PRESCRIBED FIRE EFFECTS ON EASTERN BOX TURTLES IN SOUTHWESTERN
MICHIGAN
By
Tracy Ann Melvin

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Fisheries and Wildlife – Master of Science
2017

ABSTRACT
PRESCRIBED FIRE EFFECTS ON EASTERN BOX TURTLES IN SOUTHWESTERN
MICHIGAN
By
Tracy Ann Melvin
Limited information exists on eastern box turtles (Terrepene carolina carolina) and prescribed
fire. My thesis focuses on the effects of prescribed fire, nesting ecology, hatchling movements,
and efficacy of post-burn surveys for eastern box turtles in a 1226 ha recreation area in
Michigan. Using a combination of tracking techniques, I monitored 34 female and 6 male adult
eastern box turtles, and 58 hatchlings between 2013 and 2015. In Chapter 1, I summarized the
nesting behavior of females, resultant clutch sizes and success, and hatchling movements using
summary statistics and general linear mixed models to find that average clutch size was 6 (SE =
0.47), total clutch success ranged from 45.5% to 53.6% of the total clutch emerging, hatchling
movements ranged from 3.3 to 123.7 m within the first two weeks of emerging, and that
vegetation cover type had a significant effect on clutch success. In Chapter 2, I tested the
efficacy of post fire surveys for box turtles and found that average detection probability 48 hrs
after a fire was low (0.11) (SE = 0.05) and highly variable among surveyors (range = 0.00 –
0.50). I estimated that 26 hour-long surveys per ha would be required to reliably (95%
confidence) detect turtles that were present in burned areas. In Chapter 3, I observed the direct
behavioral effects of a growing season prescribed fire on radio tagged box turtles (n = 4).
Behaviors included burying and actively negotiating flame fronts. I documented 1 post fire
mortality. My results suggest prescribed fire should not be applied annually in grassland areas to
minimize hatchling mortality and slow moving, patchy growing season fires should be
considered in dry-mesic southern forests to minimize adult box turtle mortality.

Copyright by
TRACY ANN MELVIN
2017

ACKNOWLEDGEMENTS

Many people contributed their time, wisdom, expertise, equipment or emotional support to the
work described herein. Without these contributions, this work would have been impossible. I
would to thank the Michigan Society of Herpetologists, Joint Fire Science Program, National
Box Turtle Conservation Committee, Michigan State University, and MSU College of
Agriculture and Natural Resources Alumni Association for providing academic year funding. I
would like to thank members of the Stewardship Unit, Parks and Recreation Division of the
Michigan Department of Natural Resources for funding the equipment of this project and for
their professional support and expertise. I am perpetually appreciative of their time and effort.
This includes Glenn Palmgren and Bob Clancy for their plant identification skills, assistance in
my fire qualifications, endless fire communication, ordering equipment, ignition planning, and
serving as incredible liaisons between fire personnel and researchers. I would like to thank Heidi
Frei for her support in providing volunteers and additional research activities for research
technician enjoyment. Thank you to Ray Fahlsing for agreeing that this research was generally a
good idea and should be given funding. I would especially like to thank Alicia Ihnken for her
tremendous expertise in all things related to this project, expert guidance and idea generation,
general moral support, field support, technician support, GIS support, managing equipment
needs and occasionally buying me dinner and chocolate. I thank the staff of Fort Custer State
Recreation Area for allowing me access to the park and our "equipment shed" pit toilet. I have
worked with many MDNR technicians who were invaluable to this project and would like to
especially note the hard work of A. Beals, H. Tripp, H. Sission, S. Bristow, C. Burden, and M.
Bonnau. Thank you to committee members Jim Harding and Jessica Miesel for their support and

iv

expertise with prescribed fire and herpetofauna. I would like to thank my advisor, Dr. Gary
Roloff, for his ability to see the potential in his students, statistical expertise and scientific
understanding and for his patience and guidance. I am forever grateful to my family who taught
me how to follow my passions knowing that someone always has by back, especially my parents,
for always encouraging me no matter what I chose to do. Thanks mom (Joanne Swem), dad
(Gary Swem), brother (Trevor Swem), sister (Darcy Caswell), my Lansing family (Pat and Leo
Gundrum) and my Vicksburg family (Tim and Lynne Melvin, Nate and Angie Melvin, Aubrey
and Robbie Gray). Lastly, I am forever grateful to the love of my life, Peter Melvin, for
negotiating the rollercoaster of emotion that is a three-year M.S. program, for being my very best
field technician and friend, for really making me believe I can do anything, and for building all
those cages and the transmitter removal tool: the world's greatest invention.

v

TABLE OF CONTENTS

LIST OF TABLES

viii

LIST OF FIGURES

x

INTRODUCTION
LITERATURE CITED

1
4

CHAPTER 1
8
NESTING ECOLOGY, HATCHLING BEHAVIOR AND REPRODUCTIVE SUCCESS OF
EASTERN BOX TURTLES IN SOUTHWESTERN MICHIGAN
8
Abstract
8
1.1. Introduction
9
1.2. Methods
12
1.2.1. Study Area
12
1.2.2. Radio Telemetry of Adult Females
14
1.2.3. Nest Protection and Monitoring
15
1.2.4. Hatchling Movements
15
1.2.5. Data Analysis
16
1.3. Results
17
1.3.1. Nesting Behavior
17
1.3.2. Reproductive Female Mortality
18
1.3.3. Nest Sites
18
1.3.4. Nest Site Fidelity
19
1.3.5. Clutch Size, Incubation Period, and Clutch Success
19
1.3.6. Hatchling Movements, Behavior, and Known Sources of Mortality
21
1.4. Discussion
22
1.5. Management Implications
27
1.6. Acknowledgements
27
APPENDIX
28
LITERATURE CITED
36
CHAPTER 2
RELIABILITY OF POST-FIRE SURVEYS FOR EASTERN BOX TURTLES
Abstract
2.1. Introduction
2.2. Methods
2.2.1. Study Area
2.2.2 Research Burn and Survey Area Descriptions
vi

42
42
42
43
44
44
46

2.2.3. Radio-telemetry and Fire Observation of Eastern Box Turtles
2.2.4. Detection Probability Experimental Design
2.2.5. Data Analysis
2.3. Results
2.3.1. Fire Characteristics and Turtle Behavior
2.3.2. Detection Probability
2.4. Discussion
2.5. Management Implications
2.6. Acknowledgements
APPENDIX
LITERATURE CITED

46
47
48
48
48
49
50
52
53
54
58

CHAPTER 3
DIRECT AND INDIRECT EFFECTS OF A GROWING SEASON PRESCRIBED FIRE ON
EASTERN BOX TURTLES IN SOUTHWESTERN MICHIGAN
62
Abstract
62
3.1. Introduction
63
3.2. Methods
66
3.2.1. Study Area
66
3.2.2. Vegetation and Fire Data Collection
68
3.2.3. Radio-telemetry and Observation of Eastern Box Turtles During the
Burn
68
3.2.4. Data Analysis
69
3.3. Results
70
3.3.1. Fire Characteristics
70
3.3.2 Direct Effects of Fire on Woody Vegetation and Leaf Litter
70
3.3.3. Direct Fire Effects on Box Turtle Behavior
71
3.3.4. Indirect Fire Effects on Box Turtles
73
3.4. Discussion
74
3.5. Management Implications
78
3.6. Acknowledgements
78
APPENDIX
80
LITERATURE CITED
87
CONCLUSIONS

93

vii

LIST OF TABLES

Table 1.1.

Radio-tagged female eastern box turtles at Fort Custer State Recreation Area,
southern Michigan, and years they were successfully tracked to a nesting site
where eggs were laid, 2013-2015. Within a year, 1 = tracked to nesting site, 0 =
not tracked to nesting site, and “.” = unknown nesting status.
29

Table 1.2.

Eastern box turtle nests monitored in the Fort Custer State Recreation Area,
southwestern Michigan, 2013-2015, including vegetation type containing the nest,
date of nest initiation, clutch size, number of emerged hatchlings, and information
on hatchling movements.
30

Table 1.3.

Generalized linear mixed effects model results for yearly clutch size, number of
developed eggs, and number of emerged hatchlings of eastern box turtles at Fort
Custer State Recreation Area, southwestern Michigan, 2013-2015.
32

Table 2.1.

Number of post fire marked eastern box turtles (Terrapene carolina carolina)
available for detection (ni), visually detected by observers (mi), and detection
probability (βi) by survey and survey area, Fort Custer State Recreation Area,
southwestern Michigan, USA, 2015. Detection probability (βi) was calculated
using the marked subsample method (Lancia et al. 2005:121).
55

Table 3.1.

Pre- and post- fire average stem counts (SE) within 2 m radius circle plots (n=52)
by diameter size classes for non-native woody species. These plots carried an
early growing season (May 20, 2015) prescribed fire in a 31.6 ha management
unit at Fort Custer State Recreation Area, Michigan, USA. Pre-fire vegetation
sampling was conducted in June 2014, and plots were visually inspected in May
2015 to verify that major changes to the plant community (e.g., conversion) did
not occur. Post-fire vegetation sampling was completed in September, 2015.
81

Table 3.2.

Pre- and post- fire average stem counts (SE) within 2 m radius circle plots (n=52)
by diameter size classes for native, fire adapted woody species. These plots
carried an early growing season (May 20, 2015) prescribed fire in a 31.6 ha
management unit at Fort Custer State Recreation Area, Michigan, USA. Pre-fire
vegetation sampling was conducted in June 2014, and plots were visually
inspected in May 2015 to verify that major changes to the plant community (e.g.,
conversion) did not occur. Post-fire vegetation sampling was completed in
September, 2015.
82

Table 3.3.

Pre- and post- fire average stem counts (SE) within 2 m radius circle plots (n=52)
by diameter size classes for native, woody species not adapted to fire. These plots
carried an early growing season (May 20, 2015) prescribed fire in a 31.6 ha
management unit at Fort Custer State Recreation Area, Michigan, USA. Pre-fire
vegetation sampling was conducted in June 2014, and plots were visually
viii

inspected in May 2015 to verify that major changes to the plant community (e.g.,
conversion) did not occur. Post-fire vegetation sampling was completed in
September, 2015.
83

ix

LIST OF FIGURES

Figure 1.1.

Management unit boundaries used for studying nesting ecology of eastern box
turtles in of Fort Custer State Recreation Area, southwestern Michigan.
Boundaries are portrayed in bold.
33

Figure 1.2.

Predator exclosures used at Fort Custer State Recreation Area to protect eastern
box turtle nests. Exclosures were 0.61 x 0.61 x 0.31 m bottomless wooden frames
wrapped in 0.6 cm wire mesh, preserved and camouflaged with an acrylic-based
solid stain in olive drab green. Exclosures featured a removable lid attached with
(4) 3 cm deck screws. Exclosures were dug 5 cm into the soil surface around the
nest.
34

Figure 1.3.

Five nesting sites by vegetation type in Fort Custer State Recreation Area,
Michigan, USA. A = big bluestem (Andropogon gerardii) dominated tallgrass
prairie in active restoration, B = herbicide applied area of big bluestem
(Andropogon gerardii) dominated tallgrass prairie in active restoration, C = forb
and lichen dominated field, D = agricultural field (corn), E = grass and forb
dominated abandoned gravel pit.
35

Figure 2.1.

Management unit (yellow polygon) at Fort Custer State Recreation Area in
southwestern Michigan, USA, 2015. Post fire visual-encounter survey areas A
(0.75 ha) and B (1.0 ha) delineated by white polygons.
56

Figure 2.2.

Post fire concealment of eastern box turtles (Terrapene carolina carolina)
monitored using radio telemetry at Fort Custer State Recreation Area in
southwestern Michigan, USA, 2015. A = 20% concealed under logs in burned
area, B = 80% concealed, 20 m into unburned wetland, C = 10% concealed at the
edge of unburned wetland, D = 5% concealed leaf litter and live vegetation in
unburned patch of oak and black cherry (Prunus serotina).
57

Figure 3.1.

A 31.6 ha management unit consisting of a dense patch of invasive black locust
(A), a 12.5 ha emergent marsh (B), dry-mesic southern (oak-hickory) forest (C),
open grassland (D), red pine plantation (E), and mixed lowland hardwoods (mesic
southern forest, maple-beech dominated) (F) at Fort Custer State Recreation Area,
Michigan, USA.
85

Figure 3.2.

An example of a set of 3 Omega TL-10 adhesive, non-reversible labels that
indicated temperature (OMEGA Engineering, INC., Stamford, Connecticut)
ranging from 87 – 260 ° C. Labels were affixed to a rectangle (11 cm x 20 cm) of
aluminum roof flashing. The rectangle was bent to a 90-degree angle and fixed
into the soil so that the side with the temperature labels ran parallel to and 4 cm
above the soil surface. This set of labels indicated temperatures reached 88° C.
86
x

INTRODUCTION

The Eastern Box Turtle, Terrapene carolina carolina, is a small terrestrial emydid turtle that
ranges from New England to the southern Great Lakes region south to the Carolinas and
Georgia; traditionally it is considered to intergrade with additional subspecies to the west and
south. A recent genetic study (Martin et al. 2013) suggested elevating the subspecies of
Terrapene carolina to full species, a scheme adapted by Powell et al. 2016. Although relatively
few long-term population studies have been completed, it is generally accepted that eastern box
turtle populations are in decline (Williams and Parker 1987; Stickel and Bunck 1989; Schwartz
and Schwartz 1991; Hall et al. 1999; Nickerson and Pitt 2012). Substantial declines are most
likely the result of anthropogenic stressors including habitat loss and fragmentation, road
mortality, collection for the pet trade, nest predation, and disease (Stickel 1978; Williams and
Parker 1987; Warwick 1993; Belzer 1997; Hall et al. 1999; Gibbons et al. 2000; Dodd 2001).
These anthropogenic stressors are exacerbated by delayed sexual maturity, relatively low clutch
sizes, and potentially high hatchling mortality rates (Hall et al. 1999). As with many long-lived
ectotherms, eastern box turtles exhibit delayed sexual maturity and extended longevity, with
individuals becoming sexually mature at 5-10 years and living beyond 50 years in the wild
(Congdon et al. 1993; Dodd 2001). Hence, eastern box turtles are susceptible to continuous
population declines if subjected to any short- or long-term disturbances that negatively affect
breeding adult survival, nest success, or hatchling survival; changes that can go unnoticed due to
their longevity (Madden 1975; Klemens 1989; Doroff and Keith 1990; Dodd 2001, Ernst and
Lovich 2009).

1

Prescribed fire can be an effective and inexpensive tool for vegetation management
(Knapp et al. 2009), but the direct (including injury and mortality) and indirect effects on
wildlife (including changes in prey, body condition and animal movements or interactions) can
be difficult to quantify. Fire effects on reptiles and amphibians have been studied and
summarized (e.g., Keyser et al. 2004; Russell et al. 2009; Roloff and Hmielowski 2014;
O’Donnell et al. 2015), and the literature generally suggests that prescribed fire has few
significant direct effects. However, most of these historical studies relied on animal counts or
mark-recapture data that were uncorrected for detection probability (Mazerolle et al. 2007).
Furthermore, reptiles and amphibians are often grouped into herpetofauna for fire studies (e.g.,
Means and Campbell 1981; Smith 2000; Floyd et al. 2002; Renken 2005), potentially biasing
results to those species that are readily observable and abundant. For cryptic, mobile, or rare
species like eastern box turtles, the error associated with failure to detect when present may have
substantial conservation ramifications (Gu and Swihart 2004; Refsnider et al. 2011). Research
suggests that losing even small numbers of breeding adults from k-selected, long-lived
populations like box turtles can result in irreparable harm to population viability (Congdon et al.
1993; Dodd et al. 2015). Although eastern box turtles may have evolved with fire, current turtle
population sizes, fire regimes, and habitat configurations differ from historical conditions and
frequently repeated prescribed fires may result in long-term population declines. The direct
effects of prescribed fire have been described in past studies with observations of high rates of
mortality or injury from exposure to fire (Allard 1949; Babbitt and Babbitt 1951; Bigham et al.
1964; Dolbeer 1969, Gibson 2009).
Indirect effects of prescribed fire may include effects on prey abundance, vegetation used
for thermoregulation, camouflage, and insulating duff layers during overwintering, and potential

2

disruption in behavior or range (Gibson 2009). Box turtles are particularly sensitive to
environmental variables that affect ground cover (including vegetation, litter, and subsurface
soils) because their entire life history depends on ground conditions (Dodd 2001).
My thesis focuses on the direct and indirect effects of prescribed fire, nesting ecology and
hatchling movements, and the efficacy of post-burn surveys on eastern box turtles in a 1226 ha
recreation area in Michigan. Using a combination of tracking techniques, I monitored 34 female
adult eastern box turtles, 6 males, and 58 hatchlings between the fall of 2013 and spring of 2015.
In Chapter 1, I summarized the nesting behavior of females, resultant clutch sizes and success,
and hatchling movements using summary statistics and general linear mixed models. In Chapter
2, I tested the efficacy of post fire surveys for box turtles. In Chapter 3, I observed the direct
behavioral effects of a prescribed fire on box turtles. Behaviors included burying and negotiating
flame fronts. I also documented the short-term effects of growing season fires on woody
vegetation and litter. This study is one of the first to test the efficacy of growing season
prescribed fires on both vegetation and eastern box turtle behavior and survival. Results of this
research provide insight on eastern box turtle behavior when using prescribed fire and can inform
management practices.

3

LITERATURE CITED

4

LITERATURE CITED

Allard, H. A. 1949. The eastern box turtle and its behavior. Journal of the Tennessee
Academy of Science 24:146–152.
Babbitt, L. H., and C. H. Babbitt. 1951. A herpetological study of burned-over areas in
Dade County, Florida. Copeia 1951:79.
Belzer, W. 1997. Box turtle conservation issues. Reptile and Amphibian Magazine 49:32-35.
Bigham, S. R., J. L. Hepworth, and R. P. Martin. 1964. A casualty count of wildlife following
a fire. Proceedings of the Oklahoma Academy of Science 45:47–50.
Congden, J. D., A. E. Dunham, and R. C. Van Loben Sels. 1993. Delayed sexual maturity and
demographics of Blanding’s turtles (Emydoidea blandingi): implications for
conservation and management of long-lived organisms. Conservation Biology 7:826833.
Dodd, C. K., Jr., 2001. North American box turtles: a natural history. University of Oklahoma
Press, Norman, USA.
Dodd, C. K., Jr., V. Rolland, and M. K. Oli. 2015. Consequences of individual removal on
persistence of a protected population of long-lived turtles. Animal Conservation
19(4):369-379.
Dolbeer, R. A. 1969. Population density and home-range size of the eastern Box Turtle
(Terrapene c. carolina) in eastern Tennessee. ASB Bulletin 16:49.
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.
Ernst, C. H., and J. E. Lovich. 2009. Turtles of the United States and Canada. Johns Hopkins
University Press, Baltimore, MD, USA.
Floyd, T. M., K. R. Russell, C. E. Moorman, D. H. Van Lear, D. C. Guynn Jr., and J. D. Lanham.
2002. Effects of prescribed fire on herpetofauna within hardwood forests of the Upper
Piedmont of South Carolina: a preliminary analysis. U.S. Department of Agriculture,
Forest Service, Southern Research Station, Ashville, North Carolina, General
Technical Report SRS–48.
Gibbons, J. W., D. E. Scott., T. J. Ryan, K. A. Buhlmann, T. D. Tuberville, B. S. Metts, J. L.
Greene, T. Mills, Y. Leiden, S. Poppy, and C. T. Winne. 2000. The global decline of
reptiles, dĂŠjĂ  vu amphibians. Bioscience. 50(8):653-666.

5

Gibson, J. 2009. Influence of prescribed fire on a Midwestern population of the eastern box
turtle, Terrapene c. carolina. Thesis, Purdue University, Indiana, USA.
Gu, W., and R. K. Swihart. 2004. Absent or undetected? Effects of non-detection of species
occurrence on wildlife-habitat models. Biological Conservation 116:195–203.
Hall, R. J., P. F. Henry, and C. M. Bunck. 1999. Fifty-year trends in a box turtle population in
Maryland. Biological Conservation 88:165-172.
Keyser, P. D., D. J. Sausviller, W. M. Ford, D. J. Schwab, and P. H. Brose. 2004. Prescribed fire
impacts to amphibians and reptiles in shelterwood-harvested oak-dominated forests.
Virginia Journal of Science 55(4):159-168.
Klemens, M. W. 1989. The methodology of conservation. Pages 1-4 in I. R. Swingland and
M.W. Klemens, editors. The Conservation Biology of Tortoises. Occasional Papers
IUCN/SSC 5.
Knapp, E. E., B. L. Estes, C. N. Skinner. 2009. Ecological effects of prescribed fire season: a
literature review and synthesis for managers. U.S. Department of Agriculture, Forest
Service, Pacific Southwest Research Station, Albany, California, General Technical
Report PSW-GTR-224.
Madden, R. C. 1975. Home range, movements, and orientation in the eastern box turtle
Terrapene carolina carolina. Dissertation, City University of New York, New York,
USA.
Martin, B. T., N. P. Bernstein, R. D. Birkhead, J. F. Koukl, S. M. Mussmann, and J. S. Placyk.
2013. Sequence-based molecular phylogenetics and phylogeography of the American
box turtles (Terrapene spp.) with support from DNA barcoding. Molecular
Phylogenetics and Evolution 68:119–134.
Mazarolle, M. J., L. L. Bailey, W. L. Kendall, J. A. Royle, S. J. Converse, and J. D. Nichols.
2007. Making great leaps forward: accounting for detectability in herpetological field
studies. Journal of Herpetology 41:672–689.
Means, D. B., and H. W. Campbell. 1981. Effects of prescribed burning on amphibians and
reptiles. Pages 89-97 in G. W. Wood, editor. Prescribed fire and wildlife in southern
forests. The Belle W. Baruch Forest Science Institute, Clemson University,
Georgetown, South Carolina, USA.
Nickerson, M. A., and A. L. Pitt. 2012. Historical turtle population decline and community
changes in an Ozark River. Bulletin of the Florida Museum of Natural History 51(4):
257–267.

6

O’Donnell, K. L., F. R. Thompson III, and R. D. Semlitsch. 2015. Prescribed fire and timber
harvest effects on terrestrial salamander abundance, detectability, and microhabitat
use. Journal of Wildlife Management 79(5):766-775.
Powell, R., R. Conant, and J. T. Collins. 2016. Peterson field guide to reptiles and amphibians
of eastern and central North America. Fourth edition. Houghton Mifflin Harcourt,
New York, New York, USA.
Refsnider, J. M., T. M. Mitchell, H. M. Streby, J. T. Strickland, D. A. Warner, F. J. Janzen. 2011.
A generalized method to determine detectability of rare and cryptic species using the
ornate box turtle as a model. Wildlife Society Bulletin 35(2):93-100.
Renken, R. 2005. Does fire affect amphibians and reptiles in eastern U.S. oak forests? Pages
158–166 in Proceedings of Fire in Eastern Oak Forests: Delivering Science to Land
Managers. Proceedings of a Conference, 15 November–18 November 2005,
Columbus, Ohio, USA.
Roloff, G. J. and T. L. Hmielowski. 2014. Ecological effects of fire in Great Lakes savannas and
prairies: literature review. Tallgrass Prairie Publication TPOS LR01-2014.
Russell, R. E., J. A. Royle, V. A. Saab, J. F. Lehmkuhl, W. M. Block, and J. R. Sauer. 2009.
Modeling the effects of environmental disturbance on wildlife communities: avian
responses to prescribed fire. Ecological Applications 19(5):1253-1263.
Schwartz, E. R., and C. W. Schwartz. 1991. A quarter-century study of survivorship in a
population of three-toed box turtles in Missouri. Copeia 1991:1120-1123.
Smith, J. K., editor. 2000. Wildland fire in ecosystems: effects of fire on fauna. U.S. Department
of Agriculture, Forest Service, Rocky Mountain Research Station, Ogden, Utah,
General Technical Report RS-GTR-42-vol. 1.
Stickel, L. F. 1978. Changes in a box turtle population during three decades. Copeia 1978: 221–
225.
Stickel, L. F., and C. M. Bunck. 1989. Growth and morphometrics of the box turtle,
Terrapene c. carolina. Journal of Herpetology 23: 216–223.
Warwick, C. 1993. The decline of North American box turtles. New York Turtle and Tortoise
Society Newsnotes 4(1):15-16.
Williams, E. C., and W. S. Parker. 1987. A long-term study of a box turtle (Terrapene carolina)
population at Allee Memorial Woods, Indiana, with emphasis on survivorship.
Herpetology 43: 328–335.

7

CHAPTER 1
NESTING ECOLOGY, HATCHLING BEHAVIOR AND REPRODUCTIVE SUCCESS
OF EASTERN BOX TURTLES IN SOUTHWESTERN MICHIGAN

Abstract
Limited information exists on local productivity and recruitment of eastern box turtles
(Terrapene carolina carolina), a species of special concern, in landscapes heavily impacted by
humans. I documented the nesting behavior, nesting success, hatchling behavior, and known
sources of mortality for box turtles at Fort Custer State Recreation Area in Augusta, MI, from
2013 – 2015. I used radio telemetry on adult females (n=34), trailing string and transmitters on
hatchlings (n=58), and field surveys of nests that were protected from predation (n=25). I
modeled the relationships between: 1) clutch size and physical characteristics of the
corresponding female turtle, 2) number of eggs that developed into hatchlings as a function of
year and habitat that contained the nest, and 3) number of hatchlings that emerged from the nest
by year and habitat. I found that females staged for nesting in late May, and I observed nesting
between 2 to 14 June. Turtles nested in a wide variety of habitats, including an abandoned gravel
pit, grassland restoration sites, an active disk golf course, and agricultural fields. At the time of
nesting, all of these habitats provided suitable nesting substrate with an open vegetation canopy.
By the time that hatchlings started to emerge in the fall, some of these habitats (e.g., agricultural
fields) differed substantially in vegetation cover, likely affecting hatchling success. Female
turtles tended to demonstrate fidelity to nesting habitat patches, but distances to nest locations
within a patch varied annually (average = 215.4 m, SE = 61.0, range = 11.7 - 411.4 m). Average
clutch size was 6.0 (SE = 0.47), and I found that ≥1 hatchling emerged from the nest cavity for

8

15 of 22 nests (68%). Hatching success (i.e., hatchlings emerged/clutch size) for nests protected
from predation ranged from 46% to 54%. After fall emergence from the nest cavity and removal
of predator exclosures, hatchlings moved an average of 1.22 (SE = 0.15) m/day, with minimum
and maximum straight line distances from the nest cavity of 3.3 and 123.7 m, respectively, to
their overwintering sites. I failed to find a significant effect of female physical characteristics on
clutch size, and I found no effect of year or habitat on the number of eggs that developed into
hatchlings. I found a significant habitat effect on the number of hatchlings that emerged from the
nest cavity, with a gravel pit producing significantly more emergent hatchlings than agricultural
fields (β = 6.208, t = 2.76, p = 0.03). Although female box turtles used agricultural fields for
nesting, those nests failed to produce any emergent hatchlings suggesting that these fields serve
as population sinks. My modeling results indicated that female physical characteristics were not
reliable predictors of clutch size, and that habitat factors were the primary determinant of
hatchling recruitment to the turtle population. Results from my study indicated that high quality
nesting sites are an important recruitment factor that managers can directly address. High quality
nest sites provide low mortality risk to adult females, conditions (e.g., soils, vegetation, solar
radiation) that allow eggs to develop into hatchlings, and hiding cover for emerged hatchlings.
Furthermore, after fall emergence from nests, my results indicated that hatchlings were
vulnerable to land management practices that affect open nesting areas into the following spring.
1.1. Introduction
Eastern box turtles (Terrapene carolina carolina) were historically common throughout the
eastern United States but may now be locally extirpated (Dodd 2001). Although few long-term
population studies have been completed, it is generally accepted that eastern box turtle
populations are in decline range-wide (Williams and Parker 1987; Stickel and Bunck 1989;

9

Schwartz and Schwartz 1991; Hall et al. 1999; Nickerson and Pitt 2012). Substantial declines are
most likely the result of anthropogenic stressors including habitat loss and fragmentation, road
mortality, collection for the pet trade, nest predation, and disease (Stickel 1978; Williams and
Parker 1987; Warwick 1993; Belzer 1997; Hall et al. 1999; Gibbons et al. 2000; Dodd 2001).
Effects of these anthropogenic stressors on box turtle populations are further exacerbated by
delayed sexual maturity, relatively low clutch sizes, and potentially high hatchling mortality rates
(Hall et al. 1999).
As with many long-lived ectotherms, eastern box turtles exhibit delayed sexual maturity
and extended longevity, with individuals becoming sexually mature at 5-10 years and living
beyond 50 years in the wild (Congdon et al. 1993; Dodd 2001). Hence, eastern box turtles are
susceptible to continuous population declines if subjected to disturbances that negatively affect
breeding adult survival, as nest success or hatchling survival is generally considered low
(Madden 1975; Klemens 1989; Doroff and Keith 1990; Dodd 2001). During the incubation
stage, low reproductive success is often linked to predation by skunks (Mephitis mephitis), foxes
(Urocyon cinereoargenteus, Vulpes vulpes), raccoons (Procyon lotor), crows (Corvus spp.),
snakes (Heterodon spp., Lampropeltis spp.), and ants (Ernst et al. 1994; Dodd 2001; Flitz and
Mullin 2006). Changes in weather can also result in nest failures (Hallgren-Scaffidi 1986;
Brooks et al. 1991; Ernst et al. 1994; Belzer 2002). From hatchling emergence to shell
ossification, eastern box turtles are subjected to predation by a variety of insects, mammals,
amphibians and birds (Madden 1975; Ernst et. al. 1994; Dodd 2001; Belzer 2002), including
crows, vultures (Cathartes aura), barn owls (Tyto alba), shrews (Blarina spp.), ground squirrels
(Ictidomys spp.) and eastern chipmunks (Tamias striatus) (Ernst et al. 1994; Harding 1997;
Belzer et al. 2002). Studies suggest that to maintain stable populations, eastern box turtles rely on

10

high survivorship of reproductive-aged adults (Klemens 1989, 2000; Congdon and Gibbons
1990; Doroff and Keith 1990; Congdon et al. 1993; Hall et al. 1999; Dodd 2001), assuming that
at least some of the hatchlings will periodically recruit into the breeding population. However,
with exceptionally high predation rates on eggs in nests, and land management practices that
potentially result in hatchling mortality, concerns over recruitment of young are justified.
Although landscape alteration is arguably the most critical factor in turtle population
declines across North America (Mitchell and Klemens 2000; Dodd 2001), information is lacking
on the recruitment of eastern box turtles in areas subjected to intense human activity (Doroff and
Keith 1990; Kipp 2003; Budischak et al. 2006), like that found in parks or recreational areas.
Human activity can exacerbate negative stressors on local turtle populations by increasing road
or trail mortality, turtle collection, exposure to pollution, the presence of subsidized predators,
and mortality from household pets (Wilcox and Murphy 1985; Williams and Parker 1987; Dodd
et al. 1989; Belzer and Steisslinger 1999; Mitchell and Klemens 2000). Furthermore, the
maintenance and ecological restoration activities of certain parks and recreational areas may pose
threats to extant box turtle populations through mowing, use of industrial equipment, herbicide
application and prescribed fire.
Although box turtle nesting ecology has been studied across North America (e.g., Allard
1935; Kipp 2003; Wilson and Ernst 2005; Flitz and Mullin 2006; Burke and Capitano 2011a;
Willey and Sievert 2012), limited information exists on nest success and hatchling movements
even though that information is critical to conservation (Burke et al. 2000; Kipp 2003;
Rosenburg and Swift 2013). Nesting ecology studies have mainly focused on substrates, clutch
sizes and nesting times (Allard 1935; Kipp 2003; Wilson and Ernst 2005; Flitz and Mullin 2006;
Burke and Capitano 2011a). Studies on turtle hatchlings have generally been limited to

11

describing sizes, physiological factors (Dinkelacker et al. 2005), and aspects of overwintering
(Breitenbach et al. 1984; Ultsch et al. 2007; Burke and Capitano 2011a, b). These studies have
improved our understanding of terrestrial or semi-terrestrial turtle neonate ecology and
movements (Butler and Graham 1995; Keller et al. 1997; Forsythe et al. 2004), yet multi-year
studies that directly link nesting ecology or hatchling movements and behavior to habitat
management are rare.
My goal was to better understand potential recruitment vulnerabilities of eastern box
turtles in a managed recreation area of southwest Michigan. I used a combination of radio
telemetry on adult female box turtles, trailing string and telemetry on hatchlings, and field
surveys of predator protected nests to determine nesting dates, nest locations, clutch sizes and
success rates, and hatchling movements during summers of 2013 - 2015. My objectives were to:
1) document when and where adult female eastern box turtles nested, 2) evaluate nest success, 3)
quantify short-term hatchling movements, and 4) identify factors influencing productivity and
recruitment. In the absence of nest predation, my results provide insight into the factors that
potentially limit eastern box turtle populations in recreation areas that are intensively managed
for restoration and used heavily by the public.
1.2. Methods
1.2.1. Study Area
This study was conducted within the boundaries of Fort Custer State Recreation Area (FCSRA;
1,226 ha), located in the southern lower peninsula of Michigan (Kalamazoo and Calhoun
Counties), USA. This area is characterized by heterogeneous vegetation patterns from
topographic, climate, and human influences including high levels of agricultural and urban
development (Albert 1995; Eagle et al. 2005). Most of the soils are calcareous and loamy,

12

derived from underlying limestone, shale, and sandstone bedrock (Eagle et al. 2005). Glacial till
deposits are primarily loams, silt loams and clay loams (Eagle et al. 2005). The climate is
considered humid continental, although FCSRA is strongly influence by the Maritime Tropical
air mass and proximity of Lake Michigan, resulting in a warmer climate, moderated inland
temperature fluctuations, and induced lake-effect snow (Eichenlaub 1979; Denton 1985; Albert
et al. 1986; Eichenlaub et al. 1990). The average length of the growing season was 154 days
(Albert et al. 1986). Mean monthly (May to October) minimum and maximum temperatures
during the study ranged from -1.7 – 7.2° C and 22.2 – 34.4° C, respectively.
Southwestern Michigan was historically dominated by fire-dependent oak (Quercus spp.)
savanna and prairie (Albert 1995), and contained the only extensive areas of mesic prairie found
in Michigan (Kost 2004). In the early 1800’s, plant communities included dry and dry-mesic
southern forest (oak-hickory (Carya spp.)), oak barrens (mixed oak savanna), emergent marsh,
and southern (mixed hardwood) swamp (Palmgren 2004). Currently, intensive agriculture is
concentrated in this region because of the relatively mild climate and productive soils (Eagle et
al. 2005). Natural vegetation in this region is broadly classified as black oak-white oak (Q.
velutina-Q. alba) savannas and forests, and beech-sugar maple (Fagus grandifolia-Acer
saccharum) forests (Palmgren 2004). Currently, degraded patches of oak barrens and prairie
openings, scattered oak-hickory forests, and large blocks of non-native black locust (Robinia
pseudoacacia) patches occur on FCSRA (Palmgren 2004).
I chose FCSRA for this project because it has been the focus of ecological restoration
activities since 1997. FCSRA is located within the most northerly range of the eastern box turtle.
My study area was centered on Management Units 2, 4, 9, 14, 17 and 18 (Figure 1.1). The
dominant restoration management regime in grassland areas was medium-scale (< 111 ha) spring

13

or fall prescribed burning (2-year intervals), followed by herbicide treatments in some areas. A
large area (52 ha) of mature black locust experienced windthrow in 2001 (Figure 1.1, Palmgren
2004). This area was in active restoration to open barrens and prairie during my study.
1.2.2. Radio Telemetry of Adult Females
I used visual encounter surveys and turtle detector dogs to find adult eastern box turtles in May
of 2012 and June of 2013. Surveys were conducted during daylight hours, when box turtles are
known to be active. Captured adult box turtles found during these surveys (and others found
fortuitously during routine radio telemetry) were radio-tagged and tracked using Holohil R1-2B
14.5g Transmitters (Holohil Systems Ltd., Carp, Ontario, Canada) attached to the right or left
anterior pleural carapacial scutes using multi-purpose 5-minute set epoxy putty (LoctiteÂŽ,
Henkel 289 Corporation, Cary, NC, USA). Minimum age was estimated by counting annual
rings on the carapace (Legler 1960). Turtles were also individually marked using the Cagle
method (Cagle 1939). Radio-tagged adult female box turtles were located 1-3 times per week
during the active seasons (April through October) from May 2012 to August of 2015 using
portable telemetry receivers (Advanced Telemetry Systems, Isanti, MN, USA). During active
nesting periods (mid-May through June), females were located nightly between the hours of
1800 and 2200 to determine nesting status. If a female was found to be alert after 1900, she was
subsequently checked for digging behavior until 2200. Once a female was found to be digging,
her location was marked using a GPS. The site was left alone until morning when sites were then
checked for egg deposition by gently digging into the soil until the surface of at least one egg
was observed. Turtle capture and handling was conducted by Michigan Department of Natural
Resources (MDNR) employees or volunteers and followed MDNR protocols for animal welfare.

14

Additionally, all animals were handled in compliance with Michigan State University’s
Institutional Animal Care and Use Committee guidelines (IACUC Approval # 04/14-077-00).
1.2.3. Nest Protection and Monitoring
Within 24-72 hrs of initiation of female nesting activity, box turtle nests were covered with 0.61
x 0.61 x 0.31 m bottomless wooden-frames wrapped in 0.64 cm wire mesh (Figure 1.2). These
exclosures were preserved and camouflaged with an acrylic-based solid stain in olive drab green
and featured a removable lid attached with 4 - 3.2 cm deck screws. Exclosures were dug 5 cm
into the soil surface around the nest. Nests were checked daily starting August 1st when 152 x 91
x 22 mm sponges were placed in the corner farthest from the nest cavity and wetted daily to
provide moisture for the hatchlings upon emergence. Emergence was determined by
identification of an emergence hole near the nest cavity. After emergence in 2013 and 2014,
nests were excavated to count eggs through reconstruction of eggshell evidence. Evidence of
predation included egg shells scattered outside of the cavity, cavities that were disturbed by
digging, or an abundance of ants on the eggs. I calculated the number of developed hatchings and
emergence success and, if feasible, determined causes of hatching and emergence failure. If nests
did not emerge in the fall, they were left caged and excavated in June of the following year.
Emergence monitoring was paused from November – April in these cases.
1.2.4. Hatchling Movements
After emergence, hatchlings were weighed using a Micro-Line 10 g x 0.1 g spring scale
(PesolaÂŽ, Baar, Switzerland), and individually marked with a notch on a marginal scute with a
nail clipper (Cagle 1939). Each hatchling was fitted with 13 cm long trailing string of orange fly
line backing (RIOÂŽ Products, Idaho Falls, ID) by threading it through the left or right 11th
marginal scute with a sterilized hand needle and tying with an improved clinch knot (Breder

15

1927; Stickel 1950). A subset (n = 7) of hatchlings found in 2014 was also fitted with 0.52 g BD2X transmitters (Holohil Systems Ltd., Carp, Ontario, Canada) attached to the carapace with
silicone aquarium sealant (MarinelandÂŽ, Spectrum Brands, Blacksburg, VA). We released
hatchlings within 0.5 m of each nest cavity after they were weighed and measured. Hatchlings
were relocated after dusk using black lights (for the string) and telemetry, every 24-48 hrs.
Hatchling locations were recorded using a handheld GPS. Burying behavior into mineral soil was
recorded, along with vegetation type, percent concealment, and type of concealment used.
1.2.5. Data Analysis
I calculated summary statistics for nest site fidelity, clutch sizes, incubation period, number of
emerged hatchlings per nest, and hatchling movements. I used a nonparametric Kruskal-Wallis
test to determine if nest initiation date, clutch sizes, clutch success (emerged hatchling/egg), or
incubation days varied by year (Corder and Foreman 2009), and a Dunn test for multiple
comparisons (Benjamini-Hochberg 1995; adjustment to the p-value) to identify which years
differed. Zar (2010) noted that the Dunn test is appropriate for groups with unequal numbers of
observations.
I used generalized linear mixed effects models (GLM) with a Poisson distribution in
program R 3.0.2 (R Development Core Team 2014) to estimate the effects of independent
covariates on clutch size, the number of developed eggs, and the number of emerged hatchlings.
Independent covariates for the clutch size model included female mass (g) taken after she laid
eggs in all but two cases (β1), minimum female age (β2), whether the female showed signs of fire
scarring (β3), and year (β4). Independent covariates for the number of developed eggs model
included year (β1) and vegetation type that contained the nest (β2). Vegetation types were
modeled as factors and included tallgrass prairie, forb-dominated prairie, corn fields, and an

16

abandoned gravel pit (Figure 1.3). Independent covariates for the number of emerged hatchlings
model included year (β1), vegetation type (β2), and clutch size (β3). Individual female turtles
were treated as random effects in all models.
1.3. Results
1.3.1. Nesting Behavior
I monitored 34 female eastern box turtles from 2013 – 2015 (Table 1.1). Fifteen of those were
monitored for 3 years, and 18 for 2 years (Table 1.1). In 2013, 8 of 31 (26%) turtles were tracked
to their nesting site and produced a clutch. I found nests and documented clutches for 10 of 19
(53%) and 5 of 17 (29%) females in 2014 and 2015, respectively (Table 1.1). I caution that these
results should not be viewed as a measure of clutches per female in any given year because there
were two possible outcomes for females with unknown nesting status: 1) they truly did not nest,
or 2) they nested but the activity went undetected by the research team.
Radio-tagged female box turtles were not observed staging earlier than 29 May in 2014
and 2015. Staging is the process of moving to an area with open vegetation and exposed mineral
soil to begin searching for and subsequently digging a nest cavity to lay eggs. Nesting follows
staging, and I observed nesting between 2 to 14 June. The initiation of nesting varied by year
(Kruskal-Wallis test, χ2 = 13.02, df = 2, p = 0.001), with nesting occurring earlier in 2013
compared to other years (Dunn test, 2013 -2014, z = 3.39, p = 0.001; 2013 – 2015, z = 2.51, p =
0.009; 2014 – 2015, z = -0.262, p = 0.40). All nesting behaviors (e.g., digging, laying eggs,
covering eggs) were observed between 1830 and 0700. Females would finish laying eggs in a
single night.
In addition to the 23 nests with clutches that I found using telemetry (Table 1.2), I also
encountered 2 untagged females while they were nesting. These opportunistic females were not

17

measured or marked but I placed an exclosure over their nests. Hence, I monitored 25 nests from
2013-2015. I lost 3 of those nests to predation. The predator exclosure protecting one nest was
removed (likely by a human) and this nest was subsequently depredated. An additional nest was
partially depredated by a predator that tunneled under the exclosure, and another nest was
predated before the exclosure was deployed (<24 hrs after the female finished). Another nest was
lost when the female left before covering her eggs, all of which were inside an open nest cavity
when checked 24 hrs later, and presumed non-viable. By deploying predator exclosures 24 to 36
hrs after females completed nesting, I was able to monitor 84% of the potential nests that I found
for hatchling development and emergence.
1.3.2. Reproductive Female Mortality
I documented mortality for 2 of 34 (6%) radio-tagged adult females during the 3 years of
my study. One adult female may have died as a result of long-term infection during the course of
this study. She was staging to nest in a tallgrass prairie during an herbicide application during
May 2014. For the remainder of the active season, I noted varying levels of swollen eyes and eye
discharge. I found this female dead in early March 2015; she had failed to bury into mineral soil
during the winter. The other adult female appeared to die as a result of injuries sustained from a
prescribed fire.
1.3.3. Nest Sites
Nest sites (n=25) were dispersed throughout FCSRA in 7 areas representing different vegetation
cover types. Four nests were located in tallgrass prairie dominated by big bluestem (Andropogon
gerardii) that was currently undergoing active restoration (Table 1.2). One of these nests was in
an area of recent herbicide application with no live vegetation. Four nests occurred in a corn
field, 2 in a forb-lichen dominated radio-antenna field, 5 in a grass and forb dominated

18

abandoned gravel pit, 2 in mowed areas of a disc golf course, and 8 in a different active prairie
restoration site that was co-dominated by big bluestem (Andropogon gerardii), spotted knapweed
(Centaurea maculosa) and common mullen (Verbascum thapsus) with autumn olive (Elaeagnus
umbellata) and staghorn sumac (Rhus typhina) dispersed throughout. All nests were in loose
sandy soil.
1.3.4. Nest Site Fidelity
One turtle was successfully tracked to her nesting sites in all 3 years of my study (Turtle 38,9,10; Table 1.1). For this female, the average distance between yearly consecutive nest sites
was 388.3 m (SE = 23.2). All three of these sites were in the same open active prairie restoration
area that was co-dominated by big bluestem, spotted knapweed and common mullen, with
autumn olive and staghorn sumac dispersed throughout (92 ha). I collected data on nest site
fidelity for 6 additional turtles for 2 years (Table 1.1). Average distance between yearly
consecutive nests was 215.4 m (SE = 61.0). No nest locations were <11.7 m from a prior nest
location. Radio-tagged females were never observed nesting or staging in patches of vegetation
different from previous years. My results indicated that female turtles exhibited high fidelity to
patches of vegetation and not specific nest locations, if those patches contain areas with sparse
vegetation cover and exposed mineral soil.
1.3.5. Clutch Size, Incubation Period, and Clutch Success
Average clutch size was 6.0 (SE = 0.47), and clutch sizes did not differ by year (Kruskal-Wallis
Test, χ2 = 0.58, df = 2, p = 0.75). Mean incubation period for hatchlings (counted from nest date
to emergence date in the fall) was 113.8 days (SE = 5.02, range = 89-141), and there was no
difference among years (Kruskal-Wallis Test, χ2 = 0.41, df = 2, p = 0.81).

19

For nests protected from predation, ≥ 1 egg hatched in 15 of 22 (68%) nests (Table 1.2).
Of the 7 nests that did not produce emergent hatchlings, 2 (29%) had part of the clutch not break
out of their eggs, and the other part break free of the eggs but die within the cavity. Three of the
6 nests that failed to emerge from eggs occurred in a corn field. At the time of nest initiation in
June, the corn was approximately 0.30 m tall.
The average number of hatchlings that emerged from protected nests was 3.0 (SE = 0.62)
and did not differ by year (Kruskal-Wallis Test, χ2 = 0.32, df = 2, p = 0.85). I found that mean
clutch success, defined as the number of hatchlings successfully emerged from the nest cavity
divided by clutch size, was 0.49 hatchlings/egg (SE = 0.09). Clutch success did not differ by year
(Kruskal-Wallis Test, χ2 = 0.28, df = 2, p = 0.87). For nests that contained eggs that were not
predated but still did not emerge (n=16; Table 1.2), I found that 36-47% of the eggs were
undeveloped or had hatchlings that died in the egg, and 1-4% that hatched but died in the nest
cavity. My results indicated that most clutches (83%), if predation loss is excluded, experienced
egg or hatchling loss.
I failed to identify significant parameters in the clutch size and count of developed eggs
models (Table 1.3). Although the Kruskal-Wallis test indicated no year effect on clutch sizes, the
generalized linear model suggested some support for a year effect (p = 0.06), with higher clutch
sizes later in the study (Table 1.3). Additionally, weak support for fire scarring (p = 0.10) was
identified, with higher clutch sizes from fire scarred females (Table 1.3). The number of emerged
hatchlings varied by cover type, with nests in the gravel pit producing significantly more
emerged hatchlings compared to nests in the corn field (Table 1.3). There was also some support
(p = 0.11) for nests in Eagle Prairie producing more hatchlings than nests in the cornfield (Table
1.3). Year and clutch size did not influence the number of emerged hatchlings (Table 1.3).

20

1.3.6. Hatchling Movements, Behavior, and Known Sources of Mortality
For clutches laid in 2013, I found that hatchlings (n = 26 from 4 nests) emerged between
September 7th and October 26th, 2013, and into the following spring (n = 5 from 3 nests) between
April 30th and May 30th, 2014. One nest emerged as an entire clutch in the spring after
overwintering in the nest cavity (n = 2) on April 30th, 2014, whereas 1 nest had a portion of the
clutch emerge in the fall (n = 4), and the remainder in the spring (n = 3, emerged between May
15-30th, 2014). For clutches laid in 2014, I observed hatchlings emerge in the fall (n = 28 from 5
nests) between September 12-30th. For clutches laid in 2015, I observed hatchlings emerge in the
fall (n = 7 from 2 nests) between September 7th and October 15th, 2015. Two nests emerged as
an entire clutch between May 6-15th, 2016, after overwintering in the nest cavity (n = 8
hatchlings).
Of the 75 hatchlings I encountered during this study (Table 1.2), I successfully relocated
38 individuals at least once. I experienced a wide range of success in relocating hatchlings,
ranging between <1 to 231 days after emergence for individuals. I followed 10 individuals (from
6 nests) to their overwintering locations. Six of these 10 individuals were radio-tagged. The
individuals that I tracked to overwintering locations were subsequently located the following
spring in approximately the same location (Âą1 m). Hatchlings remained in overwintering
locations and dispersed between May 9th and May 15th (n = 2) in 2013 and between May 1st and
May 9th (n = 7) in 2014.
I tracked hatchlings using the monofilament-UV light method for an average of 12.8 days
(SE = 11.2) for an average total distance of 12.8 m (SE = 11.9, range = 0.76 – 55.86). I tracked
hatchlings with both the UV monofilament and radio tags for an average of 37.7 days (SE = 2.8),
for a total distance of 37.9 m (SE = 15.7, range = 3.26 - 123.7). From both methods combined,

21

hatchlings moved an average of 1.2 m/day (SE = 0.2) in the fall. I observed a wide range of
hatchling movement patterns; one hatchling overwintered 3.3 m from the nest cavity while
another moved at least 123.7 m away within 3 days.
I documented movement by hatchlings consistently away from the nest site without
returning. I found that telemetered hatchlings that emerged in the fall in prairie areas and that
were successfully tracked to their overwintering sites (n = 9) remained in this vegetation type
throughout the winter and into the spring. Hatchlings were highly concealed (greater than 75%
concealed) in 268 of 320 relocations (84% of relocations). Concealment materials included dead
and live vegetation, fallen leaves, root base thickets of big bluestem, and mineral soil. Hatchlings
seemed to only use mineral soil as concealment when beginning to overwinter. I observed
hatchlings beginning to overwinter as early as October 19th, 2013 and October 5th, 2014, which I
defined as burying completely into the mineral soil and remaining in the same location for more
than 1 week. Hatchlings overwintered between 1.5 and 3 cm below the soil surface. In the
process of overwintering, hatchlings would usually turn 180 degrees into the soil to bury.
I was unable to determine the fates for 51 of 58 emerged hatchlings, attesting to their
cryptic life history strategy and difficulties in tracking individuals for long periods of time.
Known sources of mortality on hatchlings included a vehicle (n = 1), prescribed fire applied in
the spring (n = 1), mammalian predation (burrowing rodent, n = 1), and avian predation (n = 4).
1.4. Discussion
I determined nesting dates, clutch sizes and clutch success, and hatchling behaviors of eastern
box turtles over 3 seasons in a recreation area in southwestern Michigan. Collectively, results
from this study provide insights into recruitment vulnerabilities of this species in recreational
landscapes that are being restored with active management that included prescribed fire. Similar

22

to other box turtle studies at more northerly latitudes, I found that females consistently staged for
nesting in late May, and nested in early June. For example, Doroff and Keith (1990) found that
ornate box turtles (Terrapene ornata ornata) laid eggs between 29 May and 26 June and hatched
after 79-84 days in south-central Wisconsin. Kipp (2003) documented nest (n = 39) excavation
by eastern box turtles between 27 May and 11 July in Delaware, and Willey and Sievert (2012)
documented nests (n = 34) in Massachusetts between 27 May and 10 July, with peak nesting in
early June. At northerly latitudes, eastern box turtles seem to consistently nest from the end of
May into early July.
My observed clutch sizes were similar to clutch sizes documented by others studying box
turtles. Doroff and Keith (1990) documented average clutch sizes for ornate box turtles ranging
between 4.1 (nests = 8) and 2.8 (nests = 8) in southern Wisconsin. Kipp (2003) found that clutch
sizes ranged between 1 and 9 (mean = 4.6; nests = 53) in Delaware, and Willey and Sievert
(2012) found 3-10 eggs/clutch (mean = 5.8; nests = 31) in Massachusetts. In my study of nests
protected from predation, the majority (68%) of nests successfully hatched at least 1 egg, and
between 45 and 55% of the clutch emerged from the nest cavity on any given year. Others have
also found that <70% of clutches produce emerged hatchlings, ranging from 24 to 36% (Kipp
2003; not all nests protected from predation), 55% (Willey and Sievert 2012; all nests protected
from predation), and up to 69% (Doroff and Keith 1990; not all nests protected from predation)
Adult female eastern box turtles demonstrated site fidelity to patches of open nesting
areas, but the specific location of nest sites within each patch varied considerably. This result is
consistent with Kipp (2003), who found that female box turtles generally nested within 295 m of
previous nest locations. Fidelity to patches of nesting habitat may be a result of limited habitat
availability. For example, adult females in a similar study northwest of FCSRA consistently

23

staged in the same open area from year to year, migrating from many locations (P. Laarman,
pers. comm.). This patch was believed to be the only suitable nesting habitat reasonably
available. Females at FCSRA seemed to migrate to the nearest open area, and were never seen
staging in other open areas during subsequent nesting seasons. As nesting habitat becomes
increasingly fragmented from vegetation succession and human activities (like roadways), box
turtles may increasingly use less than ideal habitat for nesting, like agricultural fields (Kipp
2003; this study). The fidelity of eastern box turtles to suitable nesting habitat and the tendency
to use agricultural fields (that appear to act as population sinks) highlight the importance of
focusing management on providing open nesting areas.
Once emerged, hatchling box turtles moved approximately 1.2 m/day away from the nest
cavity, but individual daily movement varied substantially. Most hatchlings remained in the
vegetation patch containing the nest cavity, and burrowed into vegetation or soil (also observed
by Burke and Capitano 2011b). Hatchlings entered overwinter sites in early to mid-October and
emerged in early to mid-May. Hatchlings overwintering and subsequently emerging in
grasslands adjacent to oak, or dry-mesic southern forests may be particularly vulnerable to spring
prescribed fires through these grasslands and woodlands, primarily because they tend to hide in
herbaceous vegetation and leaf litter, and when buried are close (<2 cm) to the soil surface (Ernst
et al. 1995). This is particularly problematic for ecosystem restoration programs that require
annual prescribed burning as recruitment classes may never get a chance to enter the population.
As a k-selected species, the conservation of eastern box turtles depends on the protection
of breeding adults (Congdon et al. 1993). Anthropogenic activities having negative effects on
breeding adults can result in local population extirpation, albeit gradual and perhaps
imperceptible (Doroff and Keith 1990). Although I documented an apparently low mortality rate

24

for adult females (6% over 3 years), even low levels of adult female mortality in eastern box
turtles can have significant detrimental effects on population persistence (Williams and Parker
1987; Congdon et al. 1993). One source of confirmed mortality in my study occurred as a result
of prescribed fire where the turtle failed to evade an advancing fire front (even though other
turtles successfully negotiated the fire; Melvin 2017:Chapter 3). Prior to the burn, this female did
not exhibit any outward signs of impairment or stress, but she was slow in responding to the
flame front compared to other adult turtles. The cause of the other documented adult mortality
could not be unequivocally confirmed.
Environmental factors such as surface temperature, ground temperature, and vegetation
cover can be correlated with nest success of box turtles (Morjan and Valenzuela 2001; Kipp
2003). Incubation period can depend on soil characteristics such as temperature, soil type and
moisture (Allard 1948). Although I did not measure soil characteristics, I observed that nest
location and the associated vegetation cover type was the most significant factor in determining
hatchling emergence. For example, I observed that females nested in agricultural fields when the
cover crop (corn) was short and sunlight directly reached the soil. These nests all contained eggs
that showed some level of developed neonates, but all of these clutches failed to produce
emerged hatchlings. It seems reasonable to presume that agricultural fields can act as sinks to
local box turtle populations. The threats include exposing adult females to heavy equipment or
chemical use, but perhaps more importantly the lack of recruitment presumably caused by cover
crops shading nests to the point of incubation failure.
Similar to other studies (e.g., Kipp 2003; Willey and Sievert 2012), I observed direct
predation effects on eastern box turtle recruitment. Predator exclosures played a critical role in
the clutch successes that I documented. Although I had a limited sample of unprotected nests, all

25

experienced depredation and resulted in complete loss of the clutch. Similar to other landscapes
that support box turtles, FCSRA supports a diversity of reptilian egg predators including
raccoons, skunks, and opossums (Didelphis virginiana). It is important to note that my results on
nesting and hatchling emergence success represent potential eastern box turtle recruitment as the
data were collected from protected nests. Anecdotally, my data suggested that the presence of
predators in the study area likely has a strong negative effect on box turtle recruitment.
Managing box turtle habitat can be challenging. Too much sunlight can desiccate the soil
and reduce foraging opportunities (Saumure and Bider 1998; Kipp 2003), whereas too little
sunlight may prevent eggs from developing. The effects of changing environmental conditions
may negatively influence reproductive potential and cause the decline of local populations (Kipp
2003; Currylow et al. 2012). If sufficient nesting habitat does not exist within the home ranges of
box turtles, nesting may occur in locations where eggs or hatchlings may not survive (Kipp
2003). This may also be true for tallgrass prairie restoration that utilizes mowing, prescribed fire,
and herbicide application to create thick, homogenous communities of native grasses, such as big
bluestem. I found that sparsely vegetated areas within these types of communities are important
to successful nesting.
Here, I observed potential stressors to recruitment of eastern box turtles that included
predation, habitats that appeared to act as population sinks, and hatchling vulnerabilities to
prescribed fire. Understanding the nesting ecology and subsequent hatchling box turtle survival,
movements and behavior in altered or fragmented landscapes is a necessary precursor to
developing management strategies for box turtle conservation. I suggest that future research on
recruitment and nest success focus on the role of agricultural fields of any cover crop, better
understanding micro-scale differences in soil temperature and moisture in restored tallgrass

26

prairies, and potential manipulation of site fidelity exhibited by females with man-made nesting
areas.
1.5. Management Implications
Land managers using prescribed fire should carefully consider the frequency of burns in nesting
habitat, as hatchling turtles tended to remain in close proximity to these habitats through the fall
and into the following spring. Recognizing that prescribed fire is an important component of box
turtle habitat conservation, I recommend that annual burns be avoided, preferentially timing
burns every 3 to 4 years and with consideration for the minimum amount of fire necessary to
maintain grasslands. Managers should also recognize that portions of the adult female population
may be nesting in agricultural fields, potentially requiring a shift in field locations or change in
agricultural practices. For example, suitable nesting habitat in buffers along agricultural fields
may intercept females from entering the area of the cover crop. Furthermore, I suggest targeted
control of predator populations in areas known to support nesting box turtles.
1.6. Acknowledgements
Summer and academic funding for this project was provided by Michigan State University and
the Joint Fire Science Program under Project JFSP 14-3-01-30. Funding for equipment and
supplies was provided by the Michigan Department of Natural Resources - Parks and Recreation
Division, Michigan Society of Herpetologists, The North American Box Turtle Conservation
Committee, and The Safari Club International – Michigan Involvement Committee. I thank my
entire field crew from 2013 to 2015 for assistance in the collection of field data. Special thanks
to P. Laarman from Grand Valley State University for designing predator exclosures, to A.
Ihnken, G. Palmgren, R. Clancy and R. Fahlsing for FCSRA and box turtle expertise, and P
Melvin for equipment construction.

27

APPENDIX

28

Table 1.1. Radio-tagged female eastern box turtles at Fort Custer State Recreation Area, southern
Michigan, and years they were successfully tracked to a nesting site where eggs were laid, 20132015. Within a year, 1 = tracked to nesting site, 0 = not tracked to nesting site, and “.” =
unknown nesting status.
Monitoring Year
2013
2014
2015
1
1
1
1
1
0
1
1
0
0
1
1
0
1
1
0
1
1
0
1
0
0
1
0
0
1
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
¡
0
0
¡
0
0
¡
1
¡
¡
1
¡
¡
1
¡
¡
1
¡
¡
0
¡
¡
0
¡
¡
0
¡
¡
0
¡
¡
0
¡
¡
0
¡
¡
0
¡
¡
0
¡
¡
0
¡
¡
¡
0
¡
¡
¡
1
¡
¡
0

Turtle ID
3-8,9,10
2,3,8-0
2-9,10
1,2-3
1,9-9,11
2,9-3
1,2,11-1
2,8,9-0
2,9-9
1,2-2
1,2,10-9
1,3,8-12
1,3,9-1
1,3,9-2
8--0
1-1,2
1,11-2
1,2,10-3
1,2,10-1
1,2,3-1
1,2,3-8
1,3-1
1,11-1
1,2,10-2
1,2,3-2
1,3,8-1
1,3,8-2
1,3-1
2,10,11-0
2,9,11-0
8,9-0
2,3-1
0-2,3,10
0-2,8,9
Total Monitored (Total Nests)

31 (8)
29

19 (10)

17 (5)

Table 1.2. Eastern box turtle nests monitored in the Fort Custer State Recreation Area,
southwestern Michigan, 2013-2015, including vegetation type containing the nest, date of nest
initiation, clutch size, number of emerged hatchlings, and information on hatchling movements.

Nest
ID
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
a

Vegetation Typea
Disk Golf Course
Prairie Restoration 1
Prairie Restoration 1
Gravel Pit
Gravel Pit
Gravel Pit
Corn Field**
Corn Field
Corn Field
Corn Field
Prairie Restoration 2
Prairie Restoration 2
Gravel Pit
Prairie Restoration 1
Prairie Restoration 1
Prairie Restoration 1
Prairie Restoration 1
Prairie Restoration 1
Antenna Field
Disk Golf Course
Prairie Restoration 2
Prairie Restoration 2
Gravel Pit
Antenna Field
Prairie Restoration 1

Date
Clutch
Nested
Size
6/4/2013
2
6/9/2013
9
6/11/2013
6
6/10/2013
5
6/14/2013
9
6/10/2013
3
6/6/2013
2
6/13/2013
6
6/10/2013
6
6/8/2013
8
6/2/2014
9
6/7/2014
2
6/2/2014
3
6/2/2014
10
6/3/2014
8
6/8/2014
6
6/7/2014
7
6/6/2014
5
6/7/2014
5
6/2/2014
5
6/3/2015
4
6/4/2015
7
6/8/2015
7
6/5/2015
7
6/5/2015
8

Emerged
Hatchlings
1
8b
6
5
8
0
2b
0
0
0
0
2
0c
6
8
6
6
0
0
0c
2
0c
6
3
6

Hatchling Movements
Number
Average Daily
Recorded
(m/day) (SE)
d
¡
¡
10
1.91 (0.52)
19
0.80 (0.12)
20
1.06 (0.08)
38
1.47 (0.25)
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
60
1.07 (0.08)
47
1.20 (0.12)
61
0.25 (0.03)
61
6.25 (1.52)
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡
¡

Disk Golf Course = mowed portion of disk golf course; Prairie restoration 1 = active prairie

restoration site that was co-dominated by big bluestem (Andropogon gerardii), spotted knapweed
(Centaurea maculosa) and common mullen (Verbascum thapsus) with autumn olive (Elaeagnus
umbellata) and staghorn sumac (Rhus typhina) dispersed throughout; Prairie restoration 2 = big
bluestem dominated in active restoration; Gravel Pit = grass and forb dominated abandoned
gravel pit; Corn field = actively farmed corn field (** denotes a nest in a 3 m wide agricultural
30

Table 1.2. (cont’d).
buffer area adjacent to the corn field); Antenna field = forb and lichen dominated radio-antenna
field; Herbicide Area = herbicided area in a big bluestem dominated tallgrass prairie in active
restoration.
b

Nest contained at least one hatchling that overwintered in the nest cavity and emerged the

following May.
c

Clutch lost to predation.

d

No data.

31

Table 1.3. Generalized linear mixed effects model results for yearly clutch size, number of developed eggs, and number of emerged
hatchlings of eastern box turtles at Fort Custer State Recreation Area, southwestern Michigan, 2013-2015.
Model

Parameter

Estimate

SE

T

p-value

Clutch Size

Female mass (β1)

-0.006

0.009

-0.585

0.58

Female age (β2)

0.190

0.234

0.0811

0.44

Fire scarring (β3)

2.464

1.28

1.92

0.10

Year (β4)

1.676

0.754

2.22

0.06

Year (β1)

-0.240

0.911

-0.264

0.80

Eagle Prairie

0.445

1.965

0.226

0.83

Gravel Pit

3.045

2.284

1.333

0.22

Tallgrass Prairie

-1.626

2.513

-0.647

0.54

-0.666

0.915

-0.728

0.49

Eagle Prairie

4.110

2.279

1.803

0.11

Gravel Pit

6.208

2.247

2.762

0.03

Tallgrass Prairie

1.501

2.795

0.537

0.67

0.376

0.268

1.405

0.20

Count of Developed Eggs

Vegetation Type (β2)
Corn

Count of Emerged Hatchlings

Year (β1)
Vegetation Type (β2)
Corn

Clutch Size (β3)

32

Figure 1.1. Management unit boundaries used for studying nesting ecology of eastern box turtles in of Fort Custer State Recreation
Area, southwestern Michigan. Boundaries are portrayed in bold.

33

Figure 1.2. Predator exclosures used at Fort Custer State Recreation Area to protect eastern box
turtle nests. Exclosures were 0.61 x 0.61 x 0.31 m bottomless wooden frames wrapped in 0.6 cm
wire mesh, preserved and camouflaged with an acrylic-based solid stain in olive drab green.
Exclosures featured a removable lid attached with (4) 3 cm deck screws. Exclosures were dug 5
cm into the soil surface around the nest.

34

Figure 1.3. Five nesting sites by vegetation type in Fort Custer State Recreation Area, Michigan, USA. A = big bluestem (Andropogon
gerardii) dominated tallgrass prairie in active restoration, B = herbicide applied area of big bluestem (Andropogon gerardii)
dominated tallgrass prairie in active restoration, C = forb and lichen dominated field, D = agricultural field (corn), E = grass and forb
dominated abandoned gravel pit.

B

A

D

C

E

35

LITERATURE CITED

36

LITERATURE CITED

Albert, D. A. 1995. Regional landscape ecosystems of Michigan, Minnesota, and Wisconsin: A
working map and classification. General Technical Report NC-178. St. Paul, MN:
USDA, Forest Service, North Central Forest Experiment Station, St. Paul, USA.
Albert, D. A., S. R. Denton, and B. V. Barnes. 1986. Regional landscape ecosystems of
Michigan. University of Michigan, School of Natural Resources, Ann Arbor, USA.
Allard, H. A. 1935. The natural history of the box turtle. Scientific Monthly 41:325-338.
Allard, H. A. 1948. The eastern box turtle and its behavior. Journal of the Tennessee Academy of
Science 23:307–321.
Belzer, W. R. 1997. Box turtle conservation issues. Reptile and Amphibian Mag. (Jul/Aug):3235.
Belzer, W. R. 2002. A nine-year study of eastern box turtle courtship with implications for
reproductive success and conservation in a translocated population. Turtle and
Tortoise Newsletter 6:17-26.
Belzer, W. R., and M. B. Steisslinger. 1999. The box turtle: Room with a view on species
decline. The American Biology Teacher 61:510–513.
Belzer, W. R., S. Seibert, and B. Atkinson. 2002. Putative chipmunk predation of juvenile
eastern box turtles. Turtle and Tortoise Newsletter 5:8-9.
Benjamini, Y., and Y. Hochberg. 1995. Controlling the false discovery rate: a practical and
powerful approach to multiple testing. Journal of the Royal Statistical Society B 57:
289-300.
Breder, R. B. 1927. Turtle trailing: a new technique for studying the life habits of certain
Testudinata. Zoologica 9:231-243.
Breitenbach, G. L., J. D. Congdon, and R.C. Van Loben Sels. 1984. Winter temperatures of
Chrysemys picta nests in Michigan: effects on hatchling survival. Herpetologica
40:76–81.
Brooks, R. J., G. P. Brown, and D. A. Galbraith. 1991. Effects of a sudden increase in natural
mortality of adults on a population of the common snapping turtle (Chelydra
serpentina). Canadian Journal of Zoology 69:1314–1320.

37

Budischak, S. A., J. M. Hester, S. J. Price, and M. E. Dorcas. 2006. Natural history of terrapene
carolina (box turtles) in an urbanized landscape. Southeastern Naturalist 5(2):191204.
Burke, R. L., and W. Capitano. 2011a. Nesting ecology and hatching success of the eastern box
turtle, Terrapene carolina, on Long Island, New York. American Midland Naturalist
165:137–142.
Burke, R. L., and W. Capitano. 2011b. Eastern box turtle, Terrapene carolina, neonate ecology
on Long Island, New York. Chelonian Conservation Biology 10(2):256-259.
Burke, V. J., J. E. Lovich, and J. W. Gibbons. 2000. Conservation of freshwater turtles. Pages
156-179 in M. W. Klemens, editor. Turtle conservation. Smithsonian Institution
Press, Washington, D.C., USA.
Butler, B. O., and T. E. Graham. 1995. Early post-emergent behavior and habitat selection in
hatchling Blanding’s Turtles, Emydoidea blandingii, in Massachusetts. Chelonian
Conservation Biology 1:187–196.
Cagle, F. R. 1939. A system of marking turtles for future identification. Copeia 1939:170–173.
Congdon, J. D., and J. W. Gibbons. 1990. The evolution of turtle life histories. Pages 45–54, in
J.W. Gibbons, editor. Life history and ecology of the slider turtle. Smithsonian
Institution Press, Washington, D.C., USA.
Congdon, J. D., A. E. Dunham, and R. C. Van Loben Sels. 1993. Delayed sexual maturity and
demographics of Blanding’s turtles (Emydoidea blandingi): implications for
conservation and management of long-lived organisms. Conservation Biology 7:826–
833.
Corder, G. W., and D. I. Foreman. 2009. Nonparametric Statistics for Non-Statisticians. John
Wiley and Sons, Hoboken, New Jersey, USA.
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):e40473.
Denton, S. R. 1985. Ecological climatic regions and tree distributions in Michigan. Dissertation,
University of Michigan, Ann Arbor, USA.
Dinkelacker, S. A., J. P. Costanzo, and R. E. Lee, Jr. 2005. Anoxia tolerance and freeze tolerance
in hatchling turtles. Journal of Comparative Physiology B 175:209–217.
Dodd, C. K., Jr., 2001. North American Box Turtles: A Natural History. University of Oklahoma
Press, Norman, USA.

38

Dodd, C. K., Jr., K. M. Enge, and J. N. Stuart. 1989. Reptiles on the highways in north-central
Alabama, USA. Journal of Herpetology 23:197–200.
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.
Eagle, A. C., E. M. Hay-Chmielewski, K. T. Cleveland, A. L. Derosier, M.E. Herbert, and R.A.
Rustem, editors. 2005. Michigan's Wildlife Action Plan. Michigan Department of
Natural Resources. Lansing, Michigan, USA.
Eichenlaub, V. L. 1979. Weather and the climate of the Great Lakes Region. The University of
Notre Dame Press, Notre Dame, USA.
Eichenlaub, V. L, J. R. Harman, F. V. Nurnberger, and H. J. Stolle. 1990. The climatic atlas of
Michigan. The University of Notre Dame Press, Notre Dame, USA.
Ernst, C. H., J. E. Lovich, and R.W. Barbour. 1994. Turtles of the United States and Canada.
Smithsonian Institution Press, Washington, D.C., USA.
Ernst, C. H., T. P. Boucher, S. W. Sekscienski, and J. C. Wilgenbusch. 1995. Fire ecology of the
Florida box turtle, Terrapene carolina bauri. Herpetological Review 26(4):185-187.
Flitz, B. A. and S. J. Mullin. 2006. Nest-site selection in the Eastern Box Turtle (Terrapene
Carolina Carolina) in Illinois. Chelonian Conservation and Biology 5(2):209-312.
Forsythe, P., B. A. Flitz, and S. J. Mullin. 2004. Radio telemetry and post-emergent habitat
selection of neonate box turtles (Emydidae: Terrapene carolina) in central Illinois.
Herpetological Review 35:333–335.
Gibbons, J. W., D. E. Scott., T. J. Ryan, K. A. Buhlmann, T. D. Tuberville, B. S. Metts, J. L.
Greene, T. Mills, Y. Leiden, S. Poppy, and C. T. Winne. 2000. The global decline of
reptiles, dĂŠjĂ  vu amphibians. Bioscience 50(8):653-666.
Hall, R. J., P. F. Henry, and C. M. Bunck. 1999. Fifty-year trends in a box turtle population in
Maryland. Biological Conservation 88:165-172.
Hallgren-Scaffidi, L. 1986. Habitat, home range and population study of the eastern box turtle
(Terrapene carolina). Thesis, University of Maryland, College Park, USA.
Harding, J. 1997. Eastern box turtle: Terrapene carolina carolina. Pages 194-201 in M.M.
Douglas, editor. Amphibians and reptiles of the Great Lakes Region. University of
Michigan Press, Ann Arbor, USA.
Keller, C., C. Diaz-Paniagua, and A. Andreu. 1997. Post-emergent field activity and growth rates
of hatchling spur-thighed tortoises, Testudograeca. Canadian Journal of Zoology
75:1089–1098.
39

Kipp, R. L. 2003. Nesting ecology of the Eastern Box Turtle (Terrapene Carolina Carolina) in a
fragmented landscape. Thesis, University of Delaware, Newark, USA.
Klemens, M. W. 1989. The methodology of conservation. Pages 1-4 in I.R. Swingland and M.W.
Klemens, editors. The Conservation Biology of Tortoises. Occas. Papers. IUCN/SSC
5.
Klemens, M. W. editor. 2000. Turtle conservation. Smithsonian Institution Press, Washington,
D.C., USA.
Kost, M. A. 2004. Natural community abstract for mesic prairie. Michigan Natural Features
Inventory, Lansing, USA.
Legler, J. M. 1960. Natural history of the ornate box turtle, Terrapene ornata ornata. University
of Kansas Publications, Museum of Natural History 11(10):527-669.
Madden, R. C. 1975. Home range, movements, and orientation in the eastern box turtle
Terrapene carolina carolina. Dissertation, City University of New York, USA.
Melvin, T. A. 2017. Prescribed fire effects on eastern box turtles in southwestern Michigan.
Thesis, Michigan State University, East Lansing, USA.
Mitchell, J. C., and M. W. Klemens. 2000. Primary and secondary effects of habitat alteration.
Pages 5-32 in M.W. Klemens, editor. Turtle conservation. Smithsonian Institution
Press, Washington, D.C., USA.
Morjan, C. L., and N. Valenzuela. 2001. Is ground-nuzzling by female turtles associated with
soil surface temperatures? Journal of Herpetology 35:668-672.
Nickerson, M. A., and A. L. Pitt. 2012. Historical turtle population decline and community
changes in an Ozark River. Bulletin of the Florida Museum of Natural History 51(4):
257–267.
Palmgren, G. 2004. Phase III natural resources management plan: Fort Custer State Recreation
Area (draft review). Parks and Recreation Bureau, Department of Natural Resources.
Lansing, USA.
R Development Core Team. 2014. R: A language and environment for statistical computing. R
Foundation for Statistics Computing, Vienna, Austria.
Rosenburg, D., and R. Swift. 2013. Post-emergence behavior of hatchling western pond turtles
(Actinemys marmorata) in Western Oregon. American Midland Naturalist 169:111121.

40

Saumure, R. A., and J. R. Bider. 1998. Impact of agricultural development on a
population of wood turtles (Clemmys insculpta) in southern QuĂŠbec, Canada.
Chelonian Conservation and Biology 3:37-45.
Schwartz, E. R., and C. W. Schwartz. 1991. A quarter-century study of survivorship in a
population of three-toed box turtles in Missouri. Copeia 1991:1120-1123.
Stickel, L. F. 1950. Populations and home range relationships of the box turtle, Terrapene c.
carolina (Linnaeus). Ecological Monographs 20(4):351-378.
Stickel, L. F. 1978. Changes in a box turtle population during three decades. Copeia 1978:221–
225.
Stickel, L. F., and C. M. Bunck. 1989. Growth and morphometrics of the box turtle,
Terrapene c. carolina. Journal of Herpetology 23:216–223.
Ultsch, G. R., M. Draud, and B. Wicklow. 2007. Post-emergence movements and overwintering
of snapping turtle, Chelydra serpentine, hatchlings in New York and New Hampshire.
Canadian Field Naturalist 121:178-181.
Warwick, C. 1993. The decline of North American box turtles. New York Turtle and Tortoise
Soc. Newsnotes 4(1):15-16.
Wilcox, B. A., and D. D. Murphy. 1985. Conservation strategy: The effects of fragmentation on
extinction. American Midland Naturalist 125:879–887.
Willey, L. L., and P. R. Sievert. 2012. Notes on the nesting ecology of eastern box turtles near
the northern limit of their range. Northeastern Naturalist 19(3):361-372.
Williams, E. C., and W. S. Parker. 1987. A long-term study of a box turtle (Terrapene carolina)
population at Allee Memorial Woods, Indiana, with emphasis on survivorship.
Herpetology 43:328–335.
Wilson, G. L. and C. H. Ernst. 2005. Reproductive ecology of Terrapene carolina carolina
(Eastern Box Turtle) in Central Virginia. Southeastern Naturalist 4(4):689-702.
Zar, J. H. 2010. Biostatistical Analysis. Fifth Edition. Pearson Prentice Hall, Upper Saddle River,
New Jersey, USA.

41

CHAPTER 2
RELIABILITY OF POST-FIRE SURVEYS FOR EASTERN BOX TURTLES

Abstract
Natural resource managers may rely on post-fire surveys to understand the direct effects of
prescribed fires on wildlife. It is generally assumed that post-fire surveys are effective at
quantifying fire-related mortality or injury, as individuals should be readily observed within
burned areas. I estimated detection probability during post-fire surveys for eastern box turtles
(Terrapene carolina carolina) in southwestern Michigan. Immediately prior to a May (growing
season) fire, I confirmed that 7 adult box turtles fitted with radio transmitters occupied the
proposed burn area. Two days after the burn I reconfirmed turtle locations and subsequently
conducted 6 independent visual encounter surveys through 2 burned areas (0.75 and 1.0 ha) that
contained telemetered turtles. For these 12 surveys, I found that average detection probability per
survey was low (0.11) (SE = 0.05) and highly variable among surveyors (range = 0.00 – 0.50).
Based on this average detection probability, I estimated that 26 hour-long surveys per ha would
be required to reliably (95% confidence) detect turtles that were present in burned areas. I also
found that individual turtles directly exposed to fire remained buried for up to 12 hrs after the
fire was extinguished and were thus unavailable for detection immediately after the burn. Further
confounding post-fire survey results, buried turtles rapidly moved to unburned areas after
emerging from their subterranean refugia. My results suggested that typical visual encounter
surveys conducted for eastern box turtles after prescribed burning do not accurately reflect
occupancy status or fire caused mortality.

42

2.1. Introduction
Prescribed fire can be an effective and inexpensive tool for vegetation management (Knapp et al.
2009), but the direct (including injury and mortality) and indirect effects on wildlife (including
changes in body condition and animal movements or interactions) can be difficult to quantify.
Fire effects on reptiles and amphibians have been studied and summarized (e.g., Keyser et al.
2004, Russell et al. 2009, Roloff and Hmielowski 2014, O’Donnell et al. 2015), and the literature
generally suggests that prescribed fire has few significant direct effects. However, most of these
studies relied on animal counts or mark-recapture data that were uncorrected for detection
probability (Mazerolle et al. 2007). Furthermore, reptiles and amphibians are often grouped into
herpetofauna for fire studies (e.g., Means and Campbell 1981; Smith 2000; Floyd et al. 2002;
Renken 2005), potentially biasing results to those species that are readily observable and
abundant. For cryptic, mobile, or rare species like eastern box turtles (Terrapene carolina
carolina), the error associated with failure to detect when present may have substantial
conservation ramifications (Gu and Swihart 2004, Refsnider et al. 2011).
Eastern box turtles tend to occur in fire maintained ecosystems (Russell et al. 1999) and,
although once common (Ernst and Lovich 2009), are now recognized as a species of
conservation concern throughout much of their geographic range (van Dijk 2011). Post-fire
visual encounter surveys for eastern box turtles are used by some land management agencies and
researchers to estimate fire impacts on turtle populations (e.g., Platt et al. 2010). Results from
these surveys are also used to inform implementation of burn programs. Most burn plans with the
potential to impact species of conservation concern make reference to minimizing fire effects on
these organisms. To substantiate this claim, agency personnel sometimes conduct post-fire
surveys that typically involve walking through burned areas looking for animals or animal

43

remains. However, box turtles are difficult to locate and observe as they are well camouflaged,
exhibit cryptic behavior such as remaining buried under leaf litter or in dense thickets for long
periods of time (Dodd 2001), and can remain motionless when predators or researchers are near
(T.A. Melvin, pers. obs.). Additionally, box turtles are generally regarded as crepuscular, with
most activity occurring in early morning or evening (Dodd 2001), including nesting, which
generally takes place after dark (Flitz and Mullin 2006).
My goal was to better understand the effectiveness of post-fire surveys for eastern box
turtles. I quantified detection probability shortly after a prescribed fire using visual encounter
surveys in areas that were knowingly occupied by radio-tagged turtles. I also estimated the
number of post-fire surveys needed to reliably detect box turtles, and described how post-fire
behaviors of turtles influenced their availability for detection. My results can be used to improve
the post-fire survey process for eastern box turtles.
2.2. Methods
2.2.1. Study Area
My study was conducted at Fort Custer State Recreation Area (FCSRA; 1,226 ha), located in the
southern lower peninsula of Michigan (Kalamazoo and Calhoun Counties), USA. This area was
characterized by heterogeneous vegetation patterns resulting from varied topography, climate,
and human influence including high levels of agricultural and urban development (Albert 1995,
Eagle et al. 2005). Most of the soils are calcareous and loamy, derived from underlying
limestone, shale, and sandstone bedrock (Eagle et al. 2005). Glacial till deposits are primarily
loams, silt loams and clay loams (Eagle et al. 2005). The climate is considered humid
continental, although FCSRA is strongly influence by the Maritime Tropical air mass and
proximity to Lake Michigan, resulting in a warmer climate, moderated inland temperature

44

fluctuations, and induced lake-effect snows (Eichenlaub 1979, Denton 1985, Albert et al. 1986,
Eichenlaub et al. 1990). The average length of the growing season was 154 days (Albert et al.
1986). Mean monthly (May to October) minimum and maximum temperatures during the study
ranged from -1.7 – 7.2° C and 22.2 – 34.4° C, respectively.
The region containing FCSRA was historically dominated by fire-dependent savanna and
prairie (Albert 1995), and contained the only extensive areas of mesic prairie found in Michigan
(Kost 2004). In the early 1800’s, plant communities included dry and dry-mesic southern forest
(oak-hickory; Quercus spp. – Carya spp.), oak barrens (mixed oak savanna), emergent marsh,
and southern (mixed hardwood) swamp (Palmgren 2004). Currently, FCSRA supports degraded
patches of oak barrens and prairie openings, scattered oak-hickory forests, and large blocks of
non-native black locust (Robinia pseudoacacia), surrounded primarily by agriculture (Palmgren
2004, Eagle et al. 2005). Restoration activities on FCSRA included use of prescribed fire to
maintain and improve savanna and prairie ecosystems. Prior to my study, prescribed fires were
generally small to medium-scale (1.6 to 414.4 ha), conducted in spring (before leaf out, typically
during the months of March and April), and occurred in 1 to 4-year intervals followed by
herbicide treatments in some areas.
My research focused on a 31.6 ha management unit in FCSRA (Figure 2.1). Within the
unit, I delineated 2 (0.75 and 1 ha) visual encounter survey areas that overlapped transmittered
box turtle home ranges (Figure 2.1). Survey area A consisted of a sloping hill (slope 19%, aspect
270-330) of mixed oak-hickory forest with a pre-burn understory consisting of red-osier
dogwood (Cornus sericea), sassafras (Sassafras albidum), wild black raspberry (Rubus
occidentalis) and multiflora rose (Rosa multiflora) with duff layers 4-7 cm deep. Survey Area B
consisted of dense stands of black locust and mixed oak-hickory forest, with multiflora rose in

45

the understory, and open grassland dominated by spotted knapweed (Centaurea stoebe) and
sweetclover (Melilotus officinalis) with little bluestem (Schizachyrium scoparium), goldenrod
(Solidago spp.) and milkweed (Asclepias spp.) throughout.
2.2.2 Research Burn and Survey Area Descriptions
I used a grid of semi-permanent 2 m radius circular plots spaced 40 m apart (n = 80) to describe
fire characteristics throughout the eastern half of the management unit. At the center of each plot,
I placed a 1.2 m untreated wooden stake to estimate flame heights above the soil surface. I also
placed a set of 3 Omega TL-10 adhesive, non-reversible fire temperature labels (OMEGA
Engineering, INC., Stamford, Connecticut) that recorded temperatures between 87 – 260° C.
Temperatures were recorded to the nearest 6° C. Labels were affixed to 11 x 20 cm piece of
aluminum roof flashing that was bent to a 90° angle and anchored into the soil so that the side
with the temperature labels ran parallel to and 4 cm above the soil surface. I recorded the times
that the flame front reached different plot centers to estimate rate of spread, and estimated flame
heights by observing how high flames reached when they first contacted the wooden stakes.
2.2.3. Radio-telemetry and Fire Observation of Eastern Box Turtles
Turtles were initially captured in May 2012 and June 2013 using visual encounter surveys and
turtle detector dogs. Adult box turtles were radio-tagged with a Holohil R1-2B 14.5g transmitter
(Holohil Systems Ltd., Carp, Ontario, Canada) attached to the right or left anterior pleural
carapacial scutes using multi-purpose 5-minute set epoxy putty (LoctiteÂŽ, Henkel 289
Corporation, Cary, NC, USA). I estimated minimum age by counting annual rings on the
carapace (see Wilson et al. 2003). I also marked turtles by notching the shell (Cagle 1939).
Radio-tagged adult box turtles were monitored during the active seasons (April through October)
from May 2012 to August 2015 using portable telemetry receivers (Advanced Telemetry

46

Systems, Isanti, MN, USA), which allowed me to delineate core use areas within the
management unit. I located telemetered turtles 24 hrs before the research burn at which point
transmitters were spray-painted with RUST-OLEUM™ High-Heat Spray to prevent fire damage.
The carapace was covered with a spray guard to prevent heat resistant paint from adhering to the
turtle. All animals were handled in compliance with Michigan State University’s Institutional
Animal Care and Use Committee guidelines (IACUC Approval # 04/14-077-00).
After ignition of the prescribed fire, I relocated radio-tagged turtles from behind the
flame front to observe fire-related behaviors. I stayed 5-10 m away from the turtles to minimize
my influence on behaviors. I visually inspected telemetered turtles for direct injury and mortality
after the fire was extinguished and for 12, 24 and 48 hrs thereafter. After 48 hrs, I checked the
turtles 1-3 times per week.
2.2.4. Detection Probability Experimental Design
After the fire was completely extinguished (~48 hrs after ignition), I spaced parallel transects 10
m apart in the 2 detection survey areas (Figure 2.1). Transects were marked using spray paint
applied to tree stems at eye-level. Surveyors slowly walked the transects and scanned for box
turtles within 5 m of their position. Each surveyor (n=6) made one pass on all transects in each
survey area at roughly 1-hour intervals, between 1:30 and 7:00 p.m. Surveyors had similar
amounts of experience in negotiating the study site and had comparable training in visually
identifying eastern box turtles. Surveyors had no knowledge of the number of transmittered
turtles in the survey areas and did not disclose the results of their survey to other surveyors.
I located turtles using radio telemetry prior to each new hourly survey to confirm
availability of individuals for detection. Each time I relocated one of the marked turtles, I noted
the position and level of concealment. During surveys, each observer also recorded the locations

47

of detected box turtles using a handheld GPS unit, and recorded behavior and percent of the
carapace (top shell) obscured by vegetation, soils, or other natural objects when viewed from
straight above the turtle.
2.2.5. Data Analysis
I quantified post-fire detection probability of box turtles using the marked subsample
method (Lancia et al. 2005, Refsnider et al. 2011). I calculated box turtle detection probability,
βi, for each of the 12 survey periods using equation 12 in Lancia et al. (2005:121):
𝛽𝑖 =𝑚𝑖
𝑛𝑖

where mi was the number of transmittered box turtles observed during the i visual-encounter
survey and 𝑛𝑖 was the number of transmittered box turtles available to be observed during the i
survey period. I used the average detection probability over 6 surveys to characterize within
survey area detection probability, and averaged across all 12 surveys to portray overall detection
probability, 𝛽̂ . I estimated the number of surveys required to reliably (with 95% confidence)
detect box turtles using equation (4) in K𝑒́ ry (2002:331):
𝑁 min =

log 𝛼
log(1 − 𝜌)

where 𝛼 = 0.05 and 𝜌 was equivalent to my overall detection probability, 𝛽̂ .
2.3. Results
2.3.1. Fire Characteristics and Turtle Behavior
Average flame lengths of the fire were approximately 8 cm (SE = 0.63), average rate of spread
was approximately 1.5 m/minute (SE = 0.19), and average maximum temperatures were 149.3°
C (SE = 7.9). Approximately 51% of the management unit burned resulting in a mosaic of

48

burned and unburned upland habitats and an unburned emergent marsh. Within the survey areas,
90% and 65% burned in A and B, respectively.
Prior to fire ignition 11 telemetered turtles occupied the management unit but 6 were in a
wetland that did not burn. An additional marked turtle was within a dense black locust stand that
did not burn. The other 4 turtles were directly subjected to the flame front. These turtles
exhibited varying behaviors in response to the fire. One began to bury into mineral soil and
stopped once the flame front slowed and self-extinguished before reaching her location (roughly
3 m away), 2 turtles completely buried into mineral soil (approximately 13 cm down), and one
remained above ground and moved 1-2 m from her original location. This individual was
overtaken by the flame front, burning 90-100% of the carapace. Subsequently, this turtle moved
to an unburned black locust patch within 12 hrs and then to a wetland where she remained until
dying two weeks later.
I monitored these marked turtles for 2 years, and before the prescribed fire I only
observed burying behavior into mineral soil when individuals were entering or exiting brumation
in the late fall (October – November) or early spring (March-April). The 2 turtles that I observed
burying into mineral soil in advance of the flame front remained buried for 12 hrs following fire
extinguishment and subsequently moved to an adjacent wetland. Regardless of their exposure to
the flame front, marked turtles remained soaking in the adjacent wetland, or within 3-5 m of the
wetland edge, for 1-3 weeks following the fire. When moving back to burned areas from the
wetland, turtles seemed to remain within 1-2 m of unburned habitats that included decaying logs,
patches of leaf litter, or other areas unaffected by fire. They remained in these areas until postburn vegetation growth provided cover.
2.3.2. Detection Probability

49

During the post-burn visual-encounter surveys, 7 transmittered turtles (5 in area A, 2 in B; Figure
2.1) were confirmed to be within the survey areas. Observers detected 2 of the 7 turtles. One
observer detected 2 individuals, three observers detected one individual, and the remaining two
observers failed to detect turtles (Table 2.1). There were no unmarked turtles detected during the
surveys. The detection probability for individual surveys ranged between 0.00 and 0.50 (Table
2.1). Average detection probability for 12 surveys was 0.11 (SE = 0.05), and I estimated that 26
hour-long surveys per ha would be required to have 95% confidence that all available turtles
were detected.
Of the 7 turtles that occupied areas used for post-fire detection surveys, 3 were directly
subjected to the flame front and 4 were within the wetland that did not burn. These 7 turtles did
not move from their initial pre-survey locations for the duration of the 6 surveys in each area.
Although the 7 turtles were above ground 48 hrs after the fire was extinguished, they appeared to
seek concealment under live and dead herbaceous vegetation or downed wood (e.g., Figure 2.2).
Overhead concealment ranged from 5 to 80% among turtles (Figure 2.2). The two detected
individuals were concealed by woody debris. One was found in a fire blackened area (5%
concealed), and another in a patch of unburned vegetation (20% concealed). Concealment for the
5 undetected individuals ranged from 10 to 80%. These turtles were in unburned live vegetation.
2.4. Discussion
A typical post-fire survey for eastern box turtles involves walking the burned area and visually
identifying animals or their remains, usually 24-48 hrs after the fire is extinguished. Using visual
detection surveys, I found that average detection probability for adult eastern box turtles 48 hrs
after a prescribed fire was low. Detection probability by individual surveyor varied considerably
but never exceeded 0.50. I found that turtle behaviors following the fire contributed to low

50

detectability. These behaviors included burying for up to 12 hrs after the fire, and movements off
the burned area within 4 hrs if turtles did not bury, or within 4 hrs after surfacing from burying.
Given average surveyor capability, a growing season fire that resulted in a mosaic of burned and
unburned patches, highly variable concealment of turtles, and post-fire turtle behavior, I
estimated that 26 hour-long surveys would be needed to have 95% confidence that turtles in the
area were detected. My results indicated that visual encounter surveys of burned areas
underestimate direct and indirect fire effects on box turtles.
Although several terrestrial turtle species are of conservation concern worldwide, little is
known about survey efficacy. For 8 visual encounter surveys, detection probability was 0.03 for
telemetered ornate box turtles (Terrapene ornata) in a sand prairie in northwestern Illinois
(Refsnider et al. 2011). Detection probability can apparently be improved by using dogs. For
example, Kapher et al. (2012) found that detection varied between 0.33 and 0.67 for eastern box
turtles using detector dogs in lowland and upland deciduous forest in North Carolina. My results
are consistent with Refsnider et al. (2011) that found low visual detection probability for
terrestrial turtle species.
Following prescribed fire, I found that turtles moved into or remained within unburned
wetland areas and tended to remain in or near these areas until herbaceous vegetation started to
recover in burned areas. I purposefully expanded my survey plots to include these unburned
areas but none of the turtles were detected during the visual encounter surveys. I only
documented direct flame contact for one of the telemetered turtles during the fire, and this turtle
moved off the burned area and soaked in an adjacent wetland prior to death. Turtles severely
injured by fire may be more prone to hide, seek shelter, or soak (Dodd 2001). My results
suggested that shortly after prescribed fire, box turtles 1) move away from the burned areas, 2)

51

seek refugia that can make detection difficult 3) remain hidden in refugia for hours after the fire
has passed, and 4) can appear healthy only to later succumb to fire-related injuries. Collectively,
these behaviors can lead natural resource managers to conclude that box turtles were not present
in the burned area, or that turtles temporarily left the burned area, and therefore the fire had little
effect on movement, injury, or health.
My study is one of few that quantitatively evaluated post-fire behaviors and detection
probability for marked terrestrial turtles. Eastern box turtles tend to occupy ecosystems that were
historically maintained by fire, thus fire is often prescribed as a restoration tool. For species of
conservation concern, like box turtles, managers are often tasked with acknowledging fire effects
but tend to lack empirically based guidance on appropriate survey techniques. Although my
study provides that guidance, I caution that my results are based on a single prescribed fire in one
management unit during the early growing season. The efficacy of post-fire surveys for eastern
box turtles likely varies by fire intensity and the amount and spatial distribution of vegetation
and debris so I encourage additional detection studies in other fire and vegetation conditions.
2.5. Management Implications
Failure to account for detection probability and post-fire behaviors can result in erroneous
management decisions for eastern box turtles with respect to fire frequency, season, intensity,
and ignition type. During the growing season (defined here as after leaf-out) when box turtles are
active, I suggest that visual surveys can be more effective by watching for turtles as they move to
refugia during prescribed fires, and by focusing post-fire searches on refugia like large downed
logs, areas of unburned vegetation, and adjacent unburned areas (particularly wetlands). When
planning a growing season fire, I encourage managers to purposefully retain unburned refugia
within the burn area and ensure that turtles can negotiate the flame front to this refugia. Fires

52

should be slow and creeping backfires, lit in a way that maintains a single linear flame front. I
further suggest that growing season fires should be prevented from spreading into grasslands,
where faster fire spread rates may prohibit turtles from evading the flame front.
2.6. Acknowledgements
Funding for this project was provided by Michigan State University, by a Graduate Research
Innovation. (GRIN) grant from the Joint Fire Science Program under project JFSP 14-3-01-30,
Michigan Department of Natural Resources - Parks and Recreation Division, Michigan Society
of Herpetologists, The North American Box Turtle Conservation Committee, and the Michigan
Involvement Committee of Safari Club International. I thank My field crew for assistance in
field data collection. Special thanks to C. Burden from Michigan State University for assisting
with surveys, to A. Ihnken, G. Palmgren, R. Clancy and R. Fahlsing for their expertise on Fort
Custer State Recreation Area and box turtles.

53

APPENDIX

54

Table 2.1. Number of post fire marked eastern box turtles (Terrapene carolina carolina) available for detection (ni), visually detected
by observers (mi), and detection probability (𝛽𝑖 ) by survey and survey area, Fort Custer State Recreation Area, southwestern
Michigan, USA, 2015. Detection probability (𝛽𝑖 ) was calculated using the marked subsample method (Lancia et al. 2005:121).

Survey area A

Survey area B

Marked turtles

Detected

Detection probability

Marked turtles

Detected

Detection probability

(ni)

(mi)

(𝛽𝑖 )

(ni)

(mi)

(𝛽𝑖 )

1

5

1

0.20

2

0

0

2

5

1

0.20

2

1

0.50

3

5

0

0

2

0

0

4

5

1

0.20

2

0

0

5

5

0

0

2

0

0

6

5

1

0.20

2

0

0

Survey

Mean (SE)

0.13 (0.04)

55

0.08 (0.08)

Figure 2.1. Management unit (yellow polygon) at Fort Custer State Recreation Area in
southwestern Michigan, USA, 2015. Post fire visual-encounter survey areas A (0.75 ha) and B
(1.0 ha) delineated by white polygons.

B

A

0

150

300 Meters

56

Figure 2.2. Post fire concealment of eastern box turtles (Terrapene carolina carolina) monitored
using radio telemetry at Fort Custer State Recreation Area in southwestern Michigan, USA,
2015. A = 20% concealed under logs in burned area, B = 80% concealed, 20 m into unburned
wetland, C = 10% concealed at the edge of unburned wetland, D = 5% concealed leaf litter and
live vegetation in unburned patch of oak and black cherry (Prunus serotina).

57

LITERATURE CITED

58

LITERATURE CITED

Albert, D. A. 1995. Regional landscape ecosystems of Michigan, Minnesota, and Wisconsin: A
working map and classification. USDA, Forest Service, North Central Forest
Experiment Station, St. Paul, Minnesota, General Technical Report NC-178.
Albert, D. A., S. R. Denton, and B. V. Barnes. 1986. Regional landscape ecosystems of
Michigan. University of Michigan, School of Natural Resources, Ann Arbor,
Michigan, USA.
Cagle, F. R. 1939. A system of marking turtles for future identification. Copeia 1939:170–173.
Denton, S. R 1985. Ecological climatic regions and tree distributions in Michigan. Dissertation,
University of Michigan, Ann Arbor, USA.
Dodd, Jr., C. K. 2001. North American box turtles: a natural history. University of Oklahoma
Press, Norman, Oklahoma, USA.
Eagle, A. C., E. M. Hay-Chmielewski, K. T. Cleveland, A. L. Derosier, M. E. Herbert, and R. A.
Rustem, editors. 2005. Michigan's Wildlife Action Plan. Michigan Department of
Natural Resources. Lansing, Michigan, USA.
Eichenlaub, V. L. 1979. Weather and the climate of the Great Lakes Region. The University of
Notre Dame Press, Notre Dame, Indiana, USA.
Eichenlaub, V. L, J. R. Harman, F. V. Nurnberger, and H. J. Stolle. 1990. The climatic atlas of
Michigan. The University of Notre Dame Press, Notre Dame, Indiana, USA.
Ernst, C. H., and J. E. Lovich. 2009. Turtles of the United States and Canada. Johns Hopkins
University Press, Baltimore, MD, USA.
Flitz, B. A., and S. J. Mullin. 2006. Nest-site selection in the eastern box turtle, Terrapene
carolina carolina, in Illinois. Chelonian Conservation and Biology 5:309-312.
Floyd, T. M., K. R. Russell, C. E. Moorman, D. H. Van Lear R, D. C. Guynn Jr., and J. D.
Lanham. 2002. Effects of prescribed fire on herpetofauna within hardwood forests of
the Upper Piedmont of South Carolina: a preliminary analysis. U.S. Department of
Agriculture, Forest Service, Southern Research Station, Ashville, North Carolina,
General Technical Report SRS–48.
Gu, W., and R. K. Swihart. 2004. Absent or undetected? Effects of non-detection of species
occurrence on wildlife-habitat models. Biological Conservation 116:195–203.
59

Kapher, J. M., D. J. MuĂąoz, and T. Tomasek. 2012. Use of wildlife detector dogs to study
eastern box turtle (Terrapene carolina carolina) populations. Herpetological
Conservation and Biology 7(2):169-175.
Kéry, M. 2002. Inferring the absence of a species—a case study of snakes. Journal of Wildlife
Management 66:330–338.
Keyser, P. D., D. J. Sausviller, W. M. Ford, D. J. Schwab, and P. H. Brose. 2004. Prescribed fire
impacts to amphibians and reptiles in shelterwood-harvested oak-dominated forests.
Virginia Journal of Science 55(4):159-168.
Knapp, E. E., B. L. Estes, C. N. Skinner. 2009. Ecological effects of prescribed fire season: a
literature review and synthesis for managers. U.S. Department of Agriculture, Forest
Service, Pacific Southwest Research Station, Albany, California, General Technical
Report PSW-GTR-224.
Kost, M. A. 2004. Natural community abstract for mesic prairie. Michigan Natural Features
Inventory, Lansing, USA.
Lancia, R. A., W. L. Kendall, K. H. Pollock, and J. D. Nichols. 2005. Estimating the number of
animals in wildlife populations. Pages 106–153 in C. E. Braun, editor. Techniques for
wildlife investigations and management. The Wildlife Society, Bethesda, Maryland,
USA.
Mazarolle, M. J., L. L. Bailey, W. L. Kendall, J. A. Royle, S. J. Converse, and J. D. Nichols.
2007. Making great leaps forward: accounting for detectability in herpetological field
studies. Journal of Herpetology 41:672–689.
Means, D. B., and H. W. Campbell. 1981. Effects of prescribed burning on amphibians and
reptiles. Pages 89-97 in: G. W. Wood, editor. Prescribed fire and wildlife in southern
forests. The Belle W. Baruch Forest Science Institute, Clemson University,
Georgetown, South Carolina, USA.
O’Donnell, K. L., F. R. Thompson III, and R. D. Semlitsch. 2015. Prescribed fire and timber
harvest effects on terrestrial salamander abundance, detectability, and microhabitat
use. Journal of Wildlife Management 79(5):766-775.
Palmgren, G. 2004. Phase III natural resources management plan: Fort Custer State Recreation
Area (draft review). Parks and Recreation Bureau, Department of Natural Resource,
Lansing, USA.
Platt, S. G., H. Liu, and C. K. Borg. 2010. Fire ecology of the Florida box turtle (Terrapene
Carolina Bauri Taylor) in pine rockland forests of the lower Florida Keys. Natural
Areas Journal 30(3):254-260.

60

Refsnider, J. M., T. M. Mitchell, H. M. Streby, J. T. Stickland, D. A. Warner, and F. J. Janzen.
2011. A generalized method to determine detectability of rare and cryptic species
using the ornate box turtle as a model. Wildlife Society Bulletin 35(2):93-100.
Renken, R. 2005. Does fire affect amphibians and reptiles in eastern U.S. oak forests? Pages
158–166 in Proceedings of Fire in Eastern Oak Forests: Delivering Science to Land
Managers. Proceedings of a Conference, 15 November–18 November 2005,
Columbus, Ohio, USA.
Roloff, G. J. and T. L. Hmielowski. 2014. Ecological effects of fire in Great Lakes savannas and
prairies: literature review. Tallgrass Prairie Publication TPOS LR01-2014.
Russell, K. R., Van Lear, D. H., and D. C. Guynn Jr. 1999. Prescribed fire effects on
herpetofauna: review and management implications. Wildlife Society Bulletin 27:
374-384.
Russell, R. E., J. A. Royle, V. A. Saab, J. F. Lehmkuhl, W. M. Block, and J. R. Sauer. 2009.
Modeling the effects of environmental disturbance on wildlife communities: avian
responses to prescribed fire. Ecological Applications 19(5):1253-1263.
Smith, J. K., editor. 2000. Wildland fire in ecosystems: effects of fire on fauna. U.S. Department
of Agriculture, Forest Service, Rocky Mountain Research Station, Ogden, Utah,
General Technical Report RS-GTR-42-vol. 1.
van Dijk, P.P. 2011. Terrapene carolina. (errata version published in 2016). The IUCN Red List
of Threatened Species 2011. < http://www.iucnredlist.org/details/21641/0>. Accessed
10 February 2017.
Wilson, D.S., C. R. Tracy, and C. R. Tracy. 2003. Estimating age of turtles from growth rings: a
critical evaluation of the technique. Herpetologica: 59(2):178-194.

61

CHAPTER 3
DIRECT AND INDIRECT EFFECTS OF A GROWING SEASON PRESCRIBED FIRE
ON EASTERN BOX TURTLES IN SOUTHWESTERN MICHIGAN

Abstract
I examined the effects of a slow moving, growing season fire on the direct behavior and
subsequent survival of adult eastern box turtles in southwestern Michigan, 2015. The fire
occurred during the active season for box turtles (May), when adults have been active and above
soil for at least two weeks. I used plot-based pre- and post-burn measurements to quantify fire
characteristics and resulting effects on woody vegetation. I also directly observed behaviors of
telemetered turtles during the fire and documented mortality and habitat use for the remainder of
the active season. My results indicated that a growing season prescribed fire significantly
reduced small (≤1 cm diameter) woody stems, and resulted in cover of leaf litter <10% in burned
areas. The significant reduction of woody stems was consistent across all groups of vegetation,
including non-native, native fire-adapted, and native species that were not adapted to fire.
Greatest losses were observed for native, non-fire adapted (76% loss) and non-native (70% loss)
species. Fire adapted species were less impacted (53% loss). The growing season fire did not
result in a significant reduction of vegetation >1 cm diameter, although some individual species
declined (e.g., autumn olive, black locust, glossy buckthorn, hickory, black cherry, and
sassafras). Box turtles exhibited varying behaviors in response to the fire including burying and
actively evading and negotiating the flame front. One turtle died from injuries sustained in the
fire and females within the management unit where prescribed fire was applied were not
observed nesting that year.

62

3.1. Introduction
Prescribed fire is used to delay vegetation succession, restore historical disturbance processes,
recycle nutrients, manage wildlife habitat, and control exotic invasive vegetation (Knapp et al.
2009). The severity, uniformity, and spatiotemporal extents of prescribed fire influences
vegetation pattern, productivity, and corresponding behavioral responses of fauna located within
treated areas (Smith 2000). Although prescribed fire can be an effective and inexpensive tool for
vegetation management, direct (including injury and mortality) and indirect (including changes
in body condition and animal movements or interactions) effects on herpetofauna are poorly
understood (Russell et al. 1999; Keyser et al. 2004). Studies to date have generally focused on
short-term pre- and post-fire abundance and species richness (Griffiths and Christian 1996; Cole
et al. 1997; Moseley et al. 2003; Keyser et al. 2004; Greenburg and Waldrop 2008). Some
studies indicated that fires have few direct effects on local herpetofauna (Means and Campbell
1981; Russell et al. 1999; Floyd et al. 2002), and that species historically occurring in fire
dependent ecosystems have adaptations that reduce negative fire impacts (Erwin and Stasiak
1979; Means and Campbell 1981; Driscoll and Henderson 2008; Smith 2000). In drawing these
conclusions, researchers have frequently grouped reptiles and amphibians into herpetofauna
(Means and Campbell 1981; Smith 2000; Floyd et al. 2002; Renken 2005; Platt et al. 2010). This
grouping tends to ignore differences in behavioral responses by individual species that may
affect survivorship. More importantly, this grouping fails to acknowledge the ramifications of
losing individuals from populations of long-lived species with delayed reproductive maturity and
potentially biases the analysis towards the responses (often abundance) of r-selected species.
Eastern box turtles (Terrapene carolina. carolina) historically occupied habitats
subjected to fire throughout eastern North America (Russell et al. 1999), including grasslands,

63

oak (Quercus spp.) dominated woodlands, and more mesic ecosystems (Dodd 2001; Spencer and
Thompson 2003). Although once common (Ernst and Lovich 2009), eastern box turtles are now
recognized as a species of conservation concern throughout much of their geographic range
(Swarth and Hagood 2004; van Dijk 2011). In Michigan, eastern box turtles are documented to
occur in 36 counties, with last confirmed occurrence dating to 1933 for some observations
(MNFI 2017). In some of these counties, areas are managed with prescribed fire to restore firedependent ecosystems, control invasive vegetation, and hinder woody encroachment.
Research suggests that losing even small numbers of breeding adults from k-selected,
long-lived populations like box turtles can result in irreparable harm to population viability
(Congdon et al. 1993; Dodd et al. 2015). Although eastern box turtles may have evolved with
fire, current turtle population sizes, fire regimes, and habitat configurations differ from historical
conditions. To accomplish ecosystem restoration goals, managers often use prescribed fire
regimes that differ from historical regimes in frequency, time of year, and spatial extent. These
fire regimes have direct effects on eastern box turtles including high rates of mortality and injury
from exposure to fire (Allard 1949; Babbitt and Babbitt 1951; Bigham et al. 1964; Dolbeer 1969;
Gibson 2009). Indirect effects may include changes in prey abundance, vegetation used for
thermoregulation, camouflage, and insulating duff layers during overwintering, and potential
disruption in behavior or range (Gibson 2009). Box turtles are particularly sensitive to
environmental variables that affect ground cover (including vegetation, litter, and subsurface
soils) because their entire life history depends on ground conditions (Dodd 2001). For box
turtles, thermoregulation determines individual responses to habitat changes (Currylow et al.
2012). For example, Currylow et al. (2012) evaluated box turtle movement and thermal ecology

64

for two years following timber harvest and found that movements were shorter and more
frequent after timber harvest.
Because prescribed fire is a popular management technique in Michigan, understanding
both the efficacy of fires in achieving vegetation goals and in minimizing box turtle injury,
mortality, and survivorship is fundamental to conservation of this species. The general consensus
regarding fire and eastern box turtle conservation is to restrict burning to the early spring months,
before turtles have emerged from overwinter hibernacula (see Woodley 2013), or to late fall once
box turtles have entered subterranean hibernacula. As early spring is ambiguous, and annual
weather conditions vary, spring prescribed fires in Michigan can potentially coincide with box
turtle emergence. When emerging from hibernacula, box turtles are lethargic and have low
energy reserves (Dodd 2001; Woodley 2013). Woodley (2013:23) found that box turtles in
Michigan “often stayed within 1 to 5 meters of their burrow for 1 to 2 weeks after emerging –
seeming to indicate that turtles need a substantial interval of time to recover full mobility after
emerging in the spring”, and that they would be “especially vulnerable during this time to any
surface disturbance”. Early spring prescribed fires conducted in landscapes where box turtles
may overwinter, such as upland oak ecosystems with heavy duff layers, can be fast moving and
intense. Thus, early spring fires, if timed to box turtle emergence, can have substantial negative
impacts on turtle populations. Fires later in spring or into the growing season, after box turtles
have become fully active, may allow turtles to evade slow moving flame fronts.
I examined the effects of a slow moving, growing season fire on the behavior and
subsequent survival of adult eastern box turtles. This fire occurred during the active season for
box turtles, when adults have been active and above soil, outside hibernacula for at least two
weeks. I used pre- and post-burn behavioral observations from telemetered turtles, and fire and

65

vegetation plots to quantify fire effects. My research objectives were to: 1) quantify the efficacy
of the growing season burn in reducing woody succession and controlling invasive plants, 2)
describe behaviors of individual turtles as the flame front advanced, 3) assess direct mortality of
turtles from the burn, and 4) quantify short-term indirect fire effects, such as changes in behavior
and survivorship. My results represent one of the only known studies on eastern box turtle
behaviors during a growing season prescribed fire, and offer insights into post-fire habitat use
and survivorship.
3.2. Methods
3.2.1. Study Area
This study was conducted at Fort Custer State Recreation Area (FCSRA; 1,226 ha), located in
the southern lower peninsula of Michigan between the cities of Battle Creek and Kalamazoo,
USA. The FCSRA had heterogeneous vegetation caused by differences in topography, climate,
and human influence including high levels of agricultural and urban development (Eagle et al.
2005; Albert 1995). Most of the soils are calcareous and loamy, derived from underlying
limestone, shale, and sandstone bedrock (Eagle et al. 2005). Glacial till deposits are primarily
loams, silt loams and clay loams (Eagle et al. 2005). The climate is considered humid
continental, although FCSRA is strongly influenced by the Maritime Tropical air mass and
proximity to Lake Michigan, resulting in warmer temperatures, moderated inland temperature
fluctuations, and induced lake-effect snow (Eichenlaub 1979; Denton 1985; Albert et al. 1986;
Eichenlaub et al. 1990). The average length of the growing season was 154 days (Albert et al.
1986). The mean minimum and maximum temperatures during the study (May to October, 2013
- 2015) ranged from -1.1 – 7.2°C and 22.2 – 34.4°C, respectively. FCSRA and the adjacent Fort

66

Custer Training Center (a 3,063 ha federally owned military reservation) represent one of the
largest contiguous blocks of public land in southwest Michigan (Cohen et al. 2009).
The FCSRA was historically dominated by fire-dependent oak (Quercus spp) savanna
and prairie (Albert 1995), and contained the only extensive areas of mesic prairie found in
Michigan (Kost 2004). In the early 1800’s, plant communities included dry and dry-mesic
southern forest (oak-hickory (Carya spp)), oak barrens (mixed oak savanna), emergent marsh,
and southern (mixed hardwood) swamp (Palmgren 2004). Currently, intensive agriculture is
concentrated around FCSRA because of the relatively mild climate and productive soils (Eagle et
al. 2005). Natural vegetation in this region is broadly classified as black oak-white oak (Q.
velutina-Q. alba) savannas and forests on droughty soils, and beech-sugar maple (Fagus
grandifolia-Acer saccharum) forests on more mesic sites (Palmgren 2004). Currently, FCSRA
supports degraded patches of oak barrens and prairie openings, scattered oak-hickory forests, and
large blocks of non-native black locust (Robinia pseudoacacia), surrounded primarily by
agriculture (Palmgren 2004; Eagle et al. 2005). Restoration activities on FCSRA included use of
prescribed fire to maintain and improve savanna and prairie ecosystems. Prescribed fires were
generally small to medium-sized (1.6 to 414.4 ha), conducted in spring (before leaf out, typically
during the months of March and April), and occurred in 1 to 4-year intervals followed by
herbicide treatments in some areas.
The FCSRA is located within the most northerly range of the eastern box turtle. I chose
FCSRA for this project because it has been the focus of ecological restoration activities using
prescribed fire since 1997. My research focused on a 31.6 ha management unit in FCSRA
(Figure 3.1). The unit included dense patches of invasive black locust, emergent marsh, drymesic southern (oak-hickory) forest, open grassland, red pine (Pinus resinosa) plantation, and

67

mixed lowland hardwoods (mesic southern forest, maple-beech dominated). This unit included a
12.5-hectare emergent marsh with open water (Figure 3.1). The unit was previously burned
during early spring (typically the month of March) in 2001, 2003, 2006, and 2008. It was also
partially burned in 2007.
3.2.2. Vegetation and Fire Data Collection
Pre- and post-fire vegetation information was collected using a grid of semi-permanent 2 m
radius circular plots spaced 40 m apart (n = 80). I marked the center of each plot using a wooden
stake and recorded the location with a handheld GPS. I recorded woody stem count per species
and growth class; growth classes included ≤1 cm, 1 to ˂3 cm, and ≥3 cm. I also recorded litter
depth. At the center of each vegetation plot, I placed a set of 3 Omega TL-10 adhesive, nonreversible labels that recorded fire temperatures (OMEGA Engineering, INC., Stamford,
Connecticut) ranging from 87 – 260° C (Figure 3.2). Temperature labels were affixed to a
rectangle (11 x 20 cm) of aluminum roof flashing. The rectangle was bent to a 90-degree angle
and fixed into the soil so that the side with the temperature labels ran parallel to and 4 cm above
the soil surface. I recorded rate of fire spread by observing the time at which the flame front
reached different plot centers. Average char height for each wooden stake was recorded in
centimeters and flame height was estimated using direct observation during the fire.
Completeness of burn was recorded as percentage of charred vegetation at each vegetation plot
48 hrs after the burn. Pre-fire vegetation sampling was completed in June 2014, and post-fire
vegetation sampling was conducted in September 2015.
3.2.3. Radio-telemetry and Observation of Eastern Box Turtles During the Burn
I used meandering visual encounters and wildlife detector dogs to find adult eastern box turtles
during daylight hours in May of 2012 and June of 2013. Captured adult box turtles were radio-

68

tagged with a Holohil R1-2B 14.5g Transmitter (Holohil Systems Ltd., Carp, Ontario, Canada)
attached to the right or left anterior pleural carapacial scutes using multi-purpose 5-minute set
epoxy putty (LoctiteÂŽ, Henkel 289 Corporation, Cary, NC, USA). I estimated minimum age by
counting annual rings on the carapace (Legler 1960), and permanently marked turtles by
notching the shell (Cagle 1939). I visually located radio-tagged turtles 1-3 times per week during
the active seasons (April through October) from May 2012 to August 2015 using portable
receivers (Advanced Telemetry Systems, Isanti, MN, USA).
I located telemetered turtles 24 hrs before the research burn at which point transmitters
were spray-painted with RUST-OLEUM™ High-Heat Spray to prevent fire damage. I covered
the carapace with a spray guard to prevent heat resistant paint from adhering to the turtle. Upon
fire ignition, I relocated research animals from behind the flame front using radio-telemetry and
subsequently observed fire-related behaviors. I observed turtles from 5-10 m away to minimize
influence on behavior. I subsequently relocated the turtles 12, 24 and 48 hours after the burn and
assessed each for direct injury and mortality; thereafter I checked turtles 1-3 times per week. All
animals were handled in compliance with Michigan State University’s Institutional Animal Care
and Use Committee guidelines (IACUC Approval # 04/14-077-00).
3.2.4. Data Analysis
I summarized fire characteristics using descriptive statistics. I also reported on average stem
counts by species for 3 size classes; ≤1 cm, >1 and <3 cm, and ≥3 cm for pre- and post-fire
conditions. To test for stem count differences between pre- and post-fire, I used a Hotelling’s T2
that tests for the differences in a multivariate vector (in this case species counts) across treatment
types (Mardia et al. 1979).

69

3.3. Results
3.3.1. Fire Characteristics
The MDNR implemented the research burn on May 20, 2015, corresponding to leaf out and the
early part of the growing season. Average flame lengths were 8 cm (SE = 0.63), average rate of
spread was 1.5 m/min (SE = 0.19), and average maximum temperature was 149.3° C (SE = 7.9).
Approximately 16.2 ha of the 31.6 ha management unit burned, resulting in a mosaic of burned
and unburned upland habitats and an unburned emergent marsh. Additionally, fire failed to carry
into patches of dense locust and mixed lowland hardwoods where little fine fuels were available.
3.3.2 Direct Effects of Fire on Woody Vegetation and Leaf Litter
Pre-fire woody vegetation (based on 52 plots) was a mixture of non-native (Table 3.1), native
fire-adapted (Table 3.2), and other native species found in dry mesic southern forests (Table 3.3).
On average, the most prevalent small diameter (≤1 cm) stems included native species; red osier
dogwood, black cherry, Viburnum spp., black raspberry, and hickory (Table 3.2). Medium-sized
(1 < diameter < 3 cm) stems were mostly dominated by non-native species including autumn
olive, black locust, honeysuckle, hickory, and glossy buckthorn (Table 3.1). The most prevalent
largest woody stems (≥3 cm diameter) were autumn olive, black locust, black cherry, hickory,
and white oak (Tables 3.1, 3.2, 3.3). Leaf litter cover in pre-fire plots that eventually carried fire
was 84.5% (SE = 3.3, range = 5-100).
Only 52 of 82 (63%) vegetation plots burned during the May 2015 prescribed fire. The
fire reduced average leaf litter cover by 77.3%, resulting in average post-fire leaf litter cover of
7.3% (SE = 1.6, range 0-50%). Non-native stems ≤1 cm were reduced by 59% (Table 3.1), and
this was a significant reduction (Hotelling’s T2, F = 2.25, P = 0.01). Counts of larger non-native

70

stems (1-3 cm and ≥3 cm diameter) were reduced (e.g., 24% reduction for 1-3 cm stems), but
this was not a significant decline (Hotelling’s T2, F < 0.43, P > 0.91).
Stem counts of native, fire adapted species ≤1 cm diameter were significantly reduced
(Table 3.2; Hotelling’s T2, F = 3.64, P < 0.001). No differences in stem counts of native, fireadapted species were identified for medium or large stems (Hotelling’s T2, F < 0.39, P > 0.79).
Lastly, stems counts of native, non-fire adapted species ≤1 cm were significantly reduced (Table
3.3; Hotelling’s T2, F = 3.64, P < 0.001), whereas 1-3 cm and ≥3 cm were not significantly
reduced (Hotelling’s T2, F < 0.11, P > 0.97). Total counts of small stems (≤1 cm diameter) were
reduced by 67%. Total counts of medium stems (1-3 cm diameter) were reduced by 27%. Total
counts of larger stems (≥3 cm diameter) were reduced by 17%. I did not measure the effects of
vegetative decreases on prey abundance or thermoregulation directly.
3.3.3. Direct Fire Effects on Box Turtle Behavior
Prior to fire ignition, 11 turtles occupied the management unit but 6 of those were in a 12.5-ha
emergent marsh and unaffected by the flame front. An additional radio-tagged turtle was within a
dense black locust patch that did not burn. The other 4 turtles were directly subjected to the
flame front. These turtles exhibited varying behaviors in response to the fire. One began to bury
into mineral soil and stopped once the flame front slowed and self-extinguished before reaching
her location (roughly 3 m away), 2 turtles completely buried into mineral soil (approximately 13
cm down), and one remained above ground and moved 1-2 m from her original location. This
individual was overtaken by the flame front, burning 90-100% of her carapace. Subsequently,
this turtle moved to an unburned black locust patch within 12 hrs and then to the emergent marsh
where she remained until dying two weeks later.

71

The 3 box turtles that were observed digging into mineral soil reacted to the advancing
flame front when it was 10-20 m away. Two of these turtles were in grassland within the
management unit, and one was on the edge of a southern mesic forest and grassland. These
turtles actively negotiated the flame front by moving away quickly and then began to dig into
mineral soil. These behaviors resulted in one female outrunning the flame front to an area of low
fuels and partially burying into mineral soil in a mesic forest, and one male completely burying
into mineral soil in a grassland area. A female that actively negotiated the flames was eventually
overtaken by two flame fronts that converged on her from fires that were lit separately in a
grassland area. A firefighter picked up this female and placed her in an area that had already
burned to prevent her from being completely overtaken by flames. Once in the burned area she
immediately buried herself in mineral soil under an unburned decaying log. The fourth female
remained in form, or within its shell, until the flame front was within 5 m and was overtaken by
the flame front in a grassland area.
I failed to observe any turtle mortalities on the day of the burn. My observations indicated
that when flame fronts were slow moving (1.5 m/minute), adult eastern box turtles detected
oncoming flame fronts 10-20 m away. Whether they detected oncoming flame fronts by sight,
sound, or smell is unclear, although, I observed the individual in the mesic forest becoming alert
and arching her head out of her shell when it seemed she could only detect the fire by its
crackling sound, or smell, as it was 20 m away and had short (8 cm) flame lengths. I observed
adult turtles successfully negotiating and evading this slow-moving fire, and I also observed one
adult turtle completely fail to react to the fire, resulting in death. If burn refugia (soft mineral
soil, or stumps and logs that will not burn) are accessible, converging flame fronts do not restrict

72

movements (fires are not lit using a strip method), and fires remain creeping (not grassland fires),
some box turtles can safely negotiate prescribed fires during the active season.
3.3.4. Indirect Fire Effects on Box Turtles
I defined indirect effects of prescribed fire on eastern box turtles as potential changes in behavior
or space use not associated directly with exposure to flames. Disruptions in behavior or space use
could be indirectly related to fire effects on prey abundance and vegetation used for
thermoregulation or concealment (Gibson 2009), for example. I observed behavioral changes in
adult box turtles that included long-term burying, potential long-term avoidance of burned areas,
and failure to stage or nest. Adult box turtles that exhibited a burying strategy to cope with the
fire (n=3) remained buried for 12 hrs following fire extinguishment. These adults subsequently
began migrating to the emergent marsh, which occurred within the home ranges of these turtles.
Adults traveled between 0 to 160.6 m to reach this marsh. Regardless of their exposure to the
flame front, marked turtles remained soaking in the emergent marsh, or within 3-5 m of its edge,
for 1-3 weeks following the fire.
When moving back to burned areas from the emergent marsh, turtles remained within 1-2
m of unburned habitats that included decaying logs, patches of leaf litter, or other areas
unaffected by fire. They also seemed to remain in these areas until post-burn vegetation growth
provided cover. Soaking in the wetland in late May and early June was not abnormal behavior, as
I observed similar behavior in years prior to the burn. The year of the prescribed fire, female
turtles that were in the management unit during the fire did not nest. Typical nesting season
started in early to mid-June.

73

3.4. Discussion
My results indicated that a growing season prescribed fire significantly reduced small (≤1 cm
diameter) woody stems, and resulted in cover of leaf litter <10%. The significant reduction of
small woody stems was consistent across all species groupings of vegetation, including nonnative, native fire-adapted, and native species that were not adapted to fire. Greatest losses were
observed for native, non-fire adapted (76% loss) and non-native (70% loss) species. Fire adapted
species were less impacted (53% loss). The growing season fire did not result in a significant
reduction of vegetation >1 cm diameter, although some individual species declined (e.g., autumn
olive, black locust, glossy buckthorn, hickory, black cherry, and sassafras). It is generally
accepted that prescribed fire is not a reliable tool for killing belowground perennating tissues
(Rice and Smith 2008), and that fire frequency integrated with other management techniques
such as herbicide or manual removal is the most effective strategy for controlling non-native
species. Treatment efficacy, however, is related to both the biology of the individual species,
treatment(s) timing, and spatial extent of the invasive species (Richburg and Patterson 2003).
Growing season prescribed fires occur when plants are divesting energy into aboveground
structures and hence may be more effective at controlling certain species. Growing season fires
can help mimic the historical seasonality and range of variability of fire in certain landscapes. As
part of an overall burn regime, growing season fires can increase pyrodiversity, or the mix of fire
seasons, burn intensity, and frequency. This can ultimately lead to higher biodiversity (Howe
1994; Copeland et al. 2002). Repeated early spring burns in dry-mesic southern forests may fail
to set back woody species and even cause rigorous re-sprouting of certain woody species if the
fire is conducted before leaf out (Cohen et al. 2015). Managers seeking to manipulate restoration
regimes to include growing season fires should carefully consider the heterogeneity of fuels

74

within management units, as growing season fires that spread to grasslands can cause box turtle
mortality (Melvin 2017: Chapter 1).
Although variable within the management area, I observed that growing season burns
conducted during leaf-out coincided with some adult eastern box turtles occupying wetlands.
Hence, these turtles were not directly impacted by the burn. For those turtles in the burn area,
fires conducted when the turtles were fully active allowed responsive behaviors such as burying
and seeking refugia in decayed logs and wet areas. I observed successful evasive behavior when
flame lengths averaged approximately 8 cm and average rate of spread was approximately 1.5
m/min. My observations indicated that eastern box turtles reacted to fire and could evade fire,
but I also observed mortality. Box turtles were unable to evade fast-moving fires, or gain access
to fire refugia when converging flame fronts restricted movements.
Because this was a relatively small, single burn in a heterogeneous landscape that directly
affected a small number of turtles, further work that determines the indirect and direct effects of
fire in different vegetation types is important. Longer term studies on the relationships between
eastern box turtles and fire seasonality, ignition type, ignition time of day (box turtles are mainly
active during morning and evening hours in hot summer months), and intensity are crucial to
conservation in areas subjected to fire. Research relating fire with indirect effects such as
changes in disease prevalence, diminishing health, winter refugia site selection, nesting habits
and home range are critically needed. One of the main issues with prescribed fire and box turtle
conservation is how long-term mortality relates to fire frequency.
Given that the home range of an animal should, at least partially be an expression of
fitness (Roloff and Haufler 1997, 2002), it follows that changes in home range space use can be
used to measure animal response to environmental perturbations. The habitat quality for box

75

turtles appears to directly link to the abundance of leaf litter (Dodd 2001; Weiss 2009) and
hence, fire removal of leaf litter should lower the quality of box turtle habitat, at least in the short
term. If frequent burns maintain low leaf litter throughout a box turtle home range, it seems
reasonable to assume a negative fitness consequence. Lack of significant changes in home
ranges before and after prescribed fire events might indicate that turtles essentially remained in
unburned patches surrounded by temporarily unsuitable burned habitat. Gibson (2009) noted that
turtles not directly injured by fire maintained the extent of their home ranges following fire, but
these turtles changed their space use within the home range to restricted patches of unburned leaf
litter, which I also observed. This behavior could result in lower body index values throughout
the growing season after a fire and behavior changes in selecting winter refugia sites, movements
the following year, susceptibility to disease and indirect mortality.
Little data exist on the effects of prescribed fire on terrestrial turtle nests or hatchlings; a
research topic that is hindered by lack of a cost-effective methodology for detecting terrestrial
hatchlings in heavily managed areas. One of the biggest challenges facing box turtle
conservation is documenting and understanding the factors that affect hatchling survival.
Because hatchling box turtles are cryptic, secretive, and small, little data exists on seasonal
movements, detection rates, or prescribed fire mortality. Hatchlings appear to hide under litter,
which exposes them to fire, rather than burrowing or creating forms (Ernst et al. 1995), which I
also observed. Because box turtles typically nest in open, grassy areas that can be subjected to
frequent fire, and because hatchlings tend to stay relatively close to their nest site, further
understanding seasonal movements and fire mortality is critical to overall conservation of this
species.

76

My study is one of few that observed direct fire effects, post-fire behaviors, and
vegetation effects of a growing season fire on marked terrestrial turtles. Eastern box turtles tend
to occupy ecosystems, at least partially during the year, that were historically maintained by fire
thus fire is often prescribed as a restoration tool in these ecosystems. Some managers may argue
that because box turtles are found in areas once maintained by prescribed fire, they must
inherently be a fire-adapted species. Although I observed evasive behavior to oncoming flame
fronts, I also observed an individual not react. I caution that this study only observed 4
individuals and their reactions to fire. In this case, 25% of turtles encountering flame fronts did
not react. I also caution that fire-adapted species, or exhibiting fire adapted behaviors, does not
necessarily constitute eastern box turtles being fire dependent. For example, Willey and Sievert
(2012:367) found large ranges of canopy cover (0-65%) within 5 m of eastern box turtle nests (n
= 34) in Massachusetts, and noted that turtles “spent more time near the nest area if woody or
herbaceous material in which to hide and forage was present”. Eastern box turtles may nest in
smaller forest openings when larger grasslands are not available, and making broad overarching
consensuses that box turtles are fire dependent may actually result in more direct fire-related
mortality of breeding adult females.
For species of conservation concern, like box turtles, managers are often tasked with
restoring vegetation types to earlier successional stages, but tend to lack empirically based
guidance on how these management techniques affect this species. Although my study provides
that guidance, I caution that my results are based on a single prescribed fire in one management
unit during the early growing season. The effects of growing season fire on eastern box turtles
and vegetation likely varies by fire intensity and the amount and spatial distribution of vegetation

77

and debris. I encourage additional long-term studies on known cohorts of box turtles, in other
fire and vegetation conditions, and studies that include effects on body condition.
3.5. Management Implications
Managers should avoid burning grasslands when adult female box turtles are seeking nesting
sites, between mid-May and mid-June for southwestern Michigan. I also recommend that
managers avoid firing types that result in converging flame fronts or fast-moving fronts when
burning dry-mesic southern forests. Managers should also be aware that the seasonality of fire
likely plays a large role in minimizing box turtle mortality. Early spring fires, if timed to box
turtle emergence, have the potential to cause mortality when box turtles are lethargic after
emerging from winter hibernacula. Fires later in spring or into the growing season, after box
turtles have become fully active, may allow turtles to evade slow moving flame fronts. When
planning for growing season fires, managers should consider plant phenological characteristics,
such as the emergence and growth stages of leaves on certain deciduous tree species, the
appearance and flowering of specific forest floor species along with specific weeks or days of the
month when labeling fire seasonality as “spring” or “growing season”, as the timing of these
events can change between years. I also recommend patchy fires that leave refugia such as
decaying logs and unburned patches around seeps or wetland areas to accommodate the indirect
behavioral effects of seeking long-term refuge.
3.6. Acknowledgements
Summer and academic funding for this project was provided by Michigan State University and
the Joint Fire Science Program under project JFSP 14-3-01-30. Funding for equipment and
supplies was provided by the Michigan Department of Natural Resouces - Parks and Recreation
Division, Michigan Society of Herpetologists, The North American Box Turtle Conservation

78

Committee, and the Michigan Chapter of The Safari Club International. I thank my entire field
crew from 2013 to 2015 for assistance in the collection of field data and the Michigan DNR
Parks and Recreation Division fire crew for their incredible help and for running our research
burn. Special thanks to M. Bonnau from Michigan State University and H. Tripp from Auburn
University for assisting with box turtle work, to A. Ihnken, G. Palmgren, R. Clancy and R.
Fahlsing for FCSRA and box turtle expertise. A special thanks to P. Melvin for assistance with
creating field equipment and for moral support, and Dr. G. Roloff for assistance in data analysis
and writing.

79

APPENDIX

80

Table 3.1. Pre- and post- fire average stem counts (SE) within 2 m radius circle plots (n=52) by diameter size classes for non-native
woody species. These plots carried an early growing season (May 20, 2015) prescribed fire in a 31.6 ha management unit at Fort
Custer State Recreation Area, Michigan, USA. Pre-fire vegetation sampling was conducted in June 2014, and plots were visually
inspected in May 2015 to verify that major changes to the plant community (e.g., conversion) did not occur. Post-fire vegetation
sampling was completed in September, 2015.
Woody Stem Diameter Class
≤ 1 cm
Non-native Species

≥ 3 cm

1 < diameter < 3 cm

Pre-fire

Post-fire

Pre-fire

Post-fire

Pre-fire

Post-fire

Autumn olive

Elaeagnus umbellata

0.90 (0.40)

0.33 (0.15)

0.69 (0.24)

0.56 (0.20)

0.48 (0.26)

0.42 (0.21)

Black locust

Robinia pseudoacacia

0.41 (0.18)

0.06 (0.04)

0.27 (0.11)

0.21 (0.10)

0.23 (0.15)

0.23 (0.15)

Glossy buckthorn

Frangula alnus

1.57 (0.29)

0.76 (0.21)

0.13 (0.07)

0.04 (0.03)

0.06 (0.04)

0.04 (0.03)

Common buckthorn

Rhamnus cathartica

1.45 (0.83)

0.96 (0.76)

0.02 (0.02)

0.02 (0.02)

0.00

0.00

Bush honeysuckle

Lonicera spp.

0.55 (0.18)

0.08 (0.08)

0.19 (0.14)

0.08 (0.05)

0.00

0.00

Multiflora rose

Rosa multiflora

0.55 (0.23)

0.12 (0.07)

0.04 (0.04)

0.04 (0.04)

0.00

0.00

Bittersweet

Celastrus orbiculatus

0.18 (0.12)

0.02 (0.02)

0.00

0.00

0.00

0.00

81

Table 3.2. Pre- and post- fire average stem counts (SE) within 2 m radius circle plots (n=52) by diameter size classes for native, fire
adapted woody species. These plots carried an early growing season (May 20, 2015) prescribed fire in a 31.6 ha management unit at
Fort Custer State Recreation Area, Michigan, USA. Pre-fire vegetation sampling was conducted in June 2014, and plots were visually
inspected in May 2015 to verify that major changes to the plant community (e.g., conversion) did not occur. Post-fire vegetation
sampling was completed in September, 2015.
Woody Stem Diameter Class
≤ 1 cm
Native, Fire Adapted Species

≥ 3 cm

1 < diameter < 3 cm

Pre-fire

Post-fire

Pre-fire

Post-fire

Pre-fire

Post-fire

Shagbark hickory

Carya ovata

1.63 (0.63)

1.04 (0.21)

0.17 (0.09)

0.13 (0.07)

0.13 (0.06)

0.12 (0.05)

Red oak

Quercus rubra

0.75 (0.19)

0.33 (0.13)

0.02 (0.02)

0.00

0.12 (0.07)

0.12 (0.07)

White oak

Quercus alba

1.10 (0.41)

0.57 (0.17)

0.08 (0.08)

0.08 (0.08)

0.12 (0.05)

0.12 (0.05)

Red osier dogwood

Cornus sericea

6.49 (1.29)

1.78 (0.48)

0.00

0.00

0.00

0.00

82

Table 3.3. Pre- and post- fire average stem counts (SE) within 2 m radius circle plots (n=52) by diameter size classes for native,
woody species not adapted to fire. These plots carried an early growing season (May 20, 2015) prescribed fire in a 31.6 ha
management unit at Fort Custer State Recreation Area, Michigan, USA. Pre-fire vegetation sampling was conducted in June 2014, and
plots were visually inspected in May 2015 to verify that major changes to the plant community (e.g., conversion) did not occur. Postfire vegetation sampling was completed in September, 2015.
Woody Stem Diameter Class
≤ 1 cm
Native, Species Not Adapted to Fire

≥ 3 cm

1 < diameter < 3 cm

Pre-fire

Post-fire

Pre-fire

Post-fire

Pre-fire

Post-fire

Black cherry

Prunus serotina

2.43 (0.61)

0.82 (0.22)

0.12 (0.06)

0.12 (0.06)

0.15 (0.08)

Wild black raspberry

Rubus occidentalis

1.69 (0.43)

0.18 (0.10)

0.00

0.00

0.00

0.00

Choke cherry

Prunus virginiana

1.49 (0.37)

0.25 (0.14)

0.00

0.00

0.00

0.00

Cottonwood

Populus deltoides

0.10 (0.07)

0.10 (0.07)

0.06 (0.06)

0.06 (0.06)

0.06 (0.04)

0.06 (0.04)

Black currant

Ribes americanum

0.37 (0.14)

0.04 (0.04)

0.00

0.00

0.00

0.00

Hophornbeam

Ostrya virginiana

0.51 (0.17)

0.20 (0.09)

0.00

0.00

0.00

0.00

Hawthorn

Crataegus spp.

0.00

0.00

0.02 (0.02)

0.02 (0.02)

0.00

0.00

Mulberry

Morus rubra

0.02 (0.02)

0.00

0.00

0.00

0.00

0.00

83

0.12 (0.06)

Table 3.3. (cont’d).
Woody Stem Diameter Class
≤ 1 cm
Native, Species Not Adapted to Fire

≥ 3 cm

1 < diameter < 3 cm

Pre-fire

Post-fire

Pre-fire

Post-fire

Pre-fire

Post-fire

Sassafrass

Sassafras albidum

0.31 (0.13)

0.12 (0.08)

0.10 (0.08)

0.02 (0.02)

0.04 (0.03)

0.02 (0.02)

Serviceberry

Amelanchier arborea

0.29 (0.27)

0.00

0.00

0.00

0.00

0.00

Staghorn sumac

Rhus typhina

0.59 (0.35)

0.24 (0.18)

0.10 (0.10)

0.10 (0.10)

0.02 (0.02)

0.02 (0.02)

Sugar maple

Acer saccharum

0.69 (0.22)

0.02 (0.02)

0.06 (0.06)

0.06 (0.06)

0.00

0.00

Viburnum

Viburnum spp.

2.33 (0.75)

0.94 (0.55)

0.02 (0.02)

0.02 (0.02)

0.00

0.00

American elm

Ulmus americana

1.10 (0.41)

0.00

0.04 (0.03)

0.02 (0.02)

0.02 (0.02)

0.02 (0.02)

Wild grape

Vitis spp.

0.71 (0.27)

0.00

0.00

0.00

0.00

0.00

84

Figure 3.1. A 31.6 ha management unit consisting of a dense patch of invasive black locust (A),
a 12.5 ha emergent marsh (B), dry-mesic southern (oak-hickory) forest (C), open grassland (D),
red pine plantation (E), and mixed lowland hardwoods (mesic southern forest, maple-beech
dominated) (F) at Fort Custer State Recreation Area, Michigan, USA.

0

150

300 Meters

85

Figure 3.2. An example of a set of 3 Omega TL-10 adhesive, non-reversible labels that indicated
temperature (OMEGA Engineering, INC., Stamford, Connecticut) ranging from 87 – 260 ° C.
Labels were affixed to a rectangle (11 cm x 20 cm) of aluminum roof flashing. The rectangle was
bent to a 90-degree angle and fixed into the soil so that the side with the temperature labels ran
parallel to and 4 cm above the soil surface. This set of labels indicated temperatures reached 88°
C.

86

LITERATURE CITED

87

LITERATURE CITED

Albert, D. A. 1995. Regional landscape ecosystems of Michigan, Minnesota, and Wisconsin: A
working map and classification. USDA, Forest Service, North Central Forest
Experiment Station, St. Paul, Minnesota, General Technical Report NC-178.
Albert, D. A., S. R. Denton, and B. V. Barnes. 1986. Regional landscape ecosystems of
Michigan. University of Michigan, School of Natural Resources, Ann Arbor, USA.
Allard, H. A. 1949. The Eastern Box Turtle and its behavior. Journal of the Tennessee Academy
of Science 24:146–152.
Babbitt, L. H., and C. H. Babbitt. 1951. A herpetological study of burned-over areas in Dade
County, Florida. Copeia 1951:79.
Bigham, S. R., J. L. Hepworth, and R. P. Martin. 1964. A casualty count of wildlife following a
fire. Proceedings of the Oklahoma Academy of Science 45:47–50.
Cagle, F. R. 1939. A system of marking turtles for future identification. Copeia 1939:170–173.
Cohen, J. G., R. P. O’Connor, B. J. Barton, D. L. Cuthrell, P. J. Higman, and H. D. Enander.
2009. Fort Custer vegetation and natural features survey 2007-2008 report. Michigan
Natural Features Inventory, Report Number 2009-04, Lansing, MI, USA.
Cohen J. G., P. J. Badra, H. D. Enander, Y. Lee, J. M. Lincoln, and M. J. Monfils. 2015. Natural
Features Inventory and Management Recommendations for Lost Nation State Game
Area. Michigan Natural Features Inventory Report Number 2015-05, Lansing, MI,
USA.
Cole, E. C., W. C. McComb, M. Newton, C. L. Chambers, and J. P. Leeming. 1997. Response of
amphibians to clearcutting, burning, and glyphosate application in the Oregon coast
range. Journal of Wildlife Management 61(3):656-664.
Congden, J. D., A. E. Dunham, and R. C. Van Loben Sels. 1993. Delayed sexual maturity and
demographics of Blanding’s turtles (Emydoidea blandingi): implications for
conservation and management of long-lived organisms. Conservation Biology 7:826833.
Copeland, T. E., W. Sluis, and H. F. Howe. 2002. Fire season and dominance in an Illinois
tallgrass prairie restoration. Restoration Ecology 10:315-323.
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):e40473.

88

Denton, S. R 1985. Ecological climatic regions and tree distributions in Michigan. Dissertation,
University of Michigan, Ann Arbor, USA.
Dodd, C. K. Jr. 2001. North American box turtles: a natural history. University of Oklahoma
Press, Norman, USA.
Dodd, C. K., Jr., V. Rolland, and M. K. Oli. 2015. Consequences of individual removal on
persistence of a protected population of long-lived turtles. Animal Conservation
19(4):369-379.
Dolbeer, R. A. 1969. Population density and home-range size of the Eastern Box Turtle
(Terrapene c. carolina) in eastern Tennessee. ASB Bulletin 16:49.
Driscoll, D. A., and M. K. Henderson. 2008. How many common reptile species are fire
specialists? A replicated natural experiment highlights the predictive weakness of a
fire succession model. Biological Conservation 141:460-471.
Eagle, A. C., E. M. Hay-Chmielewski, K. T. Cleveland, A. L. Derosier, M. E. Herbert, and R. A.
Rustem, editors. 2005. Michigan's Wildlife Action Plan. Michigan Department of
Natural Resources. Lansing, Michigan, USA.
Eichenlaub, V. L. 1979. Weather and the climate of the Great Lakes Region. The University of
Notre Dame Press, Notre Dame, USA.
Eichenlaub, V. L, J. R. Harman, F. V. Nurnberger, and H. J. Stolle. 1990. The climatic atlas of
Michigan. The University of Notre Dame Press, Notre Dame, USA.
Ernst, C. H., T. P. Boucher, S. W. Sekscienski, and J. C. Wilgenbusch. 1995. Fire ecology of the
Florida box turtle, Terrapene carolina bauri. Herpetological Review. 26(4):185-187.
Ernst, C. H., and J. E. Lovich. 2009. Turtles of the United States and Canada. Johns Hopkins
University Press, Baltimore, MD, USA.
Erwin, W. J. and R. H. Stasiak. 1979. Vertebrate mortality during the burning of reestablished
prairie in Nebraska. American Midland Naturalist 101:247-249.
Floyd, T. M., K. R. Russell, C. E. Moorman, D. H. Van Lear, D. C. Guynn Jr., and J. D. Lanham.
2002. Effects of prescribed fire on herpetofauna within hardwood forests of the Upper
Piedmont of South Carolina: a preliminary analysis. U.S. Department of Agriculture,
Forest Service, Southern Research Station, Ashville, North Carolina, General
Technical Report SRS–48.
Gibson, J. 2009. Influence of prescribed fire on a Midwestern population of the eastern box
turtle, Terrapene c. carolina. Thesis, Purdue University, Indiana, USA.

89

Greenberg C. H., and T. A. Waldrop. 2008. Short-term response of reptiles and amphibians to
prescribed fire and mechanical fuel reduction in a southern Appalachian upland
hardwood forest. Forest Ecology and Management 255:2883-2893.
Griffiths, A. D., and Christian, K. A. 1996. The effects of fire on the frillneck lizard
(Chlamydosaurus kingii) in northern Australia. Australian Journal of Ecology
21:386–398.
Howe, H. F. 1994. Managing species diversity in tallgrass prairie: Assumptions and implications.
Conservation Biology 8:691-704.
Keyser, P. D., D. J. Sausviller, W. M. Ford, D. J. Schwab, and P. H. Brose. 2004. Prescribed fire
impacts to amphibians and reptiles in shelterwood-harvested oak-dominated forests.
Virginia Journal of Science 55(4):159-168.
Knapp, E. E., B. L. Estes, C. N. Skinner. 2009. Ecological effects of prescribed fire season: a
literature review and synthesis for managers. U.S. Department of Agriculture, Forest
Service, Pacific Southwest Research Station, Albany, California, General Technical
Report PSW-GTR-224.
Kost, M. A. 2004. Natural community abstract for mesic prairie. Michigan Natural Features
Inventory, Lansing, USA.
Legler, J. M. 1960. Natural history of the ornate box turtle, Terrapene ornata ornata. University
of Kansas Publications, Museum of Natural History 11(10):527-669.
Mardia, K. V., J. T. Kent, and J. M. Bibby. 1979. Multivariate Analysis. Academic Press. New
York, USA.
Means, D. B., and H. W. Campbell. 1981. Effects of prescribed burning on amphibians and
reptiles. In: Wood, G.W., editor, Prescribed fire and wildlife in southern forests. The
Belle W. Baruch Forest Science Institute, Clemson University, Georgetown, South
Carolina, USA.
Melvin, T.A. 2017. Prescribed fire effects on eastern box turtles in southwestern Michigan.
Thesis, Michigan State University, East Lansing, USA.
Michigan Natural Features Inventory [MNFI]. 2017. Rare Species Explorer (Web Application).
<http://mnfi.anr.msu.edu/explorer> Accessed Mar 13, 2017.
Moseley, K. R., S. B. Castleberry, and S. H. Schweitzer. 2003. Effects of prescribed fire on
herpetofauna in bottomland hardwood forests. Southeastern Naturalist 2(4): 475-486.
Palmgren, G. 2004. Phase III natural resources management plan: Fort Custer State Recreation
Area (draft review). Parks and Recreation Bureau, Department of Natural Resource,
Lansing, USA.
90

Platt, S. G., H. Liu, and C. K. Borg. 2010. Fire ecology of the Florida box turtle (Terrapene
carolina bauri) in the pine rockland forests of the Lower Florida Keys. Natural Areas
Journal 30:254–260.
Renken, R. 2005. Does fire affect amphibians and reptiles in eastern U.S. oak forests? Pages
158–166 in Proceedings of Fire in Eastern Oak Forests: Delivering Science to Land
Managers. Proceedings of a Conference, 15 November–18 November 2005,
Columbus, Ohio, USA.
Rice, P. M., and J. K. Smith. 2008. Chapter 4: Use of Fire to Manage Populations of Nonnative
Invasive Plants. USDA Forest Service Gen. Tech. Rep. RMRS-GTR-42-vol. 6.
Richburg, J. A., and W. A. Patterson III. 2003. Can Northeastern Woody Invasive Plants Be
Controlled with cutting and Burning Treatments? Pages 1-3 in Proceedings of Using
Fire to Control Invasive Plants: What’s New, What Works in the Northeast. Natural
Resource Network, 24 January 2003, Portsmouth, New Hampshire, USA.
Roloff, G. J., and J. B. Haufler. 1997. Establishing population viability planning objectives based
on habitat potentials. Wildlife Society Bulletin 25:895-904.
Roloff, G. J., and J. B. Haufler. 2002. Modeling habitat-based viability from organism to
population. Pages 673-685 in M. C. Scott, P. Heglund, M. L. Morrison, J. B. Haufler,
M. G. Raphael, W. A. Wall, and F. B. Samson, editors. Predicting species
occurrences: issues of accuracy and scale. Island Press, Washington, D.C., USA.
Russell, K. R., D. H. Van Lear, and D. C. Guynn Jr. 1999. Prescribed fire effects on
herpetofauna: review and management implications. Wildlife Society Bulletin
27:374-384.
Smith, J. K., editor. 2000. Wildland fire in ecosystems: effects of fire on fauna. U.S. Department
of Agriculture, Forest Service, Rocky Mountain Research Station, Ogden, Utah,
General Technical Report RS-GTR-42-vol. 1.
Spencer, R. J., and M. B. Thompson. 2003. The significance of predation in nest site selection of
turtles; an experimental consideration of macro- and microhabitat preferences. Oikos
102:592–600.
Swarth, C., and S. Hagood. 2004. Introduction. P. 32 in Swarth, C. and S. Hagood, editors.
Summary of the Eastern Box Turtle Regional Conservation Workshop. The Humane
Society of the United States.
Weiss, J. A. 2009. Demographics, activity, and habitat selection of the eastern box turtle
(Terrapene c. carolina) in West Virginia. Thesis, Marshall University, West Virginia,
USA.

91

Willey, L. L., and P. R. Sievert. 2012. Notes on the nesting ecology of eastern box turtles near
the northern limit of their range. Northeastern Naturalist 19(3):361-372.
Woodley, C. L. 2013. Predicting spring emergence in a northern population of the Eastern box
turtle (Terrapene c. carolina). Thesis, Purdue University, Indiana, USA.
van Dijk, P.P. 2011. Terrapene carolina. (errata version published in 2016). The IUCN Red List
of Threatened Species 2011. < http://www.iucnredlist.org/details/21641/0>. Accessed
10 February 2017.

92

CONCLUSIONS

My thesis focused on the direct and indirect effects of prescribed fire, nesting ecology and
hatchling movements, and the efficacy of post-burn surveys on eastern box turtles in a 1226 ha
recreation area in Michigan, USA. I summarized the nesting behavior of females, resultant clutch
sizes and successes, and hatchling movements using summary statistics and general linear mixed
models. I tested the efficacy of post fire surveys and I observed the direct behavioral effects of a
prescribed fire on research animals. This study is one of the first to test the efficacy of growing
season prescribed fires on both vegetation and animal behavior and survival. The strengths of
this research are that: 1) a known cohort of research animals were observed over 3 years, 2) the
nesting ecology and hatchling movement research was conducted in relation to current tallgrass
prairie restoration practices, 3) the research burn was conducted in relation to a novel prescribed
fire season, 4) this is the first research that I am aware of to document post fire detection
probability and direct behavioral effects of prescribed fire on eastern box turtles. The limitations
of my study include: 1) small sample size (n = 4) of adult turtles that were directly subjected to
the flame front, 2) inability to directly ascertain whether a female nested or was not detected
nesting, 3) inability to determine the known longer-term fates of 51 of 58 hatchlings, and
maximum straight line distances from the nest cavity and overwintering sites for 49 hatchlings,
and 4) small sample size (n = 7 turtles) to determine detection probability.
In Chapter 1, I used radio telemetry on adult females (n=34), trailing string and
transmitters on hatchlings (n=58), and field surveys of nests (n=25). I modeled the relationships
between: 1) clutch size and physical characteristics of the associated female turtle, 2) number of
eggs that developed into hatchlings as a function of year and habitat that contained the nest, and
3) number of hatchlings that emerged from the nest by year and habitat. I failed to find a
93

significant effect of female physical characteristics on clutch size, and I found no effect of year
or habitat on the number of eggs that developed into hatchlings. I found a significant habitat
effect on the number of hatchlings that emerged from the nest cavity, with a gravel pit producing
significantly more emergent hatchlings than agricultural fields. Although female box turtles used
agricultural fields for nesting, those nests failed to produce any emergent hatchlings suggesting
that these fields serve as population sinks. My results indicated that female physical
characteristics were not reliable predictors of clutch size, and that habitat factors were the
primary determinant of hatchling recruitment to the turtle population, emphasizing the
importance of preserving and restoring vegetation types and sites that are appropriate for
hatchling emergence.
In Chapter 2, I found that average detection probability per survey was low and highly
variable among surveyors. I estimated that 26 hour-long surveys per ha would be required to
reliably (95% confidence) detect turtles that were present in burned areas. I also found that
individual turtles directly exposed to fire remained buried for up to 12 hrs after the fire was
extinguished and then rapidly moved to unburned areas. My results suggested that typical visual
encounter surveys conducted for eastern box turtles after prescribed burning did not accurately
reflect occupancy status or fire caused mortality.
In Chapter 3, I found that the growing season fire reduced leaf litter, and woody
vegetation <1 cm diameter, including invasive plant species. Box turtles exhibited varying
behaviors in response to the fire including burying and actively evading and negotiating of the
flame front. One turtle died from injuries sustained in the fire and females within the
management unit where prescribed fire was applied were not observed nesting that year. My

94

findings emphasized the importance of considering the seasonality, fire frequency, and type of
fire and land management techniques in general for the conservation of this species.
I recommend that future research focuses on potential stressors to recruitment of eastern
box turtles including predation, habitats that appear to act as population sinks, and hatchling
vulnerabilities to prescribed fire. Understanding the nesting ecology and subsequent hatchling
box turtle survival, movements and behavior in altered or fragmented landscapes is a necessary
precursor to developing management strategies for their protection at the northern boundary of
their range. I suggest that future research on recruitment and nest success focus on the role of
agricultural fields of any cover crop, better understanding micro-scale differences in soil
temperature and moisture in restored tallgrass prairies, and potential manipulation of site fidelity
by females with man-made nesting areas. The efficacy of post-fire surveys for eastern box turtles
likely varies by fire intensity and the amount and spatial distribution of vegetation and debris, so
I encourage additional detection studies in other fire and vegetation conditions. The direct effects
of growing season fire on eastern box turtles and vegetation also likely varies by fire intensity
and the amount and spatial distribution of vegetation and debris. I encourage additional longterm studies on known cohorts of box turtles, in other fire and vegetation conditions, studies
which include effects on body condition, and studies that ascertain the triggers box turtles might
have in order to detect oncoming flame fronts, whether this is heat detection, smelling smoke,
seeing the flame front, or hearing the flame front. In summary, applied research that recognizes
the needs of land managers using prescribed fire to maintain many early successional vegetation
types, and recognizes that these management activities may irreparably harm box turtles if not
done with care and thoughtfulness, is paramount.

95