Factors influencing snowshoe hare (Lepus americanus) in Michigan
By
David Michael Burt

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Fisheries and Wildlife – Master of Science
2014

ABSTRACT

Factors influencing snowshoe hare (Lepus americanus) in Michigan
By

David Michael Burt
The goal of my thesis was to identify the climatic and vegetation factors that influence
snowshoe hare occupancy in Michigan. My objectives were to: 1) quantify snowshoe hare
occupancy as related to climate change and 2) quantify the relationships between patch-level
snowshoe hare occupancy and land cover change over time, current land cover, habitat
structure, and mesocarnivore presence. In Chapter 1, I determined the most efficient transect
configurations for winter track surveys and found that transects 150m in length with 100 or
75m spacing, or 125m in length with 75m spacing provided reliable estimates of hare
occupancy. In Chapter 2, I researched climatic variables that are potentially linked to snowshoe
hare population performance and assessed whether those factors affected the localized
extinction of snowshoe hares. I found that localized extinction was influenced by maximum
temperature from May 15 – January 19; as temperature increased the likelihood of localized
extinction increased. I also found that the total number of days with measurable snow on the
ground affected localized extinction; as number of days with snow on the ground decreased the
likelihood of localized extinction increased. In Chapter 3, I evaluated whether land cover change
over time, current land cover, and habitat factors affected snowshoe hare occupancy and
habitat use. I found that land cover change over time did not affect hare occupancy. Rather, a
current land cover covariate (the ratio between forest and open edge) and the habitat
covariates of visual obstruction at 1.0-1.5m above snow level and stem density were important;
all 3 parameters were positive but only the transect-level covariates were significant.

ACKNOWLEDGEMENTS

Funding for this project was provided by the Michigan Department of Natural Resources
– Wildlife Division through the Michigan Federal Aid in Wildlife Restoration program grant
F13AF01268 in cooperation with the U.S. Fish and Wildlife Service, Wildlife and Sport Fish
Restoration Program, and Safari Club International Michigan Involvement Committee. I greatly
appreciate the mentoring provided by my advisor, Gary Roloff, who always made time to
provide guidance, helped me advance in my career path, and was always supportive of what I
was trying to accomplish. Thank you to my committee members Rique Campa, III, Barbara
Lundrigan, and Dwayne Etter for their support and involvement with this project.
I am grateful for the help, effort, advice and encouragement of all my previous and
current lab mates, lab employees, and field crews. Without their generous support, I would
have had trouble accomplishing the goals I set out to achieve. I would like to especially thank
Steven Gray, Alex Killion, Daniel Linden, Robert Montgomery, Paul Nelson, Clint Otto, andTracy
Swem. I appreciate the numerous interviewees who graciously made themselves and their
knowledge available for identifying historic hare sites.
I would like to thank Michigan Department of Natural Resources - Wildlife Division
employees J. Belman, D. Beyer, D. Brown, J. Lukowski, K. Sitar, K. Swanson, T. Swearingen, and
C. VanHorn for their assistance with fence construction and/or data collection at the Cusino
Wildlife Research Area. In addition, thanks to R. Aikens and B. Sillet of the Sault Ste. Marie
Tribe of Chippewa for their help in fence construction at Cusino, the replication of the research
at Kinross, and their commitment as a partner to this research project. I am also thankful for

iii

the assistance of MDNR - Wildlife Division employees D. Brown, T. Lyon, T. Maples, V.
Richardson, B. Rollo, J. Valentine, and M. Wegan for helping with field data collection. Also,
thanks to S. Beyer from MDNR for helping manage the social and political aspects of the
project. Thanks to T. Minzey for coordinating my attendance at sportsperson coalition
meetings and M. Donovan for help brainstorming analysis methodologies and ways to map land
cover.
Last, but not least, thank you to my family and friends who supported and encouraged
me throughout this state of my life. In particular Ashley Burt, Kristen Burt, Michael Burt,
Kathleen Morris, Daniel O’Connor, Mitzi Palms-Burt, Lisa Rumsey, and Karalyn Yermak.

iv

TABLE OF CONTENTS

LIST OF TABLES

vii

LIST OF FIGURES

ix

INTRODUCTION
LITERATURE CITED

1
6

CHAPTER 1
RELIABILITY OF WINTER TRACK COUNTS FOR QUANTIFYING SNOWSHOE HARE OCCUPANCY
ABSTRACT
1.1. INTRODUCTION
1.2. STUDY AREA
1.3. METHODS
1.3.1. Data Analysis
1.4. RESULTS
1.5. DISCUSSION
1.6. ACKNOWLEDGEMENTS
APPENDIX
LITERATURE CITED

9
9
10
12
13
14
16
17
20
21
28

CHAPTER 2
CLIMATIC FACTORS INFLUENCING THE DISTRIBUTION OF SNOWSHOE HARES
ABSTRACT
2.1. INTRODUCTION
2.2. STUDY AREA
2.3. METHODS
2.3.1. Site Selection
2.3.2. Field Sampling
2.3.3. Climate Covariates
2.3.4. Data Analysis
2.4. RESULTS
2.5. DISCUSSION
2.6. ACKNOWLEDGEMENTS
APPENDIX
LITERATURE CITED

32
32
33
36
37
37
38
39
40
40
42
45
47
54

v

CHAPTER 3
LAND COVER AND VEGETATION FACTORS INFLUENCING SNOWSHOE HARE OCCUPANCY IN
MICHIGAN
ABSTRACT
3.1. INTRODUCTION
3.2. STUDY AREA
3.3. METHODS
3.3.1 Site Selection
3.3.2 Field Sampling
3.3.3 Data Analysis
3.4. RESULTS
3.4.1 Land Cover Change Over Time
3.4.2 Current Land Cover and Habitat Structure
3.5. DISCUSSION
3.6. MANAGEMENT IMPLICATIONS
3.7. ACKNOWLEDGEMENTS
APPENDIX
LITERATURE CITED

60
60
61
64
65
65
66
69
70
71
72
73
76
78
79
91

CONCLUSIONS
LITERATURE CITED

96
99

vi

LIST OF TABLES

Table 1.1.

Snowshoe hare densities, track survey efforts, location of surveys, and year(s)
that surveys were completed in ~6.1ha enclosures, Upper Peninsula of Michigan,
USA.
22

Table 1.2.

Transect dimensions, the number of transects in a ~6.1ha study site, and
distance traversed during snow track surveys for snowshoe hares. Table includes
the 10 combinations of transect length and spacing that produced ≥90%
accuracy for estimating site occupancy. The efficiency is determined by the total
distance walked for a complete survey.
23

Table 2.1.

Climate covariates identified from the literature review, associated citation, and
observed relationship to snowshoe hare population demographics.
48

Table 2.2.

Candidate model set for estimating the likelihood of a localized snowshoe hare
population going extinct in Michigan. LL = log-likelihood, ∆AIC = difference in AIC
value from top-ranking model, k = model parameters, ωi – Akaike weight of
evidence.
49

Table 2.3.

Means (SE) and ranges of climate covariates used for estimating the likelihood of
a localized snowshoe hare population going extinct in Michigan.
50

Table 3.1.

Sample level, means (SE), and ranges of land cover and habitat covariates used
for estimating localized snowshoe hare occupancy (site level only) or habitat use
(transect level) in Michigan.
80

Table 3.2.

Candidate model set for estimating the likelihood of a localized site being
occupied by snowshoe hare in Michigan. ∆AIC = difference in AIC value from
top-ranking model, k = model parameters, ωi – Akaike weight of evidence.
81

Table 3.3.

Top 6 ranking models with variables from site and transect levels from the 41
candidate models (Table 3.5) for estimating the likelihood of a site being
currently occupied by snowshoe hare in Michigan. ∆AIC = difference in AIC value
from top-ranking model, k = model parameters, ωi – Akaike weight of evidence.
82

Table 3.4.

The relationship between visual obstruction and stem densities in 3 forest types.
Compiled from stem densities and visual obstruction measured at 117 sites in
the northern Lower and Upper Peninsulas of Michigan, winter of 2013.
83

vii

Table 3.5.

List of all 41 candidate models used in current occupancy modeling for
estimating the likelihood of a site being currently occupied by snowshoe hare in
Michigan.
84

viii

LIST OF FIGURES

Figure 1.1.

The geographic range of snowshoe hares (diagonal striping), with Michigan
outlined in bold black (map inset), and the locations of Cusino and Kinross hare
enclosures, Upper Peninsula of Michigan, winters of 2012-2014. The geographic
range of hares was adapted from the International Union for Conservation of
Nature (www.iucnredlist.org/technical-documents/spatial-data#mammals). 24

Figure 1.2.

Schematic of how snowshoe hare tracks in enclosures were subsampled
including (A) the outline of the enclosures (black) with 9 transects spaced 25m
apart (gray), (B) snowshoe hare tracks during 1 survey as black circles, (C)
random locations generated on each transect to denote the start of transect
segments for 1 iteration as black 4-pointed stars, (D) transect segments of 25m
on each transect as black lines originating from the black 4-pointed stars, (E)
snowshoe hare tracks that intercepted transect segments, and (F) example of
the results of 1 iteration with 100m spacing.
25

Figure 1.3.

Proportion of 10,000 iterations that a site was correctly deemed as occupied for
all transect configurations at hare densities of (A) 1.5, (B) 1.2, (C) 0.8, (D) 0.5, and
(E) 0.2 hares/ha. The 90% accuracy threshold is designated by the black
horizontal line.
26

Figure 1.4.

The average number of transects occupied at each of the 5 hare densities for the
3 most efficient transect configurations: 125 m length x 75 m spacing, 150 m
length x 75 m spacing, and 150 m length x 100 m spacing.
27

Figure 2.1.

Location and occupancy status of snowshoe hare survey sites throughout the
northern Lower and Upper Peninsulas of Michigan, USA, winter 2013. The
geographic range of snowshoe hares and Michigan are portrayed in the map
inset. The geographic range was derived from the International Union for
Conservation of Nature (www.iucnredlist.org/technical-documents/spatialdata#mammals).
51

Figure 2.2.

Number of study sites and year (5-year groupings through 2010) of last
confirmed snowshoe hare occupancy in the northern Lower and Upper
Peninsulas of Michigan, USA.

Figure 2.3.

52

Localized snowshoe hare extinction probability and (a) mean maximum
temperature during May 15 – January 19, (b) number of days with snow on the
ground, and (c) mean snow depth in Michigan, USA and the 95% confidence
intervals. The points along the x-axes represent the values of each study site. 53

ix

Figure 3.1.

Location and occupancy status of snowshoe hare survey sites throughout the
northern Lower and Upper Peninsulas of Michigan, USA, winter 2013. The
geographic range of snowshoe hares and Michigan are portrayed in the map
inset as adapted from the International Union for Conservation of Nature
(www.iucnredlist.org/technical-documents/spatial-data#mammals).
88

Figure 3.2.

Number of study sites by 5-year groupings (through 2010) of last confirmed
snowshoe hare occupancy in the northern Lower and Upper Peninsulas of
Michigan, USA.

Figure 3.3.

89

Snowshoe hare site occupancy probability and (A) forest to open edge ratio, and
transect level probability of use by (B) visual obstruction 1.0-1.5m above snow
level, and (c) equivalent stem density in Michigan, USA and the 95% confidence
intervals. The points along the x-axes represent the values of each study site. 90

x

INTRODUCTION

Culturally, snowshoe hares (Lepus americanus) are part of the hunting heritage in
Michigan. In the 1970s over 100,000 individuals hunted hares each year, but this number
declined to just under 20,000 in 2007 (Frawley 2008). Small‐game hunter surveys indicate that
snowshoe hare harvest and presumably abundance have consistently declined statewide over
the past few decades (Frawley 2008). In addition to being culturally important in Michigan,
snowshoe hares are also an important component of predator-prey communities, particularly
at northern latitudes. Hares are known prey for numerous mesocarnivores and raptors,
including Canada lynx (Lynx canadensis), red fox (Vulpes vulpes), American marten (Martes
americana), fisher (Martes pennanti), coyote (Canis latrans), bobcat (Lynx rufus), red-tailed
hawk (Buteo jamaicensis), and great horned owl (Bubo virginianus) (Carreker 1985). In some
ecosystems (e.g. boreal forests of Canada and Alaska) predator population dynamics closely
correlate with hare populations, underscoring the importance of hares to ecosystem function
(Brand et al. 1976). The importance of lagomorphs to community function is not limited to
hare-lynx systems. For example, generalist predators (e.g. coyotes) have lower predation rates
on white-tailed deer when snowshoe hares exist in high abundance (Patterson and Messier
2000, Hurley and Garton 2011).
Snowshoe hares can be used as an indicator of climate change. The hare population in
Michigan is at the southernmost boundary of the species range in the Lake States region (Figure
1.1). With a warming and drying climate, the range will likely shift northward because of
reduced duration of snow cover and depth (Buehler and Keith 1982). Mechanisms causing this
1

range shift are not understood. Plausible hypotheses relate to direct effects on population vital
rates and indirect effects on individual life history strategies (Bardsen et al. 2011). For example,
molting at the proper time is essential for hare survival. This topic has drawn an increase of
attention lately with researchers studying molting times and causes of snowshoe hares (Mills et
al. 2013, Zimova et al. 2014).
A coarse, broad‐scale habitat assessment corresponding to the late 1990s and early
2000s indicated that hare habitat quality in the Upper Peninsula of Michigan ranged from poor
to marginal with only small, isolated areas of higher quality habitat (Linden et al. 2011).
Snowshoe hares occupy continuous forested areas with a dense understory (>60% visual
obstruction) that provides food, safety from predators and thermal cover. Hare survival rate is
lower in areas with sparse horizontal cover (Litvaitis et al. 1985, Hodges 2000, Berg et al. 2012).
For my study on Michigan snowshoe hares, I used stakeholders to identify potential
study sites. The use of individuals familiar with local history and ecology to supplement
research and management is common and often referred to as local ecological knowledge (or
traditional ecological knowledge if the stakeholders are indigenous to the area). Use of local
ecological knowledge has increased since 1980, with 421 papers being published between 1980
and 2004. Of these papers, over half used interviews as the process for generating local
ecological knowledge (Brook and McLachlan 2008). Gilchrist et al. (2005) found that local
ecological knowledge was valuable to complement empirical data. Traditional ecological
knowledge has proven reliable compared to more western science based approaches for
generating information and can be used in conjunction with biological assessments (e.g.
resource selection functions; Jacqmain et al. 2007, Polfus et al. 2014). Using stakeholders to
2

supplement traditional scientific approaches is also the basis for citizen science research and
projects (Dickinson et al. 2010). The data stakeholders generate in these studies has proven
reliable (Jordan et al. 2012). From each stakeholder I requested a relatively low level of local
knowledge; stakeholders were asked for when and where they saw snowshoe hares, not
specifics such as the exact day, sex, or abundance. As a testament to the repeatability of data
generated from stakeholder interviews, independently interviewed stakeholders in my study
provided the same exact locations.
Given that: 1) snowshoe hare harvest and presumably abundance have declined
throughout Michigan, 2) hares are important in many predator-prey communities, 3) hares are
a potential indicator of climate change, and 4) Michigan contains low quality and fragmented
habitats, this research was designed to study factors influencing snowshoe hare occupancy in
Michigan. My thesis consists of 2 distinct questions: 1) what is the optimal snow-tracking
configuration for effectively and efficiently sampling a snowshoe hare population and 2) what
factors are driving snowshoe hare localized extinctions, current occupancy, and habitat use in
Michigan?
In Chapter 1 I describe the experimental design I implemented, with assistance from the
Michigan Department of Natural Resources – Wildlife Division, and results of a captive hare
study that I used to determine the optimal transect configuration for snow-tracking surveys. I
implemented the captive hare study at 2 locations in the Upper Peninsula of Michigan by
building ~6.1ha enclosures to control hare densities, trapping and radio-collaring hares near the
enclosure and releasing them inside, and implemented snow track surveys as I manipulated
hare densities. I simulated different transect configurations by varying transect length and
3

spacing. The optimal transect configuration was based on accuracy (needed to correctly
designate the site as occupied >90% of the time) and efficiency (minimize the distance
traversed by a surveyor).
In chapter 2, I assessed the effects of climate change on the extinction probability (over
time ranging from 1955 to 2012) of snowshoe hares across Michigan. I compiled historically
occupied sites using stakeholder interviews and surveyed a subset of those sites (using the
optimal transect configuration from Chapter 1) to determine current occupancy status. I
reviewed the published literature on snowshoe hares and identified 7 climate variables that
other researchers denoted as important to snowshoe hare demographics. I used a logistic
regression to determine which of the 7 climate variables were affecting localized snowshoe
hare extinction in Michigan.
In Chapter 3, I evaluated how land cover and habitat factors influenced snowshoe hare
occupancy and small-scale habitat use. I conducted 2 separate analyses: 1) the influence of
historical land cover change on site level occupancy and 2) the effects of current land cover and
habitat structure on site level occupancy and transect level use. Land cover covariates were
mapped from aerial photography ranging from 1977 to 2012. I calculated the net change of
land cover variables and used a logistic regression to determine which variables were
significantly affecting snowshoe hare localized occupancy over time. Current land cover
(mapped photography) and habitat variables (collected on transects) were modeled using the
2-level occupancy model in the package “Unmarked” in R to determine which variables were
significantly influencing snowshoe hare occupancy and habitat use in Michigan. Additionally, I
provide management recommendations based on the relationships between stem densities
4

and visual obstruction for 3 different forest types (conifer dominated, deciduous dominated,
and mixed).

5

LITERATURE CITED

6

LITERATURE CITED

Bardsen, B., J. Henden, P. Fauchald, T. Tveraa, and A. Stien. 2011. Plastic reproductive allocation
as a buffer against environmental stochasticity – linking life history and population
dynamics to climate. Oikos 120:245-257.
Berg, N. D., E. M. Gese, J. R. Squires, and L. M. Aubry. 2012. Influence of forest structure on the
abundance of snowshoe hares in western Wyoming. Journal of Wildlife Management
76:1480-1488.
Brand, C. J., L. B. Keith, and C. A. Fischer. 1976. Lynx responses to changing snowshoe hare
densities in central Alberta. Journal of Wildlife Management 40:416-428.
Brook, R. K., and S. M. McLachlan. 2008. Biodiversity and Conservation 17:3501-3512.
Buehler, D. A. and L. B. Keith. 1982. Snowshoe hare distribution and habitat use in Wisconsin.
Canadian Field-Naturalist 96:19-29.
Carreker, R. G. 1985. Habitat suitability index models: snowshoe hare. Biological Report
82:10.01.
Dickinson, J. L., B. Zuckerberg, and D. N. Bonter. 2010. Citizen science as an ecological research
tool: challenges and benefits. Annual Review of Ecology, Evolution, and Systematics
41:149-172.
Frawley, B. J. 2008. 2007 small game harvest survey. Michigan Department of Natural
Resources Wildlife Division Report Number 3493, Lansing, MI, USA.
Gilchrist, G., M. Mallory, and F. Merkel. 2005. Can local ecological knowledge contribute to
wildlife management? Case studies of migratory birds. Ecology and Society 10:20.
Hodges, K. E. 2000. Ecology of snowshoe hares in southern boreal and montane forests. Pages
163-207 in L.F. Ruggiero, K.B. Aubry, S.W. Buskirk, G.M. Koehler, C.J. Krebs, K.S.
McKelvey and J.R. Squires editors. Ecology and Conservation of Lynx in the United
States. University Press of Colorado, Boulder, CO, USA.
Hurley, M. A., and E. O. Garton. 2011. Demographic response of mule deer to experimental
reduction of coyotes and mountain lions in southeastern Idaho. Wildlife Monographs
178:1-33.

7

Jacqmain, H., L. Bélanger, S. Hilton, and L. Bouthillier. 2007. Bridging native and scientific
observations of snowshoe hare habitat restoration after clearcutting to set wildlife
habitat management guidelines on Waswanipi Cree land. Canadian Journal of Forest
Research 37:530-539.
Jordan, R. C., W. R. Brooks, D. V. Howe, and J. G. Ehrenfeld. 2012. Evaluating the performance
of volunteers in mapping invasive plants in public conservation lands. Environmental
Management 49:425-434.
Linden, D. W., H. Campa, III, G. J. Roloff, D. E. Beyer, Jr., and K. F. Millenbah. 2011. Modeling
habitat potential for Canada lynx in Michigan. Wildlife Society Bulletin 35:20-26.
Mills, L. S., M. Zimova, J. Oyler, S. Running, J. T. Abatzoglou, P. M. Lukacs. 2013. Camouflage
mismatch in seasonal coat color due to decreased snow duration. Proceedings of the
National Academy of Science 110:7360-7365.
Litvaitis, J. A., J. A. Sherburne, and J. A. Bissonette. 1985. Influence of understory characteristics
on snowshoe hare habitat use and density. Journal of Wildlife Management 49:866-873.
Patterson, B. R., and F. Messier. 2000. Factors influencing killing rates of white-tailed deer by
coyotes in eastern Canada. Journal of Wildlife Management 64:721-732.
Polfus, J. L., K. Heinemeyer, M. Hebblewhite, and Taku River Tlingit First Nation. 2014.
Comparing traditional ecological knowledge and western science woodland caribou
habitat models. Journal of Wildlife Management 78:112-121.
Zimova, M., L. S. Mills, P. M. Lukacs, and M. S. Mitchell. 2014. Snowshoe hares display limited
phenotypic plasticity to mismatch in seasonal camouflage. Proceedings of the Royal
Society B 281:20140029; doi:10.1098/rspb.2014.0029.

8

CHAPTER 1

RELIABILITY OF WINTER TRACK COUNTS FOR QUANTIFYING SNOWSHOE HARE OCCUPANCY

ABSTRACT
I determined the optimum transect length and spacing for quantifying snowshoe hare (Lepus
americanus) occupancy in a fixed area from snow track surveys. I also evaluated the utility of
the most reliable and efficient designs for indexing hare density. I constructed enclosures
(~6.1ha) at 2 locations in the Upper Peninsula of Michigan, USA, and populated the enclosures
with radio-collared hares. Hare densities ranged from 0.2 – 1.5 hares/ha in the enclosures,
comparable to low and high densities recorded at the southern extent of snowshoe hare range.
I conducted snow track surveys along 9 transects spaced 25m apart in the enclosures 12-65
hours after a snowfall and mapped the location of every track that intersected a transect. After
standardizing the track maps by time since last snowfall, I simulated different transect lengths
and spacings and evaluated if hares were documented on the resultant transect segments. I
deemed transect configurations reliable if >90% of the 10,000 simulations correctly denoted
the site as occupied. Of the 28 possible transect configurations, only 10 combinations provided
reliable estimates of hare occupancy. I refined the 10 reliable configurations based on
efficiency, where efficiency was based on the distance traversed by a surveyor. I recommend
using transects that are 150m in length with 100 or 75m spacing, or 125m in length with 75m
spacing to reliably and efficiently survey a fixed area for snowshoe hares. Each of these 3
configurations was non-linearly related to the range of snowshoe hare densities evaluated in
the study.
9

Key Words
Lepus americanus, Michigan, occupancy, snowshoe hare, snow track surveys, survey
methodology, transect configuration
1.1. INTRODUCTION
Snowshoe hares (Lepus americanus) are vulnerable to climate change (Buehler and
Keith 1982, Hoving et al. 2013, Mills et al. 2013), thus interest in documenting occupancy and
abundance across large areas has increased. With models projecting warmer and drier climates
during the 21st century for many parts of the world (IPCC 2013), knowing how to efficiently
sample populations of vulnerable species is critical to documenting range shifts and
implementing conservation activities. Efficient sampling of wildlife populations across large
spatial extents relies on methods that can be quickly implemented during a single site visit and
produce reliable results.
Selection of appropriate survey techniques in any wildlife study depends on the spatial
and temporal extents of the research question(s), logistics of data collection, and the study
system (organism and environment). For members of the genus Lepus, common techniques for
documenting occupancy, abundance, and habitat use include live trapping (often used with
mark-recapture), visual detection surveys along transects, fecal pellet counts, and snow track
surveys (Marcström et al. 1989, Koehler 1990, Litvaitis et al. 1985, Shimizu and Shimano 2010,
Lu 2011); combining these techniques may be appropriate depending on survey objectives (e.g.,
Roy et al. 2010). Some techniques are proven reliable, like using live trapping and fecal pellet
counts to index hare abundance (e.g., Litvaitis et al. 1985, Hartman 2009), but these techniques

10

generally require multiple visits to the same location and are labor intensive. Fecal pellet counts
can also be difficult to interpret in areas where multiple members of the Leporidae co-exist.
Snow track surveys are another proven technique for estimating hare habitat use and
abundance (e.g., Marcström et al. 1989, Shimizu and Shimano 2010). Arguably, snow track
surveys are better suited for larger areas than live trapping or fecal pellet counts, particularly in
remote and difficult to traverse (like swamps) areas. The primary benefit of snow track surveys
is that they can produce reliable results with a single site visit (Brocke 1975, Conroy et al. 1979,
Thompson et al. 1989), however transect dimensions and configurations are inconsistent
among studies so information on the most efficient design is lacking. To date, studies on
snowshoe hare have used varying transect dimensions including a single 70m transect per study
site (Roy et al. 2010), 2 – 1km transects spaced >1km apart (Thompson et al. 1989) and multiple
transects spaced 50m apart (Conroy et al. 1979), with little guidance on efficacy of the various
designs.
I sought a reliable and efficient survey methodology to estimate snowshoe hare
occupancy. I sought a technique that could be: 1) used to reliably assess localized extinction, 2)
implemented across a large spatial extent, 3) conducted with a single site visit, and 4)
implemented with minimal equipment and survey time. Given the available techniques for
estimating snowshoe hare occupancy and abundance, snow track surveys satisfied these
criteria. My objective was to determine the optimum (i.e., most reliable and efficient) transect
length and spacing for quantifying snowshoe hare occupancy from snow track surveys.
Secondarily, I also evaluated the utility of the most reliable and efficient designs for indexing

11

hare abundance. I used a known number of radio-collared hares in ~6.1ha enclosures to
evaluate a variety of transect lengths and spacings.
1.2. STUDY AREA
I conducted this study at 2 locations in the Upper Peninsula of Michigan: the Cusino
State Wildlife Research Area (hereafter referred to as Cusino; 46°20′54″N 86°28′13″W) in Alger
County, and on tribal land in Chippewa County near the town of Kinross (hereafter referred to
as Kinross; 46°16’52”N 84°34’36”W). Cusino and Kinross are located >300 km north of the
southern edge of hare range in Michigan (Fig. 1.1), and both locations were known to
historically support snowshoe hares. Both locations have low topographic relief yet high
landform complexity because of glacial activity (Bailey 1995). On average, Cusino received
7.0cm of precipitation monthly from December to March, resulting in snow depths that ranged
between 60 and 120cm yearly (NOAA Weather Station Id MI205690; data from 2000-2009).
Average minimum temperatures during winter at Cusino averaged -10°C (NOAA Weather
Station Id MI205690; 2000-2009). Average monthly precipitation in Kinross from December
through March was 5.7cm, resulting in 40 to 85 cm of snow yearly (NOAA Weather Station Id
MI207190; data from 2000-2009). Minimum average temperature during December through
March averaged -11°C at Kinross. Vegetation at the study sites included overstory alder (Alnus
spp.), black spruce (Picea glauca), white cedar (Thuja occidentalis), and hemlock (Tsuga
canadensis) in low-lying areas and balsam fir (Abies balsamifera.), pine (Pinus spp.), maple (Acer
spp.), birch (Betula spp.), and aspen (Populus spp.) on drier areas.

12

1.3. METHODS
I constructed ~6.1ha enclosures at Cusino and Kinross using 2.4m high chicken wire
supported by posts at approximately 2.4m spacing. I selected ~6.1ha for the enclosure based
on the amount of suitable habitat available at Cusino, the logistics of fence building, and to
encompass the average home range of a snowshoe hare (Keith 1990). Depressions along the
bottom of the fence were filled with woody debris. After >30cm of snow accumulation, the
bottom of the fence was impervious to hare movements. My goal was to control snowshoe
hare numbers (i.e., prevent escape or immigration) during the winter sample period. I ensured
that the pen was unoccupied prior to the start of the study by repeatedly walking the enclosure
and looking for tracks.
I used wooden box traps to live-trap snowshoe hares in habitats <32 and <5km from the
enclosures at Cusino and Kinross, respectively. Traps were baited with fresh alfalfa and apples,
set in and under suitable cover, and checked daily. Each captured hare was sexed, weighed, and
for those >900g fitted with a 20g radio collar (with mortality sensor; Advanced Telemetry
Systems, Isanti, MN). Radio-collared hares were subsequently released at a random location in
the enclosure. The maximum hare density within an enclosure (1.5 hares/ha) corresponded to
the maximum density documented in the southern portion of snowshoe hare range (Hodges
2000). After hares were released into the enclosure, I allowed them to acclimate for a minimum
of 7 days before track sampling occurred.
At the beginning of each track sampling event, the observer confirmed that the released
hares were in the enclosure (via the radio signals) and alive (via the mortality sensors). I
conducted the track surveys along 9 north‐south transects, spaced 25m apart and from the
13

fence edge, that ranged in length from 212 to 256m (Fig. 1.2A). I established transects by using
a meter tape and hanging a labeled flag every 10m to facilitate mapping track locations. Track
surveys were initiated >12hrs after a snowfall event that was substantive enough to cover old
tracks (Brocke 1975). Observers (1 or 2) would walk transects 12-65 hours after a snowfall
event and record the location where every hare track crossed a transect (Fig. 1.2B). All 9
transects were surveyed in a single day (usually within 4 hours).
After ≥1 track survey at a specific hare density, I removed hare(s) from the enclosure
and the track survey(s) was repeated. For hares that were re-captured inside the pen, I
removed radio collars and returned them to their capture origin. Sampling and hare removals
continued when snow conditions permitted and until I could no longer ensure population
closure (based on radio-collar and track evidence along the perimeter of the pen). I replicated
this study in 2012 and 2013 at Cusino and in 2014 at Kinross, but starting hare densities varied
at each location. All trapping and animal handling procedures were conducted by the Michigan
Department of Natural Resources – Wildlife Division or Sault Ste. Marie Tribe of Chippewa
Indians following internal animal use and care guidelines that were consistent with the
American Society of Mammalogists Animal Care and Use Guidelines (Sikes et al. 2011).
1.3.1. Data Analysis
To help control for track accumulation based solely on time since the last snowfall, I
standardized the number of tracks from each survey to a 12-hour period. For example, if a
survey occurred 16 hours after snowfall and there were 40 tracks, the total number of tracks
was randomly reduced by 25% to 30 tracks (i.e. 40 * (12 hours/16 hours) = 30 tracks). Once the
track maps from all surveys were standardized to 12 hours (Fig. 1.2B), I subsampled the 21214

256m full transects into segments that were 10, 25, 50, 75, 100, 125, and 150m long. To
subsample the data, I generated random locations along each transect and then created the
appropriate transect segment from that random location (Figs. 1.2C,D). I then determined if any
snowshoe hare tracks occurred on that resulting segment (Fig. 1.2E). A track on any segment
indicated that the site was occupied (i.e., at least 1 track was documented on at least 1
transect).
I also simulated different transect spacing of 25, 50, 75, and 100m by removing
transects from the analysis, creating 28 total transect configurations. For example, at a 25m
spacing, all 9 transects were used (Fig. 1.2E), but at 100m spacing, only the 1st, 5th, and 9th
transects were used (Fig. 1.2F). I simulated changes in spacing because I was interested in the
number of transects that were needed to accurately sample a fixed area. The combination of
transect length and distance traversed between transects ultimately defined sampling
efficiency. I iterated each combination of transect length and spacing 10,000 times. I also tallied
the number of occupied transects as a potential index to density. Analyses were performed in
the statistical software R v. 3.1.0 (R Core Team 2014).
In the study, optimal transect configuration had 2 components: efficiency and reliability.
I defined a configuration as reliable if occupancy status was correctly designated for >90% of
the 10,000 iterations. I quantified efficiency based on the distance a surveyor would traverse as
defined by transect length and spacing. Longer transects are less efficient to survey than
shorter transects, and tightly spaced transects less efficient than further spaced transects
(because more transects are needed to survey a fixed area). Given the parameters of the study,

15

10m transects spaced 100m apart would theoretically be the most efficient study design if
accuracy was >90%.
1.4. RESULTS
I conducted track surveys at hare densities ranging from 1.5 hares/ha (9 in the
enclosure) to 0.2 hares/ha (1 in the enclosure; Table 1.1). Starting hare densities varied at each
enclosure because of annual variations in snow conditions, timing of the enclosure being
completed, and availability of field technicians to run the study. I generally completed ≥2
surveys at each hare density, except for 0.5 hares/ha (Table 1.1). All surveys, except for the 5
surveys at 0.2 hares/ha (a single male), were conducted with hares from both sexes in the pen.
The shortest transect (10m) was generally inaccurate (i.e., ≤90% correct designation of
occupancy) regardless of hare density, except for 25m spacing at 1.5 hares/ha (Fig. 1.3A).
Generally, as spacing increased the shorter transects became more inaccurate, particularly at
low hare densities (Figs. 1.3D,E). At high hare densities (1.5 to 1.2 hares/ha) several transect
configurations satisfied the accuracy criterion (Figs. 1.3A,B), hence the decision on which
configuration to use depended on length and spacing. At higher hare densities 50m transects
spaced 100m apart was the most efficient configuration for estimating site occupancy (Figs.
1.3A,B). The more relevant question, particularly for surveys at the southern edge of hare
distribution, is how to optimally sample at low hare densities. At 0.2 hares/ha, transects ≥75m
long were required to satisfy the accuracy criterion (Fig. 1.3E). The most efficient design was a
150m long transect at 100m spacing (Fig. 1.3E). This configuration requires surveyors to
traverse 1,150m to accurately survey ~6.1ha (Table 1.2).

16

Although my ability to rigorously evaluate the relationship between hare density and
the number of occupied transects was limited to a low number of transects at wide spacing
(n=3), my data suggest that counts of occupied transects correlate with hare density when a
broad range of hare densities is evaluated (Fig. 1.4). This pattern was consistent for the 3 most
efficient transect configurations and warrants further evaluation (Fig. 1.4). At low hare densities
typical of the southern range (i.e., 0.2 to 0.5 hares/ha), counts of occupied transects were not
consistently correlated with density (Fig. 1.4).
1.5. DISCUSSION
Given recent attention to snowshoe hares as a species vulnerable to climate change (Hoving et
al. 2013, Mills et al. 2013), interest in accurate and efficient survey techniques that can be
implemented across large areas has increased (Fortin et al. 2005). Results from the snow track
surveys and associated data simulations offer guidance to researchers on how to optimize
survey methodologies for quantifying snowshoe hare site occupancy. Given the ~6.1ha survey
area and a 12-65-hour survey window following a snow event, I recommend 3 transect designs
as being most efficient for sampling hares at low densities that include 150m length with 100 or
75m spacing, and 125m length with 75m spacing (Table 1.2). I based my estimate of efficiency
on the total distance traveled by a surveyor (Table 1.2), but caution that selection of a specific
configuration should also be based on the spatial pattern of potential habitats. If potential
habitats are large and blocky (e.g., expansive coniferous swamps), the 150m long with 100m
spacing is the most efficient design. However, if habitats are patchy or narrow (e.g., riparian
zones) a configuration based on shorter transects may be more desirable to help ensure that
survey effort is concentrated in areas that likely support hares. My recommendation that
17

multiple transect configurations are suitable for quantifying occupancy is consistent with others
that evaluated hare survey techniques. For example, Hodges and Mills (2008) found that
multiple designs were effective for fecal pellet surveys.
MacKenzie et al. (2006:7) noted that proper inference from sampling animal
populations depends on 2 critical components: the ability to capture spatial variation and
detectability. Spatial variation is important because in large scale monitoring programs, like
those needed to identify range shifts in widely distributed species, investigators cannot survey
entire areas (MacKenzie et al. 2006). Instead, rigorous experimental designs based on accurate
and efficient sampling are required (Fortin et al. 2005). For snowshoe hares, I found that snow
tracking could produce reliable data on occupancy and could be efficiently implemented across
large areas if snow and ice conditions were favorable (see Burt 2014: Chapters 2,3).
Like other techniques used to measure hare occupancy, abundance, and habitat use,
snow track surveys also have limitations. Although Hartman (1960) found that snow track
indices for hares in southern Ontario were relatively consistent between early and late winter,
the number of tracks encountered on transects can be affected by the hare population cycle,
weather, predators, and social interactions (Thompson et al. 1989). For example, during the low
phase of the population cycle snowshoe hares may only occupy prime habitats (Keith and
Windberg 1978). Conroy et al. (1979) found that distance to lowland coniferous-hardwood
habitats and habitat interspersion were the 2 most important vegetation factors determining
snowshoe hare activity (and hence track deposition) in Michigan. I contend that variability in
track deposition resulting from population and environmental factors supports the use of

18

lengthy transects to increase the likelihood of encountering a localized movement and likely
limits the population response variable of interest to site occupancy.
Although I did not incorporate the effects of short-term weather on the track survey
results, I encourage researchers implementing track surveys to document the weather between
the end of snowfall and the start of the track survey. The surveys were conducted during
periods of low wind and without precipitation so that deposited tracks were not obscured.
Snowshoe hare movements during winter in the Laurentian Mountains of Quebec was
positively influenced by lack of wind, overcast skies, and warmer temperatures (Bider 1961),
consistent with research from Minnesota where hares remained close to forms (e.g., in or
under hollow logs and under fallen trees) during inclement weather (i.e., high winds, rain,
snowstorms; Aldous 1937). Hartman (1960) also found that hares reduced movements during
rain, and further hypothesized that hare movements were inversely correlated with the
abundance of suitable food.
For single species monitoring, the most commonly used state variable of interest is
abundance or population size (MacKenzie et al. 2006:6). Hence, monitoring programs that
accurately quantify occupancy across large spatial extents (to assist with documenting range
shifts) while simultaneously estimating or indexing abundance are desirable. Limited data from
my study suggests that hare densities and the results of track surveys are positively related,
though the relationship appears to be non-linear and only applicable to a broad range of hare
densities (Fig. 1.4). If transects are spaced far enough apart (75 to 100m in my study) in a fixed
area, the number of occupied transects should theoretically relate to the number of hares using
that area. Although further research is needed to verify whether the observed relationship
19

between occupied transect counts and hare density is robust, I caution that such an approach
should be viewed as an index to abundance and not used to estimate density, primarily because
I likely double-count individual hares during snow transect surveys that are based on a grid
design.
1.6. ACKNOWLEDGEMENTS
I am thankful for the assistance of MDNR-Wildlife Division employees J. Belman, D. Beyer, D.
Brown, J. Lukowski, K. Sitar, K. Swanson, T. Swearingen, and C. VanHorn for their assistance
with fence construction and/or data collection at the Cusino Wildlife Research Area. Also,
thanks to S. Beyer from MDNR for helping me manage the social and political aspects of the
project. In addition, thanks to R. Aikens and B. Sillet of the Sault Ste. Marie Tribe of Chippewa
for their help in fence construction at Cusino, the replication of the research at Kinross, and
their commitment as a partner to this research project. Funding for this project was provided
by the Michigan Department of Natural Resources – Wildlife Division through the Michigan
Federal Aid in Wildlife Restoration program grant F13AF01268 in cooperation with the U.S. Fish
and Wildlife Service, Wildlife and Sport Fish Restoration Program and Safari Club International
Michigan Involvement Committee.

20

APPENDIX

21

Table 1.1. Snowshoe hare densities, track survey efforts, location of surveys, and year(s) that
surveys were completed in ~6.1ha enclosures, Upper Peninsula of Michigan, USA.
Hare Density
Number of Surveys
Location
Year completed
(hares/ha)
1.5
2
Cusino
2012
1.2

4

Cusino

2012, 2013

0.8

2

Cusino

2013

0.5

1

Cusino

2013

0.2

5

Kinross

2014

22

Table 1.2. Transect dimensions, the number of transects in a ~6.1ha study site, and distance
traversed during snow track surveys for snowshoe hares. Table includes the 10 combinations of
transect length and spacing that produced ≥90% accuracy for estimating site occupancy. The
Efficiency is determined by the total distance walked for a complete survey.
Transect Dimensions

a

Length

Spacing

Number of Transects

Distance Traversed (m)a

75

25

33

3275

100

25

25

3100

125

25

20

2975

150

25

17

2950

100

50

13

1900

125

50

10

1700

150

50

9

1750

125

75

7

1325

150

75

6

1275

150

100

5

1150

((Length + Spacing) * Number of Transects-1) + Length)

23

Figure 1.1. The geographic range of snowshoe hares (diagonal striping), with Michigan outlined
in bold black (map inset), and the locations of Cusino and Kinross hare enclosures, Upper
Peninsula of Michigan, winters of 2012-2014. The geographic range of hares was adapted from
the International Union for Conservation of Nature (www.iucnredlist.org/technicaldocuments/spatial-data#mammals).

24

Figure 1.2. Schematic of how snowshoe hare tracks in enclosures were subsampled including
(A) the outline of the enclosures (black) with 9 transects spaced 25m apart (gray), (B) snowshoe
hare tracks during 1 survey as black circles, (C) random locations generated on each transect to
denote the start of transect segments for 1 iteration as black 4-pointed stars, (D) transect
segments of 25m on each transect as black lines originating from the black 4-pointed stars, (E)
snowshoe hare tracks that intercepted transect segments, and (F) example of the results of 1
iteration with 100m spacing.

25

Figure 1.3. Proportion of 10,000 iterations that a site was correctly deemed as occupied for all
transect configurations at hare densities of (A) 1.5, (B) 1.2, (C) 0.8, (D) 0.5, and (E) 0.2 hares/ha.
The 90% accuracy threshold is designated by the black horizontal line.

26

Figure 1.4. The average number of transects occupied at each of the 5 hare densities for the 3
most efficient transect configurations: 125 m length x 75 m spacing, 150 m length x 75 m
spacing, and 150 m length x 100 m spacing.

27

LITERATURE CITED

28

LITERATURE CITED

Aldous, C. M. 1937. Notes on the life history of the snowshoe hare. Journal of Mammalogy
18:46-57.
Bailey, R. G. 1995. Laurentian mixed forest province. Description of ecoregions of the United
States. <http://www.fs.fed.us/land/ecosysmgmt/>. Accessed 18 May 2014.
Bider, J. R. 1961. An ecological study of the hare Lepus americanus. Canadian Journal of Zoology
39:81-102.
Brocke, R. H. 1975. Preliminary guidelines for managing snowshoe hare habitat in the
Adirondacks. Transactions of the Northeast Fish and Wildlife Conference 32:46‐66.
Buehler, D. A., and L. B. Keith. 1982. Snowshoe hare distribution and habitat use in Wisconsin.
Canadian Field‐Naturalist 96:19‐29.
Burt, D. M. 2014: in press. Factors influencing snowshoe hare occupancy in Michigan. Thesis,
Michigan State University, Michigan, USA.
Conroy, M. J., L. W. Gysel, and G. R. Dudderar. 1979. Habitat components of clear‐cut areas for
snowshoe hares in Michigan. Journal of Wildlife Management 43:680‐690.
Fortin, M. J., T. H. Keitt, B. A. Maurer, M. L. Taper, D. M. Kaufman, and T. M. Blackburn. 2005.
Species’ geographic ranges and distributional limits: pattern analysis and statistical
issues. Oikos 108:7-17.
Hartman, F. H. 1960. Census techniques for snowshoe hares. Thesis, Michigan State University,
East Lansing, MI, USA.
Hodges, K. E. 2000. Ecology of snowshoe hares in southern boreal and montane forests. Pages
163-207 in L.F. Ruggiero, K.B. Aubry, S.W. Buskirk, G.M. Koehler, C.J. Krebs, K.S.
McKelvey and J.R. Squires editors. Ecology and Conservation of Lynx in the United
States. University Press of Colorado, Boulder, CO, USA.
Hodges, K. E., and L. S. Mills. 2008. Designing fecal pellet surveys for snowshoe hares. Forest
and Ecology Management 256:1918-1926.

29

Hoving, C. L., Y. M. Lee, P. J. Badra, and B. J. Klatt. 2013. Changing climate, changing wildlife: a
vulnerability assessment of 400 species of greatest conservation need and game species
in Michigan. Michigan Department of Natural Resources Wildlife Division Report
Number 3564, Lansing, MI, USA.
IPCC. 2013. Summary for Policymakers. In: Climate Change 2013: The Physical Science Basis.
Contribution of Working Group I to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge,
United Kingdom and New York, NY, USA.
Keith, L. B., and L. A. Windberg. 1978. A demographic analysis of the snowshoe hare cycle
Wildlife Monographs 58:3-70.
Keith, L. B. 1990. Dynamics of snowshoe hare populations. Pages 119-195 in H. H. Genoways,
editor. Current mammalogy. Penum Press, New York, NY, USA.
Koehler, G. M. 1990. Population and habitat characteristics of lynx and snowshoe hares in north
central Washington. Canadian Journal of Zoology 68:845‐851.
Litvaitis, J. A., J. A. Sherburne, and J. A. Bissonette. 1985. A comparison of methods used to
examine snowshoe hare habitat use. Journal of Wildlife Management 49:693‐695.
Lu, X. 2011. Habitat use and abundance of the woolly hare Lepus oiostolus in the Lhasa
mountains, Tibet. Mammalia 75:35-40.
Marcström, V., L. B. Keith, E. Engren, and J. R. Cary. 1989. Demographic responses of arctic
hares (Lepus timidus) to experimental reductions of red foxes (Vulpes vulpes) and
martens (Martes martes). Canadian Journal of Zoology 67:658-668.
MacKenzie, D. I., J. D. Nichols, J. A. Royle, K. H. Pollock, L. L. Bailey, and J. E. Hines. 2006.
Occupancy Estimation and Modeling: Inferring Patterns and Dynamics of Species
Occurrence. Academic Press, New York, USA.
Mills, L. S., M. Zimova, J. Oyler, S. Running, J. T. Abatzoglou, and P. M. Lukacs. 2013. Camouflage
mismatch in seasonal coat color due to decreased snow duration. Proceedings of the
National Academy of Science 110:7360-7365.
R Core Team. 2014. R: A language and environment for statistical computing. R foundation for
Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.
Roy, C., L. Imbeau, and M. J. Mazerolle. 2010. Transformation of abandoned farm fields into
coniferous plantations: Is there enough vegetation structure left to maintain winter
habitat of snowshoe hares? Canadian Journal of Zoology 88:579-588.
30

Shimizu, R., and K. Shimano. 2010. Food and habitat selection of Lepus brachyurus lyoni Kishida,
a near-threatened species on Sado Island, Japan. Mammal Study 35:169-177.
Sikes, R. S., W. L. Gannon, and The Animal Care and Use Committee of the American Society of
Mammalogists. 2011. Guidelines of the American Society of Mammalogists for the use
of wild mammals in research. Journal of Mammalogy 92:235-253.
Thompson, I. D., I. J. Davidson, S. O’Donnell, and F. Brazeau. 1989. Use of track transects to
measure relative occurrence of small boreal mammals in uncut forest regeneration
stands. Canadian Journal of Zoology 67:1816‐1823.

31

CHAPTER 2
CLIMATIC FACTORS INFLUENCING THE DISTRIBUTION OF SNOWSHOE HARES

ABSTRACT
I aimed to: (1) compare the effects of different climate factors on snowshoe hare (Lepus
americanus) localized extinction probability in Michigan, USA; (2) inform recently published
vulnerability assessments; (3) identify the most influential climate variables affecting snowshoe
hare distribution; and (4) provide insights into population-level mechanisms causing localized
extinctions. I implemented track surveys at historically occupied sites that were identified by
interviewing snowshoe hare hunters and wildlife biologists. I identified climate variables that
potentially described population demographics of snowshoe hares. I modeled the likelihood
that a localized (approximately 7.5 ha) snowshoe hare population went extinct from year of last
known occupancy. I created a candidate model set from all possible combinations of
uncorrelated climate variables and used Akaike Information Criterion to rank the models and
95% confidence internals to estimate parameter significance. The top-ranking model for
describing localized snowshoe hare extinction in Michigan was based on the maximum
temperature from May 15 – January 19. As maximum temperature increased the likelihood of
localized extinction increased. The 2nd ranking model was based on total number of days with
measurable snow on the ground. As the number of days with snow on the ground decreased,
the likelihood for localized extinction increased. I found that the current distribution of
snowshoe hares at the southern edge of the range shifted northward by ~45 km over the last
20 years. Average climate conditions over time were correlated with localized extinction of
32

snowshoe hare in Michigan. Warmer temperatures and fewer days with snow on the ground
were correlated with a northward shift of snowshoe hare distribution. Population mechanisms
potentially linked to the localized effects of these climate variables include reduced litter sizes
and increased predation. Interactions with additional factors (e.g., habitat quality, predatorprey relationships) will likely exacerbate this northward shift in snowshoe hare distribution.
Key Words
Climate change, geographical range, Lepus americanus, logistic model, Michigan, snowshoe
hare, snow track surveys, species distribution
2.1 INTRODUCTION
Climatic factors are often thought to limit species ranges and these limits shift with
changes in climate (Araújo and Rozenfeld 2014). Models project warmer and drier climates
during the 21st century for many parts of the world (IPCC 2013). Warmer and drier climates
negatively affect species that are uniquely adapted to snowy or cold environments (Thomas et
al. 2004, IPCC 2007, Pereira et al. 2010), such as American marten (Martes americana),
American pika (Ochotona princeps), Arctic lemming (Dicrostonyx torquatus), Canada lynx (Lynx
Canadensis), fisher (Martes pennanti), moose (Alces alces), and snowshoe hare (Lepus
americanus). For example, Wasserman et al. (2012) predicted that by 2080, climate change
would reduce American marten habitat connectivity by over 50% in the Rocky Mountains, USA,
resulting in decreased genetic diversity and a constantly declining population. Similarly,
American marten and lynx populations in the northern Appalachian and Acadian Ecoregion
(eastern USA) are expected to decline 40% and 59%, respectively, by 2055 because of climate
change (Carroll 2007). Carroll (2007) posits that the additive effects of logging and trapping will
33

exacerbate vulnerability for these species (Carroll 2007). The population of Arctic lemmings in
northern Russia has declined since the Last Glacial Maximum and future climate is projected to
further reduce the population, eliminate genetic diversity, and cause local extinctions (Prost et
al. 2010).
The process of localized extinction and colonization is often a precursor to range
contraction or expansion, respectively. For winter-adapted species, localized extinction caused
by warming or drying climates can be difficult to document because the effects are potentially
confounded by other stressors like changes in food abundance, competitors, or mismatched
timing with host species (Cahill et al. 2013). Some species (e.g. moose) exhibit direct behavioral
responses (e.g., changes in foraging) to changing climatic conditions that ultimately can be
linked to population-level processes (van Beest et al. 2012). Generally, evidence is sparse for
direct relationships between species distributions or population performance and climate
change (Cahill et al. 2013). For most species, climate affects are inferred from the spatial
distribution of occupied and unoccupied sites over time.
Snowshoe hares are adapted to environments with abundant snowfall and prolonged
winters and hence populations are sensitive to climate change (Mills et al. 2013). Michigan is at
the southernmost boundary for the range of snowshoe hare in the Lake States region, North
America (Fig. 2.1). Buehler and Keith (1982) noted that hare populations in the Lake States
region will likely shift northward in response to a warming and drying climate because of
reduced duration of snow cover and depth. The population-level mechanisms causing this
potential range shift are not understood. Plausible hypotheses relate to direct effects on
population vital rates and indirect effects on individual life history strategies (Bardsen et al.
34

2011). For example, some have suggested that hare productivity is lower during years with
warmer summers, falls, and winters (Meslow and Keith 1971, Kielland et al. 2010; Table 2.1);
others have indicated that survival is lower during winters with more snow-free days (Kielland
et al. 2010, Mills et al. 2013; Table 2.1). Most support to date exists for the survival hypothesis
as it relates to higher predation rates (Hone et al. 2011).
During various phases of the 10-year population cycle at southern latitudes, mortality of
snowshoe hares from predation can be >90% (Hodges 2000). Concealment is recognized as a
leading evolutionary force that influences pelage coloration among mammals (Caro 2005);
molting at the proper time is critical to the survival of individual hares (Mills et al. 2013). Molt
initiation dates for hares in Maine and Montana (USA) began in late September and early April,
and research to date indicates that onset of the fall molt exhibits low plasticity (Severaid 1945,
Mills et al. 2013, Zimova et al. 2014). Climate data suggest that the number of days with snow
on the ground throughout much of the southern distribution of hare range has decreased by up
to 25 days from 1972-2004 (Choi et al. 2010). By the end of the 21st century, duration of snow
cover throughout the entire range of snowshoe hares is expected to significantly decrease
(Brown and Mote 2009). Hence, the number of days with snow on the ground is likely an
important determinant of localized extinction for snowshoe hares.
I compared the effects of different climate factors on snowshoe hare distribution in
Michigan, USA. I modeled how changes in climatic variables influenced localized (7.5 ha)
extinction probability at sites throughout the state. My results have relevance to recently
published vulnerability assessments (e.g., Hoving et al. 2013) and provide indirect insights into

35

potential population-level mechanisms causing localized extinctions that ultimately lead to
changes in broad-scale distribution of snowshoe hares.
2.2 STUDY AREA
My study occurred in the Upper and northern Lower Peninsulas of Michigan on
publically owned lands during the winter of 2013 (Fig. 2.1). This spatial extent corresponds to
the known historic distribution of snowshoe hares in Michigan (Fig. 2.1). My study consists of
three broad regions: 1) western Upper Peninsula, 2) eastern Upper Peninsula, and 3) northern
Lower Peninsula. Landforms of the western Upper Peninsula consist of glacial moraines, lake
plains, outwash channels and plains, and glacially scoured bedrock ridges (Albert 1995). Primary
vegetation types include northern hardwoods, aspen (Populus spp.), pine (Pinus spp.), and
conifer swamps (Albert 1995). The western Upper Peninsula experiences the most extreme
winters and shortest growing seasons of the three regions I studied (Albert 1995). Monthly
average temperatures ranged from ‐10⁰C to 17⁰C, with average annual precipitation of 90 cm,
including 435 cm of snowfall (NOAA 2012).
Landforms of the eastern Upper Peninsula consist mostly of flat lake plain with areas of
exposed bedrock (Albert 1995). Vegetation types include northern hardwoods, upland conifers
(white pine (P. strobus), red pine (P. resinosa), jack pine (P. banksiana)), hardwood‐conifer
swamps, rich conifer swamps, and northern wet meadows (Albert 1995). Monthly average
temperatures vary from ‐10⁰C to 18⁰C with average annual precipitation of 80 cm, including
194 cm of snowfall (NOAA 2012).
Landforms in the northern Lower Peninsula (NLP) consist of large, sandy outwash plains
and large glacial moraines (Albert 1995). The climate of coastal areas in the NLP is moderated
36

by Lakes Huron and Michigan resulting in warmer and cooler temperatures in winter and
summer, respectively. Interior areas are subjected to more extreme temperature fluctuations.
Forest types include northern hardwoods, aspen, oak (Quercus spp.), pine, and lowland conifer
(Albert 1995). Monthly average temperatures range from ‐8⁰C in January to 20⁰C in July, with
average annual precipitation of 90 cm, including 169 cm of snowfall (NOAA 2012).
2.3. METHODS
2.3.1. Site Selection
I interviewed 45 individuals that had knowledge of historical snowshoe hare locations in
Michigan. Interviewees included Michigan Department of Natural Resources employees, hare
hunters, and tribal members. I identified potential interviewees by contacting various hunt
clubs, attending sportsperson coalition meetings, talking with colleagues, and by individuals
providing contact information for hare hunters they personally knew. Each interviewee was
asked to map locations that were unequivocally occupied by snowshoe hares sometime in the
past. Unequivocal evidence included tracks, harvested hares, or observed hares. Each
interviewee was asked to estimate the year of last confirmed occupancy within 5 yrs. I also
checked museum records (e.g. VertNet or MaNIS) for historical snowshoe hare collection
locations, but opted not to use these records because the locational error was generally large
(i.e. >1km). Islands of Michigan (e.g. Drummond Island or North Manitou Island) were not
added to the list of potential study locations. I compiled potential study sites from the
interviews and selected a subset of locations that encompassed broad spatial (throughout hare
range in Michigan; Fig. 2.1) and temporal (1955-2010; Fig. 2.2) domains. The 7.5 ha study sites
were separated by >1.6 km (mean = 10.2 km, SE = 0.7). My interview protocol was reviewed
37

and approved by the Human Research Protection Program, Institutional Review Board, at
Michigan State University (IRB# x11-805).
2.3.2. Field Sampling
Winter track counts along transects are commonly used to document snowshoe hare
occupancy and habitat use (Brocke 1975, Conroy et al. 1979, Thompson et al. 1989). However,
transect dimensions and configurations are inconsistent among studies. I evaluated the effects
of transect number (2-9), length (10-150m), and spacing (25-125m) on my ability to accurately
portray site occupancy using a known number of telemetered snowshoe hares in a 6 ha
enclosure during the winters of 2012 and 2013. Prior to releasing telemetered hares, I surveyed
the enclosure to ensure that no hares were present. I demarcated 9 transects, spaced 25 m
apart, ranging in total length from 225-250 m. Track surveys were conducted 12-65 hours after
fresh snowfall. Track locations along transects were mapped to the nearest 5m and random
sub-sampling of transect lengths and spacing was used to identify the most efficient transect
configuration, where efficiency was defined as the highest probability of correctly denoting sitelevel occupancy. I replicated transect sampling at densities ranging from 0.2 to 1.5 hares/ha.
These densities represented low to high snowshoe hare densities recorded at the southern
edge of hare range (Hodges 2000). Snowshoe hare trapping and radio-collaring were conducted
by the Michigan Department of Natural Resources – Wildlife Division following internal animal
use and care guidelines that were consistent with the American Society of Mammalogists
Animal Care and Use Guidelines (Sikes et al. 2011).
I implemented the most efficient transect methodology from the enclosure study at a
subset of locations selected from the interview process (Fig. 2.1). Sampling for snowshoe hares
38

occurred on a single day and was completed 12-72 hours after fresh snowfall to allow for track
accumulation. If a track was detected on ≥1 transect at a site, then the site was designated as
occupied. Sites without tracks were deemed unoccupied and presumed to have experienced
localized extinction since the time of last known occupancy.
2.3.3. Climate Covariates
A priori, I conducted a review of scientific publications that evaluated how climate
affected snowshoe hare population demographics. I identified 4 precipitation and 3
temperature variables (Table 2.1). I note that most of the observed climate effects on hares
were indirect, often based on speculation as to the physiological mechanism(s) causing the
observed population response. The precipitation variables included mean snow depth (SD) in
an annual snow season (approximately November through April), total number of days with
measurable snow on the ground (DSoG), total number of days with a snowfall event (DSE), and
total spring precipitation from March 20 – June 20 (spring precipitation; Table 2.1).
Temperature variables included mean minimum temperature from February 4 – April 24
(minimum winter temperature), mean maximum temperature from May 15 – January 19
(maximum temperature), and mean winter temperature from December 21 to March 19 (mean
winter temperature; Table 2.1).
I compiled climate covariates for each snowshoe hare survey site from the closest
National Oceanic and Atmospheric Administration (NOAA) weather station that had archived
data for all 7 variables dating back to the year of last known hare occupancy
(http://www.ncdc.noaa.gov/; Fig. 2.2). I used 35 weather stations and evaluated multiple
approaches for summarizing climate data across time (e.g., year-to-year variation, total
39

difference, and deviance) and, based on model weight of evidence, found that average values
calculated from the year of last known occupancy to 2012 were most useful in portraying hare
extinction probability. I note that although climate data from some sites came from the same
weather station, times of last known occupancy often varied among those sites so the average
climate conditions often differed.
2.3.4. Data Analysis
I analyzed all combinations of climate covariates for colinearity using Pearson’s
correlation coefficient (Sokal and Rohlf 2011); covariates were considered correlated if r ≥
|0.15| and p ≤ 0.05. I modeled the likelihood that a localized (approximately 7.5 ha) snowshoe
hare population went extinct from year of last known occupancy using logistic regression. I
included a random effect for each weather station to account for autocorrelation among sites
that used the same weather station data. I created a candidate set of models (n=9; Table 2.2)
from all possible combinations (additive terms) of uncorrelated climate variables. Models were
ranked based on Akaike Information Criterion (AIC) and parameter estimates were deemed
significant if the 95% confidence intervals did not overlap 0 (Burnham and Anderson 2002,
Nakagawa and Cuthill 2007).
2.4. RESULTS
My interviews resulted in 386 potential study sites with year of last known hare
occupancy ranging from 1955 to 2010; climate data were compiled for 134 of those sites (Fig.
2.1). Most climate variables were correlated except for spring precipitation and DSoG and
spring precipitation and SD. Regardless of the climate variable I documented a broad range of
minimum and maximum values and considerable variation in measurements across sites (Table
40

2.3). These results typify the temperature and precipitation variation often found at small
scales in Michigan that are caused by a combination of lake effect, topography, and prevailing
direction of weather fronts.
From the captive hare study, I found that reliable (i.e., occupancy probability correctly
designated >95% of the time) estimates of hare occupancy were obtained by surveying 9 –
125m long transects spaced 75m apart. After implementing this transect configuration on the
134 sites spread throughout Michigan, I found that hare populations experienced localized
extinction on 52 sites (39%) statewide. In the northern Lower Peninsula 36 of 74 sites (~49%)
were unoccupied during 2013. Only 16 out of 60 (~27%) sites were unoccupied in the Upper
Peninsula. My results indicate almost a two-fold increase in localized extinction probability from
north to south in Michigan.
All candidate models for describing the localized extinction of snowshoe hares received
some support (i.e., ωi, ≥ 0.01), but a clear top-ranking model emerged based on weight of
evidence (Table 2.2). My top-ranking model accounted for 52% weight of evidence and was
based on maximum temperature (Table 2.2). As maximum temperature increased the
likelihood of localized extinction significantly increased (β = 0.043, 95% CI = 0.014 – 0.073; Table
2.2). The next ranked model (ΔAIC = 2.6) accounted for 15% weight of evidence and was based
on DSoG (Table 2.2). The DSoG indicated that as the number of days with snow on the ground
decreased, the likelihood for localized extinction significantly increased (β = -0.026, 95% CI = 0.026 – -0.004; Table 2.3). All other models received minimal support (i.e., ωi, ≤0.11) and only
one additional significant covariate was identified (SD). As SD decreased, the likelihood for
localized snow hare extinction increased (β = -0.004, 95% CI = -0.0078 - -0.0003; Table 2.3).
41

Mean maximum temperature from time of last known hare occupancy across the study
sites ranged from 13 to 17°C and corresponded to a 3-fold increase in localized extinction
probability (Fig. 2.3A). Average days with snow on the ground in the study ranged from 60 to
140; sites with fewer DSoG were >3 times more likely to go extinct (Fig. 2.3B). Average snow
depths in the study ranged from 100 to 600mm; hares occupying sites with greater snow depth
were ~1.5 times less likely to go extinct (Fig. 2.3C). Because these three climate covariates were
correlated and significant in my models, my results suggest that maximum temperature, DSoG,
and SD have an integrated effect on localized extinction probabilities for snowshoe hares.
2.5. DISCUSSION
I sampled a broad spatial and temporal domain throughout the range of snowshoe
hares in Michigan and found that climate variables were significantly correlated to localized
extinction probability. My findings substantiate assessments that hares are vulnerable to
climate change (Hoving et al. 2013; Mills et al. 2013). I demonstrated a greater proportion of
extinct sites closer to the southern range periphery, but also found extinct sites in more
centrally located areas suggesting that factors other than climate are also affecting localized
snowshoe hare extinctions. Sites with warmer temperatures from summer to early winter,
fewer days with snow on the ground, and shallower snow depths were more likely to go
extinct. Population mechanisms potentially linked to localized hare extinction include reduced
litter sizes from higher temperatures (Meslow and Keith,1971), increased predation from fewer
days with snow on the ground (Kielland et al. 2010, Mills et al. 2013), and less food availability
during periods of low snow depths (Bider 1961, Meslow and Keith 1971, Conroy et al. 1979).

42

My results are consistent with other studies that have explored climate-related
mechanisms associated with population declines in other species. For example Lee et al. (2000)
found that warmer weather and wetter winters resulted in fewer reindeer (Rangifer tarandus)
calves in northern Finland and Norway. Higher total rainfall in spring was correlated with a
smaller proportion of wild pig (Sus scrofa scrofa L.) females to breed the following season
(Sabrina et al. 2009). While deep snows impose higher energetic costs to hares (Hodges et al.
2006), similar to findings observed for Alpine red deer (Cervus elaphus sp.; Schmidt 1993,
Rivrud et al. 2010) and moose (Dussault et al. 2005), my results indicate that deeper snows may
benefit snowshoe hares by increasing the availability of browse on portions of plants that were
not available during the summer or early winter (Bider 1961, Meslow and Keith 1971, Conroy et
al. 1979).
Although my results suggest that multiple climatic factors influence snowshoe hare
extinction probability, these factors likely interact with predator-prey relationships, vegetation
structure, spatial arrangement of habitats, and food availability. Predators are the primary
cause of mortality for snowshoe hares and early winter is a time of increased predation from
mammalian predators (Hodges 2000). With a changing climate that results in fewer days with
snow on the ground, coupled with a relatively fixed timing of the fall molt in hares (Mills et al.
2013), the camouflage advantages of white pelage are negated. Increased predation pressure
resulting from compromised camouflage can also have negative indirect effects on hare body
condition and fecundity (Hodges et al. 1999). Increased predation pressure, indirect effects of
predation on fitness, and decreased litter sizes associated with warmer maximum temperatures
likely interact to favor localized extinction of hares.
43

Climate can also potentially affect snowshoe hares indirectly through climate induced
changes to forest communities. Increasing temperatures in northern latitudes are expected to
change forest composition from needle-leaved (e.g. Abies spp., Betula spp., Picea spp., Pinus
banksiana, and Populus spp.) to broad-leaved (e.g. Fagus spp., Picea spp., and Quercus spp.)
landscapes and decrease forest productivity and total biomass (Prentice et al. 1991, Prentice et
al. 1993, Lenihan et al. 2003, Duveneck et al. 2014). The potential loss of conifer dominated
forests to deciduous forests could negatively impact hares because conifers are 3 times more
important to hares (Litvaitis et al. 1985). A meta-analysis found that the range limits of 99
species have moved northwards by 6.1 km per decade on average and forests in northern
Michigan may experience conditions similar to those about 2˚ south causing reduced species
regeneration success (Reed and Desanker 1992, Parmesan and Yohe 2003). Correspondingly,
this northward transition of critical hare habitat components may also result in a northern shift
in hare distribution. A changing climate may also fragment habitats (Root et al. 2003). Hare
habitat in the Upper Peninsula of Michigan is already considered fragmented (Linden et al.
2010) and climate change may exacerbate those negative effects on hare populations.
Most of the Upper Peninsula of Michigan is estimated as poor to marginal snowshoe
hare habitat (Linden et al. 2010). Snowshoe hare populations tend to thrive in continuously
forested areas with dense understories (>60% visual obstruction) that provide food, safety from
predators, and thermal cover (Litvaitis et al. 1985). Hare survival rates are lower in areas with
sparse horizontal cover (Litvaitis et al. 1985). Sparse understory cover further exacerbates
higher predation pressures resulting from fewer days with snow on the ground. Additionally,
lower mean snow depths can restrict access to forage during critical winter months (Bider 1961,
44

Meslow and Keith 1971, Conroy et al. 1979). Although food availability does not appear to limit
hares during winter (Conroy et al. 1979, Carreker 1985), low levels of understory cover coupled
with decreased snow depths potentially concentrates hares and restricts access to high quality
forage. The effective browse height for snowshoe hares is about 46 cm and browsing efficiency
is maximized when snow depths increase throughout the winter making new browse available
(Bider 1961).
My results indicate that future climate change will move the southern edge of snowshoe
hare range northward. Abundant and diverse predators and poor habitat conditions will likely
exacerbate this change. In Michigan, I found that the current distribution of snowshoe hares at
the southern edge of the range shifted northward by ~45 km over the last 20 years. An
important unanswered question is whether management focused on reducing predator
numbers or improving hare habitat quality and connectivity can help reduce the negative
impacts of climate change. The loss of snowshoe hares from Michigan would have important
ecological and cultural ramifications and hence developing and testing strategies for conserving
this species during periods of rapid climate change are needed.
2.6. ACKNOWLEDGEMENTS
I thank P. Nelson for serving as a field technician during the winter of 2013 and K.
Wilson for collecting climate data. I am also thankful for the assistance of Michigan Department
of Natural Resources - Wildlife Division employees D. Brown, T. Lyon, T. Maples, V. Richardson,
B. Rollo, J. Valentine, and M. Wegan for helping with field data collection. I appreciate the
numerous interviewees who graciously made themselves and their knowledge available for
identifying historic hare sites. Also, thanks to T. Minzey for coordinating my attendance at
45

sportsperson coalition meetings and S. Beyer for helping me manage the social and political
aspects of the project. Funding for this project was provided by the Michigan Department of
Natural Resources – Wildlife Division through the Michigan Federal Aid in Wildlife Restoration
program grant F13AF01268 in cooperation with the U.S. Fish and Wildlife Service, Wildlife and
Sport Fish Restoration Program and Safari Club International Michigan Involvement Committee.

46

APPENDIX

47

Table 2.1. Climate covariates identified from the literature review, associated citation, and observed relationship to snowshoe hare
population demographics.
Variable
Publication(s)
Importance to hares
Mean snow depth (SD)

Bider 1961, Meslow and Keith
1971, Conroy 1979, Hodges et al.
2006, Kielland et al. 2010

Increased snow depth creates different browse
availability and increased energetic costs

Total number of days with
measureable snow on the
ground (DSoG)

Kielland et al. 2010, Mills et al.
2013

Incorrect pelage coloration for crypsis with lack of snow
on the ground during winter months, presumably
decreasing survival

Total number of days with a
snowfall event

Kielland et al. 2010, Meslow and
Keith 1971

Decreased juvenile and adult survival with more
snowfall

Total spring precipitation
(March 20 – June 20)

Kielland et al. 2010

Duration of decline phase associated with increased
spring precipitation

Mean minimum temperature
(February 4 – April 24)

Meslow and Keith 1971

Warmer minimum temperatures during this time
period, the greater the adult survival

Mean maximum temperature
(May 15 – January 19)

Meslow and Keith 1971

Colder temperatures during this period were associated
with larger litter sizes in subsequent breeding seasons

Mean winter temperature
(December 21 – March 19)

Kielland et al. 2010

Warmer temperatures during this time period
associated with the decline phase of the snowshoe
hare cycle

48

Table 2.2. Candidate model set for estimating the likelihood of a localized snowshoe hare
population going extinct in Michigan. LL = log-likelihood, ∆AIC = difference in AIC value from
top-ranking model, k = model parameters, ωi – Akaike weight of evidence.
Candidate Modela
LL
∆AIC
K
ωi
Maximum temperature

-85.23

0.0

3

0.52

Days snow on ground (DSoG)

-86.51

2.6

3

0.15

DSoG + Spring precipitation

-85.80

3.1

4

0.11

Snow Depth (SD)

-87.04

3.6

3

0.09

SD + Spring precipitation

-86.59

4.7

4

0.05

Days snowfall event (DSE)

-88.10

5.7

3

0.03

Mean winter temperature

-88.34

6.2

3

0.02

Spring precipitation

-88.42

6.4

3

0.02

Minimum winter temperature

-89.10

7.7

3

0.01

a

Maximum temperature = mean maximum temperature between May 15 and January 19; Days
snow on ground = Total number of days with measurable snow on the ground in an annual
snow season (approximately November through April); Spring precipitation = total spring
precipitation from March 20 thru June 20; Snow depth = Mean snow depth in an annual snow
season; Days snowfall event = Total number of days with a snowfall event in an annual snow
season; Mean winter temperature = Mean winter temperature from December 21 thru March
19; Minimum winter temperature = Mean minimum temperature from February 4 thru April
24.

49

Table 2.3. Means (SE) and ranges of climate covariates used for estimating the likelihood of a
localized snowshoe hare population going extinct in Michigan.
Variable
Mean
SE
Range
Mean snow depth (SD; mm)

234.40

9.55

107 - 562

Total number of days with measureable snow on the
ground (DSoG)

108.26

1.59

61 - 145

Total number of days with a snowfall event

49.81

1.52

20 - 90

Total spring precipitation (March 20 – June 20; mm)

221.80

29.69

150.1 - 310.3

Mean minimum temperature (February 4 – April 24; °C)

-6.24

0.14

-9.8 - -2.7

Mean maximum temperature (May 15 – January 19; °C)

14.96

0.12

12.6 – 17.6

Mean winter temperature (December 21 – March 19; °C)

-5.52

0.14

-9.2 - -1.4

50

Figure 2.1. Location and occupancy status of snowshoe hare survey sites throughout the
northern Lower and Upper Peninsulas of Michigan, USA, winter 2013. The geographic range of
snowshoe hares and Michigan are portrayed in the map inset. The geographic range was
derived from the International Union for Conservation of Nature
(www.iucnredlist.org/technical-documents/spatial-data#mammals).

51

Figure 2.2. Number of study sites and year (5-year groupings through 2010) of last confirmed
snowshoe hare occupancy in the northern Lower and Upper Peninsulas of Michigan, USA.
30

Count of Study Sites

25
20
15
10
5
0
2012

2010

2005

52

2000

1995

1990

1985

1980

1975

1970

1965

1960

1955

Year

Figure 2.3. Localized snowshoe hare extinction probability and (a) mean maximum temperature
during May 15 – January 19, (b) number of days with snow on the ground, and (c) mean snow
depth in Michigan, USA and the 95% confidence intervals. The points along the x-axes represent
the values of each study site.

53

LITERATURE CITED

54

LITERATURE CITED

Albert, D.A. 1995. Regional landscape ecosystems of Michigan, Minnesota, and Wisconsin: a
working map and classification. United States Department of Agriculture, Forest Service
General Technical Report NC-178, North Central Forest Experiment Station, St. Paul,
Minnesota, USA.
Araújo, M. B., and A. Rozenfeld. 2014. The geographic scaling of biotic interactions. Ecography
37:406-415.
Bardsen, B., J. Henden, P. Fauchald, T. Tveraa, and A. Stien. 2011. Plastic reproductive allocation
as a buffer against environmental stochasticity – linking life history and population
dynamics to climate. Oikos 120:245‐257.
Bider J. R. 1961. An ecological study of the hare Lepus americanus. Canadian Journal of Zoology
39:81-103.
Brocke, R. H. 1975. Preliminary guidelines for managing snowshoe hare habitat in the
Adirondacks. Transactions of the Northeast Fish and Wildlife Conference 32:46‐66.
Brown, R. D., and P. W. Mote. 2009. The response of northern hemisphere snow cover to a
changing climate. Journal of Climate 22:2124-2144.
Buehler, D. A., and L. B. Keith. 1982. Snowshoe hare distribution and habitat use in Wisconsin.
Canadian Field‐Naturalist 96:19‐29.
Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference. Springer,
New York, NY, USA.
Cahill, A. E., M. E. Aiello-Lammens, M.C. Fisher-Reid, X. Hua, C. J. Karanewsky, H. Y. Ryu, G. C.
Sbeglia, F. Spagnolo, J. B. Waldron, O. Warsi, and J. J. Wiens. 2013. How does climate
change cause extinction? Proceedings of the Royal Society B: Biological Sciences
280:20121890.
Caro, T. 2005. The adaptive significance of coloration in mammals. BioScience 55:125-136.
Carreker, R. G. 1985. Habitat suitability index models: snowshoe hare. United States
Department of Interior, Fish and Wildlife Service, Biological Report 82:10.101, Fort
Collins, CO, USA.

55

Carroll, C. 2007. Interacting effects of climate change, landscape conversion, and harvest on
carnivore populations at the range margin: marten and lynx in the northern
Appalachians. Conservation Biology 21:1092-1104.
Choi, G., D. A. Robinson, S. Kang. 2010. Changing northern hemisphere snow seasons. Journal of
Climate 23:5305-5310.
Conroy, M. J., L. W. Gysel, and G. R. Dudderar. 1979. Habitat components of clear‐cut areas for
snowshoe hares in Michigan. Journal of Wildlife Management 43:680‐690.
Dussault, C., J. Ouellet, R. Courtois, J. Huot, L. Breton, and H. Jolicoeur. 2005. Linking moose
habitat selection to limiting factors. Ecography 28:619-628.
Duveneck, M. J., R. M. Scheller, M. A. White, S. D. Handler, and C. Ravenscroft. 2014. Climate
change effects on northern Great Lake (USA) forests: a case for preserving diversity.
Ecosphere 5: article 23. doi:10.1890/ES13-00370.1
Hodges, K. E., C. J. Krebs, and A. R. E. Sinclair. 1999. Snowshoe hare demography during a cyclic
population low. Journal of Animal Ecology 68:581-594.
Hodges, K. E. 2000. Ecology of snowshoe hares in southern boreal and montane forests. Pages
163-207 in L.F. Ruggiero, K.B. Aubry, S.W. Buskirk, G.M. Koehler, C.J. Krebs, K.S.
McKelvey and J.R. Squires editors. Ecology and Conservation of Lynx in the United
States. University Press of Colorado, Boulder, CO, USA.
Hodges, K. E., R. Boonstra, and C. J. Krebs. 2006. Overwinter mass loss of snowshoe hares in the
Yukon: starvation, stress, adaptation or artefact? Journal of Animal Ecology 75:1-13.
Hone, J., C. J. Krebs, and M. O’Donoghue. 2011. Is the relationship between predator and prey
abundances related to climate for lynx and snowshoe hares? Wildlife Research 38:419425.
Hoving, C. L., Y. M. Lee, P. J. Badra, and B. J. Klatt. 2013. Changing climate, changing wildlife: a
vulnerability assessment of 400 species of greatest conservation need and game species
in Michigan. Michigan Department of Natural Resources Wildlife Division Report
Number 3564, Lansing, MI, USA.
IPCC. 2007. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I
to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change (ed.
by S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Avery, M.Tignor and H.L.
Miller), Cambridge University Press, Cambridge, United Kingdom and New York, NY,
USA.

56

IPCC. 2013. Summary for Policymakers. In: Climate Change 2013: The Physical Science Basis.
Contribution of Working Group I to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change (ed. by T.F. Stocker, D. Qin, G.K. Plattner,
M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley),
Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.
Kielland, K., K. Olson, and E. Euskirchen. 2010. Demography of snowshoe hares in relation to
regional climate variability during a 10-year population cycle in interior Alaska. Canadian
Journal of Forest Research 40:1265-1272.
Lee, S. E., M. C. Press, J. A. Lee, T. Ingold, and T. Kurttila. 2000. Regional effects of climate
change on reindeer: a case study of the muotkatunturi region in Finnish Lapland. Polar
Research 19:99-105.
Lenihan, J. M., R. Drapek, D. Bachelet, and R. P. Neilson. 2003. Climate change effects on
vegetation distribution, carbon, and fire in California. Ecological Applications 13:16671681.
Linden, D. W., H. Campa, III, G. J. Roloff, D. E. Beyer, Jr., and K. F. Millenbah. 2010. Modeling
habitat potential for Canada lynx in Michigan. Wildlife Society Bulletin 35:20-26.
Litvaitis, J. A., J. A. Sherburne, and J. A. Bissonette. 1985. Influence of understory characteristics
on snowshoe hare habitat use and density. Journal of Wildlife Management 49:866-873.
Meslow, E. C., and L. B. Keith. 1971. A correlation analysis of weather versus snowshoe hare
population parameters. The Journal of Wildlife Management 35:1-15.
Mills, L. S., M. Zimova, J. Oyler, S. Running, J. T. Abatzoglou, P. M. Lukacs. 2013. Camouflage
mismatch in seasonal coat color due to decreased snow duration. Proceedings of the
National Academy of Science 110:7360-7365.
Nakagawa, S., and I. C. Cuthill. 2007. Effect size, confidence interval and statistical significance:
a practical guide for biologists. Biological Reviews 82:591-605.
National Oceanic and Atmospheric Administration [NOAA]. 2012. NOWData – NOAA online
weather data. <http://www.noaa.gov/>. Accessed 15 February 2012.
Parmesan, C., and G. Yohe. 2003. A globally coherent fingerprint of climate change impacts
across natural systems. Nature 421:37-42.

57

Pereira, H. M., P. W. Leadley, V. Proença, R. Alkemade, J. P. W. Scharlemann, J. F. FernandezManjarrés, M. B. Araújo, P. Balvanera, R. Biggs, W. W. L. Cheung, L. Chini, H. D. Cooper,
E. L. Gilman, S. Guénette, G. C. Hurtt, H. P. Huntington, G. M. Mace, T. Oberdorff, C.
Revenga, P. Rodrigues, R. J. Scholes, U. R. Sumaila, and M. Walpole. 2010. Scenarios for
global biodiversity in the 21st century. Science 330:1496-1501.
Prentice, I. C., M. T. Sykes, and W. Cramer. 1991. The possible dynamic response of northern
forests to global warming. Global Ecology and Biogeography Letters 1:129-135.
Prentice, I. C., M. T. Sykes, and W. Cramer. 1993. A simulation model for the transient effects of
climate change on forest landscapes. Ecological Modeling 65:51-70.
Prost, S., N. Smirnov, V. B. Fedorov, R. S. Sommer, M. Stiller, D. Nagel, M. Knapp, and M.
Hofreiter. 2010. Influence of climate warming on Arctic mammals? New insights from
ancient DNA studies of the collared lemming Dicrostonyx torquatus. PLoS ONE
5(5):e10447. doi:10.1371/journal.pone.0010447.
Reed, D. D., and P. V. Desanker. 1992. Ecological implications of projected climate change
scenarios in forest ecosystems in northern Michigan, USA. International Journal of
Biometeorology 36:99-107.
Rivrud, I. M., L. E. Loe, and A. Mysterud. 2010. How does local weather predict red deer home
range size at different temporal scales? Journal of Animal Ecology 79:1280-1295.
Root, T. L., J. T. Price, K. R. Hall, S. H. Schneider, C. Rosenzweig, and J. A. Pounds. 2003.
Fingerprints of global warming on wild animals and plants. Nature 421:57-60.
Sabrina, S., G. Jean-Michel, T. Carole, B. Serge, and B. Eric. 2009. Pulsed resources and climateinduced variation in the reproductive traits of wild boar under high hunting pressure.
Journal of Animal Ecology 78:1278-1290.
Schmidt, K. 1993. Winter ecology of nonmigratory Alpine red deer. Oecologia 95:226-233.
Severaid, J. H. 1945. Pelage changes in the snowshoe hare (Lepus americanus struthopus
Bangs). Journal of Mammalogy 26:41-63.
Sikes, R. S., W. L. Gannon, and The Animal Care and Use Committee of the American Society of
Mammalogists. 2011. Guidelines of the American Society of Mammalogists for the use
of wild mammals in research. Journal of Mammalogy 92:235-253.
Sokal, R. R., and F. J. Rohlf. 2011. Biometry. Fourth edition. Freeman, New York, USA.

58

Thomas, C. D., A. Cameron, R. E. Green, M. Bakkenes, L. J. Beaumont, Y. C. Collingham, B. F. N.
Erasmus, M. F. De Siqueira, A. Grainger, and L. Hannah. 2004. Extinction risk from
climate change. Nature 427:145–148.
Thompson, I. D., I. J. Davidson, S. O’Donnell, and F. Brazeau. 1989. Use of track transects to
measure relative occurrence of small boreal mammals in uncut forest regeneration
stands. Canadian Journal of Zoology 67:1816‐1823.
van Beest, F. M., B. Van Moorter, and J. S. Milner. 2012. Temperature-mediated habitat use and
selection by a heat-sensitive northern ungulate. Animal Behaviour 84:723-735.
Wasserman, T. N., S. A. Cushman, A. S. Shirk, E. L. Landguth, and J. S. Littell. 2012. Simulating
the effects of climate change on population connectivity of American marten (Martes
americana) in the northern Rocky Mountains, USA. Landscape Ecology 27:211-225.

59

CHAPTER 3
LAND COVER AND VEGETATION FACTORS INFLUENCING SNOWSHOE HARE OCCUPANCY IN
MICHIGAN

ABSTRACT
Snowshoe hares (Lepus americanus) depend on dense forest vegetation for
concealment, escape, and thermal cover. To quantify how surrounding land cover and finer
scale vegetation factors influenced snowshoe hare occupancy and habitat use in Michigan, I
conducted line-transect track surveys in 117 historically occupied (i.e. 1955-2010) sites during
winter 2012-2013. I found that 62% of the sites were still occupied. I created an 812m buffer
around my survey transects (~332 ha), compiled historical data from aerial photographs from
1978 to 2012, and computed net change over time for 5 land cover covariates. I evaluated 17
candidate models of various covariate combinations of land cover change over time. All land
cover change models had AIC weight >0.01 and ≤12 (ΔAIC <4.0) and none of the parameters
estimates were significant, indicating that land cover change over time was having a negligible
effect on current occupancy status for hares. I subsequently hypothesized that in forestdominated landscapes, current land cover and vegetation structure were more important
determinants of hare occupancy because this species is short-lived, highly mobile, and evolved
in disturbance-prone environments. I analyzed 2012 land cover covariates in combination with
transect-level covariates of woody stem densities and visual obstruction. I analyzed 41 models
using a 2-level occupancy model in the package “Unmarked”. My top-ranking model (AIC
weight=0.51) for predicting snowshoe hare occupancy and use included a land cover covariate
(the ratio between forest and open edge) and the transect-level covariates of visual obstruction
60

at 1.0-1.5m above snow level and total stem density; all 3 parameters were positive and only
the transect-level covariates were significant. My results indicated that historical land cover
changes were not related to current site-level occupancy of snowshoe hares. Rather, the
preponderance of current forest relative to open areas was the most influential land cover
variable on occupancy. At a fine scale, woody stem density and visual obstruction were
important determinants of snowshoe hare habitat use.
Key Words
Forest patch sizes, visual obstruction, land cover, Lepus americanus, occupancy, snowshoe
hare, snow track surveys, Program Unmarked.
3.1. INTRODUCTION
Snowshoe hares (Lepus americanus) depend on dense forest vegetation (i.e., >60%
visual obstruction) for concealment, escape, and thermal cover (Litvaitis et al. 1985, Sievert and
Keith 1985, Hodges 2000, Jacqmain et al. 2007, Berg et al. 2012). For example, hare survival was
lower in areas with sparse horizontal cover in Maine (Litvaitis et al. 1985), and hare populations
were less likely to go extinct in areas with dense vegetation cover in Idaho (Thornton et al.
2013). Orr and Dodds (1982) found that hare habitat use was higher in areas dominated by
conifers, with total canopy cover <60%, and with trees <12m tall in Nova Scotia. Presumably
this vegetation structure provided high amounts of ground cover and abundant palatable food
sources (Orr and Dodds 1982). Recent assessments have identified snowshoe hares as
vulnerable to climate change (Hoving et al. 2013, Mills et al. 2013), and wildlife agencies have
expressed interest in managing habitats to potentially mitigate negative climate impacts.

61

In areas of fragmented snowshoe hare habitat, especially with small patch sizes and
sparse cover, the probability of localized extinction increases primarily due to predators
(Buehler and Keith 1982, Sievert and Keith 1985, Keith and Bloomer 1993, Keith et al. 1993,
Hodges 2000). These fragmented patches can provide adequate cover from predators if they
consist of dense spruce (Picea spp.) or willow-alder (Salix spp. – Alnus spp.) thickets (Wolff
1980). Multi-layered habitat structure that provides both vertical and horizontal cover is
important to snowshoe hares. Berg et al. (2012) found that multi-storied forests dominated by
spruce and fir (Abies spp.) supported higher densities of snowshoe hares compared to evenaged stands. In addition, Ivan et al. (2014) recommended managing for multi-layered mature
spruce-fir and early seral lodgepole pine (Pinus contorta) to benefit snowshoe hares in
Colorado. In some locations, these multi-layered habitat conditions take decades to develop
following timber harvest or fire. For example, Thompson et al. (1989) and Newberry and Simon
(2005) found that snowshoe hares were considerably more abundant in 30-year-old boreal
mixedwood clearcuts compared to clearcuts ≤20 years in Ontario and Labrador, Canada,
respectively, with stands <5 years having the lowest abundance. Collectively these studies
indicate that land cover type, especially those types that contain conifers, and habitat structure
in the form of vertical and horizontal cover are important to hares. Less is known on how land
management history and resultant patterns in land cover types influence hare.
Snowshoe hare habitats on the southern edge of the range distribution in the Great
Lakes region of the United States are generally fragmented in small (relative to more northward
landscapes) patches (Buehler and Keith 1982, Sievert and Keith 1985, Keith et al. 1993). For
example, a coarse, broad‐scale habitat assessment corresponding to the late 1990s and early
62

2000s in the Upper Peninsula of Michigan indicated that hare habitat quality ranged from poor
to marginal with only small, isolated areas of higher quality habitat (Linden et al. 2011). This
model portrayed potential hare densities in Michigan as about half of what is considered high
density in the southern habitats of hare range (Hodges 2000, Linden et al. 2011). The isolated
spatial arrangement of high quality habitats that occur in a matrix of low quality habitats,
coupled with a patchy distribution of hare densities across the landscape suggests that hare
populations in Michigan may have a metapopulation structure. In metapopulations localized
extinction and colonization tend to be affected by patch size and isolation (MacArthur and
Wilson 1967).
In Michigan, the current southern edge of snowshoe hare distribution has apparently
moved northward by ~45km over the last 20 years, presumably due to the negative impacts of
climate change (Burt 2014: Chapter 2). However, Burt (2014: Chapter 2) also documented
localized hare extinctions over the past 60 years throughout hare range in Michigan, indicating
that vegetation and climate are likely interacting to determine site occupancy. Whereas climate
change apparently affects hares over multi-year time frames (Burt 2014: Chapter 2), the time
frames resulting in isolation and fragmentation of hare habitats are less understood. My
objectives were to: 1) quantify the effects of historical land cover changes on localized
snowshoe hare occupancy, and 2) rank the ability of current land cover and within patch
habitat measures to predict localized occupancy. I used a combination of local ecological
knowledge, air photo interpretation, GIS analyses, logistic regression, and Akaike Information
Criteria (AIC). My results help inform land management and conservation decisions for a
climate vulnerable species.
63

3.2. STUDY AREA
The field portion of my study occurred in the Upper and northern Lower Peninsulas of
Michigan on publically owned lands during the winter of 2013 (Fig. 3.1). The spatial extent of
my study corresponds to the known historic distribution of snowshoe hares in Michigan (Fig.
3.1). The study area included 3 broad biogeoclimatic regions: 1) western Upper Peninsula, 2)
eastern Upper Peninsula, and 3) northern Lower Peninsula. Landforms of the western Upper
Peninsula consist of glacial moraines, lake plains, outwash channels and plains, and glacially
scoured bedrock ridges (Albert 1995). Primary vegetation types include northern hardwoods,
aspen (Populus spp.), pine (Pinus spp.), and conifer swamps (Albert 1995). The western Upper
Peninsula experiences the most extreme winters and shortest growing seasons of the three
regions I studied (Albert 1995). Monthly average temperatures ranged from ‐10⁰C to 17⁰C, with
average annual precipitation of 90 cm, including 435 cm of snowfall (NOAA 2012).
Landforms of the eastern Upper Peninsula consist mostly of flat lake plain with areas of
exposed bedrock (Albert 1995). Vegetation types include northern hardwoods, upland conifers
(white pine (P. strobus), red pine (P. resinosa), jack pine (P. banksiana)), hardwood‐conifer
swamps, rich conifer swamps, and northern wet meadows (Albert 1995). Monthly average
temperatures vary from ‐10⁰C to 18⁰C with average annual precipitation of 80 cm, including
194 cm of snowfall (NOAA 2012).
Landforms in the northern Lower Peninsula (NLP) consist of large, sandy outwash plains
and large glacial moraines (Albert 1995). The climate of coastal areas in the NLP is moderated
by Lakes Huron and Michigan resulting in warmer and cooler temperatures in winter and
summer, respectively. Interior areas are subjected to more extreme temperature fluctuations.
64

Forest types include northern hardwoods, aspen, oak (Quercus spp.), pine, and lowland conifer
(Albert 1995). Monthly average temperatures range from ‐8⁰C in January to 20⁰C in July, with
average annual precipitation of 90 cm, including 169 cm of snowfall (NOAA 2012).
3.3. METHODS
3.3.1. Site Selection
I interviewed 45 individuals that had knowledge of historical snowshoe hare locations in
Michigan. Interviewees included Michigan Department of Natural Resources employees, hare
hunters, and tribal members. I identified potential interviewees by contacting various hunt
clubs, attending sportsperson coalition meetings, talking with colleagues, and by individuals
providing contact information for hare hunters they personally knew. Each interviewee was
asked to map locations that to their knowledge were unequivocally occupied by snowshoe
hares sometime in the past. Unequivocal evidence included tracks, harvested hares, or
observed hares. Each interviewee was asked to estimate the year of last confirmed occupancy
within 5 yrs. I also checked museum records (e.g. VertNet or MaNIS) for historical snowshoe
hare collection locations, but opted not to use these records because the locational error was
generally large (i.e. >1km). I compiled potential study sites from the interviews and selected a
subset of locations that encompassed broad spatial (throughout hare range in Michigan; Fig.
3.1) and temporal (1955-2010; Fig. 3.2) domains. The 332 ha study sites were separated by >1.6
km (mean = 10.2 km, SE = 0.7). My interview protocol was reviewed and approved by the
Human Research Protection Program, Institutional Review Board, at Michigan State University
(IRB# x11-805).

65

3.3.2. Field Sampling
Winter track counts along transects are commonly used to document snowshoe hare
occupancy and habitat use (Brocke 1975, Conroy et al. 1979, Thompson et al. 1989). However,
transect dimensions and configurations are inconsistent among studies. I evaluated the effects
of transect number (2-9), length (10-150m), and spacing (25-125m) on my ability to accurately
portray site occupancy using a known number of telemetered snowshoe hares in a 6 ha
enclosure during the winters of 2012 and 2013 (Burt 2014: Chapter 1). Prior to releasing
telemetered hares, I surveyed the enclosure to ensure that no hares were present. I
demarcated 9 transects, spaced 25 m apart, ranging in total length from 225-250 m (Burt 2014:
Chapter 1). Track surveys were conducted 12-65 hours after fresh snowfall; survey results were
standardized to 12 hours (Burt 2014: Chapter 1). Track locations along transects were mapped
to the nearest 5m and random sub-sampling of transect lengths and spacing was used to
identify the most efficient transect configuration, where efficiency was defined as the highest
probability of correctly denoting site-level occupancy with minimal distance traveled (Burt
2014: Chapter 1). I replicated transect sampling at densities ranging from 0.2 to 1.5 hares/ha.
These densities represented low to high snowshoe hare densities recorded at the southern
edge of hare range (Hodges 2000). Snowshoe hare trapping and radio-collaring were
conducted by the Michigan Department of Natural Resources – Wildlife Division following
internal animal use and care guidelines that were consistent with the American Society of
Mammalogists Animal Care and Use Guidelines (Sikes et al. 2011).
I implemented the most efficient transect methodology from my enclosure study at a
subset of locations selected from the interview process (Fig. 3.1). My selected transect
66

configuration (9 transects, 125m length, 75m spacing) resulted in the correct designation of
occupancy status 95% of the time based on my enclosure study results (Burt 2014: Chapter 1).
My transect configuration was centered on the historic location of hares from the interviews. I
then randomly oriented the set of transects and visually confirmed that transects occurred on
public lands. Sampling for snowshoe hares occurred on a single day and was completed 12-72
hours after fresh snowfall to allow for track accumulation. If a track was detected on ≥1
transect at a site, I designated the site as occupied. Sites without tracks were deemed
unoccupied and presumed to have experienced localized extinction since the time of last
known occupancy. For transect-level analyses, I designated transects as used (a track was
documented) or unused (no tracks documented) by hares.
At all sites, I measured woody stem density and species composition (i.e., conifer or
hardwood) as important habitat elements for hare (Litvaitis et al. 1985). I conducted counts of
stem densities for each site along a 6 x 2m transect at the first hare track location or at a
randomly generated location along each transect if no hares were present. In addition to direct
measurements of stem density, I also calculated equivalent stem density (3x coniferous stem
count + 1x deciduous stem count; Litvaitis et al. 1985). I quantified visual obstruction using a
modified Robel pole at 3 different height classes above the snow (Robel et al. 1970). The Robel
pole was placed at the center of the sample location and subsequently viewed from 4m away
from 2 opposing directions along the transect. I tallied the number of 10cm demarcations;
represented by alternating colors along the pole, visible at 0.0-0.5m, 0.51-1.0m, and 1.01-1.5m
above the snow. I also tallied the number of snowshoe hare predators encountered along each
transect by families (canids, felids, and mustelids) as an index to potential predation pressure.
67

A priori, I identified 5 site-level and 7 transect-level variables potentially related to
snowshoe hare occupancy (Table 3.1). To generate the site level covariates, I buffered the 9
transects at each site by 812m, resulting in ~332ha around the transects. I selected 812m based
on the reported dispersal distance of snowshoe hare (Gillis and Krebs 1999). I used National
Agricultural Imagery Program (NAIP) infrared aerial imagery to map patches of 6 different land
cover types within the buffer; 1) coniferous, 2) deciduous, 3) mixed coniferous and deciduous,
4) water, 5) urban, and 6) open. These land cover types were consistent with Michigan
Department of Natural Resources mapping guidelines (Michigan Department of Natural
Resources 2009). I calculated average patch sizes for the 3 forested land cover types and the
proportion of total forest in the buffer, where the forest designation was based on a visible tree
canopy regardless of canopy height. The 3 non-forested vegetation types were grouped
together as a single “open” category. Lastly, I used program Fragstats v. 4.2 (McGarigal et al.
2012) to calculate the forest to open edge ratio. An edge was mapped as a boundary between
land cover type stands; this could be a transitional zone between two separate cover types or
when there was a noticeable difference between the same cover type, such as differing age
classes of stands. The same process was used with the open category. I hypothesized that this
ratio declined over time as openings from timber management and land conversion became
more prevalent.
I calculated all land cover metrics from aerial photos that most closely corresponded to
the year of last known snowshoe hare occupancy (historical) and from 2012 photos (current).
Historical photos included 2005 and 2010 from the Remote Sensing and GIS Laboratory at
Michigan State University (http://www.rsgis.msu.edu/) and 1977, 1987, and 1997 from the
68

Michigan Department of Natural Resources aerial photography library. I hypothesized that
forested patch size decreased over time, consistent with a trend in smaller-sized clearcuts and
greater forest fragmentation that potentially results in localized snowshoe hare extinction. I
further hypothesized that forest type changed to a more deciduous dominated landscape in
some locations as a result of past land management and climate effects. Current photos were
from 2012 NAIP. I evaluated 2 scenarios for site-level effects on snowshoe hare occupancy: 1)
the net change in land cover variables over time, and 2) current land cover.
3.3.3. Data Analysis
I analyzed possible combinations of site and transect-level covariates for collinearity
using Pearson’s correlation coefficient (Sokal and Rohlf 2011); covariates were considered
correlated if r ≥ |0.15| and p ≤ 0.05. For land use change over time (Table 3.1), I modeled the
likelihood that a localized (approximately 7.5 ha) snowshoe hare population went extinct from
year of last known occupancy using logistic regression. I created a candidate set of models
(n=17; Table 3.2) from all possible combinations (additive terms) of uncorrelated site variables.
Models were ranked based on Akaike Information Criterion (AIC) and parameter estimates
were deemed significant if the 95% confidence intervals did not overlap 0 (Burnham and
Anderson 2002, Nakagawa and Cuthill 2007).
For current land cover, I modeled the likelihood of snowshoe hare occupancy at a site
and use at the transect level using a 2-level occupancy model (MacKenzie et al. 2002), where
the first level corresponded to site occupancy probability and the second level corresponded to
transect level habitat use. I assumed minimal detection error, i.e., if a transect was occupied my
ability to detect the track was high. I implemented the model in the R package “unmarked” (R
69

Core Team 2014, Fiske et al. 2014). The 2 levels in my model were parameterized with 6 site
level variables and 7 transect level variables (Table 3.1). I created a candidate set of models
(n=41; Table 3.5) from possible combinations (additive terms) of uncorrelated site and transect
variables. Potential models included 3 scenarios: 1) site level variables only, with transect set as
intercept-only, 2) transect level variables only, with site set as intercept-only, and 3) site level
and transect level variables combined. Models were ranked based on Akaike Information
Criterion (AIC) and parameter estimates were deemed significant if the 95% confidence
intervals did not overlap 0 (Burnham and Anderson 2002, Nakagawa and Cuthill 2007).
3.4. RESULTS
My interviews resulted in 386 potential study sites with year of last known hare
occupancy ranging from 1955 to 2010; land cover and vegetation data were compiled for 117 of
those sites (Fig. 3.1). I documented a range of minimum and maximum values for land cover
change, current land cover conditions, and transect-level covariates (Table 3.1). Relative to the
size of my assessment area (332ha), change in forest proportion over time exhibited the
broadest range of impacts (loss of 14% to gain of 64%) but the average change was low (4%
gain) with low variability (SE=1%). The historical changes in patch sizes only varied from -12.6 to
7.8ha, influencing <4% of my assessment area at each site (Table 3.1). On average, I observed a
reduction in patch sizes for all forest cover types I evaluated, but that reduction was small
(<1.7ha). On average, forest to open edge ratio reflected a gain in forest that was generally
consistent across sites (mean=0.40, SE=0.16). My historical land cover change analysis
collectively indicated that Michigan forest patch sizes in snowshoe hare range have declined
slightly over the last ~35 years and that the amount of forest has slightly increased.
70

Current land cover indicated that all sites had ≥25% forest (Table 3.1). Most sites had
high forest amounts (mean=84%, SE=2%; Table 3.1). The deciduous patch sizes tended to be
higher than coniferous and mixed forest patch sizes with deciduous forests occurring at all sites
(Table 3.1). My sites usually contained more forest edge compared to open edge, with the
average being nearly 3 times more forest edge (Table 3.1). My results indicate that deciduous
forests are a dominant component of the sites I surveyed.
At the transect-level, conifer stem counts (averages and maximums) were typically
lower than deciduous stem counts by a factor of 4-5 (Table 3.1). On average the equivalent
stem density was higher than the deciduous stem counts, suggesting that transects generally
contained a mixture of coniferous and deciduous stems (Table 3.1). Predators were typically
absent from transects (mean families detected=0.22), but all 3 predator families were present
at some locations. Visual obstruction averages were low, but some sites exhibited complete
obstruction at all 3 height strata (Table 3.1).
Statewide, I found that 45 of 117 sites (38%) became unoccupied by snowshoe hares
over the time span of my study (1955-2013). In the northern Lower Peninsula 32 of 66 sites
(~48%) were unoccupied during 2013. Only 13 out of 51 (~25%) sites were unoccupied in the
Upper Peninsula. My results indicate almost a two-fold decrease in localized occupancy from
south to north in Michigan, with the greatest concentration of unoccupied sites at the southern
edge of the range (Fig. 3.1).
3.4.1. Land Cover Change Over Time
All candidate models for describing localized occupancy of snowshoe hares based on
changes in land cover over time received little support (i.e., ∆AIC ≤ 4.0 and ωi ≤ 0.12), with no
71

clear top-ranking model (Table 3.2). Additionally, all candidate models received some minimal
level of support (i.e., ωi, ≥ 0.02). No significant parameters were identified in any of the
candidate models. My results indicate that historic changes in land cover at a ~332ha scale did
not affect changes in localized snowshoe hare occupancy.
3.4.2. Current Land Cover and Habitat Structure
Of the 41 candidate models for describing current snowshoe hare occupancy, 6 received
some support (i.e., ωi, ≥ 0.01), but a clear top-ranking model emerged that included a
combination of transect and site-level covariates (Table 3.3). The top 4 models included both
site and transect level covariates, with the 5th and 6th models containing only transect level
covariates (Table 3.3). My top-ranking model accounted for 51% weight of evidence and
consisted of the site level covariate forest to open edge ratio (β = 0.33, 95% CI = -0.006 – 0.669)
and the transect level covariates of visual obstruction at the 1.0-1.5m height strata (β =0.25,
95% CI = 0.18 – 0.31) and equivalent stem density (β =0.019, 95% CI = 0.008 – 0.030; Table 3.3).
My results indicate that in the absence of edge (e.g., 100% coniferous forest) site level
occupancy for snowshoe hares was >0.40 (Fig. 3.3A). However, forest edge (e.g., a conifer and
deciduous mix) increased occupancy probability to near 1 indicating that diverse forest
conditions enhance site-level suitability for hares (Fig. 3.3A).
Visual obstruction between 1.0-1.5m above the snow had the greatest effect on
snowshoe hare transect use, and this effect was almost linear (Fig. 3.3B). Without visual
obstruction the likelihood of transect level habitat use was approximately 0.20 (Fig. 3.3B),
indicating that visual obstruction is not a necessity for hare habitat use. High visual obstruction
alone can result in transect-level habitat use >0.80 (Fig. 3.3B). The effect of equivalent stem
72

density on transect use was positive but less pronounced than the observed relationship for
visual obstruction (Fig. 3.3C). My results indicate that in a forested patch, snowshoe hares will
use areas without coniferous or deciduous stems (~0.30 use probability), presumably if
sufficient horizontal cover is provided, such as brush piles or windfalls.
The second ranked model (∆AIC = 1.9) accounted for 19% weight of evidence and was
based on the same variables as the top-ranking model but also included number of predator
families (Table 3.3). Similar to the top-ranking model, the forest to open edge ratio was not
significant (β = 0.33, 95% CI = -0.007 – 0.670) but visual obstruction (β = 0.24, 95% CI = 0.18 –
0.31) and equivalent stem density (β = 0.019, 95% CI = 0.008 – 0.030) were significant. The
number of predator families was not significant (β = -0.031, 95% CI = -0.351 – 0.289), but the
sign of the parameter estimate suggested that with increased predator presence, the likelihood
of snowshoe hare occupancy decreased. My results suggest that the primary determinants of
snowshoe hare occupancy and habitat use in Michigan relate to habitat structure and
composition at local scales and not on larger scale land cover variables.
3.5. DISCUSSION
My results indicate that land cover change over time in the area encompassed by
average snowshoe hare dispersal distance was having no effect on current occupancy status for
hares in Michigan. Given this finding, I contend that in forest-dominated landscapes, current
land cover and habitat structure are more important determinants of hare occupancy because
this species is relatively short-lived (e.g., 5 years; Kurta 1995), highly mobile (maximum
documented dispersal distances of 20 km; Hodges 2000), and evolved in disturbance-prone
environments (Thompson et al. 1989, Newberry and Simon 2005). Snowshoe hares are an
73

obligate forest species and hence some minimum forest amount across larger landscapes is
likely important; my sites apparently were not below that minimum amount. My study
landscapes were dominated by forests (84% forested on average) and most timber harvest
operations (except large jack pine cuts) were relatively small scale (8-20ha). Coupled with the
relatively large area of my analysis (~332ha), the change in patch sizes and configurations over
time at my sites was generally subtle potentially explaining why historical changes in land cover
was not related to localized hare occupancy. Thornton et al. (2012) also observed that the
relatively small size of forest cutting in Idaho did not cause local hare populations to become
extirpated.
Although not significant, my results suggest that more forest edge (as opposed to open
edge) has a positive influence on snowshoe hare occupancy. I was somewhat surprised that
current land cover metrics were not more important determinants of occupancy. Others have
found land cover metrics to be correlated with hare occurrence. For example, Buehler and Keith
(1982) found that the proportion of forest at a site influenced hare occurrence and mean
number of hare tracks found in Wisconsin. The primary determinant of snowshoe hare
occupancy depends on whether an area is forested with a conifer component that provides
cover close to the ground (Wolff 1980, Buehler and Keith 1982, Berg et al. 2012, Ivan et al.
2014). Snowshoe hares typically avoid openings, but can often be found near forest edges
(Pietz and Tester 1983). While land cover of conifers is important to snowshoe hares, the
vegetation structure (i.e. vegetation density or height) within the patch is critical (Brocke 1975).
The most influential variables on snowshoe hare occupancy in my study were fine-scale
measures of habitat structure and composition. Consistent with other studies (Litvaitis 1985), I
74

found positive relationships between stem density and visual obstruction and small-scale
habitat use by hares. Areas exhibiting highest use contained conifer-dominated or mixed
understories with equivalent stem densities of ~37,066/ha. In some instances, abundant
horizontal cover (typically in the form of trees that were blown down) was observed to
compensate for lower stem counts, suggesting that both habitat elements interact to effect
hare space use. My results that forest types with adequate cover 1.0-1.5m above the snow
support snowshoe hare are consistent with Wolfe et al. (1982) who also found that this
horizontal cover measure accounted for 85% of winter habitat use by hares. Equivalent stem
density and horizontal cover were not correlated in my study, presumably because a small
number of conifers can create abundant visual obstruction (depending on branching
characteristics and snow depth).
Some researchers have suggested that hares in the southern part of the distribution are
subjected to higher mortality rates from predation because a warming climate and increased
anthropogenic activities have allowed predators not adapted to deep snows to expand into
hare range (Gese et al. 2013, Dowd et al. 2014). My simplified assessment of predators lends
support to this observation in Michigan, with the number of predator families detected on a
site being negatively related to hare occupancy. I caution that although my predator covariate
occurred in the second-ranked model (19% weight of evidence), the parameter estimate was
not significant and hence this relationship warrants further evaluation. It is generally accepted
that stem densities and horizontal cover influences predation rates on snowshoe hares (Sievert
and Keith 1985, Keith et al. 1993). With recent attention devoted to the coloration mismatch of
hares to their environment under current climate change scenarios (Mills et al. 2013, Zimova et
75

al. 2014), vegetation management may be one of the few management actions available to
reduce predation rates. In addition to protection from predators, dense vegetation also
provides thermal cover, allowing hares to thermoregulate with less movement (Brocke 1975,
Sievert and Keith 1985, Keith et al. 1993).
I suspect that my observations of localized changes in occupancy status are likely an
interaction between a changing climate (Burt 2014: Chapter 2), habitat structure, and
predators. The geographical range of snowshoe hares is moving northward, likely related to
direct and indirect effects of climate change, with predators identified as a likely proximate
cause of localized extirpations (Sievert and Keith 1985, Hodges 2000). The relationships among
climate, patch configuration, habitat structure, and predators are complex, with managers
having direct influence over patch configuration, habitat structure, and predator abundance. I
contend that large patch sizes with suitable habitat structure can help ameliorate negative
predator effects on snowshoe hares in the southern part of their range. Future research should
focus on whether modifying vegetation can help alleviate the negative effects on snowshoe
hare occupancy apparently linked to climate change (Burt 2014: Chapter 2) and predation.
3.6. MANAGEMENT IMPLICATIONS
My results stress the importance of habitat structure on snowshoe hare occupancy. To
achieve >90% probability of snowshoe hare occupancy for 3 stand categories (conifer
dominated, deciduous dominated (including shrubs like Rubus spp.), and an equal mix of
conifer and deciduous; Table 3.4), managers can focus on equivalent stem density and visual
obstruction. In the Lake States region, visual obstruction should be provided at 3m height
above the ground to account for varying snow conditions. This recommendation is consistent
76

with Brocke (1975) who recommended dense vegetation at 3.5m height. Visual obstruction
does not have to be provided by living vegetation, other alternatives include jack-strawing or
hinge-cutting trees. Brush piles may also be used, but these alone may not provide enough
effective refugia for survival compared to sites without brush piles. (Cox et al. 1997).
Land managers most frequently implement prescriptions based on tree stem densities.
Based on my data, I related equivalent stem density to horizontal cover as a means to provide
general guidance to land managers on when forests should be supplemented with horizontal
cover (Table 3.4). For example, in a dense conifer-dominated patch (>37,000 stems/ha) visual
obstruction is high (9-10 demarcation on the Robel pole are covered) and thus no additional
horizontal cover is needed (Table 3.4). Conversely, in dense deciduous dominated patches
(59,305 stems/ha), 6 Robel demarcations are visible and managers should consider
supplementing the patch with horizontal cover (Table 3.4). I caution that my deciduous
densities include trees and shrubs and thus managers should have some knowledge of
understory deciduous shrub abundance to effectively use Table 3.4.
In the Lake States region, I contend that the stem density table could be used in
coordination with management of other species (e.g., Kirtland’s warbler (Setophaga kirtlandii)
or ruffed grouse (Bonasa umbellus)) to potentially improve those habitats for snowshoe hare.
For example, Cade and Sousa (1985) found a suitability index of 1 at an equivalent stem density
of ~4,900 – 16,000 stems per hectare for ruffed grouse habitat. After factoring in shrubs being
valued at 0.25 times a tree stem, this range falls within my recommended stem densities (Table
3.4). The Michigan Department of Natural Resources et al. (2014) recommends a jack-pine (i.e.
conifer) density of at least 3,588 trees per hectare for Kirtland’s warbler management, which is
77

~62% of our lowest conifer estimate, but does not count shrub stems. Moreover, horizontal
obstruction recommended for snowshoe hare could provide drumming logs for ruffed grouse,
which benefit from at least 3 logs per hectare for a potential drumming site (Sargent and Carter
1999).
3.7. ACKNOWLEDGEMENTS
I thank P. Nelson for serving as a field technician during the winter of 2013. I appreciate
the numerous interviewees who graciously made themselves and their knowledge available for
identifying historic hare sites. Also, thanks to T. Minzey for coordinating my attendance at
sportsperson coalition meetings and S. Beyer for helping me manage the social and political
aspects of the project. Funding for this project was provided by the Michigan Department of
Natural Resources – Wildlife Division through the Michigan Federal Aid in Wildlife Restoration
program grant F13AF01268 in cooperation with the U.S. Fish and Wildlife Service, Wildlife and
Sport Fish Restoration Program and Safari Club International Michigan Involvement Committee.

78

APPENDIX

79

Table 3.1. Sample level, means (SE), and ranges of land cover and habitat covariates used for estimating localized snowshoe hare
occupancy (site level only) or habitat use (transect level) in Michigan.
Change from Historical to Current
Current (2012)
Variable

Levela

Mean

SE

Range

Mean

SE

Range

Forest Proportion
Average Deciduous Patch Size (ha)
Average Coniferous Patch Size (ha)
Average Mixed Forest Patch Size (ha)
Forest to Open Edge Ratio
Conifer Stem Density (stems/12m2)
Deciduous Stem Density (stems/12m2)
Total Predator Genera
Visual Obstruction 0.0-0.5mc
Visual Obstruction 0.5-1.0mc
Visual Obstruction 1.0-1.5mc
Equivalent Stem Density (stems/12m2)

Site
Site
Site
Site
Site
Transect
Transect
Transect
Transect
Transect
Transect
Transect

0.04
-1.71
-0.81
-1.31
0.40
.b
.
.
.
.
.
.

0.01
0.55
0.42
0.34
0.16
.
.
.
.
.
.
.

-0.14 – 0.64
-37.64 – 7.89
-27.02 – 5.28
-12.68 – 6.69
-2.67 – 8.80
.
.
.
.
.
.
.

0.84
9.98
7.12
5.88
2.92
2.36
12.32
0.22
2.20
1.55
1.37
19.41

0.02
0.73
0.55
0.64
0.20
0.12
0.59
0.02
0.09
0.08
0.08
0.63

0.25 – 0.99
0.23 – 72.77
0 – 30.48
0 – 42.55
0.73 – 15.13
0 – 32
0 – 206
0–3
0 – 10
0 – 10
0 – 10
0 - 206

a

Site = ~332ha area surrounding a set of 9 transects; Transect = 125m long.
Historical data on transect level covariates unavailable.
c
Values range from 0-10 with 0 being no visual obstruction 10 being complete visual obstruction.
b

80

Table 3.2. Candidate model set for estimating the likelihood of a localized site being occupied by snowshoe hare in Michigan. ∆AIC =
difference in AIC value from top-ranking model, k = model parameters, ωi – Akaike weight of evidence.
Candidate Model
∆AIC
K
ωi
Forest Proportion + Deciduous Patch Size
Deciduous Patch Size
Forest Proportion
Deciduous Patch Size + Mixed Patch Size
Null Model
Deciduous Patch Size + Forest to Open Edge Ratio
Forest:Open Edge Ratio
Mixed Patch Size
Mixed Patch Size + Forest to Open Edge Ratio
Deciduous Patch Size + Coniferous Patch Size
Deciduous Patch Size + Coniferous Patch Size + Mixed Patch Size
Coniferous Patch Size
Deciduous Patch Size + Coniferous Patch Size + Mixed Patch Size + Forest to Open Edge Ratio
Deciduous Patch Size + Coniferous Patch Size + Forest to Open Edge Ratio
Coniferous Patch Size + Forest to Open Edge Ratio
Coniferous Patch Size + Mixed Patch Size
Coniferous Patch Size + Mixed Patch Size + Forest to Open Edge Ratio

81

0.0
0.1
0.2
0.3
0.5
0.9
1.3
1.4
1.9
2.0
2.3
2.5
2.7
2.9
3.3
3.4
3.9

3
2
2
3
1
3
2
2
3
3
4
2
5
4
3
3
4

0.118
0.114
0.105
0.099
0.093
0.075
0.062
0.058
0.045
0.042
0.037
0.034
0.030
0.028
0.023
0.022
0.016

Table 3.3. Top 6 ranking models with variables from site and transect levels from the 41 candidate models (Table 3.5) for estimating
the likelihood of a site being currently occupied by snowshoe hare in Michigan. ∆AIC = difference in AIC value from top-ranking
model, k = model parameters, ωi – Akaike weight of evidence.
Candidate Modela
∆AIC
K
ωi
Site: Forest to Open Edge Ratio
Transect: Visual Obstruction 1.0-1.5m + Equivalent Stem Density

0.0

5

0.51

Site: Forest to Open Edge Ratio
Transect: Visual Obstruction 1.0-1.5m + Equivalent Stem Density + Predators

1.96

6

0.19

Site: Forest Proportion
Transect: Visual Obstruction 1.0-1.5m + Equivalent Stem Density

2.38

5

0.16

Site: Forest Proportion
Transect: Visual Obstruction 1.0-1.5m + Equivalent Stem Density + Predators

4.35

6

0.058

Transect: Visual Obstruction 1.0-1.5m + Equivalent Stem Density

4.43

4

0.056

Transect: Visual Obstruction 1.0-1.5m + Equivalent Stem Density + Predators

6.40

5

0.021

a

Site = ~332ha area surrounding a set of 9 transects; Transect = 125m long.

82

Table 3.4. The relationship between visual obstruction and stem densities in 3 forest types. Compiled from stem densities and visual
obstruction measured at 117 sites in the northern Lower and Upper Peninsulas of Michigan, winter of 2013.
Visual Obstructiona
Conifer Dominated
Mixed Stand
Deciduous Dominated
(Stems per Hectare) b
(Stems per Hectare) b
(Stems per Hectare)b
10
5,765
11,530
16,472
9
9,059
17,297
27,182
8
13,178
24,711
37,889
7
16,472
32,124
48,596
6
19,768
39,537
59,305
5
23,885
45,302
70,012
4
27,182
52,715
80,719
3
30,475
60,128
91,429
2
34,595
67,541
102,136
1
37,889
74,954
112,843
0
41,183
81,545
127,727
a
b

The number of Robel pole demarcations (10cm) that are visible.
Includes shrub species.

83

Table 3.5. List of all 41 candidate models used in current occupancy modeling for estimating the likelihood of a site being currently
occupied by snowshoe hare in Michigan.
Model Ranka
Site-level Covariates
Transect-level Covariates
1
Forest to open edge ratio
Visual obstruction 1.0-1.5m; Equivalent stem density
2

Forest to open edge ratio

Visual obstruction 1.0-1.5m; Equivalent stem density; Total predator genera

3

Forest proportion

Visual obstruction 1.0-1.5m; Equivalent stem density

4

Forest proportion

Visual obstruction 1.0-1.5m; Equivalent stem density; Total predator genera

5

.b

Visual obstruction 1.0-1.5m; Equivalent stem density

6

.b

Visual obstruction 1.0-1.5m; Equivalent stem density; Total predator genera

7

.b

Visual obstruction 1.0-1.5m

8

.b

Visual obstruction 1.0-1.5m; Total predator genera

9

.b

Visual obstruction 0.5-1.0m

10

.b

Deciduous stem count; Visual obstruction 0.5-1.0m

11

.b

Visual obstruction 0.5-1.0m; Total predator genera

12

.b

Deciduous stem count; Visual obstruction 0.5-1.0m; Total predator genera

13

.b

Visual obstruction 0.0-0.5m

14

.b

Visual obstruction 0.0-0.5m; Total predator genera

84

Table 3.5. (cont’d)
15

.b

Conifer stem count

16

.b

Conifer stem count; Total predator genera

17

.b

Equivalent stem density

18

.b

Equivalent stem density; Total predator genera

19

Forest to open edge ratio

.c

20

Forest proportion

.c

21

Average conifer patch size;
Average mixed forest patch size

.c

22

Average mixed forest patch size

.c

23

.d

.d

24

Forest proportion; Average
mixed forest patch size

.c

25

Forest to open edge ratio;
Average mixed forest patch size

.c

26

.b

Total predator genera

27

Average conifer patch size

.c

28

.b

Deciduous stem count
85

Table 3.5. (cont’d)
29
Average conifer patch size;
Average mixed forest patch size

Visual obstruction 1.0-1.5m; Equivalent stem density

30

.b

Deciduous stem count; Total predator genera

31

Average mixed forest patch size

Visual obstruction 1.0-1.5m; Equivalent stem density

32

Average conifer patch size;
Average mixed forest patch size

Visual obstruction 1.0-1.5m; Equivalent stem density; Total predator genera

33

Average mixed forest patch size

Visual obstruction 1.0-1.5m; Equivalent stem density; Total predator genera

34

Average deciduous patch size;
Average mixed forest patch size

.c

35

Forest to open edge ratio;
Average deciduous patch size;
Average mixed forest patch size

.c

36

Forest proportion; Average
deciduous patch size; Average
mixed forest patch size

.c

37

Average deciduous patch size

.c

38

Average deciduous patch size;
Average conifer patch size;
Average mixed forest patch size

.c

86

Table 3.5. (cont’d)
39
Average deciduous patch size;
Average conifer patch size

.c

40

Forest to open edge ratio;
Average deciduous patch size

.c

41

Forest proportion; Average
deciduous patch size

.c

a

Based on AIC.
Only transect-level covariates.
c
Only site-level covariates.
d
Null model.
b

87

Figure 3.1. Location and occupancy status of snowshoe hare survey sites throughout the
northern Lower and Upper Peninsulas of Michigan, USA, winter 2013. The geographic range of
snowshoe hares and Michigan are portrayed in the map inset as adapted from the International
Union for Conservation of Nature (www.iucnredlist.org/technical-documents/spatialdata#mammals).

88

Figure 3.2. Number of study sites by 5-year groupings (through 2010) of last confirmed
snowshoe hare occupancy in the northern Lower and Upper Peninsulas of Michigan, USA.

89

Figure 3.3. Snowshoe hare site occupancy probability and (A) forest to open edge ratio, and
transect level probability of use by (B) visual obstruction 1.0-1.5m above snow level, and (c)
equivalent stem density in Michigan, USA and the 95% confidence intervals. The points along
the x-axes represent the values of each study site.

90

LITERATURE CITED

91

LITERATURE CITED

Albert, D.A. 1995. Regional landscape ecosystems of Michigan, Minnesota, and Wisconsin: a
working map and classification. United States Department of Agriculture, Forest Service
General Technical Report NC-178, North Central Forest Experiment Station, St. Paul,
Minnesota, USA.
Berg, N. D., E. M. Gese, J. R. Squires, and L. M. Aubry. 2012. Influence of forest structure on the
abundance of snowshoe hares in western Wyoming. Journal of Wildlife Management
76:1480-1488.
Brocke, R.H. 1975. Preliminary guidelines for managing snowshoe hare habitat in the
Adirondacks. Transactions of the Northeast Fish and Wildlife Conference 32:46‐66.
Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference. Springer,
New York, NY, USA.
Buehler, D. A., and L. B.Keith. 1982. Snowshoe hare distribution and habitat use in Wisconsin.
Canadian Field-Naturalist 96:19-29.
Burt, D. M. 2014. Factors influencing snowshoe hare occupancy in Michigan. Thesis, Michigan
State University, Michigan, USA.
Cade, B. S., and P. J. Sousa. 1985. Habitat suitability index models: ruffed grouse. Fish and
Wildlife Service Biological Report 82, Washington, DC, USA.
Conroy, M. J., L. W. Gysel, and G. R. Dudderar. 1979. Habitat components of clear‐cut areas for
snowshoe hares in Michigan. Journal of Wildlife Management 43:680‐690.
Cox, E. W., R. A. Garrott, and J. R. Cary. 1997. Effect of supplemental cover on survival of
snowshoe hares and cottontail rabbits in patchy habitat. Canadian Journal of Zoology
75:1357-1363.
Dowd, J. L. B., E. M. Gese, and L. M. Aubry. 2014. Winter space use of coyotes in high-elevation
environments: behavioral adaptations to deep-snow landscapes. Journal of Ethology
32:29-41.
Fiske, I., R. Chandler, D. Miller, A. Royle, and M. Kery. 2014. Package ‘unmarked:’ Models for
data from unmarked animals. Version 0.10-4. URL http://cran.at.rproject.org/web/packages/unmarked/unmarked.pdf

92

Gese, E. M., J. L. B. Dowd, and L. M. Aubry. 2013. The influence of snowmobile trails on coyote
movements during winter in high-elevation landscapes. PLoS ONE 8:e82862.
doi:10.1371/journal.pone.0082862
Gillis, E. A., and C. J. Krebs. 1999. Natal dispersal of snowshoe hares during a cyclic population
increase. Journal of Mammalogy 803:933-939.
Hodges, K. E. 2000. Ecology of snowshoe hares in southern boreal and montane forests. Pages
163-207 in L.F. Ruggiero, K.B. Aubry, S.W. Buskirk, G.M. Koehler, C.J. Krebs, K.S.
McKelvey and J.R. Squires, editors. Ecology and Conservation of Lynx in the United
States. University Press of Colorado, Boulder, CO, USA.
Hoving, C. L., Y. M. Lee, P. J. Badra, and B. J. Klatt. 2013. Changing climate, changing wildlife: a
vulnerability assessment of 400 species of greatest conservation need and game species
in Michigan. Michigan Department of Natural Resources Wildlife Division Report
Number 3564, Lansing, MI, USA.
Ivan, J. S., G. C. White, and T. M. Shenk. 2014. Density and demography of snowshoe hares in
central Colorado. Journal of Wildlife Management 78:580-594.
Jacqmain, H., L. Bélanger, S. Hilton, and L. Bouthillier. 2007. Bridging native and scientific
observations of snowshoe hare habitat restoration after clearcutting to set wildlife
habitat management guidelines on Waswanipi Cree land. Canadian Journal of Forest
Research 37:530-539.
Keith L. B., and S. E. M. Bloomer. 1993. Differential mortality of sympatric snowshoe hares and
cottontail rabbits in central Wisconsin. Canadian Journal of Zoology 71:1694-1697.
Keith L. B., S. E. M. Bloomer, and T. Willebrand. 1993. Dynamics of a snowshoe hare population
in fragmented habitat. Canadian Journal of Zoology 71:1385-1392.
Kurta, A. 1995. Mammals of the Great Lakes Region. University of Michigan Press. Ann Arbor,
USA.
Linden, D. W., H. Campa, III, G. J. Roloff, D. E. Beyer, Jr., and K. F. Millenbah. 2011. Modeling
habitat potential for Canada lynx in Michigan. Wildlife Society Bulletin 35:20-26.
Litvaitis, J. A., J. A. Sherburne, and J. A. Bisonette. 1985. Influence of understory characteristics
on snowshoe hare habitat use and density. Journal of Wildlife Management 49:866-873.
MacArthur, R. H., and E. O. Wilson. 1967. The theory of island biogeography. Princeton
University Press. Princeton, New Jersey, USA.

93

MacKenzie, D. I., J. D. Nichols, G. B. Lachman, S. Droege, J. Andrew Royle, and C. A. Langtimm.
2002. Estimating site occupancy rates when detection probabilities are less than one.
Ecology 83:2248-2255.
McGarigal, K., S. A. Cushman, and E. Ene. 2012. FRAGSTATS Version 4: Spatial Pattern Analysis
Program for Categorical and Continuous Maps. Computer software program produced
by the authors at the University of Massachusetts, Amherst.
<http://www.umass.edu/landeco/research/fragstats/fragstats.html> Accessed 1
January 2014.
Michigan Department of Natural Resources. 2009. IFMAP Manual. Appendix A.
<https://www.michigan.gov/documents/dnr/1IFMAPManualTableOfContents_363236_7.pdf> Accessed 18 October 2013.
Michigan Department of Natural Resources, US Fish and Wildlife Service, and US Forest Service.
2014. Kirtland’s Warbler Breeding Range Conservation Plan.
<https://www.michigan.gov/documents/dnr/Kirtlands_Warbler_CP_457727_7.pdf>
Accessed 15 October 2014.
Mills, L. S., M. Zimova, J. Oyler, S. Running, J. T. Abatzoglou, and P. M. Lukacs. 2013. Camouflage
mismatch in seasonal coat color due to decreased snow duration. Proceedings of the
National Academy of Science 110:7360-7365.
Nakagawa, S., and I. C. Cuthill. 2007. Effect size, confidence interval and statistical significance:
a practical guide for biologists. Biological Reviews 82:591-605.
National Oceanic and Atmospheric Administration [NOAA]. 2012. NOWData – NOAA online
weather data. <http://www.noaa.gov/>. Accessed 15 February 2012.
Newberry, T. L., and N. P. P. Simon. 2005. The effects of clearcutting on snowshoe hare (Lepus
americanus) relative abundance in central Labrador. Forest Ecology and Management
210:131-142.
Orr, C. D., and D. G. Dodds. 1982. Snowshoe hare habitat preferences in Nova Scotia spruce-fir
forests. Wildlife Society Bulletin 10:147-150.
Pietz, P. J., and J. R. Tester. 1983. Habitat selection by snowshoe hares in north central
Minnesota. The Journal of Wildlife Management 47:686-696.
R Core Team. 2014. R: A language and environment for statistical computing. R foundation for
Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.

94

Robel, R. J., J. N. Briggs, A. D. Dayton, and L. C. Hulbert. 1970. Relationships between visual
obstruction measurements and weight of grassland vegetation. Journal of Range
Management 23:295-297.
Sargent, M. S., and K. S. Carter. 1999. Managing Michigan wildlife: a landowner’s guide.
<http://www.michigandnr.com/publications/pdfs/huntingwildlifehabitat/Landowners_
Guide/Species_Mgmt/Ruffed_Grouse.htm> Accessed 15 October 2014.
Sievert, P. R., and L. B. Keith. 1985. Survival of snowshoe hares at a geographical range
boundary. Journal of Wildlife Management 49:854-866.
Sikes, R.S., W. L. Gannon, and The Animal Care and Use Committee of the American Society of
Mammalogists. 2011. Guidelines of the American Society of Mammalogists for the use
of wild mammals in research. Journal of Mammalogy 92:235-253.
Sokal, R. R., and F. J. Rohlf. 2011. Biometry. Fourth edition. Freeman, New York, USA.
Thompson, I. D., I. J. Davidson, S. O’Donnell, and F. Brazeau. 1989. Use of track transects to
measure relative occurrence of small boreal mammals in uncut forest regeneration
stands. Canadian Journal of Zoology 67:1816‐1823.
Thornton, D. H., A. J. Wirsing, J. D. Roth, and D. L. Murray. 2012. Complex effects of site
preparation and harvest on snowshoe hare abundance across a patchy landscape.
Forest Ecology and Management 280:132-139.
Thornton, D. H., A. J. Wirsing, J. D. Roth, and D. L. Murray. 2013. Habitat quality and population
density drive occupancy dynamics of snowshoe hare in variegated landscapes.
Ecography 36: 610-621.
Wolfe, M. L., N. V. Debyle, C. S. Winchell, and T. R. McCabe. 1982. Snowshoe hare cover
relationships in northern Utah. Journal of Wildlife Management 46:662-670.
Wolff, J. O. 1980. The role of habitat patchiness in the population dynamics of snowshoe hares.
Ecological Monographs 50:111-130.
Zimova, M., L. S. Mills, P. M. Lukacs, and M. S. Mitchell. 2014. Snowshoe hares display limited
phenotypic plasticity to mismatch in seasonal camouflage. Proceedings of the Royal
Society B 281:20140029.

95

CONCLUSIONS

My thesis addresses 2 distinct questions: 1) what is the optimal transect configuration
for effectively and efficiently sampling a snowshoe hare population using snow track surveys,
and 2) what factors are driving snowshoe hare localized extinctions, current occupancy, and
current habitat use in Michigan. Chapter 1 of my work is the first to experimentally manipulate
hare densities and determine the accuracy of track surveys. As such, Chapter 1 adds to the
literature on using snow track surveys to monitor snowshoe hare populations (Hodges and Mills
2008). Chapter 2 focused on climate covariates that have been linked to snowshoe hare
demographics. I found that climate correlated with localized extinctions and offered some
potential mechanisms for how climate is negatively affecting hares. In Chapter 3 I concentrated
on how various land cover and habitat factors were associated with localized hare occupancy
over time. The primary contributions of Chapter 3 included an assessment of historical and
current land cover and which is more important for influencing snowshoe hare occupancy and a
ranking of land cover and habitat structure variables to help guide where management should
be focused.
In Chapter 1, my results indicated that snow track surveys, if configured correctly, can
be efficiently implemented and provide accurate data on snowshoe hare occupancy status. I
recommended 3 designs for use in broad-scale snowshoe hare monitoring programs.
Additionally, results from the 3 transect configurations appeared to correlate with snowshoe
hare density across the full range of densities I evaluated. This relationship was not observed

96

over a narrower range of densities (e.g., like those found at the southern distribution). I used
transects that were 125m in length with 75m spacing for Chapters 2 and 3 of my thesis.
In Chapter 2, I found that of 134 sites sampled in Michigan, 39% of the sites experienced
localized snowshoe hare extinction, with higher probabilities in the Lower Peninsula of
Michigan compared to the Upper Peninsula. I found that maximum temperature between May
15 and January 19 and the number of days with snow on the ground were the primary factors
influencing localized extinction. The hypothesized mechanisms of these covariates were a
decrease in litter size and coat color mismatch to the environment, respectively. Study sites
with the highest maximum temperatures and/or fewest days with snow on the ground were
approximately 3 times more likely to experience extinction.
In Chapter 3, I found that out of the 117 sites that I compiled land cover and habitat
data for, 62% of the sites were occupied in 2013. Results suggested that long-term land cover
change over time at a 332ha scale did not affect snowshoe hare occupancy status. I also found
that occupancy status was most influenced by a single site and 2 transect level covariates.
While the 1 site covariate (forest to open edge ratio) appeared in the top-ranked model, it was
not significant, with current snowshoe hare occupancy in Michigan being primarily influenced
by the transect-level covariates of visual obstruction 1.0-1.5m above snow level (~3m) and an
equivalent stem density (3x conifer stems + 1x deciduous stems). These results are consistent
with a growing body of research that show hares mostly depend on dense vegetation for
survival (Litvaitis et al. 1985, Sievert and Keith 1985, Hodges 2000, Berg et al. 2012). I provided
specific recommendations that can help managers succeed in increasing snowshoe hare
occupancy in forests throughout Michigan.
97

The relationships among climate, land cover, vegetation structure, and predators are
complex. Future research is needed to assess if directly managing patch configuration,
vegetation structure, or predator abundance can help mitigate the apparent extinction process
related to climate change. I contend that while the range of snowshoe hares will continuously
contract northward due to climate change, the rate of this contraction could be slowed with
designated habitats being targeted to aid snowshoe hares. Overall, this research provides firm
guidelines for increasing snowshoe hare occupancy probabilities.

98

LITERATURE CITED

99

LITERATURE CITED

Berg, N. D., E. M. Gese, J. R. Squires, and L. M. Aubry. 2012. Influence of forest structure on the
abundance of snowshoe hares in western Wyoming. Journal of Wildlife Management
76:1480-1488.
Hodges, K. E. 2000. Ecology of snowshoe hares in southern boreal and montane forests. Pages
163-207 in L.F. Ruggiero, K.B. Aubry, S.W. Buskirk, G.M. Koehler, C.J. Krebs, K.S.
McKelvey and J.R. Squires editors. Ecology and Conservation of Lynx in the United
States. University Press of Colorado, Boulder, CO, USA.
Hodges, K. E., and L. S. Mills. 2008. Designing fecal pellet surveys for snowshoe hares. Forest
and Ecology Management 256:1918-1926.
Litvaitis, J. A., J. A. Sherburne, and J. A. Bisonette. 1985. Influence of understory characteristics
on snowshoe hare habitat use and density. Journal of Wildlife Management 49:866-873.
Sievert, P. R., and L. B. Keith. 1985. Survival of snowshoe hares at a geographical range
boundary. Journal of Wildlife Management 49:854-866.

100