for either males or females in VA or WV (Tables 4.1 & 4.2, Fig. 4.2).

For turtle points from 2011-2014, ANOVAs with state and sex as fixed effects and individual turtle as the random effect found significant differences between both states and sexes for HumT and HumGr (Tables 4.1 & 4.4). Wilcoxon rank sum tests found HumVeg to differ between states, but not sexes, at both turtle points and random points (Tables 4.1 & 4.4). Values were higher in VA than WV for HumT, HumGr, and HumVeg at turtle points, while both HumT and HumGr were higher for females than males (Table 4.3, Fig. 4.2).

In separate analyses for each state using 2011-2014 data (for HumT α =

0.0166, HumGr α = 0.0063), in VA HumT differed between males and females

(F1,41 = 4.69, p = 0.0362), as did HumGr (F1,42 = 6.39, p = 0.0154); similarly, in WV

there were no differences between males and females for either HumT (F1,43 =

1.010, p = 0.321) or HumGr (F1,40 = 0.263, p = 0.611) (Table 4.1, Fig. 4.2).

HumGr and humidity under litter significantly differed (paired t-test: t =

6.196, df = 57, p < 0.0001); mean humidity measured under litter for 58 VA and

WV points during 2011-2014 was 70.1% (se = 1.33), while the concurrent mean

HumGr at those same points was 62.3% (se = 1.67). Mean surface temperature in

153 the shade (TempGr) taken concurrently at those 58 points was 25.7°C (se = 0.43), while the mean temperature under litter was 24.8°C (se = 0.37); TempGr and temperature under litter significantly differed (paired t-test: t = 5.739, df = 57, p <

0.0001). Relative humidity under leaf litter did not differ between turtle points (72.7

± 1.9) and random points (72.8 ± 2.3) for a subset of concurrently measured

(within ca. 25 minutes of each other) 2011-2014 VA-WV points (paired t-test: t = -

0.057, df = 24, p = 0.955). Relative humidity under leaf litter did not differ between concurrently measured turtle and random points for a small subset of 2011-2012

VA-WV points (measured when turtles were radio-located; paired t-test: t = -0.406, df = 14, p = 0.691), nor for a subset of 2013-2014 VA-WV points (measured during veg work; paired t-test: t = -0.344, df = 13, p = 0.736).

There were significant inverse correlations between HumGr and TempGr for

2011-2014 in VA (Pearson test: t = -9.336, df = 254, r = -0.505, p < 0.0001) and

WV (Pearson test: t = -3.614, df = 98, r = -0.343, p = 0.0005), as well as for 2011-

2012 in both VA (Spearman test: S = 777289, rho = -0.468, p <0.0001) and WV

(Pearson test: t = -5.420, df = 82, r = -0.514, p = <0.0001) (Table 4.1).

In WV, transformed slope aspect was lower (suggesting warmer conditions or greater atmospheric evaporative demand) at turtle points than at random points

(paired Wilcoxon signed rank test: V = 338, p = 0.0273), whereas in VA transformed slope aspects did not differ between turtle points and random points

(paired Wilcoxon signed rank test: V = 1231, p = 0.399) (using 2011-2012 data;

Table 4.5).

154 Water Balance at Turtle and Random Points

Values for water deficits, potential evapotranspiration, and actual evapotranspiration were consistently greater in Virginia than West Virginia (Table

4.6, Figs. 4.3 & 4.4). Of the 12 test routines for each state, 8 of these found significant differences, six of which involved the turtle and paired random point comparisons (Tables 4.7 & 4.8).

In both VA and WV, turtle points and their paired random points differed for

AET in June. Turtle points and their paired random points differed for DEF25 in July and August in WV, but not in VA. There were also significant differences between turtle points and their paired random points for July AET in WV, but not in VA

(Table 4.7).

Turtle points and GIS-generated random points differed for AET for July in

WV and for June in VA. DEF25 was marginally different between turtle points and random points for June in WV (Table 4.8).

Correlations Between Ground Level Relative Humidity and Water Balance Values at

Turtle and Random Points

For each site (VA and WV) I ran separate monthly correlation tests for turtle points (males and females pooled) and their paired random points for both HumGr

and HumVeg, testing each of these values against those for DEF25, DEF100, PET, and

AET. Overall, there were weak correlations between the broad-scale water balance metrics and the fine-scale microhabitat measurements. With a Bonferonni correction for multiple tests, adjusted significance level for each month for HumGr

155 RPs alpha = 0.0125, for WT points HumGr α = 0.0063, for WT and RPs HumVeg α

= 0.0083. Using these corrected levels, there were no significant correlations of

HumGr or HumVeg with water balance attributes in either VA or WV; the closest to significant p-values were for HumGr with AET. The results reported below are those tests with p-values ca. 0.05 or under.

For VA monthly data, the strongest correlation (inverse) occurred in June between HumGr and AET at random points (Spearman test: S = 464, rho = -0.6215, p = 0.031). There were also inverse correlations at June random points for HumVeg

and DEF25 (S = 43972, rho = -0.2850, p = 0.029) and HumVeg and PET (S = 43811, rho = -0.2803, p = 0.032); with marginally significant results for August HumGr and AET at turtle points (S = 4948, rho = 0.3070, p = 0.073) and August HumVeg and AET at turtle points (S = 3601, rho = 0.3400, p = 0.057).

For WV monthly data, the strongest correlation occurred in July between

HumGr and AET at turtle points (Spearman test: S = 5340, rho = 0.4157, p =

0.009). Otherwise, the tests resulted in a correlation (inverse) for July HumGr and

DEF100 at turtle points (S = 12378, rho = -0.3544, p = 0.029); with marginally significant results for August HumT and AET at turtle points (S = 4948, rho =

0.3070, p = 0.073) and August HumVeg and AET at random points (S = 904, rho =

0.4129, p = 0.063).

Conditional Logistic Regression with AET Values for 2011-2012

The overall logistic regression results indicate that water balance conditions at a site, particularly AET, are relatively weaker influences on Wood Turtle habitat

156 preference than are finer-scale factors such as canopy openness and ground level

relative humidity. DEF25, however, was a factor in habitat preference for West

Virginia Wood Turtles.

For Virginia turtles with the three summer months pooled, the model- averaged coefficients indicate that VA Wood Turtles preferred forest sites with relatively higher ground level humidity, a more open canopy, and lower surface temperatures than were available at random sites (Table 4.9). Two competing

models had similar support (i.e., ∆AICc < 0.5) and the best model, with an AICc weight of 0.22, had the single variable HumGr (Table 4.10).

For West Virginia turtles with the three summer months pooled, model- averaged coefficients indicate that WV Wood Turtles preferred forest sites with higher ground level humidity, higher AET, and lower ground level temperatures than were available at random sites (Table 4.9). Only one model was well

supported (i.e., the other models had ∆AICc > 2) and this model, with an AICc weight of 0.70, was the same as the top model in VA, with the two variables

HumGr and Gapsize (Table 4.10).

For Virginia turtles with models run on a monthly basis, the model-averaged coefficients indicate that VA Wood Turtles preferred forest sites with a more open canopy in June, July, and August, higher ground level humidity in July and August, more canopy gaps in June, and lower surface temperatures in July than were available at random sites (Table 4.11). AET was a factor only in marginally well- supported models in June and July. In June, the two competing models with similar

157 support each contained a single variable, Gapsize or Canopy (Table 4.12); June models with HumGr or TempGr would not converge. For July, all the well- supported models contained the variable HumGr. For August, the two models with similar support each contained a single variable, Canopy or HumGr.

For West Virginia turtles with models run on a monthly basis, model- averaged coefficients indicate that WV Wood Turtles preferred forest sites with higher ground level humidity in July and August, higher AET in July, a more open canopy in August, and lower ground level temperatures in June than were available at random sites (Table 4.11). AET was a factor in all well-supported models in July.

In June, only one model was well supported and contained the single variable,

TempGr (Table 4.12); June models with HumGr or AET would not converge. For

July, the top models contained the variable HumGr. For August, two models had similar support (∆AICc < 0.5), and each contained a single variable, Canopy or

HumGr.

Conditional Logistic Regression with DEF25 Values for 2011-2012

Just as with the AET results, for Virginia turtles with the three summer months pooled, the model-averaged coefficients indicate that VA Wood Turtles in summer preferred forest sites with lower surface temperatures, higher ground level humidity, and a more open canopy than were available at random sites (Table

4.13). Two models were well supported, with a third marginally so (∆AICc = 2.04)

(Table 4.14). The best model, with an AICc weight of 0.38, had the two variables

HumGr and Canopy.

158 For West Virginia turtles with the three summer months pooled, the model- averaged coefficients indicate that WV Wood Turtles during the summer preferred

forest sites with greater DEF25, higher ground level humidity, and lower ground level temperatures than were available at random sites (Table 4.13). Two models

were well supported and the best, with an AICc weight of 0.59, contained the two

variables DEF25 and HumGr (Table 4.14).

For Virginia models run on a monthly basis, DEF25 was a factor in well- supported models in June, July, and August. The model-averaged coefficients indicate that VA Wood Turtles preferred forest sites with a more open canopy in

June, July, and August, higher ground level humidity in July and August, more area under canopy gaps in June, and lower surface temperatures in July than were available at random sites (Table 4.15). In June, the two competing models with similar support each contained a single variable, Gapsize or Canopy (Table 4.16);

June models with HumGr or TempGr would not converge. For July, all the well- supported models contained the variable HumGr. For August, the two models with similar support each contained a single variable, Canopy or HumGr.

For West Virginia turtles with models run on a monthly basis, DEF25 was a factor in well-supported models in July and a marginally well-supported model in

August. The model-averaged coefficients indicate that WV Wood Turtles preferred

forest sites with greater DEF25 in June, higher ground level humidity in July and

August, lower surface temperatures in June and July, and a more open canopy in

August than were available at random sites (Table 4.15). In June, only one model

159 was well supported and contained the single variable, TempGr (June models with

HumGr would not converge; Table 4.16). For July, all the well-supported models contained the variable HumGr. For August, the two top models had similar support and each contained a single variable, Canopy or HumGr.

Discussion

The measurements of ground-level relative humidity presented stronger support for my hypothesis than did the water balance metrics. The results suggest stronger preference by female Wood Turtles than males for more humid ground surface conditions that may change over short durations and at fine spatial scales.

The lack of difference between concurrently measured HumT and HumGr indicates that during diurnal hours the Wood Turtles are spending a lot of time in shaded conditions. This is corroborated by the fact that only 10.1% of 337 terrestrial locations of 86 adult Wood Turtles encountered during this four-year study were in the direct sun, while most were in shade, dappled conditions, or under litter. Under litter microsites had significantly higher humidity than those on the surface. That humidity under litter did not differ between turtle points and random points indicates that sites serving as relatively humid refugia are available to the turtles throughout the forest here – as long as the litter layer remains intact and is not removed (such as by burning).

Statistical analyses of the monthly water balance model outputs resulted in ambiguous overall support for my hypothesis. The results suggest that conditions

related to site moisture and evaporative demand (viz., AET and DEF25) may

160 influence the use of VA Wood Turtles’ activity areas in June and WV Wood Turtles’ activity areas in June, July, and August (Tables 4.9-4.16). There was some tendency for AET at turtle sites to be higher than at random locations (Tables 4.6-4.9, Fig.

4.3), which suggests a preference for sites of higher productivity. Contrary to my

hypothesis, however, the differences in DEF25/100 between turtle points and random points (Tables 4.6 & 4.7) and the logistic regression results for the pooled summer months indicate that WV Wood Turtles may have chosen sites with higher water deficit conditions than were randomly available (Tables 4.13 & 4.15, Figs. 4.4 &

4.5).

Due to the interactions of heat and moisture gradients resulting from climatic and topographic controls, greater water balance deficit (drier site conditions) can arise from such factors as reductions in the amount of water held in the soil (from edaphic or topographic properties) or increased evaporative demand

(greater evapotranspiration due to influences such as aspect or slope on insolation).

For example, xeric habitats can result from high insolation (due to a southerly aspect), slope inclination (steeper slopes resulting in a more perpendicular angle of incidence for solar radiation), or coarse textured soils (reduced water storage capacity) (Cantlon 1953, McCormick and Platt 1980). Turtles were using a broad range of sites, with both greater and lesser deficit and evapotranspiration than those available at random. The interactions of soil moisture, evaporative demand, and topography could exacerbate individual effects or counteract them, in that a site with a high amount of water held in the soil might have a higher evaporative

161 demand while a site with lower amounts of water held in the soil might also have lower evaporative demand, thus serving to equalize stress at the two locations. For instance, plants on south-facing slopes may face greater stress, not necessarily from reduced AWC, but due to higher atmospheric moisture demand caused by higher ambient temperatures (Lipscomb and Nilsen 1990, Huebner 1995). All these factors make patterns of Turtle preference regarding water balance difficult to unravel.

The logistic regression results indicate that water balance conditions at a site are a relatively weaker influence on Wood Turtle habitat preference than are finer- scale factors. The modeling consistently found ground-level relative humidity and canopy openness to be the stronger factors. At a site in the Ridge and Valley province of Pennsylvania, Strang (1983) found Wood Turtle activity (daily path length) to correlate positively with relative humidity. At my study site, tests found little correlation between ground-level relative humidity and water balance metrics.

The regression models with months pooled, however, did show a preference by

WV turtles for sites with higher actual evapotranspiration. The monthly modeling results also suggested that turtles’ microhabitat preferences may vary over the summer; e.g., WV turtles showed a preference for sites with higher AET only in

July. The monthly modeling consistently indicated a preference for relatively higher ground level humidity in July and August in both states. As August is typically drier than the other months, with 82mm of precipitation on average (vs. 93mm in June and 91mm in July), and July is hotter, with a mean of 23.4°C (vs. 21.0°C in June

162 and 22.5°C in August) (NCDC 2013), it makes sense that spatial preferences related to humidity and temperature may be more pronounced in those two months. The modeling also indicated a preference for lower ground temperatures in July, particularly in Virginia, which is understandable considering the inverse correlation between temperature and humidity. Temperatures are influenced not just by topography and overlying vegetation, but also, due to their effect on evaporation and transpiration, by soil moisture and near-ground humidity (Fridley 2009).

Throughout the summer, VA turtles preferred sites with a more open canopy than were available at random, whereas for WV turtles the regression modeling suggested higher canopy openness was preferred only in August. That result may be due to open canopy conditions being generally more available in WV than VA; random points in WV were on average 16.0% open vs. 12.5% in VA (Table 4.5).

Though varying degrees of canopy openness were available, the study sites overall as well as the great majority of Wood Turtle locations had high degrees of forest cover. Although reductions in the overstory canopy generally result in reduced relative humidity and moisture at the ground surface and increased mean temperature, temperature fluctuations, and solar radiation (Collins et al. 1985,

Currylow et al. 2012), it is also true that greater canopy openness does not necessarily translate to more open conditions or higher temperatures on the ground

(i.e., at the scale/level that a Wood Turtle lives); the density of overlying ground vegetation can actually increase under conditions of greater canopy openness

(Krichbaum pers. obs.). In this way, sites such as small canopy gaps could provide

163 both shaded and exposed microhabitats that allow for a range of humidity and temperature conditions, thus allowing for efficient osmo- and thermo-regulatory shuttling. In addition, by allowing for a greater range of forest floor light levels and temperature regimes, gaps allow for more floristic richness and/or abundance

(Goldblum 1997, Anderson and Leopold 2002), i.e., foraging opportunities. That

Wood Turtles react to fine-scale gradients is evident in their regular use of small natural canopy gaps (Remsburg et al. 2006, Krichbaum this study).

In sum, there may be some habitat preference/selection by Wood Turtles based upon ground level humidity or conditions related to water balance. That there were few significant differences between water balance metrics at turtle locations and those at the GIS-generated random points suggests that, from a water balance perspective, habitat suitable for Wood Turtles is available throughout the two study sites. In fact, in some cases conditions outside of the turtle points had

lower water deficits (see August DEF25 at Fig. 4.5). During this study, I commonly observed Wood Turtles using comparatively dry macrohabitats (e.g., Virginia Pine or Chestnut Oak sites, warmer aspects; Chapters 5 and 6). In addition, turtles, particularly those in WV, tended to use slopes with warmer aspects than were randomly available (Table 4.5); south-facing slopes generally experience higher temperatures and lower moisture than north-facing slopes (Cantlon 1953). The preference by Wood Turtles for higher ground surface relative humidity than is available at random is ostensibly in conflict with the result of the water balance

comparisons that found no differences in DEF25 between turtle points and their

164 paired random locations for some months. But this too is evidence that forested sites throughout, even those considered dry or submesic, are heterogeneous at a fine scale and can provide suitable microclimates.

Multiple factors may account for the ambiguous within-activity area preference by Wood Turtles regarding broad-scale water balance conditions. Wood

Turtles, at least at some spatial scales, are habitat generalists (sensu Paterson et al.

2012) in that they used a broad range of slopes and aspects (Table 4.5) as well as ten different forest types (Table 5.4). Deciduous, pine, and mixed pine-deciduous tracts were used, all of which have different water balance signatures (Anning et al.

2014). Perhaps the Turtle’s omnivory serves to drive it into a broad range of site conditions. Alternatively, turtles at these study sites may actually not prefer relatively more mesic conditions. Ernst’s 1968 study on EWL is often cited in support of this purported preference by this species. That study, however, to control for size in comparisons among three chelonian species, used juvenile Wood Turtles as experimental subjects. But such smaller individuals might experience substantially higher rates of evaporative water loss than do adults and so do not accurately represent constraints on habitat use by adults. Then again, it may be that the adult turtles here do prefer mesic conditions, but that suitable moisture conditions are broadly distributed (meaning that conditions at turtle points and their paired random points are similar) so that turtles generally can move at will within their activity area. Moreover, the ostensible lack of preference may be an artifact of the spatial scale of such preferred conditions, meaning that these

165 conditions occur at smaller spatial scales than that used herein as the basis of analysis for the water balance model (viz., 400m2). There may be a gross mismatch between the fine-scale habitat patch chosen by an animal and the more coarse- scale vegetation cover types within which these small sites are embedded; e.g., a patch of mesic deciduous forest might occur within a larger mapped stand of more xeric pine, and preferred patches may even occur at much smaller scales, such as moist conditions affiliated with a topographic depression or large downed bole within a 400m2 plot. Though relatively more mesic conditions at a fine-scale may be nested within 400m2 tracts throughout the study area (in tracts with both high or low DEF or AET), they are not necessarily homogeneously distributed; this may help account for the results of the analysis of ground-level humidity between turtle and paired random points. The water balance modeling used herein is based in part upon broad-scale spatial areas (e.g., tracts with a consistent soil type) and long-term conditions of climate (e.g., thirty year averages of monthly precipitation).

In contrast, the ground-level humidities measured and analysed were determined by the vicissitudes of weather (e.g., recent rainfall) and very localized microhabitat conditions (such as amounts of humidity-trapping shade, i.e., canopy openness).

These more fine-scale temporal and spatial patterns might override or counter or mitigate the effects of the water balance regime. Because turtles, unlike adult floristic organisms, are mobile, they can travel to and between preferred microclimatic sites that are found within or outside larger patches regardless of these patches’ overall water balance characteristics.

166 As a related matter, though water balance conditions within activity areas may not be a strong influence on spatial ecology at this scale, it may be important at a different order of selection. It could be that populations select areas with suitable water balance conditions out of the broader area of available habitat (so- called 2nd order selection; Johnson 1980); the non-significant results for comparisons between turtle points and the GIS generated random points suggest this possibility (Table 4.8). Perhaps part of the reason that populations of Wood

Turtles occupy these sites is due to the overall availability there of preferred water balance conditions.

The lack of statistically significant differentiation between turtle and random points could also be due to factors involving the application of the water balance methodology itself, any of which could confound results. The values of individual pixels were pooled and averaged into a single grid cell for analysis (the 400m2 plots). This mean vector from the individual 9m2 locations may not adequately represent a heterogeneous plot as a whole, i.e., a significant loss of information. In addition, though 9m2 qualifies as a fine grain in many ecological applications, even this may be too coarse to resolve the differences in habitat conditions too which turtles may be responding. For instance, leaf litter of varying depths and distributions can provide different micro-climates and micro-habitats, as can large woody debris and micro-topography such as small depressions; these conditions can occur at scales less than 9m2. Further, deficit was calculated based on the mapping of soil types, whereby identical values of soil depth and texture are

167 ascribed to an entire block of a particular soil type to estimate AWC. The areal extent of a particular type of soil can be tens of hectares in size (Fig. 4.6), but it can be expected that the depth and texture of soil over an extent of this scale will vary, thereby producing finer scale heterogeneity in AWC of which the model is incapable of reflecting. Moreover, water storage capacity of many Central

Appalachian soils can be severely restricted due to high rock content

(Armson1979), which can vary at multiple spatial scales. Also, all the cells in the entire study site were assigned the same monthly temperature and precipitation values. However, climatic differences can occur at fine spatial scales, particularly in mountainous terrain, due to factors such as divergent or convergent topographies and cold air drainage (Fridley 2009, Dobrowski 2010). Furthermore, the solar radiation values used to calculate PET with the Turc method were collected at a location approximately 80km from the study site, which also could have conceivably skewed results.

Because an animal’s use of habitat is the outcome of multiple interacting factors (Morris 1992, Edwards et al. 2003, Storch 2003, Bollmann et al. 2005,

Girvetz and Greco 2009, Owen-Smith et al. 2010, Baxley et al. 2011, Del Vecchio et al. 2011), there is much more than site moisture patterns influencing Wood

Turtle spatial ecology and their occurence points, such as human disturbance, predator avoidance, prey/forage distribution, forest structure and composition, and thermoregulatory opportunities. The availability, for example, of such forage items as herbs or fungi may be occurring at finer scales than the model’s grain or be due

168 to factors other than site moisture regimes (e.g., amounts of LWD on site or past disturbance history). As omnivores, Wood Turtles perhaps do not respond directly to soil moisture patterns, but instead respond to the flora or fauna that are influenced by such patterns. For example, heterogeneity in soil fertility and drainage, slope, aspect, and within-stand moisture gradients are important influences on herbaceous diversity (Hicks 1980, Rogers 1982, Nichols et al. 1998).

Hence, detection of habitat preference is confounded by numerous fine-scale subtleties (Edge 2010).

Faunal habitat suitability modeling often uses coarse-grained vegetation cover types (i.e., broad classifications such as “forest” or “cropland”) to construct spatial predictions (Edwards et al. 2003, Schlossberg and King 2009). However, for some species, such as those of small size or limited vagility (e.g., the Wood Turtle), the identification, modeling, and analysis of the habitat mosaic must be made with a finer resolution (Dillard et al. 2008, Suggitt 2011, Farallo and Miles 2016). The coarseness of scale as well as the lack of spatial explicitness that typify many wildlife habitat relation models or habitat suitability indices (that ascribe varying scores to different habitats based upon their suitability) are overcome with this GIS- based water balance model. Spatially explicit pixels with a resolution of 9m2 were used for the entirety of a 10 X 8 km study area. Such a fine scale is likely to be appropriate to many organisms of focus, in that it is of a grain size that is both detectable and selectable. Moreover, the environmental variables associated with

169 site water balance may have direct ecological relationships with the Wood Turtle or other focal organisms.

Regardless of its limited revelation of habit differences in this particular application, the water-balance method still has potential for modeling habitat relationships. Application of the water balance model would be most useful in situations where fine-scale data such as I collected here are not available. A clear advantage to this method is its use of readily available remotely sensed data; the model is practical, measurable, and understandable. It may provide a way to accurately predict potential Wood Turtle or other species occurrences at other locales or at other time periods. The water balance model provides forest managers a means to identify measurable soil- and moisture-related components suitable for a focal organism’s occurrence and allows for an assessment of the distribution and availability of such habitat patches (see, e.g., Fig. 4.5). This can be useful for habitat conservation decision-making as well as provide a means for focusing field survey locations or identifying potential relocation sites. Finally, since it incorporates quantified values for temperature and precipitation in addition to topography and soil available water capacity, the model is able with precision to evaluate and model biotic responses to changing climate.

170 Table 4.1

Values for ground-level relative humidity at turtle and paired random points Values for ground-level relative humidity (%) and temperature (C°) variables obtained at turtle points and paired random points in Virginia (VA) and West Virginia (WV) during June-August 2011-2014; reported in descending order are means, standard errors, and ranges. HumT = humidity immediately beside turtles when located, HumGr/TempGr = humidity/temperature in shade when turtles were located, HumVeg = humidity in shade measured when vegetation data were obtained at plots. FWT = female Wood Turtle points, FRP = female random points, MWT = male Wood Turtle points, MRP = male random points.

VA WV

FWT FRP MWT MRP FWT FRP MWT MRP HumT 65.1 – 63.9 – 63.6 – 59.6 – 2011-2012 1.4 – 3.0 – 2.7 – 2.6 – 38-89 – 38-81 – 31-84 – 41-83 –

HumT 70.4 – 65.3 – 60.8 – 58.2 – 2011-2014 1.0 – 1.6 – 1.7 – 1.8 – 38-96 – 38-91 – 31-94 – 38-86 –

HumGr 65.0 59.1 62.6 56.3 63.8 54.4 60.4 55.3 2011-2012 1.4 1.8 3.4 2.9 2.4 2.7 3.0 3.0 38-89 32-89 31-81 36-77 44-84 31-79 42-83 27-80

HumGr 70.1 – 64.2 – 61.4 – 60.3 – 2011-2014 1.0 – 1.7 – 1.5 – 2.3 – 38-96 – 31-91 – 39-94 – 38-93 –

HumVeg 76.7 73.1 76.7 74.2 60.1 53.7 65.6 59.1 2013-2014 1.1 1.3 1.7 1.8 2.2 2.0 3.2 3.4 47-95 42-93 54-96 52-89 36-96 35-91 40-93 37-91

HumVeg 72.8 69.0 72.6 70.1 58.6 52.9 61.1 56.4 2011-2014 1.1 1.1 1.7 1.8 2.0 1.7 2.3 2.3 32-95 26-93 32-96 37-89 31-96 27-91 27-93 32-91

TempGr 25.8 26.8 26.4 27.4 27.9 29.3 28.1 28.7 2011-2012 2.5 3.2 2.9 2.3 2.6 3.0 2.1 2.5 19-31 20-35 18-30 22-32 23-34 25-36 24-32 24-35

TempGr 25.1 – 25.6 – 26.8 – 27.0 – 2011-2014 2.8 – 2.7 – 2.9 – 2.9 – 16-33 – 18-30 – 21-34 – 21-35 –

171 Table 4.2

Results of paired t-tests comparing humidity at turtle and random points Results of paired t-tests for ground surface relative humidity variables obtained at turtle points and paired random points in Virginia (VA) and West Virginia (WV) during June-August 2011-2014; reported in descending order are p values, t statistic values, and degrees of freedom. HumGr = humidity in shade when/where turtles were located, HumT = humidity immediately adjacent to turtles when located, HumVeg = humidity in shade measured when vegetation data were obtained at plots. FWT = female Wood Turtle points, FRP = female random points, MWT = male Wood Turtle points, MRP = male random points. Comparisons with significant results are in bold. HumVeg test for VA females used paired Wilcoxon signed rank test. The HumT-HumGr comparisons used only turtle points. Tests involving HumGr 2011-2012 had α = 0.025, HumGr 2011-2014 α = 0.0063, HumVeg 2013-2014 α = 0.0083. VA WV FWT-FRP MWT-MRP FWT-FRP MWT-MRP HumGr 2011-2012 p <0.00001 0.0132 0.00018 0.244 t 5.698 -2.766 4.595 -1.212 df 52 17 20 15

HumVeg 2013-2014 p <0.0001 0.0100 0.0002 0.0091 t 2316 -2.744 -4.647 -2.847 df – 31 48 23

HumT-HumGr p 0.193 0.130 0.2978 0.2223 2011-2014 t 1.310 1.538 1.050 1.244 df 131 51 61 33

172 Table 4.3

Ground level relative humidity at pooled turtle and paired random points Values for ground level relative humidity variables obtained at turtle points and random points in Virginia (VA) and West Virginia (WV) during June-August 2011- 2014; reported in descending order are means, standard errors, and ranges. HumT = humidity immediately adjacent to turtles when located, HumGr = humidity in shade when/where turtles were located, HumVeg = humidity in shade measured when vegetation data were obtained at plots. WT = Wood Turtle points (sexes pooled), RP = random points (sexes pooled), FWT = female Wood Turtle points, MWT = male Wood Turtle points. VA WV VA & WV

WT RP WT RP FWT MWT WT RP HumT 69.0 – 59.8 – 67.1 62.0 65.5 – 0.87 – 1.25 – 0.94 1.27 0.77 – 38-97 – 31-94 – 31-97 38-91 31-97 –

HumGr 68.4 – 61.0 – 67.3 62.6 65.8 – 0.89 – 1.27 – 0.88 1.41 0.76 – 31-97 – 38-94 – 38-97 31-93 31-97 –

HumVeg 72.7 69.3 59.7 54.3 67.9 67.1 67.6 63.3 0.92 0.97 1.50 1.38 1.09 1.55 0.89 0.91 32-96 26-93 27-96 27-91 31-96 27-96 27-96 26-93

173 Table 4.4

ANOVA results for ground level relative humidity variables Results of two-way ANOVAs, with state and sex as fixed effects and individual turtle as the random effect, comparing ground level relative humidity variables obtained at turtle points (WT) in Virginia and West Virginia during June-August 2011-2014. Reported are p values, F statistics, and degrees of freedom. HumT = humidity immediately adjacent to turtles when located, HumGr = humidity in shade when/where turtles were located, HumVeg = humidity in shade measured when vegetation data were obtained at plots. As HumVeg data were not normal, individual Wilcoxon rank sum tests were used in separate tests for sex and state, and separate Kruskal-Wallis rank sum tests for turtle (WT) and random points (RP) were used to test for differences in the groups VFT-VMT-WFT-WMT and VTR-VMR-WFR-WMR; V = Virginia, W = West Virginia, FT = female Wood Turtles, MT = male Wood Turtles, FR = female random points, MR = male random points. Comparisons with significant results are in bold. Tests involving HumT had alpha = 0.0166, HumGr alpha = 0.0063, HumVeg alpha = 0.0083. State Sex State:Sex

HumT (WT) p <0.0001 0.0224 0.491 F 29.709 5.416 0.479 df 1, 84 1, 84 1, 84

HumGr (WT) p <0.001 0.0273 0.207 F 18.260 5.050 1.615 df 1, 82 1, 82 1, 82

HumVeg (WT) p <0.00001 0.574 <0.00001 W/K-W 15.372 10041 49.996 df - - 3

HumVeg (RP) p <0.00001 0.996 <0.00001 W/K-W 15997 9296 66.955 df - - 3

174 Table 4.5

Values for environmental variables at turtle points and paired random points for 2011-2012 Values for variables obtained at turtle points and paired random points in Virginia and West Virginia during June-August 2011-2012; reported in descending order are means, standard errors, and ranges. HumGr = ground level relative humidity (%) in shade when turtles were located, TempGr = ground level temperature (°C) in shade when turtles were located, Canopy = amount of overhead canopy openness (%), Gapsize = ground area (m2) in 400m2 plot under canopy gaps ≥ 9m2, Aspect = orientation of slope (Beers transform of degrees results in values from 0 to 2, with 2 being the coolest (NE aspects) and 0 the warmest (SW aspects)), Slope = inclination in degrees. WT = Wood Turtle points (males and females pooled), RP = paired random points (male and female points pooled). Virginia West Virginia WT RP WT RP HumGr 64.4 58.4 61.0 54.8 1.4 1.5 1.9 2.0 31-89 32-89 29-84 27-80

TempGr 25.9 27.0 28.0 29.0 0.3 0.4 0.4 0.4 18-31 20-35 23-34 25-36

Canopy 20.6 12.5 17.0 16.0 1.9 0.5 1.1 1.3 6-78 6-40 8-40 10-66

Gapsize 53.1 11.0 27.2 14.7 9.7 2.4 6.0 3.7 0-352 0-85 0-216 0-160

Aspect 0.92 1.02 0.67 1.07 0.1 0.1 0.1 0.1 0-2 0-2 0-2 0-2

Slope 6.5 10.9 9.1 19.4 0.7 0.7 1.2 1.3 1-31 2-28 1-31 2-38

175 Table 4.6

Water balance values (AET, DEF, PET) at 400m2 plots Values (mm) estimated with Dyer water balance model for means (± standard errors) of water deficit (DEF25/100), potential evapotranspiration 2 (PET), and actual evapotranspiration (AET, calculated with DEF100) in 400m plots at two forested montane sites in VA and WV. WT = turtle plots, TR = paired random plots, RP = random plots generated with GIS. “n” refers to number of plots, not number of individual turtles.

DEF25 DEF100 PET AET VA June WT (n = 63) 30.2 ± 0.3 7.4 ± 0.4 123.5 ± 0.3 116.1 ± 0.4 June TR (n = 63) 29.8 ± 0.4 8.6 ± 0.4 123.1 ± 0.4 114.4 ± 0.4 June RP (n = 2087) 30.4 ± 0.1 8.8 ± 0.1 123.6 ± 0.1 114.8 ± 0.1

July WT (n = 70) 37.6 ± 0.3 21.4 ± 0.7 128.8 ± 0.3 107.4 ± 0.6 July TR (n = 70) 37.7 ± 0.3 21.8 ± 0.7 129.0 ± 0.3 107.2 ± 0.7 July RP (n = 2087) 37.7 ± 0.1 20.7 ± 0.1 128.9 ± 0.1 108.2 ± 0.1

August WT (n = 38) 24.4 ± 0.7 18.2 ± 0.9 105.9 ± 0.7 87.7 ± 0.8 August TR (n = 38) 26.3 ± 0.6 19.9 ± 0.9 108.1 ± 0.6 88.2 ± 0.6 August RP (n = 2087) 24.8 ± 0.1 17.4 ± 0.1 106.8 ± 0.1 89.4 ± 0.1

WV June WT (n = 34) 23.4 ± 0.8 5.9 ± 0.4 117.3 ± 0.7 111.4 ± 0.6 June TR (n = 34) 21.7 ± 1.2 7.1 ± 0.6 115.4 ± 1.1 108.2 ± 0.6 June RP (n = 760) 23.3 ± 0.2 7.5 ± 0.2 117.0 ± 0.2 109.5 ± 0.4

July WT (n = 48) 30.9 ± 0.6 16.0 ± 0.7 122.2 ± 0.6 106.2 ± 0.5 July TR (n = 48) 27.2 ± 1.1 15.9 ± 1.0 118.4 ± 1.0 102.5 ± 0.4 July RP (n = 760) 30.7 ± 0.2 19.1 ± 0.2 121.9 ± 0.2 102.8 ± 0.1

August WT (n = 29) 21.8 ± 1.3 16.5 ± 1.3 103.5 ± 1.5 87.0 ± 0.7 August TR (n = 29) 15.8 ± 1.7 11.8 ± 1.4 97.4 ± 1.8 85.5 ± 1.2 August RP (n = 760) 20.3 ± 0.3 16.3 ± 0.3 102.0 ± 0.3 85.8 ± 0.4

176 Table 4.7 Wilcoxon test results for water balance at turtle and paired random points P-values (with V-statistic values below) for Wilcoxon paired signed rank tests between turtle points and paired random points (males and females combined). n = number of turtle points and the equal number of paired random points. α = 0.025 for each test.

DEF25 AET VA June (n = 63) 0.835 0.004 1039 1385

July (n = 70) 0.788 0.721 1289 1049 August (n = 38) 0.016 0.126 193 250 WV

June (n = 34) 0.197 0.0016 374 458 July (n = 48) 0.009 <0.0001 840 1018

August (n = 29) 0.0129 0.214 331 276

177 Table 4.8 Wilcoxon test results for water balance at turtle points and GIS-generated random points P-values (with W-statistic values below) for Wilcoxon rank sum tests between turtle points and GIS generated random points within 295m buffer zones around main streams. n = number of turtle points (males and females combined); number of random points = 2087 in VA, 760 in WV. α = 0.025 for each test.

DEF25 AET VA June (n = 78) 0.277 <0.0001 75503 104317 July (n = 103) 0.200 0.098 115516 97127 August (n = 75) 0.403 0.080 70783 68972 WV June (n = 51) 0.056 0.355 16288 20880 July (n = 74) 0.074 <0.0001 24582 45696 August (n = 51) 0.552 0.480 20345 20523

178 Table 4.9

Best conditional logistic regression model variables – using AET Conditional logistic regression model variables that best explain habitat preference for summer months pooled by Wood Turtles at sites in Virginia and West Virginia, USA in June-August 2011-2012, using AET in models. Measured values were obtained at turtle plots and paired random plots. Model coefficient values were obtained through model averaging. Positive values for variable coefficients indicate preference, while negative values indicate avoidance. Variables listed had coefficients that did not overlap zero.

Measured values (mean ± se) Model coefficient ± se Odds ratio Unit increase Variable Turtle Plot Random Plot Virginia Wood Turtles HumGr 65.1 ± 1.51 58.6 ± 1.71 0.20 ± 0.07 1.221 1 % TempGr 26.0 ± 0.31 27.2 ± 0.37 -0.76 ± 0.32 0.468 1° Canopy 20.8 ± 2.21 12.3 ± 0.60 0.13 ± 0.05 1.139 1 %

West Virginia Wood Turtles AET 102.8 ± 2.0 98.4 ± 1.80 0.17 ± 0.07 1.185 1 mm HumGr 59.7 ± 2.30 54.1 ± 2.24 0.32 ± 0.12 1.377 1 %

179 Table 4.10

Well-supported conditional logistic regression models – using AET and months pooled Well-supported conditional logistic regression models of Wood Turtle habitat preference for summer months pooled at sites in Virginia and West Virginia, USA in 2011-2012, based on habitat variables at turtle and random plots. AET = actual evapotranspiration (mm) estimated with Dyer water balance model using top 100cm of soil depth for available water capacity, Canopy = amount of canopy openness (estimated with a spherical densitometer), Gapsize = amount of ground area (m2) in the 400m2 plot under canopy gaps, HumGr = ground level relative humidity (%) in the shade, TempGr = ground level temperature (ºC) in the shade. Models in bold are the best models. LogLik = model log-likelihood; K = number of parameters in model; ∆AICc= difference in Akaike Information Criterion corrected for small sample size from the top model; higher AICc weights denote models that are better supported among the set of candidate models; cumulative weight (Cum. wt.) is the running sum of the individual model weights.

LogLik K ∆AICc AICc Cum.wt. Model weight

Virginia Wood Turtles HumGr+Gapsize -22.06 2 0 0.27 0.27 HumGr -23.32 1 0.45 0.22 0.49 TempGr+Gapsize -22.75 2 1.37 0.14 0.63 AET+HumGr+Gapsize -21.70 3 1.39 0.14 0.76 Canopy+TempGr -22.80 2 1.48 0.13 0.89 AET+HumGr -23.03 2 1.95 0.10 0.99 1 [Null model] -38.12 0 29.19 0.00 1.00

West Virginia Wood Turtles AET+HumGr+Gapsize -10.20 3 0 0.51 0.51 AET+HumGr -11.52 2 0.44 0.40 0.91 HumGr+Gapsize -13.38 2 4.17 0.06 0.97 1 [Null model] -24.95 0 23.14 0.00 1.00

180 Table 4.11

Best conditional logistic regression model variables – using monthly AET Conditional logistic regression model variables that best explain monthly habitat preference by Wood Turtles at sites in Virginia and West Virginia, USA in 2011- 2012; some candidate models included AET (except for June in WV – models with AET or HumGr would not converge). Measured values were obtained at turtle plots and paired random plots. Model coefficient values were obtained through model averaging. Positive values for variable coefficients indicate preference, while negative values indicate avoidance. Variables in bold are those with coefficients that did not overlap zero.

Measured values (mean ± se) Model coefficient ± se Odds ratio Unit increase Variable Turtle Plot Random Plot Virginia Wood Turtles June Canopy 16.4 ± 2.05 10.8 ± 0.71 0.36 ± 0.25 1.433 1 % Gapsize 61.6 ± 26.7 13.3 ± 6.59 0.08 ± 0.05 1.083 1 m2 July HumGr 66.1 ± 2.25 59.7 ± 2.51 0.15 ± 0.08 1.162 1 % TempGr 26.4 ± 0.42 27.2 ± 0.47 -0.51 ± 0.33 0.600 1 ° Canopy 20.3 ± 2.96 12.4 ± 1.15 0.11 ± 0.07 1.116 1 % August HumGr 65.9 ± 2.80 61.9 ± 2.96 0.16 ± 0.12 1.174 1 % Canopy 24.9 ± 5.33 13.4 ± 0.61 0.11 ± 0.11 1.116 1 %

West Virginia Wood Turtles June Canopy 13.7 ± 0.54 19.4 ± 6.72 0.12 ± 0.36 1.127 1 %

July AET 108.1 ± 0.89 102.1 ± 0.74 0.41 ± 0.24 1.507 1 mm HumGr 62.4 ± 3.12 56.2 ± 3.13 0.29 ± 0.23 1.336 1 % August Canopy 22.0 ± 2.31 13.7 ± 0.96 0.23 ± 0.15 1.259 1 % HumGr 52.1 ± 3.12 48.0 ± 4.05 0.40 ± 0.32 1.492 1 %

181 Table 4.12

Well-supported conditional logistic regression models – using monthly AET Well-supported conditional logistic regression models of monthly Wood Turtle habitat preference at sites in Virginia and West Virginia, USA in 2011-2012, based on habitat variables at turtle and random plots. AET = actual evapotranspiration estimated with Dyer water balance model using top 100cm of soil depth for available water capacity, Canopy = proportion of canopy openness (estimated with a spherical densiometer), Gapsize = amount of ground area (m2) in the 400m2 plot under canopy gaps, HumGr = ground level relative humidity (%) in the shade, TempGr = ground level temperature (ºC) in the shade. Models in bold are the best models. LogLik = model log-likelihood; K = number of parameters in model; ∆AICc= difference in Akaike Information Criterion corrected for small sample size from the top model; higher AICc weights denote models that are better supported among the set of candidate models; cumulative weight (Cum. wt.) is the running sum of the individual model weights.

LogLik K ∆AICc AICc Cum.wt. Model weight

Virginia Wood Turtles June [no models included HumGr or TempGr] Gapsize -4.40 1 0 0.40 0.40 Canopy -4.59 1 0.38 0.33 0.73 1 [Null model] -8.32 0 5.64 0 02 0.98 July HumGr + Canopy -10.68 2 0 0.22 0.22 HumGr -12.14 1 0.76 0.15 0.38 HumGr + Gapsize -11.40 2 1.44 0.11 0.49 AET + HumGr + Canopy -10.56 3 2.01 0.08 0.57 1 [Null model] -18.02 0 10.45 0.00 1.00 August Canopy -9.02 1 0 0.14 0.14 HumGr -9.10 1 0.16 0.13 0.28 HumGr + Canopy -8.24 2 0.71 0.10 0.38 Gapsize -9.49 1 0.95 0.09 0.47 HumGr + Gapsize -8.45 2 1.12 0.08 0.55 TempGr + Gapsize -8.80 2 1.83 0.06 0.61 TempGr + Canopy -8.82 2 1.87 0.06 0.66 1 [Null model] -11.78 0 3.41 0.03 0.90

West Virginia Wood Turtles June [no models included AET or HumGr] TempGr -1.52 1 0 0.76 0.76 1 [Null model] -5.55 0 5.76 0.04 0.96 July AET + HumGr -3.04 2 0 0.27 0.27 AET + HumGr + Gapsize -2.46 3 1.24 0.15 0.42 AET -4.97 1 1.60 0.12 0.55 AET + Gapsize -3.85 2 1.62 0.12 0.67 1 [Null model] -11.78 0 13.10 0.00 1.00 August Canopy -3.83 1 0 0.25 0.25 HumGr -4.21 1 0.76 0.17 0.43 HumGr + Canopy -3.33 2 1.44 0.12 0.55 1 [Null model] -7.62 0 5.38 0.02 0.94

182 Table 4.13

Best conditional logistic regression model variables – using DEF25 and months pooled Conditional logistic regression model variables that best explain habitat preference for summer months pooled by Wood Turtles at sites in Virginia and West Virginia,

USA in 2011-2012, using DEF25 in models. Measured values were obtained at turtle plots and paired random plots. Model coefficient values were obtained through model averaging. Variables with positive values for coefficients indicate preference, while negative values indicate avoidance. Variables in bold are those with coefficients that did not overlap zero.

Measured values (mean ± se) Model coefficient ± se Odds ratio Unit increase Variable Turtle Plot Random Plot

Virginia turtles HumGr 65.1 ± 1.51 58.6 ± 1.71 0.19 ± 0.06 1.209 1 % TempGr 26.0 ± 0.31 27.2 ± 0.37 -0.76 ± 0.32 0.468 1 º Canopy 20.8 ± 2.21 12.3 ± 0.60 0.10 ± 0.06 1.105 1 %

West Virginia turtles DEF25 26.9 ± 1.34 22.5 ± 1.76 0.14 ± 0.07 1.150 1 mm HumGr 59.7 ± 2.30 54.1 ± 2.24 0.30 ± 0.10 1.350 1 % TempGr 28.3 ± 0.46 29.1 ± 0.47 -0.52 ± 0.25 0.595 1 º

183 Table 4.14

Well-supported conditional logistic regression models – using DEF25 and months pooled Well-supported conditional logistic regression models of Wood Turtle habitat preference for summer months pooled at sites in Virginia and West Virginia, USA in

2011-2012, based on habitat variables at turtle and random plots. DEF25 = water deficit (mm) estimated with Dyer water balance model using only top 25cm of soil depth for available water capacity, Canopy = proportion of canopy openness (estimated with a spherical densitometer), Gapsize = amount of ground area (m2) in the 400m2 plot under canopy gaps, HumGr = ground level relative humidity (%) in the shade, TempGr = ground level temperature (ºC) in the shade. Models in bold are the best models. LogLik = model log-likelihood; K = number of parameters; ∆AICc= difference from the top model in Akaike Information Criterion corrected for small sample size; higher AICc weights denote models that are better supported among the set of candidate models; cumulative weight (Cum. wt.) is the running sum of the individual model weights.

LogLik K ∆AICc AICc Cum.wt. Model weight

Virginia Wood Turtles HumGr + Canopy -21.04 2 0 0.38 0.38 DEF25 + HumGr + Canopy -20.94 3 1.91 0.14 0.52 HumGr + Gapsize -22.06 2 2.04 0.14 0.66 1 [Null model] -38.12 0 30.05 0.00 1.00

West Virginia Wood Turtles DEF25 + HumGr -11.25 2 0 0.59 0.59 DEF25 + HumGr + Gapsize -10.82 0 1.32 0.31 0.90 1 [Null model] -24.95 0 23.23 0.00 1.00

184 Table 4.15

Best conditional logistic regression model variables – using monthly DEF25 Conditional logistic regression model variables that best explain monthly habitat preference by Wood Turtles at sites in Virginia and West Virginia, USA in 2011-

2012, some models included DEF25. Measured values were obtained at turtle plots and paired random plots. Model coefficient values were obtained through model averaging. DEF25 = water deficit (mm) estimated with Dyer water balance model using only top 25cm of soil depth for available water capacity, Canopy = proportion of canopy openness (%) at plot, Gapsize = amount of ground area (m2) in the 400m2 plot under canopy gaps, HumGr = ground level relative humidity (%) in the shade, TempGr = ground level temperature (ºC) in the shade. Positive values for coefficients indicate preference, while negative values indicate avoidance. Variables in bold are those with coefficients that did not overlap zero.

Measured values (mean ± se) Model coefficient ± se Odds ratio Unit increase Variable Turtle Plots Random Plots

Virginia Wood Turtles June Canopy 16.4 ± 2.05 10.8 ± 0.71 0.38 ± 0.26 1.462 1 % Gapsize 61.6 ± 26.7 13.3 ± 6.59 0.09 ± 0.06 1.094 1 sq.m.

July HumGr 66.1 ± 2.25 59.7 ± 2.51 0.16 ± 0.08 1.174 1 % TempGr 26.4 ± 0.42 27.2 ± 0.47 -0.52 ± 0.34 0.595 1 º Canopy 20.3 ± 2.96 12.4 ± 1.15 0.11 ± 0.07 1.116 1 %

August HumGr 65.9 ± 2.80 61.9 ± 2.96 0.17 ± 0.12 1.185 1 % Canopy 24.9 ± 5.33 13.4 ± 0.61 0.10 ± 0.10 1.105 1 %

West Virginia Wood Turtles June DEF25 26.1 ± 0.67 23.9 ± 2.34 0.95 ± 0.84 2.586 1 mm

July HumGr 62.4 ± 3.12 56.2 ± 3.13 0.55 ± 0.54 1.733 1 %

August Canopy 22.0 ± 2.31 13.7 ± 0.96 0.22 ± 0.15 1.246 1 % HumGr 52.1 ± 3.12 48.0 ± 4.05 0.40 ± 0.33 1.492 1 %

185 Table 4.16

Well-supported conditional logistic regression models – using monthly DEF25 Well-supported conditional logistic regression models of monthly Wood Turtle habitat preference at sites in Virginia and West Virginia, USA in 2011-2012, based on habitat variables at turtle and random plots. DEF25 = water deficit (mm) estimated with Dyer water balance model using only top 25cm of soil depth for available water capacity, Canopy = proportion of canopy openness (estimated with a spherical densiometer), Gapsize = amount of ground area (m2) in the 400m2 plot under canopy gaps, HumGr = ground level relative humidity (%) in the shade, TempGr = ground level temperature (ºC) in the shade. Models in bold are the best models. LogLik = model log-likelihood; K = number of parameters; ∆AICc= difference in Akaike Information Criterion corrected for small sample size from the top model; higher AICc weights denote models that are better supported among the set of candidate models; cumulative weight (Cum. wt.) is the running sum of the individual model weights.

LogLik K ∆AICc AICc Cum.wt. Model weight

Virginia Wood Turtles June [no models included HumGr or TempGr] Gapsize -4.40 1 0 0.34 0.34 Canopy -4.59 1 0.38 0.28 0.61 DEF25 + Gapsize -3.66 2 0.91 0.21 0.83 DEF25 + Canopy -4.03 2 1.65 0.15 0.97 1 [Null model] -8.32 0 5.64 0.02 0.99 July HumGr + Canopy -10.68 2 0 0.21 0.21 HumGr -12.14 1 0.76 0.15 0.36 HumGr + Gapsize -11.40 2 1.44 0.10 0.46 DEF25 + HumGr + Canopy -10.37 3 1.63 0.09 0.56 1 [Null model] -18.02 0 10.45 0.00 1.00 August Canopy -9.02 1 0 0.13 0.13 HumGr -9.10 1 0.16 0.12 0.25 HumGr + Canopy -8.24 2 0.71 0.09 0.34 Gapsize -9.49 1 0.95 0.08 0.42 DEF25 + HumGr -8.38 2 0.99 0.08 0.49 HumGr + Gapsize -8.45 2 1.12 0.07 0.57 1 [Null model] -11.78 0 3.41 0.02 0.95

West Virginia Wood Turtles June [no models included HumGr] TempGr -1.52 1 0 0.58 0.58 1 [Null model] -5.55 0 5.76 0.03 0.95 July DEF25 + HumGr -2.88 2 0 0.57 0.57 DEF25 + Humgr + Gapsize -2.86 3 2.37 0.17 0.74 1 [Null model] -11.78 0 13.42 0.00 1.00 August Canopy -3.83 1 0 0.25 0.25 HumGr -4.21 1 0.76 0.17 0.43 HumGr + Canopy -3.33 2 1.44 0.12 0.55 DEF25 + Canopy -3.74 2 2.24 0.08 0.62 1 [Null model] -7.62 0 5.38 0.02 0.93 ______

186 CHAPTER 5: VEGETATIVE INDICATORS FOR WOOD TURTLE (GLYPTEMYS

INSCULPTA) HABITAT USE IN CENTRAL APPALACHIAN FORESTS

Introduction

Numerous interacting factors at multiple spatial scales generate structural and compositional heterogeneity in forests (Braun 1950, Runkle 1991b, Franklin et al. 2002, McEwan et al. 2010). Broad- and fine-scale distributional patterns of understory and overstory forest vegetation result from synergies of site-specific physical conditions, disturbance regimes, and biotic interactions (Watt 1947, Braun

1950, Swanson et al. 1988, DeMars and Runkle 1992, Callaway 1997, Pickett and

Rogers 1997, Hutchinson et al. 1999, Angelstam 2003, Dyer 2006, Dyer 2010,

Matlack and Schaub 2011, McEwan and Muller 2011, Chapman and McEwan

2012, Anning et al. 2014). Because plants affect environmental conditions and resource availability within areas where animal activities take place, vegetation patches pattern animal habitat use in manifold ways (Doak et al. 1992, Baxley and

Qualls 2009).

Forests provide not only food resources for resident animal species, but also refugia from predators, osmoregulatory opportunities (such as humid microclimates), thermoregulatory opportunities (shade and basking), cover from elements, reproductive staging areas, and nesting sites. Thus, spatial ecology may not be strongly linked to forage/prey abundance or composition, particularly for taxa such as turtles that do not have high rates of energy expenditure and may be process rather than resource limited (Congdon et al. 1989). Though vegetation

187 structure may be a more important driver of animal habitat preference than taxonomic composition (DeGraaf et al. 1998, Carter et al. 1999, Kearns et al. 2006,

Waldron et al. 2008, McCoy et al. 2013), for some chelonians specific floristic composition, not just amounts of general vegetation, may be a factor in microhabitat selection (Del Vecchio et al. 2011).

Previous studies have examined habitat use by turtles (Kaufmann 1992a,

Kazmaier et al. 2001, Compton et al. 2002, Refsnider and Linck 2012, McCoy et al.

2013), but most deal with vegetation composition and structure with the use of broad-scale cover type categories (e.g., deciduous forest or woodland), few have gone into detail on evidence of preference for specific taxa of trees or forbs (but see

Del Vecchio et al. 2011, McKnight 2011, McCoard et al. 2016b). Habitat is not necessarily synonymous with a vegetative “cover type”; many factors influence habitat selection by animals and a cover type such as “oak-hickory forest” may contain tracts of different ages, species, structural attributes, and microhabitats. At these heterogeneous tracts specific vegetative taxa may be directly preferred as forage or they may be associated with other preferred food resources (such as invertebrate prey) or habitat conditions (e.g., cover or microclimates). For this reason, to be of value for practical conservation application, habitat selection studies of taxa with activity areas of limited extent and fine-scale specificity of preferences, such as small turtles, must be performed at spatial and categorical scales of congruent detail.

188 The objective of this study was to determine if forest dwelling North

American Wood Turtles (Glyptemys insculpta) were associating with specific overstory tree taxa and ground floor flora. Identification of such species can serve to identify or predict locations that are suitable habitat for the turtles, i.e., serve as

“management indicator species” useful for protecting Wood Turtle populations and their habitat. A better understanding of Wood Turtle spatial ecology, informed by empirical data and statistical analyses, will help focus conservation efforts, especially where commercial logging, recreational activities, road construction, vehicular traffic, and other anthropogenic disturbances may occur (Gardner et al.

2007). Due to the heterogeneous nature of forest patches and the Wood Turtles’ omnivory, I hypothesized that some sites would be preferred more than others and predicted that a subset of the floristic taxa identified in the field would be correlated with Wood Turtle occurrences. Due to my observations of Wood Turtles at various locations over the course of eight years, I also predicted that most of the floristic taxa correlated with Wood Turtle occurrences would not be hydrophytes.

Focal Species

Wood Turtles (Glyptemys insculpta) are amphibious emydids found in deciduous, coniferous, and mixed forests in the northeastern United States (see

“Focal Species” in Chapter 1 for more details on their natural history). Wood Turtle foraging and ingestion occur in both terrestrial and aquatic settings, including underwater (Carroll 1999, Krichbaum pers. obs.). As omnivores, they use a wide variety of foods (Strang 1983, Kaufmann 1992, Niederberger and Seidel 1999, Ernst

189 2001, Compton et al. 2002, Walde et al. 2003, Ernst and Lovich 2009, Jones 2009,

Krichbaum pers. obs.). Turtle habitat use may be in response to fine-scale presence or abundances of litter invertebrates, fungi, or herbs that are distributed non- randomly in the forest (Meier et al. 1995, Caldwell 1996, Hanula 1996,

Hutchinson et al. 1999, Rubino and McCarthy 2003, Van de Poll 2004, Kappes

2006, Gilliam 2007). Strang (1983) and Kaufmann (1995) noted seasonal differences in terrestrial habitat use apparently in response to the variance in availability of fungi, herbs, berries, and slugs.

Though I have few direct personal observations of feeding, I assumed that ground floor plant taxa found at a site may be an important driver of turtle use of those sites. I have observed Wood Turtles feeding, or observed evidence of feeding

(such as pieces of foodstuffs on their faces), only 39 times from 2006 to 2015 in VA and WV (ca. 4% of my encounters with turtles); these observations took place in

March to October from 9:35-20:30. Almost half of these occasions (18) involved herbaceous leaves, with the only identifiable taxon being Viola spp. My other foraging observations involved mushrooms (7), earthworms (5), insects (3), slugs or snails (3), fruit (2: blackberries and Skunk Cabbage), and a crustacean (crayfish).

Others have reported Wood Turtles feeding on cinquefoil (Potentilla spp.), wood sorrel (Oxalis spp.), greenbrier (Smilax spp.), Partridgeberry (Mitchella repens), Prairie Ragwort (Senecio plattensis), violets (Viola spp.), fruit of Skunk

Cabbage (Symplocarpus foetidus), grasses, blueberries (Vaccinium spp.), blackberries (Rubus spp.), leaves and fruit of strawberries (Fragaria spp.), leaves of

190 willows (Salix spp.) and alders (Alnus spp.) and birches (Betula spp.), new growing tips of ferns, moss, algae, fungi (e.g., Amanita, Boletus, Cortinarius, Russula,

Suillus), invertebrates (e.g., earthworms, slugs, snails, crustaceans, beetles, millipedes, caterpillars, and leeches), carrion, and tadpoles (see pg. 260 of Ernst and Lovich 2009 for literature citations for the above listed foods). Slugs (221 times) and other invertebrates (25), fungi (27), including the genera Russula and Lactarius, and the leaves of Jewelwort (I. capensis) (30) were salient taxa in Jones’ (2009) extensive feeding observations in Massachusetts.

Methods

Study Area

My study was set toward the southern edge of the species range in montane forests of Virginia and West Virginia. See “Study Area” in Chapter 1, Fig. 1.1,

Tables 5.3-5.5, and Appendix 1 for detailed description of the study area.

Field Procedures

I used radio-telemetry to locate turtles during the summer months when

Wood Turtles are most terrestrially active (see “Field Procedures” at Chapter 1 for general information). I compared turtle locations to randomly identified points to to identify habitat use by Wood Turtles. The same habitat features were measured, counted, or characterized on the ground at turtle locations as at paired random points. Each turtle location was paired with a random point designated in the field at a random compass orientation and random distance 23-300m from turtle points; geographic coordinates for these were generated using a Garmin ETrex GPS unit.

191 These distances represent the spatial range that could easily be within an individual’s summer seasonal activity area (ca. 0.2-10ha in size), as well as ensuring that random point plots did not overlap turtle plots. Over the four field seasons habitat features were measured, counted, or characterized (such as diameter-breast high (dbh) of canopy trees and slope aspect) at 640 plots (394 in

VA, 246 in WV), evenly distributed between adult turtle and random points.

At each turtle and random point I used nested plots (1m2 and 400m2) to capture vegetation composition and structure of differentially scaled patches

(Barbour et al. 1986, Stromberg 1995, Peet et al. 1998). In 2011-2014 at each

Turtle point, I positioned a square 1m2 plot (Daubenmire frame) using the animal’s location as the center; the same procedure was followed at random points. To assess microhabitat use, in these plots I visually estimated percent ground cover of forbs, grass, woody vegetation, coarse woody debris, moss, fungi, rock, sand, bare soil, and leaf litter. During 2013-2014 herbaceous plants (non-gramineous) and woody seedlings (those ≤ 25cm in height) were identified to the species- or genus- levels within the 1m2 plots to estimate microhabitat compositional preference.

Botanical nomenclature herein follows Weakley et al. 2012.

To capture meso-scale ecological community data, habitat features for each

Wood Turtle location (and paired random site) were measured in 400m2 circular plots with the turtle and random points at the center of each plot (11.3m radius =

400m2 = 1/25 of a hectare = ca. 1/10 of an acre). These plot dimensions can adequately capture representation of understory herbaceous species as well as

192 overstory canopy tree species (Barbour et al. 1986, Peet et al. 1998) and have been used to identify ecological communities on the GWNF (Coulling and Rawinsky

1999, Fleming and Coulling 2001). In 2011-2013 I identified non-gramineous herbaceous plants to the species- or genus-levels within these 400m2 plots. In 2011-

2014 all trees ≥ 10cm in diameter at breast height (“dbh”) were counted, identified, and measured for dbh.

Analytic Procedures

Herbaceous and Woody Seedling Taxa

I used the R package “indicspecies” (R Development Core Team 2015) to examine the herbaceous taxa and woody seedling taxa found at the plots. At both spatial scales (1m2 and 400m2), most taxa were found in a limited number of plots.

Taxa that are not somewhat common, easily found, and readily identifiable have limited pragmatic utility for management and conservation purposes. Therefore, I reduced the original presence-absence tabulations to datasets for the indicator species analyses at both the 400m2 and 1m2 scales that included only those taxa that were found in at least 10% of any of the plot types in a state. The six plot type alternatives for each state used in all the indicator value and association coefficient analyses were: female Wood Turtles = FWT, female random points = FRP, male

Wood Turtles = MWT, male random points = MRP, WT = males and females combined, RP = male and female random points combined. The significance level for reported p-values was alpha ≤0.05.

193 The “indicspecies” computational output includes an indicator value index for each individual taxon that indicates the degree of its association with site groups (e.g., the plot types identified above). Such indices are useful for assessing the predictive values of the taxa as indicators of the conditions prevailing in site groups. This index is the product of two components, denoted A and B (De Cáceres and Legendre 2009): 1) A is the probability that a site belongs to the target group

(e.g., MWT) given that the specific herbaceous taxon was found; this is called the

“specificity” of the species as an indicator. 2) B is the probability of finding the species in the sites belonging to the individual site group (i.e., the plot types); this component is called the “sensitivity” of the species as an indicator. Hence, in this framework the perfect indicator species would be one that is only found in plots of the target site group (a specificity of 1.00) and that is found in 100% of the target site group plots (a sensitivity of 1.00).

In addition, for the 1m2 plots I computed the indicator value index for combinations of two taxa (De Cáceres et al. 2012). This procedure evaluated 171 species pair combinations for the 1m2 plots in Virginia, and 210 such combinations for West Virginia. The species combinations indicator values among site groups are examined in the same way as for individual taxa (i.e., calculating A and B).

With the “indicspecies” package I also examined the association between the herbaceous species and the turtle plot types by computing Pearson’s phi coefficient of association (Chytry ́ et al. 2002). This correlation index is similar to but somewhat different from the indicator value index. It is useful for determining

194 the ecological preferences of species among a set of alternative site groups (in this case the FWT/FRP/MWT/MRP/WT/RP plot types). An advantage of the phi coefficient is that it can identify avoidance of a particular taxon through negative values. The computations corrected for the fact that groups had unequal numbers of points.

As described above for the herbaceous taxa, I also used “indicspecies” to examine the woody seedling taxa found at the 1m2 plots.

I also computed Pearson’s phi coefficient of association between the herbaceous taxa and the forest type groups I designated to the random point 400m2 plots. Using tree importance values calculated for each random plot and the forest typing system used by the US Forest Service (USFS undated), I designated a forest type for each plot (e.g., a Chestnut Oak – Scarlet Oak type or a Virginia Pine type).

I then aggregated these more finely resolved designations into six more broadly defined forest type groups: oligotrophic oak (Od), sub-mesic oak (Om), dry mixed pine and deciduous (M), pine (P), mesic deciduous (Dm), and mesic mixed pine and deciduous (Mm) (Table 5.3). These six forest type groups were used as the alternative site groups in a coefficient of association analysis.

To ascertain whether Wood Turtles were using wetland or upland habitats I tabulated the hydrophytic status of the taxa used in the analyses. A standard classification system is used for hydrophytes, which are defined as “plants growing in water or on a substrate that is at least periodically deficient in oxygen due to excessive wetness” (Tiner 2006). The presence of various types of hydrophytes is an

195 essential feature for defining wetlands (Tiner 2006). The classifications reflect the frequency of a species’ occurrence in wetlands: I) obligate (OBL): >99% of time in wetlands), 2) facultative wetland (FACW): 67 99% in wetlands, 3) facultative

(FAC): 34-66% (equally likely to occur in wetlands or non-wetlands), 4) facultative upland (FACUP): 1-33% (usually occur in non-wetlands) and 5) upland (UPL): occur in wetlands <1% of the time. Plants in classes 1, 2, and 3 in this scheme are considered to be hydrophytic; the classifications used herein for taxa were taken from Lichvar et al. (2014).

Tree Taxa, Forest Types, and Seral Stages

To examine Wood Turtles’ affinity for or association with canopy tree species I calculated “importance values” (IVs) for the taxa of trees ≥ 10cm dbh in each 400m2 plot. Trees of this size generally form the overstory and midstory canopy and are also those that are typically removed during commercial logging operations (being so-called “pole timber” (10-24cm dbh) and “saw timber” (≥25cm dbh)). Importance values can range from 0 to 100 and are a combination of two metrics: the number of trees of a given taxon and the total basal area of each taxon.

Basal area is calculated for each tree by dividing its dbh by two, squaring this quotient, and multiplying by pi (πr2), then the basal areas of individuals of a specific taxon are summed within site to give the overall basal area for that taxon. These basal areas for each taxon are compared to the total basal area of all the measured trees in the plot to give a proportionate number for the individual taxa. Similarly, the numbers of individual trees of each taxon are compared to the total number of

196 measured trees in the plot to give a proportionate abundance for the individual taxa. Finally, these two numbers (proportions of basal area and number of trees) are added together, divided by two, and multiplied by 100 to give the importance value of each taxon in the plot. For example, given that there are a total of ten trees in a plot and their total basal area is 10,000cm2 and five of these trees are Sugar

Maples with a total basal area of 3000cm2, then, the importance value for Sugar

Maple in the plot is 40: (((5/10) + (3000/ 10,000))/2) X 100 = 40.

Using the tree taxa typically dominant in the forest type groups, as well as taxonomic groupings (i.e., oaks, maples, pines) and mixtures of these groupings

(deciduous taxa, mesic taxa, and mixtures of these), I formulated fifteen models that were used in conditional logistic regressions of the importance values for tree taxa at turtle plots (used habitat) versus those at paired random plots (available habitat;

Appendix 5). The global model was the one with the smallest number of taxa that accounted for at least 90% of the mean importance value for the total plots in each state. Due to differences in species composition and sample size that resulted in some “failures to converge” for the regression process in R, the final models I ran were slightly modified between Virginia and West Virginia and between sexes.

Data for Virginia and West Virginia were analysed separately due to obvious differences in forest composition, and because of the geographic proximity of the study sites I also performed analyses with pooled data. I used the R packages

“Survival” for the regressions and “AICcmodavg” for AICc values and model averaged coefficients (R Development Core Team 2015).

197 Using the IVs for the 400m2 plots, to visually examine habitat preferences I used classification and regression trees (CART) to determine how effectively overstory composition partitioned turtle and random plots. Due to obvious differences in forest composition, data for Virginia and West Virginia were analysed separately. For the CART analyses I constructed trees for each state using pooled male and female data. In the IV datasets for tree constructions, I used only those taxa with an overall mean importance value > 0.8 in any of the four plot types

(FWT/FRP/MWT/MRP in each state): 20 taxa in Virginia and 16 in West Virginia

(see mean importance values at Table 5.2). Due to differences in species composition and sample size (leading to failures of log likelihood convergence), the final models were slightly modified between Virginia and West Virginia. The model for Virginia contained the following taxa (see Appendix 5 for acronyms):

WP+WO+RM+SM+NRO+HICK+CO+BG+VP+SYC+ELM+BW+WA+

TP+SO+BL+BLBIR+SERV+BO+IW; while the model for West Virginia contained:

WP+WO+RM+SM+NRO+HICK+CO+BG+VP+SYC+ELM+BW+WA+

BLCH+DW+CEDAR. I identified optimal tree sizes by examining cross-validation graphs that plotted change in relative error against tree size. I simplified the trees using pruning code to find the tree closest to two terminal leaves (equal to the number of categories: WT-RP) with the lowest misclassification rate. I used the R package “rpart” (R Development Core Team 2015).

198 After tree pruning I calculated a K statistic (Dellinger et al. 2007) to assess

("#$) the strength of the optimal trees relative to chance classification: K = (&#$)

A = # of actual observations correctly classified by a tree,

B = # of observations correctly classified by chance on average (number of observations divided by number of classification categories),

C = # of observations correctly classified by a perfect tree.

The values of K can be used to gauge the strength of the optimal trees (Landis and

Koch 1977): < 0 poor, 0-0.20 slight, 0.21-0.40 fair, 0.41-0.60 moderate, 0.61-0.80 substantial, 0.81-1.00 almost perfect.

To examine forest composition (forest type) and structure (seral stage) at a larger spatial scale than the 400m2 plots I used the stand inventory data supplied by the USFS. This database supplies a forest type, age, and site index for every delineated “stand” on the GWNF, stands being generally 5-20ha in size. Most paired turtle and random points were in a delineated GWNF stand; points found in private land inholdings were not used in this analysis. Using the Forest Service categorized forest types for the stands I examined turtle points and random points with G-tests, comparing frequencies of forest type groups of observed (turtle) points with those of expected (random) points; the expected proportions for the turtle points were the proportions calculated from the random point frequencies. In the same way, with the forest type characterizations for the 400m2 plots (based on the plots’ calculated importance values) I used G-tests to compare frequencies of forest

199 type groups of turtle points (the observed numbers) with those of random points

(the expected numbers); the expected proportions for the turtle points were the proportions calculated from the random point frequencies. With G-tests I also examined the turtle and random points by comparing their forest type group characterization at the 400m2 scale (the observed numbers) with their characterization at the stand scale; the expected proportions for the plots were the proportions calculated from the stand frequencies. I reasoned that if the stand level characterization was accurate for the entire stand (i.e., forest composition was homogeneously distributed), then there should be no difference between the expected and observed frequencies.

All the G-tests were performed in Excel, were two-tailed, used a Williams continuity correction, and had degrees of freedom set for an intrinsic hypothesis because I was using my data to generate expected proportions (McDonald 2014).

When the G-test results were significant, exact binomial tests of turtle and random points for specific forest type groups were done in R.

Using the USFS inventory stand ages I categorized the seral stage of stands.

Early successional habitat (“esh”) were stands aged 0-35 years, mid-successional habitat (“mid-suc.”) were those aged 36-75 years, mature stands (“mature”) were aged 75 years to the minimum age for old growth for specific forest types (“FT”).

Stands were considered “old growth” at a minimum age of 100 years for FT 33;

110 for FT 60; 120 for FTs 42 and 45; 130 for FTs 10, 52, 53, and 54; and 140 years for FTs 9, 41, 50, 56 (age figures from USDA FS 1997) (see Table 5.3 for FT

200 nomenclature and enumeration). Based on personal observation and the ages of adjacent GWNF stands, the private lands in VA were included in the mid- successional and mature data, while those in WV were included in the mature and old growth data. Using the categorized seral stages I examined turtle points and random points with G-tests, comparing frequencies of seral stages of turtle points with those of random points; the expected proportions for the turtle points were the proportions calculated from the random point frequencies.

Results

Forest Types and Seral Stages of Stands and Plots

Though only separated by ca. 20km, forest composition clearly differed between the two study sites (Tables 5.1-5.4, Figs. 5.1-5.3); see, e.g., the IVs for

VRPs and WRPs at Table 5.2. The 400m2 plots at the VA site were mostly composed of six broad forest type groups: oligotrophic oak (Od), sub-mesic oak

(Om), dry mixed pine and deciduous (M), pine (P), mesic deciduous (Dm), and mesic mixed pine and deciduous (Mm), with a small number of points located in two additional types, seeps and brushy (ruderal) habitats. Almost 90% of the total

VA turtle and random plots were composed of three forest type groups, Dm

(21.8%), Od (31.0%), and Om (37.6%). Plots at the WV site were also mostly composed of the same six broad forest types, with the addition of a small number of points in brushy and Eastern Red Cedar (Juniperus virginiana) habitats. Around

81% of the total WV turtle and random plots were composed of three forest type

201 groups, but unlike in VA the most prevalent groups were M (45.1%), P (19.1%), and Om (16.7%).

At the stand spatial scale, turtle and random points occurred in thirteen different forest types, but only one of these occurred in both states, White Oak –

Northern Red Oak – Hickory (FT 53) (Table 5.4). Stands at the VA site were composed of the same forest type groups as the plots, except M stands were absent.

Almost 90% of the total VA turtle and random points were in stands composed of the two oak forest type groups, Om (81.0%) and Od (8.6%). Stands at the WV site were composed of just three forest type groups, but unlike in VA there were no points in Od stands, instead, many points were in M and P stands: Om (39.8%), M

(35.9%), P (24.3%).

Ninety-five percent of turtle location points were within the 295m buffer zone around the VA main stream. This zone included 560ha of National Forest; these stands comprised 379.6ha (67.8%) of Om, 89.4ha (16.0%) of Od, 26.0ha

(5.1%) of M, 24.6ha (4.4%) of Mm, 20.2ha (3.6%) of P, 4.5ha (0.8%) of Dm, 7.1ha

(1.3%) of brushy ruderal, and 6.1ha (1.1%) were not given a forest type designation. There were also 26.4ha of private lands without stand data (Fig. 5.4).

Ninety-five percent of turtle location points were within a 290m buffer zone around the WV main stream. This zone included 148ha of National Forest; these stands comprised 70.7ha (47.9%) of M, 45.3ha (30.7%) of P, 31.7ha (21.4%) of

Om, and 0.07ha (0.04%) of Od. There were 71.5ha of private lands without stand data (Fig. 5.5).

202 G-tests detected a difference between the frequency of turtle points and random points for forest type groups characterized at the 400m2-plot scale (Fig. 5.2) in both WV (G = 32.06, df = 4, p < 0.0001) and VA (G = 111.13, df = 4, p <

0.0001).

A G-test detected no difference between the frequency of turtle points and random points for forest type groups characterized at the stand level in WV (G =

1.42, df = 1, p = 0.233) (Fig. 5.3). In VA, however, there was a difference between the frequency of turtle points and random points for forest type groups characterized at the stand level (G = 41.18, df = 3, p < 0.0001) (Fig. 5.3). Exact binomial tests found the differences to lie with the Mm (p < 0.00001) and Od (p <

0.00001) forest type groups.

G-tests detected a difference between the frequency of points for forest type groups characterized at the 400m2-plot and stand scales in both WV (G = 361.83, df = 5, p < 0.0001) and VA (G = 596.55, df = 5, p < 0.0001) (Fig. 5.6).

Most of the 2011-2014 plots wherein I calculated importance values were in older forest (mature and old growth seral stages) (Table 5.5, Figs. 5.7 & 5.8). In VA,

83.7% of turtle points (males and females pooled) were in stands of older forest, while 70.6% of random points were. One hundred percent of turtle points and random points were in stands of older forest at the WV site, which had no esh or mid-successional stands. In the VA buffer zone, the GWNF stands comprosed

89.4ha (16.0%) of esh, 95.8ha (17.1%) of mid-successional, 247.4ha (44.2%) of mature, and 127.2ha (22.7%) of old growth (Fig. 5.8). GWNF stands in the WV

203 buffer zone comprised 109.5ha (74.1%) of mature and 38.3ha (25.9%) of old growth.

A G-test detected a difference between the frequency of turtle points and random points for seral stages characterized at the stand scale in VA (G = 21.91, df

= 2, p < 0.0001) (Fig. 5.7). Turtles used esh (exact binomial test, p = 0.0120) and mid-successional (p = 0.0053) less than was available, while using OG (p =

0.0042) more than was available at random; there was no difference of frequency between turtle and random points for the mature seral stage (p = 0.0960). Because all the sites in WV were in stands of older forest, I did not test for differences in frequency for seral stages there.

Overstory Trees

Importance values were calculated from 7098 trees ≥10cm dbh at 396 plots

(144 female plots and 54 male plots and their paired RP plots) in Virginia and from

5024 trees ≥10cm dbh at 246 plots (74 female plots and 49 male plots and their paired RP plots) in West Virginia. I performed separate analyses for the VA and WV sites because of clear differences between the states in the proportionate composition of overstory trees.

I identified 41 overstory tree taxa; importance values (IVs) were calculated for 39 taxa in Virginia and 31 taxa in West Virginia. In the regression model formulations, I used only those taxa with an overall mean importance value ≥ 0.8 in regression models for any of the four plot types (FWT/FRP/MWT/MRP): 18 taxa in Virginia and 18 in West Virginia. See Tables 5.1-5.3 for tree taxa used in

204 modeling, definitions of acronyms, and importance values. For trees with a dbh ≥

10cm in 400m2 plots in 2011-2014 there was no difference in taxa richness between turtle points and paired random points in either state (Tables 5.6 & 5.7).

Paired Logistic Regression

There was some overlap in the tree taxa that best explain habitat preferences for Virginia females and males (Tables 5.9 & 5.10). Both sexes showed a positive affinity for sites with relatively higher values for Sugar Maple (Acer saccharum) and

Serviceberry (Amelanchier spp.), while both sexes showed a tendency to avoid sites with higher importance values for Chestnut Oak (Quercus montana) and Scarlet

Oak (Q. coccinea). Virginia females showed a positive affinity for sites with higher values for White Oak (Q. alba) and White Ash (Fraxinus americana), and a tendency to avoid sites with high values for Black Oak (Q. velutina). In contrast,

Virginia males showed a tendency to avoid sites with higher importance values for

White Oak, Red Maple (A. rubra), White Pine (Pinus strobus), and Ironwood (O. virginiana).

Both West Virginia females and males tended to avoid sites with higher importance values for Chestnut Oak (Tables 5.9 & 5.10). West Virginia males also tended to avoid sites with higher importance values for Sugar Maple and White

Pine. West Virginia females showed a positive affinity for sites with higher values for Red Maple, Hickories (Carya spp.), and Sycamore (Platanus occidentalis).

The top two conditional regression models for explaining habitat selection by Virginia females were the mesic deciduous (Dm) and sub-mesic oak (Om)

205 models, with eleven and nine taxa respectively (Table 5.8). The dry mixed pine and deciduous (M) and oaks’ taxa models were the top two models for Virginia males, with seven and five taxa respectively. The top two models for West Virginia females were the global (with ten taxa) and mesic species (with six taxa) models.

The global (with ten taxa) and dry mixed pine and deciduous (M, with eight taxa) models were the top two models for West Virginia males.

There was limited commonality in the tree taxa that best explain habitat preferences for pooled females or pooled males from both states (Table 5.12). Both sexes showed a tendency to avoid sites with higher importance values for Chestnut

Oak and Scarlet Oak. In addition, some taxa were preferred by one sex but avoided by the other, such as with Sugar Maple and Elm. Females showed a tendency to avoid sites with higher importance values for Ironwood and Elm (Ulmus spp.), while males exhibited a tendency to avoid sites with higher importance values for

Sugar Maple and seven other taxa. Females showed a positive affinity for sites with higher values for White Oak, Sugar Maple, White Ash, and Serviceberry. In contrast, males showed a preference for sites with higher importance values for Elm and Basswood (Tilia americana).

The only well-supported model for explaining habitat selection by pooled

Virginia and West Virginia females was the mesic deciduous (Dm) model, with ten taxa (Table 5.11). The oaks model (with five taxa) was the best model for pooled

Virginia and West Virginia males; the global model (with twelve taxa) was also well supported. The only well-supported model for explaining habitat selection by all

206 Virginia and West Virginia Wood Turtles combined was the global model (Table

5.11).

When all Wood Turtles of both sexes from both states were examined together, only White Ash and Serviceberry were strongly preferred (Table 5.12).

The models indicated turtles tended to avoid sites with higher importance values for White Pine, Chestnut Oak, Scarlet Oak, and Northern Red Oak (Q. rubra).

CART

The CART results suggest VA Wood Turtles preferred sites with relatively higher importance values for White Ash, Sugar and Red Maples, and White Oak and relatively lower values for Chestnut, Scarlet, and Black Oaks. For WV Wood

Turtles the results indicate a preference for sites with relatively higher importance values for Red Maple, Sycamore, and Hickories and lower values for Chestnut Oak and White Pine. These CART results are generally congruent with the conditional logistic regression results.

Virginia. For the Virginia CART, using pooled female and male turtles and their random points, when all three of the initial trees were pruned, only Chestnut

Oak and Scarlet Oak were used in tree construction (Fig. 5.9). The pruned trees resulted in 73.4% of points correctly classified by the IV threshold values for CO and SO; K = 0.47. Twelve taxa were used in tree construction without priors using the “information” split: all five Oaks (Black, Chestnut, Northern Red, Scarlet, and

White), White and Virginia Pine, Red and Sugar Maple, White Ash, Hickories, and

Ironwood. Using the “information” split with priors, the taxa used in tree

207 construction were three Oaks (Chestnut, Scarlet, and White), Red and Sugar Maple,

White Ash, and Black Birch (Betula lenta). When trees were constructed using a

“gini” split with priors, the only taxa used were Chestnut and Scarlet Oaks and Red and Sugar Maples.

Because a substantial number of the VA turtle points had a component of

Chestnut Oak (63 out of 197), I ran further CART analyses on this subset of data

(i.e., turtle plots with CO and their paired random points) to see which taxa partitioned turtle and random points. The model used did not include Chestnut or

Scarlet Oaks or taxa with a mean IV less than 1 (see Appendix 5 for acronyms):

WP+WO+RM+SM+NRO+HICK+BG+TP+SERV+BO+WA. The taxa used in tree construction using the “information” split without priors and the “gini” split with priors were the same: Sugar Maple, White, Northern Red and Black Oaks, and

Hickories. When both of these trees were pruned, only Sugar Maple, White Oak, and Hickories were used in tree construction (Fig. 5.10). The pruned trees resulted in 65.3% of points correctly classified by the IV threshold values; K = 0.31.

Using this same data subset I also constructed trees with models that included Scarlet Oak as a variable, but not Chestnut Oak. The taxa used in tree construction using the “information” split without priors and the “gini” split with priors were the same: Sugar Maple, four Oaks (White, Scarlet, Northern Red and

Black), and Hickories. When the “information” split tree was pruned, Sugar Maple,

White Oak, and Hickories were used in tree construction, while the pruned “gini” tree used Scarlet Oak in addition to those three taxa (Fig. 5.11). The pruned “gini”

208 tree resulted in 69.4% of points correctly classified by the IV threshold values; K =

0.39.

West Virginia. For the West Virginia CARTs, using pooled female and male turtles and their paired random points, the taxa actually used in tree construction were the same using the “information” split without priors, the “information” split with priors, and the “gini” split with priors: Red Maple, Chestnut and White Oak,

Sycamore, Hickories, and White Pine. When these three trees were pruned, only

Red Maple, Hickories, and Sycamore were used in tree construction (Fig. 5.12).

The pruned trees resulted in 71.1% of points correctly classified by the IV threshold values for Red Maple, Hickories, and Sycamore; K = 0.42.

Though Sycamore was consistently used in tree construction, most of the turtle points did not have a component of Sycamore (107 out of 123). Therefore, I ran further CART analyses on this subset of data (viz., the turtle points without

Sycamore and their paired random points) to see which taxa classified turtle and random points. The model used did not include Sycamore (see Appendix 5 for acronyms): WP+WO+RM+SM+VP+NRO+HICK+CO+BLCH+ELM+BG+BW+WA.

When all three of the initial trees were pruned, only Chestnut Oak and White Pine were used in tree construction (Fig. 5.13).

The pruned information tree resulted in 68.2% of points (K = 0.36), and the pruned gini tree 69.2% of points (K = 0.38), correctly classified by the IV threshold values for CO and White Pine. These results indicate that Wood Turtles at WV sites without a component of Sycamore preferred sites with relatively lower values for

209 Chestnut Oak and White Pine. The taxa used in tree construction using the

“information” split without priors were Red Maple, White, Northern Red and

Chestnut Oaks, Hickories, and Virginia and White Pines. The taxa used in tree construction using the “information” split with priors were the same, with the addition of Black Cherry. The taxa used in tree construction using the “gini” split with priors were Red Maple, White and Chestnut Oaks, and Virginia and White

Pines. The initial trees indicated that Wood Turtles at WV sites without a component of Sycamore preferred sites with relatively higher importance values for

Red Maple and Hickories.

Herbaceous Flora

400m2 plots

I recorded 3596 presences of 128 native taxa at 311 plots in VA (with an additional 142 presences of 7 alien species and 214 presences of unknowns) and

1523 presences of 88 native taxa at 159 plots in WV (with an additional 99 presences of 4 alien species and 202 presences of unknowns) in the 400m2 plots in 2011-2013.

The reduced herbaceous dataset for VA included 40 taxa, as did that for WV

(see Table 5.13); of the 53 total taxa 28 were common to both states. Due to these differences in composition, I performed separate analyses for the VA and WV sites.

Thirty-eight of the 53 taxa had significant indicator value (p-values ≤ 0.05) for at least one group (Tables 5.13 & 5.14) and 37 taxa had significant phi coefficients of association for at least one group (Tables 5.15 & 5.16); thirty-four taxa were useful

210 in both analyses. Virginia had a total of 28 taxa useful as indicators, while WV had

22.

Thirty taxa were indicators for the presence of Wood Turtles (either the

FWT, MWT, or FWT+MWT groups) in at least one state (Tables 5.13 & 5.14). Of these 30 taxa, eight were useful in both VA and WV as indicators for pooled males and females: Amphicarpaea bracteata (Hog Peanut), Circaea lutetiana (Enchanter’s

Nightshade), Eurybia divaricata (White Wood Aster), Galium triflorum (Bedstraw),

Oxalis spp. (Wood Sorrel), Potentilla spp. (Cinquefoils), Viola spp. (Violets), and the alien Microstigeum vimineum (Stiltgrass); with E. divaricata (White Wood Aster) and Oxalis spp. (Wood Sorrel) having marginally significant p-values in WV (Table

5.14). Three of the 30 WT taxa were ferns: Christmas Fern (Polystichum arostichoides) in WV and Sensitive (Onoclea sensibilis) and New York

(Parathelypteris noveboracensis) Ferns in VA. Twenty-seven taxa had significant phi coefficients of association for Wood Turtles (Tables 5.5 & 5.16). Most of these were useful in the indicator species analyses, with the addition of White Snake Root

(Ageratina altissima) and Wild Geranium (Geranium maculatum).

Nine taxa (p <0.051) were indicators for random points of one group or another (FRP, MRP, or FRP+MRP; Tables 5.13 & 5.14), suggesting turtles tend to avoid these taxa (see “Study rationale” in Chapter 6). One taxon was a fern, Ebony

Spleenwort (Asplenium platyneuron). Many of the nine taxa, such as Trailing

Arbutus (Epigaea repens), Tick Trefoil (Desmodium spp.), and Round-lobed

Hepatica (Anemone americana), also had significant phi coefficients of association

211 for random points (Tables 5.15 & 5.16). In addition, Panicled Hawkwort

(Hieracium paniculatum) and Goldenrods (Solidago spp.) had significant phi values, but not indicator species values for random points.

One taxon was useful in both VA and WV as an indicator for pooled male and female random plots: Chimaphila maculata (Spotted Wintergreen). Dittany

(Cunila origanoides) had marginally significant phi values for FRPs in both states

(Table 5.16). Two species, Uvularia perfoliata (Perfoliate Bellwort) and the alien

Perilla frutescens (Beefsteak Plant), were indicators for pooled males and females in

WV, but indicators for pooled male and female random plots in VA. Stiltgrass was by far the most common alien herbaceous species, present in 172 plots while the other seven alien taxa pooled occurred in 74 plots.

Some taxa were indicators for turtles at one scale, but not the other; e.g.,

Arisaema triphyllum (Jack-in-the-pulpit), Mitchella repens (Partridgeberry), and

Wood Sorrel at the 400m2 but not the 1m2 scale. Pussytoes (Antennaria spp.) indicated male random points at the 1m2 but not the 400m2 scale. Some common taxa were useful indicators individually at neither scale; e.g., Dioscorea villosa

(Wild Yam), Galium circaezans (Wild Licorice), and Packera obovata (Round- leaved Ragwort). Several taxa presented somewhat contradictory results in that they were indicators for turtles at one scale, but random points at the other scale, with interstate variation as well for the first two: viz., Gaultheria procumbens (Teaberry) for WWT 400m2 – VFR 1m2, Parthenocissus cinquefolia (Virginia Creeper) for VMR

212 400m2 phi – WWT 1m2, and Hieracium venosum (Hawkwort) for WRP 400m2 –

WMT 1m2.

Thirteen of the 30 taxa (43%) that were indicators for turtles (males, females, or M+F) in the 400m2 plots can be considered hydrophytic (nine of the thirteen were FACW or OBL). Fifteen (29%) of the 53 common plant taxa used in the indicator analyses for the 400m2 plots were wetland plants (i.e., hydrophytes)

(Table 5.13).

In both states in 2011-2013, 400m2 turtle plots had greater herbaceous taxa richness than did random plots, except for WV males (Tables 5.6 & 5.7, Fig. 5.14).

Using pooled males and females: VA paired Wilcoxon signed rank test: V = 8298, p < 0.00001; WV paired t-test: t = 4.2659, df = 78, p < 0.0001.

Virginia 400m2. Three groups possessed taxa with significant indicator values: one taxon for the MWT, twenty-one for FWT-MWT, and four for the FRP-

MRP group. One additional taxon had a p-value ≤ 0.07 (Table 5.14). Five of the six site groups possessed taxa with significant phi association coefficients, only the

MRP group did not. Except for the FWT+MWT group, for which 18 taxa were significantly associated, the other four groups only had one or two associated taxa per group (Table 5.16). For all site groups there was also one more taxon with a p- value < 0.09.

West Virginia 400m2. There were significant indicators for four groups; one taxon for the FRP group, three for the MWT, thirteen for FWT-MWT, and four for the FRP-MRP. The taxon for FRP was Desmodium spp. (Tick Trefoil), while those

213 for the MWT group were Lycopus spp. (Buglewort), Impatiens capensis (Jewelwort), and Scuttellaria spp. (Skullcap). There were also three more taxa with p-values ≤

0.1 in the FWT-MWT and FRP-MRP groups (Table 5.14).

All six groups possessed taxa with significant phi coefficients; two taxa for the FWT group, one for the FRP, three for the MWT, two for the MRP, seven for the

FWT-MWT, and two for the FRP-MRP. There were also four more taxa with p- values ≤ 0.1 in the FWT-MWT and FRP-MRP groups (Table 5.16).

1m2 plots

I recorded 510 presences of 65 native herbaceous taxa at 246 plots (with an additional 42 presences of 1 alien species and 13 presences of unknowns) in VA and

390 presences of 59 native herbaceous taxa at 152 plots (with an additional 29 presences of 2 alien species and 28 presences of unknowns) in WV in 2013-2014.

The reduced herbaceous dataset for VA included 18 taxa, while that for WV had 20 (Table 5.17). Although the quantities of taxa were similar for each state, only

10 of the 28 total taxa were common to both states. None of the taxa were ferns.

Fourteen of the 28 herbaceous taxa had significant indicator value (p-values ≤ 0.05) for at least one group and twelve of these taxa also had significant phi coefficients of association for at least one group (Tables 5.17 & 5.18).

Eleven taxa had significant indicator value (p-values ≤ 0.05) for at least one turtle group (FWT, MWT, or FWT+MWT) (Table 5.17). Six taxa were useful for turtle groups in both VA and WV: Cinquefoil, Hog Peanut, Bedstraw (G. triflorum),

Jewelwort, Violets, and the alien Stiltgrass. The same eleven taxa, except for Blue-

214 stemmed Goldenrod (Solidago caesia) which had a marginally significant value, had significant phi coefficients of association for turtles (Table 5.18). The eleven taxa useful at the 1m2 scale were also useful at the 400m2 scale, except for the

Blue-stemmed Goldenrod (Solidago was not identified to the species level in the

400m2 plots) and the two taxa mentioned above with contradictory results between scales (Hawkwort and Virginia Creeper).

Three taxa (p ≤0.05) were indicators for random points of one group or another (FRP or FRP+MRP; Table 5.17) in Virginia. No taxa were indicators for random points in West Virginia, except for Pussytoes, of marginal significance for

MRP. By inference, turtles tend to avoid these taxa (see “Study rationale” in

Chapter 6). Of the three indicator taxa, Teaberry and Bellworts (Uvularia spp.) had significant phi coefficients of association for random points (Table 5.18), but

Dittany did not; Pussytoes had marginal significance for MRP.

Three of the 11 taxa (27%) that were indicators for turtles were hydrophytic.

Five (18%) of the 28 common plant taxa used in the indicator analyses for the 1m2 plots were wetland plants (i.e., hydrophytes) (Table 5.17).

In both states in 2013-2014, except for WV females 1m2 turtle plots had greater herbaceous taxa richness than did random plots (Tables 5.6 & 5.7, Fig.

5.15).

Virginia 1m2. There were significant indicators for three site groups; one taxon for the FRP group, four for the MWT, and seven for the FWT-MWT (Table

5.17). Eleven of these 12 taxa also had significant phi coefficients for the same

215 groups; except Potentilla was in the VMT group instead of VWT, and one more taxon (Solidago caesia) had a p-value = 0.086 (Table 5.18).

Three groups had significant indicator indices for species pairs. Four taxa were involved for the FWT group, 14 taxa for the MWT group, and 11 taxa for the

FWT-MWT group. All of the pairs involved taxa that were significant indicators individually, with the addition of P. quinquefolia, Viola spp., A. bracteata, S. caesia,

Galium pilosum (Bedstraw), and Oxalis spp. for MWT, M. vimineum, A. altissima,

G. triflorum, and Solidago spp. for FWT-MWT, and I. capensis, P. quinquefolia,

Viola spp., and D. villosa for FWT. Out of 171 examined combinations, 51 had significant p-values.

West Virginia 1m2. The only group for which there were significant indicators was the FWT-MWT combination; the taxa involved were Stiltgrass,

Violets, and Bedstraw (G. triflorum). There were also four more taxa with p-values

< 0.09 in the MWT, MRP, and FWT-MWT groups (Table 5.17). Only two groups had significant phi coefficients; the associated taxa were Potentilla spp. for the

MWT group, and M. vimineum for the FWT-MWT combination group. There were also seven more taxa with p-values ≤ 0.10 in the MWT, MRP, and FWT-MWT groups (Table 5.18).

Three groups had significant indicator indices for species pairs. The taxa involved for the MRP group were Potentilla spp. and Antennaria spp. The taxa involved for the MWT group were Galium circaezans (Wild Licorice), M. vimineum, Viola spp., G. triflorum, Potentilla spp., and P. quinquefolia. The taxa

216 involved for the FWT-MWT group were the same three that were significant indicators individually. Out of 210 examined combinations, only eleven had significant p-values.

Woody Seedlings

I recorded 686 presences of 19 native woody seedling taxa in VA (with an additional 2 presences of unknowns) and 230 presences of 10 native seedling taxa

(with an additional 4 presences of unknowns) in WV at the 1m2 plots in 2013-2014.

The reduced woody seedling dataset for VA included 15 taxa, while that for

WV had 10 (Table 5.19). Eight of the 17 total taxa were common to both states.

Overall, in the 1m2 plots ten taxa had significant indicator value (p-values ≤ 0.05) for at least one group and eight of these ten taxa also had significant coefficients of association for at least one group (Tables 5.19 & 5.20).

The same six taxa (four in VA, two in WV) had significant indicator values as well as significant phi coefficient of association values (p-values ≤ 0.05) for at least one turtle group (FWT, MWT, or FWT+MWT); one was useful in both VA and WV,

Rubus spp. (blackberries) (Table 5.19). Three taxa had significant indicator value for female + male random point groups; two of these were useful in both VA and

WV, Chestnut Oak and Vaccinium spp. (blueberries). The same three taxa had significant phi coefficient of association values for at least one random point group

(FRP or FRP+MRP); only Chestnut Oak was useful in both VA and WV (Table 5.20).

217 Three of the 8 taxa (37%) that were indicators for turtles were hydrophytic.

Five (29%) of the 17 common woody seedling taxa used in the indicator analyses can be considered hydrophytes (they were all “facultative”) (Table 5.19).

There was no difference in woody seedling taxa richness between turtle and random points in either state in 2013-2014 (Tables 5.6 & 5.7).

Virginia 1m2

There were significant indicators for four site groups; one taxon each for the

FWT and MWT groups, two for the FWT-MWT, and four for the FRP-MRP (Table

5.19). The sole taxon for FWT plots was Lindera benzoin (Spicebush) while that for

MWT plots was Ostraya virginiana (Ironwood). Of significant indicator value for the

FWT-MWT group were blackberries and Smilax spp. (Greenbrier). Significant indicators for the FRP-MRP sites were Red Maple, Chestnut Oak, Northern Red

Oak, and blueberries. As was the case for the indicator value index, four groups had significant phi coefficients: one taxon each for the FWT, MWT, and FWT-MWT groups, and three for the FRP-MRP (Table 5.20). There was also one taxon (Rubus spp.) with a p-value = 0.08.

West Virginia 1m2

There were significant indicators for only the MWT and FRP-MRP groups; taxa for MWT plots were Amelanchier spp. (Serviceberry) and Rubus spp., while those for FRP-MRP plots were Chestnut Oak and blueberries (Table 5.19). Two groups had significant phi coefficients; the taxa for MWT plots were Amelanchier

218 spp. and Rubus spp., while the significant taxon for FRP plots was Q. montana

(Table 5.20). There was also one taxon (O. virginiana) with a p-value = 0.105.

Herbaceous Richness, Forest Types, Seral Stages, Herbaceous Cover

The herbaceous taxa association analyses for random point 400m2 plots categorized by forest type groups in each state found few taxa to be indicators for specific forest type groups; of these taxa most were associated with the Mm (6 taxa) or Dm (5 taxa) groups, with one taxon (Hieraceum paniculatum) having a significant Pearson’s phi coefficient of association for the M group (Table 5.21).

There were no indicator taxa for the Od, Om, or P forest type groups.

Most of the 2011-2013 400m2 plots used in the herbaceous indicator species analyses were in older forest (mature and old growth seral stages), 89.5% of aggregated (both states) turtle points and 77.4% of aggregated random points (Table

5.5). At the WV site, which had no esh or mid-successional stands, 82.3% of turtle points were in mature stands and 17.7% were in old growth, whereas 77.2% of random points were in mature stands and 22.8% were in old growth. In VA, 84.0% of turtle points (males and females pooled) were in stands of older forest, while

65.6% of pooled random points were.

The number of herbaceous taxa in VA random point 400m2 plots did not

differ between seral stages (Fig. 5.16) (ANOVA: F(3,148) = 0.594, p = 0.620), but the number did differ between forest type groups (Fig. 5.17) (Mm was not used as the

sample size was only two): (ANOVA: F(3,146) = 10.965, p < 0.00001). Mean number of taxa: Dm = 14.1 ± 1.7, M = 6.1 ± 1.7, Mm = 35 ± 0, Od = 7.2 ± 0.6, and Om =

219 12.1 ± 0.8. TukeyHSD tests (with alpha = 0.0083) found the differences to lie between Dm-M (p = 0.0075), Dm-Od (p = 0.00025), and Od-Om (p = 0.00001).

In contrast, in WV the number of herbaceous taxa in random point 400m2 plots did not differ between forest type groups (Fig. 5.17) (Od and Dm were not

included as the sample size was only three for each): (ANOVA: F(3,69) = 2.485, p =

0.0679); mean number of taxa: M = 11.9 ± 1.2, Mm = 18.2 ± 3.9, Om = 15.6 ±

1.5, P = 10.8 ± 1.7, Dm = 18.0 ± 2.5, Od = 8.0 ± 2.1.

There was some consistency of pattern among both states (Fig. 5.17). In VA random plots the Od and M forest type groups had the lowest number of herbaceous taxa, while in WV the Od and P plots had the lowest (with M third from the bottom). In both states random plots in Dm or Mm forest types had the highest number of taxa, with Om plots occupying the mid-range of richness. With this in mind, I ran correlation tests and regressions between number of herb taxa in random plots and the importance values of some tree taxa typically associated with the different forest type groups.

For Spearman correlation tests between number of herb taxa and importance values in VA I used Chestnut Oak, Scarlet Oak, and White Ash.

Number of herb taxa in a plot and the tree importance value were negatively correlated for SO (S = 767469, p < 0.0001, rho = -0.364) and less so for CO (S =

674400, p = 0.0146, rho = -0.199), while WA was positively correlated (S =

320911, p < 0.00001, rho = 0.429) (Fig. 5.18). However, the regressions using

220 these taxa all had low R2 values (ca. 0.1), indicating that the importance value of a single tree species did not predict much of the variation in herbal richness at sites.

For the WV data I used Chestnut Oak, Virginia Pine, and Sugar Maple.

Number of herb taxa in a plot and tree importance value were negatively correlated for both VP (S = 1008113, p < 0.00001, rho = -0.371) and CO (S =

963726, p < 0.0001, rho = -0.311), while SM was positively correlated (S =

419342, p < 0.00001, rho = 0.430) (Fig. 5.19). As in VA, the regressions using these taxa all had low R2 values (ca. 0.1).

In both states in 2011-2014, 1m2 turtle plots had significantly more herbaceous cover (forbs and grass combined) than did random plots (Tables 5.6 &

5.7, Fig. 5.20). Amount of herbaceous ground cover had a weak negative correlation with distance from the main streams (S = 595000, p < 0.0001, rho = -

0.362).

Discussion

I evaluated relationships between floristic composition and structure and

Wood Turtle habitat use. The underlying question for a study such as this is: What life-history requirements are met by the use of a habitat (Beyer et al. 2010)? Prudent choices are necessary in order to obtain adequate energy, find refuge from predators, and avoid environmental extremes; these choices and activities may increase or decrease the use of available habitats (Halstead et al. 2009, Willems and Hill 2009). Wood Turtles preferred or avoided specific herbaceous and woody taxa and forest types, but these sometimes differed between the states. Both

221 structural and compositional characteristics of forest ground floor habitat may directly affect the turtles’ foraging success, vulnerability to or avoidance of predators, and osmo- and thermo-regulatory options; overstory canopy structure and composition may directly affect or indicate these potentialities as well.

Herbaceous and Seedling Taxa

Over thirty herbaceous and woody taxa were indicators for Wood Turtles at the 400m2 and/or 1m2 scales at these VA and WV study sites (Tables 5.13-5.20), including most of the taxa referenced above (in “Focal Species”). The herbaceous indicator taxa discerned by this study, as well as the overstory tree taxa of note, are found throughout the Wood Turtle’s range or at least a large portion of it (Fernald

1950, Newcomb 1977, Burns and Honkala 1990). Localized edaphic, topographic, and moisture conditions determine the presence of these flora at finer spatial scales within their broad-scale distributions (Braun 1950, Cantlon 1953).

There was solid concordance between the results of the indicator species analyses and those of the phi coefficient of association analyses (Tables 5.13-5.18).

With few exceptions, the same taxa were significant in each methodology, a congruence that militates for the accuracy and usefulness of the results. There was also general concordance between the results of the analyses at the 1m2 and 400m2 scales. There was limited congruence, however, between the states as to the taxa that indicated preference by Wood Turtles (Tables 5.13 & 5.17). Only eight indicator taxa were common to turtles in both VA and WV at the 400m2 or 1m2 scales. This evidence suggests habitats composed of different forest types be

222 evaluated separately, even when they are in close geographic proximity.

Notwithstanding differences in composition, in both states Wood Turtle plots at both spatial scales had greater herbaceous taxa richness than did random plots

(Figs. 5.14 & 5.15), which also was the case at a WV river site (McCoard et al.

2016b). Another salient result involved herbaceous cover which was clearly greater at turtle points than random points (Fig. 5.20). It is also noteworthy that the great majority of taxa and presences at my forest sites were of native species.

Another pattern is the relative paucity of wetland species both in the overall herbaceous species lists as well as those that serve as indicators. Though the Wood

Turtle is often characterized as a riparian species or denizen of wet areas (see, e.g.,

McCoard et al. 2016b or USDA FS $$$$), they clearly use dry uplands a great deal.

Most of the indictors were upland or facultative upland taxa (Tables 5.13, 5.17, 5.19).

Many of the species with which the Wood Turtles were commonly associated are found in sites of intermediate soil moisture, while some, such as Ageratina altissima,

C. oreganoides, D. villosa, E. divaricata, G. triflorum, Maianthemum racemosa, M. repens, P. obovata, Polygonatum biflorum, and Potentilla spp. can be regularly found in dry settings (Hutchinson et al. 1999, Weakley et al. 2012). Because many of the species used in the indicator analyses have somewhat broad habitat niches (at least when within relatively undisturbed forests), it is not surprising that only a limited number were useful as indicators for specific forest type groups (Table 5.21). In

Beech-Maple and Maple-Basswood forests in Michigan and Minnesota, Rogers

(1981) similarly found broad overlap in herbaceous composition in stands of different

223 forest types.

Only one species was an indicator for random points for both states at the

400m2 scale, Spotted Wintergreen (C. maculata). In addition, one alien species,

Beefsteak Plant, served as an indicator for turtle sites in WV, but for random sites in

VA. McCoard and colleagues (2016b) elsewhere in WV also found Bedstraw

(Gallium spp.) to indicate Wood Turtle sites. One taxon that served as an indicator for female Wood Turtle sites in VA, Boehmeria cylindrical (False Nettle), was also preferred by Eastern Box Turtles (Terrapene carolina) at a coastal plain site in

Maryland (McKnight 2011).

As with the herbs, most of the woody seedling taxa found in turtle plots were facultative upland or upland taxa; only three of the seedling indicator taxa were classified as wetland species (Table 5.19). This is further evidence that Wood

Turtles regularly use habitats far outside of riparian or wet areas. Though Spicebush

(useful for VA females) prefers moister site conditions (Weakley et al. 2012), most of the seedling taxa useful as indicators for Wood Turtles (Ironwood, Greenbriar,

Serviceberry, and Blackberry) are found in various types of forested settings

(Hutchinson et al. 1999, Burns and Honkala 1990, Weakley et al. 2012); further indication that Wood Turtles commonly use different types of forest. Turtles may have been feeding on the leaves of seedlings (such as Greenbrier) or on the fruits of species found in the shrub or tree layer (e.g., Serviceberry or Spicebush) that had seedlings in the understory.

224 Overstory Tree Composition and Structure

The results of the conditional logistic regressions, CARTs, and G-tests were concordant; the same species and species groups consistently showed up as important drivers of habitat preference. Although some general tendencies are apparent, particularly the avoidance of sites with a high component of Chestnut

Oak, the somewhat low discriminatory power exhibited by the some of the CARTs

(only “fair” K values), the low number of regression coefficients that did not overlap zero, and the fact that most tree taxa were not indicative of habitat partitioning or preference, all signify that Wood Turtles use sites with many different tree taxa and proportions of tree taxa (i.e., different forest types). The indicator species of import

(e.g., White and Chestnut Oaks, Red and Sugar Maples, White Pine) as well as the other taxa are broadly distributed in eastern North America and most are found throughout or in large portions of the Wood Turtle’s range (see maps in Burns and

Honkala 1990 and Turtle Taxonomy Working Group 2017).

In the United States Wood Turtle distribution occurs in the following broad- scale forest type regions as defined by Braun (1950) and Dyer (2006):

Mesophytic Region (VA, MD, DE),

Appalachian Oak Section of the Mesophytic Region (VA, WV, MD,

PA, NJ, NY, CN, RI, MA, NH, ME),

Beech – Maple – Basswood Region (PA, NY, WS, MN, IA),

Northern Hardwoods – Hemlock Region (PA, NY, CN, MA, VT, NH, ME),

Northern Hardwoods – Red Pine Region (MI, WS, MN).

225 As with the herbaceous flora, there is a great deal of overlap of tree species among the different large-scale eco-regions where the Wood Turtle occurs (Ashe 1922,

Braun 1950, Newcomb 1977, McNab and Avers 1994, Woods et al. 1999, Burns and Honkala 1990, Dyer 2006, McNab et al. 2007). Generally, throughout the turtle’s range a relatively large number of tree species (16-35) predominate at any particular location (Dyer 2006, Rentch 2006); this study area comports with this pattern (Table 5.2).

Just as at the broadest scale (i.e., the regional species distribution level),

Wood Turtles at the local scale use a range of different forest types. Overall, at least ten forest types were used at this study area (Table 5.4), though some were used more than others, with the oligotrophic oak sites (Od) in particular being used less than expected based upon their availability (Figs. 5.1 & 5.2). In concordance with this finding, the analyses of individual taxa indicated Wood Turtles tended to avoid sites with high importance values for Chestnut and Scarlet Oaks (Table 5.9, Figs.

5.9 & 5.13). Domination by these taxa is generally indicative of nutrient poor sites

(oligotrophic) (Burns and Honkala 1990, Fleming and Coulling 2001, Weakley et al. 2012). At plots with a component of Chestnut Oak, the CART results indicated

VA Wood Turtles preferred sites with relatively higher importance values for Sugar

Maple and White Oak and relatively lower values for Hickories (Fig. 5.10).

The preference for Sugar Maple in VA and Red Maple in WV may have to do with site productivity (higher nutrient availability) and/or moisture regimes. The forests are different in each state, with Red Maple, a generalist, typically found in

226 more mesophytic associations at the WV site, while in VA it is more widespread, being abundant also in drier oligotrophic sites as well as in tracts of both older

(mature and old growth) and younger age (early and mid-successional seres). The preference for sites with high importance values for Virginia Pine by WV females may be due in part to the high amounts of grass cover often found in these tracts

(cover of 10.7% ± 3.4 in 1m2 plots at random point sites with a VP IV > 25 vs.

3.7% ± 0.9 at RP sites with a VP IV < 25; mean grass cover in 1m2 plots at WV female turtle points was 9.9% ± 1.8; Fig. 5.21 illustrates an example). Locations with high importance values for White Oak, White Ash, Sugar Maple, Elm,

Basswood, and Serviceberry can be mesic or sub-mesic sites that are generally productive (Burns and Honkala 1990, Fleming and Coulling 2001, Weakley et al.

2012) and have high herbaceous species richness (Fig. 5.17). In contrast, the turtles’ tendency to avoid sites with high importance values for White Pine in WV

(Fig. 5.7) may be reflective of types of forest, pine or mixed pine/deciduous, with low productivity and low amounts of grass cover (mean cover in 1m2 plots at random points with WP IV > 49.7 was 1.85 ± 0.39 vs. 6.81 ± 1.51 in RPs with WP

IV < 49.7); mean amount of grass cover in 1m2 turtle plots in WV was 10.46% ±

1.40. As with edible herbaceous flora, the abundance or presence of mushrooms or invertebrate prey or the amounts of ground cover (facilitating avoidance of predators) might also be correlated to the predominance of different tree taxa (e.g., increased earthworm activity on sites with higher pH). The ostensible avoidance by male Wood Turtles of White Oaks and Red Maple in Virginia and of Sugar Maple

227 and White Pine in West Virginia is difficult to reconcile with site conditions where such taxa occur. This result may be a statistical artifact that does not reflect biological necessity.

Forest Types and Seral Stages – Patch Scale

My study exemplifies that the scale at which habitat patches are categorized can be a significant factor in accurately portraying habitat use. When the difference in forest type groups were characterized at the stand scale (patches generally 10-

20ha) there were no significant differences between WV turtle and random points, though differences did occur for two of the five groups in VA (Fig. 5.3). Such results could be interpreted as showing that Wood Turtles do not exhibit habitat preferences with regard to types of forest. When forest types were examined at the scale of 400m2 plots, however, significant differences between turtle and random points consistently occurred (Fig. 5.2), with turtles exhibiting a preference for mesic deciduous (Dm), mesic mixed pine and deciduous (Mm), and sub-mesic oak (Om), while showing a tendency to avoid tracts of oligotrophic oak (Od) and pine (P).

Clearly, simply using easily downloaded forest type characterizations made at the stand spatial scale is not sufficient for inferring habitat selection by an animal with small home ranges or activity areas (0.2-13.3ha for this study, with a mean of 2.3ha for 59 telemetered Wood Turtles during June-August in 2011-2014; see Chapter 1).

If stand-scale characterizations were accurate for entire stands (i.e., within- stand forest composition was homogeneously distributed), then there would be no difference between the observed frequencies of turtle or random points when

228 characterized at the stand or plot scales. Which was not the case; there were consistently different frequencies for points characterized at the two different scales

(Fig. 5.4), in large part due to discrepancies within the Om and P forest type groups. In VA, numerous 400m2 plots within stands characterized as Om were actually Od or Dm, while numerous plots in WV within stands characterized as P were actually M and many of those in stands characterized as Om were actually M,

Dm, or Mm. Submergent properties at spatial scales of relevance to Wood Turtle habitat use are not evident at coarse grained stand categorization where pertinent discriminatory details of pattern are lost. While the use of stand-scale or more coarsely defined land cover characterizations may be suitable for modeling Wood

Turtle population distribution across broad landscapes, this study indicates a finer resolution is required for ascertaining preferential use of habitats by individuals within populations.

As for coarse-filter structural differences (viz., seral stages), Wood Turtles exhibited a tendency to avoid mid-successional and early successional habitat

(Figs. 5.7 & 5.8); by esh I refer here to young stands of forest regenerating after logging, not to shrubby or grassy ruderal or natural openings in the forest. Wood

Turtles showed a preference for small natural or anthropogenic clearings dominated by shrubs (e.g., Autumn Olive (Elaeagnus umbellata)) or herbaceous ground cover, including dry as well as moist sites (i.e., seeps) (Fig. 5.2). These types of shrubby or grassy clearings with few or no overstory trees are lumped into the early successional rubric under some land cover classifications. However, they are

229 structurally and often compositionally different than young sites of regenerating forest with high stem densities of saplings. Managerial mishaps are possible when relevant distinctions of pattern go unrecognized within categorical coalescence.

Synthesis

This study is consistent with others in finding that Wood Turtles have an affinity for microhabitats with an abundance of herbaceous cover (Compton et al.

2002, Akre and Ernst 2006, McCoard 2016a). Such sites can be found far from streams and are not limited to floodplains or riparian areas (this study). For instance, small natural canopy gaps are regularly used by Wood Turtles (Remsburg et al. 2006, this study) and are important for sustaining herbal growth, richness, and persistence (Goldblum 1997, Anderson and Leopold 2002). By allowing for a greater range of forest floor light levels and temperature regimes, gaps allow for more floristic richness or abundance and enhanced thermoregulatory opportunities.

The forest growth at a particular site is due to the complex interplay of numerous physical factors, disturbance regimes, and biotic interactions (Ashe

1922, Gleason 1926). Localized factors such as disturbance, elevation, slope, aspect, topographic position, slope configuration, and moisture availability are primary influences on the composition of tree taxa at forested tracts in the Central

Appalachians (Lawrence et al. 1997). As they mature, seedlings sort out on slopes along gradients of light, moisture, and nutrient availability. For instance, McCarthy et al. (1984) found that oak species (Quercus rubra, prinus, and coccinea) were distributionally replaced based on relative resistance to low soil moisture and

230 nutrients. At montane sites in western Virginia, differences in soil moisture and depth, aspect, and topography explained differences in vegetation on upper and lower slopes (Stephenson and Mills 1999). In an eastern Kentucky deciduous forest, north aspect slopes had higher productivity (McEwan and Muller 2011). Shifts in floristic composition can be particularly facilitated where disturbances occur in the presence of advanced regeneration (tree seedlings that are already present on site)

(Goins et al. 2013).

In both states, both 400m2 and 1m2 turtle plots contained significantly more herbaceous taxa than did random plots (Figs. 5.14 & 5.15). If certain herbaceous plants are an important foraging resource, Wood Turtles may be differentially using forest tracts dominated by different tree taxa. This study provides some evidence that the number of herbaceous taxa in the 400m2 plots at these VA and WV sites varied with forest type; Fleming and Coulling (2001) reported a similar finding in montane VA, as did Hutchinson and colleagues (1999) in the Appalachian region of SE Ohio. This may be a reason the Wood Turtles tended to avoid sites dominated by Chestnut and Scarlet Oaks or Virginia and White Pines. Perhaps this is a bet-hedging stategy in that at any given time sites with greater herbaceous richness are more likely to have some taxa one can use. However, since forest development is a complex of climatic, topographic, and edaphic influences, complicated by biotic interactions (competition, predation, symbioses) and historical contingency, any correlations which can be discerned involving specific tree and herbaceous taxa must be accepted with reservations.

231 In West Virginia Rentch and colleagues (2005) found the lowest herbal richness in Chestnut Oak forests; these tracts were associated with acidic conditions, which typify my study sites (Fleming 2012). In Kentucky, oak plots were in the driest most nutrient poor sites, while maple species were in the more mesic and nutrient rich sites (McEwan and Muller 2011). The maple communities had significantly greater richness than the oak communities, though the floristic diversity within these oak or maple communities was indistinguishable over the growing season. During the growing season plants can regenerate the consumed parts, so faunal movements can be curtailed and foraging can continue within the same area for extended periods (Owen-Smith et al. 2010). Understory species composition and richness in SE Ohio oak forests were correlated with soil moisture and pH (DeMars and Runkle 1992, Hutchinson et al. 1999). The age (seral stage) of forest tracts influences the herbaceous community present there; disturbance sensitive species are underrepresented in secondary forests (DeMars and Runkle

1992, Dyer 2010, Matlack and Schaub 2011). However, in SE Ohio forests Olivero and Hix (1998) found herbaceous plant assemblages to vary between aspects, but not with stand age. Composition in Kentucky forests correlated with light flux and soil moisture related to canopy openness, evaporation, and aspect (Adkison and

Gleeson 2004). In Ohio, herb species richness was higher on south aspect slopes, but density was greater on north aspect slopes (Small and McCarthy 2003). Aspect can correlate with pH and other soil fertility measures (McEwan et al. 2005). Fine- scale (tens of meters) variation in nitrogen and light availability affects understory

232 communities (Frelich et al. 2003). The herbaceous layer responds to disturbances in each of the three major vertical layers in a forest ecosystem, viz., the overstory canopy, understory vegetation, and the forest floor and soil (Roberts 2004). As an outcome of timber harvesting, local floristic distributions can be reduced or altered

(Meier et al. 1995). Some fauna in turn could be forced into smaller fragments of intact forest which could potentially increase interspecific competition or exposure to predators (Hagan et al. 1996). Habitat fragmentation and loss may potentially increase or decrease either the frequency or the distance of individual movements

(Fahrig 2007). As the above citations make clear, multiple physical factors affect floristic composition and distribution, thus Wood Turtle habitat use can be directly or indirectly influenced as well.

The Wood Turtle data used in this study lack a discrete behavioral context.

Although animal behaviors can occur synchronously, i.e., multi-tasking (Fortin et al. 2004), many behaviors are asynchronous since the habitat characteristics associated with meeting different needs are spatially segregated (Roever et al.

2014). Behaviors can have opposing habitat selection patterns, thus obscuring the detection of selection in pooled models (Roever et al. 2014). For instance, a patch may provide optimal cover from predators or osmo-regulatory benefits, but sub- optimal provision of food resources (Downes 2001). Selection of food patches can be examined in terms of which plant species are accepted for feeding when encountered (Owen-Smith et al. 2010). Precise measures of foraging benefits to

Wood Turtles would involve close-range observations on the precise types of plants

233 consumed, their age or condition, estimates of their nutritional value, and the food intake rate.

Though herbaceous taxa richness and cover were positively correlated with

White Ash, and VA Wood Turtles showed a preference for sites with high importance values for White Ash, it may be difficult to use this species as a management indicator. When I visited the VA site in the summer of 2016, every mature White Ash I observed was dead. Loss of White Ash due to the Emerald Ash

Borer (Agrilus planipennis) will open up niche space and cause a shift in tree composition across wide areas of the Turtle’s distribution. Based on similar affinities for elevation, soil fertility/pH, and moisture regimes (Burns and Honkala

1990, Mueller 2000), taxa such as Tulip Tree, White Oak, Sugar Maple, Red

Maple, Basswood, Black Cherry, Elms, or Cucumber Magnolia (Magnolia acuminata) may increase in dominance at sites vacated by White Ash.

It must be remembered that just because a site has high amounts of White

Pine or Chestnut or Scarlet Oaks does not mean that Wood Turtles cannot or do not use it; e.g., nearly a third of VA turtle points had a substantial component of

CO. In fact, at the plot with the highest importance value for any single tree species, a value of 98 for Chestnut Oak, a Wood Turtle was present. Such sites can have habitat attributes that the turtles prefer, such as LWD, abundant mushrooms, particular forbs, or dense understories. Though the CART analysis results for VA found that ca. 70% of the turtle points could be discerned from random points on the basis of the low importance value of Chestnut Oak alone, this still leaves 30%

234 of the points that were not distinguished. For a rare and vulnerable species this degree of uncertainty is particularly important, meaning that precaution must be exercised when devising forest management prescriptions based on turtle preferential tendencies regarding importance values or forest type groups. There are complexities involved even for a category as ostensibly fine-grained as a forest type utilized by the USFS or other agency; for example, multiple types of “Chestnut Oak forest” can be distinguished by differences in their understory flora (Fleming and

Coulling 2001).

This study presents evidence that some forest types are avoided or preferred more than others. If this is indeed factual, one must not automatically conclude that it is acceptable to cut down stands of a relatively avoided forest type (e.g., Chestnut

Oak or White Pine). Wood Turtles are labile in their use of sites with different tree taxa and proportions of tree taxa (i.e., different forest types); for example, ca. 16% of Wood Turtle plots were pine or oligotrophic oak (Table 5.3). Moreover, the scale at which forests are typically intensively logged (individual tracts of 10-20ha or more) is not the spatial scale at which Wood Turtles typically move about in the summer (ca. 1-2 ha). Stands that may be of a non-preferred type can have many inclusions of smaller tracts of preferred forest. So, since forest “stands” are not homogeneous (Fig. 5.6), this scale should not be used when managing Wood Turtle habitat from a silvicultural perspective. If a decision is made that forested tracts of a certain composition can be logged without negative impacts to Wood Turtles because they are of a non-preferred type (e.g., Chestnut Oak or White Pine), such

235 habitat removal must still be accomplished at the appropriate scale, meaning small tracts of individual selection or small group selection, and appropriate time

(implemented only during winter months when Wood Turtles are totally aquatic).

Conservation Recommendations

Though often characterized as a riparian species (see, e.g., McCoard 2016b or USDA FS $$$$), Wood Turtles regularly range far afield in dry upland habitats; ca. 95% of turtle location points were within 295 meters of the main streams, with some turtles ranging out as far as 500-700m. The understory and overstory analyses of this study clearly show that Wood Turtles in the summer regularly use a broad range of forested upland habitats, not just those nearby streams or dominated by mesic or hydric flora. A 300 meter minimal disturbance zone on both sides of occupied perennial streams would protect areas and conditions essential to their survival. The taxa and factors identified herein can be used for well-informed decisions regarding management practices, protective measures, and habitat enhancement/restoration (e.g., fabrication of small canopy gaps), as well as make predictions as to the suitability of sites as potential or current Wood Turtle habitat.

The findings of this study in the central Appalachians are of particular relevance and applicability across areas with similar ecological conditions, i.e., ecoregions with forests of similar composition (Dyer 2006, McNab et al. 2007).

Ecoregions by definition share similar species compositions, topography, and climate, thus serve as a reasonable mechanism for extrapolation (Omernik and

Bailey 1997). If extrapolation using ecoregional commonality is reasonable for

236 Wood Turtles, then the results from this study may be applicable to other sites in

Virginia, West Virginia, Maryland, Pennsylvania, New York, New Jersey, and

Massachusetts. The oak forest habitat deemed to be suitable at the broad-scale that is currently found in these states could also greatly expand in extent under some climate change scenarios (Iverson et al. 2008).

An understanding of the ecological processes under which vegetational communities develop is needed in order to determine the effects of management practices upon them and in turn upon Wood Turtles. At any site, multiple successional pathways are possible post-disturbance (Egler 1954, Connel and

Slatyer 1977). Various factors are responsible for this (e.g., site-specific physical conditions or the abundance of browsers), but it partially depends upon the starting point (see “initial floristic composition” in Egler 1954, Roberts 2004). For trees in particular this means the existence of a seed bank and advanced regeneration (the seedlings already growing at a particular site). The types and amounts of these are important for determining precisely if or where to subject an area to anthropogenic disturbance.

In recognition of our poor understanding of the precise mechanisms of extirpation, habitat selection, and community development, the results of this study suggest that where Wood Turtle populations occur in this ecoregion simply letting forests within the 300m buffer zones develop mature and old-growth conditions under a natural disturbance regime, regardless of their type, would be the best and least expensive course. Habitat complexity generally increases as forests age

237 (Franklin et al. 2002) and, amongst other benefits, this complexity provides refugia from predators (Finke and Denno 2006). A body of research indicates that canopy gaps, herbaceous vegetation, mushrooms, invertebrate richness or abundance, snags, and large woody debris amounts are generally more abundant in older forest habitats (Whitney and Foster 1988, Meier et al. 1995, Greenberg and Forrest 2003,

Van de Poll 2004, Ziegler 2004, Webster and Jenkins 2005, Keeton et al. 2007,

Scheff 2014). For instance, the stand-initiation and stem-exclusion stages of seral development (sensu Oliver and Larson 1996) (i.e., early successional habitat with high density of saplings) is commonly characterized by a depauperate herbaceous layer (Halpern and Spies 1995, Roberts 2004).

This precautionary approach (Cooney 2004) of minimizing human impacts and allowing old-growth forest conditions to develop through natural processes

(i.e., restoration by “purposeful inaction”, Trombulak 1996) is beneficial to not only

Wood Turtles. The flowers, fruits, nuts, leaves, roots, bark, and sap of many of the herbaceous, woody, and fungal taxa found where Wood Turtles occur have significant human nutritional, medicinal, and application value (Angier 1974, Horn and Cathcart 2005, Strauss 2014, United Plant Savers 2017). Aside from their ecological functionality, these non-timber resources can provide significant economic and social benefit without commercial logging taking place. See Chapter

6 and “Small Streams, Springs, and Seepages” and “Hardwood Forests” modules in

Mitchell et al. (2006) for other general habitat management guidelines.

238 Prescribed fire has been suggested as a management tool for eastern deciduous forests. Fire may not be necessary, however, for maintaining and regenerating northeastern oak forests and increased frequency of burning could potentially reduce forest herb and shrub diversity (Elliot et al. 2004, Matlack 2013).

Decay processes generally tend to mesify microsites while fire tends to xerify them

(Van Lear 1996). Hence, burning is of concern for Wood Turtles, not only due to the potential for direct mortality (for fires that occur outside of their hibernation period) and the deleterious alteration of forest composition and structure, but also because it tends to make sites hotter, drier and more open, thereby exposing turtles and other small organisms to more predators and desiccation.

The diminishment of the importance value of oaks in some eastern forests is often characterized in a negative light, however, this so-called mesification of forests (resulting in, e.g., relatively more maples) can be of benefit to some taxa, including the Wood Turtle (see Figs. 5.1, 5.2, 5.9-5.13). The mesification of oak forests is not ubiquitous, but is a trajectory dependent upon various topographical and ecological gradients (McEwan and Muller 2006, Iffrig et al. 2008, Loewenstein

2008). Increased frequency of fires may have allowed oak dominance to expand into mesic sites (White and White 1996). The general increase in mesic conditions over time is considered a natural process (“xerarch succession” in Braun 1950,

Foster et al. 1996), particularly to be expected after the unnatural expansion of oak domination facilitated by various direct and indirect anthropogenic disturbances at

239 multiple scales (e.g., even-age logging and increased burning) (Foster et al. 1996,

McEwan et al. 2010).

Because Wood Turtles regularly range far afield in dry upland habitats, providing protection to a narrowly delineated stream buffer zone, such as typically used 10-30 meters wide riparian strips (Lee et al. 2004, USFS 2004), while at the same time degrading or destroying other used habitat, does little to preserve or enhance populations. In other words, narrow or inadequately protected riparian buffer zones often fail to effectively protect the “core habitat” of Wood Turtles and a host of other species (Semlitsch and Bodie 2003, Crawford and Semlitsch 2007,

Sterrett et al. 2011). Biologically realistic expansive protected zones are needed to accommodate movements and reduce edge effects such as predation (Burke and

Gibbons 1995, Joyal et al. 2001, Steen et al. 2012); for example, predation on artificial Wood Turtle nests (Rutherford et al. 2016) and neonates (Dragon 2015) decreased as distance from rivers increased. Perennial stream courses occupied by

Wood Turtles in this and similar ecoregions should be buffered on both sides by at least a 300m minimal disturbance zone in order to mitigate for impacts to Turtle population viability and protect areas and conditions essential to their survival. This

300m zone is a bare minimum as it may not be expansive enough to include extensive pre-nesting movements of females or connectivity to other populations; conversely, in some situations the 300m standard could be reduced due to ownership patterns, topography, or habitat type (e.g., cliffs or already existent agricultural sites).

240 Table 5.1

Importance values of overstory and midstory (dbh ≥ 10cm) tree taxa present in 400m2 plots at turtle and paired random points Mean importance values of overstory and midstory (dbh ≥ 10cm) tree taxa present in 400m2 plots at Wood Turtle points and random points in Virginia and West Virginia during June-August 2011-2014. Values for minor species not shown. VFT = Virginia females (n = 144 [points, not turtles]), VFR = Virginia female random points (n = 144), VMT = Virginia males (n = 53), VMR = Virginia male random points (n = 53), WFT = West Virginia females (n = 74), WFR = West Virginia female random points (n = 74), WMT = West Virginia males (n = 49), WMR = West Virginia male random points (n = 49). “n” refers to numbers of plots, not numbers of individual turtles. See Appendix 5 for common names of trees. VFT VFR VMT VMR WFT WFR WMT WMR Taxon Acer rubra 15.5 15.0 13.2 19.7 4.4 1.9 4.3 1.0 Acer saccharum 2.7 0.7 5.0 1.8 6.2 2.9 4.2 6.1 Amelanchier spp. 1.4 0.8 0.6 0.4 0.6 0.6 0.4 0.3 Betula lenta 1.6 1.4 0.9 1.3 0.0 0.0 0.0 0.0 Carya spp. 4.0 5.1 4.4 4.0 13.0 9.9 12.6 12.8 Cornus florida 0.0 0.4 0.5 0.5 0.8 0.4 0.4 0.5 Fraxinus americana 7.6 1.7 7.7 3.2 0.4 0.0 1.1 0.4 Liriodendron 10.6 10.0 7.0 2.0 0.6 0.0 0.2 0.0 tulipifera Nyassa sylvatica 5.4 5.7 4.4 4.3 1.1 0.9 0.6 0.1 Ostraya virginiana 0.6 0.6 0.9 1.5 0.5 0.0 0.1 0.0 Pinus rigida 0.4 0.3 0.2 0.2 0.5 0.4 1.1 0.4 Pinus strobus 1.9 1.6 4.2 3.2 26.1 27.7 29.6 36.0 Pinus virginiana 1.0 0.9 4.6 2.5 13.6 16.8 13.2 12.9 Platanus 0.4 0.0 0.9 0.0 2.7 0.1 1.3 0.0 occidentalis Prunus serotina 0.3 0.6 0.2 0.4 1.2 1.0 1.7 0.6 Quercus alba 20.1 12.3 23.7 16.3 17.0 18.0 20.2 16.3 Quercus coccinea 3.6 7.7 4.6 12.4 0.4 0.8 0.0 0.4 Quercus montana 10.6 23.2 6.0 16.6 3.5 8.2 1.2 4.9 Quercus rubra 6.8 6.9 6.2 5.7 4.1 6.1 3.9 6.7 Quercus velutina 1.3 3.0 0.6 2.3 0.3 0.6 0.0 0.0 Robinia pseudo- 1.5 0.9 1.2 0.4 0.0 0.1 1.6 0.0 acacia Tilia americana 0.1 0.4 1.1 0.7 1.0 0.1 0.4 0.1 Ulmus spp. 0.6 0.1 0.9 0.4 0.7 0.1 1.3 0.5

241 Table 5.2

Importance values of tree taxa in 400m2 plots at pooled turtle and paired random points in VA and WV Mean importance values of common overstory and midstory (dbh ≥ 10cm) tree taxa present in 400m2 plots at Wood Turtle points and random points in Virginia and West Virginia during June-August 2011-2014. Values for minor species not shown. FWT = female turtles (n = 218), FRP = female random points (n = 218), MWT = male turtles (n = 102), MRP = male random points (n = 102), VWT = Virginia turtles (n = 197), VRP = Virginia random points (n = 197), WWT = West Virginia turtles (n = 123), WRP = West Virginia random points (n = 123), WT = turtle points (n = 320), RP = random points (n = 320). “n” refers to numbers of plots, not numbers of individual turtles. Totals do not equal 100 because the IVs of 20 minor taxa were excluded. See Appendix 5 for common names of trees. FWT MWT FRP MRP VWT WWT VRP WRP WT RP Taxon Acer rubra 11.8 9.0 10.6 10.6 15.1 4.1 16.2 1.6 10.9 10.6 Acer saccharum 4.1 4.7 1.5 3.8 3.2 5.8 1.0 3.9 4.3 2.2 Amelanchier spp. 1.1 0.5 0.8 0.4 1.2 0.5 0.7 0.5 0.9 0.3 Betula lenta 1.0 0.5 0.4 0.7 1.4 0.0 1.4 0.0 0.9 0.6 Carya spp. 7.2 8.2 6.8 8.1 4.2 13.7 4.8 10.5 7.6 7.2 Cornus florida 0.2 0.5 0.4 0.5 0.1 0.7 0.4 0.4 0.3 0.3 Fraxinus americana 5.2 4.7 1.1 1.6 7.6 0.6 2.0 0.2 5.0 1.3 Liriodendron tulipifera 7.3 3.6 6.3 1.0 9.5 0.5 7.8 0.0 6.1 4.6 Nyassa sylvatica 4.0 2.5 4.1 2.4 5.1 0.9 5.3 0.8 3.5 3.6 Ostraya virginiana 0.6 0.5 1.0 0.7 0.7 0.4 0.8 0.0 0.6 0.9 Pinus rigida 0.5 0.7 0.3 0.4 0.4 0.8 0.3 0.4 0.5 0.3 Pinus strobus 9.5 16.8 10.3 18.9 2.5 27.0 2.0 30.8 11.8 13.1 Pinus virginiana 5.5 8.8 6.4 7.6 2.4 14.2 1.3 16.3 6.6 6.8 Platanus occidentalis 1.3 1.1 0.4 0.0 0.5 1.8 0.0 0.1 1.3 0.3 Prunus serotina 0.7 0.9 0.8 0.6 0.3 1.3 0.6 0.7 0.8 0.4 Quercus alba 18.8 21.7 14.2 16.5 20.7 18.4 13.4 17.1 19.7 14.9 Quercus coccinea 2.6 2.5 5.4 6.9 4.2 0.2 9.2 0.6 2.5 5.9 Quercus montana 8.3 3.8 18.3 11.1 9.1 2.4 21.5 7.8 6.9 16.0 Quercus rubra 5.9 5.0 6.7 6.2 6.7 3.9 6.6 5.9 5.6 6.5 Quercus velutina 1.0 0.3 0.8 1.1 1.1 0.2 2.8 0.5 0.8 0.9 Robinia pseudo- 1.0 1.5 0.3 0.2 1.4 0.6 0.8 0.1 1.1 0.2 acacia Tilia americana 0.3 0.8 0.6 0.4 0.4 0.8 0.5 0.1 0.5 0.5 Ulmus spp. 0.6 1.1 2.0 0.4 0.6 0.4 0.2 0.2 0.8 1.5

Totals 98.5 99.7 99.5 100.0 98.4 99.2 99.6 98.5 99.0 98.9

242 Table 5.3

Forest type groups of 400m2 plots at turtle and paired random points Forest type groups of 400m2 plots at Wood Turtle points and paired random points in Virginia and West Virginia during June-August 2011-2014 (based on importance values of all trees with dbh ≥ 10cm in each plot); for each turtle group row, numbers on top indicate counts of plots of that forest type, with proportions (%) of total plots in each turtle group below. Forest types used to define groups (USFS terminology): 3 = White Pine, 10 = White Pine – Upland Hardwoods, 33 = Virginia Pine, 39 = Table Mountain Pine, 41 = Cove Hardwoods – White Pine, 42 = Upland Hardwoods – White Pine, 45 = Chestnut Oak – Scarlet Oak – Yellow Pine, 52 = Chestnut Oak, 53 = White Oak – Northern Red Oak – Hickory, 54 = White Oak, 56 = Tulip Poplar – White Oak – Northern Red Oak, 59 = Scarlet Oak, 60 = Chestnut Oak – Scarlet Oak. Forest type groups: Br = brushy (ruderal), Dm = mesic deciduous (includes forest type 56), M = dry mixed pine and deciduous (FTs 10, 42, 45), Mm = mesic mixed pine and deciduous (FT 41), Od = oligotrophic oak (FTs 52, 59, 60), Om = sub-mesic oak (FTs 53, 54), P = pine (FTs 3, 33, 39), Seep = sparse canopy with saturated soil, Ced = Eastern Red Cedar. Turtle groups: V = Virginia, W = West Virginia, FT = female turtle locations, FR = female random points, MT = male turtle locations, MR = male random points, WT = turtle locations (males and females pooled), TR = random points locations (points for males and females pooled). Forest Type Groups

Br Dm M Mm Od Om P Seep VFT 2 49 3 5 25 58 0 2 (n=144) 1.4 34.0 2.0 3.5 17.4 40.3 0 1.4 VFR 1 21 4 1 68 49 0 0 (n=144) 0.7 14.6 2.8 0.7 47.2 34.0 0 0 VMT 4 12 1 4 6 23 2 1 (n=53) 7.5 22.6 1.9 7.6 11.3 43.4 3.8 1.9 VMR 0 4 6 2 23 18 0 0 (n=53) 0 7.5 11.3 3.8 43.4 34.0 0 0 VWT 6 61 4 9 31 81 2 3 (n=197) 3.1 31.0 2.0 4.6 15.7 41.1 1.0 1.5 VTR 1 25 10 3 91 67 0 0 (n=197) 0.5 12.7 5.1 1.5 46.2 34.0 0 0

Br Dm M Mm Od Om P Ced WFT 0 5 33 15 0 11 10 0 (n=74) 0 6.8 44.6 20.3 0 14.9 13.5 0 WFR 0 1 32 6 2 15 16 2 (n=74) 0 1.4 43.2 8.1 2.7 20.3 21.6 2.7 WMT 2 2 21 6 0 8 10 0 (n=49) 4.1 4.1 42.9 12.2 0 16.3 20.4 0 WMR 0 2 25 3 1 7 11 0 (n=49) 0 4.1 51.0 6.1 2.0 14.3 22.5 0 WWT 2 7 54 21 0 19 20 0 (n=123) 1.6 5.7 43.9 17.1 0 15.4 16.3 0 WTR 0 3 57 9 3 22 27 2 (n=123) 0 2.4 46.3 7.3 2.4 17.9 22.0 1.6

243

Table 5.4

Forest types of stands at turtle and random points Forest types of stands at turtle points and random points in Virginia and West Virginia during June-August 2011-2014 (based on USFS stand inventory); for each row, top numbers indicate counts of points in that forest type, with proportions (%) of total points in each turtle group below. Forest types used to define groups: 3 = White Pine, 10 = White Pine – Upland Hardwoods, 33 = Virginia Pine, 39 = Table Mountain Pine, 41 = Cove Hardwoods – White Pine, 42 = Upland Hardwoods – White Pine, 45 = Chestnut Oak – Scarlet Oak – Yellow Pine, 52 = Chestnut Oak, 53 = White Oak – Northern Red Oak – Hickory, 54 = White Oak, 56 = Tulip Poplar – White Oak – Northern Red Oak, 59 = Scarlet Oak, 60 = Chestnut Oak – Scarlet Oak. Forest type groups: Dm = mesic deciduous (includes forest type 56), M = dry mixed pine and deciduous (FTs 10, 42, 45), Mm = mesic mixed pine and deciduous (FT 41), Od = oligotrophic oak (FTs 52, 59, 60), Om = mesic oak (FTs 53, 54), P = pine (FTs 3, 33, 39). Turtle groups: FT = female turtle locations, FR = female random points, MT = male turtle locations, MR = male random points, WT = turtle locations (males and females aggregated), RP = random points locations (points for males and females aggregated). Turtle groups Virginia West Virginia Forest FT FR MT MR WT RP FT FR MT MR WT RP types Dm 56 4 4 0 1 4 5 0 0 0 0 0 0 3.7 3.1 2.0 2.5 2.8 - - M 10 0 0 0 0 0 0 3 3 2 1 5 4 3.8 4.2 3.9 2.1 3.9 3.3 42 0 0 0 0 0 0 24 22 15 18 39 40 30.4 30.5 29.4 36.7 30.0 33.1 45 0 0 0 0 0 0 0 2 0 0 0 2 - 2.8 1.65 Mm 41 4 1 11 4 15 5 0 0 0 0 0 0 3.7 0.8 22.0 8.2 9.4 2.8 - - Od 52 0 3 1 5 1 8 0 0 0 0 0 0 2.3 2.0 10.2 0.6 4.5 - - 59 0 2 0 2 0 4 0 0 0 0 0 0 1.5 4.1 2.3 - - 60 3 8 2 3 5 11 0 0 0 0 0 0 2.8 6.2 4.0 6.1 3.1 6.2 - - Om 53 75 93 31 29 106 122 32 23 20 25 52 48 68.8 72.1 62.0 59.2 66.7 68.5 40.5 31.9 39.2 51.0 40.0 39.7 54 20 16 5 4 25 20 0 0 0 0 0 0 18.3 12.4 10.0 8.2 15.7 11.2 - - P 3 0 0 0 1 0 1 0 0 0 0 0 0 2.0 0.6 - - 33 0 0 0 0 0 0 20 22 14 5 34 27 25.3 30.6 27.5 10.2 26.2 22.3 39 3 2 0 0 3 2 0 0 0 0 0 0 2.7 1.6 1.9 1.1 Point totals 109 129 50 49 159 178 79 72 51 49 130 121

244 Table 5.5

Seral stage of stands at turtle points and random points Seral stage of stands at turtle points and random points in Virginia and West Virginia during June-August of 2011-2014 and 2011-2013 (based on USFS stand inventory ages); for each turtle group row, top numbers indicate counts of points in that seral stage, with proportions (%) of total stands in each turtle group below. Seral stages: esh = early successional habitat (0-35 years old), mid = mid-successional habitat (36- 75 years old), mature = mature forest habitat (76-140 years old, depending on forest type), OG = old-growth forest (> 100-140 years old, depending on forest type). Turtle groups: FT = female turtle locations, FR = female random points, MT = male turtle locations, MR = male random points, WT = turtle locations (males and females aggregated), TR = random points locations (points for males and females aggregated). n = number of points in VA, WV. Seral stages Virginia West Virginia Turtle esh mid mature OG esh mid mature OG group 2011-2014 FT 7 19 88 30 NA NA 63 11 (n=144,74) 4.9 13.2 61.1 20.8 - - 85.1 14.9 FR 16 28 82 18 NA NA 55 19 (n=144,74) 11.1 19.4 56.9 12.5 - - 74.3 25.7 MT 4 2 38 8 NA NA 45 4 (n=52,49) 7.7 3.8 73.1 15.4 - - 91.8 8.2 MR 6 8 33 6 NA NA 42 7 (n=53,49) 11.3 15.1 62.3 11.3 - - 85.7 14.3 WT 11 21 126 38 NA NA 108 15 (n=196,123) 5.6 10.7 64.3 19.4 - - 87.8 12.2 TR 22 36 115 24 NA NA 97 26 (n=197,123) 11.1 18.3 58.4 12.2 - - 78.9 21.1

2011-2013 FT 2 19 71 26 NA NA 37 10 (n=118,47) 1.7 16.1 60.2 22.0 - - 78.7 21.3 FR 16 25 60 17 NA NA 33 14 (n=118,47) 13.6 21.2 50.8 14.4 - - 70.2 29.8 MT 2 1 24 5 NA NA 28 4 (n=32,32) 6.3 3.3 75.0 15.6 - - 87.5 12.5 MR 4 7 19 3 NA NA 28 4 (n=33,32) 12.1 21.2 57.6 9.1 - - 87.5 12.5 WT 4 20 95 31 NA NA 65 14 (n=150,79) 2.7 13.3 63.3 20.7 - - 82.3 17.7 TR 20 32 79 20 NA NA 61 18 (n=151,79) 13.2 21.2 52.4 13.2 - - 77.2 22.8

245 Table 5.6

Floristic richness and cover in plots at turtle and paired random points Values for floristic richness and cover in plots at turtle points and paired random points in Virginia and West Virginia during June-August; forb and seedling taxa in 1m2 plots were counted in 2013-2014, forb taxa in 400m2 plots were counted in 2011- 2013, herbaceous cover (%) in 1m2 plots and tree taxa (≥10cm dbh) in 400m2 plots were measured in 2011-2014. Reported in descending order are means, standard errors, and ranges. Herbaceous cover is forb cover and grass cover combined. Site groups: FWT = female Wood Turtles, FRP = female random points, MWT = male Wood Turtles, MRP = male random points. Virginia West Virginia

Variable FWT FRP MWT MRP FWT FRP MWT MRP 1m2 Forb taxa 2.72 1.28 4.33 0.92 2.67 2.63 4.42 2.50 0.27 0.22 0.46 0.32 0.30 0.35 0.47 0.46 0-9 0-10 0-9 0-9 0-7 0-9 1-8 0-8

Seedling taxa 3.10 3.36 2.83 3.31 1.92 1.69 2.00 1.50 0.21 0.20 0.30 0.30 0.22 0.19 0.36 0.26 0-10 0-11 0-7 0-7 0-5 0-5 0-8 0-5

Herbaceous 14.6 3.3 14.0 1.3 14.4 8.1 15.9 8.0 cover 2.05 0.74 2.57 0.65 2.13 1.72 2.35 1.53 0-100 0-79 0-73 0-28 0-75 0-81 0-76 0-53 400m2 Forb taxa 14.3 10.2 16.1 9.9 16.7 11.4 16.8 15.1 0.71 0.64 1.33 1.28 0.95 0.98 1.26 1.23 0-31 0-35 1-30 2-35 5-33 0-29 5-31 1-29

Tree taxa 5.42 5.69 5.81 5.75 5.68 5.35 5.41 5.12 0.15 0.14 0.25 0.26 0.15 0.18 0.23 0.18 1-11 1-10 2-9 2-12 3-9 3-10 3-9 3-8

246 Table 5.7

Test results comparing floristic variables at turtle and paired random points Results of paired Wilcoxon signed rank tests and paired t-tests comparing floristic variables obtained at turtle points and random points in Virginia (VA) and West Virginia (WV) during June-August; reported in descending order are p values, V or t statistic values, and degrees of freedom. Forb and woody seedling taxa in 1m2 plots were counted in 2013-2014, forb taxa in 400m2 plots were counted in 2011-2013, herbaceous cover (%) in 1m2 plots and tree taxa (≥10cm dbh) in 400m2 plots were measured in 2011-2014. FWT = female Wood Turtle points, FRP = female random points, MWT = male Wood Turtle points, MRP = male random points. Comparisons with significant results are in bold. VA WV FWT-FRP MWT-MRP FWT-FRP MWT-MRP 1 m2 Forb taxa richness p <0.0001 <0.0001 0.360 0.0012 V 1584 431 417 192

Woody seedling taxa p 0.373 0.198 0.966 0.272 V 1258 183 477 93

Herbaceous cover p <0.0001 <0.0001 0.0104 0.0026 V 4827 74 1465 264

400m2 Forb taxa richness p <0.0001 0.0064 0.00016 0.165 V or t 4971 2.92 4.10 1.42 df 32 46 31

Tree taxa richness p 0.198 0.715 0.131 0.387 V 3252 599 1049 474

247 Table 5.8

Well-supported conditional logistic regression models – using tree importance values Best of fifteen conditional logistic regression models of tree importance values at Wood Turtle sites in Virginia and West Virginia, USA in 2010-2014, based on proportions in 400m2 plots. Type of model in brackets (see Appendix 5). LogLik = model log-likelihood; K = number of parameters; ∆AICc= difference in Akaike Information Criterion corrected for small sample size from the top model; higher AICc weights denote models that are supported among the set of candidate models; cumulative weight (Cum. wt.) is the running sum of the individual model weights (listed are those models with a cumulative weight ≥ 0.95). See Appendix 5 for identification of tree acronyms. LogLik K ∆AICc AICc Cum.wt. Models weight

Virginia females

WO+NRO+SM+RM+BW+TP+WA+BG+ HICK+BLBIR+ELM -71.31 11 0 0.61 0.61 WO+CO+NRO+RM+SM+HICK+BG+TP+WA [Om]-74.42 9 1.91 0.23 0.84 WO+CO+NRO+SO+BO -79.51 5 3.66 0.10 0.94 WO+CO+SO+BO+BG+HICK -79.56 6 5.84 0.03 0.97 1 [Null model] -99.81 0 34.05 0.00 1.00

Virginia males

WO+CO+NRO+SO+BO [Oaks] -24.39 5 0 0.34 0.34 CO+SO+BO+BG+HICK+WO -23.55 6 0.56 0.26 0.60 WP+VP+WO+CO+RM+SO+BG -22.50 7 0.75 0.23 0.83 WP+CO+SM+RM+WO+BG+WA+TP+SO+ NRO+HICK+VP -17.33 12 2.57 0.09 0.92 1 [Null model] -37.43 0 15.48 0.00 1.00

West Virginia females

WP+WO+CO+SM+RM+BLCH+SYC+ NRO+HICK+VP -34.45 10 0 0.50 0.50 BW+BLCH+ELM+SM+RM+SYC [Mesic] -39.21 6 0.52 0.38 0.88 WP+VP+WO+CO+RM+BG+SO+HICK -39.34 8 5.22 0.04 0.92 WO+CO+NRO+RM+SM+HICK+BG -40.82 7 5.93 0.03 0.94 RM+SM -46.42 2 6.42 0.02 0.96 1 [Null model] -51.29 0 12.09 0.00 1.00

West Virginia males

WP+VP+CO+SM+RM+WO+BLCH+ ELM+NRO+HICK -10.40 10 0 0.47 0.47 WP+VP+WO+CO+RM+BG+WA+HICK [M] -12.89 8 0.06 0.45 0.92 WO+CO+NRO+RM+SM+HICK+BG -16.02 7 3.96 0.06 0.99 1 [Null model] -33.96 0 24.60 0.00 1.00

248 Table 5.9

Best conditional logistic regression model variables – using IVs Conditional logistic regression model variables that best explain Wood Turtle meso- scale habitat selection based on tree importance values at sites in Virginia and West Virginia, USA in 2010-2014. Measured values are based on counts and dbh of trees in 400-m2 sampling plots. Model coefficients were obtained through model averaging. Variables with positive values for coefficients are preferred, while negative values indicate avoidance. Variables in bold denote those with coefficients that did not overlap zero. See Appendix 5 for identification of tree acronyms. Measured values (mean ± se) Model coefficient ± se Odds ratio Unit increase

Variable Turtle Plots Random Plots

Virginia females WO 20.1 ± 1.88 12.3 ± 1.30 0.03 ± 0.01 1.030 1 unit CO 10.7 ± 1.74 23.2 ± 1.99 -0.02 ±0.01 0.980 1 unit SO 3.7 ± 0.92 7.7 ± 1.13 -0.02 ±0.01 0.980 1 unit SM 2.7 ± 0.67 0.7 ± 0.29 0.06 ± 0.03 1.062 1 unit WA 7.7 ± 1.07 1.7 ± 0.54 0.07 ± 0.02 1.073 1 unit SERV 1.4 ± 0.37 0.8 ± 0.20 0.07 ± 0.05 1.073 1 unit BW 0.1 ± 0.06 0.4 ± 0.13 -0.22 ± 0.12 0.803 1 unit BO 1.3 ± 0.36 3.0 ± 0.50 -0.06 ± 0.03 0.942 1 unit ELM 0.6 ± 0.15 0.1 ± 0.07 0.20 ± 0.13 1.221 1 unit

Virginia males WP 4.2 ± 1.59 3.2 ± 1.08 -0.08 ±0.04 0.923 1 unit CO 6.0 ± 1.64 16.9 ± 2.55 -0.05 ±0.03 0.951 1 unit SO 4.6 ± 1.73 12.6 ± 2.49 -0.06 ±0.03 0.942 1 unit BO 0.6 ± 0.41 2.3 ± 0.75 -0.11 ±0.06 0.896 1 unit RM 13.2 ± 1.45 19.5 ± 1.86 -0.09 ±0.05 0.914 1 unit SM 5.0 ± 1.48 1.9 ± 0.83 0.10 ± 0.07 1.105 1 unit SERV 0.6 ± 0.25 0.4 ± 0.21 0.35 ± 0.20 1.419 1 unit BL 1.2 ± 0.60 0.4 ± 0.16 0.15 ± 0.11 1.162 1 unit IW 0.9 ± 0.37 1.4 ± 0.40 -0.22 ±0.14 0.803 1 unit

West Virginia females CO 2.1 ± 0.79 9.8 ± 2.77 -0.07 ±0.03 0.932 1 unit RM 5.2 ± 0.92 2.5 ± 0.84 0.11 ± 0.05 1.116 1 unit HICK 10.1 ± 2.07 10.3 ± 1.36 0.05 ± 0.02 1.051 1 unit VP 10.5 ± 2.41 17.9 ± 3.07 0.02 ± 0.02 1.020 1 unit SYC 3.4 ± 1.59 0.04 ± 0.04 0.40 ± 0.23 1.492 1 unit

West Virginia males CO 1.0 ± 0.55 5.2 ± 1.84 -0.62 ± 0.25 0.538 1 unit SM 5.6 ± 2.23 6.5 ± 2.05 -0.26 ± 0.16 0.771 1 unit WP 29.1 ± 3.94 38.3 ± 4.13 -0.11 ± 0.10 0.896 1 unit

249 Table 5.10

Synopsis of best conditional logistic regression model variables – using IVs of various turtle groups Conditional logistic regression model variables that best explain Wood Turtle meso- scale habitat selection based on taxon importance values at sites in Virginia and West Virginia, USA in 2010-2014. Taxon importance values are based on counts and dbh of trees ≥ 10cm dbh in 400m2 sampling plots. Model coefficients were obtained through model averaging. An X of positive sign denotes variables that are preferred, while a –X indicates avoidance. Variables with exes are only those with coefficients that did not overlap zero for that particular site group. See Appendix 5 for identification of tree acronyms. Type of Site VFT VMT WFT WMT FT MT WT Taxon

WP -X -X -X

VP -X

WO +X +X

CO -X -X -X -X -X -X

SO -X -X -X -X

NRO -X

RM -X +X -X

SM +X +X -X

WA +X +X +X

SERV +X +X

HICK +X -X

BW

BO -X

ELM -X

250 Table 5.11

Well-supported conditional logistic regression models – using IVs and pooled turtle groups Best of fifteen conditional logistic regression models of tree importance values at combined Wood Turtle sites in Virginia and West Virginia, USA in 2010-2014, based on values in 400m2 plots. LogLik = model log-likelihood; K = number of parameters; ∆AICc= difference in Akaike Information Criterion corrected for small sample size from the top model; higher AICc weights denote models that are supported among the set of candidate models; cumulative weight (Cum. wt.) is the running sum of the individual model weights (best models are those with a cumulative weight > 0.95). See Appendix 5 for identification of tree acronyms. LogLik K ∆AICc AICc Cum. Model weight weight

Virginia and West Virginia females

WO+NRO+SM+RM+BW+TP+WA+BG+ HICK+ELM [Dm] -118.90 10 0 0.91 0.91 WO+CO+NRO+RM+SM+HICK+BG+TP+WA -122.87 9 5.97 0.05 0.96 1 [Null model] -151.11 0 43.89 0.00 1.00

Virginia and West Virginia males

WP+VP+WO+CO+SM+RM+BG+WA+TP+ SO+NRO+HICK -47.39 12 0 0.50 0.50 WO+CO+NRO+SO+BO [Oaks] -55.23 5 0.35 0.42 0.93 WO+CO+SO+BO+BG+HICK -56.55 6 5.11 0.04 0.97 1 [Null model] -141.40 0 20.98 0.00 1.00

Virginia and West Virginia Wood Turtles

WP+VP+WO+CO+SM+RM+BG+WA+TP+ SO+NRO+HICK [Global] -196.41 12 0 0.93 0.93 WO+CO+NRO+RM+SM+HICK+BG+TP+WA -202.39 9 5.75 0.05 0.98 1 [Null model] -241.76 0 66.04 0.00 1.00

251 Table 5.12

Best conditional logistic regression model variables – using IVs and pooled turtle groups Conditional logistic regression model variables that best explain Wood Turtle meso-scale habitat selection based on tree importance values at combined sites in Virginia and West Virginia, USA in 2010-2014. Measured values are based on counts and dbh of trees in 400m2 sampling plots. Model coefficients were obtained through model averaging. Variables with positive values for coefficients are preferred, while negative values indicate avoidance. Variables in bold denote those with coefficients that did not overlap zero. See Appendix 5 for identification of tree acronyms. Measured values (mean ± se) Model coefficient ± se Odds ratio Unit increase Variable Turtle Plots Random Plots VA and WV females WO 18.8 ± 1.37 14.2 ± 1.17 0.02 ± 0.01 1.020 1 unit CO 8.3 ± 1.22 18.3 ± 1.50 -0.02 ±0.01 0.980 1 unit SO 2.6 ± 0.62 5.4 ± 0.79 -0.03 ±0.01 0.970 1 unit SM 4.1 ± 0.70 1.5 ± 0.37 0.05 ± 0.02 1.051 1 unit WA 5.2 ± 0.75 1.1 ± 0.36 0.08 ± 0.02 1.083 1 unit SERV 1.1 ± 0.25 0.6 ± 0.17 0.18 ± 0.08 1.197 1 unit IW 0.6 ± 0.13 1.0 ± 0.26 -0.05 ±0.04 0.951 1 unit ELM 0.6 ± 0.14 2.0 ± 0.34 -0.12 ±0.04 0.887 1 unit

VA and WV males WP 16.8 ± 2.02 18.9 ± 2.33 -0.10 ±0.04 0.905 1 unit VP 8.8 ± 1.73 7.6 ± 1.61 -0.08 ±0.04 0.923 1 unit CO 3.8 ± 0.95 11.1 ± 1.64 -0.10 ±0.05 0.905 1 unit SO 2.5 ± 0.96 6.9 ± 1.46 -0.09 ±0.05 0.914 1 unit NRO 5.0 ± 1.00 6.2 ± 0.86 -0.09 ±0.06 0.914 1 unit BO 0.3 ± 0.22 1.1 ± 0.40 -0.10 ±0.06 0.905 1 unit RM 9.0 ± 1.00 10.6 ± 1.36 -0.11 ±0.05 0.896 1 unit SM 4.7 ± 0.96 3.8 ± 0.85 -0.12 ±0.05 0.887 1 unit TP 3.6 ± 0.95 1.0 ± 0.43 -0.07 ±0.05 0.932 1 unit HICK 8.2 ± 1.21 8.1 ± 1.16 -0.09 ±0.04 0.914 1 unit ELM 1.1 ± 0.36 0.4 ± 0.15 0.13 ± 0.10 1.139 1 unit BW 0.8 ± 0.31 0.4 ± 0.20 0.11 ± 0.08 1.116 1 unit

VA and WV Wood Turtles WP 11.8 ± 0.96 13.1 ± 1.13 -0.02 ±0.01 0.980 1 unit CO 6.9 ± 0.89 16.0 ± 1.16 -0.03 ±0.01 0.970 1 unit SO 2.5 ± 0.52 5.9 ± 0.71 -0.04 ±0.01 0.961 1 unit NRO 5.6 ± 0.58 6.5 ± 0.55 -0.03 ±0.01 0.970 1 unit WA 5.0 ± 0.59 1.3 ± 0.30 0.05 ± 0.02 1.051 1 unit SERV 0.9 ± 0.18 0.3 ± 0.06 0.15 ± 0.06 1.162 1 unit IW 0.6 ± 0.11 0.9 ± 0.19 -0.04 ±0.03 0.961 1 unit ELM 0.8 ± 0.15 1.5 ± 0.24 -0.06 ±0.03 0.942 1 uni

252 Table 5.13

Herbaceous taxa used in indicator species analyses – 400m2 plots Taxa used in indicator analyses of herbaceous taxa present in 400m2 plots at turtle points and random points in Virginia and West Virginia during June-August 2013-2014. Taxa present in a state are marked with a . Taxa with significant indicator values (by themselves, not in combination with other taxa) for a site group are marked with an X (alpha ≤ 0.05); o = an indicator for that site group (0.05 < alpha ≤ 0.11); – = not an indicator; denotations are in the order VA/WV. Site groups: FWT = female Wood Turtles, FRP = female random points, MWT = male Wood Turtles, MRP = male random points. Wetland class: FAC = facultative, FACU = facultative upland, FACW = facultative wetland, OBL = obligate wetland, UPL = upland. State Site Groups Wetland VA WV FWT FRP MWT MRP FWT+ FRP+ Taxon Common name class MWT MRP Ageratina altissima White Snake Root FACU X/- Agrimonia grypsophela Agrimony FACU X/- Alium cernum Wild Onion FACU Amphicarpaea Hog Peanut FAC X/X bracteata Anemone americana Rnd-lobed Hepatica UPL -/X Antennaria spp. Pussytoes UPL Arisaema triphyllum Jack-in-the-pulpit FACW X/- Aster spp. Asters na -/X Boehmeria cylindrica False Nettle FACW X/- Chimaphila maculata Spotted Wintergrn. UPL X/X Cicuta maculata Water Hemlock OBL X/- Circaea lutetiana Ench.’s Nightshade FACU X/X Collinsonia canadensis Stone Root FAC X/- Cunila origanoides Dittany UPL X/- Desmodium spp. Tick Trefoil UPL -/X (glabellum) Dioscorea villosa Wild Yam FAC Epigaea repens Trailing Arbutus UPL -/X Eurybia divaricata White Wood Aster FACU/ X/o UPL Galium triflorum Bedstraws FACU X/X Galium circaezan Wild Licorice UPL Gaultheria procumbens Teaberry FACU /X Geranium maculatum Wild Geranium FACU -/X Goodyera pubescens Rattlesnake Plantain FACU Hieracium paniculatum Panicled Hawkwort UPL Hieracium venosum Hawkwort UPL -/o Impatiens capensis Jewelwort FACW -/X X/- Isotria verticillata Whorled Pogonia FACU Lespedeza spp. Bush Clover UPL (procumbens) Lycopus spp. Bugleworts OBL -/X X/- Maianthemum Plume Lily FACU racemosa Medeola virginiana Indian Cuc. Root FAC Mitchella repens Partridgeberry FACU X/- Nabalus spp. Gall-of-the-Earth FACU -/X (albus, altissimus) Oxalis spp. Wood Sorrel FACU X/o Packera obovata Rnd-leaved Ragwort FACU Parthenocissus Virginia Creeper FACU quinquefolia Persicaria Tearthumb OBL X/- sagittatum/arifolium Pilea pumila Clearwort FACW X/- Polygonatum biflorum Solomon’s Seal FACU Potentilla spp. Cinquefoil FACU X/X Prunella vulgaris Heal-All FACU X/- Scuttellaria spp. Skullcap (lateriflora) FACW -/X Solidago spp. Goldenrods na Uvularia perfoliata Perfoliate Bellwort FACU -/X o/- Uvularia spp. Bellworts (puberula) FACU X/- Veronica officinalis Speedwell FACU Viola spp. Violets na X/X Perilla frutescens Beefsteak Plant FACU - /X X/- Microstigeum Stiltgrass FAC X/X vimineum Asplenium platyneuron Ebony Spleenwort FACU -/X Onoclea sensibilis Sensitive Fern FACW X/- Parathelypteris NY Fern FAC X/- noveboracensis Polystichum Christmas Fern FACU - /X arostichoides

253 Table!"#$%&!' 5.14 !"#$%&'($)*+",'-'"./,0"+/12/3/('2,"/24/(',)%"5'6*$+"7&8",0$9+$65$+:"2),"/2" ()9&/2',/)2"./,0"),0$%",'-';"<%$+$2,"/2"=>>?9@"<6),+"',",*%,6$"<)/2,+"'24"%'24)9" Herbaceous taxa with significant indicator values <)/2,+"/2"A/%1/2/'"'24"B$+,"A/%1/2/'"4*%/21"C*2$?D*1*+,"@>EE?@>EFG"%$<)%,$4"'%$" Herbaceous taxa with significant indicator values (by themselves, not in combination with other taxa) present in 400m2 plots,0$"/24/(',)%"5'6*$+" at turtle points and7D"-"H;" random./,0",0$"< points in ?Virginia5'6*$+"&$2$',0"7'+"('6(*6',$4"&8" and West Virginia during JuneI"<'(J'1$"-August 2011-2013; reported are the indicatorK/24/(L<$(/$+ valuesM ;!"L/,$"1%)*<+N"AOP"Q"AD"9'6$"B))4"P*%,6$+:"BOP"Q"BA"9'6$"B))4"(A x B) with the p-values beneath (as calculated by R package “indicSpecies”). Site groups: VMT = VA maleP*%,6$+:" WoodABP"Q"AD"B))4"P*%,6$+:"BBP"Q"BA"B))4"P*%,6$+:"AIR"Q"AD"%'24)9" Turtles, WMT = WV male Wood Turtles, VWT = VA Wood Turtles, WWT = WV Wood Turtles, VRP = VA random<)/2,+:"BIR"Q"BA"%'24)9"<)/2,+ points, WRP = WV random points,:"BSI"Q"B$+,"A/%1/2/'"3$9'6$"%'24)9"<)/2,+ WFR = West Virginia female random points.!" P'-'"Taxa with an * had a 0.05 < p- value./,0"'2"T"0'4"'"< ≤ 0.1 for a group.?5'6*$"U">!>V "3)%"'"1%)*

"""""""" """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""./"!0!!!!!!!!!!1/"!!!!!!!!!!!.1"!!!!!!!!!!!11"!!!!!!!!!.23!!!!!!!!!!123! "#4*5!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!6*77*5!5#7&!!!!!!!!!!!!!!!!!!!!!!!!!!182!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!9/:8;!!!!!!!!!9/:8;!!!!!!!9/:8;!!!!!!9/:8;" !"#$%&'(%)%*&'++',%) ) B0/,$"L2'J$"I)),"" " " " >!VW=" " " " " " " " " >!>@F" " " !"$',-('%)"$./+-/0#*%) D1%/9)28"XAOPY" """""""""""""""">!F=Z" " " " " " """""""""""""""">!>>E" " " !,/0'1%$/%#%)2$%1&#%&%) #)1"R$'2*,"" " " " >!WFZ" >!=V=" " " " " " " " " >!>>E" >!>@[" !$'+%#,%)&$'/0.**3,) ) C'(J?/2?,0$?<*6!=ZZ" " " " " " " " " >!>>E" !+&#$) ) ) ) ) ) ) ) ) >!=[=" " " " " " " " " " >!>>V" " " 4-#0,#$'%)1.*'(5$'1%)" S'6+$"\$,,6$"" " " " >!V=>" " " " " " " " " >!>>E" " ) 60',%/0'*%),%13*%&%) L<),,$4"B/2,$%1%$$2"" " " " " >!W@=" >!V@E" " " " " " " " " " " >!>>E" >!>@E" 6'13&%),%13*%&%) ) B',$%"#$96)(J""" " " " >!F=V" " " " " " " " " >!>>Z" " " 6'$1%#%)*3&#&'%(%" ) ]2(0'2,$%^+"\/10,+0'4$"" " " >!V>F" >!=[>" " " " " " " " " >!>>E" >!>>W" " 6-**'(+-('%)1%(%5#(+'+) L,)2$"I)),"" " " " >!=E_" " " " " " " " " >!>>F" " " 63('*%)-$'"%(-'5#+) ) `/,,'28"" " " " " " " >!FWF" " " " " " " " " " " >!>FZ" " 7#+,-5'3,)+W" " " " """""""" " " " " " """""""""""""""""" """"""""""""""""">!>VE" " " " " " ) 8/'"%#%)$#/#(+) ) P%'/6/21"D%&*,*+"" " " " " " " >!F[F" " " " " " " " " " " " >!>@W" 83$.2'%)5'9%$'1%&%) ) B0/,$"B))4"D+,$%"T" " " " >!WFE" >!=ZF" " " " " " " " " >!>>E" >!>V_" :%*'3,)&$';*-$3,)" "" H$4+,%'." " " " " >!V=E" >!V_Z" " " " " " " " " >!>>F" >!>EZ" :%3*&0#$'%)/$-13,2#(+) P$'&$%%8""" " " " " >!FZ_" " " " " " " " " " >!>FZ) :#$%('3,),%13*%&3,) B/64"a$%'2/*9"" " " " " >!=EZ" " " " " " " " " " >!>VE) <#/%&'1%)%,#$'1%(%) ) I)*24?6)&$4"#$<',/('"" " " " " " >!=V_" " " " " " " " " " " " >!>@>) <'#$%1'3,)9#(-+3,) ) #'.J.)%,"T" " " " " " " >!=W=" " " " " " " " " " " " >!E>" ) =,/%&'#(+)1%/#(+'+) " C$.$6.)%,""" " " >!FWE" >!V@_" " " " " " " " >!>>=" >!>>E" " " " ) >.1-/3+)+!VFW" " " " " " " " >!>>E" >!>>E" " " " " ?'&10#**%)$#/#(+)) ) R'%,%/41$&$%%8"" " " " >!FZE" " " " " " " " " " >!>E>" @%2%*3+)+!>@>" " " BC%*'+)+!=Z_" " " " " " " " " >!>>_" >!>_E" D'*#%)/3,'*%) ) b6$'%.)%,""" " " " >!FV@" " " " " " " " " >!>>Z" " " D-*."-(3,)+%"'&&%&3,E%$';-*'3,)P$'%,0*9&"" " " " >!F=F" " " " " " " " " >!>>[" " "

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

254 Table 5.15

Herbaceous taxa used in association analyses – 400m2 plots Taxa used in association analyses of herbaceous taxa present in 400m2 plots at turtle points and random points in Virginia and West Virginia during June-August 2013-2014. Taxa present in a state are marked with an X. Taxa with significant Pearson’s phi coefficients of association (by themselves, not in combination with other taxa) for a site group are marked with an X (alpha ≤ 0.05); o = an indicator for that site group (0.05 < alpha ≤ 0.11); - = not an indicator; denotations are in the order VA/WV. Site groups: FWT = female Wood Turtles, FRP = female random points, MWT = male Wood Turtles, MRP = male random points. Wetland class: FAC = facultative, FACU = facultative upland, FACW = facultative wetland, OBL = obligate wetland, UPL = upland. State Site Groups Wetland VA WV FWT FRP MWT MRP FWT+ FRP+ Taxon Common name class MWT MRP Ageratina altissima White Snake Root FACU o/- Agrimonia grypsophela Agrimony FACU X/- Alium cernum Nod. Wild Onion FACU Amphicarpaea Hog Peanut FAC X/o bracteata Anemone americana Rnd-lobed Hepatica UPL -/X Antennaria spp. Pussytoes UPL Arisaema triphyllum Jack-in-the-pulpit FACW X/- Aster spp. Asters na -/X Boehmeria cylindrica False Nettle FACW X/- Chimaphila maculata Spot. Wintergreen UPL X/o Cicuta maculata Water Hemlock OBL X/- Circaea lutetiana Ench.’s Nightshade FACU X/X Collinsonia canadensis Stone Root FAC X/- Cunila origanoides Dittany UPL o/o Desmodium spp. Tick Trefoil UPL -/X (glabellum) Dioscorea villosa Wild Yam FAC Epigaea repens Trailing Arbutus UPL -/X Eurybia divaricata White Wood Aster FACU/ X/- UPL Galium triflorum Bedstraws FACU X/o Galium circaezans Wild Licorice UPL Gaultheria procumbens Teaberry FACU -/X Geranium maculatum Wild Geranium FACU Goodyera pubescens Rattlesnake Plantain FACU Hieracium paniculatum Panicled Hawkwort UPL X/- Hieracium venosum Hawkwort UPL Impatiens capensis Jewelwort FACW -/X X/- Isotria verticillata Whorled Pogonia FACU Lespedeza spp. Bush Clover UPL (procumbens) Lycopus spp. Bugleworts OBL -/X X/- Maianthemum Plume Lily FACU racemosa Medeola virginiana Ind. Cucumber Root FAC Mitchella repens Partridgeberry FACU X/- Nabalus spp. Gall-of-the-Earth FACU -/o (albus, altissimus) Oxalis spp. Wood Sorrel FACU -/o X/- Packera obovata Rnd-leaved Ragwort FACU Parthenocissus Virginia Creeper FACU o/- quinquefolia Persicaria Tearthumb OBL X/- sagittatum/arifolium Pilea pumila Clearwort FACW X/- Polygonatum biflorum Solomon’s Seal FACU Potentilla spp. Cinquefoil FACU X/X Prunella vulgaris Heal-All FACU X/- Scuttellaria spp. Skullcap (lateriflora) FACW -/X Solidago spp. Goldenrods na -/X Uvularia perfoliata Perfoliate Bellwort FACU -/X o/- Uvularia spp. Bellworts (puberula) FACU X/- Veronica officinalis Speedwell FACU

Viola spp. Violets na X/X Perilla frutescens Beefsteak Plant FACU X/- - /X Microstigeum Stiltgrass FAC X/X vimineum Asplenium platyneuron Ebony Spleenwort FACU -/X Onoclea sensibilis Sensitive Fern FACW X/- Parathelypteris NY Fern FAC noveboracensis Polystichum Christmas Fern FACU - /X arostichoides

255 Table!"#$%&! 5.16' !"#$%&'($)*+",'-'"./,0"1$'%+)23+"40/"()$55/(/$2,+")5"'++)(/',/)2"6&7"

,0$8+$9:$+;"2),"/2"()8&/2',/)2"./,0"),0$%",'-'<"4%$+$2,"/2"=>>8?"49),+"',",*%,9$" Herbaceous taxa with significant association values – 400m2 plots Herbaceous4)/2,+"'2@"%'2@)8"4)/2,+"/2"A/%B/2/'"'2@ taxa with significant Pearson’s phi "coefficientsC$+,"A/%B/2/'"@*%/2B"D*2$ of association (byE F*B*+,"?>GGthemselves, notE in combination with other taxa)?>GHI"%$4)%,$@"'%$",0$" present in 400m2 plots()$55/(/$2, at turtle points":'9*$+" and6F"-"J< random"./,0",0$"4 points in EVirginia:'9*$+"&$2$',0"6'+" and West Virginia during June-August 2011- 2013;('9(*9',$@"&7" reported areK"4'(L'B$"M the coefficient/2@/(N4$(/$+ values withO!>V"W"4 E:'9*$"W">!G"5)%"'"B%)*4!"" """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""()*+,-!

"""""""" """""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""./"!0!!!!!!!!!1/"!!!!!!!!.1"!!!!!!!!!!!11"!!!!!!!!!.23!!!!!!!!!!123! "#4*5!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!6*77*5!5#7&!!!!!!!!!!!!!!!!!!!!!!!!!!!182!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!9/:8;!!!!!!!!!9/:8;!!!!!!!9/:8;!!!!!!9/:8;" !"#$%&'(%)%*&'++',%) ) C0/,$"N2'L$"K)),"" " " " " " " " " " " " " " " " !"$',-('%)"$./+-/0#*%) FB%/8)27"XAQRY" " >!?=Z" " " " " " " >!>>?" " " !,/0'1%$/%#%)2$%1&#%&%) #)B"1$'2*,"U"" " " " >!H=>" >!G[V" " " " " " " " " >!>>G" >!>\?" !(#,-(#)%,#$'1%(%) ) K)*2@E9)&$@"#$4',/('"QK1" " " " " " >!??V" " " " " " " " " " " " >!>G\) !$'+%#,%)&$'/0.**3,) ) D'(LE/2E,0$E4*94/," " " " >!?\H" " " " " " " " " >!>>G" !+&#$)+44!) ) ) ) ) ) ) ) ) >!?]]" " " " " " " " " " >!>>\" " " 4-#0,#$'%)1.*'(5$'1%)" T'9+$"^$,,9$"" " " " >!HH]" " " " " " " " " >!>>G" " ) 60',%/0'*%),%13*%&%) N4),,$@"C/2,$%B%$$2"U"" " " " " >!HHH" >!G[V" " " " " " " " " " " >!>>G" >!>]]" 6'13&%),%13*%&%) ) C',$%"#$89)(L""" " " " >!?V>" " " " " " " " " >!>>G" " " 6'$1%#%)*3&#&'%(%" ) _2(0'2,$%3+"^/B0,+0'@$"" " " >!?\Z" >!?=]" " " " " " " " " >!>>?" >!>G>" " 6-**'(+-('%)1%(%5#(+'+) N,)2$"K)),"" " " " >!?G\" " " " " " " " " >!>>]" " " 63('*%)-$'"%(-'5#+) ) `/,,'27"" " " >!G]>" " " " ATK" " " " " " " >!>VV" " " " " " 7#+,-5'3,)+44!" " R/(L"R%$5)/9"XCTKY" " >!?G=" " " " """""""" " " " " " " " >!>?V" " " " " " ) 8/'"%#%)$#/#(+) ) R%'/9/2B"F%&*,*+"" " " " " " " >!G[[" " " " " " " " " " " " >!>=Z" 83$.2'%)5'9%$'1%&%) ) C0/,$"C))@"F+,$%"" " " " >!??Z" " " " " " " " " " >!>>=" " :%*'3,)&$';*-$3,)" "" J$@+,%'."U" " " " >!G\[" >!G[?" " " " " " " " " >!>H?" >!>![>" :%3*&0#$'%)/$-13,2#(+) R$'&$%%7""CTR" " " " " >!?H\" " " " " " " " " " >!>GH) :#$%('3,),%13*%&3,) C/9@"a$%'2/*8"" " " " " " " " " " " " " " " ) <'#$%1'3,)/%('13*%&%) 1'2/(9$@"#'.L.)%,""ATK" >!G\>" " " " " " " " " " " " " >!>H[" " " " " " ) =,/%&'#(+)1%/#(+'+) " D$.$9.)%,""" " " >!H?>" >!V?[" " " " " " " " >!>>G" >!>>G" " " " ) >.1-/3+)+44!" " J*B9$.)%,+"" " " >!H?>" >!VH]" " " " " " " " >!>>G" >!>>G" " " " " ?'&10#**%)$#/#(+)) ) 1'%,%/@B$&$%%7"" " " " >!G]V" " " " " " " " " " >!>=V" @%2%*3+)+44!" " a'99E)5E,0$E_'%,0"U" " " " " >!G\G" """"""""6%*23+A)%*&'++',3+<" " " " " " " >!>Z]" " " BC%*'+)+44!) ) C))@"N)%%$9"U" " " " >!G\=" " " " " " " " " " >!>?]" " D%$&0#(-1'++3+)1'(E3#;-*'%) A/%B/2/'"b%$$4$%"UAQK" >!G=[" " " " " " " >!>[=" D#$+'1%$'%)+%"'&&%&3,F)" R$'%,0*8&"" " " " >!G\G" " %$';-*'3," " " " " " " >!>=\" " "

!"#$%&'()"#%& & !"#$%&'%())) ) ) ) *+,-.) ) ) ) ) ) ) ) ) *+*//) ) ) !*+$,+"##%&011+& & !2345#6'2"))) ) ) ) *+/*.) *+/*.) ) ) ) ) ) ) ) ) *+*,7) *+*/8) !-(,$##%&.(#/%-"0& & 9#$":;"")) <=>) ) *+,-*) ) ) ) ) ) ) ) ) ) ) *+*/7) ) ) ) ) 12(++$##%-"%&011+& & ?@5""A$1)) ) ) ) *+/8B) ) ) ) ) ) ) ) *+*,,) ) & 1*#"3%/*&011+) ) C'"D#3%'D0)EFG) ) *+/**) & & & & & & *+*8/) 4.(#%-"%&'$-5*#"%+%& & H#%6'"2$(#)I#""&'%()) E=>) ) ) ) *+/7J) ) ) ) ) ) ) ) ) ) ) *+*,,) ) & 4.(#%-"%&011+& & I#""&'%(0) ) ) ) ) ) ) *+K-/) ) ) ) ) ) ) ) ) ) ) *+*/B) ) ) 6"*#%&011+) ) ) <2'"#(0)) ) ) ) ) *+K*,) *+7/8) ) ) ) ) ) ) ) ) *+**,) *+**,) !$-"##%&5-(+$02$,0& & I##60(#$@)H"$3()<=G) ) ) ) ) *+77/) *+7*-) ) ) ) ) ) ) ) ) ) *+*/,) *+**,) ) ) 7"2-*0+"/$()&.")",$()& ?(2"(L%$00))) ) ) ) *+KK*) *+7,8) ) ) ) ) ) ) ) ) *+**,) *+**,) 80'#$,"()&'#%+9,$(-*,& MN'3O)?1"##3&'%()) ) ) ) ) ) ) *+/8B) ) ) ) ) ) ) ) ) ) ) ) *+**K& !%-%+:$#9'+$-"0&,*.$;*-%2$,0"0&&PQ)=#%3)) )) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) <,*2#$%&0$,0";"#"0& & ?#302(2R#)=#%3)))))) ) ) ) *+/8J) ) ) ) ) ) ) ) ) *+**,) ) ) !*#90+"2:()&%2-*0+"2:*"3$0& !S%20(T$0)=#%3)) ) ) ) ) *+/8,) ) ) ) ) ) ) ) ) ) *+**-) ______)

256 Table 5.17

Herbaceous taxa used in indicator species analyses – 1m2 plots Taxa used in indicator analyses of herbaceous taxa present in 1m2 plots at turtle points and random points in Virginia and West Virginia during June-August 2013-2014. Taxa present in a state are indicated with a . Taxa with significant indicator values (by themselves, not in combination with other taxa) for a site group are marked with an X (alpha ≤ 0.05); o = an indicator for that site group (0.05 < alpha ≤ 0.084); – = not an indicator; denotations are in the order VA/WV. Site groups: FWT = female Wood Turtles, FRP = female random points, MWT = male Wood Turtles, MRP = male random points. Wetland class: FAC = facultative, FACU = facultative upland, FACW = facultative wetland, OBL = obligate wetland, UPL = upland. State Site Groups Wetland VA WV FWT FRP MWT MRP FWT+ FRP+ Taxon Common name class MWT MRP Ageratina altissima White Snake Root FACU X/- Alium cernum Nodding Wild Onion FACU Amphicarpaea bracteata Hog Peanut FAC X/o Antennaria spp. Pussytoes UPL -/o Arisaema triphyllum Jack-in-the-pulpit FACW Circaea canadensis Ench.’s Nightshade FACU X/- Convolvulus spp. Bindweed FACU Cunila origanoides Dittany UPL X/- Dioscorea villosa Wild Yam FAC Eurybia divaricata White Wood Aster FACU/ X/- UPL Galium circaezans Wild Licorice UPL

Galium pilosum Bedstraw UPL Galium triflorum Bedstraw FACU X/- -/X Gaultheria procumbens Teaberry FACU X/- Hieracium venosum Hawkwort UPL -/o Impatiens capensis Jewelwort FACW -/X X/- Mitchella repens Partridgeberry FACU Nabalus spp. Gall-of-the-Earth FACU (albus, altissimus) Oxalis spp. Wood Sorrel (stricta) FACU Packera obovata Round-leaved FACU Ragwort Parthenocissus Virginia Creeper FACU X/- quinquefolia Potentilla spp. Cinquefoil FACU X/- Prunella vulgaris Heal-All FACU -/o Solidago spp. Goldenrods Solidago caesia Blue-stemmed FACU X/- Goldenrod Uvularia spp. Bellworts (puberula) FACU X/- Viola spp. Violets - X/X Microstigeum vimineum Stiltgrass FAC X/- -/X

257 Table 5.18

Herbaceous taxa used in association analyses – 1m2 plots Taxa used in association analyses of herbaceous taxa present in 1m2 plots at turtle points and random points in Virginia and West Virginia during June-August 2013-2014. Taxa present in a state are marked with a . Taxa with significant Pearson’s phi coefficients of association (by themselves, not in combination with other taxa) for a site group are marked with an X (alpha ≤ 0.05); o = an indicator for that site group (0.05 < alpha ≤ 0.10); - = not an indicator; denotations in the site groups are in the order VA/WV. Site groups: FWT = female Wood Turtles, FRP = female random points, MWT = male Wood Turtles, MRP = male random points. Wetland class: FAC = facultative, FACU = facultative upland, FACW = facultative wetland, OBL = obligate wetland, UPL = upland. State Site Groups Wetland VA WV FWT FRP MWT MRP FWT+ FRP+ Taxon Common name class MWT MRP Ageratina altissima White Snake Root FACU X/- Alium cernum Nodding Wild FACU Onion Amphicarpaea Hog Peanut FAC -/o X/- bracteata Antennaria spp. Pussytoes UPL -/o Arisaema triphyllum Jack-in-the-pulpit FACW Circaea canadensis Ench.’s Nightshade FACU X/- Convolvulus spp. Bindweed FACU Cunila origanoides Dittany UPL -/o Dioscorea villosa Wild Yam FAC Eurybia divaricata White Wood Aster FACU/UPL X/- Galium circaezans Wild Licorice UPL Galium pilosum Bedstraw UPL Galium triflorum Bedstraw FACU X/- -/o Gaultheria Teaberry FACU X/- procumbens Hieracium venosum Hawkwort UPL -/o Impatiens capensis Jewelwort FACW X/- Mitchella repens Partridgeberry FACU Nabalus spp. Gall-of-the-Earth FACU (albus, altissimus) Oxalis spp. Wood Sorrel (stricta) FACU Packera obovata Round-leaved FACU Ragwort Parthenocissus Virginia Creeper FACU X/- quinquefolia Potentilla spp. Cinquefoil FACU X/X Prunella vulgaris Heal-All FACU -/o Solidago spp. Goldenrods na Solidago caesia Blue-stemmed FACU o/- Goldenrod Uvularia spp. Bellworts (puberula) FACU X/- Viola spp. Violets na X/o Microstigeum Stiltgrass FAC X/- -/X vimineum

258 Table 5.19

Woody seedling taxa used in indicator species analyses – 1m2 plots Taxa used in indicator analyses of woody seedling taxa present in 1m2 plots at turtle points and random points in Virginia and West Virginia during June-August 2013-2014. Taxa present in a state are marked with a . Taxa with significant indicator values (by themselves, not in combination with other taxa) for a site group are marked with an X (alpha ≤ 0.05); - = not an indicator; denotations under site groups are in the order VA/WV. Site groups: FWT = female Wood Turtles, FRP = female random points, MWT = male Wood Turtles, MRP = male random points. Wetland class: FAC = facultative, FACU = facultative upland, FACW = facultative wetland, OBL = obligate wetland, UPL = upland. State Site Groups Wetland VA WV FWT FRP MWT MRP FWT+ FRP+ Taxon Common name class MWT MRP Acer rubra Red Maple FAC X/- Amelanchier spp. Serviceberry FAC -/X Carya spp. (glabra, tomentosa) Hickories FACU Fraxinus americana White Ash FACU Hamamelis virginiana Witch Hazel FACU Hypericum prolificum Bush St. John’s Wort FACU Lindera benzoin Spicebush FAC X/- Liriodendron tulipifera Tulip Tree FACU Ostrya virginiana Ironwood FACU X/- Pinus strobus White Pine FACU Quercus alba White Oak FACU Quercus montana Chestnut Oak UPL X/X Quercus rubra Northern Red Oak FACU X/- Rhododendron Pinkster Azalea FAC periclymenoides Rubus spp. Blackberries na -/X X/- Smilax spp. Greenbriar FAC X/- Vaccinium spp. Blueberries FACU X/X

259 Table 5.20

Woody seedling taxa used in association analyses – 1m2 plots Taxa used in association analyses of woody seedling taxa present in 1m2 plots at turtle points and random points in Virginia and West Virginia during June-August 2013-2014. Taxa present in a state are marked with a . Taxa with significant Pearson’s phi coefficients of association (by themselves, not in combination with other taxa) for a site group are marked with an X (alpha ≤ 0.05); o = an indicator for that site group (0.05 < alpha ≤ 0.08); - = not an indicator; denotations under site groups are in the order VA/WV. Site groups: FWT = female Wood Turtles, FRP = female random points, MWT = male Wood Turtles, MRP = male random points. Wetland class: FAC = facultative, FACU = facultative upland, FACW = facultative wetland, OBL = obligate wetland, UPL = upland. State Site Groups Wetland VA WV FWT FRP MWT MRP FWT+ FRP+ Taxon Common name class MWT MRP Acer rubra Red Maple FAC X/- Amelanchier spp. Serviceberry FAC -/X Carya spp. Hickories FACU Fraxinus americana White Ash FACU Hamamelis virginiana Witch Hazel FACU Hypericum prolificum Bush St. John’s Wort FACU Lindera benzoin Spicebush FAC X/- Liriodendron tulipifera Tulip Tree FACU Ostrya virginiana Ironwood FACU X/- Pinus strobus White Pine FACU Quercus alba White Oak FACU Quercus montana Chestnut Oak UPL -/X X/- Quercus rubra Northern Red Oak FACU Rhododendron Pinkster Azalea FAC periclymenoides Rubus spp. Blackberries na -/X o/- Smilax spp. Greenbriar FAC X/- Vaccinium spp. Blueberries FACU X/-

260 Table 5.21

Herbaceous taxa used in association analyses – 400m2 plots and forest type groups Taxa used in association analyses of herbaceous taxa present in 400m2 plots at random points in Virginia and West Virginia during June-August 2013-2014. Taxa present in a state are marked with a . Taxa with significant Pearson’s phi coefficients of association (by themselves, not in combination with other taxa) for a site group are marked with an X (alpha ≤ 0.05); o = associated with that site group (0.05 < alpha ≤ 0.1); − = not associated; denotations are in the order VA/WV. Site groups: Dm = mesic deciduous, M = dry mixed pine and deciduous, Mm = mesic mixed pine and deciduous, Od = oligotrophic oak, Om = mesic oak, P = pine. State Site Groups Wetland VA WV Dm M Mm Od Om P Taxon Common name class Ageratina altissima White Snake Root FACU Agrimonia grypsophela Agrimony FACU X/- Alium cernum Nod. Wild Onion FACU Amphicarpaea Hog Peanut FAC o/- bracteata Anemone americana Rnd.-lobed Hepatica UPL -/X Antennaria spp. Pussytoes UPL Arisaema triphyllum Jack-in-the-pulpit FACW X/- Aster spp. Asters na Boehmeria cylindrica False Nettle FACW o/- Chimaphila maculata Spotted Wintergreen UPL o/ Cicuta maculata Water Hemlock OBL Circaea lutetiana Ench.’s Nightshade FACU Collinsonia canadensis Stone Root FAC o/- Cunila origanoides Dittany UPL Desmodium spp. Tick Trefoil UPL (glabellum) Dioscorea villosa Wild Yam FAC -/X Epigaea repens Trailing Arbutus UPL Eurybia divaricata White Wood Aster FACU/UPL o/- Galium triflorum Bedstraws FACU -/o Galium circaezans Wild Licorice UPL Gaultheria procumbens Teaberry FACU Geranium maculatum Wild Geranium FACU -/X Goodyera pubescens Rattlesnake Plantain FACU Hieracium paniculatum Panicled Hawkwort UPL X/- Hieracium venosum Hawkwort UPL Impatiens capensis Jewelwort FACW o/- Isotria verticillata Whorled Pogonia FACU Lespedeza spp. Bush Clover UPL (procumbens) Lycopus spp. Bugleworts OBL o/- Maianthemum Plume Lily FACU racemosa Medeola virginiana Ind. Cucumber Root FAC o/- Mitchella repens Partridgeberry FACU Nabalus spp. Gall-of-the-Earth FACU (albus, altissimus) Oxalis spp. Wood Sorrel FACU Packera obovata Rnd.-leaved Ragwort FACU Parthenocissus Virginia Creeper FACU quinquefolia Persicaria Tearthumb OBL X/- sagittatum/arifolium Pilea pumila Clearwort FACW X/- Polygonatum biflorum Solomon’s Seal FACU Potentilla spp. Cinquefoil FACU X/- Prunella vulgaris Heal-All FACU Scuttellaria spp. Skullcap (lateriflora) FACW Solidago spp. Goldenrods na o/- Uvularia perfoliata Perfoliate Bellwort FACU -/X Uvularia spp. Bellworts (puberula) FACU Veronica officinalis Speedwell FACU Viola spp. Violets na -/o Perilla frutescens Beefsteak Plant FACU Microstigeum Stiltgrass FAC o/- vimineum Asplenium platyneuron Ebony Spleenwort FACU Onoclea sensibilis Sensitive Fern FACW X/- Parathelypteris NY Fern FAC o/- noveboracensis Polystichum Christmas Fern FACU -/X arostichoides

261 CHAPTER 6: MULTI-SCALE HABITAT PREFERENCES OF WOOD TURTLES

(GLYPTEMYS INSCULPTA) IN CENTRAL APPALACHIAN FORESTS

Introduction

What controls who lives where and concomitant issues of scale are of fundamental theoretical and empirical importance for ecologists (Levin 1992,

Jackson et al. 2001). In a word, the focus is on habitat, an area with the resources and conditions that allow a particular species’ occupancy, survival, and reproduction (Hall et al. 1997, Mitchell and Powell 2003, Kearney 2006, Morrison et al. 2006, Beyer et al. 2010). All habitats are patchily distributed in geographical space (Aarts et al. 2008) and habitat selection emerges because organisms are better adapted to live and reproduce in some places than in others (Morris et al.

2008). Knowledge of habitat associations is critical for maintaining the spaces essential to organismal conservation.

Used habitat is any place occupied by the animal (Boyce et al. 2002).

Prudent choices are necessary to acquire adequate energy, obtain refuge from predators, and avoid environmental extremes. These necessities may increase or decrease the use of available habitat (Halstead et al. 2009), which is any habitat actually accessible by the animal (Beyer et al. 2010). The ease with which an individual can reach a point in geographic space is a complex function of behavioral and environmental factors that might constrain access (Garshelis 2002).

The comparison of used and available habitats allows us to identify habitats that are used non-randomly, i.e., disproportionate to their availability to the animal

262 (Aarts et al. 2008). Estimates of this non-random habitat preference are contingent upon and sensitive to the samples of used and available habitat (Beyer et al. 2010).

“Habitat selection” sensu stricto refers to the behavioral process of choosing to use habitat if it is encountered (Lele et al. 2013), while habitat preference is an attempt to quantify selection by a statistical description of habitat used relative to a particular sample of available habitat (Beyer et al. 2010). Such habitat suitability models assess the relationship between a species’ presence or abundance and a suite of ecological predictor variables (Bollman et al. 2005, McDonald et al. 2013).

The fundamental assumption of such models is that an organism’s distribution reflects its ecological requirements, it being present in suitable habitats and absent from unsuitable ones (Guissan and Zimmerman 2000, Hirzel and Le May 2008).

The objective of this study was to determine a small set of predictor variables that best explain summer habitat use by North American Wood Turtles

(Glyptemys insculpta) at two neighboring forested montane study sites in Virginia and West Virginia. The pertinent habitat attributes can be related to food availability, security from predation, thermo- and osmo-regulation, reproductive opportunities, or shelter from environmental extremes. Knowledge of the conditions underlying habitat preference can assist in identifying the factors influencing population distribution, abundance, or persistence. A better understanding of Wood Turtle habitat preference, informed by empirical data, model construction, and statistical analyses, will help focus conservation efforts, especially where commercial logging, recreational activities, road construction,

263 vehicular traffic, or other anthropogenic disturbances may occur (Gardner et al.

2007). The findings from this study could contribute to an understanding of proximate ecological consequences arising from alterations of Wood Turtle habitat by human activities. Because the resources and conditions that allow survival and reproduction of a particular taxon are often not distributed uniformly, I hypothesized that habitat quality at the study sites was spatially heterogeneous and predicted that a subset of the environmental variables measured in the field at used and available locations would be correlated with Wood Turtle occurrences and thereby indicate habitat preference.

Focal Species

Wood Turtles are amphibious emydid turtles of eastern deciduous, coniferous, and mixed forests (see “Focal Species” in Chapter 1 for more detail).

Wood Turtle foraging and ingestion occur in both terrestrial and aquatic settings, including underwater (Carroll 1999, Krichbaum pers. obs.). Wood Turtles forage on herbaceous plant leaves (e.g., Viola), fruits (e.g., Rubus), mushrooms (e.g., Boletus and Amanita), earthworms, slugs, crayfish, beetles, and millipedes (see, e.g., Strang

1983, Kaufmann 1992a, Niederberger and Seidel 1999, Ernst 2001a, Compton et al. 2002, Walde et al. 2003, Ernst and Lovich 2009, Jones 2009, Krichbaum pers. obs.). Turtle habitat use may be in response to fine-scale presence or abundances of litter invertebrates, fungi, or herbs that are distributed non-randomly in the forest

(Meier et al. 1995, Caldwell 1996, Hanula 1996, Hutchinson et al. 1999, Rubino and McCarthy 2003, Van de Poll 2004, Kappes 2006, Gilliam 2007).

264 Due to their long lives and exhibited philopatry (Kaufmann 1995, Ernst

2001a&b, Arvisais et al. 2002, Tuttle and Carroll 2003, Akre and Ernst 2006,

Willoughby et al. 2008, Parren 2013, Krichbaum this study), an adult Wood Turtle is presumably familiar with its activity area and the location of resources and clumped habitat patches therein, so its location at any time can be reasonably assumed to represent selection (McLellan 1986). A turtle’s movement patterns thus implicitly include biotic and abiotic interactions (Kearney 2006).

Methods

Study Area

I studied two adjacent populations of Wood Turtles at their southern range periphery in Virginia and West Virginia. See “Study Area” in Chapter 1, Fig. 1.1,

Tables 5.3-5.5, and Appendix 1 for details.

Field Procedures

Radio-telemetry and detailed sampling of turtle and random points at three spatial scales (stands, 400m2 and 1m2 plots) were used to evaluate habitat selection by Wood Turtles. See “Field Procedures” at Chapters 1 and 5 for details on turtles, radio-telemetry, and data collection. Detailed habitat variables lists and definitions are provided in Tables 6.1 and 6.2. In addition to the 1m2 plot centered on the individual’s location, in 2011-2012 I positioned four 1m2 plots at the perimeter of each 400m2 plot located at the four cardinal directions from the animal’s location.

Each turtle and random point was also located within and assigned to a specific “stand” as defined by the US Forest Service. For management and

265 inventory purposes the entire GWNF is divided into tracts that average ca. 10-20ha in size. Each of these delineated “stands” is consigned an age in years and specific

“forest type” based on canopy composition and when previous stand-replacing logging took place (see Tables 5.4 and 5.5) (USDA Forest Service undated). As grouping of related forest types improves classification accuracy (Baker 2015), I aggregated these thirteen forest types into six forest type groups: Dm = mesic deciduous (composed of FT 56), M = dry mixed pine and deciduous (FTs 10, 42,

45), Mm = mesic mixed pine and deciduous (FT 41), Od = oligotrophic oak (FTs

52, 59, 60), Om = sub-mesic oak (FTs 53, 54), and P = pine (FTs 3, 33, 39).

Depending upon their age, stands were aggregated into four different seral stages: esh = early successional habitat (0-35 years old), mid-suc. = mid-successional forest (36-75 years old), mature = forest at least 75 years old, but less than the lower age limit for designation as “old growth”, OG = old growth (minimum years of age [90-140] depends upon forest type – see USDA FS 1997). These forest composition (forest type) and structure (seral stage) categories at the stand level provided for a third scalar grain of spatial analysis.

Analytic Procedures

I used conditional logistic regressions to analyze the habitat data at the different plot scales. In this method variables at each individual turtle point are compared with those at the paired random point. Results from these tests were assessed through an information-theoretic approach.

266

Akaike’s information criterion (AICc), modified for small sample sizes relative to estimated parameters, was used to rank models, select the most

parsimonious models, and for averaging of habitat variable coefficients. AICc values

were rescaled to ∆i values for ease of interpretation and ranking, and Akaike

weights (wi) were calculated to give the approximate probability of each model being the best model in the set (Anderson and Burnham 2002). All models with

AICc values within two of the minimum value were considered well supported

(Burnham and Anderson 1998) and of the well-supported models the one with the smallest number of variables can be considered to be the best model (Arnold

2010).

Different analyses were run for each state and gender as well as by gender for both states combined. The acronyms used herein are: VA female Wood Turtles

= VFT, VA female random points = VFR, VA male Wood Turtles = VMT, VA male random points = VMR; WV female Wood Turtles = WFT, WV female random points = WFR, WV male Wood Turtles = WMT, WV male random points = WMR; female Wood Turtles = FWT, female random points = FRP, male Wood Turtles =

MWT, male random points = MRP.

Analyses were conducted with the R statistical program (R Development

Core Team 2015), specifically, the “Survival” and “mclogit” packages were used

for the conditional logistic regressions and the “AICcmodavg” package for AICc values and model averaged coefficients.

267 Microhabitat Preference

This was assessed using data from the 1m2 plots. Variables used in modeling were chosen based on their lack of correlation with each other; only the least correlated variables (Spearman or Pearson correlation coefficients ≤ 0.50) were retained for multivariate candidate models. Except for sand and rock for Virginia male turtles, none of the initial variables were strongly correlated so all were retained. Some variables had to be dropped from various model formulations due to lack of convergence in the regressions.

Fifteen candidate models were formulated with different combinations of inorganic, herbaceous living, non-herbaceous living, and organic non-living cover variables (Table 6.3). Some initial variables were dropped due to their miniscule

(<0.5%) mean overall amounts of coverage. In addition, as leaf litter cover was the matrix element for ground cover and co-varied directly with the proportion of cover provided by the other variables, it was not included in any models to avoid the interpretation of a forest dwelling turtle avoiding leaf litter. Inorganic variables for ground cover were sand/gravel and rock; herbaceous living variables included forb and grass; non-herbaceous living cover variables were woody vegetation and moss; organic non-living cover variables included coarse woody debris (cwd) and bare soil. Models were the same for the different analyses except that rock was retained and sand dropped from the model formulations for Virginia males.

268 Mesohabitat Preference

Variables from the 400m2 plots were retained for multivariate candidate models based on their lack of correlation (Spearman or Pearson correlation coefficients ≤ 0.50). None of the initial variables were strongly correlated so all were retained. Some variables were dropped from some model formulations due to lack of convergence in the regressions.

Twenty-two candidate models were formulated using different combinations of seventeen structural, compositional, and topographical variables obtained in

2011-2014 (Table 6.4). The ten structural and four compositional variables were each divided into two groups, encompassing attributes found at either overhead or ground levels. The structural overhead variables were number of large trees, number of medium trees, percent of horizontal vegetation obscurity at human-eye level, number of snags, percent of canopy openness, and stand age in years.

Structural ground variables included percent of horizontal obscurity at turtle level, area of plot under canopy gaps, and number of pieces of large woody debris (LWD) in two size classes. The four compositional variables were likewise divided. The compositional overhead variables were the number of tree taxa and number of shrub taxa and the compositional ground level variables were the number of seedling taxa and the number of herbaceous taxa. I used the herbaceous taxa metric only in separate modeling and analyses performed with 2011-2013 data.

The four topographical variables were distance to the main stream, slope inclination, slope aspect, and elevation. In addition, for analyses using pooled

269 males and females from both VA and WV, I formulated nine additional models using the “importance values” of tree taxa as variables (see models 13-17 and 28-

31 in Table 6.4); in these models I dropped number of large trees, number of medium trees, and number of tree taxa as variables. Importance values for individual taxa were calculated from the dbh and counts of every tree ≥ 10cm dbh in the 400m2 plots (see Chapter 5).

Macrohabitat Preference

Using the stand data obtained from the USFS and the aggregated (Virginia and

West Virginia turtles combined) male and female mesohabitat (400m2 plots) data sets, I ran mixed conditional logistic regressions with stand forest type as the random factor. Using my calculated importance values and US Forest Service descriptions and definitions for forest types (USDA FS undated), I designated a forest type for each 400m2 plot. Using the two aggregated (Virginia and West

Virginia turtles combined) male and female data sets again, I also ran mixed conditional logistic regressions with plot forest type as the random factor. For these mixed conditional logistic regressions involving forest types I used the 2011-2013 datasets that included the number of herbaceous taxa as a variable and used the same model formulations (without importance values) as with the other regressions of meso-habitat data.

In addition, I also ran mixed conditional logistic regressions for males and females using the Virginia 2011-2014 data with stand seral stage as the random factor. I did not do such a seral analysis for the West Virginia data due to the fact

270 that all the points there were either in mature or old growth stands. Forest seral stage is an indirect measure of some age-related environmental attributes (e.g., woody debris, canopy cover, litter depth, and cool, moist, equable microclimatic conditions) that contribute to determining whether a site is suitable for a given species (Welsh 1990, deMaynadier and Hunter 1995).

Study Rationale

In this study, all the random points (i.e., the posited available habitat) were accessible to the individual Wood Turtles for the following reasons. The random points were nearby the known location points of the various individuals; subsequent post hoc delineations of activity areas using minimum convex polygons showed the paired random points were within or nearby each individual’s estimated summer activity area (Chapter 1). The random points were within or nearby the general zone (viz., the area within 300 meters of the main streams) known to be of use by Wood Turtles. Topographic barriers were not apparent, nor were obvious environmental factors that could have served to filter turtles out of forest communities at these lower elevations in the mountains. All the random points were in a proximity (within ca. 300m of a turtle point) that could easily be reached by a turtle, i.e., healthy animals were physically capable of reaching them.

Wood Turtles are mobile at small spatial scales, with mean daily straight-line distances moved during the summer of ca. 28.3m (computed based on the number of days separating the greatest straight-line distances between location points in 59 radio-tracked individuals’ activity areas). They can move at a speed of at least

271 0.32km/hr (Woods 1945) or 1.1km/hr over short distances (Swanson 1940). In

2010 I radio-tracked an adult female in VA who traversed at least 1km in ca. 26 hours (Wood Turtles are typically diurnal so it is doubtful she was walking at night).

The greatest known distance moved in a day by an individual during this study was ca. 520m (straight-line distance between location points) by an adult male in

Virginia.

This study is of a “site attribute” design which compares ecological attributes at sites actually used by animals to those at random locations (the proxy for unused sites) (Garshelis 2002). The dependent variable is simply whether a site is used or unused. The fine-scale problem arises, however, when habitat differences might not be detected if the random sites (available habitat) are too similar to the used sites. My study is also a “design III” protocol where availability was sampled for each individual (Manly et al. 2002).

The chosen variables were hypothesized to be of biological importance to the animal, with a focus on resources and conditions available at a finer-scale, not coarse-grained land cover. There can be a gross mismatch between the fine-scale habitat patch chosen by an animal and the coarse-scale mapped vegetation cover types within which these smaller sites are embedded (e.g., a patch of mesic deciduous forest within a larger mapped stand of more xeric pine or a canopy gap in an area of forest). Because of the short duration of the overall field study period

(3.3 years), the restricted seasonality for the field work (summer only), and the long- term successional nature of these forest ecosystems (Shugart and West 1981),

272 overall habitat quality and conditions were presumed to be somewhat constant at the study sites.

Used-available data do not give probabilities of use, rather, the results are proportional to these; the density of observations is modeled instead of a probability. A typical use-availability design compares number of locations (or time spent) in each habitat type to the relative amount or area of each type (a proportional use study). But frequently used habitats are certainly not selected against even if they are widely available. Likewise, infrequent use is not necessarily indicative of lack of suitability. Therefore, it is unlikely that selection can be accurately assessed “just by comparing relative use to the relative area of different habitats” (Garshelis 2002 at pg. 131). Instead, one must look at differential use, rather than use in terms of sheer availability. Observed actual use, such as is used in a site attribute design, is a stronger indicator of habitat selection than inference based on relative use (Garshelis 2002).

In this study sites classified as “used” were undoubtedly actually used; i.e., the Wood Turtles were definitely found at those points. Of course, the mere occurrence in a habitat, especially of mobile organisms, does not necessarily indicate strong ecological links to that habitat (Lövei et al. 2006). Further, there exists the issue of contaminated controls, wherein some random/available points may actually be used points. The relative proportions of such used and unused points are unknown (Beyer et al. 2010). Although some sites might eventually be used, a representative sample that does not show presence is still an unbiased

273 sample of use and non-use (Boyce et al. 2002). In this study, unless the Turtles were very well hidden, the random points were not being used by Wood Turtles when habitat metrics were obtained in the field; a thread trail revealed that a gravid female had passed through one random plot.

In use-availability and site attribute designs, the exponential logistic model is often used and fitted using logistic regression to evaluate the relative probabilities of use (by using the maximum-likelihood values of the model coefficients) (Beyer et al. 2010). Available points are not true zeroes, but numerical tricks to approximate the integral of the likelihood function. It must be remembered that conditional logistic regression coefficients are interpreted as relative differences in the habitat and not as absolute values (Compton et al. 2002, Row and Blouin–Demers 2006); i.e., extreme negative values for certain habitat variables does not mean that turtles necessarily avoid such habitats. The conditional logistic regression approach is purely descriptive, it cannot extract causality. In other words, it is correlative; only an experimental approach can test the existence of causal links (Connell 1983).

Hence, this method does not examine the fundamental niche, but only the realized niche within a specific geographic area.

If model coefficients indicate non-preference (i.e., their confidence intervals include zero), one should not automatically infer that a habitat is unimportant: at some level of availability, preference is expected to be zero. A habitat attribute may be selected, but because it is widely available at the study site then preference is not reflected. Such an apparent lack of preference may be a phenomenon of scale.

274 Preference can change with scale because the relative availability of vegetation types or other environmental attributes changes across those scales (Beyer et al.

2010). Perhaps a preference for a certain attribute would become manifest if a broader extent of study was used, such as at the scale of Johnson’s (1980) first or second order selection. That is to say, one of the reasons a population of organisms is found at a certain finer-scale locale may be because of the widespread availability there of particular habitat attributes. The coefficient in a logistic regression model decreases as availability increases and can even change sign, i.e., changes in the spatial scale of availability can lead to drastic changes in perceived preference (Mysterud and Ims 1998, Beyer et al. 2010). The coefficient for a habitat used a lot can even be negative if that habitat is common, and, conversely, low levels of use can have positive coefficients if the used habitat is rare. Despite such anomalous results, even if functional responses exist, resource selection function coefficients for a particular locality can still be modeled (Boyce et al. 2002).

Results

Preference for Microhabitat Features

There were salient differences between proportions of ground cover in the

1m2 plots at turtle points and random points (Fig. 6.1). Turtles in both states and of both sexes consistently were found in microhabitats with higher amounts of coverage by forbs, grass, and coarse woody debris compared to paired random locations (see Table 6.5 for summary statistics of the metrics in 1m2 plots). VA

275 females and WV males also preferred higher woody vegetation coverage, whereas

WV females preferred lower amounts of coverage.

For both females and males in Virginia, two regression models were well supported (∆AICc < 2), and contained similar combinations of cover variables

(Tables 6.7 & 6.8, Fig. 6.2). Coefficients of the averaged models suggest that VA male Wood Turtles preferred microhabitats similar to those of females, with the addition of lower amounts of bare soil.

For West Virginia females, only one model, with four variables, was well supported. In addition to the variables stated previously, WV females preferred microhabitats with higher amounts of bare soil cover compared to paired random locations (Table 6.8, Fig. 6.2). West Virginia males’ preference was more complex, six competing models had similar support and contained various combinations of six cover variables (Table 6.7).

In both states in 2013-2014, 1m2 turtle plots had greater herbaceous taxa richness than did random plots, except for WV females (Tables 6.5 and 5.7, Fig.

5.15). In both states in 2011-2014, 1m2 turtle plots had significantly more herbaceous cover (forbs and grass combined) than did random plots (Tables 6.5 and 5.7, Fig. 5.20): VA paired Wilcoxon signed rank test: V = 9552, p < 0.00001;

WV Wilcoxon signed rank test: V = 4418, p < 0.0001. Mean herbaceous coverage:

VA turtle plots 14.4% ± 1.7, VA RPs 2.5% ± 0.6; WV turtle plots 15.0% ± 1.6, WV

RPs 8.1% ± 1.2.

276 In both states in 2011-2012, the proportional amount of herbaceous cover at

1m2 plots at the center of the 400m2 plots did not differ from that at the four 1m2 plots at the perimeter (using the mean value of the four), for either turtle points or random points (Table 6.5, Fig. 6.3) – VA female turtles paired Wilcoxon signed rank test: V = 708, p = 0.2222, VA female random points paired Wilcoxon signed rank test: V = 504, p = 0.528, VA male turtles: V = 70, p = 0.609, VA male random points: V = 15, p = 0.119; WV female turtles paired Wilcoxon signed rank test: V =

125, p = 0.974, WV female random points paired Wilcoxon signed rank test: V =

88, p = 0.948, WV male turtles: V = 141, p = 0.797, WV male random points: V =

135, p = 0.509.

In both states in 2011-2012, the five 1m2 plots at 400m2 turtle plots had significantly more herbaceous cover (forbs and grass combined) than did the five

1m2 plots at 400m2 random plots (using the mean of all five for comparisons): VA paired Wilcoxon signed rank test: V = 2184, p < 0.00001; WV Wilcoxon signed rank test: V = 860, p < 0.0005. Mean herbaceous coverage of the five plots: VA turtle points 13.7% ± 2.1, VA random points 2.3% ± 0.5; WV turtle points 11.2% ±

1.3, WV RPs 5.5% ± 1.2.

Preference for Mesohabitat Features

Female Wood Turtles in both states consistently preferred mesohabitats with higher levels of turtle-level obscurity, the larger size class of LWD, medium size trees, snags, canopy openness, and shrub taxa, along with gentler slopes and lower elevations, compared to paired random locations (Fig. 6.4). Females in WV also

277 preferred mesohabitats with greater herbaceous richness (see Table 6.6 for summary statistics of the metrics in 400m2 plots). Whereas VA females preferred warmer aspects, WV females preferred cooler aspects. Habitat preferences by male

Wood Turtles were not as obvious and in some cases were contradictory to those for females. Virginia males showed marginal preference for mesohabitats with more canopy openness and fewer herbaceous taxa, compared to random sites. West

Virginia males showed some preference for higher amounts of turtle-level obscurity, lower amounts of snags, and gentler slopes compared to paired random locations. See Table 6.17 for a synopsis of important mesohabitat variables for the different turtle groups.

In addition to counts of pieces of LWD, I measured the distance of individual turtles to the closest piece of LWD10 when they were located and did the same for the random points. The mean distance for 320 turtle points was 5.5m

(se = 0.29m), while that for the 320 paired random points was 9.7m (se = 0.36m).

A paired Wilcoxon signed rank test found turtle points to be significantly closer to

LWD than were random points (p < 0.000001, V = 40203).

Virginia

For Virginia females, only one model, with ten structural, compositional, and topographic variables, was well supported (Table 6.9). Coefficients of the averaged models suggest that female Wood Turtles in Virginia prefer mesohabitats that have higher amounts of turtle-level obscurity, the large size class of LWD, medium size trees, snags, canopy openness, and shrub taxa, along with fewer tree

278 taxa, gentler slopes, warmer aspects, and lower elevations, compared to paired random locations (Table 6.10, Fig. 6.5).

When I ran the same models for Virginia females with data from 2011-2013 that included the number of herbaceous taxa in a plot, the only well supported model was the same model formulation as was the top model for the 2011-2014 data (Table 6.11); moreover, the top five models were the same as resulted from the

2011-2014 data. Important coefficients of the averaged models for 2011-2013 were the same as the results from the 2011-2014 data, except without the larger size class of LWD as a factor (Table 6.12, Fig. 6.6). The number of herbaceous taxa was also not a factor in any of the top models, nor did it show up as an important coefficient from model averaging.

Male Wood Turtles in VA were not as clear as females in their preferences for mesohabitat conditions relative to random sites (Table 6.10, Fig. 6.5). Four competing models had similar support (∆AICc< 2) and contained combinations of

8-10 habitat variables (Table 6.9).

When I ran models for Virginia males with data from 2011-2013 that included the number of herbaceous taxa in a plot, only one model, with four structural variables, was well supported (Table 6.11). Coefficients of the averaged models suggest that male Wood Turtles in Virginia prefer mesohabitats with more canopy openness and fewer woody seedling and herbaceous taxa, compared to random sites (Table 6.12, Fig. 6.6). The number of herbaceous taxa was not a factor in the top model.

279 West Virginia

Female Wood Turtles in West Virginia prefer mesohabitats that have higher values for the larger size class of LWD, medium-size trees, snags, turtle-level and eye-level obscurity, and shrub taxa, with cooler aspects, lower elevations, and locations closer to the main stream when compared to paired random locations

(Table 6.10, Fig. 6.5). Only one model, with twelve structural, compositional, and topographic variables, was well supported (Table 6.9). Female Wood Turtles in

West Virginia preferred mesohabitats that have higher amounts of herbaceous taxa and snags compared to random sites (Table 6.12, Fig. 6.6) when the same models were run with data from 2011-2013 that included the number of herbaceous taxa in a plot (Table 6.11).

Male Wood Turtles in West Virginia prefer mesohabitats with fewer numbers of large trees and snags, and gentler slopes, compared to paired random locations (Table 6.10, Fig. 6.5). Only one model, with eight structural, compositional, and topographic variables, was well supported (Table 6.9). Male

Wood Turtles in WV preferred mesohabitats that have higher amounts of turtle- level obscurity and fewer numbers of snags compared to paired random locations

(Table 6.12, Fig. 6.6) when the same models were run with data from 2011-2013 that included the number of herbaceous taxa in a plot (Table 6.11).

Pooled VA and WV Turtles

Both female and male Wood Turtles consistently preferred mesohabitats with higher values for canopy openness and turtle-level obscurity, along with

280 gentler slopes, compared to paired random locations. Otherwise, there was not much intersexual overlap in preferred habitat conditions. In addition, females preferred sites with more medium-size trees, while males preferred fewer. There was no intersexual overlap in preference for importance values of overstory tree taxa (Table 5.3). Moreover, females preferred sites with higher values for Red

Maple (Acer rubra) and Sugar Maple (Acer saccharum), while males preferred lower values. The number of herbaceous taxa was a marginally important variable for female turtles, but not males.

Females. Female Wood Turtles prefer mesohabitats that have higher levels of canopy openness, the larger size class of LWD, medium-size trees, snags, turtle- level and eye-level horizontal obscurity, and shrub taxa, along with fewer tree taxa, gentler slopes, and warmer aspects, compared to paired random locations (Table

6.14, Fig. 6.7). The models with importance values indicate that females preferred sites with higher values for White Oak (Quercus alba), Red Maple, Sugar Maple, and Tulip Tree (Liriodendron tulipifera), than random sites (Fig. 6.8). For Virginia and West Virginia female turtles combined, only one model, with nine structural, compositional, and topographic variables, was well supported (Table 6.13).

Important coefficients of the averaged models for females were similar to the results from the 2011-2014 data (Table 6.16, Fig. 6.7) when the same models for were run with data from 2011-2013 that included the number of herbaceous taxa in a plot. Only one model, with nine structural, compositional, and topographic variables, was well supported (Table 6.15); this was the same model formulation as

281 was the top model for the 2011-2014 data. The number of herbaceous taxa was not a factor in the top model, though it did show up as a marginally important coefficient from model averaging. The models with importance values indicate that females prefer sites with higher values for White Oak, Red Maple, Sugar Maple, and White Ash (Fraxinus americana), compared to random sites (Fig. 6.8).

Males. Male Wood Turtles prefer mesohabitats that have higher amounts of turtle-level horizontal obscurity, canopy openness, and large-size trees, with fewer medium-size trees, gentler slopes, and closer to the main streams, compared to paired random locations (Table 6.14, Fig. 6.7). The models with importance values indicate that males prefer sites with lower values for Chestnut Oak (Quercus montana), Virginia Pine (Pinus virginiana), White Pine (Pinus strobus), Sugar Maple, and Tulip Tree, compared to random sites (Fig. 6.8). For combined VA and WV males, only one model, with thirteen structural and topographic variables, was well supported (Table 6.13).

Male Wood Turtles preferred mesohabitats that have higher amounts of turtle-level horizontal obscurity, canopy openness, large-size trees, and tree taxa, with gentler slopes, compared to paired random locations (Table 6.16, Fig. 6.7) when the same models were run with data from 2011-2013 that included the number of herbaceous taxa in a plot. The models with importance values indicate that males prefer sites with higher values for Black Gum (Nyassa sylvatica), and lower values for White Pine, Red and Sugar Maple, and Hickories (Carya spp.), than random sites (Fig. 6.8). Only one model, with seven structural and

282 topographic variables, was well supported (Table 6.15). The number of herbaceous taxa was not a factor in the top model, nor did it show up as an important coefficient from model averaging.

Preference for Mesohabitat Features Using Random Effect Models

Using data from 2011-2013 that included number of herbaceous taxa, neither forest type of stands, forest type of 400m2 plots, nor stand seral stage was a significant factor in well-supported regression models wherein it was used as a random effect, except for stand forest type for female turtles. Significant coefficients in the best models were consistent with those obtained from averaging of conditional logistic regression models without random effects; e.g., preference for mesohabitats that have higher amounts of turtle-level horizontal obscurity, the larger size class of LWD, and canopy openness or area under gaps, with gentler slopes and warmer aspects, compared to paired random locations.

Seral stage of stands in VA did not arise as a significant factor in any of the regressions wherein it was used as a random effect. For VA male Wood Turtles, only one model, with seven structural, compositional, and topographic variables, was well supported (Table 6.18). Model coefficients with significant p-values suggest that male turtles preferred mesohabitats with more canopy openness and gentler slopes compared to random locations (Table 6.19). For VA female turtles, two models, each with ten structural, compositional, or topographic variables, were well supported (Table 6.18). Model coefficients with significant p-values suggest that female Wood Turtles preferred mesohabitats with more turtle-level

283 horizontal obscurity, snags, shrub taxa, and medium-sized trees, with gentler slopes, compared to random locations (Table 6.19).

Of the 22 models using stand forest type as the random effect for Virginia and West Virginia female turtles combined, two models had similar support

(∆AICc< 2) (Table 6.18). The top model contained six structural and topographic variables. Model coefficients with significant p-values suggest that female Wood

Turtles preferred mesohabitats that had higher amounts of turtle-level horizontal obscurity and area under canopy gaps, with gentler slopes and warmer aspects

(with higher amounts of LWD10 being of marginal significance), compared to paired random locations (Table 6.19). Stand forest type was a significant factor in the top model.

For Virginia and West Virginia male turtles combined, of the 22 models using stand forest type as the random effect only one was well-supported (Table

6.18). This model had 8 structural, compositional, and topographic variables variables, including many of those in the top models for the females. Model coefficients with significant p-values suggest that male Wood Turtles prefer mesohabitats that have higher amounts of turtle-level horizontal obscurity (with distance from the main stream being of marginal significance), compared to paired random locations (Table 6.19). Stand forest type did not arise as a significant factor in any of the male turtle regressions.

Forest type of the 400m2 plots did not arise as a significant factor in any of the regressions wherein it was used as a random effect. Of the 22 models for

284 Virginia and West Virginia female turtles combined, two models were well supported (Table 6.18). The best model had six structural and topographic variables and was the same formulation as the top model when forest type of stands was used as the random effect. Model coefficients with significant p-values suggest that female Wood Turtles prefer mesohabitats that have higher amounts of turtle-level horizontal obscurity along with gentler slopes and warmer aspects (with higher amounts of LWD10 and area under gaps being of marginal significance), compared to paired random locations (Table 6.19). For pooled Virginia and West Virginia male turtles only one model, with five structural and topographic variables, was well supported (Table 6.18). Model coefficients with significant p-values suggest that male Wood Turtles prefer mesohabitats that have higher amounts of turtle-level horizontal obscurity and warmer aspects, compared to paired random locations

(Table 6.19).

Discussion

Tradeoffs in resource allocation to various compartments of their energy budget, result in habitat-mediated choices (manifest as habitat preferences) that ultimately affect reproductive output and fitness (Congdon 1989, Huey 1991,

Penick et al. 2002). Structural, compositional, and topographical characteristics of forest habitat can impact Wood Turtle fitness either directly or indirectly through foraging success, predator vulnerability or avoidance, and thermo- and osmo- regulatory options. I identify Wood Turtle habitat preference through a subset of multi-scalar environmental variables at the southern periphery of their range.

285 In the 1m2 plots female and male Wood Turtles in both VA and WV exhibited the same preferences for microhabitats with greater forb, grass, CWD, and woody ground cover than available at random points. This comports with the well documented Wood Turtle omnivory as well as their apparent preference for more dense ground vegetation in the 400m2 plots (see “Turtle-level obscurity”,

“LWD”, and “Herbaceous richness and cover” subsections below), perhaps using this to avoid predation or high temperatures. Wood Turtles seem more risk averse than syntopic Box Turtles (Terrapene carolina): Except for gravid females during the nesting season, I never observed Wood Turtles on roads or roadsides (in the ten years of observing turtles at this study area), whereas I observed Box Turtles sitting on closed roads out in the open many times. Presumably the adaptive advantage of having a closable shell makes Box Turtles less averse to being in open habitats where detection by predators is more likely. Costa (2014) found differences in flight distance between two species of Emydid turtles and Lopez et al. (2005) found escape decisions by the Mediterranean Terrapin (Mauremys leprosa) to be influenced by habitat-related visibility.

In contrast to the 1m2 plots, there was limited congruence between turtle groups (VFT, VMT, WFT, and WMT) as to the habitat variables that were preferred or avoided at the 400m2 plots (Table 6.17, Figs. 6.5 & 6.6). Whereas several variables were consistently significant in various model formulations female turtles, the model results for males were more ambiguous, suggesting that males did not select the sampled meso-habitat conditions as strongly as did the females (Tables

286 6.10-6.12, 6.14, 6.16, 6.19). All these results may be at least partially explained by the fact that male turtles generally did not disperse into the forest as far as females, i.e., they were located closer to the main streams. The habitats available there may be less variable than the habitats used by the more widely ranging females (e.g., relatively less variation in slope, aspect, soil moisture, or forest types). If this is the case, then perhaps it is not that the males are less selective, it is just that preferred habitats are relatively more available for them, i.e., random points are analytically indistinguishable from the turtle points.

Due to differences in life history characteristics, habitat preferences of animals at the periphery of their geographic ranges can be expected to differ from those at the core (Kapfer et al. 2008). For example, though sites that are too open in

the region of this study can be potentially problematic with regard to CTmax for

Wood Turtles in the summer (Chapter 3), such risk changes with the season, geographic location, and behavior. Habitat of limited suitability for use by adults in the summer due to excessive temperatures, such as roadsides or anthropogenic openings, may nevertheless provide valuable nesting sites earlier in the year

(Krichbaum pers. obs.). In addition, sites detrimental in the south or summer may be more suitable in the north or spring and vice-versa.

Habitat Variables

Canopy Openness

Amount of canopy cover was a key factor in preferred habitats for these

Central Appalachian Wood Turtles; Michigan Wood Turtles also showed such an

287 affinity (Remsberg et al. 2006). Broken canopies and gaps provide space for the development of internal habitat patchiness and edges (Noss 1991) and allow enough light and warmth for the ground floor herbaceous layer and associated insect communities (trophic resources) (Jennings et al. 1999, Bollman et al. 2005).

Gaps are important for sustaining herbal growth, richness, and persistence

(Goldblum 1997, Anderson and Leopold 2002). Thick herbaceous growth also provides cover from predators (Bollman et al. 2005), thus may enhance fitness

(survival). Gap formation from a large tree falling also supplies LWD with the below-described benefits.

Wood Turtles, however, did not prefer large canopy openings such as logging cuts, roads, or roadsides. On those few occasions when VA turtles were located in early successional habitat at recent large logging cuts, with one exception (a female eating blackberries) they were at the very edge of the cutting unit (within ca. 10-15m of the surrounding uncut forest). Wood Turtles did use small fabricated “wildlife openings” with grassy and herbaceous ground floors and shrubby overstories (e.g., Rubus spp.); unlike regenerating logging openings, these did not have high densities of saplings. These types of habitats with small numbers of trees are typically lumped into the early successional category. However, they are structurally and often compositionally different than young sites of regenerating forest with high stem densities of saplings. Managerial mishaps are possible when relevant distinctions of pattern go unrecognized; i.e., fabricating large tracts of esh

288 with dense numbers of saplings, when what would actually be beneficial are small tracts dominated by herbs/grass/shrubs.

For ectotherms, thermoregulatory or energetic requirements may be a primary driver of habitat selection (Lagory et al. 2009, Kapfer et al. 2010); thus, canopy closure may serve as a coarse surrogate metric of opportunities for thermoregulation (Pringle et al. 2003). Utilization of open areas, forest edges, or other habitats could be correlated with thermoregulatory behaviors to maximize physiological states (Row and Blouin-Demers 2006, Halstead et al. 2009). Sites with high degree of canopy openness, such as roadsides or recently logged sites, have the highest temperatures and greatest temperature variance (Collins et al.

1985, Currylow et al. 2012, Chapter 3), while sites with very closed canopies, such as a regenerating clearcut with high stem density, offer the least thermal diversity

(Chapter 3). Canopy gaps along with their associated LWD and other ground cover provide basking opportunities for turtles where sunlight reaches the ground (Dodd

2001, Krichbaum pers. obs.), as well as thermal refugia that allow escape from high midday temperatures, thus allowing for thermoregulatory precision via fine-scale shuttling.

In Appalachian forests 2.5-9.6% of the canopy may be in some stage of recovery from natural gap formation (Boerner 2006). Gaps in old-growth Tennessee mesic deciduous forest tended to be larger, more variable in area, and have greater variation in microclimates than the undisturbed forest floor (Clebsch and Busing

1989). Perhaps due to cooler, moister microclimates, the abundance of the largest

289 macroarthropod size class was similar in mature closed-canopy controls and unlogged natural gaps (Greenberg and Forrest 2003); hence, intensive cutting could result in declines of ground-occurring macroarthropods (Greenberg and Forrest

2003) that are preferred food of Wood Turtles.

Turtle- and Eye-level Obscurity

Horizontal obscurity can arise from multiple sources that are not mutually exclusive and can be additive: herbaceous vegetation, as well as woody vegetation, alive and dead (i.e., woody debris). The obscurity board can be conceived as a surrogate metric of exposure to predation; a greater degree of shrub, herbaceous, or grass cover may reduce the risk of predation by taxa that rely on visual cues when hunting. Aside from offering cover, obscurity can be associated with foraging opportunities, either directly (such as consumption of forbs) or indirectly (such as invertebrates associated with herbaceous understories or woody debris). The area beneath low-lying vegetation may provide distinct microclimates with the lower ambient temperatures and increased soil moisture and relative humidity that Wood

Turtles prefer (Chapters 3 & 4), conditions that might also favor terrestrial invertebrate populations (Trainor et al. 2007).

Eye-level obscurity was not nearly as important a variable as was turtle- level, though it was weakly important for female turtles (Fig. 6.8, Table 6.14). Eye- level and turtle-level horizontal obscurity were only weakly correlated. The potential mismatch between the perception of cover by turtles and humans (Bowne

2008) is important to consider when characterizing a forested site. The term “open

290 habitat” is often employed, but a location that has the appearance of openness to an upright human observer can be densely vegetated at a turtle’s level, and vice- versa. It is essential to discern and communicate these distinctions when making recommendations for management practices that have the potential to alter site structural conditions.

Slope Inclination

It is important to recognize that the avoidance of steeper slopes was not an absolute result, but relative to the site. In both states slope inclination at random sites was around 50% more that at turtle locations; however, the average slope inclination of turtle locations in WV was around 50% more that at random locations in VA (Table 6.6). Turtles were sometimes located on very steep sites, up to 31° and 33° in VA and WV respectively. Not only could steeper slopes be more metabolically expensive to traverse, there could be a physical limit to a turtle’s ability to use steep slopes; Box Turtles (T. carolina and T. ornata) had trouble maintaining their position on slopes greater than 40° (Muegel and Claussen 1994,

Claussen et al. 2002). Though Wood Turtles are larger, have stronger limbs (Abdala et al. 2008), and are more adept at climbing than Box Turtles (Pope 1939,

Krichbaum pers. obs.), there could still be metabolic, physical, and behavioral constraints upon their use of steep slopes.

Aspect

Because Wood Turtles are one of the more northerly distributed chelonians and are considered to be a cold-adapted species (Stephens and Wiens 2009), I

291 expected them to exhibit a clear preference for north aspect sites, which are generally cooler and of higher humidity (Cantlon 1953, Smith and Smith 2002).

This was not the case. When the states were examined separately, the logistic regression results suggest that Wood Turtle males did not prefer a specific aspect, as was the case for Box Turtles in WV (Weiss 2009). Males preferred plots with warmer aspects when plot forest type was a factor. And though both males and females used slopes of all aspects, ranging from the warmest (SW orientations with a value of 0) to the coolest (NE orientations with a value of 2), VA females preferred relatively warmer aspects, while WV females preferred relatively cooler aspects

(Table 6.6). Thermoregulatory patterns may have differed because temperatures were lower in the higher elevation VA site than in WV (Chapter 3).

Overall diurnal surface temperatures of microhabitats in WV were ca. 1.2°C higher than those in VA, a not unexpected result given that elevations in VA were on average 180m higher than those in WV (Chapter 3). Based on temperatures of turtles’ carapaces in July and August of 2014, females in VA maintained slightly higher diurnal temperatures than males (22.8°C vs. 21.3°C) (Chapter 3), perhaps due to the differential energetics of reproduction. Due to aspect, height of canopy vegetation, and the degree of slope inclination, some ground-level sites never receive direct solar radiation. So, to bask in direct sunlight a turtle (such as VA females) would have to select certain aspects. In addition, as slopes were generally steeper in WV than VA, warmer aspects in WV can be expected to be warmer than similar aspects in VA (due to greater angle of incidence). Moreover, aspect can

292 influence plant assemblages. In Ohio, herb species richness was higher on south aspect slopes, but density was greater on north aspect slopes (Small and McCarthy

2003) and in a Kentucky deciduous forest, north facing slopes had higher productivity (McEwan and Muller 2011). All these metabolic, ecological, and physiographic differences may contribute to the differences between sexes and states in the use of slope aspects.

Woody Debris

Wood Turtles had a propensity for associating with sites with relatively higher abundance of the larger size-class (diameter ≥ 25cm) of woody debris

(LWD10). Turtles also showed a preference at the microhabitat scale (1m2 plots) for sites with relatively higher amounts of coarse woody debris. Furthermore, turtles were also located closer to LWD than were random points. These three pieces of evidence are strong indication of the Turtles’ preference for associating with woody debris.

Many reptile and amphibian and other vertebrate taxa show a close association with woody debris (Whiles and Grubaugh 1996). For instance, Trainor and colleagues (2007) found a positive relationship between woody debris amounts and jumping mouse abundance and survival. Coarse woody debris can be an important substrate for herbaceous plants (Roberts 2004) on which turtles might forage. Woody debris also provides substrate and refuge for macroinvertebrates that are important food items (Harmon et al. 1986, Caldwell 1996) and for small vertebrates can provide cover from predation (Everett and Ruiz 1993, Manning and

293 Edge 2004). Predators can affect the choice of habitat by prey species (Power et al.

1985, Ripple and Beschta 2004, Willems and Hill 2009) and such alteration of habitat use can have life history and fitness consequences (Huffaker 1958, Jackson et al. 2001). Spaces under large woody debris can also provide thermal and hydric refugia with moister and cooler conditions that can be more favorable than those available on nearby ambient soil and litter (Rittenhouse et al. 2008, Krichbaum unpub. data). For instance, at 11:56 on Aug. 8, 2011 the surface temperature where an adult male WV Wood Turtle was sitting in the sun was 35.2°C, while the temperature in the space under an upraised log ca. 0.5m away was 27.2°C (Fig.

6.9).

Various mushroom species are important elements of the Turtle’s diet

(Strang 1983, Kaufmann 1992a, Tuttle 1996, Compton et al. 2002, Walde et al

2003, Ernst and Lovich 2009, Jones 2009, Krichbaum pers. obs.). Macrofungal and myxomycete fungi richness was positively correlated with log size and amounts of coarse woody debris at old age oak and mixed mesic forest study sites in Ohio

(Rubino and McCarthy 2003). Similarly, in New Hampshire all sites with above average CWD cover had above average numbers of species of macro-fungi, with mean mushroom diversity in old growth sites being 2.5 times the amount in non- old growth sites (Van de Poll 2004). Box Turtles (T. carolina) can be important dispersal vectors of fungal spores (Jones et al. 2007) and Wood Turtles, being facultative mycovores, can be inferred to supply similar ecological function (viz., endozoospory and fecal facilitation of fungi reproductive success).

294 Wood Turtles also relish slugs, snails, and earthworms (Ernst and Lovich

2009, Jones 2009, Krichbaum pers. obs.). Slug densities and land snails are positively correlated with the presence of coarse woody debris (Caldwell 1996,

Kappes 2006). After intensive logging (which removes trees that would become large dead trees) it can take many decades for loadings of large woody debris to recover on sites (Hedman et al. 1996, McMinn and Hardt 1996, Webster and

Jenkins 2005, Keeton et al. 2007). Snail assemblages and densities are also positively correlated with litter composition and depth (Martin and Sommer 2004).

Other invertebrates, such as beetles, millipedes, and earthworms that Wood Turtles are known to eat (Krichbaum pers. obs.), are associated with forest floor litter or

LWD (Caldwell 1996, Hanula 1996, Hendrix 1996, Ulyshen and Hanula 2009).

Snags

Standing dead trees were perhaps affiliated with macro-invertebrates or saprophytic fungi upon which the turtles feed. Of course, snags also can be associated with both LWD on the ground and canopy openness, although statistical tests indicated that these conditions were not strongly correlated at these study sites. Though female turtles in both VA and WV exhibited a preference for sites with relatively greater numbers of snags, male turtles in WV preferred fewer snags relative to random points. Though the situation regarding females has discernible underlying biological explanations, I have no explanation for that involving males, i.e., the results may be a statistical artifact.

295 Herbaceous Richness and Cover

Turtle 400m2 and 1m2 plots had greater herbaceous richness than did random plots in both states (Figs. 6.4 & 5.15, Chapter 5); which was also the case at a WV Wood Turtle river site (McCoard et al. 2016b). Though I have few direct personal observations of feeding, I assumed that ground floor plant taxa found at a site may be an important driver of turtle use of those sites (see pg. 260 of Ernst and

Lovich 2009 for literature citations for foraging observations). I have only observed

Wood Turtles feeding, or observed evidence of feeding (such as pieces of foodstuffs on their faces), 39 times from 2006 to 2017 in VA and WV. Almost half of these

(18) involved herbaceous leaves, with the only identifiable taxon being Viola spp.

In both states 1m2 plots positioned at turtle points had significantly more herbaceous cover (combining both forbs and grass) than did those at random points. The lack of difference in amounts of herbaceous cover between 1m2 plots at the center of 400m2 plots and the four placed at the perimeter of the 400m2 plots suggests that the turtles are selecting for higher levels of cover at the meso-scale as well as the micro. Meso-scale preference is also corroborated by the five 1m2 plots at turtle points (the central one and four peripheral) having significantly more cover than the five at random points. In addition to providing edible herbaceous flora, high levels of herbaceous ground cover may be correlated to abundance or presence of invertebrate prey as well as facilitate the avoidance of predators.

296 Shrub Taxa

Sites with more shrub taxa may be preferred because they are more likely to provide foraging opportunities; perhaps Turtles were feeding on the fruits or leaves of seedlings of some species found in the shrub layer, such as Spicebush (Lindera benzoin) and Serviceberry (Amelanchier spp.). Turtles preferred sites with greater obscurity at the ground- and/or eye-levels; brushier sites with greater eye-level obscurity could have had more taxa, though strong correlations did not exist between these conditions here.

Tree Taxa

Females preferred sites with higher importance values for White Oak, Red

Maple, Sugar Maple, and White Ash or Tulip Tree, while males showed an aversion for sites with higher values for Chestnut Oak and Virginia Pine, as well as maples and Tulip Tree. Domination by Chestnut and Scarlet Oaks and Virginia

Pine is generally indicative of nutrient poor sites (oligotrophic) (Ashe 1922, Burns and Honkala 1990, Fleming and Coulling 2001, Weakley et al. 2012). Greater importance values for Sugar Maple, Red Maple, White Ash, and White Oak at sites may have to do with site productivity (higher soil nutrient availability) and/or moisture regimes (Burns and Honkala 1990). If herbaceous plants are an important foraging resource, Wood Turtles may be differentially using forest tracts dominated by different tree taxa; this study provides some evidence that the number of herbaceous taxa in the 400m2 plots at these VA and WV sites varied with forest type (Chapter 5). The logistic regressions indicated that female Wood Turtles prefer

297 sites with higher numbers of taxa of both larger trees (≥ 10cm dbh) and shrubs

(2.5cm ≤ dbh < 10cm); at a WV river study area McCoard and colleagues (2016a) also reported greater tree species richness at Wood Turtle locations than random locations.

It must be kept in mind that just because a site has high amounts of a certain taxon, such as Chestnut or Scarlet Oaks, does not mean that Wood Turtles cannot or do not use it. Such sites can have habitat attributes that the turtles prefer, such as

LWD, abundant mushrooms, particular forbs, or dense understories. In fact, at the plot with the highest importance value for any single tree species, a value of almost

98 for Chestnut Oak, a Wood Turtle was present. Because of this welter of interacting confounding conditions, Wood Turtles overall are labile in their use of sites with different tree taxa and proportions of tree taxa (i.e., different forest types).

Forest Type of Stand and Plot

Except at the stand scale for female turtles, forest type was a significant factor at neither the scale of a stand (generally ca. 5-20ha in size) nor the 400m2 scale when used as a random factor in mixed conditional logistic regressions (Table

6.13). The nonsignificant result may be due to two reasons, one being the size of the stands (generally 10ha or more). Since the summer activity areas of the turtles were generally small (ca. 2ha) and the random points were located within 300 meters of the turtle points, many random points would fall within the same stand as the turtle points. Secondly, the resolution of stand categorization is too coarse to reflect actual on-the-ground variation, even if the random point fell within a

298 different stand from the turtle point. For instance, an entire 20ha stand may be validly designated as a Scarlet Oak forest type since overall across the stand that taxon is predominant, but embedded within it there may be tracts that are dominated by White Oaks or maples. Because female turtles disperse farther from the main streams than males, they are more likely to encounter stands of different forest types and seral stages. These differences in overall composition and structure may influence or be in response to the other environmental conditions available on site.

The 400m2 plot-scale designations of forest type can be expected to more closely pick up actual on-the-ground variation in forest composition that is meaningful to the spatial scale at which Wood Turtles move about. However, when the forest type of the specific 400m2 plots was used as a random effect in the mixed conditional logistic regressions it was not a significant factor for either males or females (Table 6.17). This anomalous result is difficult to reconcile with the often clear differences between the stand- and plot-scale designations of forest types for turtle and random points (Chapter 5).

Both the stand- and plot-scale regression results suggest that forest type composition is of minor importance for Wood Turtles, particularly males. However, as the findings regarding importance values show, these forest type results should not be strictly interpreted as showing sites of a certain species composition are either preferred or avoided relatively more than others. It is just that the environmental variables used in this particular mixed modeling (e.g., number of

299 tree taxa) may vary little with forest type here, or if they do are of little influence on

Wood Turtle preference.

Stand Age and Seral Stage

Stand age was of marginal significance (at best) only for male Wood Turtles

(VA and WV combined) in the model averaged coefficients for the 2011-2014 data

(Table 6.14, Fig. 6.7); the negative sign for the males’ age coefficient indicated a preference for stands relatively younger than the random points. This outcome for the males may be due to locations of several Virginia males inside recent “modified shelterwood” logging sites (this type of even-age logging removes around 90% of that of a clearcut). In each case, however, the Turtles were located where a mature

“leave tree” was still standing, along with its relatively undisturbed understory, at the very edge of the cutting unit (within ca. 10-15m of the surrounding uncut forest)

(Fig. 6.10). Overall, Wood Turtles tended to avoid early successional habitat of regenerating sites of recent intensive logging. Seral stage did not arise as a significant factor in any of the regressions wherein it was used as a random effect

(Table 6.18). The majority of the turtle points (83.7%) as well as the random points

(70.5%) in Virginia were in older forest stands (mature and old growth seral stages).

It is possible that if more random points had been in early successional stands that seral stage would have emerged as a significant factor. As an example, in Maine

Wood Turtles were considered to not use regeneration sites of the forest-types they inhabit (Bryan 2007 at pg. 62).

300 General Issues

In general, animals should be trying to leave poor-quality habitats and trying to stay in higher-quality habitats (Garshelis 2002). The consistently observed philopatry and high survival of Wood Turtles at these study sites (this study, T. Akre unpub. data) are indicative of the high habitat quality found there. When a Wood

Turtle is located repeatedly at the same site it is not known whether this is due to frequent return to the site or to prolonged stay there. Either way, stay periods within a particular area generally indicate the favorability of that habitat for some performed activity (Owen-Smith et al. 2010). Perceived predation risk can alter the use of habitat by prey species (Carrascal et al. 1992, Calsbeek and Cox 2010, Costa

2014), though for herbivores the primary influence on home range occupation patterns may be the spatio-temporal availability of resources (Mueller and Fagan

2008).

Summer activity areas of 64 Wood Turtles radio-tracked during this study averaged 2.2ha (se = 0.35, range 0.1-13.3ha) (Chapter 1). Abundant, spatially concentrated, and predictable resources promote small home ranges; though the outcomes of interactions between food availability, habitat structure and complexity, and predator presence are complex and poorly understood (Huffaker

1958, Ritchie and Johnson 2009). Productivity decline or more patchily distributed or spatially unpredictable resources can result in home range expansion (i.e., lowering the population density in the landscape); in Ontario, Smith (2002) found

Wood Turtle density to be negatively correlated with the size of home ranges.

301 Resources that are both spatially and temporally unpredictable across years can result in different occupation patterns in different years; hence, there is no single best temporal scale to sample Wood Turtle spatial ecology.

According to “ideal free distribution”, as animal density increases in preferred habitats, less becomes available for each individual, so habitat quality for each diminishes as well (Fretwell and Lucas 1970). Thus, population density can influence habitat selection and in turn reproductive success and potential evolutionary change (Fortin et al. 2008). For Wood Turtles at these study sites, however, it is difficult to see how their apparently low population densities (e.g., an estimated 0.7 adults/ha in WV based upon an adult population size of 76 and a site area of 110ha; Chapter 3) could have any appreciable impact on available foraging, cover, or thermoregulatory resources; therefore, activity area size or habitat selection here is probably not influenced by conspecific densities. It is not clear which and to what extent potential interspecific competitors may be affecting

Wood Turtle spatial ecology. Perhaps herbivores, omnivores, or insectivores such as White-tailed Deer (Odocoileus virginianus), mice, voles, shrews, Wild Turkeys

(Melleagris gallopavo), or Box Turtles (T. carolina) are influencing Wood Turtles’ habitat use.

Wood Turtles are neither herd animals nor territorial (Kaufmann 1992b), though during their winter inactive period they may congregate at underwater stream hibernacula (Harding and Bloomer 1979, Parren 2013). Particularly in spring and fall, they also can be attracted to conspecifics for mating purposes in

302 aquatic habitats, but during the summer they apparently move independently and have overlapping home ranges/activity areas (Fig. 1.2). On several occasions, however, I found the same pairs of females very close to one another (within 1 meter) on different days at successive terrestrial locations separated by hundreds of meters. I also occasionally found males adjacent to females at various distances from the main streams. Those particular occasions aside, differences in habitat use between males and females such as revealed by this study have been observed previously in Wood Turtles as well as other chelonian species (Tingley et al. 2010,

Millar and Blouin-Demers 2011, Brown et al. 2016).

Differences in habitat preferences between Virginia and West Virginia turtles may be due to differences in vegetation, topography, or underlying soil types between the two sites. The West Virginia site is more sharply incised, with generally steeper slopes (see “Slope” at Table 6.6) and more topographic relief than in an area of similar size in Virginia (see Fig. 1.1). The relatively flat benches associated with the main streams are generally far wider in Virginia than in West

Virginia. In addition, the forests in West Virginia present a much greater dominance by conifers, indicating somewhat more oligotrophic conditions there than in

Virginia. Due to all these differences, there may be subtle though distinct variation in the distributional patterns of beneficial and adverse conditions, resulting in turn in different patterns of Wood Turtle behavior, selective pressures, and spatial ecology. For instance, overall the canopy is more open in West Virginia, so suitable

303 basking sites open to direct sun are more widely available and more easily attained by thermoregulating female turtles.

Conservation Recommendations

Habitat suitability models such as those resulting from my study contribute to conservation by delineating ecological requirements of species, facilitating design of conservation reserves and plans, predicting effects of habitat loss or alteration (including by climate change), providing an understanding of biogeography and dispersal barriers, directing the search for new populations or species, identifying reintroduction sites, and understanding species invasions

(Cianfrani et al. 2010). The habitat models and metrics described herein are practical, measurable, and understandable (Noss 1990). These allow conservationists and forest managers to 1) identify habitat components of high value to Wood Turtles, 2) assess the abundance, distribution, and quality of suitable habitat, 3) monitor habitat availability and quality over time, and 4) restore or protect suitable habitat and improve carrying capacity for the species (Bollman et al. 2005). Using the results from this study and others, a habitat suitability index

– a simple scoring system for evaluating the quality of known, potential, or restored habitats – could be developed for Wood Turtles. Also see “Conservation

Recommendations” in Chapter 5 and “Conservation Considerations” in Chapter 3 for relevant discussion in re Wood Turtles.

Tradeoffs exist between the predictive power and explanatory power of models (Kearney and Porter 2004); its applicability to other areas is one way the

304 robustness of a model can be measured. Some models can be applied to vast areas because habitat is similarly dominant across broad regions; e.g., the old growth

Douglas Fir forest constituting Spotted Owl (Strix occidentalis) habitat (Boyce et al.

2002). My study design does not test for broader-scale landscape effects beyond the extent of the study sites. Extrapolation beyond the region and scale where models were developed is generally to be avoided (Bollman et al. 2005, Hirzel and

Le May 2008); there is little theoretical support for believing preferences estimated in one region will be good predictors of preference in different regions (Beyer et al.

2010). The Wood Turtle models generated by this study, however, may have applicability in the region with similar ecological conditions Omernik and Bailey

1997), e.g., the Ridge and Valley physiographic province and where oak and mixed forests dominate (sensu Dyer 2006, McNab et al. 2007).

Within natural forests where this study was located, as well as in much of the northeast region where Wood Turtles range, a disturbance regime of small- scale, within-stand gap processes is the norm (Runkle 1985 & 1990, Mladenoff et al. 1993, White and White 1996, Seymour et al. 2002, Rentch 2006). These intermittent canopy disruptions occur through such mechanisms as windthrow, tree senescence, ice storms, drought, insects, American Beaver (Castor canadensis), floods, and pathogens (Braun 1950, Rentch 2006). Large “catastrophic” stand replacing events, such as hurricanes and conflagrations (canopy fires), are naturally a rare occurrence (Runkle 1990, Lorimer and White 2003). The congruence and harmonization, or lack thereof, of human disturbance (e.g., cutting regimes) with

305 the spatial and temporal parameters of the natural disturbance regime are an ongoing conservation concern throughout the Turtle’s range and elsewhere

(Franklin et al. 2002, Seymour et al. 2002, Lorimer and White 2003, Roberts 2004).

Researchers in northern Maine found individual tree selection and group selection systems to be the most obvious silvicultural analogs to the natural disturbance history (White et al. 2005). In research involving Appalachian mixed-hardwood sites in West Virginia, Miller and Kochenderfer (1998) found that that “[c]anopy openings with a minimum diameter of 170 feet (0.5 acre) provide suitable light conditions for virtually all desirable [tree] species to develop and grow to maturity”.

Intensive logging, such as even-age harvest methods that cut an entire stand, typically simplify structural diversity at sites, reduce litter and woody debris, and alter soil structure and microclimate regimes (Chen et al. 1999, Zheng et al. 2000,

Webster and Jenkins 2005, Todd and Andrews 2008). Except for severe fires, natural disturbances generally result in greater amounts of coarse woody debris

(Spies et al. 1988). Diminishment, removal, or absence of woody debris, litter, and humus can have a dramatic impact on organisms that depend on them for food and shelter (McMinn and Crossley 1996). Hence, intensive cutting and removal operations can negatively influence the abundance and species composition of turtle prey/forage such as arthropods, slugs, snails, and fruits (Shure and Phillips

1991, Caldwell 1996, Greenberg and Forrest 2003, Kappes 2006, Reynolds-

Hogland et al. 2006). This could be due to cooler, moister microclimates in

306 unlogged sites (Greenberg and Forrest 2003). For instance, slugs, especially stenoecious forest species, are highly sensitive to climatic fluctuations originating from canopy gaps or from disturbance of the leaf litter layer (Kappes 2006). A typical rationale used for timber sales is the assertion that after cutting the logged sites will have increased berry or soft mast production. However, this enhancement is only short-term (2-9 years), then the cutover sites have a very long period (30-60 years) of very low soft mast production (Reynolds-Hogland et al. 2006).

Burning is another concern for Wood Turtles, not only due to the potential for direct mortality, but also because of the potential for habitat degradation. Decay processes generally tend to mesify microsites while fire tends to xerify them (Van

Lear 1996). Burning tends to make sites hotter, drier and more open, thereby exposing organisms such as turtles to more predators and desiccation. The incineration of forest floor material (viz., woody debris, litter, humus) may also directly destroy or reduce the site quality for biota that serve as turtle prey/forage.

By reducing important components of habitat such as leaf litter, fire can degrade mesic micro-habitats (Ford et al. 1999), such as that for snails (Martin and Sommer

2004), and hinder turtle osmo- and thermos-regulation. Moreover, fire may not be necessary for maintaining and regenerating northeastern oak forests and increased frequency of burning could potentially reduce forest herb and shrub diversity (Elliot et al. 2004, Matlack 2013).

Forested habitats of different structure and composition may yield differences in foraging success, abundance of refugia from predators, or thermal

307 and hydric properties. Seral stages differ in composition and structure (Meier et al.

1995, Hardt and Swank 1997) and it can take many decades for forest composition, structure, or function to recover from human disturbance (Likens et al.

1978, Meier et al. 1995, Bellemare et al. 2002). Wood Turtles at these study sites were disinclined to use large recent even-age logging sites except at their very margins. Where Wood Turtle populations occur in this ecoregion, the results of this study suggest that simply letting forests develop mature and old-growth conditions under a natural disturbance regime (i.e., restoration by “purposeful action and inaction”, Trombulak 1996) would be the best and probably least expensive course for their conservation. Although passive restoration (purposeful inaction) is often successful (Jones et al. 2018), in some site-specific situations this general guideline of just ceasing human disturbances and letting a forest develop on its own may not be sufficient. Unlike my study sites, other Wood Turtle sites may not be relatively intact older forest. Active restoration may be necessary to counteract previous disruptions and degradations; e.g., countering inflated populations of herbivores, alien invasives, noxious pests, or harmful roads by reintroduction of larger predators, direct removal of invasives/pests, or closure of roads to public vehicular traffic.

Further, some land holders and managers can take ecologically sensitive actions that at some places could improve Wood Turtle habitat. Beyond such focused actions as, e.g., the fabrication or maintenance of nesting sites (Buhlmann and Osborn 2011), I am here referring to silvicultural techniques that fall under the

308 broad rubric of "structural complexity enhancement" (SCE) (Keeton 2006, Scheff

2014). Typical objectives of SCE include vertically differentiated canopies, elevated large snag and LWD volumes and densities, variable horizontal density (including canopy gaps), and re-allocation of tree basal area to larger diameter classes (Keeton

2006). Intensive logging, such as typical even-age harvest methods, generally simplify structural diversity at sites, reduce litter and woody debris, and alter soil structure and microclimate regimes (Chen et al. 1999, Zheng et al. 2000, Webster and Jenkins 2005, Todd and Andrews 2008). SCE, however, could accelerate the development of important older forest characteristics while allowing for an economic return (Keeton and Troy 2006). Mimicking gap-scale natural disturbance in a limited and targeted manner can fall within the range of disturbance intensities consistent with developing and maintaining old-growth structure while assisting in the regeneration and recruitment of oaks and other shade intermediate-tolerant species (Scheff 2014).

Of course, it is not just the availability of artificial openings fabricated by logging that determines whether oaks can reestablish and sustain themselves at sites in a forest (Rentch et al. 2003a, McEwan and Muller 2006, McEwan et al.

2010). Where perpetuation of a substantial oak component is a concern, oak recruitment can be facilitated by locating individual selection or small group selection harvests (during winter months when Wood Turtles are totally aquatic) in forest patches with ample advanced oak regeneration. Oak seedlings can grow and out-compete other species in small gaps or even under canopy (Beckage 2000,

309 Clinton 2003, Iffrig et al. 2008); for example, Rentch and colleagues (2003b) found oaks were able to establish and persist in gaps < 200m2 in area.

As Wood Turtles are not confined to riparian or wetland habitats, but instead regularly range far afield in dry upland habitats (Table 6.6 and Chapter 5), stream courses occupied by Wood Turtles in this and similar ecoregions should be buffered on both sides by at least a 300 meter minimal disturbance zone in order to mitigate for effects to turtle population viability and allow the natural development of conditions essential to their survival. Even when this measure is implemented, offsite effects of forest anthropogenic disturbance remain of concern. It is critical that deleterious edge effects, which translate to a form of habitat loss, receive much more explicit consideration for conservation to prove to be effective (Harris et al.

1996, Zheng and Chen 2000, Fletcher 2005, Harper et al. 2005). Because the condition of the matrix within which occupied patches reside may influence turtle abundance and population viability, effective restoration and protection must encompass even larger spatial scales (beyond the 300m zones) (Hansen and Rotella

2002, Ficetola et al. 2004, Roe and Georges 2007, Quesnelle et al. 2013). The

300m prescription should generally be considered a minimum standard (site specific conditions may obviate or preclude its implementation) as this zone may not include lengthy pre-nesting peregrinations by female Wood Turtles or connectivity to other populations. Improving or protecting the quality of other habitats outside of more strictly protected core areas can be crucial (Angermeier

310 1995, Harris et al. 1996, Browne and Hecnar 2007, Hansen and DeFries 2007,

Quesnelle et al. 2013).

For instance, stream communities at sites with stringently protected riparian buffers can still be significantly degraded by intensive development elsewhere in the catchment (Wahl et al. 2013). The cumulative effects of timber harvest on sedimentation rates last for many years, even after cutting has ceased in an area

(Frissell 1997), and erosion from roads used for logging often contributes more sediment than the land logged for timber (Box and Mossa 1999). Increased sedimention, turbidity, and/or nutrient loads from erosion are known to reduce dissolved oxygen levels (Henley et al. 2000). Oxygen levels may be a critical variable for Wood Turtle survival during winter dormancy (Graham and Forsberg

1991, Ultsch 2006, Greaves and Litzgus 2007 & 2008). See “Small Streams,

Springs, and Seepages” and “Hardwood Forests” modules in Mitchell et al. (2006) for general habitat management guidelines apropos to Wood Turtles.

The factors identified by this study can be used for well-informed decisions regarding management practices, protective measures, and habitat enhancement or restoration (e.g., fabrication of small canopy gaps), as well as make predictions as to the suitability of sites as potential turtle habitat. In short, we need to develop our understanding of the turtles, not develop their habitat. A multitude of other flora and fauna, including human communities, will benefit when we accord Wood

Turtles enhanced on-the-ground protections.

311 Table 6.1

Ground cover variables measured in 1m2 plots – microhabitat Terrestrial habitat variables measured in 1m2 plots at each Wood Turtle and associated random location in Virginia and West Virginia June-August 2011-2014. Samples taken: one centered on each turtle and random point; in 2011-2012, also one at each of the four cardinal directions at the perimeter of the 400-m2 plot circle centered on turtle/random points. Variables visually estimated as proportion (percent) of ground cover in 1m2 frame. ______Variable Description

Inorganic (In) Sand proportion of sand and/or gravel present

Rock proportion of rock present

Herbaceous vegetation (H) Forb proportion of herbaceous dicot (non-grassy) vegetation present

Grass proportion of grassy vegetation (including sedges and rushes) present

Non-herbaceous vegetation (Nh) Woody proportion of woody vegetation present

Moss proportion of ground covered by bryophytes

Fern proportion of ground covered by ferns

Organic non-living (O) CWD proportion of woody debris > 5 cm diameter present

Soil proportion of bare soil present

Litter proportion of ground covered by leaf litter

Water proportion of water (standing or flowing) present

Other proportion of mushrooms, lichens, lycopodia, or equisetums present

Herbcov proportion of ground cover in frame that is forb and grass combined

312 Table 6.2

Environmental variables measured in 400m2 plots – mesohabitat Habitat variables measured in 400m2 plots at each Wood Turtle and associated random location in Virginia and West Virginia June-August 2011-2014. ______Variable Description

Structural overhead (So) Large count of each live woody stem ≥ 25 cm dbh – taxon and dbh also recorded

Medium count of live stems ≥ 10cm & < 25cm dbh – taxon and dbh also recorded

Obs1 obscurity measured (%) at human eye-level with modified Nudds profile board placed at perimeter of plot at the four cardinal directions (relative to the Turtle point) – mean of four readings

Snags number of standing dead trees ≥ 10cm dbh in plot

Can canopy openness (%) over turtle and random points – mean of open percentages taken at each of the four cardinal directions using a spherical densitometer

Age age of stand in years as determined by USFS

Shrub count of woody stems ≥ 2.5cm dbh and <10cm dbh

BA basal area (m2/ha) of plot calculated from dbh measurements of trees ≥ 10cm dbh

Structural ground (Sg) Obs4 obscurity measured (%) at ground floor with Modified Nudds profile board placed at perimeter of plot at the four Cardinal directions (relative to the Turtle point) – mean of four readings; care was taken to not place the board behind trees ≥ 10cm dbh

Gapsize visual estimate of amount of ground area (m2) under canopy gaps ≥ 9m2 in the plot; natural or anthropogenic gaps (e.g., treefall gap or grassy game opening)

LWD25 count of woody debris pieces ≥ 25cm diameter in plot (≥ 4m in length and at least 2m of length in plot)

LWD10 count of woody debris pieces ≥ 10cm and < 25cm diameter in plot (≥ 4m in length and at least 2m of length in plot)

Compositional overhead (Co) Tree spp count of woody taxa present with stems ≥ 10cm dbh

Shrub spp count of woody taxa present with stems ≥ 2.5cm dbh and <10cm dbh

Compositional ground (Cg) Seed spp count of woody taxa < 50cm high present on ground floor

Herb spp count of ground floor forb taxa present in plot

Topographical (T) Dist distance (m) to the closest permanent stream; paced off, measured with tape, or estimated in GIS

Slope site inclination in degrees measured with a clinometer

ASPB aspect – first estimated with compass in degrees, then converted with Beers transform (this metric ranges from 0-2, 0 = SW warm aspect, 2 = NE cool aspect)

Elev elevation (m) measured with GPS unit, quad maps, or GIS

Behavioral Dlwd distance (m) of turtle/random point to the closest large woody debris ≥ 25cm diameter

313 Table 6.3

Conditional logistic regression models used – 1m2 plots Models used in conditional logistic regression analyses of proportions of cover present in 1m2 plots at turtle points and random points in Virginia and West Virginia during June-August 2011-2014. Variable types: In = inorganic (sand/gravel and rock), H = herbaceous vegetation (forb and grass), Nh = non-herbaceous vegetation (woody and moss), Or = organic non-living (coarse woody debris and bare soil).

Variable type model # model variables ______

Virginia females, West Virginia females & males

In-H mod1 Sand+Rock+Forb+Grass In-Nh mod2 Sand+Rock+Woody+Moss In-Or mod3 Sand+Rock+CWD+Soil H-Nh mod4 Forb+Grass+Woody+Moss H-Or mod5 Forb+Grass+CWD+Soil Nh-Or mod6 Woody+Moss+CWD+Soil In-H-Nh mod7 Sand+Rock+Forb+Grass+Woody+Moss In-H-Or mod8 Sand+Rock+Forb+Grass+CWD+Soil In-Nh-Or mod9 Sand+Rock+Woody+Moss+CWD+Soil H-Nh-Or mod10 Forb+Grass+Woody+Moss+CWD+Soil In-H-Nh-Or mod11 Sand+Forb+Woody+CWD In-H-Nh-Or mod12 Rock+Grass+Moss+Soil In-H-Nh-Or mod13 Sand+Forb+Moss+Soil In-H-Nh-Or mod14 Rock+Grass+Woody+CWD Global mod15 Sand+Rock+Forb+Grass+CWD+Woody+Moss+Soil

Virginia males (Sand removed)

In-H mod1 Rock+Forb+Grass In-Nh mod2 Rock+Woody+Moss In-Or mod3 Rock+CWD+Soil H-Nh mod4 Forb+Grass+Woody+Moss H-Or mod5 Forb+Grass+CWD+Soil Nh-Or mod6 Woody+Moss+CWD+Soil In-H-Nh mod7 Rock+Forb+Grass+Woody+Moss In-H-Or mod8 Rock+Forb+Grass+CWD+Soil In-Nh-Or mod9 Rock+Woody+Moss+CWD+Soil H-Nh-Or mod10 Forb+Grass+Woody+Moss+CWD+Soil In-H-Nh-Or mod11 Forb+Woody+CWD In-H-Nh-Or mod12 Rock+Grass+Moss+Soil In-H-Nh-Or mod13 Forb+Moss+Soil In-H-Nh-Or mod14 Rock+Grass+Woody+CWD Global mod15 Rock+Forb+Grass+CWD+Woody+Moss+Soil

314 Table 6.4

Conditional logistic regression models used – 400m2 plots Models used in conditional logistic regression analyses of environmental attributes present in 400m2 plots at turtle points and random points in Virginia and West Virginia during June-August 2011-2014. Variable types: So = structural overhead (Large, Medium, Obs1, Snags, Can, Age), Sg = structural ground (Obs4, Gapsizem2, LWD10, LWD4), Co = compositional overhead (Treespp, Shrubspp), Cg = compositional ground (Seedspp, Herbspp [for 2011-2013 models]), T = topographical (Dist, Slope, ASPB, Elev), Iv = importance value of tree taxa (a synthetic metric incorporating both structural and compositional attributes – models using Iv did not include Large, Medium, or Treespp). See Table 2 for description of variables. model type model # model variables ______

Virginia females

So-Sg mod1 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+ LWD4+Snags+Can So mod2 Large+Medium+Obs1+Snags+Can Sg mod3 Obs4+LWD10+LWD4+Gapsizem2 Co-Cg mod4 Treespp+Shrubspp+Seedspp Co mod5 Treespp+Shrubspp Cg mod6 Seedspp T mod7 Dist+Slope+ASPB+Elev So-Sg-Co-Cg mod8 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+ LWD4+Snags+Can+Treespp+Shrubspp+Seedspp So-Co mod9 Large+Medium+Obs1+Snags+Can+Treespp+Shrubspp So-Cg mod10 Large+Medium+Obs1+Snags+Can+Seedspp Sg-Co mod11 Obs4+LWD10+LWD4+Gapsizem2+Treespp+Shrubspp Sg-Cg mod12 Obs4+LWD10+LWD4+Gapsizem2+Seedspp So-Sg-T mod13 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+LWD4+ Snags+Can+Dist+ASPB+Slope+Elev So-T mod14 Large+Medium+Obs1+Can+Slope+ASPB+Dist+Elev Sg-T mod15 Obs4+LWD10+LWD4+Gapsizem2+Slope+ASPB+Dist+Elev Co-Cg-T mod16 Treespp+Shrubspp+Seedspp+Dist+ASPB+Slope+Elev Co-T mod17 Treespp+Shrubspp+Dist+ASPB+Slope+Elev Cg-T mod18 Seedspp+Dist+ASPB+Slope+Elev So-Co-T mod19 Large+Medium+Obs1+Snags+Can+Treespp+Shrubspp+ Dist+ASPB+Slope+Elev So-Cg-T mod20 Large+Medium+Obs1+Snags+Can+Seedspp+ Dist+ASPB+Slope+Elev Sg-Co-T mod21 Obs4+LWD10+LWD4+Gapsizem2+Treespp+Shrubspp+ Dist+ASPB+Slope+Elev Sg-Cg-T mod22 Obs4+LWD10+LWD4+Gapsizem2+Seedspp+ Dist+ASPB+Slope+Elev

Virginia males

So-Sg mod1 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+ LWD4+Snags+Can So mod2 Large+Medium+Obs1+Snags+Can Sg mod3 Obs4+LWD4 Co-Cg mod4 Treespp+Shrubspp+Seedspp Co mod5 Treespp+Shrubspp Cg mod6 Seedspp T mod7 Dist+Slope+ASPB+Elev So-Sg-Co-Cg mod8 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+ LWD4+Snags+Can+Treespp+Shrubspp+Seedspp So-Co mod9 Large+Medium+Obs1+Snags+Can+Treespp+Shrubspp So-Cg mod10 Large+Medium+Obs1+Snags+Can+Seedspp Sg-Co mod11 Obs4+LWD10+LWD4+Gapsizem2+Treespp+Shrubspp Sg-Cg mod12 Obs4+LWD4+Seedspp So-Sg-T mod13 Large+Medium+Obs1+LWD10+LWD4+Can+ Dist+ASPB+Slope+Elev

315 Table 6.4: continued So-T mod14 Large+Medium+Obs1+Can+Slope+ASPB+Dist+Elev Sg-T mod15 Obs4+LWD10+LWD4+Gapsizem2+Slope+ASPB+Dist+Elev Co-Cg-T mod16 Treespp+Shrubspp+Seedspp+Dist+ASPB+Slope+Elev Co-T mod17 Treespp+Shrubspp+Dist+ASPB+Slope+Elev Cg-T mod18 Seedspp+Dist+ASPB+Slope+Elev So-Co-T mod19 Large+Medium+Obs1+Snags+Can+Treespp+Shrubspp+ Dist+ASPB+Slope+Elev So-Cg-T mod20 Large+Medium+Obs1+Snags+Can+Seedspp+ ASPB+Slope+Elev Sg-Co-T mod21 Obs4+LWD10+LWD4+Treespp+Shrubspp+ Dist+ASPB+Slope+Elev Sg-Cg-T mod22 LWD10+LWD4+Gapsizem2+Seedspp+ Dist+ASPB+Slope+Elev West Virginia females

So-Sg mod1 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+ LWD4+Snags+Can+Age So mod2 Large+Medium+Obs1+Snags+Can+Age Sg mod3 Obs4+LWD10+LWD4+Gapsizem2 Co-Cg mod4 Treespp+Shrubspp+Seedspp Co mod5 Treespp+Shrubspp Cg mod6 Seedspp T mod7 Dist+Slope+ASPB+Elev So-Sg-Co-Cg mod8 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+LWD4+ Snags+Can+Treespp+Shrubspp+Seedspp+Age So-Co mod9 Large+Medium+Obs1+Snags+Can+Treespp+Shrubspp+Age So-Cg mod10 Large+Medium+Obs1+Snags+Can+Seedspp+Age Sg-Co mod11 Obs4+LWD10+LWD4+Gapsizem2+Treespp+Shrubspp Sg-Cg mod12 Obs4+LWD10+LWD4+Gapsizem2+Seedspp So-Sg-T mod13 Large+Medium+Obs4+Gapsizem2+LWD10+LWD4+ Snags+Can+Dist+ASPB+Slope+Elev+Age So-T mod14 Large+Medium+Obs1+Can+Slope+ASPB+Dist+Elev+Age Sg-T mod15 Obs4+LWD10+LWD4+Gapsizem2+Slope+ASPB+Dist+Elev Co-Cg-T mod16 Treespp+Shrubspp+Seedspp+Dist+ASPB+Slope+Elev Co-T mod17 Treespp+Shrubspp+Dist+ASPB+Slope+Elev Cg-T mod18 Seedspp+Dist+ASPB+Slope+Elev So-Co-T mod19 Large+Medium+Obs1+Snags+Can+Treespp+Shrubspp+ Dist+ASPB+Slope+Elev+Age So-Cg-T mod20 Large+Medium+Obs1+Snags+Can+Seedspp+ Dist+ASPB+Slope+Elev+Age Sg-Co-T mod21 Obs4+LWD10+LWD4+Gapsizem2+Treespp+Shrubspp+ Dist+ASPB+Slope+Elev Sg-Cg-T mod22 Obs4+LWD10+LWD4+Gapsizem2+Seedspp+ Dist+ASPB+Slope+Elev

West Virginia males

So-Sg mod1 Large+Medium+Obs1+Obs4+Snags+LWD10+LWD4+ Can+Gapsizem2+Age So mod2 Large+Medium+Obs1+Snags+Can+Age Sg mod3 Obs4+LWD10+LWD4+Gapsizem2 Co-Cg mod4 Treespp+Shrubspp+Seedspp Co mod5 Treespp+Shrubspp Cg mod6 Seedspp T mod7 Dist+Slope+ASPB+Elev So-Sg-Co-Cg mod8 Large+Medium+Obs1+Obs4+Gapsizem2+LWD10+LWD4+ Snags+Can+Hebcov+Treespp+Shrubspp+Seedspp+Age So-Co mod9 Large+Medium+Obs1+Snags+Can+Treespp+Shrubspp+Age So-Cg mod10 Large+Medium+Obs1+Snags+Can+Seedspp+Age Sg-Co mod11 Obs4+LWD10+LWD4+Gapsizem2+Treespp+Shrubspp Sg-Cg mod12 Obs4+LWD10+LWD4+Gapsizem2+Seedspp So-Sg-T mod13 Large+Medium+Obs1+Obs4+LWD10+Can+ Gapsizem2+ASPB+Slope So-T mod14 Large+Medium+Obs1+Can+Slope+ASPB+Dist+Elev+Age Sg-T mod15 Obs4+LWD10+LWD4+Gapsizem2+Slope+ASPB+Dist+Elev Co-Cg-T mod16 Treespp+Shrubspp+Seedspp+Dist+ASPB+Slope+Elev Co-T mod17 Treespp+Shrubspp+Dist+ASPB+Slope+Elev Cg-T mod18 Seedspp+Dist+ASPB+Slope+Elev

316 Table 6.4: continued So-Co-T mod19 Large+Medium+Obs1+Snags+Treespp+Shrubspp+ ASPB+Slope+Elev+Age So-Cg-T mod20 Large+Medium+Obs1+Snags+Can+Seedspp+ ASPB+Slope+Age Sg-Co-T mod21 Obs4+LWD10+LWD4+Gapsizem2+Treespp+Shrubspp+ Dist+ASPB+Slope+Elev Sg-Cg-T mod22 Obs4+LWD10+LWD4+Gapsizem2+Seedspp+ Dist+ASPB+Slope+Elev

Female Wood Turtles

So-Sg mod1 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+ LWD4+Snags+Can So mod2 Large+Medium+Obs1+Snags+Can Sg mod3 Obs4+LWD10+LWD4+Gapsizem2 Co-Cg mod4 Treespp+Shrubspp+Seedspp Co mod5 Treespp+Shrubspp Cg mod6 Seedspp T mod7 Dist+Slope+ASPB+Elev So-Sg-Co-Cg mod8 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+ LWD4+Snags+Can+Treespp+Shrubspp+Seedspp So-Co mod9 Large+Medium+Obs1+Snags+Can+Treespp+Shrubspp So-Cg mod10 Large+Medium+Obs1+Snags+Can+Seedspp Sg-Co mod11 Obs4+LWD10+LWD4+Gapsizem2+Treespp+Shrubspp Sg-Cg mod12 Obs4+LWD10+LWD4+Gapsizem2+Seedspp Iv-So-Sg-Co-Cg mod13 CO+SM+RM+WO+BG+WA+TP+SO+NRO+Obs1+Obs4+Snags+ LWD10+LWD4+Gapsizem2+Can+Shrubspp+Seedspp Iv-Sg-Cg mod14 CO+SM+RM+WO+BG+WA+TP+SO+NRO +Obs4+LWD10+LWD4+Gapsizem2+Seedspp Iv-So-Co mod15 CO+SM+RM+WO+BG+WA+TP+SO+NRO+ Obs1+Snags+Can+ Shrubspp Iv-Sg-Co mod16 CO+SM+RM+WO+BG+WA+TP+SO+NRO+ Obs4+LWD10+LWD4+Gapsizem2+Shrubspp Iv-So-Cg mod17 CO+SM+RM+WO+BG+WA+TP+SO+NRO+Obs1+Snags+Can+ Seedspp So-Sg-T mod18 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+LWD4 +Snags+Can+Dist+ASPB+Slope+Elev So-T mod19 Large+Medium+Obs1+Can+Slope+ASPB+Dist+Elev Sg-T mod20 Obs4+LWD10+LWD4+Gapsizem2+Slope+ASPB+Dist+Elev Co-Cg-T mod21 Treespp+Shrubspp+Seedspp+Dist+ASPB+Slope+Elev Co-T mod22 Treespp+Shrubspp+Dist+ASPB+Slope+Elev Cg-T mod23 Seedspp+Dist+ASPB+Slope+Elev So-Co-T mod24 Large+Medium+Obs1+Snags+Can+Treespp+Shrubspp+ Dist+ASPB+Slope+Elev So-Cg-T mod25 Large+Medium+Obs1+Snags+Can+Seedspp+ Dist+ASPB+Slope+Elev Sg-Co-T mod26 Obs4+LWD10+LWD4+Gapsizem2+Treespp+Shrubspp+ Dist+ASPB+Slope+Elev Sg-Cg-T mod27 Obs4+LWD10+LWD4+Gapsizem2+Seedspp+ Dist+ASPB+Slope+Elev Iv-So-Co-T mod28 CO+SM+RM+WO+BG+WA+TP+SO+NRO+Obs1+Snags+Can +Shrubspp+Dist+ASPB+Slope+Elev Iv-So-Cg-T mod29 CO+SM+RM+WO+BG+WA+TP+SO+NRO+Obs1+Snags+Can+ Seedspp+Dist+ASPB+Slope+Elev Iv-Sg-Co-T mod30 CO+SM+RM+WO+BG+WA+TP+SO+NRO+Obs4+LWD10+ LWD4+Gapsizem2+Shrubspp+Dist+ASPB+Slope+Elev Iv-Sg-Cg-T mod31 CO+SM+RM+WO+BG+WA+TP+SO+NRO+Obs4+LWD10+ LWD4+Gapsizem2+Seedspp+Dist+ASPB+Slope+Elev

Male Wood Turtles

So-Sg mod1 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+ LWD4+Snags+Can So mod2 Large+Medium+Obs1+Snags+Can Sg mod3 Obs4+LWD4 Co-Cg mod4 Treespp+Shrubspp+Seedspp Co mod5 Treespp+Shrubspp Cg mod6 Seedspp

317 Table 6.4: continued T mod7 Dist+Slope+ASPB+Elev So-Sg-Co-Cg mod8 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+ LWD4+Snags+Can+Treespp+Shrubspp+Seedspp So-Co mod9 Large+Medium+Obs1+Snags+Can+Treespp+Shrubspp So-Cg mod10 Large+Medium+Obs1+Snags+Can+Seedspp Sg-Co mod11 Obs4+LWD10+LWD4+Gapsizem2+Treespp+Shrubspp Sg-Cg mod12 Obs4+LWD4+Seedspp Iv-So-Sg-Co-Cg mod13 CO+SM+RM+WO+BG+WA+TP+SO+NRO+Obs1+LWD10+Can Iv-Sg-Cg mod14 CO+SM+RM+WO+BG+WA+TP+SO+NRO+LWD10+LWD4+ Seedspp Iv-So-Co mod15 CO+SM+RM+WO+BG+WA+TP+SO+NRO+Obs1+Snags+Can+ Shrubspp Iv-Sg-Co mod16 CO+SM+RM+WO+BG+WA+TP+SO+NRO+LWD10+LWD4+ Gapsizem2+Shrubspp Iv-So-Cg mod17 CO+SM+RM+WO+BG+WA+TP+SO+NRO+Obs1+Snags+Can So-Sg-T mod18 Large+Medium+Obs1+LWD10+LWD4+Can+ Dist+ASPB+Slope+Elev So-T mod19 Large+Medium+Obs1+Can+Slope+ASPB+Dist+Elev Sg-T mod20 Obs4+LWD10+LWD4+Gapsizem2+Slope+ASPB+Dist+Elev Co-Cg-T mod21 Treespp+Shrubspp+Seedspp+Dist+ASPB+Slope+Elev Co-T mod22 Treespp+Shrubspp+Dist+ASPB+Slope+Elev Cg-T mod23 Seedspp+Dist+ASPB+Slope+Elev So-Co-T mod24 Large+Medium+Obs1+Snags+Can+Treespp+Shrubspp+ Dist+ASPB+Slope+Elev So-Cg-T mod25 Large+Medium+Obs1+Snags+Can+Seedspp+ ASPB+Slope+Elev Sg-Co-T mod26 Obs4+LWD10+LWD4+Treespp+Shrubspp+ Dist+ASPB+Slope+Elev Sg-Cg-T mod27 LWD10+LWD4+Gapsizem2+Seedspp+ Dist+ASPB+Slope+Elev Iv-So-Co-T mod28 CO+SM+RM+WO+BG+WA+TP+SO+NRO+Obs1+Snags+ Can+Shrubspp+ASPB+Slope Iv-So-Cg-T mod29 CO+SM+RM+WO+BG+WA+TP+SO+NRO+ Obs1+Snags+ Can+ASPB+Slope Iv-Sg-Co-T mod30 CO+SM+RM+WO+BG+WA+TP+SO+NRO+LWD10+LWD4+ Shrubspp+Dist+ASPB+Slope+Elev Iv-Sg-Cg-T mod31 CO+SM+RM+WO+BG+WA+TP+SO+NRO+ LWD10+LWD4+ Seedspp+ASPB+Slope+Elev

318 Table 6.5a

Values of ground cover variables measured in 1m2 plots Values for coverage proportions (in %) of structural-compositional variables obtained at 1m2 plots at turtle points and random points (one plot per point) in Virginia and West Virginia during June-August 2011-2014; reported for each variable in descending order are means, standard errors, and ranges. See Table 6.1 for description of variables. FWT = female Wood Turtles (VA: 35 turtles/143 plots, WV: 25 turtles/82 plots), FRP = female random points (VA: 35/141, WV: 24/74), MWT = male Wood Turtles (VA: 15/52, WV: 21/51), MRP = male random points (VA: 15/52, WV: 20/49). Virginia West Virginia Variable FWT FRP MWT MRP FWT FRP MWT MRP Sand/Gravel 0.02 0.65 0.15 0 0.71 1.05 2.04 0.27 0.02 0.53 0.15 0 0.34 0.70 1.25 0.17 0-3 0-74 0-8 0 0-9 0-47 0-46 0-7

Rock 0.80 1.77 1.77 1.88 0.61 0.11 0.16 0.41 0.27 0.45 1.31 0.58 0.34 0.04 0.07 0.17 0-32 0-45 0-68 0-18 0-27 0-2 0-3 0-6

Forb 6.25 1.51 5.35 0.79 4.01 2.23 4.59 3.22 1.12 0.29 0.79 0.35 0.52 0.45 0.63 0.67 0-89 0-26 0-20 0-16 0-25 0-64 0-17 1-44

Grass 8.36 1.52 8.83 0.44 10.34 5.88 12.55 4.76 1.49 0.61 2.11 0.25 1.88 1.69 2.41 1.17 0-75 0-79 0-65 0-12 0-74 0-80 0-75 0-50

Woody 4.74 3.30 4.54 3.58 2.38 2.64 2.22 1.51 0.59 0.23 0.60 0.38 0.38 0.59 0.51 0.29 0-65 0-15 0-20 0-13 0-19 0-34 0-25 0-7

Moss 2.28 3.00 0.83 1.13 3.78 3.27 0.61 2.63 0.45 0.67 0.24 0.35 1.35 0.86 0.19 1.26 0-38 0-50 0-10 0-14 0-70 0-40 0-8 0-61

CWD 4.67 1.79 4.96 1.90 2.56 1.51 2.92 1.12 0.56 0.33 1.20 0.79 0.55 0.34 0.95 0.35 0-31 0-20 0-40 0-33 0-25 0-15 0-30 0-11

Soil 2.33 0.96 1.81 1.42 0.65 0.36 1.27 0.59 0.70 0.35 0.98 1.25 0.20 0.15 1.10 0.25 0-65 0-33 0-47 0-65 0-9 0-7 0-56 0-8

Fern 0.45 0.01 0 0 0.23 0.37 0.20 0.37 0.28 0.01 0 0 0.10 0.17 0.09 0.14 0-39 0-1 0 0 0-6 0-10 0-3 0-4

Other 0.15 0.14 0.06 0.15 0.89 0.25 0.31 0.06 0.05 0.08 0.03 0.08 0.67 0.17 0.23 0.05 0-100 0-79 0-73 0-28 0-53 0-12 0-10 0-2

Litter 69.95 85.34 71.71 88.69 72.99 82.34 73.14 85.06 2.16 1.31 3.04 1.91 2.53 2.45 3.12 2.14 0-98 0-99 5-95 15-100 2-99 13-99 10-97 16-100

Herbaceous 14.63 3.26 14.09 1.34 14.38 8.23 15.92 7.98 cover (forb 2.05 0.74 2.56 0.65 2.13 1.72 2.35 1.53 + grass) 0-100 0-79 0-73 0-28 0-75 0-81 0-76 0-53

Herb. cover 14.96 2.54 5.29 1.76 9.82 5.31 12.38 4.58 (peripheral) 2.35 0.62 1.64 0.76 1.49 1.80 2.07 1.04 2011-2012) 0-51 0-21 0-23 0-12 0-26 0-30 1-33 0-20

Forb taxa 2.72 1.28 4.33 0.92 2.67 2.63 4.42 2.50 2013-2014 0.27 0.22 0.46 0.32 0.30 0.35 0.47 0.46 (#) 0-9 0-10 0-9 0-9 0-7 0-9 1-8 0-8

319 Table 6.5b

Values of ground cover variables measured in four peripheral 1m2 plots Values for coverage proportions (in %) of structural-compositional variables obtained at four 1m2 plots at turtle points and paired random points (means of four peripheral plots per point) in Virginia and West Virginia during June-August 2011-2012; reported for each variable in descending order are means, standard errors, and ranges. See Table 6.1 for description of variables. FWT = female Wood Turtles (VA: 13 turtles/228 plots, WV: 10 turtles/116 plots), FRP = female random points (VA: 13/228, WV: 9/92), MWT = male Wood Turtles (VA: 5/72, WV: 10/100), MRP = male random points (VA: 5/72, WV: 9/96). Virginia West Virginia Variable FWT FRP MWT MRP FWT FRP MWT MRP

Sand/Gravel 0.22 0.51 4.10 1.13 2.26 1.08 3.77 0.07 0.12 0.24 1.95 0.95 0.94 0.77 1.54 0.05 0-16 0-13 0-25 0-17 0-20 0-16 0-29 0-1

Rock 2.59 1.38 2.08 1.83 2.34 0.27 5.38 1.18 0.49 0.28 0.70 0.58 0.95 0.19 1.58 0.87 0-17 0-12 0-9 0-10 0-22 0-4 0-21 0-21

Forb 5.14 1.00 2.37 1.10 3.29 1.01 3.61 1.56 0.89 0.20 0.67 0.49 0.38 0.31 0.54 0.26 0-26 0-7 0-9 0-7 0-8 0-6 0-10 0-5

Grass 9.82 1.53 2.93 0.67 5.28 4.09 8.90 3.02 1.71 0.51 1.07 0.30 1.14 1.53 1.72 0.92 0-42 0-18 0-14 0-5 0-21 0-24 0-25 0-16

Woody 3.31 2.87 4.40 3.58 1.78 2.04 1.40 1.41 0.41 0.24 0.58 0.54 0.39 0.41 0.37 0.30 0-14 1-9 1-10 2-12 0-10 0-8 0-8 0-6

Moss 3.62 2.65 2.31 3.68 2.52 3.96 3.20 4.45 0.49 0.57 0.76 0.97 0.82 1.14 0.80 1.19 0-19 0-19 0-13 0-16 0-21 0-18 0-15 0-20

CWD 2.63 3.66 3.38 2.31 3.43 3.28 1.90 2.60 0.45 0.46 0.99 0.61 0.67 0.88 0.61 0.63 0-13 0-14 0-16 0-11 0-13 0-19 0-14 0-13

Soil 1.45 3.16 2.24 2.67 1.60 0.62 2.72 0.97 0.36 0.82 1.82 1.72 0.97 0.25 1.11 0.70 0-16 0-27 0-33 0-27 0-28 0-5 0-18 0-17

Fern 0.36 0.11 0 0.01 0.34 0.28 0.17 0.43 0.11 0.06 0 0.01 0.13 0.13 0.05 0.12 0-4 0-4 0 0-0.3 0-3 0-2 0-0.8 0-2

Other 1.14 0.46 1.28 0.22 2.68 0.11 1.85 0.02 0.51 0.23 0.88 0.21 1.26 0.07 0.88 0.01 0-23 0-12 0-15 0-4 0-25 0-2 0-18 0

Litter 69.43 82.65 73.78 82.81 74.50 83.25 67.11 84.29 2.30 1.58 4.75 3.94 2.88 2.72 3.45 2.42 25-94 42-97 22-94 36-94 35-93 45-96 35-95 50-99

Herb. cover 21.49 2.65 7.39 1.39 14.08 9.58 12.38 5.83 (forb+grass) 4.26 0.83 2.79 1.11 3.79 4.17 3.32 1.40 1 center plot 0-100 0-38 0-40 0-20 0-75 0-81 0-66 0-22

Herbaceous 14.96 2.54 5.29 1.76 9.82 5.31 12.38 4.58 cover (four 2.35 0.62 1.64 0.76 1.49 1.80 2.07 1.04 peripheral) 0-51 0-21 0-23 0-12 0-26 0-30 1-33 0-20

Forb taxa 2.72 1.28 4.33 0.92 2.67 2.63 4.42 2.50 2013-2014 0.27 0.22 0.46 0.32 0.30 0.35 0.47 0.46 (#) 0-9 0-10 0-9 0-9 0-7 0-9 1-8 0-8

320 Table 6.6a

Values of environmental variables measured in 400m2 plots Values for structural, compositional, and topographical variables obtained at 400m2 plots at turtle points and random points in Virginia and West Virginia during June-August 2011-2014 (numbers of herbaceous taxa are from 2011-2013); reported in descending order are means, standard errors, and ranges. See Table 6.2 for description of variables. FWT = female Wood Turtles, FRP = female random points, MWT = male Wood Turtles, MRP = male random points. Virginia West Virginia

Variable FWT FRP MWT MRP FWT FRP MWT MRP Large trees 6.7 7.6 6.2 6.5 5.7 6.7 5.8 6.5 (#) 0.27 0.31 0.44 0.45 0.33 0.37 0.48 0.43 0-13 0-20 1-15 0-15 1-14 0-14 0-15 1-16

Medium 10.1 12.4 8.7 13.4 14.5 15.0 13.2 13.6 trees 0.60 0.60 0.80 1.07 0.82 0.88 0.86 1.13 (#) 0-39 0-36 0-32 0-38 2-33 5-48 2-28 2-36

Eye-level 26.0 18.4 32.0 20.6 27.5 19.8 29.5 21.6 obscurity 1.9 1.6 3.26 2.54 1.83 1.54 2.61 1.90 (%) 0-96 0-99 2-100 0-89 0-66 0-64 0-98 1-44

Snags 2.0 1.6 1.6 1.6 2.6 2.5 1.9 2.3 (#) 0.15 0.11 0.20 0.25 0.23 0.24 0.23 0.32 0-9 0-5 0-7 0-9 0-10 0-13 0-7 0-10

Canopy 20.1 13.6 20.0 11.9 19.8 17.9 19.1 15.0 openness 1.10 0.53 1.73 0.38 1.26 1.51 1.19 0.82 (%) 7-81 4-48 6-64 6-20 7-74 9-88 10-50 9-42

Stand age 98.1 88.6 101.4 92.2 109.9 109.6 106.2 112.2 (years) 2.98 3.34 4.40 5.26 1.33 1.48 2.18 1.55 3-142 5-162 6-142 6-142 75-141 75-142 74-123 90-131

LWD10 2.3 1.6 2.0 1.6 1.9 1.1 1.7 1.2 (#) 0.15 0.16 0.21 0.28 0.15 0.13 0.22 0.19 0-11 0-9 0-9 0-11 0-6 0-4 0-8 0-5

LWD4 3.3 4.0 2.7 3.5 3.2 3.1 4.3 4.0 (#) 0.27 0.32 0.30 0.28 0.27 0.22 0.54 0.38 0-16 0-27 0-9 0-11 0-13 0-9 0-18 0-13

Gap size 41.1 11.9 39.1 10.3 27.7 17.6 40.6 18.4 (m2) 5.35 2.69 6.70 3.79 3.8 4.2 8.1 4.7 0-352 0-255 0-200 0-175 0-168 0-192 0-225 0-186

Turtle-level 94.0 77.7 94.5 77.3 78.6 67.8 82.0 61.0 obscurity 0.86 1.51 1.15 2.40 1.77 2.34 1.95 3.28 (%) 46-100 25-100 54-100 31-100 30-100 12-100 51-100 14-100

Tree taxa 5.4 5.7 5.8 5.7 5.6 5.3 5.5 5.1 (#) 0.15 0.14 0.23 0.26 0.14 0.17 0.24 0.32 1-11 1-10 2-9 2-12 3-9 3-9 3-10 3-8

Shrub taxa 10.4 8.3 10.8 8.6 8.1 7.0 8.6 7.6 (#) 0.31 0.25 0.52 0.48 0.28 0.23 0.35 0.32 4-19 2-18 4-18 4-19 3-15 2-13 4-16 4-13

Seedling 9.1 9.4 8.3 8.8 6.1 5.0 5.5 5.2 taxa 0.30 0.27 0.36 0.35 0.33 0.31 0.51 0.41 (#) 1-19 3-19 0-14 1-14 0-15 0-12 0-17 0-12

Herbaceous 14.3 10.2 16.1 9.9 16.7 11.4 16.8 15.1 taxa 0.71 0.64 1.33 1.28 0.95 0.98 1.26 1.23 (#) 0-31 0-35 1-30 2-35 5-33 0-29 5-31 1-29

Distance to 117.8 170.5 69.6 123.3 93.3 139.4 54.0 107.3 stream 9.7 9.5 10.12 9.06 14.5 11.8 13.3 8.87 (m) 5-502 1-572 3-289 9-287 5-788 16-523 3-650 16-273

Slope 6.3 9.7 4.8 9.8 12.3 19.6 13.0 20.8 (degrees) 0.42 0.41 0.61 0.77 1.06 0.89 1.34 1.45 1-31 1-28 1-23 2-26 1-32 2-36 1-33 2-38

Aspect 1.1 1.0 1.2 1.0 0.8 0.9 0.6 1.1 0.05 0.05 0.08 0.08 0.08 0.09 0.09 0.10 0-2 0-2 0-2 0-2 0-2 0-2 0-2 0-2

Elevation 519.7 534.1 508.0 513.8 333.3 347.2 326.6 337.4 (m) 3.2 3.0 4.52 3.56 2.02 2.11 2.90 2.92 428-607 436-607 437-608 446-571 280-368 302-394 266-362 279-380

321 Table 6.6b

Values of environmental variables measured in 400m2 plots – pooled data Values for structural, compositional, and topographic variables obtained at 400m2 plots at turtle points and random points at pooled Virginia and West Virginia sites during June-August 2011-2014 (numbers of herbaceous taxa are from 2011-2013); reported in descending order for each variable are means, standard errors, and ranges. See Table 6.2 for description of variables. FWT = female Wood Turtle points (n = 218), FRP = female random points (n = 218), MWT = male Wood Turtle points (n = 102), MRP = male random points (n = 102), WT = Wood Turtle points (n = 320), RP = random points (n = 320). Virginia and West Virginia

Variable FWT FRP MWT MRP WT RP Large trees 6.4 7.3 6.0 6.5 6.3 7.0 (#) 0.22 0.24 0.34 0.31 0.18 0.19 0-14 0-20 1-15 0-16 0-15 0-20

Medium 11.5 13.2 10.9 13.3 11.3 13.2 trees 0.52 0.50 0.64 0.77 0.41 0.42 (#) 0-39 0-48 0-32 0-38 0-39 0-48

Eye-level 26.8 18.9 31.0 21.1 28.1 19.6 obscurity 1.46 1.17 2.13 1.59 1.21 0.94 (%) 0-96 0-99 2-100 0-89 0-100 0-99

Snags 2.2 1.9 1.7 1.9 2.1 1.9 (#) 0.13 0.11 0.16 0.20 0.10 0.10 0-10 0-13 0-7 0-10 0-10 0-13

Canopy 20.0 15.1 19.6 13.4 19.9 14.5 openness 0.88 0.63 1.11 0.46 0.70 0.46 (%) 7-81 4-88 6-64 6-42 6-81 4-88

Stand age 102.1 95.7 103.7 101.8 102.6 97.7 (years) 2.05 2.36 2.51 2.99 1.61 1.87 3-142 5-162 6-142 6-142 3-142 5-162

LWD10 2.1 1.5 1.8 1.4 2.0 1.4 (#) 0.12 0.12 0.16 0.17 0.09 0.10 0-11 0-9 0-9 0-11 0-11 0-11

LWD4 3.2 3.7 3.4 3.7 3.3 3.7 (#) 0.20 0.23 0.32 0.31 0.17 0.18 0-16 0-27 0-18 0-14 0-18 0-27

Gap size 37.7 13.8 39.3 14.2 38.2 13.9 (m2) 3.89 2.29 5.33 2.96 3.14 1.82 0-352 0-255 0-225 0-186 0-352 0-255

Turtle-level 88.9 74.3 88.2 69.6 88.7 72.8 obscurity 0.86 1.31 1.30 2.15 0.78 1.13 (%) 30-100 12-100 51-100 14-100 30-100 12-100

Tree taxa 5.5 5.6 5.6 5.6 5.5 5.5 (#) 0.11 0.11 0.17 0.16 0.09 0.09 1-11 1-10 2-9 2-12 1-11 1-12

Shrub taxa 9.6 7.9 9.8 8.1 9.6 7.9 (#) 0.24 0.19 0.34 0.30 0.20 0.16 3-19 2-18 4-18 4-19 3-19 2-19

Seedling 8.1 7.9 7.0 7.0 7.8 7.6 taxa 0.25 0.25 0.34 0.32 0.20 0.20 (#) 0-19 3-19 0-17 0-14 0-19 0-19

Herbaceous 15.0 10.6 16.5 12.5 15.4 11.1 taxa 0.58 0.54 0.91 0.94 0.49 0.47 (#) 0-33 0-35 1-31 1-35 0-33 0-35

Distance to 110.0 161.0 57.1 116.4 93.2 146.8 stream 8.15 7.57 6.26 6.41 6.05 5.66 (m) 5-502 1-572 3-289 9-287 3-502 1-572

Slope 8.3 13.1 8.6 15.2 8.4 13.8 (degrees) 0.50 0.51 0.82 0.97 0.43 0.47 1-31 1-36 1-33 2-38 1-33 1-38

Aspect 1.0 1.0 0.9 1.0 0.95 1.0 0.05 0.05 0.07 0.07 0.04 0.04 0-2 0-2 0-2 0-2 0-2 0-2

Elevation 456.6 470.7 419.8 429.0 444.9 457.4 (m) 6.42 6.35 9.33 9.05 5.37 5.31 280-607 302-607 266-608 279-571 266-608 279-607

322 Table 6.7

Well-supported conditional logistic regression models – using cover variables in 1m2 plots Best of fifteen conditional logistic regression models of Wood Turtle microhabitat selection at sites in Virginia and West Virginia, USA in 2011-2014, based on proportions of habitat variables in 1m2 plots. LogLik = model log-likelihood; K = number of parameters; ∆AICc= difference in Akaike Information Criterion corrected for small sample size from the top model; AICc weight denotes a model’s level of support among the set of 15 candidate models; cumulative weight (Cum. wt.) is the running sum of the individual model weights (top models are those with a cumulative weight ≥ 0.95). See Table 6.1 for definitions of variables.

Model LogLik K ∆AICc AICc wt. Cum.wt.

Virginia females Sand+Rock+Forb+Grass+CWD+Woody+Moss+Soil -57.58 8 0 0.49 0.49 Forb+Grass+CWD+Woody+Moss+Soil -59.89 6 0.38 0.41 0.90 1 [Null model] -97.73 0 63.77 0.00 1.00 Virginia males Forb+Grass+CWD+Woody+Moss+Soil -12.42 6 0 0.42 0.42 Rock+Grass+CWD+Woody -15.43 4 1.56 0.19 0.61 Rock+Forb+Grass+CWD+Woody+Moss+Soil -12.42 7 2.29 0.13 0.74 1 [Null model] -36.04 0 34.38 0.00 1.00 West Virginia females Forb+Grass+CWD+Soil -41.32 4 0 0.55 0.55 Sand+Rock+Forb+Grass+CWD+Soil -40.66 6 3.00 0.12 0.68 Sand+Forb+Woody+CWD -42.94 4 3.25 0.11 0.79 1 [Null model] -50.60 0 10.28 0.00 0.99 West Virginia males Sand+Rock+Forb+Grass+Woody+Moss -24.44 6 0 0.18 0.18 Rock+Grass+CWD+Woody -26.78 4 0.19 0.17 0.35 Sand+Rock+Forb+Grass -27.24 4 1.11 0.10 0.45 Sand+Rock+Woody+Moss -27.31 4 1.25 0.10 0.55 Forb+Grass+Woody+Moss -27.31 4 1.25 0.10 0.65 Rock+Grass+Moss+Soil -27.32 4 1.25 0.10 0.75 1 [Null model] -33.96 0 6.12 0.01 1.00

323 Table 6.8

Best conditional logistic regression model variables – cover variables in 1m2 plots Conditional logistic regression model variables that best explain microhabitat selection by Wood Turtles at sites in Virginia and West Virginia, USA in 2011-2014. Measured values are percentages of coverage in 1m2 sampling plots. Model coefficient values were obtained through model averaging. Variables with positive values for coefficients are preferred, while negative values indicate avoidance. Variables in bold denote those with coefficients that did not overlap zero. See Table 6.1 for definitions of variables.

Variable Measured values (mean ± se) Model coefficient ± se Odds ratio Unit increase Turtle Plot Random Plot Virginia females Forb 6.25 ± 1.12 1.51 ± 0.29 0.12 ± 0.06 1.127 1 % Grass 8.36 ± 1.49 1.52 ± 0.61 0.09 ± 0.03 1.094 1 % Woody 4.74 ± 0.59 3.30 ± 0.23 0.26 ± 0.08 1.297 1 % CWD 4.67 ± 0.56 1.79 ± 0.33 0.16 ± 0.04 1.174 1 %

Virginia males Forb 5.35 ± 0.79 0.79 ± 0.35 0.23 ± 0.17 1.259 1 % Grass 8.83 ± 2.11 0.44 ± 0.25 0.51 ± 0.28 1.665 1 % Woody 4.54 ± 0.60 3.58 ± 0.38 0.29 ± 0.13 1.336 1 % CWD 4.96 ± 1.20 1.90 ± 0.79 0.07 ± 0.05 1.073 1 % Soil 1.81 ± 0.98 1.42 ± 1.25 -0.08 ± 0.05 0.923 1 %

West Virginia females Forb 4.01 ± 0.52 2.23 ± 0.45 0.25 ± 0.10 1.284 1 % Grass 10.34 ± 1.49 5.88 ± 1.69 0.02 ± 0.01 1.020 1 % CWD 2.56 ± 0.55 1.51 ± 0.34 0.12 ± 0.07 1.127 1 % Soil 0.65 ± 0.20 0.36 ± 0.15 0.16 ± 0.13 1.174 1 %

West Virginia males Forb 4.59 ± 0.63 3.22 ± 0.67 0.09 ± 0.07 1.094 1 % Grass 12.55 ±2.41 4.76 ± 1.17 0.04 ± 0.03 1.041 1 % Woody 2.22 ± 0.51 1.51 ± 0.29 0.12 ± 0.10 1.127 1 % CWD 2.92 ± 0.95 1.12 ± 0.35 0.09 ± 0.06 1.094 1 % Moss 0.61 ± 0.19 2.63 ± 1.26 -0.18 ± 0.15 0.835 1 % Rock 0.16 ± 0.07 0.41 ± 0.17 -0.70 ± 0.51 0.497 1 %

324 Table 6.9

Well-supported conditional logistic regression models – using variables in 400m2 plots 2011-2014 Best of twenty-two conditional logistic regression models of Wood Turtle meso-scale habitat selection at sites in Virginia and West Virginia, USA in 2011-2014, based on habitat variables in 400m2 plots. LogLik = model log-likelihood; K = number of parameters; ∆AICc= difference in Akaike Information Criterion corrected for small sample size from the top model; AICc weight denotes a model’s level of support among the set of 22 candidate models; cumulative weight (Cum. wt.) is the running sum of the individual model weights (top models herein are those with a cumulative weight ≥ .95). See Table 6.2 for definitions of variables.

Model LogLik K ∆AICc AICc wt. Cum.wt.

Virginia females Obs4+LWD10+LWD4+Gapsizem2+Treespp+ Shrubspp+Dist+ASPB+Slope+Elev -35.43 10 0 0.88 0.88 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+ LWD4+Snags+Dist+ASPB+Slope+Elev -35.79 12 5.05 0.07 0.95 1 [Null model] -99.81 0 108.0 0.00 1.00

Virginia males Obs4+LWD10+LWD4+Gapsizem2+Seedspp+ Dist+ASPB+Slope+Elev -5.80 9 0 0.29 0.29 Large+Medium+Obs4+LWD10+LWD4+ Snags+Can+Treespp+Shrubspp+Seedspp -5.03 10 0.90 0.19 0.48 Obs4+LWD10+LWD4+Gapsizem2+Slope+ ASPB+Dist+Elev -7.60 8 1.20 0.16 0.64 Large+Medium+Obs1+Obs4+Gapsizem2+ LWD10+LWD4+Snags+Can -6.43 9 1.26 0.16 0.79 1 [Null model] -36.74 0 42.00 0.00 1.00

West Virginia females Large+Medium+Obs1+Snags+Can+Treespp+ Shrubspp+Age+Dist+ASPB+Slope+Elev -9.99 12 0 0.93 0.93 Large+Medium+Obs4+LWD10+LWD4+Snags+ Gapsizem2+Age+Slope+Elev+ASPB+Dist -13.10 12 6.22 0.04 0.97 1 [Null model] -51.29 0 56.29 0.00 1.00

West Virginia males Obs4+LWD10+LWD4+Gapsizem2+Seedspp+ Dist+ASPB+Elev -6.83 8 0 0.72 0.72 Seedspp+Dist+ASPB+Elev -13.13 4 3.43 0.13 0.86 1 [Null model] -33.96 0 36.66 0.00 1.00

325 Table 6.10

Best conditional logistic regression model variables in 400m2 plots 2011-2014 Conditional logistic regression model variables that best explain meso-scale habitat selection by Wood Turtles at sites in Virginia and West Virginia, USA in 2011-2014. Measured values were obtained in 400m2 sampling plots. Model coefficients were obtained through model averaging. Variables with positive values for coefficients are preferred, while negative values indicate avoidance. Variables in bold denote those with coefficients that did not overlap zero.

Variable Measured values (mean ± se) Model coefficient ± se Odds ratio Unit increase Turtle Plot Random Plot Virginia females Medium 10.1 ± 0.60 12.4 ± 0.60 0.06 ± 0.05 1.062 1 tree Tree spp. 5.4 ± 0.15 5.7 ± 0.14 -0.41 ± 0.15 0.667 1 spp. Shrub spp. 10.4 ± 0.31 8.3 ± 0.25 0.13 ± 0.09 1.139 1 spp. Obs4 94.0 ± 0.86 77.7 ± 1.51 0.11 ± 0.03 1.116 1 % LWD10 2.3 ± 0.15 1.6 ± 0.16 0.31 ± 0.18 1.363 1 piece Snags 2.0 ± 0.15 1.6 ± 0.11 0.47 ± 0.16 1.600 1 snag Canopy 20.1 ± 1.10 13.6 ± 0.53 0.06 ± 0.03 1.062 1 % Slope 6.3 ± 0.42 9.7 ± 0.41 -0.16 ± 0.06 0.852 1 º Elevation 519.7 ± 3.2 534.1 ± 3.0 -0.05 ± 0.02 0.951 1 meter Aspect 1.1 ± 0.05 1.0 ± 0.05 -0.62 ± 0.48 0.538 .1 unit

Virginia males [no variables had strong support]

West Virginia females Medium 14.5 ± 0.82 15.0 ± 0.88 0.29 ± 0.21 1.336 1 tree Obs1 27.5 ± 1.49 19.8 ± 1.69 0.51 ± 0.34 1.665 1 % Obs4 78.6 ± 1.77 67.8 ± 2.34 0.11 ± 0.06 1.116 1 % Shrub spp. 8.1 ± 0.28 7.0 ± 0.23 4.96 ± 3.31 1.642 .1 spp. LWD10 1.9 ± 0.15 1.1 ± 0.13 0.84 ± 0.57 2.316 1 piece Snags 2.6 ± 0.23 2.5 ± 0.24 1.40 ± 1.05 4.055 1 snag Aspect 0.8 ± 0.08 0.9 ± 0.09 3.73 ± 2.81 1.452 .1 unit Elevation 333.3 ± 2.02 347.2 ± 2.11 -0.36± 0.24 0.698 1 meter

West Virginia males Large 5.8 ± 0.48 6.5 ± 0.43 -0.49 ± 0.41 0.613 1 tree Snags 1.9 ± 0.23 2.3 ± 0.32 -0.62 ± 0.44 0.538 1 snag Slope 6.3 ± 0.42 9.7 ± 0.41 -0.12 ± 0.10 0.887 1 º

326 Table 6.11

Well-supported conditional logistic regression models – using variables in 400m2 plots, including number of herbaceous taxa 2011-2013 Best of twenty-two conditional logistic regression models of Wood Turtle meso-scale habitat selection at sites in Virginia and West Virginia, USA in 2011-2013, based on habitat variables in 400m2 plots, including number of herbaceous taxa. LogLik = model log-likelihood; K = number of parameters; ∆AICc= difference in Akaike Information Criterion corrected for small sample size from the top model; AICc weight denotes a model’s level of support among the set of 22 candidate models; cumulative weight (Cum. wt.) is the running sum of the individual model weights.

Model LogLik K ∆AICc AICc wt. Cum.wt.

Virginia females Obs4+LWD10+LWD4+Gapsizem2+Treespp+ Shrubspp+Dist+ASPB+Slope+Elev -29.79 10 0 0.78 0.78 Large+Medium+Obs4+Obs1+Gapsizem2+LWD10+ LWD4+Snags+Can+Dist+ASPB+Slope+Elev -27.95 13 2.99 0.17 0.95 1 [Null model] -81.79 0 83.03 0.00 1.00

Virginia males Large+Medium+Obs4+LWD10 -2.24 4 0 0.99 0.99 Seedspp+Herb+Dist+ASPB+Elev -6.56 5 11.0 0.01 1.00 1 [Null model] -22.87 0 32.60 0.00 1.00 West Virginia females Obs4+LWD10+LWD4+Gapsizem2+ASPB+ Dist+Slope -4.26 7 0 0.95 0.95 Dist+ASPB+Slope+Elev -11.72 4 8.06 0.02 0.97 1 [Null model] -32.58 0 41.33 0.00 1.00

West Virginia males Obs4+Gapsizem2+Shrubspp+ASPB+Slope -4.07 5 0 0.77 0.77 Obs4+LWD10+Gapsizem2+Dist+ASPB+Elev -5.11 6 4.52 0.08 0.85 1 [Null model] -22.18 0 25.18 0.00 1.00

327 Table 6.12

Best conditional logistic regression model variables in 400m2 plots 2011-2013, including number of herbaceous taxa Conditional logistic regression model variables that best explain meso-scale habitat selection by Wood Turtles at sites in Virginia and West Virginia, USA in 2011-2013. Measured values, including number of herbaceous taxa present, were obtained in 400m2 sampling plots. Model coefficients were obtained through model averaging. Variables with positive values for coefficients are preferred, while negative values indicate avoidance. Variables in bold denote those with coefficients that did not overlap zero.

Variable Measured values (mean ± se) Model coefficient ± se Odds ratio Unit increase Turtle Plot Random Plot Virginia females Medium 10.3 ± 0.66 11.9 ± 0.61 0.06 ± 0.05 1.062 1 tree Obs4 93.1 ± 1.03 77.3 ± 1.68 0.11 ± 0.04 1.116 1 % Tree spp. 5.4 ± 0.16 5.9 ± 0.15 -0.46 ± 0.17 0.631 1 spp. Shrub spp. 10.4 ± 0.35 8.4 ± 0.28 0.16 ± 0.11 1.174 1 spp. Snags 2.0 ± 0.17 1.7 ± 0.12 0.52 ± 0.20 1.682 1 snag Canopy 19.8 ± 1.29 13.7 ± 0.50 0.05 ± 0.04 1.051 1 % Aspect 1.1 ± 0.06 1.0 ± 0.06 -0.82 ± 0.52 0.921 .1 unit Slope 6.3 ± 0.49 10.1 ± 0.46 -0.16 ± 0.06 0.852 1 º Elevation 521.0 ± 3.59 534.8 ± 3.24 -0.05 ± 0.02 0.951 1 meter

Virginia males Seed spp. 7.4 ± 0.50 8.2 ± 0.45 -0.44 ± 0.26 0.644 1 spp. Herb. spp. 16.1 ± 1.33 9.9 ± 1.28 -0.13 ± 0.08 0.878 1 spp. Canopy 20.9 ± 2.46 12.2 ± 0.45 0.30 ± 0.25 1.350 1 % Dist 51.1 ± 10.9 132.2 ± 13.0 -0.05 ± 0.03 0.951 1 meter

West Virginia females Herb spp. 16.7 ± 0.95 11.4 ± 0.98 0.32 ± 0.29 1.377 1 spp. Snags 2.8 ± 0.34 2.6 ± 0.33 1.29 ± 1.04 3.633 1 snag

West Virginia males Snags 2.2 ± 0.30 2.2 ± 0.44 -1.87 ± 1.20 0.154 1 snag Obs4 81.7 ± 2.61 58.4 ± 3.89 0.56 ± 0.61 1.751 1 %

328 Table 6.13

Well-supported conditional logistic regression models – using variables in 400m2 plots and pooled data 2011-2014 Best of thirty-one conditional logistic regression models of Wood Turtle meso-scale habitat selection at combined sites in Virginia and West Virginia, USA in 2011-2014, based on habitat variables in 400m2 plots, including importance values of trees. LogLik = model log-likelihood; K = number of parameters; ∆AICc= difference in Akaike Information Criterion corrected for small sample size from the top model; AICc weight denotes a model’s level of support among the set of 31 candidate models; cumulative weight (Cum. wt.) is the running sum of the individual model weights.

Model LogLik K ∆AICc AICc wt. Cum.wt.

Virginia and West Virginia females Obs4+LWD10+LWD4+Gapsizem2+ Treespp+Shrubspp+Dist+ASPB+Slope -69.31 9 0 0.91 0.91 Obs4+LWD10+LWD4+Gapsizem2+ Slope+ASPB+Dist -74.24 7 5.72 0.05 0.97 1 [Null model] -151.11 0 146.4 0.00 1.00

Virginia and West Virginia males Large+Medium+Obs4+Obs1+LWD10+LWD4+ Snags+Can+Gapsizem2+Age+Slope+ASPB+Dist -14.42 13 0 0.93 0.93 Obs4+LWD10+LWD4+Gapsizem2+ Slope+ASPB+Dist -24.18 7 6.18 0.04 0.97 1 [Null model] -70.70 0 84.64 0.00 1.00

Virginia and West Virginia Wood Turtles Obs4+LWD10+LWD4+Gapsizem2+Treespp+ Shrubspp+Slope+ASPB+Dist -98.42 9 0 0.59 0.59 Obs4+LWD10+LWD4+Gapsizem2+ Slope+ASPB+Dist -101.23 7 1.50 0.28 0.87 1 [Null model] -221.81 0 228.5 0.00 1.00

329 Table 6.14

Best conditional logistic regression model variables in 400m2 plots, using pooled data 2011-2014 Conditional logistic regression model variables that best explain meso-scale habitat selection by Wood Turtles at combined sites in Virginia and West Virginia, USA in 2011-2014. Measured values were obtained in 400m2 sampling plots. Model coefficients were obtained through model averaging. Variables with positive values for coefficients are preferred, while negative values indicate avoidance. Variables in bold denote those with coefficients that did not overlap zero.

Variable Measured values (mean ± se) Model coefficient ± se Odds ratio Unit increase Turtle Plot Random Plot VA & WV females Medium 11.5 ± 0.50 13.2 ± 0.50 0.03 ± 0.03 1.03 1 tree Obs1 26.8 ± 1.46 18.9 ± 1.17 0.02 ± 0.01 1.020 1 % Obs4 88.9 ± 0.97 74.3 ± 1.31 0.06 ± 0.01 1.062 1 % Tree spp. 5.5 ± 0.11 5.6 ± 0.11 -0.26 ± 0.10 0.771 1 spp. Shrub spp. 9.6 ± 0.24 7.9 ± 0.19 0.10 ± 0.06 1.105 1 spp. LWD10 2.1 ± 0.12 1.5 ± 0.12 0.27 ± 0.11 1.310 1 piece Snags 2.2 ± 0.13 1.9 ± 0.11 0.11 ± 0.08 1.116 1 snag Slope 8.3 ± 0.50 13.1 ± 0.51 -0.13 ± 0.03 0.878 1 º Aspect 1.0 ± 0.05 1.0 ± 0.05 -0.57 ± 0.28 0.945 0.1 unit WO 18.8 ± 1.37 14.2 ± 1.17 0.02 ± 0.02 1.020 1 unit RM 11.8 ± 0.95 10.6 ± 0.90 0.03 ± 0.02 1.030 1 unit SM 4.1 ± 0.70 1.5 ± 0.37 0.04 ± 0.03 1.041 1 unit TP 7.3 ± 1.14 6.3 ± 1.26 0.03 ± 0.02 1.030 1 unit

VA & WV males Large 6.0 ± 0.34 6.5 ± 0.31 0.38 ± 0.20 1.462 1 tree Medium 10.9 ± 0.64 13.3 ± 0.77 -0.14 ± 0.10 0.869 1 tree Obs4 88.2 ± 1.30 69.6 ± 2.15 0.19 ± 0.08 1.209 1 % Can 19.6 ± 1.11 13.4 ± 0.46 0.25 ± 0.12 1.284 1 % Dist 57.1 ± 6.26 116.4 ± 6.41 -0.03 ± 0.01 0.970 1m Slope 8.6 ± 0.82 15.2 ± 0.97 -0.15 ± 0.07 0.861 1 º WP 16.8 ± 2.02 18.9 ± 2.33 -0.09 ± 0.07 0.914 1 unit VP 8.8 ± 1.73 7.6 ± 1.61 -0.12 ± 0.08 0.887 1 unit SM 4.7 ± 0.96 3.8 ± 0.85 -0.20 ± 0.13 0.819 1 unit SO 2.5 ± 0.96 6.9 ± 1.46 -0.10 ± 0.08 0.905 1 unit CO 3.8 ± 0.95 11.1 ± 1.64 -0.23 ± 0.11 0.795 1 unit TP 3.6 ± 0.95 1.0 ± 0.43 -0.18 ± 0.12 0.835 1 unit

VA & WV Wood Turtles Obs4 88.7 ± 0.78 72.8 ± 1.13 0.07 ± 0.01 1.073 1 % Tree spp. 5.5 ± 0.09 5.5 ± 0.09 -0.15 ± 0.08 0.861 1 spp. Shrub spp. 9.6 ± 0.20 7.9 ± 0.16 0.08 ± 0.05 1.083 1 spp. LWD10 2.0 ± 0.09 1.4 ± 0.10 0.14 ± 0.08 1.150 1 piece Can 19.9 ± 0.70 14.5 ± 0.46 0.03 ± 0.02 1.030 1 snag Slope 8.4 ± 0.43 13.8 ± 0.47 -0.11 ± 0.02 0.890 1 º Aspect 1.0 ± 0.04 1.0 ± 0.04 -0.43 ± 0.22 0.958 0.1 unit SM 4.3 ± 0.57 2.2 ± 0.37 0.04 ± 0.03 1.041 1 unit TP 6.1 ± 0.84 4.6 ± 0.88 0.02 ± 0.02 1.020 1 unit ______

330 Table 6.15

Well-supported conditional logistic regression models – using variables in 400m2 plots, including number of herbaceous taxa and IVs, pooled data 2011-2013 Best of thirty-one conditional logistic regression models of Wood Turtle meso-scale habitat selection at combined sites in Virginia and West Virginia, USA in 2011-2013, based on habitat variables in 400m2 plots, including number of herbaceous taxa present and importance values of trees. LogLik = model log-likelihood; K = number of parameters; ∆AICc= difference in Akaike Information Criterion corrected for small sample size from the top model; AICc weight denotes a model’s level of support among the set of 31 candidate models; cumulative weight (Cum. wt.) is the running sum of the individual model weights.

Model LogLik K ∆AICc AICc wt. Cum.wt.

Virginia and West Virginia females Obs4+LWD10+LWD4+Gapsizem2+ Treespp+Shrubspp+Dist+ASPB+Slope -51.05 9 0 0.95 0.95 Obs4+LWD10+LWD4+Gapsizem2+ Slope+ASPB+Dist -56.78 7 7.24 0.03 0.97 1 [Null model] -114.37 0 108.1 0.00 1.00

Virginia and West Virginia males Obs4+LWD10+LWD4+Gapsizem2+ Slope+ASPB+Dist -8.10 7 0 0.88 0.88 Obs4+LWD10+LWD4+Gapsizem2+Herb+ Seedspp+Slope+ASPB+Dist -7.95 9 4.28 0.10 0.98 1 [Null model] -45.05 0 59.00 0.00 1.00

Virginia and West Virginia Wood Turtles Obs4+LWD10+LWD4+Gapsizem2+Treespp+ Shrubspp+Slope+ASPB+Dist -67.53 9 0 0.35 0.35 Obs4+LWD10+LWD4+Gapsizem2+ Slope+ASPB+Dist -69.88 7 0.54 0.27 0.62 WO+WP+RM+HICK+VP+CO+NRO+TP+SM+ WA+BG+SO+Obs4+LWD10+LWD4+Gapsizem2+ Shrubspp+Slope+ASPB+Dist -56.40 20 1.25 0.19 0.80 Obs4+LWD10+LWD4+Gapsizem2+ Herb+Seedspp+Slope+ASPB+Dist -68.47 9 1.88 0.14 0.94 1 [Null model] -159.42 0 165.4 0.00 1.00 ______

331 Table 6.16

Best conditional logistic regression model variables in 400m2 plots, including number of herbaceous taxa and IVs, using pooled data 2011-2013 Conditional logistic regression model variables that best explain meso-scale habitat selection by Wood Turtles at combined sites in Virginia and West Virginia, USA in 2011-2013. Measured values were obtained in 400m2 sampling plots and include number of herbaceous taxa present. Model coefficients were obtained through model averaging. Variables with positive values for coefficients are preferred, while negative values indicate avoidance. Variables in bold denote those with coefficients that did not overlap zero.

Variable Measured values (mean ± se) Model coefficient ± se Odds ratio Unit increase Turtle Plot Random Plot VA & WV females Obs4 88.2 ±1.16 74.7 ± 1.51 0.05 ± 0.02 1.051 1 % Tree spp. 5.6 ± 0.13 5.8 ± 0.12 -0.34 ± 0.12 0.712 1 spp. Shrub spp. 10.0 ± 0.27 8.2 ± 0.22 0.16 ± 0.09 1.174 1 spp. Herb spp. 15.0 ± 0.58 10.6 ± 0.54 0.04 ± 0.03 1.041 1 spp. LWD10 2.2 ± 0.14 1.6 ± 0.14 0.23 ± 0.13 1.259 1 piece Snags 2.3 ± 0.16 1.9 ± 0.13 0.16 ± 0.09 1.174 1 snag Aspect 1.0 ± 0.06 1.0 ± 0.05 -0.98 ± 0.35 0.907 .1 unit WO 18.7 ± 1.49 14.4 ± 1.30 0.03 ± 0.03 1.030 1 unit RM 13.4 ± 1.15 12.0 ± 1.10 0.04 ± 0.03 1.041 1 unit SM 4.3 ± 0.82 0.9 ± 0.33 0.13 ± 0.07 1.139 1 unit WA 5.7 ± 0.89 1.1 ± 0.26 0.07 ± 0.06 1.073 1 unit

VA & WV males Large 5.9 ± 0.40 6.9 ± 0.41 0.37 ± 0.32 1.448 1 tree Obs4 88.3 ± 1.71 67.8 ± 2.15 0.20 ± 0.10 1.221 1 % Tree spp. 5.6 ± 0.21 5.4 ± 0.18 0.44 ± 0.32 1.552 1 spp. Canopy 20.0 ± 1.45 13.3 ± 0.56 0.23 ± 0.15 1.259 1 % Slope 8.0 ± 0.96 15.6 ± 1.26 -0.11 ± 0.09 0.896 1 º WP 14.4 ± 2.06 21.7 ± 2.99 -0.11 ± 0.08 0.896 1 unit RM 9.1 ± 1.22 10.1 ± 1.56 -0.15 ± 0.12 0.861 1 unit SM 5.2 ± 1.37 5.1 ± 1.25 -0.23 ± 0.16 0.795 1 unit BG 3.3 ± 0.81 1.8 ± 0.77 0.20 ± 0.18 1.221 1 unit NRO 3.8 ± 0.90 5.3 ± 1.05 -0.16 ± 0.13 0.852 1 unit HICK 7.4 ± 1.51 8.9 ± 1.48 -0.13 ± 0.10 0.878 1 unit

VA & WV Wood Turtles Obs4 88.2 ± 0.96 72.8 ± 1.32 0.08 ± 0.02 1.083 1 % Tree spp. 5.6 ± 0.11 5.7 ± 0.10 -0.19 ± 0.10 0.827 1 spp. Shrub spp. 10.1 ± 0.23 8.3 ± 0.19 0.09 ± 0.08 1.094 1 spp. Herb spp. 15.4 ± 0.49 11.1 ± 0.47 0.04 ± 0.03 1.041 1 spp. Slope 7.5 ± 0.45 13.5 ± 0.53 -0.14 ± 0.04 0.869 1 º Aspect 0.9 ± 0.05 1.0 ± 0.05 -0.89 ± 0.41 0.915 .1 unit SM 4.5 ± 0.70 2.1 ± 0.44 0.12 ± 0.05 1.127 1 unit WA 5.5 ± 0.73 1.3 ± 0.28 0.05 ± 0.04 1.051 1 unit ______

332 Table 6.17

Synopsis of best conditional logistic regression model variables in 400m2 plots for various turtle groups in VA and WV Synopsis of conditional logistic regression model variables that best explain Wood Turtle meso-scale habitat selection based on habitat variables in 400m2 sampling plots at sites in Virginia and West Virginia, USA during June-August 2010-2014. Based on model coefficients obtained through model averaging. Variables with exes are those with coefficients that did not overlap zero for that particular site group; an X of positive sign indicates variables that are preferred, while –X indicates avoidance. Variables with bs are those with coefficients that slightly overlapped zero for that particular site group; a b of positive sign indicates variables that are preferred, while a –b indicates avoidance. FT = female Wood Turtles, MT = male Wood Turtles, V = Virginia, W = West Virginia. See Table 6.2 for definition of variables. Type of site Variable VFT VMT WFT WMT FT MT WT

Large trees -b X

Medium b -b trees Obs1 b X X

Snags X b

Canopy b X b

Stand age -b

LWD10 b b X b

LWD4

Gap size

Obs4 X b X X X

Tree taxa -X -X -b

Shrub taxa b b

Seedling taxa Herbaceous b b taxa Distance to -X stream Slope -X -X -X -X

Aspect -X -X

Elevation -X -b

CO -X

SM X X

333 Table 6.18

Well-supported mixed conditional logistic regression models – using variables in 400m2 plots, including stand and plot forest type and seral stage, pooled data 2011-2013 Best of twenty-two conditional logistic regression models of Wood Turtle meso-scale habitat selection at sites in Virginia and West Virginia, USA, based on habitat variables in 400m2 plots in 2011-2013; models included stand seral stage, stand forest type, or plot forest type as a random effect. K = number of parameters; ∆AIC = difference in Akaike Information Criterion size from the top model; only models within 2 units of the top model are listed (or the top two models if only one was well supported). AIC values denote models that are well-supported among the set of 22 candidate models. See Table 6.2 for definitions of variables. Number of plots (for both turtle and random points): for seral stage VA F n = 144, VA M n = 52; for stand forest type F n = 197, M n = 89; for plot forest type F n = 330, M n = 120.

Model AIC K ∆AICc

Virginia females – seral stage Large+Medium+Obs4+Obs1+LWD10+LWD4+ Gapsize+Snags+Slope+ASPB 98.07 10 0 Large+Medium+Obs4+LWD10+LWD4+ Gapsize+Snags+Treespp+Shrubspp+Seedspp 99.57 10 1.50

Virginia males – seral stage Large+Medium+Snags+Can+Seedspp+ASPB+Slope 44.98 7 0 LWD10+Gapsize+ASPB+Slope 47.88 4 2.90

Virginia &West Virginia females – stand forest type (with Herb) Obs4+LWD10+LWD4+Gapsize+Slope+ASPB 109.42 6 0 Large+Medium+Obs4+Obs1+LWD10+LWD4+ Snags+Can+Age+Slope+ASPB 110.19 11 0.77

Virginia & West Virginia males – stand forest type (with Herb) Obs4+LWD10+LWD4+Gapsize+ Herb+Seedspp+ASPB+Dist 32.96 8 0 LWD10+Treespp+Shrubspp+ASPB+Dist 40.44 5 7.48

Virginia &West Virginia females – plot forest type (with Herb) Obs4+LWD10+LWD4+Gapsize+Slope+ASPB 133.35 6 0 Large+Medium+Obs4+Obs1+LWD10+LWD4+ Snags+Can+Age+Slope+ASPB 135.30 11 1.95

Virginia & West Virginia males – plot forest type (with Herb) Obs4+ LWD10+LWD4+Gapsize+ASPB 35.38 5 0 Large+Obs4+LWD10+LWD4+Snags+Can+Age 39.81 7 4.43 ______

334 Table 6.19

Best mixed conditional logistic regression model variables in 400m2 plots, using stand and plot forest type and seral stage, pooled data 2011-2013 Conditional logistic regression model variables that best explain meso-scale habitat selection by Wood Turtles at sites in Virginia and West Virginia, USA in 2011-2014, based on habitat variables in 400m2 plots. Coefficient values are from the well-supported models run with stand forest type, plot forest type, or stand seral stage as random effects (see Table 18). Positive values for coefficients indicate preference, while negative values indicate avoidance. Variables with coefficients that overlapped zero are denoted with an asterisk. See Table 6.2 for definitions of variables. Number of plots (for both turtle and random points): for seral stage VA F n = 144, VA M n = 52; for stand forest type F n = 197, M n = 89; for plot forest type F n = 330, M n = 120.

Variable value se p-value

Virginia females – seral stage Mod1 - Large+Medium+Obs4+Obs1+LWD10+ LWD4+Snags+Gapsizem2+Slope+ASPB Obs4 0.130 0.03 0.00001 Snags 0.483 0.15 0.0011 Slope -0.117 0.05 0.0099 Gapsizem2 0.015 0.01 0.0193 Seral 0.042 0.06 0.6100 Mod2 - Large+Medium+Obs4+LWD10+LWD4+ Snags+Gapsizem2+Treespp+Shrubspp+Seedpp Obs4 0.127 0.03 0.00001 Snags 0.452 0.14 0.0017 Medium 0.089 0.04 0.0381 Gapsizem2 0.018 0.01 0.0195 Shrub spp. 0.213 0.09 0.0198 Seral 0.014 0.01 0.8921

Virginia males – seral stage Large+Medium+Snags+Can+Seedspp+ASPB+Slope Can 0.265 0.12 0.0220 Slope -0.211 0.09 0.0179 Seral 0.400 0.33 0.2872

Virginia &West Virginia females – stand forest type (with Herb) Mod1 Obs4 0.059 0.02 0.00017 LWD10 * 0.218 0.12 0.0705 Gapsizem2 * 0.009 0.005 0.0543 ASPB -0.857 0.34 0.0113 Slope -0.183 0.04 0.00001 Stand 2.088 0.99 0.0031 Mod2 Obs4 0.052 0.02 0.0031 Can 0.045 0.02 0.0420 ASPB -0.881 0.48 0.0182 Slope -0.221 0.05 0.00001 Stand 0.589 0.48 0.1216

Virginia & West Virginia males – stand forest type (with Herb) Obs4 0.148 0.07 0.0432 Dist * -0.028 0.02 0.0695 Stand 0.102 0.04 0.9431

Virginia &West Virginia females – plot forest type (with Herb) Mod1 Obs4 0.056 0.01 0.00009 LWD10 * 0.205 0.11 0.0623 Gapsizem2 * 0.007 0.004 0.0745 Slope -0.160 0.04 0.00001 ASPB -0.718 0.30 0.01629 Plot 0.061 0.06 0.3241 Mod2 Obs4 0.054 0.02 0.0005 LWD10 * 0.203 0.12 0.0964 Can * 0.033 0.02 0.0864 Slope -0.185 0.04 0.00001 Age 0.012 0.004 0.0180 ASPB -0.771 0.33 0.0180 Plot 0.062 0.04 0.4873 Virginia & West Virginia males – plot forest type (with Herb) Mod1 Obs4 0.218 0.07 0.0015 ASPB -1.541 0.66 0.0191 Plot 0.881 0.37 0.1976 ______

335 References

Aarts, G., M. MacKenzie, B. McConnell, M. Fedak and J. Matthiopoulos. 2008. Estimating space-use and habitat preference from wildlife telemetry data. Ecography 31: 140-160.

Abdala, V., A.S. Manzano, and A. Herrel. 2008. The distal forelimb musculature in aquatic and terrestrial turtles: Phylogeny or environmental constraints? Journal of Anatomy 213: 159–172.

Adkison, G.P. and S.K. Gleeson. 2004. Forest understory vegetation along a productivity gradient. Journal of the Torrey Botanical Society 131(1): 32-44

Adolph, S.C. 1990. Influence of behavioral thermoregulation on microhabitat use by two Sceloporous lizards. Ecology 71: 315-327.

Akre, T. 2002. “Growth, Maturity, and Reproduction of the Wood Turtle, Clemmys insculpta (LeConte 1830) in Virginia”. PhD dissertation, George Mason University, Fairfax, VA. 135 pp.

Akre, T. and C. Ernst. 2006. Final Report: Population Dynamics, Habitat Use, and Home Range of the Wood Turtle, Glyptemys (Clemmys) insculpta, in Virginia, submitted to the Virginia Department of Game and Inland Fisheries, contract ed0820ce_080.

Allan, J.D. 1995. Stream Ecology: Structure and Function of Running Waters. Chapman and Hall, New York, NY.

Anderson, K.L. and D.J. Leopold. 2002. The role of canopy gaps in maintaining vascular plant diversity at a forested wetland in New York State. Journal of the Torrey Botanical Society 129(3): 238-250.

Angelstam, P. 2003. “Reconciling the linkages of land management with natural disturbance regimes to maintain forest biodiversity in Europe”, pp. 193-226 in J. Bissonette and I. Storch (eds.), Landscape Ecology and Resource Management: Linking Theory With Practice. Island Press, Washington, D.C. 463 pp.

Angermeier, P.L. 1995. Ecological attributes of extinction-prone species: Loss of freshwater fishes of Virginia. Conservation Biology 9(1): 143-158.

Angier, B. 1974. Field Guide to Edible Wild Plants. Stackpole Books, Harrisburg, PA. 255 pp.

336

Anguiletta, M.J., Jr. 2009. Thermal Adaptation: A Theoretical and Empirical Synthesis. Oxford University Press, England. 289 pp.

Anning, A.K., J.M. Dyer, and B.C. McCarthy. 2014. Tree growth response to fuel reduction treatments along a topographic moisture gradient in mixed-oak forests of Ohio, U.S.A. Canadian Journal of Forest Research 44: 413-421.

Armson, K.A. 1979. Forest Soils: Properties and Processes. University of Toronto Press. Toronto, Ontario.

Arnold, T.W. 2010. Uninformative parameters and model selection using Akaike’s information criterion. Journal of Wildlife Management 74(6): 1175-1178.

Arvisais, M., J.-C. Bourgeois, E. Levesque, C. Daigle, D. Masse, and J. Jutras. 2002. Home range and movements of a Wood Turtle (Clemmys insculpta) population at the northern limit of its range. Canadian Journal of Zoology 80: 402–408.

Arvisais, M., E. Levesque, J.-C. Bourgeois, C. Daigle, D. Masse, and J. Jutras. 2004. Habitat selection by the Wood Turtle (Clemmys insculpta) at the northern limit of its range. Canadian Journal of Zoology 82: 391–398.

Ashe, W.W. 1922. Forest types of the Appalachian and White Mountains. Journal of the Elisha Mitchell Scientific Society 37 (3-4): 183-198.

Ashton, D.T., J.B. Bettaso, and H.H. Welsh, Jr. 2011. Comparative Ecology of Western Pond Turtle (Actinemys marmorata) populations on the Free- flowing South Fork and Regulated Main Fork Trinity River: Demography, Size and Body Condition Comparisons, Thermal Ecology, and Spatial Dynamics. Final Report to the Trinity River Restoration Program. Accessed at http://iims.trrp.net/FileDatabase/Documents/Ashton

Avise, J.C. 1995. Mitochondrial DNA polymorphism and a connection between genetics and demography of relevance to conservation. Conservation Biology 9: 686–690.

Avise, J.C., B.W. Bowen, T. Lamb, A.B. Meylan, and Bermingham, E. 1992. Mitochondrial DNA evolution at a turtle’s pace: Evidence for low genetic variability and reduced microevolutionary rate in the Testudines. Molecular Biology and Evolution 9: 457-473.

Baker, W.L. 2015. Are high-severity fires burning at much higher rates recently

337 than historically in dry-forest landscapes of the Western USA? PLoS ONE 10:e0141936.

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

Barbour, M.G., J.H. Burk, and W.D. Pitts. 1987. Terrestrial Plant Ecology. Benjamin/Cummings Publishing Co., Menlo Park, CA. 634 pp.

Barrett, S.L. 1990. Home range and habitat of the Desert Tortoise (Xerobates agassizi) in the Picacho Mountains of Arizona. Herpetologica 46(2): 2012- 206.

Barzilay, S. 1980. Aggressive behavior in the Wood Turtle, Clemmys insculpta. Journal of Herpetology 14(1): 89-91.

Baxley, D.L. and C.P. Qualls. 2009. The Black Pine Snake (Pituophis melanoleucus lodingi): Spatial ecology and associations between habitat use and prey dynamics. Journal of Herpetology 43:284–293.

Baxley, D., G.J. Lipps, Jr., and C.P. Qualls. 2011. Multiscale habitat selection by Black Pine Snakes (Pituophis melanoleuca lodingi) in southern Mississippi. Herpetologica 67(2): 154-166.

Beatty, S.W. 1984. Influence of microtopography and canopy species on spatial patterns of forest understory plants. Ecology 65: 1406–1419.

Beaupre, S.J. 1995. Effects of geographically variable thermal environment on bioenergetics of mottled rock rattlesnakes, Crotalus lepidus. Ecology 76: 1655–1665.

Beckage, B., J.S. Clark, B.D. Clinton, and B.L. Haines. 2000. A long-term study of tree seedling recruitment in southern Appalachian forests: The effects of canopy gaps and shrub understories. Can. J. For. Res. 30: 1617–1631.

Bellemare, J., G. Motzkin and D.R. Foster. 2002. Legacies of the agricultural past in the forested present: An assessment of historical land-use effects on rich mesic forests. Journal of Biogeography 29: 1401–1420.

Belzer, B. 2000. A nine year study of Eastern Box Turtle courtship with implications for reproductive success and conservation in a translocated population. Turtle and Tortoise Newsletter 6: 19-26.

338 Belzer, W. and S. Seibert. 2009. How do male box turtles find mates? Turtle and Tortoise Newsletter 13: 11–21.

Beers, T.W., P.E. Dress, and L.N. Wensel. 1966. Aspect transformation in site productivity research. Journal of Forestry 64: 691-692.

Berec, L., E. Angulo, and F. Courchamp. 2007. Multiple Allee effects and population management. Trends in Ecology & Evolution 22(4): 185–191.

Berg, S.W. 2014. “Population structure, size, and the thermal ecology of Iowa wood turtles (Glyptemys insculpta): A comparison between suburban and rural populations.” Master’s thesis, University of Northern Iowa, Cedar Falls, IA; accessed at Electronic Theses and Dissertations, Paper 65.

Beyer, H.L., D.T. Haydon, J.M. Morales, J.L. Friar, M. Hebblewhite, M. Mitchell, and J. Matthiopoulos. 2010. The interpretation of habitat preference metrics under use-availability designs. Philosophical Transactions of the Royal Society B 365: 2245-2254.

Blouin-Demers, G. and P. Nadeau. 2005. The cost–benefit model of thermoregulation does not predict lizard thermoregulatory behavior. Ecology 86(3): 560–566.

Blouin-Demers, G. and P.J. Weatherhead. 2008. Habitat use is linked to components of fitness through the temperature-dependence of performance in ratsnakes (Elaphe obsoleta). Israel Journal of Ecology and Evolution (54): 361-372.

Boerner, R.E.J. 2006. Unraveling the Gordian Knot: Interactions among vegetation, topography, and soil properties in the central and southern Appalachians. Journal of the Torrey Botanical Society 133(2): 321-361.

Bogert, C.M. 1959. How reptiles regulate their body temperature. Scientific American 200: 105-120.

Bollmann, K., P. Weibel, and R.F. Graf. 2005. An analysis of central Alpine capercaillie spring habitat at the forest stand scale. Forest Ecology and Management 215: 307-318.

Bowne, D.R. 2008. Terrestrial activity of Chrysemys picta in northern Virginia. Copeia 2008 (2): 306–310.

339 Box, J.B. and J. Mossa. 1999. Sediment, land use, and freshwater mussels: Prospects and problems. Journal of the North American Benthological Society 18: 99-117.

Boyce, M.S., P.RT. Vernier, S.E. Nielsen, and F.K. Schmiegelow. 2002. Evaluating resource selection functions. Ecological Modeling 157: 281-300.

Boyer, D.R. 1965. Ecology of the basking habit in turtles. Ecology 46(1/2): 99-118.

Branch, W.R. 1984. Preliminary observations on the ecology of the angulate tortoise (Chersina angulate) in Eastern Cape Province, South Africa. Amphibia-Reptilia 5: 43-55.

Braun, E.L. 1950. Deciduous Forests of Eastern North America. The Blakiston Company: Philadelphia, PA. 596 pp.

Brenner, F.J. 1970. The influence of light and temperature on fat utilization in female Clemmys insculpta. The Ohio Journal of Science 70(4): 233-237.

Brooks, R.J., G.P. Brown, and D.A. Galbraith. 1991. Effects of a sudden increase in natural mortality of adults in a population of the common snapping turtle (Chelydra serpentina). Canadian Journal of Zoology 69: 1314-1320.

Brooks, R.J., Shilton, C.M., Brown, G.P., and N.W.S. Quinn. 1992. Body size, age distribution, and reproduction in a northern population of Wood Turtles (Clemmys insculpta). Canadian Journal of Zoology - Revue Canadienne de Zoologie 70(3): 462-469.

Brooks, R.T. and T.D. Kyker-Snowman. 2009. Forest-floor temperatures and soil moisture across riparian zones on first- to third-order headwater streams in southern New England, USA. Forest Ecology and Management 258: 2117- 2126.

Brown, G.P. and R.T. Brooks 1991. Thermal and behavioral responses to feeding in free-ranging turtles, Chelydra serpentina. Journal of Herpetology 25(3): 273- 278.

Brown, D.J., M.D. Nelson, D.J. Rugg, R.R. Buech, and D.M. Donner. 2016. Spatial and temporal habitat-use patterns of Wood Turtles at the western edge of their distribution. Journal of Herpetology 50(3): 347-356.

340 Browne, C.L. and S.J. Hecnar. 2007. Species loss and shifting population structure of freshwater turtles despite habitat protection. Biological Conservation 138: 421–429.

Bryan, R.R. 2007. Focus Species Forestry: A Guide to Integrating Timber and Biodiversity Management in Maine. Maine Audubon with Maine Department of Conservation, Professional Logging Contractors of Maine, Master Logger Program, and Small Woodland Owners Association of Maine: Falmouth, ME. 92 pp.

Buhlmann, K.A. and C.P. Osborn. 2011. Use of an artificial nesting mound by Wood Turtles (Glyptemys insculpta): A tool for turtle conservation. Northeastern Naturalist 18(3): 315-334.

Buhlmann, K.A., T.S.B. Akre, J.B. Iverson,D. Karapatakis, R.A. Mittermeier, A. Georges, A.G.J. Rhodin, P.P. van Dijk, and J.W. Gibbons. 2009. A global analysis of tortoise and freshwater turtle distributions with identification of priority conservation areas. Chelonian Conservation and Biology 8(2): 116– 149.

Burke, V.J. and Gibbons, J.W. 1995. Terrestrial buffer zones and wetland conservation: A case study of freshwater turtles in a Carolina bay. Conservation Biology 9: 1365-1369.

Burnham, K.P. and D.R. Anderson. 2002. Model selection and multimodel inference: A practical information–theoretic approach. 2nd ed. Springer- Verlag. New York. 488 pp.

Burns, R.M., and B.H. Honkala, tech. coords. 1990. Silvics of North America: 1. Conifers; 2. Hardwoods. Agriculture Handbook 654.U.S. Department of Agriculture, Forest Service, Washington, DC. Vol.2, 877 p.

Cadenasso, M.L., M.M. Traynor, and S.T. Pickett. 1997. Functional location of forest edges: Gradients of multiple physical factors. Canadian Journal of Forest Research 27: 774-782.

Caldwell, R.S. 1996. “Macroinvertebrates and their relationship to coarse woody debris: With special reference to land snails”, pp. 49-54 in McMinn, J.W. and D.A. Crossley, Jr., (eds.), Biodiversity and Coarse Woody Debris in Southern Forests, USDA FS Gen. Tech. Report SE-94. USDA FS Southeast Research Station, Asheville, NC. 146 pp.

Callaway, R.M. 1997. Positive interactions in plant communities and the individualistic-continuum concept. Oecologia 112(2): 143-149.

341 Calsbeek, R. and R.M. Cox. 2010. Experimentally assessing the relative importance of predation and competition as agents of selection. Nature 465(3 June 2010): 613-616.

Cantlon, J.E. 1953. Vegetation and microclimates on north and south slopes of Cushetunk Mountain, New Jersey. Ecological Monographs 23: 241–270.

Carmean, W.H. 1965. Black oak site quality in relation to soil and topography in southeastern Ohio. Soil Science Society of America Journal 29(3): 308-312.

Carrascal, L.M., P. Lopez, J. Martin, and A. Salvador. 1992. Basking and antipredator behaviour in a high altitude lizard: Implications of heat- exchange rate. Ethology 92: 143-154.

Carriere, M.A., N. Rollinson, A.N. Suley, and R.J. Brooks. 2008. Thermoregulation when the growing season is short: Sex-biased basking patterns in a northern population of Painted Turtles (Chrysemys picta). Journal of Herpetology 42: 206-209.

Carroll, D.M. 1999. Swampwalker’s Journal: A Wetlands Year. Houghton Mifflin Co., New York, NY. 292 pp.

Carter, S.L., C.A. Haas, and J.C. Mitchell. 1999. Home range and habitat selection of Bog Turtles in Southwestern Virginia. Journal of Wildlife Management 63: 853–860.

Castellano, C.M. 2008. “Ecology and conservation of the wood turtle, Glyptemys insculpta, at Delaware Water Gap National Recreation Area.” PhD dissertation, Fordham University, Bronx, NY. 144 pp.

Castellano, C.M., J.L. Behler, and G.R. Ultsch. 2008. Terrestrial movements of hatchling Wood Turtles (Glyptemys insculpta) in agricultural fields in New Jersey. Chelonian Conservation and Biology 7: 113–118.

Center for Biological Diversity [“CBD”]. 2012. “Petition To List 53 Amphibians And Reptiles In The United States As Threatened Or Endangered Under The Endangered Species Act”. CBD, Circle Pines, MN. Accessible at https://home.cci-dev.org/esadocs-search/

Channell, R. and M.V. Lomolino. 2000. Dynamic biogeography and conservation of endangered species. Nature 403: 84-86.

342 Chapman, J.I. and R.W. McEwan. 2012. Tree regeneration ecology of an old- growth central Appalachian forest: Diversity, temporal dynamics, and disturbance response. Journal of the Torrey Botanical Society 139(2): 194- 205.

Chen, J., S.C. Saunders, T.R. Crow, R.J. Naiman, K.D. Brosofske, G.D. Mroz, B.L. Brookshire, and J.F. Franklin. 1999. Microclimate in forest ecosystem and landscape ecology: Variations in local climate can be used to monitor and compare the effects of different management regimes. BioScience 49(4): 288-297.

Chen, T.-H. and K.Y. Lue. 2008. Thermal preference of the Yellow-Margined Box Turtle (Cuora flavomarginata) (Testudines: Geoemydidae) inhabiting a mesic lowland forest, northern Taiwan. Amphibia-Reptilia 29: 513–522.

Christian, K.A. and C.R. Tracy. 1981. The effect of the thermal environment on the ability of hatchling land iguanas to avoid predation during dispersal. Oecologia 49: 218-223.

Chytry ́, M., L. Tichy ́, J. Holt, and Z. Botta-Dukát. 2002. Determination of diagnostic species with statistical fidelity measures. Journal of Vegetation Science 13: 79–90.

Cianfrani, C., G. Le May, A.H. Hirzel, and A. Loy. 2010. Do habitat suitability models reliably predict the recovery areas of threatened species? Journal of Applied Ecology 47: 421-430.

Claussen, D.L., R. Lim, M. Kurz, and K. Wren. 2002. Effects of slope, substrate, and temperature on the locomotion of the Ornate Box Turtle, Terrapene ornata. Copeia 2002(2): 411–418.

Clebsch, E.E. and R.T. Busing 1989. Secondary succession, gap dynamics, and community structure in a Southern Appalachian cove forest. Ecology 70(3): 728-735.

Clinton, B.D. 2003. Light, temperature, and soil moisture responses to elevation, evergreen understory, and small canopy gaps in the southern Appalachians. Forest Ecology and Management 186: 243–255.

Cloudsley-Thompson, J.L. 1971. The Temperature and Water Relations of Reptiles. Merrow Publishing Co. Ltd., London. 159 pp.

Collins, B.S., K.P. Dunne, and S.T.A. Pickett. 1985. “Response of forest herbs to canopy gaps”, pp. 217-234 in S.T.A. Pickett and P.S. White (eds.), The

343 Ecology of Natural Disturbance and Patch Dynamics. Academic Press, Orlando, FL. 472 pp.

Compton, B.W., J.M. Rhymer, and M. McCollough. 2002. Habitat selection by Wood Turtles (Clemmys insculpta): An application of paired logistic regression. Ecology 83: 833-843.

Congdon, J.D. 1989. Proximate and evolutionary constraints on energy relationships of reptiles. Physiological Zoology 62: 356–373.

Congdon, J.D. and D.W. Tinkle. 1982. Reproductive energetics of the painted turtle (Chrysemys picta). Herpetologica 38:228-237.

Congdon, J.D., A.E. Dunham, and R.C. van Loben Sels. 1993. Delayed sexual maturity and demographics of Blanding’s Turtles (Emydoidea blandingii): Implications for conservation and management of long-lived organisms. Conservation Biology 7(4): 826-833.

Congdon, J.D., A.E. Dunham, and R.C. van Loben Sels. 1994. Demographics of common Snapping Turtles (Chelydra serpentina): Implications for conservation and management of long-lived organisms. American Zoologist 34: 397-408.

Connell, J.H. 1983. On the prevalence and relative importance of interspecific competition: Evidence from field experiments. American Naturalist 122: 661-696.

Connell, J.H. and R.O. Slatyer. 1977. Mechanisms of succession in natural communities and their role in community stability and organization. American Naturalist 111: 1119-1144.

Converse, S.J. and J.A. Savidge. 2003. Ambient temperature, activity, and microhabitat use by Ornate Box Turtles (Terrapene ornata ornata). Journal of Herpetology 37(4): 665–670.

Converse, S.J., J.B. Iverson, and J.A. Savidge. 2005. Demographics of an Ornate Box Turtle population experiencing minimal human-induced disturbances. Ecological Applications 15(6): 2171-2179.

Cooch, E.G and G.C. White (eds.). 2013. MARK: A Gentle Introduction, 13th Ed. Accessed at www.phidot.org

Cooney, R. 2004. The Precautionary Principle in Biodiversity Conservation and

344 Natural Resource Management: An issues paper for policy-makers, researchers and practitioners. IUCN, Gland, Switzerland and Cambridge, UK. xi + 51pp.

COSEWIC. 2007. COSEWIC assessment and update status report on the Wood Turtle Glyptemys insculpta in Canada. Committee on the Status of Endangered Wildlife in Canada, Ottawa. 42 pp. Available at www.sararegistry.gc.ca/status/status_e.cjm

Costa, Z.J. 2014. Responses to predators differ between native and invasive freshwater turtles: Environmental context and its implications for competition. Ethology 120: 1-8.

Coulling, P.P. and T.J. Rawinsky 1999. Classification of vegetation and ecological land units of the Piney River and Mt. Pleasant area, Pedlar Ranger District, George Washington National Forest, Virginia. Natural Heritage Tech. Rep. 99-03, Virginia Dept. of Conservation and Recreation, Div. of Natural Heritage, Richmond. 135 pp, plus appendices.

Crawford, K.M., J.R. Spotila, and E.A. Standora. 1983. Operative environmental temperatures and basking behavior of the turtle Pseudemys scripta. Ecology 64(5): 989-999.

Crawford, J.A. and R.D. Semlitsch. 2007. Estimation of core terrestrial habitat for stream-breeding salamanders and delineation of riparian buffers for protection of biodiversity. Conservation Biology 21(1): 159–167.

Cunnington, G.M., J. Schaefer, J.E. Cebek, and D. Murray. 2008. Correlations of biotic and abiotic variables with ground surface temperature: An ectothermic perspective. Ecoscience 15(4): 472-477.

Currylow, A.F., B.J. MacGowan, and R.N. Williams. 2012. Short-term forest management effects on a long-lived ectotherm. PLoS One 7(7): 1-12.

Curtis, J. and P. Vila. 2015. The ecology of the Wood Turtle (Glyptemys insculpta) in the eastern panhandle of West Virginia. Northeastern Naturalist 22(2): 387-402.

Daigle, C. 1997. Size and characteristics of a Wood turtle, Clemmys insculpta, population in Southern Quebec. Canadian Field-Naturalist 111: 440-444.

Daigle, C. and J. Jutras. 2005. Evidence of decline in a southern Quebec Wood Turtle (Glyptemys insculpta) population. Journal of Herpetology 39(1): 130- 132.

345

Davis, M.B. (ed.). 1996. Eastern Old-Growth Forests: Prospects for Rediscovery and Recovery. Island Press, Washington, D. C. 383 pp.

Davy, C.M. 2013. “Conservation Genetics of Freshwater Turtles.” PhD dissertation, University of Toronto, Ontario, Canada. 163 pp.

De Cáceres, M. and P. Legendre 2009. Associations between species and groups of sites: Indices and statistical inference. Ecology 90: 3566–3574.

De Cáceres, M., P. Legendre, S.K. Wiser, and L. Brotons. 2012. Using species combinations in indicator value analyses. Methods in Ecology and Evolution 3: 973–982.

DeGraaf, R.M., J.B. Hestbeck, and M. Yamasaki. 1998. Associations between breeding bird abundance and stand structure in the White Mountains, New Hampshire and Maine, USA. Forest Ecology and Management 103: 217- 233.

Del Vecchio, S., R.L. Burke, L. Rugiero, M. Capula, and L. Luiselli. 2011. The turtle is in the details: Microhabitat choice by Testudo hermanni is based on microscale plant distribution. Animal Biology 61: 249-261.

Dellinger, R.L., P.B. Wood, P.D. Keyser, and G. Seidel. 2007. Habitat partitioning of four sympatric thrush species at three spatial scales on a managed forest in West Virginia. The Auk 124(4): 1425-1436.

DeMars, B.G. and J.R. Runkle. 1992. Groundlayer vegetation ordination and site- factor analysis of the Wright State University Woods (Greene County, Ohio). Ohio Journal of Science 92(4): 98-106. deMaynadier, P.G. and M.L. Hunter Jr. 1995. The relationship between forest management and amphibian ecology: A review of the North American literature. Environmental Review 3: 230–261.

Dillard, L.O., K.R. Russell, and W.M. Ford. 2008. Site-level habitat models for the endemic, threatened Cheat Mountain salamander (Plethodon nettingi): The importance of geophysical and biotic attributes for predicting occurrence. Biodiversity Conservation 17: 1475–1492.

Doak, D.F. 1992. Lifetime impacts of herbivory for a perennial plant. Ecology 73(6): 2086-2099. do Amaral, J.P.S., G.A. Marvin, and V.H. Hutchison. 2002. Thermoregulation in

346 the box turtles Terrapene carolina and Terrapene ornata. Canadian Journal of Zoology 80: 934–943.

Dobrowski, S.Z. 2010. A climatic basis for microrefugia: The influence of terrain on climate. Global Change Biology 16: 1-14.

Dodd Jr., C.K. 2001. North American Box Turtles: A Natural History. University of Oklahoma Press, Norman, OK. 231 pp.

Donaldson, B.M. and A.C. Echternacht. 2005. Aquatic habitat use relative to home range and seasonal movement of eastern box turtles (Terrapene carolina carolina: Emydidae) in eastern Tennessee. Journal of Herpetology 39(2): 278-284.

Doroff, A.M. and L.B. Keith. 1990. Demography and ecology of an ornate box turtle (Terrapene ornata) population in south-central Wisconsin. Copeia 1990: 387-399.

Downes, S. 2001. Trading heat and food for safety: Costs of predator avoidance in a lizard. Ecology 82: 2870-2881.

Dragon, J. 2014. “Habitat selection, movement, and survival of hatchling Wood Turtles (Glyptemys insculpta) in an atypical habitat”. Master’s thesis, George Mason University, Fairfax, VA. 81 pp.

Dubois, Y., G. Blouin-Demers, and D. Thomas. 2008. Temperature selection in Wood Turtles (Glyptemys insculpta) and its implications for energetics. Ecoscience 15(3): 398-406.

Dubois, Y., G. Blouin-Demers, B. Shipley, and D. Thomas. 2009. Thermoregulation and habitat selection in Wood Turtles Glyptemys insculpta: Chasing the sun slowly. Journal of Animal Ecology 78: 1023– 1032.

Dunham, A.E., B.W. Grant, and K.L. Overall. 1989. Interfaces between biophysical and physiological ecology and population ecology of terrestrial vertebrate ectotherms. Physiological Zoology 62: 335–355.

Dyer, J.M. 2006. Revisiting the Deciduous Forests of Eastern North America. BioScience 56(4): 341-352.

Dyer, J.M. 2009. Assessing topographic patterns in moisture use and stress using a water balance approach. Landscape Ecology 24: 391-403.

347 Dyer, J.M. 2010. Land-use legacies in a central Appalachian forest: Differential response of trees and herbs to historic agricultural practices. Applied Vegetation Science 13: 195-206.

Edge, C.B., B.D. Steinberg, R.J. Brooks, and J.D. Litzgus. 2010. Habitat selection by Blanding’s Turtles (Emydoidea blandingii) in a relatively pristine landscape. Ecoscience 17: 90–99.

Edwards, Jr., T.C., G.C. Moisen, T.S. Frescino, and J.J. Lawler. 2003. “Modeling multiple ecological scales to link landscape theory to wildlife conservation”, pp. 153-172 in J. Bissonette and I. Storch (eds.), Landscape Ecology and Resource Management: Linking Theory With Practice. Island Press, Washington, D.C. 463 pp.

Egler, F.E. 1954. Vegetation science concepts. I. Initial floristic composition, a factor in old-field vegetation development. Vegetatio 4: 412-417.

Elliot, K.J., J.M. Vose, B.D. Clinton, and J.D. Knoepp. 2004. “Effects of understory burning in a mesic mixed-oak forest of the southern Appalachians”, pp. 272- 283 in R.T. Engstrom, K.E.M. Galley, and W.J. de Groot (eds.), Proceedings of the 22nd Tall Timbers Fire Ecology Conference: Fire in Temperate, Boreal, and Montane Ecosystems. Tall Timbers Research Station, Tallahassee, FL.

Enneson, J.J. and J.D. Litzgus. 2008. Using long-term data and a stage-classified matrix to assess conservation strategies for an endangered turtle (Clemmys guttata). Biological Conservation 141: 1560–1568.

Ernst, C.H. 1968. Evaporative water-loss relationships of turtles. Journal of Herpetology 2: 159-161.

Ernst, C. 1972. Temperature-activity relationship in the Painted Turtle, Chrysemys picta. Copeia 1972(2): 217-222.

Ernst, C.H. 1986. Environmental temperatures and activities of the wood turtle, Clemmys insculpta. Journal of Herpetology 20: 222–229.

Ernst, C.H. 2001a. Some ecological parameters of the Wood Turtle, Clemmys insculpta, in southeastern Pennsylvania. Chelonian Conservation and Biology 4(1): 94-99.

Ernst, C.H. 2001b. An overview of the North American turtle genus Clemmys Ritgen, 1828. Chelonian Conservation and Biology 4(1): 211-216.

348 Ernst, C.H. and J.E. Lovich. 2009. “Wood Turtle”, pp. 250-262 in Turtles of the United States and Canada. Johns Hopkins University Press, Baltimore, MD. 840 pp.

Everett, R.A. and G.M. Ruiz 1993. Coarse woody debris as a refuge from predation in aquatic communities: An experimental test. Oecologia 93: 475-486.efuge

Fagan, W.F. and E.E. Holmes. 2006. Quantifying the extinction vortex. Ecology Letters 9: 51–60.

Fahrig, L. 2003. Effects of habitat fragmentation on biodiversity. Annual Review of Ecology, Evolution, and Systematics 34: 487–515.

Fahrig, L. 2007. Non-optimal animal movement in human-altered landscapes. Functional Ecology 21: 1003–1015.

Farallo, V.R. and D.B. Miles. 2016. The importance of microhabitat: A comparison of two microendemic species of Plethodon to the widespread P. cinereus. Copeia 104(1): 67-77.

Farrell, R.F. and T.E. Graham. 1991. Ecological notes on the turtle Clemmys insculpta in northwestern New Jersey. Journal of Herpetology 25(1): 1-9.

Fernald, M.L. 1950. Gray's Manual of Botany: A Handbook of the Flowering Plants and Ferns of the Central and Northeastern United States and Adjacent Canada (8th edition). American Book Company, New York, NY. 1632 pp.

Ficetola, G.F., E. Padoa-Schioppa, A. Monti, R. Massa, F. De Bernardi, and L. Bottoni. 2004. The importance of aquatic and terrestrial habitat for the European pond turtle (Emys orbicularis): Implications for conservation planning and management. Canadian Journal of Zoology 82: 1704-1712.

Finke, D.L. and R.F. Denno. 2006. Spatial refuge from intraguild predation: Implications for prey suppression and trophic cascades. Oecologia 149: 265–275.

Finkler, M., D.L. Knickerbocker, and D.L. Claussen. 2000. Influence of population and hydric conditions during incubation on overland movement of neonatal snapping turtles. Journal of Herpetology 43: 452-455.

Finkler, M.S. 2001. Rates of water loss and estimates of survival time under varying humidity in juvenile Snapping Turtles (Chelydra serpentina). Copeia 2001(2): 521-525.

349 Flanagan, M., V. Roy-McDougal, G. Forbes, and G. Forbes. 2013. Survey methodology for the detection of Wood Turtles (Glyptemys insculpta). Canadian Field-Naturalist 127(3): 216-223.

Fleming, G.P. 2012. “The Nature of the Virginia Flora”, pp. 24-75 in Weakley, A.S., J.C. Ludwig, and J.F. Townsend, Flora of Virginia. BRIT Press, Fort Worth, TX. 1554 pp.

Fleming, G.P. and P.P. Coulling. 2001. Ecological Communities of the George Washington and Jefferson National Forests, Virginia: Preliminary classification and description of vegetation types. Natural Heritage Tech. Rep. 01-14. Virginia Dept. of Conservation and Recreation, Division of Natural Heritage, Richmond. Unpublished report submitted to the USDA Forest Service. 372 pp.

Fletcher, R.J., Jr. 2005. Multiple edge effects and their implications in fragmented landscapes. Journal of Animal Ecology 74: 342–352.

Foley, R.E. and J.R. Spotila. 1978. Effect of wind speed, air temperature, body size and vapor density difference on evaporative water loss from the turtle Chrysemys scripta. Copeia 1978(4): 627-634.

Ford, W.M., Menzel, M.A., McGill, W.W., Laerm, J., and T.S. McCay. 1999. Effects of a community restoration fire on small mammals and herpetofauna in the southern Appalachians. Forest Ecology and Management 114: 233–243.

Fortin, D., M.S. Boyce, and E.H. Merrill. 2004. Multi-tasking by mammalian herbivores: Overlapping processes during foraging. Ecology 85(8): 2312- 2322.

Foscarini, D.A. 1994. “Demography of the Wood Turtle (Clemmys insculpta) and habitat selection in the Maitland River Valley”. Master’s thesis, University of Guelph, Ontario, Canada. 139 pp.

Foster, D.R., D.A. Orwig, and J.S. Machlan. 1996. Ecological and conservation insights from reconstructive studies of temperate old-growth forests. TREE 11(10): 419-424.

Frankham, R. 2003. Genetics and conservation biology. CR Biologies 326: 22–29.

Franklin, J.F., T.A. Spies, R. Van Pelta, A.B. Carey, D.A. Thornburgh, D.R. Berge, D.B. Lindenmayer, M.E. Harmon, W.S. Keeton, D.C. Shaw, K. Bible, and J. Chen. 2002. Disturbances and structural development of natural forest

350 ecosystems with silvicultural implications, using Douglas-fir forests as an example. Forest Ecology and Management 155: 399–423.

Fredericksen, T.S. 2014. Thermal regulation and habitat use of the Eastern Box Turtle in southwestern Virginia. Northeastern Naturalist 21(4): 554-564.

Frelich, L.E., J-L Machado, and P.B. Reich. 2003. Fine-scale environmental variation and structure of understory plant communities in two old-growth pine forests. Journal of Ecology 91: 283-293.

Fretwell, S.D. and H.L. Lucas. 1970. On territorial behavior and other factors influencing habitat distribution in birds. Acta Biotheoret. 19: 16-36.

Fridgen, C., L. Finnegan, C. Reaume, J. Cebek, J. Trottier, and P.J. Wilson. 2013. Conservation genetics of Wood Turtle (Glyptemys insculpta) populations in Ontario, Canada. Herpetological Conservation and Biology 8(2): 351-358.

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

Frissell, C.A. 1997. “Ecological Principles”, pp. 96-115 in J. E. Williams, C. A. Wood, and M. P. Dombeck (eds.), Watershed Restoration: Principles and Practices. American Fisheries Society, Bethesda, MD.

Fulton, W., R. Pendall, M. Nguyen, and A. Harrison. 2001. Who sprawls most? How growth patterns differ across the U.S. The Brookings Institute Center on Urban & Metropolitan Policy, Washington D.C. 24 pp.

Futuyma, D.J. 2010. Evolutionary constraint and ecological consequences. Evolution 64(7): 1865-1884.

Galbraith, D.A. 1993. Mini review: Multiple paternity and sperm storage in turtles. Herpetological Journal 3(4): 117-123.

Galbraith, D.A., R.J. Brooks, and G.P. Brown. 1997. "Can management intervention achieve sustainable exploitation of turtles?", pp. 186-194 in J. Van Abbema (ed.), Proceedings: Conservation, Restoration, and Management of Tortoises and Turtles - An International Conference, New York Turtle and Tortoise Society, New York. 494 pp.

Gans, C., T. Krakauer, and C.V. Pagnelli. 1968. Water loss in snakes: Interspecific

351 and intraspecific variability. Comp. Biochem. Physiol. 27: 747-761.

Garber, S.D. and J. Burger. 1995. A 20-yr study documenting the relationship between turtle decline and human recreation. Ecological Applications 5(4): 1151-1162.

Gardner, T.A., J. Barlow, and C.A. Peres. 2007. Paradox, presumption and pitfalls in conservation biology: The importance of habitat change for reptiles and amphibians. Biological Conservation 138: 166-179.

Garshelis, D.L. 2002. “Delusions in habitat evaluation: Measuring use, selection, and importance”, pp. 111-164 in L. Boitani and T.K. Fuller (eds.), Research Techniques in Animal Ecology: Controversies and Consequences. Columbia University Press, New York.

Gianopulos, K.D. and J.W. Rowe 1999. Effects of short-term water temperature variation on food consumption in painted turtles (Chrysemys picta marginata). Chelonian Conservation and Biology 3: 504-507.

Gibbs, J.P. and G.D. Amato. 2000. “Genetics and Demography in Turtle Conservation”, pp. 207-217 in M.W. Klemens (ed.), Turtle Conservation. Smithsonian Institution Press, Washington D.C. 334 pp.

Gilliam, F.S. 2007. The ecological significance of the herbaceous layer in temperate forest ecoystems. BioScience 57(10): 845-858.

Girvetz, E.H. and S.E. Greco. 2009. Multi-scale predictive habitat suitability modeling based on hierarchically delineated patches: An example for yellow-billed cuckoos nesting in riparian forests, California, USA. Landscape Ecology 24: 1315-1329.

Glasgow, L.S. and G.R. Matlack. 2007. The effects of prescribed burning and canopy openness on establishment of two non-native plant species in a deciduous forest, southeast Ohio, USA. Forest Ecology and Management 238: 319–329.

Gleason, H.A. 1926. The individualistic concept of the plant association. Bulletin of the Torrey Botanical Club 53: 7-26.

Goins, S.M., J.I. Chapman, and R.W. McEwan. 2013. Composition shifts, disturbance, and canopy-accession strategy in an old-growth forest of southwestern Ohio, USA. Natural Areas Journal 33: 384-394.

352 Goldblum, D. 1997. The effects of treefall gaps on understory vegetation in New York State. Journal of Vegetation Science 8: 125–132.

Graham, T.E and J.E. Forsberg. 1991. Aquatic oxygen intake by naturally wintering Wood Turtles (Clemmys insculpta). Copeia 1991(3): 836-838.

Graham, T.E. and V.H. Hutchison. 1979. Turtle diel activity: Response to different regimes of temperature and photoperiod. Comparative Biochemistry and Physiology Part A: Physiology 63(2): 299-305.

Grant, B.W. and A.E. Dunham. 1988. Biophysically imposed time constraints on the activity of a desert lizard, Sceloporus merriami. Ecology 69: 167-176.

Grayson, K.L. and M.E. Dorcas. 2004. Seasonal temperature variation in the Painted Turtle (Chrysemys picta). Herpetologica 60(3): 325-336.

Greaves, W.F. 2008. “A cold and harsh environment: Demography and spatial ecology of a northern population of wood turtles (Glyptemys insculpta)”. Master’s thesis, Laurentian University, Sudbury, Ontario.

Greaves, W.F. and J.D. Litzgus. 2007. Overwintering ecology of Wood Turtles (Glyptemys insculpta) at the species’ northern range limit. Journal of Herpetology 41(1): 32–40.

Greaves, W.F. and J.D. Litzgus. 2008. Chemical, thermal, and physical properties of sites selected for overwintering by northern Wood Turtles (Glyptemys insculpta). Canadian Journal of Zoology 86(7): 659–667.

Greaves, W.F. and J.D. Litzgus. 2009. Variation in life-history characteristics among populations of North American wood turtles: A view from the north. Journal of Zoology 2009: 1-12.

Greenberg, C.H. and T.G. Forrest. 2003. Seasonal abundance of ground-occurring macroarthropods in forest and canopy gaps in the Southern Appalachians. Southeastern Naturalist 2(4): 591-608.

Grigg, G. and D. Kirshner. 2015. Biology and Evolution of Crocodylians. Cornell University Press, Ithaca, NY. 649 pp.

Grover, M.C. 2000. Determinants of salamander distributions along moisture gradients. Copeia 2000(1): 156–168.

Guisan, A. and N.E. Zimmermann. 2000. Predictive habitat distribution models in

353 ecology. Ecological Modeling 135(2-3): 147-186.

Gunderson, A.R. and M. Leal 2016. A conceptual framework for understanding thermal constraints on ectotherm activity with implications for predicting responses to global change. Ecology Letters 19: 111-120.

Hagan, J.M., W.M. Vander Haegen, P.S. McKinley. 1996. The early development of forest fragmentation effects on birds. Conservation Biology 10(1): 188- 202.

Hall, L.S., P.R. Krausman, and M.L. Morrison. 1997. The habitat concept with a plea for standard terminology. Wildlife Society Bulletin 25: 173–182.

Halpern, C.B. and T.A. Spies. 1995. Plant species diversity in natural and managed forests of the Pacific Northwest. Ecological Applications 5: 913–934.

Halstead, B.J., H.R. Mushinsky, and E.D. McCoy. 2009. Masticophis flagellum selects Florida scrub habitat at multiple spatial scales. Herpetologica 65(3): 268–279.

Hamernick, M.G. 2000. Home ranges and habitat selection of Blanding’s turtles (Emydoidea blandingii) at the Weaver Dunes, Minnesota. Final report submitted to the Nongame Wildlife Program, Minnesota Department of Natural Resources. 18 pp. Accessed at http://www.dnr.state.mn.us/eco/nongame/projects/research_reports/reptiles. html

Hammond, K.A., J.R. Spotila, and E.A. Standora. 1988. Basking behavior of the turtle Pseudemys scripta: Effects of digestive state, acclimation temperature, sex, and season. Physiological Zoology 61(1): 69-77.

Hampe, A. and R.J. Petit. 2005 Conserving biodiversity under climate change: The rear edge matters. Ecology Letters 8: 461-467.

Hansen, A.J. and J.J. Rotella. 2002. Biophysical factors, land use, and species viability in and around nature reserves. Conservation Biology 16(4): 1112- 1122.

Hansen, A.J. and R. DeFries. 2007. Ecological mechanisms linking protected areas to surrounding lands. Ecological Applications 17(4): 974-988.

Hanula, J.L. 1996. “Relationships of wood-feeding insects and coarse woody debris”, pp. 55-81 in McMinn, J.W. and D.A. Crossley, Jr. (eds.), Biodiversity

354 and Coarse Woody Debris in Southern Forests. General Technical Report SE-94. USDA FS Southeast Research Station, Asheville, NC. 146 pp.

Harding, J. 2002. “Clemmys insculpta North American wood turtle” at University of Michigan Museum of Zoology Animal Diversity Web. Accessible via world wide web.

Harding, J.H. and T.J. Bloomer. 1979. The Wood Turtle, Clemmys insculpta . . . A Natural History. HERP 15(1): 9-26.

Hardt, R.A. and W.T. Swank. 1997. A comparison of structural and compositional characteristics of Southern Appalachian young second-growth, maturing second-growth, and old-growth stands. Natural Areas Journal 17(1): 42-52.

Harlow, H.J., D. Purwandana, T.S. Jessop, and J.A. Phillips. 2010. Body temperature and thermoregulation of Komodo dragons in the field. Journal of Thermal Biology 35: 338-347.

Harmon, M.E., J.F. Franklin, F.J. Swanson, P. Sollins, S.V. Gregory, J.D. Lattin, N.H. Anderson, S.P. Cline, N.G. Aumen, J.R. Sedell, G.W. Lienkaemper, K. Cromack Jr., and K.W. Cummins. 1986. Ecology of coarse woody debris in temperate ecosystems. Adv. Ecol. Research 15: 133-302.

Harper, K.A., S.E. Macdonald, P.J. Burton, J. Chen, K.D. Brosofske, S.C. Saunders, E.S. Euskirchen, D. Roberts, M.S. Jaiteh, and Per-Anders Esseen. 2005. Edge influence on forest structure and composition in fragmented landscapes. Conservation Biology 19: 768-782.

Harris, L.D., T.S. Hoctor, and S.E. Gergel. 1996. “Landscape processes and their significance to biodiversity conservation”, pp. 319-347 in Rhodes, O.E. Jr., R.K. Chesser, M.H. Smith (eds.), Population Dynamics in Ecological Space and Time. University of Chicago Press, Chicago.

Harts, A.M.F., L.E. Schwanz, and H. Kokko. 2014. Demography can favor female advantageous alleles. Proceedings of the Royal Society B 281: 2014005.

Hayne, D.W. 1949. Calculation of size of home range. Journal of Mammalogy 30(1): 1-18.

Hedman, C.W., D.H. Van Lear, and W.T. Swank. 1996. In-stream large woody debris loading and riparian forest seral stage associations in the southern Appalachian Mountains. Canadian Journal of Forest Research 26: 1218- 1227.

355

Hendrix, P.F. 1996. “Earthworms, biodiversity, and coarse woody debris in forest ecosystems of the southeastern U.S.A.”, pp. 43-48 in McMinn, J.W. and D.A. Crossley, Jr. (eds.), Biodiversity and Coarse Woody Debris in Southern Forests, USDA FS Gen. Tech. Report SE-94. USDA FS Southeast Research Station, Asheville, NC. 146 pp.

Henley, W.F., M.A. Patterson, R.J. Neves, and A. Dennis Lemly. 2000. Effects of sedimentation and turbidity on lotic food webs: A concise review for natural resource managers. Reviews in Fisheries Science 8(2): 125-139.

Heppell, S.S. 1998. Application of life-history theory and population model analysis to turtle conservation. Copeia 1998: 367–375.

Heppell, S.S., H. Caswell, and L.B. Crowder. 2000. Life histories and elasticity patterns: Perturbation analysis for species with minimal demographic data. Ecology 81: 654-665.

Hicks, D.J. 1980. Intrastand distribution patterns of southern Appalachian cove forest herbaceous species. American Midland Naturalist 104: 209–223.

Hirzel, A.H. and G. Le May. 2008. Habitat suitability modeling and niche theory. Journal of Applied Ecology 45: 1372-1381.

Hirzel, A.H., G. Le May, V. Helfer, C. Randin, and A. Guisan. 2006. Evaluating the ability of habitat suitability models to predict species presences. Ecological Modeling 199: 142-152.

Holden, M.H. and E. McDonald-Madden. 2017. High prices for rare species can drive large populations extinct: The anthropogenic Allee effect revisited. Journal of Theoretical Biology 429: 170-180.

Holt, S.M. 2000. Development and evaluation of a model for turtle embryonic growth. Masters thesis, University of Guelph, Guelph, Ontario, Canada.

Homyack, J.A. 2010. Evaluating habitat quality of vertebrates using conservation physiology tools. Wildlife Research 37: 332-342.

Honnay, O., H. Jacquemyn, B. Bossuyt, and M. Hermy. 2005. Forest fragmentation effects on patch occupancy and population viability of herbaceous plant species. New Phytologist 166: 723–736.

356 Horn, D. and T. Cathcart. 2005. Wildflowers of Tennessee, the Ohio Valley, and the Southern Appalachians. Lone Pine Publishing, Auburn, WA. 496 pp.

Howey, C.A. 2014. “The Ecological Effects of Prescribed Fire on the Black Racer (Coluber constrictor)”. PhD dissertation, Ohio University, Athens, OH. 172 pp.

Huebner, C.D., J.C. Randolph, and G.R. Parker. 1995. Environmental factors affecting understory diversity in second-growth deciduous forests. The American Midland Naturalist 134(1): 155-165.

Huey, R.B. 1982. “Temperature, physiology, and the ecology of reptiles”, pp. 25- 91 in C. Gans and F. H. Pough (eds.), Biology of the Reptilia. Vol. 12. Physiology C. Academic Press, London.

Huey, R.B. 1991. Physiological consequences of habitat selection. The American Naturalist 137 Supplement (June 1991): S91-S115.

Huey, R. B. and M. Slatkin. 1976. Costs and benefits of lizard thermoregulation. Quarterly Review of Biology 51: 363-384.

Huey, R.B. and R. D. Stevenson. 1979. Integrating thermal physiology and ecology of ectotherms: A discussion of approaches. American Zoologist 19: 357- 366.

Huey, R.B., M.R. Kearney, A. Krockenberger, J.A. M. Holtum, M. Jess, and S.E. Williams. 2012. Predicting organismal vulnerability to climate warming: Roles of behaviour, physiology and adaptation. Philosophical Transactions of the Royal Society B (2012) 367: 1665-1679.

Huffaker, C.B. 1958. Experimental studies on predation: Dispersion factors and predator-prey oscillations. Hilgardia 27: 343-383.

Hunt, C.B. 1974. Natural Regions of the United States and Canada. W.H. Freeman and Co., San Francisco, CA. 725 pp.

Hutchinson, T.F., R.E.J. Boerner, L.R. Iverson, S. Sutherland, and E.K. Sutherland. 1999. Landscape patterns of understory composition and richness across a moisture and nitrogen mineralization gradient in Ohio (U.S.A.) Quercus forests. Plant Ecology 144: 177-189.

Hutchison, V.H. 1979. “Thermoregulation”, pp. 207-228 in Harless, M. and H. Morlock (eds.), Turtles: Perspectives and Research. John Wiley & Sons, New York. 695 pp.

357 Hutchison, V.H., A. Vinegar, and R.J. Kosh. 1966. Critical thermal maxima in turtles. Herpetologica 22: 32-41.

Iffrig, G.F., C.E. Trammel, and T. Cunningham. 2008. “Describing Single-Tree Selection Harvests in Missouri Ozark Forests”, pp. 49-60 in Guldin, J.M., G.F. Iffrig, and S.L. Flader (eds.), Pioneer Forest— A half century of sustainable uneven-aged forest management in the Missouri Ozarks. Gen. Tech. Rep. SRS–108. Asheville, NC: U.S. Department of Agriculture, Forest Service, Southern Research Station. 123 pp.

International Union for Conservation of Nature [IUCN]. 2017. The IUCN Red List of Threatened Species. Version 2017-2. Accessible at www.iucnredlist.org [For Wood Turtles: “Status: Endangered A2cd4c Pop. Trend: decreasing”]

Iverson, J.B. 1982. Biomass in turtle populations: A neglected subject. Oecologia 55: 69-76.

Iverson, L.R., A.N. Prasad, S.N. Matthews, and M. Peters. 2008. Estimating potential habitat for 134 eastern US tree species under six climate scenarios. Forest Ecology and Management 254: 390–406.

Jackson, M.T. and J.E. Newman. 1967. Indices for expressing differences in local climates due to forest cover and topographic differences. Forest Science 13: 60–71.

Jackson, D.A., P.R. Peres-Neto, and J.D. Olden. 2001. What controls who is where in freshwater fish communities – the roles of biotic, abiotic, and spatial factors. Canadian Journal of Fisheries and Aquatic Science 58: 157-170.

Jaeger, C.P. and V.A. Cobb. 2012. Comparative spatial ecologies of female Painted Turtles (Chrysemys picta) and Red-Eared Sliders (Trachemys scripta) at Reelfoot Lake, Tennessee. Chelonian Conservation and Biology: 11(1):59- 67.

James, C.D. 1991. Temporal variation in diets and trophic partitioning by coexisting lizards (Ctenotus: Scincidae) in central Australia. Oecologia 85(4): 553-561.

Jamieson, I.G. and F.W. Allendorf. 2012. How does the 50/500 rule apply to MVPs? Trends in Ecology and Evolution 27(10): 578-584.

Jennings, S.B., N.D. Brown, and D. Sheil. 1999. Assessing forest canopies and

358 understory illumination: Canopy closure, canopy cover, and other measures. Forestry 72: 59–73.

Johnson, D.H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61(1): 65-71.

Jones, H.P., P.C. Jones, E.B. Barbier, R.C. Blackburn, J.M. Rey Benayas, K.D. Holl, M. McCrackin, P. Meli, D. Montoya, and D. Moreno Mateos. 2018. Restoration and repair of Earth's damaged ecosystems. Proceedings of the Royal Society B: Biological Sciences 285(1873): 20172577 DOI: 10.1098/rspb.2017.2577

Jones, K.B., K.H. Riitters, J.D. Wickham, R.D. Tankersley, R.V. O’Neill, D.J. Chaloud, E.R. Smith, and A.C. Neale. 1997. An Ecological Assessment of the United States Mid-Atlantic Region: A Landscape Atlas. USEPA/600/R- 97/130. U.S. Environmental Protection Agency. Washington, D.C. 104 pp. See “Mid-Atlantic Integrated Assessment” at http://www.epa.gov/maia/html/maia-atlas.html

Jones, M. 2009. “Spatial ecology, population structure, and conservation of the wood turtle, Glyptemys insculpta, in central New England”. Ph.D. dissertation, University of Massachusetts, Amherst. Accessible at http://scholarworks.umass.edu/openaccessdissertations/39

Jones, M.T. and L.L. Willey. 2015. Status and Conservation of Wood Turtles in the Northeastern United States. Final Report to the Regional Conservation Needs (RCN) Program. Accessible at http:// rcngrants.org/sites/default/files/datasets/RCN2011-02v2.pdf

Jones, S.C., W.J. Jordan IV, S.J. Meiners, A.N. Miller, and A.S. Methven. 2007. Fungal spore dispersal by the Eastern Box Turtle (Terrapene carolina carolina). American Midland Naturalist 157: 121–126.

Joyal, L.A., M. McCollough, and M.L. Hunter Jr. 2001. Landscape ecology approaches to wetland species conservation: A case study of two turtle species in Southern Maine. Conservation Biology 15: 1755-1762.

Kapfer, J.M., J.R. Coggins, and R. Hay. 2008. Spatial ecology and habitat selection of Bullsnakes at the northern periphery of their geographic range. Copeia 2008(4): 815-826.

Kapfer, J.M., C.W. Pekar, D.M. Reineke, J.R. Coggins, and R. Hay. 2010. Modeling the relationship between habitat preferences and home-range size: A case

359 study on a large mobile colubrid snake from North America. Journal of Zoology 282: 13-20.

Kappes, H. 2006. Relations between forest management and slug assemblages (Gastropoda) of deciduous regrowth forests. Forest Ecology and Management 237: 450–457.

Kaufmann, J.H. 1992a. Habitat use by wood turtles in central Pennsylvania. Journal of Herpetology 26(3): 315-321.

Kaufmann, J.H. 1992b. The social behavior of Wood Turtles, Clemmys insculpta, in central Pennsylvania. Herpetological Monographs (6): 1-25.

Kaufmann, J.H. 1995. Home ranges and movements of wood turtles, Clemmys insculpta, in central Pennsylvania. Copeia 1995: 22–27.

Kazmaier, R.T., E.C. Hellgren, and D.C. Ruthven. 2001. Habitat selection by the Texas Tortoise in a managed thornscrub ecosystem. Journal of Wildlife Management 65(4): 653-660.

Kearney, M. 2006. Habitat, environment and niche: What are we modeling? Oikos 115(1): 186-191.

Kearney, M.R. 2013. Activity restriction and the mechanistic basis for extinctions under climate warming. Ecology Letters 16:1470-1479.

Kearney, M. and W.P. Porter 2004. Mapping the fundamental niche: Physiology, climate, and the distribution of a nocturnal lizard. Ecology 85(11): 3119- 3131.

Kearns, L.J., E.D. Silverman, and K.R. Hall. 2006. Black-Throated Blue Warbler and Veery abundance in relation to understory composition in northern Michigan forests. The Wilson Journal of Ornithology 118(4): 461-470.

Keeton, W.S. 2006. Managing for late-successional/old-growth characteristics in northern hardwood-conifer forests. Forest Ecology and Management 235: 129–142.

Keeton, W.S. and A.R. Troy. 2006. “Balancing ecological and economic objectives while managing for old-growth forest characteristics”, pp. 21-33 in L. Zahoyska (ed.), Ecologisation of economy as a key prerequisite for sustainable development. Proceedings of the international conference, Sept. 22-23, 2005, Ukrainian National Forestry University, L’viv, Ukraine.

360 Keeton, W.S., Kraft, C.E., and D.R. Warren. 2007. Mature and old-growth riparian forests: Structure, dynamics, and effects on Adirondack stream habitats. Ecological Applications 17(3): 852-68.

Kinniston, M.T. and N.G. Hairston. 2007. Eco-evolutionary conservation biology: Contemporary evolution and the dynamics of persistence. Functional Ecology 21: 444-454.

Kotliar, N.B. and J.A. Wiens. 1990. Multiple scales of patchiness and patch structure: A hierarchical framework for the study of heterogeneity. Oikos 59(2): 253-260.

Küchler, A.W. 1964. Potential Natural Vegetation of the Conterminous United States. New York: American Geographical Society.

Kuo, C-H. and F.J. Janzen. 2004. Genetic effects of a persistent bottleneck on a natural population of Ornate Box Turtles (Terrapene ornata). Conservation Genetics 5: 425–437.

Lagory, K.E., L.J. Walston, C. Goulet, R.A. van Lonkhuyzen, S. Najjar, and C.A. Andrews. 2009. An examination of scale-dependent resource use by Eastern Hognose Snakes in southcentral New Hampshire. Journal of Wildlife Management 73: 1387–1393.

Lande, R. 1993. Risks of population extinction from demographic and environmental stochasticity and random catastrophes. The American Naturalist 142: 911-927.

Lande, R. and S.J. Arnold. 1983. The measurement of selection on correlated characters. Evolution 37(6): 1210-1226.

Landis, J.R. and G.G. Koch. 1977. The measurement of observer agreement for categorical data. Biometrics 33: 159-174.

Lawrence, D.M., H.S. Adams, and S.L. Stephenson. 1997. “Upland Forest Communities in the Mid-Appalachian Region of Eastern North America”, pp. 27-52 in Hitt, T.P. (ed.), Proceedings from the 1996 Central Appalachian Ecological Integrity Conference. Heartwood, Bloomington, IN. 108 pp.

Lawson, D.P. 2006. Habitat use, home range, and activity patterns of Hingeback Tortoises, Kinixys erosa and K. homeana, in Southwestern Cameroon. Chelonian Conservation and Biology 5(1): 48-56.

361 Lee, P., C. Smyth, and S. Boutin. 2004. Quantitative review of riparian buffer width guidelines from Canada and the United States. Journal of Environmental Management 70: 165–180.

Lele, S.R., E.H. Merrill, J. Keim, and M.S. Boyce. 2013. Selection, use, choice, and occupancy: Clarifying concepts in resource selection studies. Journal of Animal Ecology 82: 1183-1191.

Levin, S.A. 1992. The problem of pattern and scale in ecology: The Robert H. MacArthur Award Lecture. Ecology 73(6): 1943-1967.

Lichvar, R.W., M. Butterwick, N.C. Melvin, and W.N. Kirchner. 2014. The National Wetland Plant List: 2014 Update of Wetland Ratings; Eastern mountains and piedmont 2014 Regional Wetland Plant List. Phytoneuron 2014-41: 1-42.

Likens, G.E., F.H. Bormann, R.S. Pierce, and W.A. Reiners. 1978. Recovery of a deforested ecosystem. Science 199(4328): 492-496.

Lindeman, P. 2000. Resource use of five sympatric turtle species: Effects of competition, phylogeny, and morphology. Canadian Journal of Zoology 78: 992–1008.

Lindenmayer, D. and M. Burgman. 2005. Practical Conservation Biology. CSIRO Publishing, Victoria, Australia. 625 pp.

Lindenmayer, D. et al. 2008. A checklist for ecological management of landscapes for conservation. Ecology Letters 11: 78-91.

Lipscomb, M.V. and E.T. Nilsen. 1990. Environmental and physiological factors influencing the natural distribution of evergreen and deciduous ericaceous shrubs on northeast- and southwest-facing slopes of the Southern Appalachian Mountains. II. Water relations. American Journal of Botany 77(4): 517-526.

Litzgus, J.D., F. Bolton, and A.I. Schulte-Hostedde. 2008. Reproductive output depends on body condition in Spotted Turtles (Clemmys guttata). Copeia 2008(1): 86–92.

Loewenstein, E.F. 2008. “Silviculture and the long-term dynamics of single-tree selection on Pioneer Forest”, pp. 43-48 in Guldin, J.M., G.F. Iffrig, and S.L. Flader (eds.), Pioneer Forest— A half century of sustainable uneven-aged forest management in the Missouri Ozarks. Gen. Tech. Rep. SRS–108.

362 Asheville, NC: U.S. Department of Agriculture, Forest Service, Southern Research Station. 123 pp.

Lookingbill, T.R. and D.L. Urban 2005. Gradient analysis, the next generation: Towards more plant-relevant explanatory variables. Canadian Journal of Forest Research 35: 1744–1753.

Lopez, P., I. Marcos, and J. Martin. 2005. Effects of habitat-related visibility on escape decisions of the Spanish Terrapin Mauremys leprosa. Amphib. Reptil. 26: 557—561.

Lorimer, C.G. and A.S. White. 2003. Scale and frequency of natural disturbances in the northeastern US: Implications for early successional forest habitats and regional age distributions. Forest Ecology and Management 185: 41–64.

Lourenco, J.M., S. Glemin, Y. Chiari, and N. Galtier. 2012. The determinants of the molecular substitution process in turtles. Journal of Evolutionary Biology 26: 38-50.

Lövei, G.L., T. Magura, B. Tóthmérész, and V. Ködöböcz. 2007. The influence of matrix and edges on species richness patterns of ground beetles (Coleoptera: Carabidae) in habitat islands. Global Ecology and Biogeography 15: 283– 289.

Lovich, J.E., C.H. Ernst, and J.F. McBreen. 1990. Growth, maturity, and sexual dimorphism in the wood turtle, Clemmys insculpta. Canadian Journal of Zoology 68: 672-677.

Lue, K-Y and T-H Chen. 1999. Activity, movement patterns, and home range of the yellow-margined box turtle (Cuora flavomarginata) in northern Taiwan. Journal of Herpetology 33: 590-600.

Luiselli, L. 2005. Aspects of comparative thermal ecology of sympatric hinge-back tortoises (Kinixys homeana and Kinixys erosa) in the Niger Delta, southern Nigeria. African Journal of Ecology 43: 64–69.

Luiselli, L. 2006. Resource partitioning in the communities of terrestrial turtles: A review of the evidences. Revue d’ Ecologie: La Terre et la Vie 61(4): 353- 365.

MacArthur, R.H. and E.R. Pianka. 1966. An optimal use of a patchy environment. The American Naturalist 100(916): 603-609.

363 Mahmood, R., R.A. Pielke, Sr., K.G. Hubbard, D. Niyogi, P.A. Dirmeyer, C. McAlpine, A.M. Carleton, R. Hale, S. Gameda, A. Beltra -́ Przekurat, B. Baker, R. McNider, D.R. Legates, M. Shepherd, J. Du, P.D. Blanken, O.W. Frauenfeld, U.S. Nair, and S. Fall. 2014. Land cover changes and their biogeophysical effects on climate. International Journal of Climatology 34: 929-953.

Manly, B.F.J., L.L. McDonald, D.L. Thomas, T.L. McDonald, and W.P. Erickson (Eds.). 2002. Resource Selection by Animals: Statistical Analysis and Design for Field Studies, second ed. Kluwer, Boston, USA. 222 pp.

Manning, J.A. and W.D. Edge. 2004. Small mammal survival and downed wood at multiple scales in managed forests. Journal of Mammalogy 85(1): 87-96.

Marsack, K. and B.J. Swanson. 2009. A genetic analysis of the impact of generation time and road-based habitat fragmentation on Eastern Box Turtles (Terrapene c. carolina). Copeia 2009(4): 647–652.

Martin, K. and M. Sommer. 2004. Relationships between land snail assemblage patterns and soil properties in temperate-humid forest ecosystems. Journal of Biogeography 31: 531–545.

Matlack, G.R. 2013. Reassessment of the use of fire as a management tool in deciduous forests of Eastern North America. Conservation Biology 26(5): 916-926.

Matlack, G.R. and J.R. Schaub. 2011. Long-term persistence and spatial assortment of nonnative plant species in second-growth forests. Ecography 34: 649-658.

McCarthy, B.C., T.L. Vierheller, and W.A. Wistendahl. 1984. Species ordination of upper slope oak-hickory stands of southeastern Ohio. Bulletin of Torrey Botanical Club 111: 56-60.

McCarthy, J. 2001. Gap dynamics of forest trees: A review with particular attention to boreal forests. Environmental Review 9: 1–59.

McCoard, K.R.P., A.A. Billings, and J.T. Anderson. 2016a. Wood Turtle home range and habitat use in the central Appalachians. Chelonian Conservation and Biology 15(2): 173-180.

McCoard, K.R.P., N.S. McCoard, P.J. Turk, and J.T. Anderson. 2016b. Habitat characteristics that influence the occurrence of wood turtles at the southern limits of their range in the Central Appalachians. Journal of Herpetology 50: 381–387.

364 McCormick, J. F. and R. B. Platt. 1980. Recovery of an Appalachian forest following the chestnut blight or Catherine Keever-you were right! American Midland Naturalist 104: 264-273.

McCoy, E.D., K.A. Basiotis, K.M. Connor, and H.R. Mushinsky. 2013. Habitat selection increases the isolating effect of habitat fragmentation on the gopher tortoise. Behavioral Ecology and Sociobiology 67: 815–821.

McDonald, J.H. 2014. Handbook of Biological Statistics (3rd ed.). Sparky House Publishing, Baltimore, Maryland. The web page contains the content of pages 53-58 in the printed version – downloadable spreadsheet for goodness-of-fit G-test accessed at http://www.biostathandbook.com/gtestgof.html

McDonald, L., B. Manly, F. Huettmann, and W. Thogmartin. 2013. Location-only and use-availability data: Analysis methods converge. Journal of Animal Ecology 82: 1120-1124.

McEwan, R.W., R.N. Muller, and B.C. McCarthy. 2005. Vegetation-Environment relationships among woody species in four canopy-layers in an old-growth mixed mesophytic forest. Castanea 70(1): 32–46.

McEwan, R.W. and R.N. Muller. 2006. Spatial and temporal dynamics in canopy dominance of an old-growth central Appalachian forest. Canadian Journal of Forestry Research 36: 1536–1550.

McEwan, R.W. and R.N. Muller. 2011. Dynamics, diversity, and resource gradient relationships in the herbaceous layer of an old-growth Appalachian forest. Plant Ecology 212: 1179-1191.

McEwan, R.W., J.M. Dyer, and N. Pederson. 2010. Multiple interacting ecosystem drivers: Toward an encompassing hypothesis of oak forest dynamics across eastern North America. Ecography 33: 1-13.

McGinnis, S.M. and W.G. Voigt. 1971. Thermoregulation in the Desert Tortoise, Gopherus agassizii. Comparative Biochemistry and Physiology Part A: Physiology 40(1): 119-126.

McKnight, D.T. 2011. Observed plant preferences of Eastern Box Turtles (Terrapene carolina carolina) in a Maryland forest. Herpetology Notes 4: 97-102.

365 McMinn, J.W. and D.A. Crossley, Jr. (eds.). 1996. Biodiversity and Coarse Woody Debris in Southern Forests. General Technical Report SE-94. USDA FS Southeast Research Station, Asheville, NC. 146 pp.

McMinn, J.W. and R.A. Hardt. 1996. “Accumulations of coarse woody debris in southern forests”, pp. 1-9 in McMinn, J.W. and D.A. Crossley, Jr., (eds.), Biodiversity and Coarse Woody Debris in Southern Forests. General Technical Report SE-94. USDA FS Southeast Research Station, Asheville, NC.

McNab, B.K. 1963. Bioenergetics and the determination of home range size. The American Naturalist 97: 133–140.

McNab, W.H. and P.E. Avers (comps.). 1994. Ecological Subregions of the United States: Section Descriptions. Administrative Publication WO-WSA-5. USDA Forest Service, Washington, DC. 267 pp.

McNab, W.H., D.T. Cleland, J.A. Freeouf, J.E. Keys, Jr., G.J. Nowacki, and C.A. Carpenter (comps.). 2007. Description of ecological subregions: Sections of the conterminous United States [CD-ROM]. Gen. Tech. Report WO-76B. Washington, DC: U.S. Department of Agriculture, Forest Service. 80 pp. and map.

Meek, R. and R.A. Avery. 1988. Mini-review: Thermoregulation in chelonians. Herpetological Journal 1: 253-259.

Meier, A.J., S.P. Bratton, and D.C. Duffy. 1995. Possible ecological mechanisms for loss of vernal-herb diversity in logged eastern deciduous forests. Ecological Applications 5: 935-946.

Millar, C.S. and G. Blouin-Demers. 2011. Spatial ecology and seasonal activity of Blanding’s Turtles (Emydoidea blandingii) in Ontario, Canada. Journal of Herpetology 45(3): 370-378.

Miller, G.W. and J.N. Kochenderfer. 1998. Maintaining species diversity in the Central Appalachians. Journal of Forestry 96(7): 28-33.

Mitchell, J.C. 1994. “Wood Turtle”, pp. 88-91 in The Reptiles of Virginia. Smithsonian Institution Press, Washington D.C. 352 pp.

Mitchell, J.C., A.R. Breisch, and K.A. Buhlmann. 2006. Habitat Management Guide- lines for Amphibians and Reptiles of the Northeastern United States.

366 Partners in Amphibian and Reptile Conservation, Technical Publication HMG-3, Montgomery, Alabama. 108 pp.

Mitchell, M.S and R.A. Powell. 2007. Optimal use of resources structures home ranges and spatial distribution of black bears. Animal Behavior 74: 219-230.

Mitchell, M.S. and R.A. Powell. 2003. “Linking fitness landscapes with the behavior and distribution of animals”, pp. 93-124 in Bissonette, J. and I. Storch (eds.), Landscape Ecology and Resource Management: Linking Theory With Practice, Island Press, Washington, D.C. 463 pp.

Mitro, M. 2003. Demography and viability analyses of a diamondback terrapin population. Canadian Journal of Zoology 81: 716-726.

Mladenoff, D.J., M.A. White, J. Pastor, and T.R. Crow. 1993. Comparing spatial pattern in unaltered old-growth and disturbed forest landscapes. Ecological Applications 3(2): 294-206.

Morris, D.W. 1992. Scales and costs of habitat selection in heterogeneous landscapes. Evolutionary Ecology 6: 412-432.

Morris, D.W., R.G. Clark, and M.S. Boyce. 2008. Habitat and habitat selection: Theory, tests, and implications. Israel Journal of Ecology & Evolution 54: 287-294.

Morrison, M.L., B.G. Marcot, and R.W. Mannan. 2006. Wildife-Habitat Relationships: Concepts and Applications. Third Edition. Island Press, Washington, D.C. 493 pp.

Moskovits D.K. and A.R. Kiester. 1987. Activity levels and ranging behaviour of the two Amazonian tortoises, Geochelone carbonaria and Geochelone denticulata, in north-western Brazil. Functional Ecology 1: 203–214.

Moulherat, S., V. Delmas, T. Slimani, E.H. El Mouden, T. Louzizi, F. Lagarde, and X. Bonnet. 2014. How far can a tortoise walk in open habitat before overheating? Implications for conservation. Journal for Nature Conservation 22: 186–192.

Muegel, L.A. and D.L. Claussen. 1994. Effects of slope on voluntary locomotor performance in the turtle, Terrapene carolina carolina. Journal of Herpetology 28(1): 6-11.

367 Mueller, R. 2000. “Stability relations in forests”, in Forests of the Central Appalachians Project, accessible at http://www.asecular.com/forests/stability.htm

Mueller, T. and W.F. Fagan. 2008. Search and navigation in dynamic environments: From individual behaviors to population distributions. Oikos 117(5): 654-664.

Mysterud, A. and R.A. Ims. 1998. Functional responses in habitat use: Availability influences relative use in trade-off situations. Ecology 79(4): 1435-1441.

National Oceanic and Atmospheric Administration. 2013. Climate data from National Climatic Data Center [“NCDC”]. Accessed at http://www.ncdc.noaa.gov/land-based-station-data/climate-normals

National Resource Conservation Service [NRCS]. Soil surveys for VA and WV. Available at https://www.nrcs.usda.gov/wps/portal/nrcs/soilsurvey/soils/survey/state/

NatureServe. 2012. “Wood Turtle” report at NatureServe Explorer: An online encyclopedia of life [web application]. Version 7.1. NatureServe, Arlington, Virginia, available at http://www.natureserve.org/explorer

Newcomb, L. 1977. Newcomb’s Wildflower Guide. Little, Brown and Company, Boston. 490 pp.

Nichols, W.F., K.T. Killingbeck, and P.V. August. 1998. The influence of geomorphological heterogeneity on biodiversity: II. A landscape perspective. Conservation Biology 12(2): 371-379.

Niederberger, A.J. and M.E. Seidel. 1999. Ecology and status of a Wood Turtle (Clemmys insculpta) population in West Virginia. Chelonian Conservation and Biology 3(3): 414-418.

Nilsen, E.B., I. Herfindal, and J.D.C. Linnell. 2005. Can intra-specific variation in carnivore home-range size be explained using remote-sensing estimatesof environmental productivity? Ecoscience 12(1): 68-75.

Noss, R.F. 1990. Indicators for monitoring biodiversity: A hierarchical approach. Conservation Biology 4: 355-364.

Noss, R.F. 1991. Effects of edge and internal patchiness on avian habitat use in an old-growth Florida hammock. Natural Areas Journal 11: 34-47.

368 Nutting, W.L. and T.E. Graham. 1993. Preferred body temperatures in five Nearctic freshwater turtles: A preliminary study. Comparative Biochemistry and Physiology A 104(2): 243-246.

O’Grady, J.J., D.H. Reed, B.W. Brook, and R. Frankham. 2004. What are the best correlates of predicted extinction risk? Biological Conservation 118: 513– 520.

Oke, T. R., 1997. “Surface climate processes”, pp. 21–43 in W. G. Bailey, T. R. Oke and W. R. Rouse (eds.), The Surface Climates of Canada. McGill- Queens’ University Press, Montreal, Quebec and Kingston, Ontario.

Oliver, C.D. and B.C. Larson. 1996. Forest Stand Dynamics, updated ed. John Wiley & Sons, New York. 544 pp.

Olivero, A.M. and D.M. Hix. 1998. Influence of aspect and stand age on ground flora of southeastern Ohio forest ecosystems. Plant Ecology 139: 177-187.

Omernik, J.M. and R.G. Bailey. 1997. Distinguishing between watersheds and ecoregions. Journal of the American Water Resource Association 33(5): 935- 949.

Owen-Smith, N., J.M. Fryxell, and E.H. Merrill. 2010. Foraging theory upscaled: The behavioral ecology of herbivore movement. Philosophical Transactions of the Royal Society B 365: 2267-2278.

Parmenter, R.R. 1980. Effects of food availability and water temperature on the feeding ecology of Pond Sliders (Chrysemys s. scripta). Copeia 1980: 503– 514.

Parmenter, R. R. 1981. Digestive turnover rates in freshwater turtles: The influence of temperature and body size. Comp. Biochem. Physiol. A Physiol. 70: 235- 238.

Parren, S.G. 2013. A twenty-five year study of the Wood Turtle (Glyptemys insculpta) in Vermont: Movements, behavior, injuries, and death. Herpetological Conservation and Biology 8(1): 176-190.

Paterson, J.E., B.D. Steinberg, and J.D. Litzgus. 2012. Generally specialized or especially general? Habitat selection by Snapping Turtles (Chelydra serpentina). Canadian Journal of Zoology 90(2): 139-149.

Peet, R.K., T.R. Wentworth, and P.S. White. 1998. A flexible, multipurpose method for recording vegetation composition and structure. Castanea 63: 262-274.

369

Penick, D.N., J. Congdon, J.R. Spotila, and J.B. Williams. 2002. Microclimates and energetics of free-living Box Turtles, Terrapene carolina, in South Carolina. Physiological and Biochemical Zoology 75(1): 57-65.

Pickett, S.T.A. and K.H. Rogers. 1997. “Patch dynamics: The transformation of landscape structure and function”, pp. 101-127 in J. Bissonette (ed.), Wildlife and Landscape Ecology: Effects of Pattern and Scale. Springer Verlag, New York. 410 pp.

Pickett, S.T.A. and M.L. Cadenasso. 1995. Spatial heterogeneity in ecological systems. Science 269: 331-334.

Pittman, S.E. 2011. Demographic and genetic status of an isolated population of bog turtles (Glyptemys muhlenbergii): Implications for managing small populations of long-lived animals. Conservation Genetics 12: 1589-1601.

Plummer, M.V. 2003. Activity and thermal ecology of the Box Turtle, Terrapene ornata, at its southwestern range limit in Arizona. Chelonian Conservation and Biology 4(3): 569–577.

Polo-Cavia, N., P. Lopez, and J. Martın. 2012. Feeding status and basking requirements of freshwater turtles in an invasion context. Physiology and Behavior 105: 1208-1213.

Pope, C.H. 1939. “Wood Turtle”, pp. 92-99 in Turtles of the United States and Canada. Alfred A. Knopf, New York. 343 pp.

Powell. R.A. 2000. “Animal home ranges and territories and home range estimators”, pp. 65-110 in L. Boitani and T.K. Fuller (eds.), Research Techniques in Animal Ecology: Controversies and Consequences. Columbia University Press, New York. 464 pp.

Power, M.E., W.J. Matthews, and A.J. Stewart. 1985. Grazing minnows, piscivorous bass and stream algae: Dynamics of a strong interaction. Ecology 66: 1448- 1456.

Primack, R. 2010. Essentials of Conservation Biology. Sinauer Associates, Sunderland, MA. 601 pp.

Pringle, R.M., J.K. Webb, and R. Shine. 2003. Canopy structure, microclimate, and habitat selection by a nocturnal snake, Hoplocephalus bungaroides. Ecology 84(10): 2668-2679.

370

Profitt, K. and B. Chance. 2003. “Status of the Wood Turtle (Clemmys insculpta) on the coastal plain of Harford County, Maryland”, pg. 109 in Swarth, C.W., W.M. Roosenburg, and E. Kiviat (eds.), Conservation and Ecology of Turtles of the Mid-Atlantic Region: A Symposium Patuxent Wildlife Refuge, Laurel, Maryland 30-31 October 1999. Bibliomania!, Salt Lake City, Utah. 122 pp.

Quesnelle, P.E., L. Fahrig, and K.E. Lindsay. 2013. Effects of habitat loss, habitat configuration and matrix composition on declining wetland species. Biological Conservation 160: 200-208.

Quinlan, M., C.W. Swarth, and M. Marchand. 2003. “Characteristics of a high- density Eastern Box Turtle population on Maryland’s coastal plain”, pg. 110 in Swarth, C.W., W.M. Roosenburg, and E. Kiviat (eds.), Conservation and Ecology of Turtles of the Mid-Atlantic Region: A Symposium Patuxent Wildlife Refuge, Laurel, Maryland 30-31 October 1999. Bibliomania!, Salt Lake City, Utah.

Quinn, N.W.S. and D.P. Tate. 1991. Seasonal movements and habitat of wood turtles (Clemmys insculpta) in Algonquin Park, Canada. Journal of Herpetology 25: 217-220.

Quintero, I. and J.J. Wiens. 2013. Rates of projected climate change dramatically exceed past rates of climatic niche evolution among vertebrate species. Ecology Letters 16, 1095-1103.

R Core Team. 2015. R: a language and environment for statistical computing. Vienna: R Foundation for Statistical Computing.

Reagan, D.P. 1974. Habitat selection in the three-toed box turtle, Terrapene carolina triunguis. Copeia 1974: 512–527.

Reed, D.H. 2005. Relationship between population size and fitness. Conservation Biology 19(2): 563-568.

Reed, D.H. and R. Frankham. 2003. Correlation between fitness and genetic diversity. Conservation Biology 17: 230–237.

Reed, M.J. and E.D. McCoy. 2014. Relation of minimum viable population size to biology, time frame, and objective. Conservation Biology 28(3): 867-870.

Reed, R.N. and J.W. Gibbons. 2003. Conservation status of live U.S. non-marine turtles in domestic and international trade. Report to US Department of the

371 Interior and US Fish and Wildlife Service. 92 pp. Accessed at www.tiherp.org/docs/Library/Turtle_trade_report.pdf

Refsnider, J.M. and M.H. Linck. 2012. Habitat use and movement patterns of Blanding’s Turtles (Emydoidea blandingii) in Minnesota, USA: A landscape approach to species conservation. Herpetological Conservation and Biology 7: 185-195. Regal, P.J. 1966. Thermophilic response following feeding in certain reptiles. Copeia 1966(3): 588-590.

Remsberg, A.J., T.L. Lewis, P.W. Huber, and K.A. Asmus. 2006. Home ranges of Wood Turtles (Glyptemys insculpta) in northern Michigan. Chelonian Conservation and Biology 5(1): 42–47.

Rentch, J.S. 2006. Structure and functioning of terrestrial ecosystems in the Eastern Rivers and Mountains Network: Conceptual models and vital signs monitoring. Technical Report NPS/NER/NRR-2006/007. National Park Service, Philadelphia, PA. Accessed at http://www.nps.gov/nero/science/

Rentch, J.S., R.H. Fortney, S.L. Stephenson, H.S. Adams, W.N. Grafton, R.B. Coxe, and H.H. Mills. 2005. Vegetation patterns within the lower Bluestone River gorge in southern West Virginia. Castanea 70(3): 184-203.

Rentch, J.S., M.A. Fajvan, and R.R. Hicks Jr. 2003a. Oak establishment and canopy accession strategies in five old-growth stands in the central hardwood forest region. Forest Ecology and Management 184: 285–297.

Rentch, J.S., M.A. Fajvan, and R.R. Hicks Jr. 2003b. Spatial and temporal disturbance characteristics of oak-dominated old-growth stands in the central hardwood forest region. Forest Science 49(5): 778-789.

Reynolds-Hogland, M.J., M.S. Mitchell, and R.A. Powell. 2006. Spatio-temporal availability of soft mast in clearcuts in the Southern Appalachians. Forest Ecology and Management 237: 103–114.

Riitters, K.H. 2007. “Forest Fragmentation”, pp. 9-15 in Ambrose, M. and B. Conkling (eds.), Forest health monitoring: 2005 national technical report. Gen. Tech Rep. SRS-104. Asheville, NC: USDA FS Southern Research Station.

Riitters, K.H., J.D. Wickham, R.V. O’Neill, K.B. Jones, E.R. Smith, J.W. Coulston, T.G. Wade, and J.H. Smith. 2002. Fragmentation of continental United States forests. Ecosystems 5: 815–822.

372

Ripple, W.J. and R.L. Beschta 2004. Wolves and the ecology of fear: Can predation risk restructure ecosystems? BioScience 54: 755–766.

Rittenhouse, T.A.G., E.B. Harper, L.R. Rehard, and R.D. Semlitsch. 2008. The role of microhabitats in the desiccation and survival of anurans in recently harvested oak–hickory forest. Copeia 2008(4): 807-814. Rivera-Ortíz, F.A.R. Aguilar, M.D.C. Arizmendi, M. Quesada, and K. Oyama. 2015. Habitat fragmentation and genetic variability of tetrapod populations. Animal Conservation 18(3):249-258.

Roberts, M.R. 2004. Response of the herbaceous layer to natural disturbance in North American forests. Canadian Journal of Botany 82: 1273-1283.

Robertson, B.A., J.S. Rehage, and A. Sih. 2013. Ecological novelty and the emergence of evolutionary traps. Trends in Ecology & Evolution 28: 552- 561.

Roe, J.H. and A. Georges. 2007. Heterogeneous wetland complexes, buffer zones, and travel corridors: Landscape management for freshwater reptiles. Biological Conservation 135: 67–76.

Roever, C.L., H.L. Beyer, M.J. Chase, and R.J. van Aarde. 2014.The pitfalls of ignoring behavior when quantifying habitat selection. Diversity and Distributions 20: 322-333.

Rogers, R.S. 1982. Early spring herb communities in mesophytic forests of the Great Lakes region. Ecology 63: 1050-1063.

Rollinson, N., R.G. Farmer, and R.J. Brooks. 2012. Widespread reproductive variation in North American turtles: Temperature, egg size and optimality. Zoology 115: 160-169.

Rooney, T.P., S.M. Wiegmann, D.A. Rogers, and D.M. Waller. 2004. Biotic impoverishment and homogenization in unfragmented forest understory communities. Conservation Biology 18(3): 787-798.

Roosenburg, W.M., K.L. Haley, and S. McGuire. 1999. Habitat selection and movements of Diamondback Terrapins, Malaclemys terrapin, in a Maryland estuary. Chelonian Conservation and Biology 3(3): 425-429.

Rosenzweig, M.L. 1968. Net primary productivity of terrestrial communities: Prediction from climatological data. The American Naturalist 102(923): 67- 74.

373 Ross, D.A., K.N. Brewster, R.K. Anderson, N. Ratner, and C.M. Brewster. 1991. Aspects of the ecology of Wood Turtles, Clemmys insculpta, in Wisconsin. Canadian Field-Naturalist 105: 363-367.

Ross, D.A. and R.K. Anderson. 1990. Habitat use, movements, and nesting of Emydoidea blandingii in central Wisconsin. Journal of Herpetology 24: 6-12.

Ross, D.A., K.N. Brewster, R.K. Anderson, N. Ratner, and K.M. Brewster. 1991. Aspects of the ecology of Wood Turtles, Clemmys insculpta, in Wisconsin. The Canadian Field Naturalist 105: 363-67.

Rossell, Jr., C.R., I.M. Rossell, and S. Patch. 2006. Microhabitat selection by Eastern Box Turtles (Terrapene c. carolina) in a North Carolina mountain wetland. Journal of Herpetology 40(2): 280–284.

Row, J.R. and G. Blouin-Demers. 2006. Thermal quality influences habitat selection at multiple scales in milksnakes. Ecoscience 13(4): 443-450.

Rowe, J.W. and S.F. Dalgarn. 2009. Effects of sex and microhabitat use on diel body temperature variation in Midland Painted Turtles (Chrysemys picta marginata). Copeia 2009(1): 85-92.

Rubino, D.L. and B.C. McCarthy. 2003. Composition and ecology of macrofungal and myxomycete communities on oak woody debris in a mixed-oak forest of Ohio. Canadian Journal of Forest Research 33: 2151–2163.

Runkle, J.R. 1985. “Disturbance regimes in temperate forests”, pp.17–33 in S.T.A. Pickett and P.S. White (eds.), The ecology of natural disturbance and patch dynamics. Academic Press, San Diego, California. 472 pp.

Runkle, J.R. 1990. Gap dynamics in an Ohio Acer-Fagus forest and speculations on the geography of disturbance. Canadian Journal of Forestry Research 20: 632-641.

Runkle, J.R. 1991a. Gap dynamics of old-growth eastern forests: Management implications. Natural Areas Journal 11(1): 19-25.

Runkle, J.R. 1991b. “Natural disturbance regimes and the maintenance of stable regional floras”, pp. 31-47 in Restoration of Old Growth Forests in the Interior Highlands of Arkansas and Oklahoma. Ouchita National Forest, Hot Springs, Arkansas. 190 pp.

Rutherford, J.L., J.S. Casper, and B. Graves. 2016. Factors affecting predation on

374 Wood Turtle (Glyptemys insculpta) nests in the upper peninsula of Michigan. Chelonian Conservation and Biology 15(2): 181–186.

Sajwaj, T.D. and J.W. Lang. 2000. Thermal ecology of Blanding’s Turtle in central Minnesota. Chelonian Conservation and Biology 3: 626–636.

Sanderson, E.W., J. Malanding, M.A. Levy, K.H. Redford, A.V. Wannebo, and G. Woolmer. 2002. The human footprint and the last of the wild. BioScience 52: 891–904.

Sarkar, S.N., K. Sarkar, P. Das, and B.R. Maiti. 1996. Photothermal effects on ovarian growth and function in the Soft-Shelled Turtle Lissemys punctata punctata. Journal of Experimental Zoology 274: 41–55.

Saumure, R.A. 1992. Clemmys insculpta (Wood Turtle) Size. Herpetological Review 23(4): 116.

Saumure, R.A. 2004. “The spatial ecology and conservation of the North American wood turtle (Glyptemys insculpta) in a fragmented agri-forest landscape.” PhD dissertation, McGill University, Montreal, Quebec, Canada.

Saunders, S.C., J. Chen, T.R. Crow, and K.D. Brosofske. 1998. Hierarchical relationships between landscape structure and temperature in a managed forest landscape. Landscape Ecology 13: 381-395.

Scheff, R.J. 2012. "The Development Of Old-Growth Structural Characteristics In Second-Growth Forests Of The Cumberland Plateau, Kentucky, U.S.A." Master’s thesis, Eastern Kentucky University, Richmond, KY. Available at Online Theses and Dissertations, Paper 116.

Schlossberg, S. and D.I. King. 2009. Modeling animal habitats based on cover types: A critical review. Environmental Management 43: 609-618.

Schmidt-Nielson, K. and P.J. Bentley. 1966. Desert tortoise Gopherus agassizii: Cutaneous water loss. Science 154: 911.

Schoener, T.W. 2011. The newest synthesis: Understanding the interplay of evolutionary and ecological dynamics. Science 331: 426-429.

Schoener, T.W. and A. Schoener. 1982. Intraspecific variation in home-range size in some Anolis lizards. Ecology 63: 809–823.

Seebacher, F. and C.E. Franklin. 2012. Determining environmental causes of biological effects: The need for a mechanistic physiological dimension in

375 conservation biology. Philosophical Transactions of the Royal Society B 367: 1607-1614.

Seebacher, F. and G.C. Grigg. 1997. Patterns of body temperature in wild Freshwater Crocodiles, Crocodylus johnstoni: Thermoregulation versus thermoconformity, seasonal acclimatization, and the effect of social interactions. Copeia 1997(3): 549-557.

Seibert, S. and W.R. Belzer. 2015. Diverse movement patterns of North America’s Eastern Box Turtle (Terrapene carolina L.). Part 3: Shifts in residency fidelity. IRCF Reptiles and Amphibians 22(2): 56-67.

Seigel, R. A. and J. T. Collins. 1993. Snakes: Ecology and Behavior. McGraw-Hill, Inc., New York, New York, USA.

Semlitsch, R.D. and J.R. Bodie. 2003. Biological criteria for buffer zones around wetlands and riparian habitats for amphibians and reptiles. Conservation Biology 17:1219–1228.

Seymour, R.S., A.S. White, and P.G. DeMaynadier. 2002. Natural disturbance regimes in northeastern North America—evaluating silvicultural systems using natural scales and frequencies. Forest Ecology and Management 155: 357–367.

Shaffer, B. et al. 2013. The western painted turtle genome, a model for the evolution of extreme physiological adaptations in a slowly evolving lineage. Genome Biology 14: R28.

Shoemaker, K.T., A.R. Breisch, J.W. Jaycox, and J.P. Gibbs. 2013. Reexamining the minimum viable population concept for long-lived species. Conservation Biology 27(3): 542-551.

Shoemaker, K.T., A.R. Breisch, J.W. Jaycox, and J.P. Gibbs. 2014. Disambiguating the minimum viable population concept: Response to Reed and McCoy. Conservation Biology 28(3): 871-873.

Shugart, H.H., Jr. and D.C. West. 1981. Long-term dynamics of forest ecosystems. American Scientist 69: 647-652.

Shure, D.J. and D.L. Phillips. 1991. Patch size of forest openings and arthropod populations. Oecologia 86(3): 325-334.

Slavenko, A., Y. Itescu, F. Ihlow, and S. Meiri. 2016. Home is where the shell is: Predicting turtle home ranges. Journal of Animal Ecology 85: 106-114.

376

Small, C.J. and B.C. McCarthy. 2003. Spatial and temporal variability of herbaceous vegetation in an eastern deciduous forest. Plant Ecology 164: 37-48.

Smith, L.M. and R.P. Cherry. 2016. Movement, seasonal activity, and home range of an isolated population of Glyptemys muhlenbergii, Bog Turtle, in the Southern Appalachians. Southeastern Naturalist 15(2): 207-219.

Smith, K.A. 2002. “Demography and spatial ecology of Wood Turtles (Clemmys insculpta) in Algonquin Provincial Park”. Master’s thesis, University of Guelph, Guelph, Ontario.

Soulé, M.E. 1987. Viable Populations for Conservation. Cambridge University Press, Cambridge, UK.

Spencer, R.J., M.B. Thompson, and I.D. Hume. 1998. The diet and digestive energetics of an Australian short-necked turtle, Emydura macquarii. Comparative Biochemistry and Physiology 121A: 341–349.

Spies, T.A., J.F. Franklin, and T.B. Thomas. 1988. Coarse woody debris in Douglas- fir forests of western Oregon and Washington. Ecology 69: 1689–1702.

Spradling, T.A., J.W. Tamplin, S.S. Dow, and K.J. Meyer. 2010. Conservation genetics of a peripherally isolated population of the wood turtle (Glyptemys insculpta) in Iowa. Conservation Genetics 11: 1667-1677.

Spray, D.C. and M.L. May. 1972. Heating and cooling rates in four species of turtles. Comparative Biochemistry and Physiology 41A: 507-522.

Steen, D.A., J.P. Gibbs, K.A. Buhlmann, J.L. Carr, B.W. Compton, J.D. Congdon, J.S. Doody, J.C. Godwin, K.L. Holcomb, D.R. Jackson, F.J. Janzen, G. Johnson, M.T. Jones, J.T. Lamer, T.A. Langen, M.V. Plummer, J.W. Rowe, R.A. Saumure, J.K. Tucker, and D.S. Wilson. 2012. Terrestrial habitat requirements of nesting freshwater turtles. Biological Conservation 150: 121-128.

Stephens, P.R. and J.J. Wiens. 2009. Bridging the gap between community ecology and historical biogeography: Niche conservatism and community structure in emydid turtles. Molecular Ecology 18: 4664-4679.

Stephenson, N.L. 1998. Actual evapotranspiration and deficit: Biologically meaningful correlates of vegetation distribution across spatial scales. Journal of Biogeography 25: 855-870.

377

Stephenson, N.L. 2011. Comment on ''Changes in Climatic Water Balance Drive Downhill Shifts in Plant Species' Optimum Elevations''. Science 334: 177-c.

Stephenson, S.L. and H.H. Mills. 1999. Contrasting vegetation of noses and hollows in the Valley and Ridge Province, southwestern Virginia. Journal of the Torrey Botanical Society 126(3): 197-212.

Sterrett, S.C., L.L. Smith, S.W. Golladay, S.H. Schweitzer, and J.C. Maerz. 2011. The conservation implications of riparian land use on river turtles. Animal Conservation 14: 38-46.

Stickell, L.F. 1950. Populations and home range relationships of the Box Turtle, Terrapene c. carolina (Linnaeus). Ecological Monographs 20(4): 351-378.

Storch, I. 2003. “Linking a multiscale habitat concept to species conservation”, pp. 303-320 in J. Bissonette and I. Storch (eds.), Landscape Ecology and Resource Management: Linking Theory With Practice, Island Press, Washington, D.C. 463 pp.

Strang, C.A. 1983. Spatial and temporal activity patterns in two terrestrial turtles. Journal of Herpetology 17(1): 43-47.

Strauss, P. 2014. The Big Herbs: The use and abuse, natural history and identification of major tree and shrub species in the Midwest and Eastern U.S. with stories and insights of a life married to farm and forest. XOXOX Press. 203 pp.

Strayer, D.L., J.A. Downing, W.R. Haag, T.L. King, J.B. Layzer, T.J. Newton, and S.J. Nichols. 2004. Changing perspectives on Pearly Mussels, North America’s most imperiled animals. BioScience 54(5): 429-439.

Stromberg, L.P. 1995. “Vegetation sampling methods for use in wildlife habitat evaluation”, pp. 133-174 in Berwick, S.H. and V.B. Saharia (eds.), The Development of International Principles and Practices of Wildlife Research and Management: Asian and American Approaches, Oxford University Press, Delhi, India. 481 pp.

Sturbaum, B.A. 1981. Responses of the Three-toed Box Turtle, Terrapene carolina triunguis, to heat stress. Comparative Biochemistry and Physiology 70A: 199-204.

Suggitt, A. J., P. K. Gillingham, J. K. Hill, B. Huntley, W. E. Kunin, D. B. Roy, and C. D. Thomas. 2011. Habitat microclimates drive fine-scale variation in

378 extreme temperatures. Oikos 120: 1-8.

Swanson, P.L. 1940. The Wood Turtle. Pennsylvania Angler 9(8): 10-11.

Swanson, F.J., T.K. Kratz, N. Caine, and R.G. Woodmansee. 1988. Landform effects on ecosystem patterns and processes. BioScience 38(2): 92-98. Tamplin, J.W. 2005. Microhabitat selection of free-ranging wood turtles (Glyptemys insculpta) in Butler County, Iowa. Permit report to Iowa DNR.

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

Tiner, R.W. 2006. Lists of potential hydrophytes for the United States: A regional review and their use in wetland identification. Wetlands 26(2): 624-634.

Tingley, R., D.G. McCurdy, M.D. Pulsifer, and T.B. Herman. 2009. Spatio-temporal differences in the use of agricultural fields by male and female Wood Turtles (Glyptemys insculpta) inhabiting an agri-forest mosaic. Herpetological Conservation and Biology 4(2): 185-190.

Tingley, R., T.B. Herman, M.D. Pulsifer, D.G. McCurdy, and J.P. Stephens. 2010. Intra-specific niche partitioning obscures the importance of fine-scale habitat data in species distribution models. Biodiversity Conservation 19: 2455-2467.

Todd, B.D. and K.M. Andrews. 2008. Response of a reptile guild to forest harvesting. Conservation Biology 22(3): 753–761.

Tracy, C.R. and K.A. Christian. 1986. Ecological relations among space, time, and thermal niche axes. Ecology 67(3): 609-615.

Traill, L.W., B.W. Brook, R. Frankham, and C.J.A. Bradshaw. 2010. Pragmatic population viability targets in a rapidly changing world. Biological Conservation 143: 28–34.

Trainor, A.M., T.M. Shenk, and K.R. Wilson. 2007. Microhabitat characteristics of Preble’s Meadow Jumping Mouse high-use areas. Journal of Wildlife Management 71(2): 469-477.

Trauger, D.L., B. Czech, J.D. Erickson, P.R. Garrettson, B.J. Kernohan, and C.A. Miller. 2003. The relationship of economic growth to wildlife conservation. Wildlife Society Technical Review 03-1. The Wildlife Society, Bethesda, MD.

379

Trombulak, S.C. 1996. “The Restoration of Old Growth: Why and How”, pp. 305- 320 in Davis, M.B. (ed.), Eastern Old-Growth Forests: Prospects for Rediscovery and Recovery, Island Press, Washington, D. C.

Turc, L. 1961. Evaluation des besoins en eau d’irrigation, evapotranspiration potentielle, formule climatique simplifice et mise a jour. Annales Agronomigue 12:13–49.

Turtle Taxonomy Working Group [TTWG : Rhodin, A.G.J., J.B. Iverson, R. Bour, U. Fritz, A. Georges, H.B. Shaffer, and P.P. van Dijk]. 2017. Turtles of the World: Annotated Checklist and Atlas of Taxonomy, Synonymy, Distribution, and Conservation Status (8th Ed.), in Rhodin, A.G.J., Iverson, J.B., van Dijk, P.P., Saumure, R.A., Buhlmann, K.A., Pritchard, P.C.H., and Mittermeier, R.A. (Eds.), Conservation Biology of Freshwater Turtles and Tortoises: A Compilation Project of the IUCN/SSC Tortoise and Freshwater Turtle Specialist Group. Chelonian Research Monographs 7:1–292. doi: 10.3854/crm.7.checklist.atlas.v8.2017

Tuttle, S. 1996. “Ecology and Natural History of the Wood Turtle (Clemmys insculpta) in Southern New Hampshire”. Master’s thesis, Antioch University New England, Keene, NH.

Tuttle, S.E. and D.M. Carroll. 2003. Home range and seasonal movements of the Wood Turtle (Glyptemys insculpta) in southern New Hampshire. Chelonian Conservation and Biology 4(3): 656-663.

Ultsch, G.R. 2006. The ecology of overwintering among turtles: Where turtles overwinter and its consequences. Biological Reviews 81: 339–367.

Ulyshen, M.D. and J.L. Hanula. 2009. Litter-dwelling arthropod abundance peaks near coarse woody debris in Loblolly Pine forests of the southeastern United States. Florida Entomologist 92(1): 163-164.

United Plant Savers. 2017. Information on native medicinal plants accessed at www.unitedplantsavers.org

USDA Forest Service [“USDA FS”]. 1997. Guidance for Conserving and Restoring Old-Growth Forest Communities on National Forests in the Southern Region; Forestry Report R8-FR-62. USDA FS, Atlanta, GA. 122 pp.

USDA Forest Service. undated. Southern Region Silvicultural Examination and Prescription Field Book For CISC Version 4.02: Codes, Descriptions,

380 Tables, and Guides that facilitate the Silvicultural Examination and Prescription Process. USDA FS, Atlanta, GA. Stored as /fsapps/fsprod/cisc/programs/CISC_Field_Book.doc

USDA Forest Service. $$$$. $$$$ [name and date withheld] Timber Sale Environmental Assessment and Biological Evaluation. George Washington National Forest, Roanoke, VA.

USDI Fish and Wildlife Service. 2015. 90-day findings on various petitions to list, reclassify, or delist fish, wildlife, or plants under the Endangered Species Act of 1973, as amended (Act) [includes the Wood Turtle]. Federal Register 80(181): 56423-432.

USGS National Map site. Accessible at https://nationalmap.gov/

Van de Poll, R. 2004. “The importance of coarse woody material in fostering fungal development”, pp. 57-59 in Bennett, K.P. (technical coordinator), Moving Toward Sustainable Forestry: Lessons from Old Growth Forests, University of New Hampshire Cooperative Extension Natural Resource Network Report. 82 pp. van Dijk, P.P. and J. Harding. 2011. Glyptemys insculpta. IUCN Red List of Threatened Species Version 2011.2. Available at http://www.iucnredlist.org van Horn, B. 1983. Density as a misleading indicator of habitat quality. Journal of Wildlife Management 47: 893-901.

Van Lear, D.H. 1996. “Dynamics of coarse woody debris in southern forest ecosystems”, pp. 10-17 in McMinn, J.W., and D.A. Crossley, Jr. (eds.), Biodiversity and Coarse Woody Debris in Southern Forests. General Technical Report SE-94.

Vellend, M. 2004. “Metapopulation dynamics following habitat loss and recovery: Forest herbs in ancient and recent forests”, pp. 163-177 in Honnay, O., K. Verhayen, B. Bossuyt, and M. Hermy (eds.), Forest Biodiversity: Lessons from History for Conservation. CABI Publishing, Wallingford, UK. 285 pp.

Vellend, M., K. Verheyen, H. Jacquemyn, A. Kolb, H. van Calster, G. Peterken, and M. Hermy. 2006. Extinction debt of forest plants persists for more than a century following habitat fragmentation. Ecology 87(3): 542–548.

381 Verdon, E. and M.A. Donnelly. 2005. Population structure of Florida Box Turtles (Terrapene carolina bauri) at the southernmost limit of their range. Journal of Herpetology 39(4): 572-577.

Virginia Department of Game and Inland Fisheries [VDGIF]. 2005. Virginia’s comprehensive wildlife conservation strategy. Virginia Department of Game and Inland Fisheries, Richmond, Virginia. 717 pp. plus appendices. Available at www.bewildvirginia.org/wildlifeplan/virginia-wildlife-action- plan.pdf

Wahl, C.M, A. Neils, and D. Hooper. 2013. Impacts of land use at the catchment scale constrain the habitat benefits of stream riparian buffers. Freshwater Biology 58: 2310-2324.

Walde, A.D., J.R. Bider, C. Daigle, D. Masse, J.-C. Bourgeios, J. Jutras, and R.D. Titman. 2003. Ecological aspects of a Wood Turtle, Glyptemys insculpta, population at the northern limit of its range in Quebec. The Canadian Field Naturalist 117: 377-388.

Waldron, J.L., S.M. Welch, and S.H. Bennett. 2008. Vegetation structure and the habitat specificity of a declining North American reptile: A remnant of former landscapes. Biological Conservation 141: 2477–2482.

Wallace, A.R. 1876. The Geographical Distribution of Animals: With a Study of the Relations of Living and Extinct Faunas as Elucidating the Past Changes of the Earth’s Surface. Harper and Brothers, New York. 574 pp.

Watt, A.S. 1947. Pattern and process in the plant community. Journal of Ecology 35: 1-22.

Weakley, A.S., J.C. Ludwig, and J.F. Townsend. 2012. Flora of Virginia. BRIT Press, Fort Worth, TX. 1554 pp.

Weatherhead. P.J., J.H. Sperry, G.L. Carfagno, and G. Blouin-Demers. 2012. Latitudinal variation in thermal ecology of North American ratsnakes and its implications for the effect of climate warming on snakes. Journal of Thermal Biology 37: 273-281.

Weathers, W.W. and F.N. White. 1971. Physiological thermoregulation in turtles. American Journal of Physiology 221(3): 704-710.

382 Webb, W., S. Szarek, W. Lauenroth, R. Kinerson, and M. Smith. 1978. Primary production and water use in native forest, grassland and desert ecosystems. Ecology 59(6): 1239-1247.

Webster, C.R. and M.A. Jenkins. 2005. Coarse woody debris dynamics in the southern Appalachians as affected by topographic position and anthropogenic disturbance history. Forest Ecology and Management 217: 319–330. Weiss, J.A. 2009. “Demographics, activity, and habitat selection of the Eastern Box Turtle (Terrapene c. carolina) in West Virginia”. Master’s thesis, Marshall University, Huntington, WV.

Welsh, Jr., H.H. 1990. Relictual amphibians and old-growth forests. Conservation Biology 4: 309–319.

West Virginia Division of Natural Resources (WVDNR). 2005. West Virginia Wildlife Conservation Action Plan. 1057 pp. Available at http://www.wildlifeactionplans.org/west_virginia.html

Whiles, M.R. and J.W. Grubaugh. 1996. “Importance of coarse woody debris to southern forest herpetofauna”, pp. 94-100 in McMinn, J.W. and D.A. Crossley, Jr. (eds.), Biodiversity and Coarse Woody Debris in Southern Forests, USDA FS Gen. Tech. Report SE-94. USDA FS Southeast Research Station, Asheville, NC. 146 pp.

White, A.S., S. Fraver, E.L. Rowland, and U. Chokkalingam. 2005. “The disturbance history of northern Maine old-growth forests”, pp. 24-25 in Bennett, K.P. (technical coordinator), Moving Toward Sustainable Forestry: Lessons from Old Growth Forests, University of New Hampshire Cooperative Extension Natural Resource Network Report. 82 pp.

White, P.S. and R.D. White. 1996. “Old-growth Oak and Oak-Hickory Forests”, pp. 178-198 in Davis, M.B. (ed.), Eastern Old-Growth Forests: Prospects for Rediscovery and Recovery, Island Press, Washington, D. C.

Whitney, G.G. and D.R. Foster. 1988. Overstorey composition and age as determinants of the understorey flora of woods of central New England. Journal of Ecology 76: 867-876.

Whittaker, R.H. 1956. Vegetation of the Great Smoky Mountains. Ecological Monographs 26: 1-80.

383 Willems, E.P. and R.A. Hill. 2009. Predator-specific landscapes of fear and resource distribution: Effects on use. Ecology 90: 546-555.

Williams, T.A. and J.L. Christiansen 1981. The niches of two sympatric softshell turtles, Trionyx muticus and Trionyx spiniferus, in Iowa. Journal of Herpetology 15(3): 303-308.

Willi, Y. and A.A. Hoffman. 2009. Demographic factors and genetic variation influence population persistence under environmental change. Journal of Evolutionary Biology 22: 124–133.

Willoughby, J., T.L. Lewis, and P.W. Huber. 2008. Site fidelity of wood turtles (Glyptemys insculpta) in the northern Lower Peninsula of Michigan. Presentation at the 93rd ESA Annual Meeting -- August 8, 2008 – abstract contained on official program CD.

Wills, C.A. and S.J. Beaupre. 2000. An application of randomization for detecting thermoregulation in Timber Rattlesnakes (Crotalus horridus) from northwest Arkansas. Physiological and Biochemical Zoology 73(3): 325-334.

Woods, G.T. 1945. Rate of travel of the wood turtle. Copeia 1945: 49.

Woods, A.J., J.M. Omernik, and D. Brown. 1999. Level III and IV ecoregions of Delaware, Maryland, Pennsylvania, Virginia, and West Virginia. U.S. Environmental Protection Agency, National Health and Environmental Effects Research Laboratory, Corvallis, Oregon.

Yagi, K.T. and J.D. Litzgus. 2013. Thermoregulation of spotted turtles (Clemmys guttata) in a beaver-flooded bog in southern Ontario, Canada. Journal of Thermal Biology 38: 205–213.

Zheng, D. and J. Chen. 2000. Edge effects in fragmented landscapes: A generic model for delineating area of edge influences (D-AEI). Ecological Modeling 132: 175-190.

Zheng, D., J. Chen, B. Song, M. Xu, P. Sneed, and R. Jensen. 2000. Effects of silvicultural treatments on summer forest microclimate in southeastern Missouri Ozarks. Climate Research 15: 45–59.

Ziegler, S.S. 2004. Composition, structure, and disturbance history of old-growth and second-growth forests in Adirondack Park, New York. Physical Geography 12(2): 152-169.

384 Zimmerman, L.C. and C.R. Tracy. 1989. Interactions between the environment and ectothermy and herbivory in reptiles. Physiological Zoology 62(2): 374-409.

385 APPENDIX 1. FLORA AT STUDY AREA

Common overstory canopy tree species include Quercus (alba, cocinna, prinus, rubra, and velutina), Pinus (rigida, strobus, and virginiana), Acer rubrum and saccharum, Betula lenta, Carya glabra, Fraxinus americana, Liriodendron tulipifera, Nyssa sylvatica, and Prunus serotina. Common midstory tree taxa include smaller individuals of the above species as well as Amelanchier spp.,

Cornus florida, Hamamelis virginiana, and Ostrya virginiana.

Common shrub and woody understory taxa include seedlings of the above taxa as well as Gaylusuchia spp., Ilex verticilata, Kalmia latifolia, Lindera benzoin,

Lyonia spp., Parthenocissus quinquefolia, Rhododendron periclymenoides, Rubus spp., Smilax spp., Vaccinium spp., and Viburnum spp.

Common herbaceous ground floor taxa include Ageratina altissima,

Amphicarpaea bracteata, Aster spp., Boehmeria cylindrica, Chimaphila maculata,

Cunila origanoides, Desmodium spp., Dioscorea villosa, Epigaea repens, Eurybia divaricata, Gallium spp., Gaultheria procumbens, Goodyera pubescens, Hieracium venosum, Impatiens capensis, Lobelia spp., Lycopus spp., Medeola virginiana,

Mitchella repens, Nabalus spp., Oxalis stricta, Pedicularis canadensis, Potentilla spp., Scutellaria spp., Smilacena recemosa, Solidago spp., Thalictrum spp., Uvularia spp., Viola spp., Carex spp. and Panicum spp.

Forest types (FT) of stands in Virginia: FT3 = White Pine, 10 = White

Pine/Upland Hardwoods, 39 = Table Mountain Pine, 41 = Cove Hardwoods/White

Pine, 52 = Chestnut Oak, 53 = White Oak – Northern Red Oak – Hickory, 54 =

White Oak, 56 = Tulip Poplar – White Oak – Northern Red Oak, 59 = Scarlet Oak,

60 = Chestnut Oak – Scarlet Oak.

The WV site has a greater proportion of relatively more-xeric pine and mixed pine-deciduous forest types: FT10 = White Pine/Upland Hardwoods, 33 =

Virginia Pine, 42 = Upland Hardwoods/White Pine, 45 = Chestnut Oak – Scarlet

Oak – Yellow Pine, 52 = Chestnut Oak, 53 = White Oak – Northern Red Oak –

Hickory (stand designations, nomenclature, and enumeration as per USFS).

386 APPENDIX 2. ANNUAL CAPTURE HISTORY (2006-2014) FOR ADULT WOOD TURTLES USED IN MARK ANALYSES

Years M F 110001000 1 0 1; 110101011 1 0 1; 111000000 1 0 1; 100000000 1 0 1; 101000010 1 0 1; 101000111 1 0 1; 100010001 0 1 1; 010000000 0 1 1; 010100000 1 0 1; 010000000 0 1 1; 010000000 1 0 1; 010000010 0 1 1; 010011111 0 1 1; 010000000 1 0 1; 010001100 0 1 1; 011111100 1 0 1; 010000000 1 0 1; 010000000 0 1 1; 001000001 0 1 1; 000100000 0 1 1; 000100000 1 0 1; 000100000 1 0 1; 000101011 1 0 1; 000100000 0 1 1; 000100000 0 1 1; 000111111 1 0 1; 000010000 1 0 1; 000001001 1 0 1; 000101110 1 0 1; 000001001 0 1 1; 000001111 1 0 1; 000001000 0 1 1; 000001011 0 1 1; 000001101 0 1 1; 000001100 0 1 1; 000001100 0 1 1; 000000100 0 1 1; 000000101 0 1 1; 000000110 0 1 1; 000000111 0 1 1; 000000111 1 0 1; 000000100 0 1 1; 000000100 0 1 1; 000000101 0 1 1; 000000011 0 1 1; 000000010 0 1 1; 000100010 0 1 1; 000000011 0 1 1; 000000011 0 1 1; 000000010 0 1 1; 000000100 1 0 1; 000000001 0 1 1; 000000001 1 0 1; 000000001 1 0 1; 000000001 1 0 1; 000000001 1 0 1; 000000011 0 1 1; 000001000 1 0 1; 000000100 1 0 1;

387 APPENDIX 3. DETAILS OF MICROHABITAT TEMPERATURE PATTERNS iButton temperatures:

Synchronous broad temperature differentials were obtained for different microhabitats at the same array site. For example, at a WV site on July 16, 2013 at

12:31, the open iB recorded a temperature of 48.6°C, while at the same time the iB under litter ca. 1.5m away recorded a temperature of 21°C, a difference of over

27°C. Diel temperature differentials of over 60°C at a single microhabitat were recorded. For instance, a WV “open” iButton recorded a temperature of 18°C at

8:01 on July 15, 2013 while the same button recorded a temperature of 79.5°C at

12:31 on the same date.

The temperature range across all microhabitats during a single diurnal period was usually greater than 15 degrees, often more than 20 (see, e.g., Fig. 3.7). The diurnal temperature range under litter was usually less than 5 degrees, often less than 3 (Figs. 3.5 & 3.7). The UL sites were cooler diurnally and warmer nocturnally than open locations (Tables 3.2 & 3.4, Figs. 3.5 & 3.8). Except for that single day in

VA, temperatures in UL microhabitats never closely approached CTmax (with a maximum of only 26.5° in WV). In WV microhabitats, UL had the greatest proportion of diurnal temperatures <17°C, while UV microhabitats had the lowest proportion (Table 3.3). In VA the opposite pattern occurred, UL had the lowest proportion of diurnal temperatures <17°C, while the UV microhabitats had the highest.

In the open microhabitats, the highest temperatures (>27°C) typically

388 occurred during the 11:00-16:00 time period, with daily maximums generally ca.

13:00 in both states (Figs. 3.5 & 3.7). The great majority of the lowest diurnal temperatures (<16°C) occurred from 8:00-10:00 (97.3% in WV, 82.7% in VA).

CTmax:

With one exception, temperatures ≥ CTmax never occurred diurnally in UL

microhabitats or nocturnally in any microhabitat. In VA, temperatures ≥ CTmax occurred in open microhabitats during 10:30-15:30, while they occurred in the UV

microhabitats only during 12:00-12:30. In WV, temperatures ≥ CTmax occurred in open microhabitats during 10:30-16:00, while in UV microhabitats they occurred only during 10:00-12:30 or 14:00-15:30. It is important to note that even when

temperatures ≥ CTmax were recorded in an open or UV microhabitat, temperatures

well below the CTmax were available only a meter or two away in the UL or UV microhabitats (Fig. 3.7).

In both states most of the temperatures ≥ CTmax were recorded in the open microhabitats (unshaded by ground vegetation). This circumstance only occurred for 2.4% of the total diurnal time period open habitat readings in WV, an average

of ca. 18 minutes per day; these temperatures ≥ CTmax occurred on 27 calendar days. In VA it occurred during 3.0% of the total diurnal time period open habitat

readings, an average of ca. 22 minutes per day; temperatures ≥ CTmax occurred on

22 calendar days. In WV ca. 0.75% of the under veg temperatures were ≥ CTmax (58

of 7,684 temperatures), an average of ca. 5.4 minutes per day; temperatures ≥ CTmax occurred on 25 calendar days. All these temperatures were from two buttons (at

389 array sites SRB2 and SRC3) under Blueberries (Vaccinium spp.) and Bush St. John

Worts (Hypericum prolificum) with scanty leaf cover on steep SW aspect slopes and high canopy openness (Table 3.1). In VA only ca. 0.12% of the under veg

temperatures were ≥ CTmax (8 of 6,781 temperatures). All eight of these were from one button (at PRC4) under Black Locust seedlings/saplings (Robinia pseudo- accacia) on a SE aspect roadside with high canopy openness (Table 3.1).

In WV, temperatures ≥ CTmax occurred in open microhabitats only during

10:30-16:00, but temperatures ≥ CTmax only occurred an average of 18 minutes per day. In the UV microhabitats it was even less (under 6 minutes per day on average)

and temperatures ≥ CTmax occurred there only during 10:00-12:30 or 14:00-15:30.

Temperatures ≥ CTmax never occurred under litter in WV. In WV during this time period (10:30-16:00), the mean temperature for these open iButtons was 26.5°C, while the mean of the maximum temperatures then was 34.1°C. As in VA, these maxima pertain to any day during the duration of this study.

Thermal patterns and environmental conditions:

VA: In VA, 83% (5 of 6) of the open iButtons recorded temperatures ≥

CTmax on at least one day, while 12.5% (1 of 8) of the under veg and 16.7% (1 of 6) of the under litter buttons did. For all the open iButtons, temperatures reached ≥

CTmax on 54 days out of a total of 210 recording days (the product of number of iButtons X number of days of readings per button) (25.7% of the days); on those 54 days this condition occurred for a mean of 86 minutes/day/button at the five buttons. Two buttons recorded 72% of the recording days (39 of 54) that had

390 temperatures ≥ CTmax, one at a shrubby anthropogenic “wildlife opening” (array site

PRA2) where temperatures were ≥ CTmax on 21 days out of 35 (60%) and the other

at a roadside (PRC4) where temperatures were ≥ CTmax on 18 days out of 35 (51%).

Both had E aspects (somewhat neutral with regard to radiation and warmth) and a higher than average degree of canopy openness (Table 3.1). One open iButton, situated in a thickly regenerated 30 years old clearcut with a WNW aspect (PRC3),

never recorded temperatures ≥ CTmax.

The only under veg and under litter iButtons that recorded temperatures ≥

CTmax were both at the roadside (PRC4). For all the UV iButtons, temperatures were

≥ CTmax on 4 days out of a total of 271 recording days (1.5%); on those 4 days this condition occurred for a mean of 60 minutes/day at the single button. For all the

UL iButtons, temperatures were ≥ CTmax on only 1 day out of a total of 210 recording days (0.5%); this condition occurred for ca. 30 minutes on that day at the roadside UL button.

For the VA open microhabitats, the highest average daily minimum and highest average daily mean temperatures were recorded by the roadside iButton

(PRC4), while the highest average daily maximum temperatures and highest standard deviation (SD) were recorded at the shrubby opening (PRA2) (Table 3.5).

The lowest average daily minimum, lowest average daily maximum, lowest average daily mean temperatures, and lowest SD for VA open microhabitats were recorded at PRC3, the thickly regenerated 30 years old clearcut.

391 For the VA under veg microhabitats the highest average daily minimum, highest average daily maximum and highest average daily mean temperatures, and highest standard deviation (SD) were recorded by the roadside iButton (PRC4). The lowest average daily maximum and average daily mean temperatures, and lowest

SD for VA UV microhabitats were recorded at an E aspect mature mixed oak forest site (PRB1).

For the VA under litter microhabitats the highest average minimum, highest average maximum and highest average mean temperatures, and highest SDs, all computed on both daily and hourly bases, were recorded at the roadside location

(PRC4). The lowest average daily maximum and average daily mean temperatures, and lowest SD for VA UL microhabitats were recorded at the E aspect mature forest site (PRB1).

WV: In WV, 88% (7 of 8) of the open iButtons recorded temperatures ≥

CTmax on at least one day during the study period, while only 22% (2 of 9) of the under veg and 0% (0 of 6) of the under litter buttons did. For all the open iButtons,

temperatures reached ≥ CTmax on 78 days out of a total of 272 recording days

(28.7% of the days); on those 78 days this condition occurred for a mean of 72 minutes/day/button at the seven buttons. Three buttons recorded 68% of the

recording days (53 of 78) that had temperatures ≥ CTmax. One was at a canopy gap at a mature mixed pine-hardwoods site with a NE aspect (but very little slope inclination) and slightly higher than average degree of canopy openness (array site

SRA1) where temperatures were ≥ CTmax on 26 days out of 35 (74%), another in

392 young mature forest with a W aspect and below average degree of canopy

openness (SRB4) where temperatures were ≥ CTmax on 16 days out of 31 (52%), and the third in mature forest with a NW aspect and below average degree of canopy

openness (SRA2) where temperatures were ≥ CTmax on 11 days out of 35 (31%).

One open iButton, situated in in mature mixed oak forest with a NNW aspect and below average degree of canopy openness (SRA3), recorded no temperatures ≥

CTmax.

For all the under veg iButtons, temperatures were ≥ CTmax on 30 days out of a total of 307 recording days (9.8%); on those days this condition occurred for a mean of 60 minutes/day/button at the two buttons. Most of these temperatures (24 of 30 recording days – 80%) were recorded by a button in mature forest with a SW aspect, steep slope, and high degree of canopy openness (SRB2); temperatures ≥

CTmax occurred there on 24 days out of 35 (69%). The only other UV iButton that

recorded temperatures ≥ CTmax was in old-growth forest that also had a SW aspect,

steep slope, and high degree of canopy openness (SRC3); temperatures ≥ CTmax occurred here on 6 days out of 35 (17%). No under litter iButtons recorded

temperatures ≥ CTmax (zero days out of 202 recording days).

For the WV open microhabitats, the highest average daily maximum and mean temperatures, lowest average daily minimum temperatures, and highest standard deviation were recorded by the iButton (SRA1) with the highest number of

temperatures ≥ CTmax (Table 3.5). The lowest average daily maximum and mean

393 temperatures, and lowest SD for WV open microhabitats were recorded at SRA3,

where no temperatures ≥ CTmax were recorded.

For the WV under veg microhabitats the highest average daily maximum and mean temperatures, and highest standard deviation (SD) were recorded by the

iButton (SRB2) that had the most under veg temperatures ≥ CTmax. The lowest average daily maximum and mean temperatures, and lowest SD for WV UV microhabitats were recorded at an E aspect mature mixed pine-hardwoods site with below average degree of canopy openness (SRB1).

For the WV under litter microhabitats, the highest average maximum and mean temperatures, and highest SDs, computed on both daily and hourly bases, were recorded at an old growth Virginia Pine (Pinus virginiana) location (SRB3) with a N aspect and the highest degree of canopy openness of any array site. The lowest average minimum, maximum and mean temperatures, and lowest SDs, computed on both daily and hourly bases, for WV UL microhabitats were recorded at a NW aspect mature mixed oak site with a below average degree of canopy openness (SRA4).

394 APPENDIX 4. WATER BALANCE CONDITIONAL LOGISTIC REGRESSION

MODELS

Models used in conditional logistic regression analyses of environmental attributes related to moisture conditions at turtle plots and random plots in Virginia and West Virginia during June-August 2011-2012. AET = actual evapotranspiration (mm) estimated with Dyer water balance model using top 100cm of soil depth for available water capacity, DEF25 = water deficit (mm) estimated with Dyer water balance model using only top 25cm of soil depth for available water capacity, Canopy = amount of canopy openness (% estimated with a spherical densiometer), Gapsize = amount of ground area (m2) in the 400m2 plot under canopy gaps (visual estimate), HumGr = ground level relative humidity in the shade, TempGr = ground level temperature in the shade. The same model formulations listed below were also run with DEF25 in place of AET.

Models model variables ______Virginia June - only those with * would converge mod1 AET+TempGr+Gapsize mod2 AET+TempGr+Canopy mod3 AET+TempGr mod4 AET+HumGr * mod5 AET+Gapsize * mod6 AET+Canopy mod7 TempGr+Canopy mod8 TempGr+Gapsize mod9 TempGr+HumGr mod10 HumGr * mod11 Canopy * mod12 AET mod13 TempGr * mod14 Gapsize July and August mod1 AET+HumGr+Canopy mod2 AET+HumGr+Gapsize mod3 AET+TempGr+Gapsize mod4 AET+TempGr+Canopy mod5 AET+TempGr mod6 AET+HumGr mod7 AET+Gapsize mod8 AET+Canopy mod9 HumGr+Canopy mod10 TempGr+Canopy mod11 TempGr+Gapsize mod12 HumGr+Gapsize mod13 HumGr mod14 Canopy mod15 AET mod16 TempGr mod17 Gapsize mod18 AET+HumGr+TempGr+Canopy ### this model only run in August mod19 HumGr+TempGr ### this model only run in August

West Virginia June AET - only those with * would converge mod1 AET+TempGr+Gapsize mod2 AET+TempGr+Canopy mod3 AET+TempGr mod4 AET+HumGr mod5 AET+Gapsize mod6 AET+Canopy * mod7 TempGr+Canopy * mod8 TempGr+Gapsize mod9 TempGr+HumGr

395 mod10 HumGr * mod11 Canopy mod12 AET Appendix 4 table: continued * mod13 TempGr * mod14 Gapsize

June DEF25 - only those with * would converge mod1 DEF25+TempGr+Gapsize mod2 DEF25+TempGr+Canopy mod3 DEF25+TempGr mod4 DEF25+HumGr mod5 DEF25+Gapsize * mod6 DEF25+Canopy * mod7 TempGr+Canopy * mod8 TempGr+Gapsize mod9 TempGr+HumGr mod10 HumGr * mod11 Canopy * mod12 DEF25 * mod13 TempGr * mod14 Gapsize July and August mod1 AET+HumGr+Canopy ### would not run in July with DEF25 mod2 AET+HumGr+Gapsize mod3 AET+TempGr+Gapsize mod4 AET+TempGr+Canopy mod5 AET+TempGr mod6 AET+Gapsize mod7 AET+Canopy mod8 AET+HumGr mod9 HumGr+Canopy mod10 TempGr+Canopy mod11 TempGr+Gapsize mod12 HumGr+Gapsize mod13 HumGr mod14 AET mod15 Canopy mod16 Gapsize mod17 TempGr

396 APPENDIX 5. IMPORTANCE VALUE CONDITIONAL LOGISTIC REGRESSION

MODELS

Models used in conditional logistic regression analyses of importance values of overstory and midstory tree taxa present in 400m2 plots at turtle points and random points in Virginia and West Virginia during June-August 2011-2014. Taxa used in models: BG (Black Gum) = Nyassa sylvatica, BL (Black Locust) = Robinia pseudo- acacia, BLBIR (Black Birch) = Betula lenta, BLCH (Black Cherry) = Prunus serotina, BW (Basswood) = Tilia americana, DW (Dogwood) = Cornus florida, ELM (Elms) = Ulmus spp., HICK (Hickories)= Carya spp., IW (Ironwood) = Ostraya virginiana, SERV (Serviceberry) = Amelanchier spp., SYC (Sycamore) = Platanus occidentalis, TP (Tulip Poplar)= Liriodendron tulipfera, WA (White Ash) = Fraxinus americana; Maples: RM (Red Maple) = Acer rubra, SM (Sugar Maple) = Acer saccharum; Oaks: BO (Black Oak) = Quercus velutina, CO (Chestnut Oak) = Quercus montana, NRO (Northern Red Oak) = Quercus rubra, SO (Scarlet Oak) = Quercus coccinea, WO (White Oak) = Quercus alba; Pines: PP (Pitch Pine) = Pinus rigida, VP (Virginia Pine) = Pinus virginiana, WP (White Pine) = Pinus strobus. Forest type groups: Dm = mesic deciduous, M = dry mixed pine and deciduous, Mm = mesic mixed pine and deciduous Od = oligotrophic oak, Om = mesic oak, P = pine.

Model type ______

Virginia females and Virginia males

Dm mod1 WO+NRO+SM+RM+BW+TP+WA+BG+HICK+BLBIR+ELM M mod2 WP+VP+WO+CO+RM+SO+BG Mm mod3 WP+VP+WA+TP+RM+BLBIR+BG Od mod4 CO+SO+BO+BG+HICK+WO Om mod5 WO+CO+NRO+RM+SM+HICK+BG+TP+WA P mod6 WP+VP+RM

Global mod7 WP+CO+SM+RM+WO+BG+WA+TP+SO+NRO+HICK+VP

Oaks mod8 WO+CO+NRO+SO+BO Maples mod9 RM+SM Pines mod10 WP+VP

Mesic spp. mod11 BW+TP+WA+BLBIR+ELM Deciduous spp. mod12 HICK+BG+BL+SERV+IW

Mixed mod13 HICK+BG+BL+SERV+IW+WP+VP Mixed mod14 HICK+BG+BL+SERV+IW+RM+SM Mixed mod15 HICK+BG+BL+SERV+IW+WO+CO+NRO+SO+BO

397

Appendix 5 table: continued West Virginia females (removed WA, TP, ELM, BL, IW; added BLCH, SYC, DW)

Dm mod1 WO+NRO+SM+RM+BW+BG+HICK+BLCH M mod2 WP+VP+WO+CO+RM+SO+BG+HICK Mm mod3 WP+VP+SM+RM+BLCH+BW Od mod4 CO+SO+BG+HICK+WO Om mod5 WO+NRO+RM+SM+HICK+BG+C0 P mod6 WP+VP+PP+BG

Global model mod7 WP+VP+WO+CO+SM+RM+BLCH+SYC+NRO+HICK [removed DW]

Oaks mod8 WO+CO+NRO+SO Maples mod9 RM+SM Pines mod10 WP+VP+PP

Mesic spp. mod11 BW+BLCH+SM+RM+SYC Deciduous spp. mod12 HICK+BG+DW+SERV

Mixed mod13 HICK+BG+DW+SERV+WP+VP+PP Mixed mod14 HICK+BG+DW+SERV+RM+SM Mixed mod15 HICK+BG+DW+SERV+WO+CO+NRO+SO

West Virginia males (removed IW, SO, SYC, TP; added BLCH, DW)

Dm mod1 WO+NRO+SM+RM+BW+BG+HICK+BLCH+ELM M mod2 WP+VP+WO+CO+RM+BG+WA+HICK Mm mod3 WP+VP+SM+RM+BLCH+BW+ELM Od mod4 CO+SO+BG+HICK+WO Om mod5 WO+NRO+RM+SM+HICK+BG+C0 P mod6 WP+VP+PP+BG

Global model mod7 WP+VP+WO+CO+SM+RM+BLCH+ELM+NRO+HICK

Oaks mod8 WO+CO+NRO Maples mod9 RM+SM Pines mod10 WP+VP+PP

Mesic spp. mod11 BW+WA+BLCH+SM+RM+ELM Deciduous spp. mod12 HICK+BG+DW+SERV

Mixed mod13 HICK+BG+DW+SERV+WP+VP+PP Mixed mod14 HICK+BG+DW+SERV+RM+SM Mixed mod15 HICK+BG+DW+SERV+WO+CO+NRO

398 APPENDIX 6: LITERATURE REVIEW AND NATURAL HISTORY

OF THE WOOD TURTLE (GLYPTEMYS INSCULPTA)

I. Description of Species

A. Taxonomy

The North American Wood Turtle (Glyptemys insculpta) is a member of the class

Reptilia, order Testudines, family Emydidae, and subfamily Emydinae (TTWG 2007).

Emydids include Box Turtles (Terrapene), Blanding’s Turtle (Emydoidea), Spotted

Turtle (Clemmys), Western Pond Turtle (Actinemys), Painted Turtles (Chrysemys),

Chicken Turtles (Dierochelys), Map Turtles (Graptemys), Diamond-backed Terrapins

(Malaclemys), Cooters (Pseudemys), and Sliders (Trachemys). Emydids are ecologically diverse and include generalists and specialists in both habitat (including terrestrial, semiterrestrial, and aquatic species) and diet (including herbivores, omnivores, and carnivores) (Stephens and Wiens 2004). The type specimens of the

Wood Turtle were collected around New York City (Mitchell 1994). It was first scientifically named and described by Le Conte in 1830.

Until recently the Wood Turtle was placed in the genus Clemmys along with the Bog Turtle (muhlenbergii), Spotted Turtle (guttata), and Western Pond Turtle

(marmorata) (Ernst et al. 1994). However, based on DNA analysis this is no longer considered a monophyletic group (TTWG 2007). While the phylogenic disposition of some of these species is still in dispute, there is general agreement that insculpta and muhlenbergii are sister-species that should be placed in the genus Glyptemys

399 (Holman and Fritz 2001, Feldman and Parham 2002, Parham and Feldman 2002,

Stephens and Wiens 2003, TTWG 2007, Thomson and Shaffer 2010).

Fossil remains of Wood Turtles have been dated to ca. 6 million years old

(Harding 2002).

B. Physical Description

Adult Wood Turtles have a keeled knobby carapace (top shell) up to around nine inches (238 mm) long that is generally brownish in color with short darker and lighter lines radiating from the centered growth whorl of each vertebral and pleural scute.

Although overall the carapace is dully earth-toned, the richly colored striae of black, mahogany, ochre, fawn, and even pale gold all serve to make the Turtle indistinguishable from the interplay of light, shadow, line, form, and pattern that characterizes both the terrestrial and aquatic realms it inhabits (Carroll 2009). The hingeless plastron (bottom shell) is buff colored with dark marginal blotches. The raised growth annuli on the scutes of adults’ carapaces often give the appearance of sculpted truncated pyramids. These annuli and ziggurats smooth out with age. The growth rings are highly reliable indicators of age up to 15-20 years (Harding and

Bloomer 1979, Lovich et al. 1990, and Akre 2002). This technique generally does not adequately estimate adult ages of long-lived species (Galbraith and Brooks 1989).

In older Turtles, growth, and thus the formation of annuli, may essentially cease; nonetheless, counting scute annuli will usually provide a reliable minimum age

(Harding 2002). The carapace posterior marginal scutes are serrated, particularly in young individuals. The head is black and the snout is blunt with a notch on the upper

400 jaw. Their chins, necks, and limbs have various shades, intensities and amounts of red, orange, or yellow coloration. Their large front legs manifest their amphibious lifestyle with the forelimb musculature being morphologically intermediate between that exhibited by strictly terrestrial or aquatic species (Abdala et al. 2008). The height silhouette of their carapace is also intermediate between the high dome displayed by the terrestrial Box Turtle (Terrapene carolina) and the more flattened appearance of aquatic species such as the Painted Turtle (Chrysemys picta) (see shell silhouettes in

Buhlmann et al. 2008).

They exhibit low annual egg production (averaging six to ten) and generally do not reach sexual maturity in the wild until 10-18 years of age (Akre 2002 and

Akre & Ernst 2006 in Virginia, Brooks et al. 1992 in Ontario). Sexual maturity seems to be related more to body size rather than to a specific age. Turtles become sexually mature at a carapace (dorsal or top shell) length of around 150-199 mm (Akre 2002,

Lovich et al. 1990, and Brooks et al. 1992). Wood Turtles in northern populations generally are older and larger at maturity than those in southern populations (Akre

2002, Walde et al. 2003). Reproductive output in females does not decrease with old age (Jones 2009).

Wood Turtles may live to be over 50 years old (Ernst 2001a, Ernst and Lovich

2009, COSEWIC 2007). “[B]ased upon an assessment of shell-wear from digital photographs,” Jones (2009) concluded that “wood turtles regularly achieve ages over

80 years”. Barker (1964) reported an individual who survived for 100 years (1839-

1939) at the London Zoological Garden. It may be that Turtles need to live for these

401 long periods in order to replace themselves in the population (Gibbons 1987, Iverson

1992, Seigel 2005). This is consistent with a ‘‘bet-hedging’’ life history strategy whereby organisms that evolve in environments where egg and hatchling mortality are unpredictable, and where adult mortality is relatively predictable, evolve extreme iteroparity to maximize the probability of successful reproduction in ‘‘good years’’

(Litzgus et al. 2008). The average age of adults (generation time, GT) has not been calculated in the literature, but an estimation based on published values for age at maturity (AM) and adult rates of mortality (MR) was found to be 35 years (COSEWIC

2007).

Maximum size varies geographically, with the largest animals (up to 238-242 mm straight- line carapace length) occurring in the northern part of the range

(Saumure 1992, Walde et al. 2003, Greaves & Litzgus 2009). There is support for the hypothesis that Turtle size is negatively correlated with the number of frost-free days (Walde et al. 2003). However, large individuals (c 215-220 mm) also occur at the very southern extremity of the range as well (Akre 2002, Greaves and Litzgus

2009, Krichbaum pers. obs.). Males may mature later and at larger sizes than females

(Lovich et al. 1990) or slightly earlier and at marginally smaller lengths than females

(Akre 2002). Large animals may approach 1.5 kg (3.3 lb) in mass (Akre 2002, Arvisais et al. 2002, Walde et al. 2003, Castellano et al. 2008, Greaves & Litzgus 2009). Color varies geographically and with age. In general, animals in the western part of the range are lighter in overall coloration. The color of very old individuals often appears duller and darker (Krichbaum pers. obs.).

402 The low carapaces of hatchlings are 28-40 mm (1.1-1.5 in) in length and 28-

36 mm in width (Farrell and Graham 1991, Tuttle and Carroll 2005, Mitchell 1994,

Akre and Ernst 2006). They weigh from 4.6-10 g (0.2-0.4 oz) (Farrell and Graham

1991, Akre and Ernst 2006). Hatchlings are generally light brown-olive with tails as long as their carapace (Krichbaum pers. obs.).

Wood Turtles are sexually dimorphic (Lovich et al. 1990, Stephens and Wiens

2009). Males reach a larger maximum size than females (Akre 2002, Lovich et al.

1990). In general, females have smaller narrower heads, relatively rounder carapaces, and shorter thinner tails (Akre 2002, Greaves & Litzgus 2009). Mature males have a concave plastron. The carapace of males is also generally more domed, but proportionately narrower than females’ (Akre 2002, Harding 2002). The male’s cloaca is located past the carapace edge whereas the female’s is further in toward the body (id.). Mature males often have larger imbricate scales on the anterior part of the forelimbs than do mature females (Mitchell 1994). The red/orange/yellow color of the limbs, chin, and neck may be paler in females than males, especially during the mating season (id.). However, females can be as or even more brightly colored than males; there are pale, vivid, and dark individuals of both sexes (Krichbaum pers. obs.). The coloration may have an epigamic function (serving to attract a mate). In contrast, procrypsis is manifest in the Turtle’s dark carapace and skin tones, nodose and striated carapace that break up its line and form, and ability to freeze (i.e., go immediately quiescent).

403 Turtle size dimorphism has been interpreted in terms of sexual selection theory: Males are larger than females when large male size evolves as an adaptation to increase success in male combat, or to enable forcible insemination of females

(Berry and Shine 1980).

See Ernst and Lovich (2009) for a more detailed description of physical characteristics.

II. Present Distribution and Status of the Wood Turtle (Glyptemys insculpta)

A. Range / Distribution

One of the more northern ranging of North American reptile species, the Wood Turtle is sporadically distributed in southeastern Canada and 17 northern states of the USA

(see range map in Ernst and Lovich 2009). Now Virginia’s Rockingham County and

West Virginia’s Pendleton County are the southernmost extent of its known global range. Most of the Wood Turtle’s current range was glaciated during the Pleistocene epoch (see Amato et al. 2007, and map in Church et al. 2003). Fossil evidence shows they (relatively) recently ranged as far south as central Tennessee and northwestern

Georgia (Parmalee and Klippel 1981), then proceeded back northward following the retreat of the Pleistocene ice-sheets (Amato et al. 2007).

The Wood Turtle’s range is fragmented and populations within it are generally geographically isolated (Ernst 2001b). The distribution of the Wood Turtle can be separated into two main areas: the Northeast and the Great Lakes. The Northeast region extends from Nova Scotia and New Brunswick, through Maine, New York, and Pennsylvania, south to northern Virginia & West Virginia. As this is the most

404 heavily populated and highly developed region in the country, the Turtle survives here in disjunct populations.

Populations in the Great Lakes region are also discontinuous, with populations in the northern Lower Peninsula of Michigan and from the central Upper

Peninsula of Michigan southwest to southern Wisconsin, and into eastern Minnesota.

Generally, these Wood Turtle populations are scattered in isolated habitat fragments, and the species is considered rare. There are also disjunct populations in northeastern Iowa, south central Quebec and southeastern Ontario. Approximately

30% of the species’ overall range is in Canada (COSEWIC 2007).

Use of range maps can significantly overestimate the actual occurrences and distribution of a species (Jetz et al. 2008). Range maps are misleading as to the actual extent of the Wood Turtle’s distribution (e.g., see map in Ernst and Lovich 2009).

Although the Turtle is wide-spread in “range”, it actually has a restricted distribution

(see habitat constraints at section III.). Occurrence of actual populations is recognized to be localized and spotty (Jones and Willey 2015). In other words, the actual “area of occupancy” is only a tiny fraction of the “extent of occurrence” based on range maps. For example, the Turtle’s occupied area was calculated to be only ca. 0.3% of the extent of its range in Canada (COSEWIC 2007).

Examples of the species’ current disjunct and sporadic distribution include

Wisconsin (“Wood turtles were once found throughout the state, except in the southwestern-most portion. Today, small scattered populations exist in isolated

405 habitat.” Wisconsin DNR 2005 Wildlife Plan), West Virginia (“[T]he occurrence of insculpta in West Virginia, even within favorable habitats such as this site, appears to be localized (Neiderberger, 1993).” Neiderberger and Seidel 1999), and Virginia

(unpublished survey work by Akre and Krichbaum).

“Research indicates that even in rivers with apparently good habitat throughout their length, the Turtles are usually patchily distributed with most of the river unoccupied except by transients (Brooks, pers. comm. 2005; Wesley, 2006).

For example, on a major river in central Ontario, virtually all sightings of Wood

Turtles along a 20-km stretch of the river occurred at 2 sites, one 1.2 km and the other 0.4 km. in length. Extrapolation from these sites over the 20-km surveyed would have given an estimate over 2000 adults, when the real population of adults is likely fewer than 150 (Brooks, pers. comm. 2005). As the Turtles are not found along entire lengths of watercourses, but occur in discontinuous patches, estimates generalized to an entire river or watershed will grossly over-estimate metapopulation size”

(COSEWIC 2007).

For habitat to be occupied it must supply the basic requirements of food, cover, thermoregulation, hydration, hibernacula, mating opportunities, and nesting sites. As for the discontinuities in the Turtle’s current distribution: “These absences could be explained by poor dispersal capabilities or key habitat requirements that are missing (Wesley 2006), or by random extinctions characteristic of small, isolated populations where isolation is accentuated in some areas by anthropogenic activity.”

(COSEWIC 2007)

406 Use of overall “range” maps also misrepresents the potential for the species’ recovery. Such a “method overestimates recovery potential for most species because they are often restricted to particular habitats that may only extend across a small proportion of the actual natural habitat within the total range map for the species.”

(Kerr and Deguise 2004)

Many factors contribute to the distribution of Wood Turtle populations within and across watersheds (Compton et al. 2002, Jones 2009, Jones and Willey 2015). At the landscape scale, Wood Turtle occurrences and abundance are positively correlated with amount of forest cover, and negatively correlated with amounts of impervious surface, urban development, and agriculture (Jones and Willey 2015). Among the most important aquatic factors are stream gradient, discharge rate, and sinuosity, as well as substrate composition and degree of suspended solids (id.) Clear, low gradient

2nd to 5th order streams with a moderate, continuous current and hard sand or gravel bottoms appear to be preferred (Ernst 2001b). Thus, compared to species such as

Painted or Snapping Turtles, Wood Turtles can be viewed as "stream specialists", occurring along those stream reaches offering suitable hibernacula and terrestrial habitats (Wesley 2007).

Additionally, because females, at least in some areas, prefer sandy soil for nesting (see, e.g., Hughes et al. 2009), the occurrence and distribution of populations may also be limited by geologic factors. Historically, in the Great Lakes Region,

Wood Turtles may have been concentrated in areas where rivers or streams flowed

407 through glacial outwash plains. These geologic features are distributed in a heterogeneous fashion there, and so are Wood Turtle populations (Buech et al.

1997). The need for clear, moderate current streams and sandy soils, combined with the linear nature of Wood Turtle habitat may partially explain why Wood Turtles appear to occur in disjunct, isolated populations (Buech et al. 1997, Ernst 2001b).

The correlations of Wood Turtle distribution with soil, geological, and physiographic types, stream morphology, forest types and ages, and other habitat features (e.g., large woody debris loadings) need further investigation. This may prove to be of value in identifying and evaluating suitable habitat for the species, predicting potential sites of occurrence, and implementing effective conservation protocols.

The southern boundary of the Turtle’s current range was found to generally correspond with the 85 F° isotherm for normal daily maximum temperature in July

(Parmalee and Klippel 1981); however, the referenced isotherm was according to

1968 data from the U. Dept. of Commerce. Studies on nest success suggest that insufficient time and degree days for the completion of incubation, along with the inability of hatchling insculpta to over-winter in the nest, delimit the distribution of northern Wood Turtle populations (Walde et al. 2007).

At northern locations where the species occurs in Canada there may only be

90-100 frost-free days a year on average (Arvisais et al. 2002, Walde et al. 2003), whereas at some lowland sites in Virginia and Maryland there are 190-195 (Akre

408 2002). There may be a significant positive correlation between Wood Turtle density and number of frost-free days (Walde 1998, Smith 2002).

B. Status

Many freshwater turtle species in eastern North America are experiencing declining populations (2005 Comprehensive Wildlife Action Plans for the states in the

Turtle’s range). The recent comprehensive status report for the Wood Turtle in the northeast stated: “Abundant evidence strongly indicates that the wood turtle has undergone widespread population declines. The wood turtle occurs primarily in small, isolated, declining populations.” (Jones and Willey 2015)

In other words, their condition has not improved in the fifteen or so years since the following statement: “Of 13 states in North America that responded to a survey on the status of wood turtles, no states reported stable or increasing populations, 8 states reported declines, and 5 states reported unknown trends. In

New Hampshire, the wood turtle was much more abundant historically. The expert panel indicated that the current outcome for this species is D range-wide and declining in the future because many populations are comprised of many old turtles and very few young ones, so as old turtles die the populations may also die.

. . . Panelists agreed that for herps, viability outcome C is not viable; only A and B are viable for these species.” (NH White Mountains National Forest 2002)

NatureServe (2008) had listed it as a G4 species (“Global Status Last

Reviewed: 27Apr2005”). However, it was “apparently declining throughout the range and may warrant G3 rank, but survey data are scanty” (id.) In November 2010,

409 NatureServe changed the categorization of the Turtle to a status of G3 (a conservation status of “vulnerable”) (NatureServe 2012).

Similarly, the International Union for the Conservation of Nature (“IUCN”) had classified the Wood Turtle as a “vulnerable” species on its Red List (Hilton-

Taylor 2000; www.iucnredlist.org or http://www.iucn-tftsg.org/red-list/ -

Glyptemys insculpta — Emydidae — Wood Turtle — VU - A1abcd+2cd). More recently, however, the IUCN changed the species’ status from vulnerable to

“Endangered” to reflect range-contraction and increased threats and impacts from human encroachment (IUCN 2011).

The Convention on International Trade in Endangered Species of Wild

Fauna and Flora (“CITES”) regulates import and export of Wood Turtles internationally, with the species listed on Appendix II in June 1992 (Buhlmann

1993; also see http://www.cites.org/eng/app/index.shtml).

In addition, the Tortoise and Freshwater Turtle Specialist Group of the IUCN in 2007 compiled Regional Top Threatened Lists to provide rough hierarchical rankings of the currently most threatened species of tortoises and freshwater turtles for various geographic regions. The Wood Turtle is listed as one of the top ten

“Turtles in Trouble: North America’s Most Endangered Tortoises and Freshwater

Turtles – 2007” (http://www.iucn-tftsg.org/trouble/).

The Turtle is considered rare, declining, and/or vulnerable in almost every state and province throughout its range. All the state Wildlife Action Plans (see

410 http://wildlifeactionplans.org) in the Turtle’s range list it as a “species in greatest conservation need” (SGCN).

State rankings and listings (from NatureServe 2012) are:

Connecticut S3 Special Concern, Delaware SR, [not on state list 2012] District of Columbia SH, Iowa S1 Endangered, Maine S4 Special Concern, Maryland S4, Massachusetts S3 Special Concern, Michigan S2/S3 Special Concern, Minnesota S2 Threatened, New Hampshire S3 Special Concern, New Jersey S2 Threatened [S3 in 2008], New York S3 Special Concern, Ohio S1, Pennsylvania S3S4, Rhode Island S2 Special Interest, Vermont S3, Virginia S2 Threatened, West Virginia S2, Wisconsin S2 Threatened In Canada the rankings are: New Brunswick S3, Nova Scotia S3, Ontario S2

Rare, and Quebec S2. [2008 & 2012]

Ranking definitions are as follows: SH: possibly extirpated, SR: reported, S1: critically imperiled, S2: imperiled, S3: vulnerable, S4: apparently secure.

Listing by S-Ranks for the States/Provinces (from NatureServe 2012):

S5 – Secure - none S4 – Apparently secure - Maine, Maryland S3/S4 - Pennsylvania S3 – Vulnerable - Connecticut, Massachusetts, New Brunswick, New Hampshire, New York, Nova Scotia, Vermont

411 S2/S3 - Michigan S2 – Imperiled - Minnesota, New Jersey, Ontario, Quebec, Rhode Island, Virginia, West Virginia, Wisconsin S1 – Critically imperiled - Iowa, Ohio (viability status of extant populations in OH is uncertain) SH – Historic District of Columbia

The state wildlife agencies of both Virginia and West Virginia consider it a

"Priority Group 1" species in their state’s wildlife conservation strategy, meaning it is a “species of greatest conservation need” (VDGIF 2005). In Virginia it is officially listed as “Threatened” under the state’s Endangered Species legislation and is considered to be “declining”. Both Virginia and West Virginia consider the Turtle to be an S2 species, meaning “very rare and imperiled”. In WV it was protected from collection for commercial purposes, but could be collected for personal-use under the State fishing regulations (WVDNR 2005). This gigantic loophole allowed West

Virginia to be used to “launder” Wood Turtles illegally procured elsewhere.

The Wood Turtle is one of New Hampshire’s “species in greatest need of conservation” (see “SC = NH species of special concern (List revised 2000), RC =

Regional conservation concern” at page 2 – 2 of New Hampshire Wildlife Action

Plan 2005). It is illegal to possess, sell, or import Wood Turtles (RSA 212-A, New

Hampshire Fish and Game (NHFG) Rules Fis 800 at 4-61). The Turtle is also a

“species of greatest conservation need” in New York

(http://www.deny.gov/animals/9406.html), Maryland, Minnesota, Wisconsin,

412 Pennsylvania (of “immediate concern”, the highest priority), and Vermont (“high priority”, the most urgent category).

Within the Great Lakes region, the Wood Turtle is considered imperiled in every state in which it occurs. The Wood Turtle is state Endangered in Iowa, state

Threatened in Minnesota and Wisconsin, and a species of Special Concern in

Michigan.

The Turtle was considered “apparently secure” in only two (Maryland and

Maine) of the 22 states and provinces where the species is known to occur

(NatureServe 2008).

It is a “focal species for conservation” in both Maine and Nova Scotia (Beazely and Cardinal 2004). In the Northeast United States the Wood Turtle is a “focal species of regional conservation concern” (Therres 1999). This publication was produced by the Northeast Endangered Species and Wildlife Diversity Technical

Committee, a working committee of the Northeastern Association of Fish and

Wildlife Agencies. It is the result of a request from Region 5 of the U. Fish and

Wildlife Service to develop a process to identify animal species within the northeastern states that may warrant federal listing, and the need for a list of species of regional concern. This list played a key role in determining which species should be SGCN in various states (e.g., Maine and Vermont; see chapter 11 of ME wildlife plan). The Turtle is certainly a regional conservation priority (Jones and Willey 2015).

413 In Canada, the Wood Turtle exists in discontinuous, small, relatively isolated populations in Nova Scotia, New Brunswick, Québec, and Ontario (COSEWIC

2007).

“[I]n Québec the species has recently been added to the list of vulnerable species (Gouvernement du Québec 2005), a status upgrade that had been expected for more than a decade (Beaulieu 1992).” (Tessier et al. 2005) The Ontario Ministry of Natural Resources designates it as “Endangered” (Greaves and Litzgus 2007).

More recently, the Turtle’s status was re-examined and it was designated as

“Threatened” throughout Canada in November 2007 (COSEWIC 2007). “A crude estimate of total population size of the Wood Turtle in Canada, based on quantitative estimates from researchers across its Canadian range, is ~6,000-12,000 adults. Wood

Turtle populations that are in areas to which people have limited access may be stable, but where there is road access many populations are declining, and the overall trend in Wood Turtle abundance over the past three generations (~100+ years) is also one of decline.” (id.)

C. Population Estimates

Roughly tens of thousands of adults survive in disjunct populations and fragmented habitat [perhaps 40,000? – Krichbaum rough estimate from population size estimates in literature and number of populations known in Virginia compared to areal extent of range]. The population of Turtles in Canada was recently estimated to be 6,000 to 12,000 individuals (COSEWIC 2007). NatureServe (2012) estimated total population size as 10,000 - 100,000 individuals.

414 D. Decline from Historic Range

The Wood Turtle is currently threatened and vulnerable and the trend is downward

(state Wildlife Plans). The available evidence suggests that the species is declining range wide (Bowen and Gillingham 2004, Jones and Willey 2015).

For instance, the US Forest Service reported: “Of 13 states in North America that responded to a survey on the status of wood turtles, no states reported stable or increasing populations, 8 states reported declines, and 5 states reported unknown trends.” (NH WMNF 2002)

Part of the reason for this threatened and declining situation are actions that occurred in the past, an “extinction debt” from past harms, degradations and diminishments.

The species is already absent from a significant part of its historic range. Many population extirpations or declines and/or a general range contraction have been documented (Ernst and McBreen 1991, Farrell and Graham 1991, Harding 1991,

Klemens 1993, Garber and Burger 1995, Lovich 1995, Litzgus and Brooks 1996,

Oldfield 1996, Dahl 1996, Levell 2000, Burke et al. 2000, Harding 2002, Daigle and

Jutras 2005, Tessier et al. 2005, Akre and Ernst 2006, COSEWIC 2007; Saumure et al. 2007, Tingley et al. 2009, Jones 2009, Jones and Willey 2015).

Wood Turtles can still be locally common where suitable habitat still remains and the animals are relatively undisturbed by the effects of human encroachment

(Harding 1997). However, “the exceptionally high vulnerability of Wood Turtle populations to anthropogenic sources of mortality means that any population to

415 which humans have access (all current known populations, to some degree) are susceptible to decline.” (COSEWIC 2007)

The Turtle’s decline across the landscape is partially accountable for its now spotty distribution. Though distributional data from the colonial era simply does not exist, there is certainly reason to believe that in the past the Turtle’s aerography was far greater in extent and/or they inhabited a much greater number of locations. Prior to European colonization and subsequent population growth and development of the landscape, there was a great deal more habitat connectivity that allowed the Turtle much greater freedom of movement. Until very recently in the species evolutionary history, rivers, creeks, and brooks -- perennial, intermittent, and ephemeral -- existed as a channel continuum (Pringle et al. 1988) without significant human disruption.

At present, there is much apparently “suitable” Wood Turtle habitat that is not inhabited (Oldfield 1995, Akre and Ernst 2006, COSEWIC 2007, Krichbaum pers. obs.). For example, there are occurrence records for the Turtle at Virginia locations from the relatively recent past (i.e., 50 years ago in southern Rockingham County) where contemporary surveys have not found any Turtles (Akre and Ernst 2006,

Krichbaum, pers. obs., Akre pers. comm.).

In addition, Canadian authors recently referred to extensive fragmentation and diminishment of the Turtle’s range in southern Ontario (Tessier et al. 2005; Amato et al. 2007) So far no studied population has been judged to be increasing (COSEWIC

2007).

416 It is evident that the Turtle’s habitat has declined and is declining over much of the species’ historic range in both Canada and the United States (Jones and Willey

2015). For example, only a restricted number of creeks and rivers retain clear water, undisturbed nesting sites, deep pools for overwintering, and undisturbed riparian zones. This habitat loss and degradation is due to agricultural activities, development, channelization, dams, contamination, roads, and forestry activities. In addition, any increase in access (by humans and/or predators) to their populations constitutes a degradation of habitat even before direct habitat modification occurs.

Due to the extent and intensity of past, present, and potential future impacts, threats, and conditions, that the Turtle has persisted over the last 100 years is not a strong indication that the species will continue to persist into the foreseeable future despite the loss of historical habitat. The fact that Turtle populations persist in one area or class of areas does not mean that threats to the species elsewhere are not significant. Nor should the limitations or uncertainty in the available population studies be used to conclude that Turtle populations are in fact viable and stable throughout most of the species’ current range; i.e., the absence of evidence of population decline does not equate to evidence of persistence (Tuscon

Herpetological Society v. Salazar (566 3d 870,879 (9th Cir.2009)). For Wood Turtles, there is persuasive evidence of widespread decline and widespread threat to population persistence.

III. Ecology, Biology, and Behavior of the Wood Turtle A. Landscape – Macrohabitat

417 Lithologically, the Wood Turtle’s range generally is underlain by steeply folded or faulted sedimentary rocks (sandstones and shales in the Appalachians of Virginia,

West Virginia, Maryland, and Pennsylvania) and by slightly to moderately tilted older sedimentary rocks (in the Lake States of Michigan, Wisconsin, Pennsylvania, and

New York), as well as metamorphic and intrusive igneous rocks (mostly in the

Northeast – New York, Massachusetts, Connecticutt, Vermont, New Hampshire, and

Maine) (Raisz in Veregin 2005). Landforms here are made up of folded mountains, hills, and moraines (id.). Most (ca. 75%) of the species’ current range was glaciated

(id.). In the main, soil orders in the Turtle’s range are incepticols (suborder ochrepts), ultisols (udults), and spodosols (orthods), with some alfisols (as per USDA soil taxonomy in Brady and Weil 1996).

The dominant “natural vegetation” of the broad region occupied by the Wood

Turtle is characterized as broadleaf deciduous trees, mixed broadleaf deciduous and needleleaf evergreen trees, and needleleaf evergreen trees (Kuchler in Veregin 2005).

Specifically, the broadleaf deciduous tree types are beech-maple, beech-tulip tree- maple-basswood, and oak-tulip tree; the mixed broadleaf deciduous and needleleaf evergreen tree types are oak-pine, maple-beech-hemlock, and maple-yellow birch- hemlock-pine; and the needleleaf evergreen tree types are spruce-fir and pine (id.).

The average annual precipitation in the Turtle’s range varies from 20-30 inches to 40-50 inches. In general, this precipitation is somewhat evenly split between the periods of November 1-April 30 and May 1-October 31. Average dates for the first killing frost in the fall occur from September 30-October 30 in the more southern

418 portions of the range and from August 30-Sept 30 in the more northern. Average dates for the last killing frost in the spring occur from March 30-April 30 in the more southern portions of the range and after May 30 in the most northern. The average annual length of the frost-free period is from 160-200 days in the more southern portions of the species’ range and 80-120 days in the most northern (all information in this paragraph from Veregin 2005).

Ecoregions are geographic areas of similar physical, chemical, and biological characteristics organized within a hierarchical conceptual framework (Omernik and

Bailey 1997). Each level of the hierarchy shares important ecological attributes such as climate, geology, landform, hydrology, soils, and vegetation. In the United States the Wood Turtle’s range (Iverson 1992, Ernst and Lovich 2009) includes the following large-scale ecoregion “provinces” and “sections” (sensu Avers et al. 1994,

Bailey et al. 1994, McNab and Avers 1994, and McNab et al. 2007; also see 2008

EPA level IV ecoregions http://www.epgov/wed/pages/ecoregions/level_iv.htm.):

Outer Coastal Plain Mixed Forest Province 232 (MD, VA) Middle Atlantic Coastal Plain 232A Southeastern Mixed Forest Province 231 (VA, MD) Southern Appalachian Piedmont 231A Central Appalachian Broadleaf Forest – Coniferous Forest – Meadow Province M221 (VA, MD, WV, PA) Northern Ridge and Valley M221A (VA, MD, WV, PA) Allegheny Mountains M221B (MD, WV, PA) Blue Ridge Mountains M221D (VA, MD, PA) Eastern Broadleaf Forest (Continental) Province 222 (PA, NY, WS, MN, IA) Erie and Ontario Lake Plain 222I (PA, NY) Southwestern Great Lakes Morainal 222K (WS) North-Central U. Driftless and Escarpment 222L (WS)

419 Minnesota and Northeastern Iowa Morainal 222M (WS, MN, IA) Eastern Broadleaf Forest (Oceanic) Province 221 (VA, WV, PA, DE, NY, NJ, CN, MA, RI, VT, ME) Lower New England 221A (NJ, NY, CN, RI, MA, NH, ME) Hudson Valley 221B (NY, MA, VT) Upper Atlantic Coastal Plain (DE, NJ) Northern Appalachian Piedmont 221D (VA, MD, PA, NJ, NY) Southern Unglaciated Allegheny Plateau 221E (PA) Western Glaciated Allegheny Plateau 221F (PA) Laurentian Mixed Forest Province 212 (PA, NY, VT, ME, MI, WS, MN) Aroostook Hills and Lowlands 212A (ME) Maine and New Brunswick Foothills and Eastern Lowlands 212B (ME) Fundy Coastal and Interior 212C (ME) Central Maine Coastal and Interior 212D (ME) St. Lawrence and Champlain Valley 212E (NY, VT) Northern Glaciated Allegheny Plateau 212F (PA, NY) Northern Unglaciated Allegheny Plateau 212G (PA) Northern Great Lakes 212H (MI, WS) Southern Superior Uplands 212J (MI, WS) Western Superior 212K (WS, MN) Northern Superior Uplands 212L (MN) Northern Minnesota Drift and Lake Plains 212N (MN) Adirondack – New England Mixed Forest – Coniferous Forest – Alpine Meadow Province M212 (NY, VT, NH, MA) White Mountains M212A (VT, NH, ME) New England Piedmont M212B (MA, VT, NH) Green, Taconic, Berkshire Mountains M212C (CN, MA, VT) Adirondack Highlands M212D (NY) Catskill Mountains M212E (NY)

Overall, as was elucidated by Kuchler (1964), “forest” is the potential natural vegetation cover that generally characterizes (or would in the absence of human modifications) the on-the-ground reality of the above ecological classification areas.

In the United States the Wood Turtle occurs in the following large-scale forest regions

(and section) as defined by Braun (1950) and refined by Dyer (2006):

Mesophytic Region (VA, MD, DE) Appalachian Oak Section of the Mesophytic Region (VA, WV, MD, PA, NJ,

420 NY, CN, RI, MA, NH, ME) Beech – Maple – Basswood Region (PA, NY, WS, MN, IA) Northern Hardwoods – Hemlock Region (PA, NY, CN, MA, VT, NH, ME) Northern Hardwoods – Red Pine Region (MI, WS, MN).

“About two-fifths of the North is in forest use, a dominant land use (64 percent) in the Northeast [Connecticut, Delaware, Maine, Maryland, Massachusetts,

New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont,

West Virginia] in contrast to that for the North Central [Illinois, Indiana, Iowa,

Michigan, Minnesota, Missouri, Ohio, Wisconsin] region (28 percent). . . .

Hardwood forest types (e.g., oak-hickory) comprise about 80 percent of total forest area in the North (fig. 5), with about equal percentages in both the Northeast and

North Central subregions. . . . The oak-hickory and maple-beech-birch forest cover types are the largest in the North, combining to represent 62 percent or 106 million acres of unreserved forest land.” (Alig and Butler 2004). See forest cover map in

DeFries et al. 2000.

B. Habitat Type/Use and Behavior

Aquatic/Terrestrial Mosaic

After the four tortoise and two Box Turtle species, the Wood Turtle is the most terrestrial of the USA’s 58 turtle species (Conant and Collins 1991, Ernst et al. 1994).

Generally associated with wooded streams, the Wood Turtle is also North America’s most amphibious turtle species, requiring a mosaic of wetland and upland habitats in which to survive and carry out its complex biphasic (aquatic & terrestrial) life

421 history. Individuals regularly use a variety of both aquatic and terrestrial habitats.

Their use of these varying habitat associations is a function of seasonal, circadian, and weather-related factors. The Turtles depend upon this diversity of habitats for foraging, nesting, basking, cover, hibernation, and other needs.

Wetland habitats they affiliate with include hard-bottomed rivers, creeks, streams, and brooks and their associated swamps, fens, seeps, wet meadows, and

Beaver ponds. Turtles generally occupy low-gradient, rocky, clear-flowing, forested streams and rivers (Ernst and Lovich 2009). Such sites are normally waters 10-130cm deep with slow to moderate currents (Ernst 2001b, Ernst and Lovich 2009). A northeast study found a 1.5% gradient is generally the limit, but size of the watercourse is also a factor: Wood Turtles appear less tolerant of large streams with a high gradient (Jones 2009). This probably has to do with the volume of water discharged and the Turtles’ avoidance of flood events that can kill, maim or displace them (Sweeten 2008, Jones 2009). This is another reason (along with the availability of nesting areas) that historically Turtles may have preferred stream segments with high sinuosity (the meanders dissipate energy) (Beuch and Nelson 1991, Buech et al.

1997a). They typically affiliate with watercourses with substrates consisting of sand, small gravel, large gravel, and/or cobble, perhaps with some interspersions of boulders or bedrock (Akre and Ernst 2006).

In various parts of their range the Turtles use such terrestrial habitats as deciduous, coniferous, and mixed deciduous/coniferous forests and associated shrubland (such as alder thickets), meadows and glades, fields, and pastures. The

422 natural habitat patchiness in the Turtle’s range is reflective of a mosaic of soil types, topographic conditions, microclimates, successional stages, and disturbance regimes

(McNab and Avers 1994, Law and Dickman 1998, Lorimer and White 2003, Rentch

2006). Habitat preferences can vary greatly, both seasonally and geographically, as well as individually (Strang 1983, Kaufmann 1992). For instance, “in the range where conifer stands or alder thickets are absent, such as in southeastern Pennsylvania and northern Virginia, deciduous woods are heavily use” (Ernst et al. 1994).

The Wood Turtle can thus be considered a mosaic species, moving amongst and dependant upon various aquatic, terrestrial, and transitional habitats and microenvironments to satisfy thermal, hydric, food, social interaction, reproductive, and cover requirements (Carroll 1999, Compton et al. 2002, Arvisais et al. 2004,

Tamplin 2006, Akre and Ernst 2006, Carroll 2009, Krichbaum unpub. data). The

Turtles do not seem to use habitats randomly, suggesting they actively select their habitat (Strang 1983, Kaufmann 1992, Compton et al. 2002, Arvisais et al. 2004,

Akre and Ernst 2006, Tingley et al. 2010). Habitat selection may reflect a trade-off between feeding and hydro- and/or thermoregulation (Strang 1983, Compton et al.

2002, Arvisais et al. 2004, Dubois et al. 2009).

For instance, Turtles can typically be found near, in, or at the edges of small canopy gaps or semi-gaps and/or in low-lying vegetation in mature deciduous forest

(Akre and Ernst 2006, Remsburg et al. 2006, Krichbaum pers. obs.). Canopy gaps are a major factor structuring understory and overstory vegetation in deciduous forests of the eastern United States (Glasgow and Matlack 2007a). Such places provide food

423 items as well as cover from predators, while the dappled sunlight and/or small size of these sites allow the Turtles to vary thermal and moisture factors with ease and efficiency, as only small movements are necessary to be in either sunlight or shade.

Canopy gaps are important for sustaining herbal growth, richness, and persistence

(Anderson and Leopold 2002). Wood Turtles display an affinity for microhabitats with an abundance of herbaceous cover (Compton et al. 2002, Akre and Ernst 2006).

Range-wide, Wood Turtles are more terrestrial in summer, but there is some evidence that northwestern populations tend to be more aquatic in general (Vogt

1981, and Harding 1991). In addition, “it was observed that individuals living at higher latitudes reached sexual maturity later and at larger body sizes (Brooks et al.

1992), had much larger home ranges, and had more seasonal movements (Quinn and Tate 1991; Brooks et al. 1992; Arvisais et al. 2002)” (Tessier 2005).

C. Meso-Habitat

Within the previously referenced broad-scale ecoregions and forest regions the Turtle occupies various forest and vegetation cover types, as well as sundry aquatic habitats. For instance, in New Hampshire the Turtles are said to inhabit Appalachian

Oak-Pine Forests, Hemlock-Hardwood-Pine Forests, Lowland Spruce-Fir Forests,

Northern Hardwood-Conifer Forests, Grasslands, and Shrublands (see Ch. 3 of 2005

NH wildlife plan); in New York, Northern Deciduous Forest and Grassland (App. A4 of 2005 wildlife plan); in Minnesota, Red Pine-White Pine, Yellow Birch-Sugar

Maple, spruce-fir-hardwood, and upland mixed forests (Brown et al. 2016). Within this matrix the Turtles use flowing waters ranging from small 1st order mountain

424 streams to large floodplain rivers (such as the Potomac) (Krichbaum pers. obs.), as well as associated lentic habitats such as marshes and swamps (Brown et al. 2016).

Generally, throughout the Turtle’s range a relatively large number of tree species (16-35) predominate at any particular location (Dyer 2006, Rentch 2006).

There is a great deal of overlap of tree species, herbaceous flora, and fauna between these large-scale regions (Braun 1950, Newcomb 1977, McNab and Avers 1994,

Woods et al. 1999, Dyer 2006, McNab et al. 2007).

A disturbance regime of small-scale, within-stand gap processes dominated the natural forests in the northeast region (Runkle 1985, Runkle 1990, Mladenoff et al. 1993, Seymour et al. 2002, Rentch 2006). For example, in northern Maine White et al. (2004) found that “in different forest types (including hardwood, mixed wood, cedar seepage, cedar swamp, mixed conifer, and spruce) . . . [i]nstead of stand- replacing events, small gaps (< 0.1 ha) have dominated the disturbance history of these plots.” Manifest is the prominence of non-catastrophic tree replacement within the forest (gap phase dynamics). This natural disturbance regime of relatively small- scale canopy disruptions occurs through such mechanisms (both biological processes and physical forces) as windthrow, tree senescence, ice storms, drought, insects, Beavers, floods, pathogens, and lightning-strike ignitions (Braun 1950,

Rentch 2006). In addition to moisture, edaphic, elevation, and topographic gradients

(McEwan and Muller 2006, Lawrence et al. 1997, Ashe 1922), and disturbance events (Runkle 1991b, Glasgow and Matlack 2007a), canopy gaps are a major factor structuring understory and overstory vegetation in deciduous forests of the eastern

425 United States (Glasgow and Matlack 2007a). In research involving Appalachian mixed-hardwood sites, Miller and Kochenderfer (1998) found that “[c]anopy openings with a minimum diameter of 170 feet (0.5 acre) provide suitable light conditions for virtually all desirable [tree] species to develop and grow to maturity”.

Such disturbances and resultant canopy gaps are certainly of the appropriate spatial and temporal scale for Wood Turtle use.

Large “catastrophic” stand replacing events, such as hurricanes and conflagrations, are naturally a rare occurrence. Most of the Turtle’s range is outside of the zones of frequent hurricanes and stand replacing fires have very long return intervals (Lorimer and White 2003) For example, a natural fire rotation in northern hardwoods was estimated to be 1070 years, the fire rotation periods in spruce- hardwoods were estimated to be 1253-1519 years, and the period for severe windthrow in the mixed spruce–hardwood forests and northern hardwood dominated forests on the better soils was 2585 years (id.).

The congruence and harmonization, and lack thereof, of human disturbance

(viz., cutting regimes) with the spatial and temporal parameters of natural disturbance and their associated biological legacies are of concern, or should be, throughout the

Turtle’s range (Flamm 1990, Franklin et al. 2002, Seymour et al. 2002, Lorimer and

White 2003, Keeton 2004). Researchers in northern Maine found that “[t]he most obvious silvicultural analogs to this [natural] disturbance history are individual tree selection and group selection systems” (White et al. 2004). The scale and intensity of logging of any type do not happen in a vacuum; even if in and of themselves such

426 activities are considered to be tolerable, their attendant indirect impacts and their cumulative impacts in conjunction with the other factors operant at a locality must be considered and weighed.

Wood Turtles can be tolerant of mild habitat alterations such as small-scale openings in the streamside canopy that may create feeding and nesting areas

(Harding 1991). However, depending on their scale and context, the benefits of such activity could be offset by effects of harvest machinery (e.g., compaction of soil, destruction of nesting habitat, the crushing of Turtles), harmful edge effects (e.g., facilitation of depredation and invasive plants), and sedimentation (Saumure et al.

2007). Habitat alterations are only tolerable when they do not have catastrophic or persistent long-term effects on habitat integrity or do not draw in subsized predators

(Akre pers. comm. 2009). Within the Turtle’s “core habitat” zones (see habitat section

V.) it is certainly oftentimes sensible to minimize human disturbance and allow the dynamic of natural processes to be expressed.

Populations are considered to be declining at numerous sites that have a variety of human disturbance (Garber and Burger 1994, Saumure and Bider 1998,

Daigle and Jutras 2005, Akre and Ernst 2006, Saumure et al. 2007, Tingley et al.

2009, Jones 2009).

Wood Turtles may benefit from habitat management and alteration that mimics Beaver activity, where ponds and meadows in various stages of succession are distributed along streams, but only when other management activities in the area do not already contribute to the degradation of stream quality and the availability of

427 relatively cool, moist microhabitats (Saumure 2004, Akre pers. comm. 2009).

As a mosaic species that travels across variously scaled ecotonal edges in its daily and seasonal peregrinations (Strang 1983, Carroll 1999, Compton et al. 2002,

Tamplin 2005), the Turtle has been described as “disturbance-dependent” (Saumure et al. 2007). However, a propensity to occupy terrestrial habitats open to various degrees of human disturbance can expose the Turtle to significant risk in the contemporary landscape (id.).

It is sometimes stated or implied that the Wood Turtle does not prefer “contiguous forested habitat” or that it is not a “wilderness” species. Such characterizations are not accurate and instead indicate problematic nomenclature, concept, or perception.

A forest can be “intact” or “contiguous” yet have numerous canopy openings due to a variety of natural disturbances (see, e.g., McCarthy 2001). In fact, this is the natural state of wild old growth forests in the Turtle’s range (Davis 1996). Such forests exhibit great structural complexity and heterogeneity at multiple spatial scales

(Franklin et al. 2002, Davis 1996, Runkle 1991b). And it is such forests that the Turtle has lived in and has adapted to over the course of its evolutionary history. Mature forests are of the age that a mosaic of habitats is gaining expression due to the operant disturbance regime (Franklin et al. 2002). And still more such niche complexity

(including canopy openings) can be expected to develop as mature forests develop into old growth (Dahir and Lorimer 1996).

428 In their character and proclivities Wood Turtles are nothing if not dynamic

(Akre pers. comm. 2009). The term “wilderness” is a shorthand referent for places and conditions where dynamic natural processes and conditions are untrammeled by humans. In the case of forests, it is where they are allowed to attain old-growth condition and express a natural disturbance regime. And it is through natural disturbances that forest ecosystems sustain and perpetuate themselves in the long- term as “shifting mosaics” at dynamic quasi-equalibria (Bormann and Lickens 1979,

Shugart and West 1981). And it is precisely to such a dynamic natural disturbance regime that the Turtle has adapted. Such “wilderness” forests (of sufficient age) typically include stands or patches dominated by young early successional forest, old early successional forest, mid-successional forest, young late successional forest, and old late successional forest (Franklin et al. 2002, Frelich and Reich 2003). In other words, “wilderness” means not just a particular place or current pattern, but additionally means incorporating the processes of change (Harris et al. 1996) that supply an interacting mosaic of vegetation associations shifting in time. The natural disturbance regime is a critical element of the forces and selection processes that shaped this particular expression of biodiversity (viz. forests) in the first place

(Hansson and Angelstam 1991), and that are necessary to maintain it in situ in the future. In the absence of human logging/cutting/clearing disruptions, maturing and old-growth forest tracts undergoing natural disturbance support numerous microhabitat patches that are used by wildlife on a fine scale (see, e.g., Fahrig and

Merriam 1994, Law and Dickman 1998). The disturbance regime in northeastern

429 forests is generally characterized by small-scale canopy disruptions (Braun 1950,

Runkle 1991a, Rentch 2006). These disturbances and resultant canopy gaps are certainly of the appropriate spatial scale for use by Wood Turtles (Krichbaum pers. obs.).

The Beaver (Castor canadensis) was an important element of the historic disturbance regime throughout the Turtle’s range. Prior to European settlement (and consequent heavy trapping), Beaver wetlands were much more abundant in North

America (Bolen and Robinson 1999). Beavers are a riparian keystone species that can provide and improve Wood Turtle habitat (Naiman et al. 1988, Carroll 1999,

Russell et al. 1999, Wright et al. 2002).

In Wisconsin, alder thickets associated with shorelines are considered to be critical habitat for hatchlings and juveniles (Wisconsin DNR 2005). In New

Hampshire, Carroll (1999) found revegetated Beaver clearings to be especially important to hatchlings and juveniles. While in Virginia, Akre (2009 pers. comm.) reports that wet/mesic meadows, glades, and seepages with thick herbaceous cover appear to be particularly important to these younger age classes. Carter et al. (1999) believed that congeneric Bog Turtles (G. muhlenbergii) in southwestern Virginia responded more to structural habitat components than to actual vegetation type.

Thus, the presence of specific habitat types, such as alder for example, may not be critical. Indeed, as Ernst and Lovich (2009) point out, in places where alder thickets are absent deciduous woods are instead heavily used.

430 Meso-Habitat Example from Virginia/West Virginia

Ecological communities/vegetation types (nomenclature as per Fleming and Couling

2001) where Wood Turtles have been observed on the George Washington National

Forest: Central Appalachian rich cove forest, White Pine - Eastern Hemlock forest, acidic oak – hickory forest, Chestnut Oak – Northern Red Oak forest, Mixed oak – heath forest, Northern Appalachian xeric oak/heath forest – Chestnut Oak/low elevation subtype, White Pine – xeric oak forest, xeric shale woodland (both

Virginia Pine and Chestnut Oak types), small-stream montane forest, basic seepage swamp, and acidic seepage swamp (Krichbaum pers. obs.).

D. Micro-Habitat

Within their broadly defined habitats (such as “Mixed Oak Forest” or “Emergent

Wetland”) Wood Turtles may actually be selecting for vegetational composition or structure on a micro scale (Compton et al. 2002). Therefore, conclusions about land cover type selection by Wood Turtles should not be made without supplemental microhabitat data (Hamernick 2000, Tinsley et al. 2010).

One correlate may be local fine-scale responses of litter arthropods, fungi, and herbs to microhabitat/microclimate conditions. Turtle microhabitat use may be in response to variations in prey/food abundances, which in turn may be correlated to seasonal changes in factors such as litter depth and rainfall. Strang (1983) and

Kaufmann (1995) noted seasonal differences in terrestrial habitat use apparently in response to the variance in availability of fungi, herbs, berries, and slugs. As Turtles may have access to foods such as berries, insects, and mushroom only seasonally or

431 intermittently, perhaps they utilize an energy minimizing strategy whereby cool microhabitats are selected to reduce overall metabolic costs when food is less abundant (such as in mid- to late summer) and energy costs are elevated by high ambient temperatures (see Penick et al. 2002 as regards Box Turtles). Observations of the closely related Spotted Turtle (Clemmys guttata) in Ontario revealed that although they may appear to be targeting aquatic vegetation, it is most likely that individuals are actually seeking the small freshwater snails and benthic invertebrates

(prey items) residing within the matrix of aquatic vegetation (Rasmussen et al. 2009).

Net energy gain (and hence growth and reproduction), and even survival, depend not only on food levels but also on the thermal environment; consequently, patches may be selected that are “suboptimal” in regard to food resources (Huey

1991). Thus, Turtle temporal and spatial activity patterns appear to be influenced not just by dietary preferences, but thermo- and hydro-regulation as well (Strang 1983,

Foscarini 1994, Compton et al. 2002, Arvisais et al. 2004, Akre and Ernst 2006).

Wood Turtles perhaps select different microhabitat features depending on their activity (e.g., open areas for basking and forest for foraging) (Compton et al. 2002).

In other words, Turtles face tradeoffs in resource allocation to various compartments of their energy budget, choices that ultimately have significant implications for reproductive output (Congdon 1989, Penick et al. 2002).

The Wood Turtle’s apparent selection of terrestrial microhabitat may be a direct response to microclimatic moisture and temperature gradients (Dubois et al.

2008 & 2009). Wood Turtles are often characterized as being associated with

432 relatively more mesic habitat conditions (Mitchell 1994) and humid microclimates

(Akre and Ernst 2006); in some regards such conditions are similar to those preferred by terrestrial woodland salamanders (sensu Davic and Welsh 2004). Wood Turtles were found to be more vulnerable to evaporative water loss than Box Turtles

(Terrapene carolina) (Ernst 1968). Wood Turtle activity (daily path length) was found to correlate positively with relative humidity (Strang 1983). The attributes (surface temperature, relative humidity, and understory plant cover) most important for defining the microhabitat of sympatric Eastern Box Turtles (T.carolina) are apparently related to minimizing water loss and to thermoregulation (Penick et al. 2002, Rossell et al. 2006).

Maintaining body temperatures within a salubrious range is a primary factor of Turtle biology. “[M]ost macrohabitats are thermally heterogeneous, and the actual

body temperature (Tb) an animal achieves depends on its behavior (e.g., choice of microhabitat, activity level, posturing), its morphology (e.g., size, color), [and] its physiology (e.g., metabolic rate, water loss)” (Huey 1991). “To maximize survivorship, selection of preferred environmental temperatures must be successfully coupled with foraging and predator avoidance behavior” (Tamplin 2006). “The main benefit of thermoregulation for freshwater turtles, increased energy gain, should be higher in climatic extremes. At high latitudes, temperature can be viewed as an ecological resource (Magnuson, Crowder & Medvick 1979, Tracy & Christian 1986) that may limit growth and reproduction of ectotherms more than food availability because food processing rate is constrained by low temperatures and short growing

433 seasons (Congdon 1989; Grant & Porter 1992; Niewiarowski 2001).” (Dubois et al.

2009)

Observations of the Turtles in the Virginias suggest a propensity for associating with large-diameter woody debris (LWD), both on land and in the water (Akre and

Ernst 2006, Krichbaum pers. obs.). Their affiliation with aquatic LWD is particularly salient (see, e.g., Jones 2009). In addition to providing a spatial setting for Wood

Turtle activity, woody structure (both dead and alive) increases the physical and biological stability and complexity of a stream. Living and dead woody structures

(i.e., tree and root masses) and large woody debris should be maintained in all rivers/streams with known Wood Turtle populations, as well as those with potential populations and the potential for dispersing individuals. Such structure should be restored to a sufficient level in those rivers/streams that have been deemed deficient.

Kaufmann (1995) found that juveniles often hide underneath riverbanks.

The lack of adequate no-harvest buffers along all classes of stream channels means the recruitment of LWD will be reduced, thus impeding the provision of this critical habitat element in streams. A similar impoverishment of course may occur in terrestrial habitat when the Turtle’s core habitat is subjected to intensive timber harvest.

Various mushroom species are important elements of the Turtle’s diet (see, e.g., Strang 1983, Kaufmann 1992, Tuttle 1996, Compton et al. 2002, Walde et al

2003). In addition to log size, macrofungal and myxomycete fungi richness was significantly positively correlated with amounts of CWD at old age oak and mixed

434 mesic forest study sites in Ohio (Rubino and McCarthy 2003). Similarly, in New

Hampshire all sites with above average coarse woody debris cover had above average numbers of species of macro-fungi, with mean mushroom diversity in old growth sites being 2.5 times the amount in non-old growth sites (Van de Poll 2004).

E. Temporal Patterns

Daily and seasonal movement patterns of turtles vary in response to environmental, physiological, and demographic conditions (Gibbons et al. 1990). Wood Turtles are mostly diurnal and exhibit seasonal shifts in activity and habitat usage (Ernst 2001a).

“Phenology” refers to the study of periodic events in biological organisms; in the

Turtle’s case many of the timings of these events are correlated with patterns of climate and weather. Phenology is an indicator of differing environmental conditions, especially weather-related conditions that, over time, can amount to climate change.

The seasonal activity of the Turtle generally involves aquatic winter dormancy, a spring semi-aquatic period, a summer terrestrial or semi-terrestrial period, and a fall semi-aquatic period (Harding and Bloomer 1979, Ernst et al. 1994,

Tamplin 2005, Sweeten 2008). Arvisais et al. (2002) divided the active season of a

Quebec population into four distinct periods. Akre and Ernst (2006) added clarification based upon the literature (e.g., Harding and Bloomer 1979, Ernst, et al.

1994, Foscarini and Brooks 1997, Niederburger and Seidel 1999, Compton et al.

2002) and their observation of Virginia populations: “1) during the hibernation period (December-February) turtles are wholly aquatic and primarily quiescent with

435 few short movements, 2) the emergence (March) is characterized by a rapid increase in the frequency and distance of movements, 3) the ‘prenesting’ period (April-May) is characterized by mostly short terrestrial distances traveled and large amounts of time spent basking on riverbanks and in riparian areas, 4) the ‘nesting’ (June) period involves the largest movements of the season and movements to nesting sites are an important reason for this, 5) the ‘postnesting’ (beginning of July to end of September) period is defined by smaller movements than the nesting period, use of aquatic and terrestrial habitat, and variable individual movements, and 6) the ‘prehibernation’

(October-November) period when turtles return to and are active primarily in the river or stream, engage in an autumn mating peak, and eventually cease movement and become dormant.” In northern Virginia and West Virginia active Wood Turtles have been observed in all months except February (Krichbaum pers. obs.).

Though the basic patterns appear to hold true, the specific time periods for these patterns differ across the Turtle’s range (Arvisais et al. 2002). “The wood turtle’s annual surface activity reported here for southeastern Pennsylvania is longer than that reported for more northern populations – late April or May to early October in

Michigan (Harding, 1991), early March to late November in northeastern New Jersey

(Farrell and Graham, 1991) – but not as long as that of insculpta at the southern extreme of the range in northern Virginia – March to January (Ernst and McBreen,

1991).” (Ernst 2001a) Jones (2009) reported the following biologically based Turtle

“seasons” for Massachusetts: “Early spring (1 March - 5 May) is defined as the period following emergence from hibernation, when turtles of both sexes [are] within or

436 immediately adjacent to the stream corridor. Spring (6 May to 25 May) is a period when turtles move away from the stream corridor, but females have not yet begun to nest. Nesting (25 May – 4 July) is bracketed by the earliest and latest observed nesting events, Summer (5 July – 21 September) is that period following nesting, before the turtles have returned to the stream corridor. Autumn (22 September – 15 November) is the period immediately prior to hibernation, when wood turtles may be found congregated in small groups near prominent structural features in the stream

(logjams, undercut banks, exposed roots).”

Daily and seasonal activity patterns are variable and influenced by environmental conditions (Harding and Bloomer 1979, Brewster 1985, Farrell and

Graham 1991). Chronology of habitat use is related to air temperature (Kaufmann

1992) and associated emergence of vegetation and fungi (Strang 1983, Arvisais et al.

2004). Certainly, nearness to water during prenesting and prehibernation activity periods facilitates control of body temperature (Foscarini 1994, Dubois et al. 2009) and hydration (Strang 1983). For ectotherms (such as the Turtle) that inhabit heterogeneous macrohabitats, an individual’s morphology, physiology, and behavior

(e.g., selection of microhabitats) serve as filters to transfer the environmental thermal regime into its actual body temperature (Huey 1991).

F. Temperature Relations

The Turtle’s thermal activity range is 3.4-32.0°C (Ernst 2001b). The maximum cloacal temperatures recorded for free-ranging Wood Turtles are 30-33°C (Ernst

1986, Farrell and Graham 1991, Ross et al. 1991, Tuttle 1996, Tamplin 2005). Turtles

437 engage in behavioral thermoregulation to sustain temperatures optimal for their activities (Tamplin 2006). The optimal temperature for digestive performance in turtles is 29-30°C and for growth rate is 30-31°C437(Dubois et al. 2008). For active

Wood Turtles in Pennsylvania, Ernst (1986) reported a mean cloacal temperature of

21.01°C, with a mean on land of 27.2°C. In a New Hampshire study, for measurements taken between April and December, the mean cloacal temperature for all activities and months was 23.4° ± 5.7°C; temperature ranged from 3.8°C for a

Turtle found dormant in water to 32.0°C for one found dormant on land (Tuttle

1996). In a New Jersey study air and substrate temperatures were significantly lower at capture sites for juveniles than at those for adults (Castellano et al. 2008). In an experimental setting the Turtle’s mean preferred temperature was 27.5°C (Nutting and Graham 1993). The “optimal” body temperature maintained by Wood Turtles in an experimental situation (ca. 30°C), however, may be closer to the thermal maximum in order to increase digestive performance and net energy available for growth and development; this was particularly the case for juveniles (Dubois et al.

2008).

The work of Tamplin (2005) in Iowa during the months of April-July is illustrative: “Mean body temperature was highest in June, and lowest in July, but was closely regulated near 25oC in all months and was maintained several degrees above environmental levels from April through June (Figure 5). In July, mean body temperature was lower than mean air temperature as the turtles limited their exposure to sunlight. Maximum body temperature observed was 33.3oC; minimum

438 body temperature of active turtles was 12.8oC, although active turtles were observed in water at temperatures below 10oC. Amount of light exposure selected by turtles peaked in April and subsequently declined each month (Figure 6) as the turtles thermoregulated by maximizing environmental heat gain when air temperatures were low and minimizing heat gain when air temperatures approached 30oC.”

G. Over-wintering

Turtles select specific water depths and substrates for over-wintering (Akre and Ernst 2006, Greaves and Litzgus 2008). They hibernate submerged in flowing water, in well-oxygenated pools deep enough to not freeze entirely, under the mud or leaves on the bottom, exposed on the river bottom, in undercut banks, or under submerged woody debris, log jams, tree roots, or root balls (Farrell and Graham

1991, Ross et al. 1991, Kaufmann 1992a, Niederburger and Seidel 1999, Ernst

2001a, Akre and Ernst 2006, Greaves and Litzgus 2008, Jones 2009).

Movements in winter may be to prevent covering by substrate in order to facilitate aerobic respiration (extrapulmonary gas exchange) (Ultsch 2006).

Particularly in more northern locations, survival is contingent upon finding hibernacula that meet a narrow range of critical variables (e.g., flowing water with high dissolved oxygen concentrations and adequate protection from predators and freezing temperatures) (Greaves and Litzgus 2008). Wood Turtles use the structural features of overhanging roots along the bank, submerged logs, coarse woody debris packs, and Beaver lodges or Muskrat burrows not only for hibernation, but also for general refugia, courtship and reproduction, and foraging/feeding (see, e.g.,

439 Kaufmann 1995). In a Virginia study, 75% of Turtles exhibited some form of overwintering refuge site fidelity by returning to within approximately 100 meters of their prior year’s hibernaculum (Akre and Ernst 2006).

H. Courtship and Reproduction

Courtship and mating primarily take place in water in the spring and fall

(Kaufmann 1992b, Ernst 2001a, Walde et al. 2003, Jones 2009, Krichbaum pers. obs.). Courtship and mating behaviors are aggressive and prolonged. These involve the males’ biting and chasing after females, as well as lengthy mountings that include bouts of shaking and thumping of their plastron against the females’ carapace

(Kaufmann 1992b). During courtship mountings, male Wood Turtles aggressively bite females when they try to escape or extend their head from the shell (Kaufmann

1992b, Walde et al. 2003). Many mountings do not result in copulation (Kaufmann

1992b).

There is no evidence that female Wood Turtles lay more than one clutch per year (Powell, 1967, Farrell and Graham, 1991, Harding 1991, Walde 1998, Akre

2002). Clutch-size generally averages 7-10 eggs (Farrell and Graham 1991, Ross et al. 1991, Brooks et al. 1992, Tuttle and Carroll 1997, Akre 2002, Jones 2009,

Krichbaum pers. obs.). Several studies have found that Wood Turtle clutch size is positively correlated to female straight-carapace length (Brooks et al. 1992, Walde et al. 2007, Akre 2002, Jones 2009). Walde et al. (2007) also found that larger females were more likely to nest in consecutive years than smaller females. Individual females may not nest every year. Approximately 75% of adult females in Ontario

440 and Virginia studies nested annually (Akre 2002). In a New England study, the proportion of years in which adult female Turtles became gravid over multiple consecutive years averaged 0.71 (Jones 2009).

Eggs are laid in late spring and early summer, late May to early July. The time of onset of nesting in insculpta shows little latitudinal variance. Nesting has been observed during June in Pennsylvania (Ernst 2001a, Kaufmann 1992b), New

Hampshire (Tuttle and Carroll 2005), New Jersey (Castellano et al. 2008), Iowa

(Tamplin 2005), Virginia (Krichbaum 2009, Akre and Ernst 2006), Massachusetts

(Jones 2009), Ontario (Brooks et al. 1992), and Québec (Walde et al. 2007).

Oviposition frequently occurs during the evening (see “90%” in Jones 2009; Carroll

2009, Krichbaum pers. obs.).

The eggs are laid in a nest cavity dug in well-drained soil (Farrell and Graham

1991). The nesting site may be a considerable distance from water, to 700 meters away (Ernst 2001b). Nest site fidelity has been reported for individuals in some populations (Walde 1998, Walde et al. 2003), and this may occur frequently in locations where optimal nesting sites persist across years, but it has been observed that females use novel nest sites in different years just as often (Harding 1991, Akre

2002). Wood Turtles may nest in a variety of vegetative and edaphic settings and nests may be spread across suitable patches on the landscape (Jones and Willey

2015). In addition, the availability of suitable and accessible nesting habitat conditions (such as an open sandy-soil area) may attract females to nest communally in relatively high density (Walde et al. 2007, Akre and Ernst 2006).

441 Within a watershed or along a stretch of river or stream with adequate terrestrial and aquatic habitat, suitable nest sites may be the most limiting facet of wood turtle habitat (Buech et al. 1997). Harding (1997) described preferred nesting habitat as exposed elevated areas with moist sand or sandy soil receiving full afternoon sun. In Virginia females appeared to preferentially select a variety of substrates, relatively dry and well-drained, on predominantly southeast or south- facing aspects with good solar exposure (Akre and Ernst 2006). Perhaps as with sympatric Eastern Box Turtles (Belzer et al. 2007), nest site selection may be influenced by soil temperature. Nests are often close to water (e.g., within 10-150 m) (Harding 1994, Buech et al. 1997a, Siart 1999, Akre and Ernst 2006) and include railroad grades, sand/gravel pits, eroding riverbanks, sand bars, utility corridors, dirt roads, and road shoulders (Brooks et al., 1992, Buech et al. 1997, Walde 1998,

Wusterbarth 2000, Krichbaum 2009, Jones and Willey 2015). However, nest sites may also be some distance away from water (for instance, hundreds of meters) (Akre

2002, Behler and Castellano 2005, Akre and Ernst 2006, Greaves 2007). Nesting may entail long-distance travel by females to reach a particular site; for example, three kilometers (Paradis et al. 2004, Krichbaum pers. obs.)

In Massachusetts, “[o]f 52 nests with known emergence rates, 35% were deposited on beaches along the stream in which the turtle over-wintered, 27% were deposited in gravel pits, 19% were deposited on sand piles or along dirt roads in pastures, 4% were deposited in powerlines, and 2% each were deposited along dirt roads and in a corn fiel Sixty-four percent were deposited in sand, 29% were

442 deposited in mixed sand and gravel; 6% were deposited in organic material such as lichen, coarse woody debris or loam or a combination of organic material and sand, and 2% were deposited in gravel.” (Jones 2009) In an Ontario study of microsite- scale preferences, nesting females chose sites with large sand grain sizes, high incubation temperatures, low moisture content, and low organic content (Hughes et al. 2009).

Nesting habitat in Virginia includes instream and stream bank sandbars, gravel bars, coarse, alluvial cobble-gravel deposits, tilled floodplain agricultural fields, fallow or grazed pasture, rotting logs, cleared utility corridors, unpaved road beds, road-cut embankments, and garden mulch beds and manure piles (Akre 2002,

Akre and Ernst, 2006, Krichbaum 2009). In Minnesota, suitable nesting sites were described as having a sand or gravel substrate, less than 40-degree slope, and low disturbance, close to water but at least 1 meter above the normal water level, vegetation cover less than 20%, and the height of woody vegetation should be less than the distance to the southern edge of the nesting area (Buech et al. 1997). Buech et al. (1997) also found that if the slope of the nesting area is less than 20 degrees then any aspect is acceptable, but if the slope is greater than 20 degrees the aspect is generally east-southeast or west-southwest. They emphasize that nesting needs are specific and that sandy soil appears to be the most limiting factor. See Tuttle and

Carroll 2005, Castellano et al. 2008.

The incubation period (number of days from egg deposition until hatchling emergence) in the wild generally varies from 60-78 days (Brewster and Brewster

443 1991, Farrell and Graham 1991, Tuttle and Carroll 2005, Castellano et al. 2008, and

Krichbaum pers. obs. [62-78 days for clutches in Virginia in 2010]). However, incubation period in Québec ranged from 60 to 116 days (Walde et al. 2007).

Incubation time length is most dependent upon temperature (Harding 1991, Walde

1998).

Jones (2009) reported an emergence rate (nest success), exclusive of depredation by mammalian predators, that ranged from 0 to 100% and averaged

41%. Foscarini (1994) reported a high probability of egg or hatchling mortality, with

80% - 83% of nests destroyed at her Ontario study site. Reported nest depredation is often high, e.g., 70-100% of nests (Buech 1991 & 1992, Harding 1992 & 2002,

Buech et al. 1993, Hunter et al. 1999, Tamplin 2005) and intact nests do not necessarily hatch out all the eggs (Walde et al. 2007). Due to a combination of insufficient solar energy and frost-free days, in the northern portions of the species range at least 50% of nests may not hatch in any given year (Compton 1999).

Hatching success of eggs may be low or virtually nil in cool years (Foscarini 1994,

Smith 2002).

Emergence of hatchlings occurs from July-September and may be staggered over several days (Harding and Bloomer 1979, Harding 1991, Akre and Ernst 2006,

Krichbaum pers. obs.). Hatchlings may overwinter in the nest (Carroll and Ultsch

2007), but in most situations they apparently do not (Akre 2002, Buech et al. 2004,

Parren and Rice 2004, Walde et al. 2007). After hatching, neonates spend time on land. When terrestrial, the neonates most frequently select dense herbaceous or

444 woody cover (Tuttle 1996, Dragon 2014). In New Hampshire, mean time to reach a brook was six days (range 1-24) during which time the hatchlings traveled straight- line distances of 28 to 445 meters (mean of 97 meters) (Tuttle and Carroll 2005). In

New Jersey it was substantially longer for the three hatchlings followed to water, with a mean of 43 days and maximum of 62 (Castellano et al. 2008). The mean total distance traveled by these three was 546 meters (range 163–1052); a distance eight times greater than the mean straight-line distance between their nests and the nearest water. Hatchlings were observed foraging on slugs (id.) In an experimental setting, juveniles (3-, 6-, and 12-months old) selected the warmest temperatures available in an aquatic thermal gradient (Tamplin 2009). Hydro-regulation or access to water may be a significant factor affecting hatchling behavior or survival. In New

Hampshire, Carroll (2009) encountered a hatchling that when offered water drank for twenty-one minutes while adjacent to the corpse of another hatchling in an open sunny are

I. Behavior

Males exhibit aggressive behavior such as chasing, biting, and lunging (Kaufmann

1992b, and Barzilay 1980). This appears to be especially prevalent during the spring and fall mating seasons, but at other times as well (Farrell and Graham 1991, Ernst

2001a, Walde et al. 2003). Walde et al. (2003) documented bleeding and injuries to male and female wood turtles resulting from aggressive mating and male–male combat. Wood Turtles even express a hierarchal social structure based on size, sex, and maturity, with male rank affecting reproductive success (Kaufmann 1992b,

445 Galbraith 1993).

Some Wood Turtles display an unusual feeding habit called “worm stomping”. This involves stomping their front feet or tamping their plastron against the ground to draw earthworms to the surface. The vibrations presumably mimic those produced by raindrops hitting the soil or by a tunneling mole. This behavior has been observed in Michigan (Harding and Bloomer 1979), Pennsylvania

(Kaufmann 1986), New Hampshire (Tuttle and Carroll 1997), Virginia (Akre pers. comm.), and West Virginia (Krichbaum pers. obs.).

Some have described whistling calls (as of a steaming teapot) made by male

Wood Turtles (Cochran 1954, Carr 1952). It is speculated that these sounds may have something to do with mating. Neither the information conveyed by the signaler to conspecifics nor the influence these sounds have on sexual interactions is known

(see Galleotti et al. 2005).

Among turtles, they show an environmental awareness that may go well beyond most other species. Tinklepaugh (1932) reported that Wood Turtles tested in a maze exhibited the learning ability of a rat. Wever and Vernon (1956) found the

Wood Turtle’s hearing to be very sensitive to low tones (< 500 hz.), in fact, even more sensitive than that of the domestic cat. Wood Turtles in captivity will also learn pertinent parts of the human daily routine; a Turtle held captive by one researcher learned its way around the house and associated the kitchen with food and the bathroom with swimming (Carr 1952, Harding and Bloomer 1979). A female Turtle in Virginia has returned to an individual’s home for 28 years in response to the

446 provision of food on a porch (Akre and Ernst 2006). The Turtle's apparent intellect boosts its popularity as a pet (Ernst 2001b).

G. insculpta are somewhat the all-terrain-vehicles of the turtle world. Their large muscular legs (Abdala et al. 2008), speed (Ernst et al. 1994), and endurance

(Stephens and Wiens 2008) attest to their versatile abilities. Their agility in clamoring over downed logs and rocks appears far beyond the capabilities exhibited by Box

Turtles or other chelonians (Krichbaum pers. obs.). Wood Turtles are renowned for their climbing and locomotor abilities, having been observed up in bushes, negotiating stairs, going over fences (Pope 1939, Harding and Bloomer 1979), and opening screen doors (Knowlton 1943).

J. Spatial Ecology – Core Habitat/Home Range

Quinn and Tate (1991) found that abrupt linear movements to late summer ranges in mid-June were common events. Although Ernst (2001a) and Arvisais et al. (2002) found no differences in activity patterns between the sexes, Tuttle and Carroll (2003) and Akre and Ernst (2006) found, in New Hampshire and Virginia populations, respectively, that females tended to be more terrestrial than males during the active season. In addition, adult females in Ontario (Foscarini and Brooks 1997) and

Massachusetts & New Hampshire (Jones 2009) moved significantly farther away from watercourses than males. However, some individuals, usually males, travel great distances (≥ 2 km) overland or by stream across seasons (Ross et al. 2000, Saumure

2004, Akre and Ernst, 2006, Remsberg et al. 2006, Walde et al 2007, Sweeten 2008,

Jones 2009, Krichbaum pers. obs.); these movements are presumably an infrequent

447 but critical component of Wood Turtle behavior and demography.

During the pre-nesting period when females move farther from the home stream than males as they foraged wider in more habitats, it is done presumably to accumulate the nutrients, calories, and an optimal body temperature range necessary for egg production and nesting (Remley and Rhymer 1997, Kuchling 1999). Similarly, during the post-nesting period females are slower to return to the streams as they seek foraging opportunities in order to replenish nutrients and fat storage used in May and June (Kuchling 1999). The search for and competition for mates appears to be a significant driver of male movement patterns (Strang 1983, Kaufmann 1992b, Jones

2009).

Some male Wood Turtles in Virginia were observed to move at least 1 km between their hibernacula and summer ranges (Ernst and McBreen 1991, Akre and

Ernst 2006). Sweeten (2008) found all Wood Turtles moving distances ≥5 km to be males (however, at least one of these moves was due to displacement from a flood event). A large adult male radio-tracked in Massachusetts moved (mostly overland) a straight-line distace of 17 km across three watersheds (Jones 2009). Females are also capable of long distance movements. Although typical pre-nesting female movement is between 200-600 m from a stream, they have been documented to move nearly 4 km (Behler and Castellano 2005, Walde et al. 2007), and movements of at least 1 km were not uncommon in Virginia (Akre 2002, Akre and Ernst 2006,

Krichbaum pers. obs.). Brewster and Brewster (1991) noted a juvenile moved a distance of 500 meters.

448 If displaced less than 2 km from their home range Wood Turtles display good homing abilities (Carroll and Erhenfeld 1978). “My study presents additional evidence that wood turtles are capable of returning to familiar sites after being displaced more than 3.5 km downstream. . . . At least one displaced wood turtle in my study returned to her initial capture site by traveling over land, indicating that wood turtles are not constrained to river corridors when returning home after floods.”

(Jones 2009) Two male Wood Turtles displaced by a 6,000 cfs flood in Rockingham

County, Virginia, initiated upstream travel following displacement, but did not return to their original capture site (Sweeten 2008).

Although a primary requirement of their life history is the presence of water, they habitually use terrestrial habitat and are certainly not confined to waterways or narrow “riparian” zones. Spring and fall are transitional periods. They start spending more time on land once spring temperatures rise. In the summer they spend much time away from water, living a solitary, terrestrial lifestyle in the same types of habitat as and behaving similar to the Eastern Box Turtle (Strang 1983, Farrell and Graham

1991). As autumn temperatures drop, the Turtles move back closer to water and become increasingly aquatic. Seasonal movements appear to be dependent on ambient temperature (Niederberger and Seidel 1999, Akre and Ernst 2006). It appears that the requirements for thermoregulation (and hydroregulation) impose constraints on the spatial distribution and temporal activity patterns of individual Turtles,

449 particularly at higher latitudes (such as the use of aquatic thermal refuges at night)

(Dubois et al. 2009).

As Conant and Collins (1991) observe: "it frequently wanders far afield through woods and meadows, across farmlands, and – often with fatal results – on roads and highways." Radio-tracking studies clearly show that they may normally range up to 200-600 meters (660-2000 feet) from the water (Kaufmann 1992, Arvisais et al. 2002, Compton et al. 2002, Akre and Ernst 2006, Remsberg et al. 2006). The normal extent of their terrestrial use of habitat varies amongst individuals, populations, and site conditions (Saumure 2004).

Turtles were found to range over 500 meters from the streams in Ontario

(Quinn and Tate 1991, Foscarini and Brooks 1997). In the Ridge and Valley ecoregion of Pennsylvania the maximum distance was found to be 600 meters

(Kaufmann 1992). In a Québec study all sightings were made within 300 meters of streams used by the Turtles (Arvisais et al. 2002). In Nova Scotia 95% of all female

Wood Turtle locations were within 235 meters of water (Tingley 2009). In West

Virginia male Wood Turtles have been observed up to ca. 400-600 meters from streams (Krichbaum per. obs.). At another West Virginia study site the greatest known distance traveled from a river was approximately 200 meters (Niederberger and

Seidel 1999). In Michigan’s Huron National Forest, “92.5% were within 200 m of the river. Ten of 29 telemetered turtles moved > 200 m from the river . . . Only 2 turtles, composing less than 4% of turtle locations (n=36), traveled more than 500 m from the river” (Remsberg et al. 2006). In Maine, 95% of Turtle activity areas were

450 within 304 m of rivers and streams (Compton et al. 2002). In New Hampshire 95% of captures and recaptures were recorded within 175 meters of water (Tuttle and

Carroll 2003). Virginia studies indicated that the maximum distance Turtles’ range from streams is ca. 650 meters (about 2,145 feet) (Akre and Ernst 2006). While a study conducted in Massachusetts and New Hampshire found 228 meters to be the

“distance representing the 75th percentile of all radio-equipped animal’s median distance traveled from water between July and August of all years.” (Jones 2009) In this same study 470 meters represented the 95th percentile median distance from water, with maximum distances being 634-932 meters (id.).

In Virginia, movement patterns at all three of Akre and Ernst’s (2006) study areas were found to be similar and consistent with other studies (e.g., Compton 1999,

Arvisais et al. 2002). At the agricultural site, 90% of all locations were within 250 m of the stream, while at the forested site, 95% of locations were within 300 m of the stream, and at the moderately altered forest site, 95% of locations were within 200 m of the stream.

These terrestrial zones that generally extend out to ca. 300 meters from waterways can be considered the “core habitat” (sensu Semlitsch and Jensen 2001,

Semlitsch and Bodie 2003) where conservation efforts for the species should be focused. Vermont recognizes that “the wood turtle uses streams and rivers for overwintering, and uses adjacent riparian areas up to 300 meters from the water’s edge for foraging, breeding, nesting, and dispersal.” (Vermont 4 – 68) And New Jersey

451 uses a 322-meter stream buffer to identify Wood Turtle habitat (NJ Landscape Project at http://www.njfishandwildlife.com/ensp/landscape/index.htm).

Individuals’ home range sizes vary throughout their range, with some evidence for those in the north to be larger (Arvisais et al. 2002). Home ranges are generally anchored on a creek, stream or river and may be elongate in shape as a result (Strang 1983, Kaufman 1995, Daigle 1997, Saumure 2004). The home ranges of males can tend to be more linear (in other words, include more stream length) than are those of females (Jones 2009). Home ranges tend to be small, but are highly variable among individuals (Quinn and Tate 1991, Daigle 1997, Arvisais et al. 2002,

Saumure 2004), both within a population and among populations inhabiting different landscape types (Akre and Ernst 2006). In this respect, Wood Turtles display an individual variability in home range size, activity, and habitat use similar to that seen in other forest-dwelling testudinids (cf., Moskovits and Kiester 1987, Lue and Chen

1999, Lawson 2006). Within a population, average home range sizes typically do not differ significantly between males and females, nor does home range size appear to be correlated to Turtle size or dominance (Ross et al. 1991, Kaufman 1995, Akre and Ernst, 2006, Remsberg et al. 2006, Sweeten 2008). In Massachusetts, Jones

(2009) found that older females had larger home ranges than younger females. See

Saumure 2004 for range differences based on primary productivity of forest vs. agri- forest settings.

Though sizes vary depending on the individual and local conditions, as well as from year to year, Turtles generally have limited home ranges: from 1-80 hectares,

452 with an average of perhaps 2-20 (see, e.g., Quinn and Tate 1991, Foscarini 1994,

Kaufmann 1995, Arvisais et al. 2002, Smith 2002, Tuttle and Carroll 2003, Akre and

Ernst 2006, Remsberg et al. 2006, Castellano 2007, Greaves 2008, Jones 2009). Too, it must be remembered that the methodologies used in different studies to derive home range estimates were not consistent (e.g., variation in the time periods used to calculate home range, the number of individual animal locations, and the range estimators employed, as well as the use of minimum convex polygon vs. adaptive kernel models) (detailed in Saumure 2004 and Jones 2009).

Home range size has also been shown fluctuate in response to environmental conditions. In Michigan, home range size increased during wet years. Wetter conditions may improve vegetation and adjunct watercourses, allowing Turtles to move further away from more permanent bodies of water. In contrast, periods of drought may restrict their movement and give the perception of smaller home ranges

(Remsberg et al. 2006).

Certainly the spatial distribution of resources affects home range size, structure, and location on the landscape (Heppell et al. 2000). So sizes may be related to productivity and habitat “quality”, with Turtles perhaps having to range farther in “poor” or fragmented habitat to satisfy food requirements or find mating opportunities (Kaufmann 1995). Or perhaps Turtles range farther in contiguous or

“good” habitat (e.g., those with a high availability of humid microclimates or food items) simply because they can, i.e., the opportunities for expansive roaming are present (Arvisais et al. 2002, Akre and Ernst 2006, Chybowski et al. 2008, Hamernick

453 2000). Constraints such as topography, which can limit or enhance access to resources, may certainly affect the sheer size of vertebrate home ranges (Strang 1983,

Powell and Mitchell 1998). For all these reasons, home range studies should span at least a period of several years.

Akre and Ernst (2006) studied Wood Turtles occupying three different habitat types: (A) agricultural, (B) forested, and (C) forested (moderately altered). Home range size was found not to be correlated with body size or sex; only at the moderately altered forest site was age found to be positively correlated with home range size.

Although home range size differed little between the three study areas, home range size at the forested site was found to be, on average, larger than study areas A and

C. This evidence supports the hypothesis that home range size in wood turtles may be reduced by habitat alteration and fragmentation (Arvisais et al. 2002). Of the three habitats studied, study area B had the least anthropogenically altered habitat.

Wood Turtles are strongly philopatric over extended time periods (Quinn and

Tate 1991, Kaufmann 1995, Ernst 2001a&b, Arvisais et al. 2002, Tuttle and Carroll

2003, Akre and Ernst 2006, Willoughby 2008, Jones 2009, Krichbaum 2009).

Individuals often occupy nearly the same home ranges over multiple years

(Kaufmann 1995, Tuttle and Carroll 2003, Arvisais et al. 2004, Akre and Ernst 2006).

There is substantial overlap of individual home ranges despite the apparent aggressive tendencies of males (Kaufmann 1995, Daigle 1997, Arvisais et al. 2002,

Tuttle and Carroll 2003, Akre and Ernst 2006). Though Wood Turtles may be the

454 most aggressive emydid yet studied, they are not demonstrably territorial (Kaufmann

1995).

Turtle density was found to be negatively correlated with the size of home ranges (Smith 2002). Differing parameters and definitions for the amounts of

“habitat” considered/used make the population density estimates derived by different researchers difficult to compare. Perhaps the most useful density estimates would be a standardized one that represents the number of adult turtles per linear unit of stream, as well as estimates based upon hectares of stream, calculated using measures of stream width (Jones 2009). There is evidence that population densities are lower at northern locations (Walde et al. 2003). Density may be positively correlated with the number of frost-free days, as well as be consistent with the general trend for lower site productivity at higher latitudes (Walde et al. 2003, Greaves &

Litzgus 2009). However, localized differences in this trend also occur. For instance, a northern mixed-deciduous site had higher productivity than a disturbed agri-forest site to the south (Saumure 2004). Geographic variation in Turtle densities may result from differences in recruitment rate, e.g., embryonic development may not occur in cold years at high latitudes (Walde 1998).

K. Primary diet

Feeding may be more common during the warmer months (Farrell and Graham

1991, Ernst 2001a, Arvisais et al. 2002), and less common during the mating seasons

(Brewster 1985, Niederberger and Seidel 1999). Optimal feeding conditions have

455 been documented when the water and air temperatures were equal to or above 17.2° and 23°C respectively (Ernst 1986).

Wood Turtles are omnivores. Foraging and ingestion occur in both terrestrial and aquatic settings, including underwater (Carroll 1999). Herbaceous plant leaves

(e.g., Viola), mushrooms (e.g., Boletus and Amanita), earthworms, fruits (e.g., raspberries Rubus), slugs, beetles, and millipedes figure prominently in researchers’ observations (see, e.g., Strang 1983, Kaufmann 1992, Niederberger and Seidel 1999,

Ernst 2001a, Compton et al. 2002, Walde et al. 2003, Jones 2009, Krichbaum 2009).

It may be that plants or animals that are infrequently encountered and ingested nonetheless supply important or even essential nutritional support beyond basic caloric needs.

Other foods consumed include Cinquefoil (Potentilla), blueberries

(Vaccinium), blackberries (Rubus), leaves and fruit of strawberries (Fragaria), fruit of

Skunk Cabbage (Symplocarpus), sorrel (Oxalis), leaves of Smilax, grasses, new growing tips of ferns, mollusks, moss, flowers, fungi, leaves of willow (Salix) and alder

(Alnus) and Paper Birch (Betula), algae, invertebrates (e.g., leeches, insects, snails, caterpillars), carrion, and tadpoles (Ernst et al. 1994).

Slugs (221 times) and other invertebrates (25), fungi (27), including the genera

Russula and Lactarius, the leaves of Jewelweed (Impatiens capensis) (30), and the fruit of Strawberry (Fragaria spp.) (13) were salient in Jones’ (2009) extensive feeding observations in Massachusetts. In New Hampshire, Carroll (1999) noted the Turtle’s propensity for feeding upon sallows (Willow shoots).

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

Thesis and Dissertation Services ! !