Aw

In!

15;:

u

.Kur'vald .3.....~I0L«u

4.. .
.. «:2 ..

“nun...“

 

 

In 5?.»

 

 

 

lt'thi):

20*}

This is to certify that the
dissertation entitled

Resource Selection of Recolonizing
Gray Wolves in Northwest Wisconsin

presented by

Paul W. Keenlance

has been accepted towards fulfillment
of the requirements for

PH.D. degreein Fisheries & Wildlife

 

 

QI (4947in MA-d 1:,

 

Major professor

Date November 27, 2002

 

MSUi: an Affirmative Action/Equal Opportunity Institution 0-1277!

 

 

i LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

A015 '1 282093

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/CIRC/DaleDue.p65-p. 15

RESOURCE SELECTION OF RECOLONIZING
GRAY WOLVES IN NORTHWEST WISCONSIN

By

Paul W. Keenlance

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Fisheries and Wildlife

2002

UMI Number: 3075028

®

UlVII

 

UMI Microform 3075028

Copyright 2003 by ProQuest Information and Learning Company.

All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.

 

ProQuest Information and Learning Company
300 North Zeeb Road
PO. Box 1346
Ann Arbor, MI 48106-1346

ABSTRACT
RESOURCE SELECTION OF RECOLONIZING
WOLVES m NORTHWEST WISCONSIN
By

Paul W. Keenlance

The gray wolf (Canis lupus) has expanded numerically and geographically in
eastern Minnesota and northwest Wisconsin since the late 19705. Based on data collected
from 40 wolves in 15 resident packs during 1992-2001, I examined resource selection by
wolves at the following scales: territory selection on the landscape, resource selection
within territories, and den placement. Road density within pack territories increased from
a mean of 0.35 km/km2 in packs established in 1992 to 1.09 km/km2 in packs established
in 1999. Mean road density in pack territories was lower (0.70 km/kmz) than in unused
areas (1.43 km/kmz, p < 0.001). Forested wetland (14%), lowland shrub (13%), and
aspen (Populus spp.) (22%) were all more prevalent (p < 0.10 in all cases) in wolf pack
territories than in unused areas (7%, 6%, and 9% respectively). County owned land
comprised larger proportions (p < 0.001) of pack territories (61%) than unused areas
(15%), while privately owned land were less prevalent (p < 0.001) in territories (19%)
than unused areas (56%). A logistic regression-based habitat model including terms for
road density, proportion of county and national forest land, and proportion of forested
wetland in the area of interest correctly classified 14 of 15 pack territories and 13 of 15
unused areas outside of the area where the data used to develop the model was collected.

A compositional analysis indicated that forested wetland, lowland shrub, and

aspen were the 3 most selected land cover categories at the landscape level of selection

while forested wetland, aspen, and oak were the 3 most selected land cover categories
within territories. The proportion of wolf telemetry locations in buffers of 25 - 250 m
around roads was lower (p < 0.10 in all cases) than the proportion of territories occupied
by these same buffers. Wolf dens fell within an arbitrarily defined territory core
occupying the inner 25% of the territory more often (78%) (p = 0.002) than random sites
(35%). Roads occurred within 1212m of fewer (p < 0.029) dens (50%) than random
points. This distance represented the average distance from a den that an alpha female
was located during the study. Pine (8%) and shrub (6%) comprised higher proportions (p
< 0.10 in both cases) of buffers around dens than random sites (3%, and 4% respectively),
while lowland shrub comprised a lower proportion (p = 0.068) of buffers around dens
(16%) than random sites (10%). A compositional analysis indicated that deciduous
forest, aspen, and lowland shrub were the 3 most selected land cover categories at wolf
den sites. Logistic regression indicated that location within the 25% inner core of a
territory was the measured variable contributing the most to classification accuracy of a
model delineating den sites versus random sites.

AS the wolf population in Wisconsin continues to increase, wolves are
establishing territories in areas with road densities that are both increasing and higher
than past paradigms of wolf habitat suitability would predict. In spite of this, wolves are
still establishing territories in areas with lower road density than unoccupied areas, and at
least during daylight, are avoiding roads within territories. Wolves appear to select land
cover and ownership category based on the avoidance of human disturbance and prey

availability. Wolves also select den sites that tend to be centered in their territories.

ACKNOWLEDGEMENTS

By the time that any of us reaches the end of a doctoral program, we owe a great
deal to a great many individuals and organizations. Though I am sure that I will miss
many whom I should be thanking, I need to express my heartfelt thanks to the following
(in no particular order):

Funding for this project and my involvement in it was provided by the Wisconsin
Department of Natural Resources Bureaus of Integrated Science Services and
Endangered Resources, the Wisconsin Department of Transportation, The Michigan State
University Fisheries and Wildlife Department, Biological Sciences and Integrated Studies
in Biology programs, and the students of the Manawa, Wisconsin School district.

I would like to thank the members of my Doctoral committee for their
enthusiastic contributions to this project, their time and their guidance: Thank you Drs.
Gary Roloff, Michael Jones, Kay Holekamp, and David Lusch. I owe a special debt of
gratitude to my major professor, Dr. Kelly Millenbah. Thank you for your guidance,
your advice, your incredible level of commitment to all your students, the example you
set as an educator and most of all, your patience. I could not have asked for a better
advisor.

I must thank Dr. Eric Anderson and his graduate students, Dr. Tom Gehring,
Doug Shelley, and Jacqui Friar. This project would not have been possible without the
enormous amount of work they expended during their M.S. projects in the early stages of
the H53 study. A special thanks to Dave Unger who contributed not only through his
MS. work on the H53 study, but who served as an invaluable field assistant during the

last 2 ‘/2 summers of my doctoral research. The aid provided by his knowledge of wolf

iv

ecology, the H53 SA, and his skills as a biologist were exceeded only by the value of his
friendship throughout my MS. and my Doctoral degrees. . ..I owe you buddy.

The personnel of the WDNR Gordon Ranger station and Douglas county forestry
departments were unfailingly supportive of this project and aided greatly in its
completion. The pilots of the WDNR Siren and Gordon hangers saved us countless hours
of ground telemetry and kept me in contact with the wolves during the time each year
when I was back in East Lansing. Thank you all very much for your efforts.

Thank you Drs. Daniel Trainer and Dave Staszak for the gentle nudge on a trip to
Australia that got me started on the road to where I am now.

Heartfelt thanks to the ladies of team Millenbah for the humor, the unswerving
support, the insight into the mysterious psyche of “the collective”, and most importantly,
for putting up with me with unwavering good humor. I could not have asked for better
labmates.

Thank you Bruce (B. Eric) Kohn. From start to finish, you were the heart and soul
of the H53 study. I can never put into words the level of gratitude I feel to you or the
amount of respect I hold for you. I think that we all look for mentors in our lives. The
first summer that we worked together I found mine. I have never met anyone who I
respect more as a person and as a professional. Eternal thanks.

Nearly lastly, I must thank my family. They have been there for me in every
segment of my life and though I am sure that they did not always understand why I was
taking as long as I did to finally get out of school, my Doctoral program was no

exception. I could not have done it without your love and support. No matter where I go

or what I do, you are the basis of all things good in my life. Mom, Dad, and Laura, thank
you for who I am. I love you all.

And now, finally, though it may be a bit clichéd, and potentially not even overly
professional, I thank the wolves of the H53SA for allowing the long ago dreams of a little

boy to come true. If only I could have learned all that you had to teach.

vi

TABLE OF CONTENTS

ABSTRACT ...................................................................................................................... ii
ACKNOWLEDGEMENTS .............................................................................................. iv
LIST OF TABLES ............................................................................................................ vi
LIST OF FIGURES ....................................................................................... xiv
CHAPTER 1: INTRODUCTION AND DISSERTATION STRUCTURE ..................... 1
INTRODUCTION ................................................................................................ 1
The Importance of Examining Resource Selection at Differing Scales. . .2
Changes in Resource Selection as Population Density Increases .......... 4
DISSERTATION STRUCTURE .......................................................................... 6
STUDY AREA ..................................................................................................... 8
LITERATURE CITED ......................................................................................... 11

CHAPTER 2: PACK TERRITORY PLACEMENT AND HABITAT
CHARACTERISTICS OF A RECOLONIZING POPULATION OF GRAY WOLVES
IN NORTHERN WISCONSIN

INTRODUCTION ................................................................................................ 14
METHODS ........................................................................................................... l7
Capture, Handling, and Radiotelemetry .................................................... 17
Territory Boundary Estimation ................................................................. 19
Evaluation of Changes in Territory Characteristics With Increasing
Population and Territory Density ............................................................. 20
Development of a Landscape Level Habitat Suitability Model ................ 23
RESULTS ............................................................................................................. 28
Evaluation of Changes in Territory Characteristics With Increasing
Population and Territory Density .............................................................. 28
Differences in Measured Habitat Variables Between Pack Territories
and Random Unused Areas ....................................................................... 32
Landscape Level Habitat Suitability Model ............................................. 32
DISCUSSION ....................................................................................................... 37

vii

Changes in Territory Characteristics With Increasing Population and

Territory Density ....................................................................................... 40

Differences in Measured Habitat Variables Between Pack Territories

and Random Unused Areas ....................................................................... 43

Landscape Level Habitat Suitability Model ............................................. 45
SUMMARY AND MANAGEMENT IMPLICATIONS ..................................... 46
LITERATURE CITED ......................................................................................... 50

CHAPTER 3: COVER TYPE AND LAND OWNERSHIP SELECTION OF
RECOLONIZING GRAY WOLVES IN NORTHWEST WISCONSIN

INTRODUCTION ................................................................................................ 55
METHODS ........................................................................................................... 57
Capture, Handling, and Radiotelemetry .................................................... 57
Territory Boundary Estimation ................................................................. 59
Determination of Territory Land Cover and Land Ownership
Composition .............................................................................................. 58
RESULTS ............................................................................................................. 61
Selection of Land Cover Category ............................................................ 61
Selection of Land Ownership Category .................................................... 67
Use of Buffers around Roads .................................................................... 67
DISCUSSION ....................................................................................................... 67
Selection of Land Cover Category ............................................................ 68
Selection of Land Ownership Category .................................................... 73
Use of Buffers around Roads .................................................................... 74
SUMMARY AND MANAGEMENT IMPLICATIONS ..................................... 76
LITERATURE CITED ......................................................................................... 78

CHAPTER 4: GRAY WOLF DEN SITE CHARACTERISTICS IN NORTHWEST
WISCONSIN: ADDITIONAL DATA

INTRODUCTION ................................................................................................ 89
METHODS ........................................................................................................... 91
Capture, Handling, and Radiotelemetry .................................................... 91
Den Site Location and Delineation ........................................................... 92
Analysis of Variables Influencing Den Site Selection .............................. 93
RESULTS ............................................................................................................. 96
Den site location and delineation .............................................................. 96

viii

Variables Influencing Den Site Selection ............................................... 96

DISCUSSION ..................................................................................................... 99
SUMMARY AND MANAGEMENT IMPLICATIONS ................................... 102
LITERATURE CITED ....................................................................................... 105

CHAPTER 5: SYNTHESIS AND FUTURE RESEARCH DIRECTIONS

SYNTHESIS ....................................................................................................... 111
FUTURE RESEARCH DIRECTION ................................................................. 1 14
LITEREATURE CITED ..................................................................................... 116

ix

LIST OF TABLES

CHAPTER 2: PACK TERRITORY PLACEMENT AND HABITAT
CHARACTERISTICS OF A RECOLONIZING POPULATION OF GRAY WOLVES
IN NORTHERN WISCONSIN

Table 2.1 Mean road density within wolf pack territories established in different
years calculated from coverages using roads and trails and roads only in
the Highway 53 Study Area in northwest WI 1992-2001 ............................... 30

Table 2.2 Land ownership category composition of gray wolf pack territories
grouped by year of territory establishment in the Highway 53 Study Area
in northwest Wisconsin 1992-2001. ................................................................ 33

Table 2.3 Land cover category composition of gray wolf pack territories grouped
by year of territory establishment in the Highway 53 Study Area in
northwest Wisconsin 1992-2001. ................................................................... 34

Table 2.4 Means and standard errors of habitat variables in wolf pack territories
and corresponding random unused areas in the Highway 53 Study Area
in northwest Wisconsin 1992-2001. ............................................................... 35

CHAPTER 3: COVER TYPE AND LAND OWNERSHIP SELECTION OF
RECOLONIZING GRAY WOLVES IN NORTHWEST WISCONSIN

Table 3.1. Mean percent land cover category composition in the Highway 53 Study
Area and within wolf pack territories and the percentage of telemetry
locations within the 14 cover types used in a compositional analysis of
land cover type selection in northwest Wisconsin 1995-2001 ........................ 63

Table 3.2. Consolidated thematic mapper based WISCLAND land cover categories
used in the compositional analysis of landscape level and within-territory
selection and corresponding selection ranks for both levels of analysis
based on year-long use data for gray wolves in the Highway 53 Study
Area in northwest Wisconsin 1995-2001 ........................................................ 64

Table 3.3. Compositional analysis selection rankings of consolidated WISCLAND
based land cover types for year-long and seasonal periods at the within-
territory level of analysis for gray wolves in the Highway 53 Study Area
in northwest Wisconsin 1995-2001. ............................................................... 66

Table 3.4. Compositional analysis selection rankings of consolidated WISCLAND
based land cover types for year-long and seasonal periods at the within-
territory level of analysis for gray wolves in the Highway 53 Study Area
in northwest Wisconsin 1995-2001. ............................................................... 67

Table 3.5. Percent of the Highway 53 Study Area and mean percent of territories
within buffers of varying distances around roads and trails compared to
percent of telemetry locations within these same buffers in northwest
Wisconsin 1995-2001. .................................................................................... 69

Appendix A. Original and consolidated land cover categories within the Highway
53 Study Area in northwest Wisconsin ...................................................... 84

Appendix B. Compositional analysis based pair-wise comparisons of landscape level
land cover category selection by gray wolves in the Highway 53
Study Area of northwest Wisconsin 1995-2001 ........................................ 85

Appendix C. Compositional analysis based pair-wise comparisons of
within-territory land cover selection by gray wolves in the Highway
53 Study Area of northwest Wisconsin 1995-2001 ................................... 86

Appendix D. Compositional analysis based pair-wise comparisons of
within-territory land cover selection by gray wolves in the Highway
53 Study Area of northwest Wisconsin during “summer” seasonal
periods (15 April-14 September) 1995-2001 ............................................ 87

Appendix E. Compositional analysis based pair-wise comparisons of
within-territory land cover selection by gray wolves in the Highway
53 Study Area of northwest Wisconsin during “winter” seasonal
periods (15 September — 14 April) 1995-2001 .......................................... 88

CHAPTER 4: GRAY WOLF DEN SITE CHARACTERISTICS IN NORTHWEST
WISCONSIN: ADDITIONAL DATA

Table 4.1 Land cover category composition in 1212m radius buffers around gray
wolf dens and corresponding random sites in the H53 SA in northwest
Wisconsin 1993-2000. ..................................................................................... 97

Table 4.2 Compositional analysis rankings of land cover category selection at gray
wolf den sites within the Highway 53 Study Area in northwest Wisconsin
1993-2000 ........................................................................................................ 100

Appendix A. Original and consolidated land cover categories used for compositional
analysis in 1212m buffers around den and random points within the
Highway 53 Study Area in northwest Wisconsin ...................................... 109

xi

Appendix B. Compositional analysis rankings of land cover category selection at gray
wolf den sites within the Highway 53 Study Area in northwest Wisconsin
1993-2000 ..................................................................................................... 110

xii

LIST OF FIGURES
CHAPTER 1: INTRODUCTION AND DISSERTATION STRUCTURE

Fig. l. The Highway 53 Study Area (H53SA) in northwest Wisconsin and east-
central Minnesota .............................................................................................. 9

CHAPTER 2: PACK TERRITORY PLACEMENT AND HABITAT
CHARACTERISTICS OF A RECOLONIZING POPULATION OF GRAY
WOLVES IN NORTHERN WISCONSIN

Fig. 2.1 Location of the Highway 53 Study Area (H53 SA) within
primary wolf range (PWR) in Wisconsin ........................................................... 18

Fig. 2.2 Year of establishment of wolf packs within the Highway 53 Study Area
in northwest Wisconsin 1992-2001. Color indicates year of initiation
and pattern indicates individual pack territories ................................................. 29

Fig. 2.3. Differences in road density within wolf pack territories established in
different years in the Highway 53 Study Area in northwest Wisconsin
1992-2001. Mean road density calculated using roads and trails was not
differentwithin pack territories established in year groups connected by a
line (p > 0.10 from Fisher LSD multiple comparison test). Groupings were
the same when road density was calculated using roads only (p > 0.10 from
Fisher LSD multiple comparison test). ............................................................. 31

Fig. 2.4 Predicted wolf habitat quality within zone 1 (primary wolf habitat) on a
scale of 0-100 using a logistic regression model developed from data
collected in the Highway 53 Study Area in the northwest Wisconsin
1992- 2001 compared to documented wolf pack territories. .............................. 38

CHAPTER 3: COVER TYPE AND LAND OWNERSHIP SELECTION OF
RECOLONIZING GRAY WOLVES IN NORTHWEST WISCONSIN

Fig. 3.1 Wolf pack territories in the Highway 53 Study Area in northwest Wisconsin
and adjacent Minnesota 1995-2001. ................................................................... 62

xiii

CHAPTER 1
INTRODUCTION AND DISSERTATION STRUCTURE

INTRODUCTION

The gray wolf (Canis lupus) was listed as an endangered species in 1974 under
the Endangered Species Act (ESA). Prior to 1974, the wolf was extirpated from its
historical range in the lower 48 states with the exceptions of northeast Minnesota and Isle
Royale in Lake Superior (Bailey 1978). At the time of listing, 550 - 650 individuals
represented the entire United States population outside of Alaska (Mech 1970, Peterson
and Page 1988, Fuller et al. 1992). Much of the reduction in wolf populations was due to
human activities, which historically have been the major source of wolf mortality (Mech
1970)

With a reduction in human induced mortality due to the species listing under the
BSA and an increasingly favorable public attitude toward wolves, the Minnesota wolf
population steadily increased during the mid-19705 and 19805 (Fuller et al. 1992). With
this increase in the number of individuals, dispersing wolves began to recolonize
northwest Wisconsin in the late-19705 (Mech and Nowak 1981, Thiel and Welch 1982).
At this time northwest Wisconsin offered an abundant prey base and large contiguous
tracts of land unoccupied by wolves. Exploitation of these factors by wolves has led to a
steady expansion of the wolf population both numerically and geographically over the
past 25 years. In 2002, the Wisconsin population was estimated at approximately 270
individuals (A. P. Wydeven, Wisconsin Department of Natural Resources (WDNR), pers.

comm.)

During this period of wolf population growth, wolves in the Great Lakes region
have dispersed into a wide variety of habitats outside of traditional wilderness areas
(Mech 1995). Expansion into these areas has increased the need for research examining
wolf ecology in human dominated, non-wilderness or semi-wild landscapes. Past
research in these areas has focused on relating wolf resource selection to landscape
features including land cover, road density, prey base, and land ownership patterns
(Mladenoff et al. 1995, 1999).

The importance of examining resource selection at differing scales

Mladenoff et al. (1995) examined resource selection at the landscape level (within
a 171,000 km2 study area) using the geographic information system (GIS) ARC/INFO to
compare the landscape attributes of Wisconsin wolf pack territories to similarly sized
unused areas. This study found that road density was the best predictor of wolf territory
placement (Mladenoff et al 1995). Wolf pack territories were also found to contain
higher percentages of publicly owned land, mixed conifer-hardwood forest, and forested
wetlands than unused areas (Mladenoff et a1. 1995). Human population density was also
found to be lower in pack territories than in unused areas (Mladenoff et a1. 1995).

Though the studies by Mladenoff et a1. (1995, 1999) provide useful information
about wolf habitat use in a semi-wild landscape, they examined resource selection at only
one scale. Mladenoff et al. (1995, 1999) examined resource selection at the level of
territory placement on the landscape (i.e., where territories occur and why). Resource
selection occurs hierarchically from the geographic range of a species, to individual home
ranges within this geographic range, to use of general features within the home range, to

the selection of particular elements within the general features (Manly et a1. 1993). This

hierarchy for wolves proceeds from selection of territories, to the use of particular habitat
types within territories, to the selection of specific locations such as den sites within
habitat types or portions of the territory. The criteria for resource selection may differ at
each level (Johnson 1980, Weins 1981). It has been suggested that as a general rule,
researchers should consider studying selection at more than one scale (Manly et al. 1993).
Thus, management guidelines based on research conducted at one scale of resource
selection may be inadequate. By examining resource selection at multiple scales based
on the organism’s biology, a more complete understanding of a species’ resource .
requirements can be gained.

The establishment and location of wolf pack territories across a landscape results
from the cumulative effects of factors often beyond the control of individual land
managers. However, the activities of individual land mangers undoubtedly influence the
landscape’s potential to support wolves. These activities typically involve the
manipulation of habitat and or human activity at a scale corresponding to individual
territories. Thus, analyses of resource selection at multiple scales (i.e., selection of
territory within the landscape, differential use of habitat within the territory, and (den)
site selection within the territory) is necessary.

To facilitate these types of analyses, land cover maps with a resolution that
corresponds to the finest scale of analysis must be used. For example, Mladenoff et a].
(1995) evaluated the regional placement of pack territories using a land cover map based
on a 16 ha resolution satellite image. This resolution was sufficient for the coarse
analysis of territory placement. In contrast, for the evaluation of habitat use and den site

placement within individual territories, a finer resolution is required. Mladenoff et a1.

(1995) acknowledged this and state “indices of landscape characteristics. . .must be
interpreted in the context of the resolution and classification specificity of the land cover
data from which they are derived.”

Changes in resource selection as population density increases

Two assumptions of resource selection theory are that a species will select
resources that best satisfy its life requirements and that higher quality resources will be
selected first (Manly et a1. 1993). If these assumptions are valid, the first pack territories
established in an area will contain the highest quality habitat and subsequent territories
will contain lower quality habitat. This process appears to be occurring in northwest
Wisconsin where local wolf populations have been intensively monitored since 1992
(Kohn et a1. 1997). The number of packs in this area has increased from 3 in 1992 to 15
in 2001.

Since the mid-1980’s road density has been the most commonly used measure of
habitat quality for wolves, with higher road densities associated with lower quality
habitat (Mech 1995). Several of the pack territories within northwest Wisconsin are in
areas with road densities identified as unsuitable in past studies (Thiel 1985, Jensen et a1.
1986, Mech et a1. 1988, Mladenoff et a1. 1995). Also, in contrast to the findings of
Mladenoff et a1. (1995), where no pack territories were bisected by major federal or state
highways, 3 pack territories in this area are bisected by U. S. Highway 53 (H53). The
establishment of territories (if persistent and productive) in areas of what is generally
considered marginal or unsuitable habitat (e. g., higher road densities, more human
activity, more Open land cover), may indicate that our perceptions of suitable habitat

within the Great Lakes region need to be revised.

In the past decade, wolves have been documented in a variety of non-traditional
settings. For example, wolves in Spain have persisted in predominantly agricultural areas
with human densities of up to 200 people/km2 (Vila et al. 1993). Fuller et a1. (1992)
found that wolves are occupying areas of higher road densities than any previously
reported in Minnesota. Mech (1995) also reported that the range of wolves in the Great
Lakes region continues to include new areas that are “much more open, accessible, and
heavily populated.”

Although recent research has provided valuable insight into wolf habitat
associations at the landscape level, including a density-dependent component in this
analysis can increase our understanding. An improved understanding of changes in wolf
resource selection as the population in the Great Lakes region continues to expand is
important for several reasons. If wolves are adapting to successfully use areas that have
been regarded as marginal quality or unsuitable habitat (e.g., higher road densities, more
human activity, differing land covers), present estimates of carrying capacity (Mladenoff
and et a1. 1997) in Wisconsin (380) may need to be updated. With state management
agencies in the Great Lakes region striving to develop wolf management guidelines and
population goals for the time period after delisting from the ESA, accurate estimates of
biological carrying capacity are increasingly more important. Another need to accurately
quantify and map wolf habitat relates to the potential for increasing conflicts between
wolves and human activities as the number of wolves in the area continues to grow.
Though wolves may be successfully colonizing areas that in the past would have been
viewed as unsuitable, the higher level of human activity in many of these areas increases

the likelihood of conflict between wolves and humans.

This study examined changes in resource selection over a 9-year period (1992-
2001) encompassing a period of recolonization and geographic expansion by gray wolves
in a semi-wild landscape. Resource selection was examined at the following scales:
territory selection on the landscape, within the territory, and at den sites.

The information gained from this study will increase our understanding of gray
wolf ecology in semi-wild landscapes and provide several benefits to resource
professionals managing areas within current or future wolf range. It will aid in more
accurately estimating biological carrying capacity within a given region. Appropriate
population objectives cannot be set by management agencies without reasonable
estimates of biological carrying capacity. Also, a greater understanding of resource use
in this region will allow us to more accurately predict areas of potential conflict between
wolves and humans as wolves expand into more developed areas. Results of this study
may also be useful for assessment of wolf habitat potential in the northeastern United
States. This area is generally regarded as the next potential area for gray wolf
recolonization or reintroduction. The northeastern United States has ecosystems, land
use, and amounts of favorable habitat more similar to the Great Lakes region than to
areas examined in most past studies dealing with gray wolf ecology (Mladenoff and
Sickley 1998). These similarities indicate that the findings from this study should be
more useful in the delineation of wolf habitat (and resultant management decisions) in the
northeast United States than the results of past studies in other portions of gray wolf
range.

DISSERTATION STRUCTURE

This Dissertation is divided into five chapters:

Chapter 1: Introduction and Dissertation Structure

Chapter 2: Pack Territory Placement and Habitat Characteristics of a
Recolonizing Population of Gray Wolves in Northern Wisconsin

Chapter 3: Cover Type and Land Ownership Selection of Recolonizing Gray
Wolves in Northwest Wisconsin

Chapter 4: Gray Wolf Den Site Characteristics in Northwest Wisconsin:
Additional Data

Chapter 5: Synthesis and Future Research Direction

Note: Images in this dissertation are presented in color.

STUDY AREA

This research was conducted in the U. S. Highway 53 study area (H53SA) in
portions of Douglas, Washbum, Burnett, Bayfield, and Sawyer counties in northwest
Wisconsin and Pine county in east-central Minnesota (Figure 1). This study area was
established by the WDNR in response to a project to expand H53 from 2 lanes to 4 lanes
in 1992. This expansion took place along a 44-mile section of the highway from Trego in
the south to Hawthorne in the north. The final 6-mile section of this expansion was
completed in the summer of 1999. This stretch of highway passes through established
wolf territories and bisects the primary wolf dispersal corridor from Minnesota into
Wisconsin (Kohn et a1. 1997). This prompted the Wisconsin Department of
Transportation (WDOT) to contract with the WDNR to conduct a study examining the
effects of the expansion on wolves in the area.

The landscape in the H53SA is primarily a mixture of upland mixed deciduous
forest (Acer rubrum, Quercus rubra, Q. alba, Betula papyrifera), aspen (Populus
tremuloides, P. grandidentata), grasslands (Pteridium aquilinum, Hieracium
aurantiacum, Phleum pratense, Poa compressa, P. pretensis, Curtis 1992), idle pasture,
mixed upland deciduous/coniferous forest (Acer rubrum, Quercus rubra, Q. alba, Betula
papyrifera, Pinus banksiana, Pinus strobus, Picea glauca), wetland deciduous shrubs
(Alnus rugosa, immature F raxinus pennsylvanica), wetland forest (Thuja occidentalis,
Picea mariana, Fraxinus nigra, Ulmus rubra), emergent/wet meadow and bogs (Scirpus
spp., Sphagnum spp; Curtis 1992) and agricultural lands. The topography in this area is

that of a rolling plain (Curtis 1992).

    
  

Hawthorne

Douglas

 
 
   

  

US Hwy.
53

 
 
 
 
 

- Pine County

I Minnesota

L.

  
 
 

Washbum County

   

Burnett County

Figure 1. The Highway 53 Study Area (H53SA) in northwest Wisconsin and east-central
Minnesota.

Available prey within the H53SA consists largely of white-tailed deer
(Odocoileus virginianus) and beaver (Castor canadensis), both of which have been
shown to be primary food sources for wolves in the Great Lakes region (Mandemack
1983, Fuller 1989). Estimates of deer densities within the H53 SA have ranged from 9.7
deer/km2 to 13.5 deer/km2 (mean = 11.68 deer/km2,) since 1995 (B. E. Kohn,WDNR,
pers. commun.). Beaver are also common in the H53 SA, with 1998 helicopter surveys in
northwest Wisconsin estimating densities at 0.61 active beaver colonies/km2 (B. E. Kohn,

pers. commun.).

10

LITERATURE CITED

Bailey, R. editor. 1978. Recovery plan for the eastern timber wolf. U. S. Fish and
Wildlife Service, Washington D. C. USA.

Curtis, J. T. 1992. The vegetation of Wisconsin: an ordination of plant communities.
The University of Wisconsin Press, Madison Wisconsin, USA.

Fuller, T.K. 1989. Population dynamics of wolves in north-central Minnesota.
Wildlife Monographs 105.

_, W. E. Berg, G. I. Radde, M. S. Lenarz, and G. B. Joselyn. 1992. A history
and current estimate of wolf numbers and distribution in Minnesota. Wildlife
Society Bulletin 20:42-55.

Jensen, W. F ., T. K. Fuller, and W. L. Roinson. 1986. Wolf (Canis lupus)
distribution on the Ontario-Michigan border near Sault Ste. Marie.

Canadian Field Naturalist 100:363-366.

Johnson, D. H. 1980. The comparison of usage and availability measurements for
evaluating resource preference. Ecology 61 :65-71.

Kohn, B. E., J. Frair, D. Unger, T. Gehring, D. Shelley, E. Anderson, and J. Ashbrenner.

1997. Impacts of highway development on northwestern Wisconsin timber
wolves, preliminary findings-May, 1992 through November, 1996. Wisconsin
Department of Natural Resources. Madison, Wisconsin, USA.

Mandernack, BA. 1983. Food habits of Wisconsin timber wolves. Thesis,

University of Wisconsin-Eau Claire. Eau Claire, Wisconsin, USA.

11

Manly, B. F., L. L. McDonald, and D. L. Thomas. 1993. Resource selection by
animals, statistical design and analysis for field studies. Chapman and Hall,
London, United Kingdom.

Mech, L. D. 1970. The wolf: the ecology and behavior Of an endangered species.
Natural History Press, Garden City, New York, USA.

_. 1995. The challenge and opportunity of recovering wolf populations.
Conservation Biology 92270-278.

_, and R. M. Nowak. 1981. Return of the gray wolf to Wisconsin. American
Midland Naturalist 105:408-409.

_, S. H. Fritts, G. Radde, amd W. J. Paul. 1988. Wolf distribution in
Minnesota relative to road density. Wildlife Society Bulletin 16:85-88.

Mladenoff, D. J., T. A. Sickley, R. G. Haight, and A. P. Wydeven. 1995. A regional
landscape analysis and prediction of favorable gray wolf habitat in the northern
Great Lakes region. Conservation Biology 9:279-294.

_, R. G. Haight, T. A. Sickley, and A. P. Wydeven. Causes and implications of

species restoration in altered ecosystems. Bioscience 47:21-32.

_, and T. A. Sickley. 1998. Assesing potential gray wolf restoration in the
northeastern United States: a spatial prediction of favorable habitat and potential
population levels. Journal of Wildlife Management 62:1-10.

_, T. A. Sickley, and A. P. Wydeven. 1999. Predicting gray wolflandscape

recolonization: logistic regression models vs. new field data. Ecological

Applications 9:37-34.

12

Peterson, R. 0., and R. E. Page. 1988. The rise and fall of Isle Royale wolves, 1975
-1986. Journal of Mammalogy 69:89-99.

Thiel, R. P. 1985. Relationship between road densities and wolf habitat suitability in

Wisconsin. American Midland Naturalist 133:404-407.

_, and R. J. Welch. 1982. Evidence of recent breeding activity in Wisconsin
wolves. American Midland Naturalist 1062401-402.

Unger, DE. 1998. A multi-scale analysis of timber wolf den and rendezvous
sites in northwestern Wisconsin and east-central Minnesota. Thesis, University
of Wisconsin-Stevens Point, Stevens Point, Wisconsin USA.

Vila, C., J. Castroviejo, and V. Urios. 1993. The Iberian wolf in Spain. Pages 104-109 in
C. Promberger and W. Schroeder, editors. Wolves in Europe: status and
perspectives. Munich Wildlife Society, Ettal, Germany.

Weins, J. A. 1981. Scale problems in avian censusing. Pages 513-521 in Ralph C. J.
and J. M. Scott, editors. Estimating numbers of terrestrial birds. Studies in Avian

Ecology Number 6.

13

CHAPTER 2
PACK TERRITORY PLACEMENT AND HABITAT CHARACTERISTICS OF A
RECOLONIZING POPULATION OF GRAY WOLVES IN NORTHERN

WISCONSIN

INTRODUCTION

Two major assumptions of resource selection theory are that a species will select
resources that will best satisfy its life requirements and that higher quality resources will
be selected first (Manly et a1. 1993). If these assumptions are valid for wolves, the first
pack territories established in an area will contain the highest quality habitat and
subsequent territories will contain lower quality habitat. This process may be occurring
in northwest Wisconsin where local wolf populations have been intensively monitored
Since 1992 (Kohn et a1. 1997). The number of packs in this part of the state has increased
from 3 in 1992 to 15 in 2001 and the habitat characteristics of pack territories established
early in this period appear to differ from those of territories established later in the
recolonization (B. E. Kohn, pers. commun.).

Although the estimated number of wolf packs in Wisconsin has varied greatly
since their recolonization of the state (2 in 1979, 52 in 2001, A. P. Wydeven, pers.
commun.), no examination of density-dependent changes in resource selection has been
conducted. Due to the strong territorial behavior of wolves, previously established packs
will limit the choice of potential territory locations for additional packs within an area
(Mech 1970). Thus, only the first pack to establish a territory in Wisconsin was
unlimited (in relation to intraspecies strife) in its Choice of territory location. This

limitation affects wolf pack territory placement and our perceptions of habitat selection

14

and suitability. An examination of the characteristics of pack territories established
sequentially throughout recolonization will provide insight into whether the attributes of
areas being colonized by wolves are changing as the wolf population and pack density
increase.

Past research in the Wisconsin has focused on relating wolf resource selection to
landscape features including land cover, road density, prey base, and land ownership
patterns (Mladenoff et al. 1995). Mladenoff et a1. (1995) examined resource selection at
the landscape level (within a 171,000 km2 study area) using the geographic information
system (GIS) ARC/INFO to compare the landscape attributes of Wisconsin wolf pack
territories to similarly sized unused areas. This study found that road density was the best
predictor of wolf territory placement (Mladenoff et al. 1995). Wolf pack territories were
also found to contain higher percentages of publicly owned land, mixed conifer-
hardwood forest, and forested wetlands than unused areas (Mladenoff et al. 1995).
Human population density was also found to be lower in pack territories than in unused
areas (Mladenoff et al. 1995).

Although past research has provided valuable insight into wolf habitat
associations at the landscape level, our understanding can be increased by examining
changes in resource use and selection as wolf populations and pack density increase. An
improved understanding of changes in wolf resource selection as the population in the
Great Lakes region continues to expand is important for several reasons. If wolves are
adapting to successfully use areas that have been regarded as marginal quality or
unsuitable habitat [e.g., higher road densities (Merril 2000), more human activity (Vila et

al. 1993), differing land covers (Mech 1995)], present estimates of carrying capacity

15

(Mladenoff and Sickley 1998) in the region may be low. With state management
agencies in the Great Lakes region striving to develop wolf management guidelines and
population goals for the time period after delisting, accurate estimates of biological
carrying capacity are increasingly important. Another need to accurately quantify and
map wolf habitat relates to the potential for increasing conflicts between wolves and
human activities as the number of wolves continues to increase. Although wolves may
be successfully colonizing areas that in the past would have been viewed as unsuitable,
the higher level of human activity in many of these areas increases the likelihood of
conflict between wolves and humans.

I conducted an analysis of the characteristics (land cover category, road density,
land ownership category, and landscape pattern) of wolf pack territories established
between 1992 and 1999 in northwest Wisconsin. My first objective was to examine
whether the characteristics of pack territories changed based on the year of pack initiation
(and the related difference in the number of packs in the study area). The assumption
underlying this analysis was that packs establishing territories later in the study would be
more limited in where they could place territories and might be forced to establish
territories in lower quality habitat. This analysis provides insight into whether pack
territory characteristics are changing as the wolf population and resultant wolf pack
density increases. My second objective was to examine differences in the characteristics
of wolf pack territories compared to corresponding unused areas. This examination
provides information on what habitat factors are influencing the placement of wolf pack
territories. My final objective was to use the data from the comparison of territories and

unused areas to develop a logistic regression-based GIS model to predict the occurrence

16

and spatial pattern of suitable wolf habitat and resulting wolf pack territory placement.
This model, while developed from data collected only in the H53SA, allowed me to
generate a habitat-based carrying capacity for wolves in the entire area I judged to be
primary wolf range (PWR) [(zone 1 in Wisconsin wolf management plan (WIDNR
1999)] (Figure 2.1) in Wisconsin.

METHODS

Capture, Handling, and Radiotelemetry

Wolves were trapped by WDNR personnel using modified #14 Newhouse leg-hold traps
(Schultz et al. 1996) May-August 1992-2000. Captured wolves were immobilized using
a mixture of ketamine hydrochloride/xylazine hydrochloride administered with a jabstick.
Any trap-related injuries were treated, blood samples were taken, an injection of
antibiotics was administered, and animals were weighed. Animals > 15kg were also
fitted with radio collars (Telonics Inc., Mesa, Arizona) and uniquely numbered ear tags.
Yohimbine was administered as a reversal agent for the xylazine hyrdrochloride and
wolves were monitored until they could move away from the trap site under their own
power.

Radio-collared animals were located 1-5 times per week using telemetry from
fixed-wing aircraft (Ballard and Dau 1983, Fuller 1989) or a vehicle-mounted 5-element
Yagi directional antenna until radio failure, death of the animal, or a permanent
movement out of the study area. Locations were collected from January 1992 — April
2001. The majority (>95%) of locations were collected during daylight hours and a
minimum of 3 bearings were collected for each location. Triangulations for locations

collected on the ground where solved using a Lenth estimator (Lenth 1981) in the

17

o¢0f 00000000.0000
J00000000000000c

0000p
000000
0000000r
0 00000r

Primary Wolf Range (PWR)

 

 

 

 

 

 

 

 

0
0000000/
0000000».
000.00000(
.000000000
0000000000
0000000000
.00000000.
..J000000
0000000.
0000000
.000r .000
0000000.
0001/000
0000/0000
0000000000
0000000/000
40/0000000000

m HSBSA

I
0 o

0’

1000000
000000000

0000000000
00000000000

00000

‘—

If.

        

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.000000000000000
0000000J000000a
O . . .JOO th/r.
. 00. i
«Z

.L

 

p

 

 

 

 

 

 

 

 

 

 

 

lSCOIlSlI’l.

“I

In

If range (PWR)

primary wo

Fig 2.1 Location of the Highway 53 Study Area (H53SA) within

18

program Locate II (Pacer, Turo Nova Scotia, Canada). Shelly and Anderson (1995)
found no significant difference between the accuracy of aerial and ground locations in the
study area using the same telemetry techniques and equipment, so locations were pooled
for territory estimation and resource selection analyses. Locations obtained for an
individual wolf were at least 4 hours apart to maintain independence (Reynolds and
Laundre 1990). Date, time, Universal Transverse Mercator (UTM) coordinates, ear tag
number, and pack were recorded for each telemetry location. Animals were classified as
belonging to individual packs based on trapping location and movement patterns after
release.

Extensive track surveys were conducted each winter throughout the study area to
estimate the number of individuals in each pack and to locate any newly formed packs.
This effort as well as additional track searching and howling surveys each summer
allowed for the identification and monitoring of all resident pup producing packs within
the study area throughout the study. Knowing where all pack territories in the H53 SA
were was crucial in the analysis or territories versus unused areas. Efforts were also
made to maintain at least one radio collared animal in each pack at all times, although
there were periods when this was not possible.

Territory Boundary Estimation

Pack territory boundaries were estimated using a fixed kernel method (Worton
1989) based on an 80% isopleth using the animal movement analysis extension (Hooge
and Eichenlaub 1997) in ArcView [Environmental Systems Research Inc. (ESRI),
Redlands CA]. A minimum of 30 locations (Fuller and Snow 1988, Seaman et al. 1999)

was used in estimating territory boundaries. Only territories of packs which had been

19

monitored for at least one year and which had produced pups were included in the
analysis. All locations from all radio-marked animals within a pack during the course of
the study were used in territory boundary estimation.

Evaluation of Changes in Territory Characteristics With Increasing Population and
Territory Density

Variables potentially affecting wolf habitat suitability based on my understanding
of current paradigms of wolf habitat use and personal experience were determined for
each pack territory and corresponding unused area. These variables were: road density,
the proportion of each pack territory classified as forested wetland, aspen, lowland shrub.
shrub, and agricultural; the proportion of each pack territory under private, county, state,
and industrial forest ownership; and 5 measures of landscape pattern [weighted mean
fractal dimension (Mandelbrodt 1977), mean patch edge, mean patch area, edge density,
and Shannon’s diversity index (Shannon and Weaver 1949)].

Road density was calculated using a line coverage based on U. S. Geologic
Survey 1:100,000 scale Digital Line Graphs in ArcView. Two measures of road density
were calculated. The first was based on all roads and trails in the area (hereafter roads
and trails). The H53SA contains an extensive system of trails that are heavily traveled by
all-terrain vehicles during the summer and snowmobiles during the winter. The current
paradigm is that increased human access provided by higher road density into an area is
the mechanism that causes road density to be a useful indicator of wolf habitat suitability
(Fuller et al. 1992, Mladenoff et al. 1995). It is probable that the system of trails in the
H53SA provides additional access into otherwise roadless areas and I chose to include

trails in this measure of analysis. The second measure of road density included all parts

20

of the road network other than trails, hereafter referred to as “roads only”. This coverage
may have included more unimproved forest roads than the one used by Mladenoff et al.
(1995, 1999) in the region, but I judged it to be an equally appropriate metric for two
reasons. This measure of road density is easily derived from existing GIS data layers and
this widespread availability increases its utility to resource managers. Road density for
both measures of roads was determined by intersecting a roads coverage with the 80%
use isopleths representing wolf pack territories and calculating the straight-line road
distance and area of the pack territory. All road densities were calculated as km/kmz.
The proportion of each pack territory classified as forested wetland, aspen,
lowland shrub, agricultural, and shrub within the WDNR’s GIS WISCLAND land cover

classification system (http://wwwdnr.stale.\Vi.us/Org/al/ct/gco/data/ \VIC.I11111) was also

 

calculated. These proportions were determined by intersecting the WISCLAND land
cover layer with territory isopleths to calculate the area of each territory in a land cover
category divided by the total area of the territory. Mladenoff et al. (1995) identified
forested wetland as compromising a higher proportion of wolf pack territories than
corresponding unused areas and agricultural lands as comprising lower proportions of
pack territories than unused areas. 1 found that both forested wetland and aspen were
highly selected by wolves at both the landscape and within-territory level of analysis
(Chapter 3 this document). This examination also revealed that lowland shrubland was
selected and that agricultural lands and shrub were avoided by wolves at the landscape
level of analysis. Based on these findings, I believed that these cover types could serve
as reasonable measures of habitat suitability and as metrics to differentiate between wolf

pack territories and unused areas.

21

I also examined the proportions of county owned, privately owned, state owned,
and industrial forest land in pack territories. Privately owned land comprised a lower
proportion and county owned land comprised a higher proportion of pack territories than
corresponding unused areas in the study by Mladenoff et a1. (1995). Other publicly
owned lands including state owned were also found to be important to wolves by
Mladenoff et al. (1995). Industrial forest land was found to be important in some pack
territories by Mladenoff et al. (1995) and was also observed to comprise large portions of
some pack territories the H53SA (Chapter 3 this document). Based on these findings, I
believed that the four land ownership categories that I included in this analysis would
serve as useful indicators of wolf habitat suitability.

It has been suggested that landscape indices may be useful as descriptors of wolf
habitat (Mladenoff et al. 1995), therefore the following measures of landscape pattern
where also included in the analysis: 1) weighted mean fractal dimension, an index (scaled
1-2) of patch boundary complexity related to patch size (Mandelbrodt 1977), 2) mean
patch edge, 3) mean patch area, 4) edge density, and 5) Shannon’s diversity index
(Shannon and Weaver 1949). These spatial indices were generated using the Patch
Analyst extension (Elkie et al. 1999) within ArcView.

Mean values of measured landscape variables were generated for each pack
territory. Pack territories were grouped based on the year of pack establishment. Packs
present in the study area before 1992 were classified as having been established in 1992.
The means for each variable were grouped by year of territory initiation and analyzed
with a one-way ANOVA (Zar 1999) (\7’ = 0.10) to determine if pack territory

composition changed as decreasing amounts of the H53 SA were available for

22

colonization by new packs. A number of unused areas equal to the number of pack
territories were generated for each year of the analysis to allow comparisons between
habitat characteristics in areas occupied by wolves and areas that were not occupied by
wolves. For example, there were 3 pack territories established in 1999, so 3
corresponding unused areas were generated and included in the analysis for 1999. These
areas were the same size as the average size of pack territories in the year of interest and
where randomly located within the H53 SA with the restriction that no part of the random
area could intersect a pack territory. These areas were generated using the pixel
aggregation process described by Roloff and Haufler (1997) and each was iterated 5
times to avoid potential bias due to the shape or position of the random area. Calculated
values for measured variables were averaged across all 5 iterations of a random site
before comparison with used sites. The same landscape variables measured in pack
territories were also measured in corresponding random areas and means for each
variable were derived. The means of pack territories and unused areas were compared
using a Wilcoxon rank-sum test (or = 0.10) (Zar 1999).
Development of Landscape Level Habitat Suitability Model

All variables examined in the year of initiation analysis were considered for
inclusion in a logistic regression analysis (Zar 1999) differentiating between pack
territories and unused areas. All variables considered for inclusion in the analysis were
checked for colinearity using 3 Pearson correlation analysis before they were entered into
the model. The final choice of which variables to include in the logistic model was based
on observed differences in mean values between pack and unused territories, a forward

selection procedure, and knowledge of wolf ecology gleaned from relevant literature and

23

personal experience. Proportions of area within pack territory or unused area classified
as forested wetland and county owned lands as well as the road density (km/kmz) of the
roads only coverage were selected for inclusion in the logistic model.

One adjustment was made to the land ownership category values input into the
logistic model. There is no national forest land in the H53 SA, but based on past
predictions of wolf habitat suitability in Wisconsin (Mladenoff et al. 1995, 1999) I
believed that national forest land would contribute to quality habitat for wolves. Since I
planned on testing the logistic model in portions of the state that included national forest,
I needed to assign a value for national forest land within the model. The best method of
selecting a value for national forest land was to assign it the same value as the most
' similar land ownership category that was present in the H53 SA that had been assigned a
value by the logistic regression procedure. The land ownership category present in the
H53 SA with the most similar management regime and structure to national forest was
county owned. Based on this, I assigned the same logistic-based value (coefficient) to
national forest land as was assigned to county owned lands by the logistic regression
procedure.

Logistic models including all possible combinations of proportion of forested
wetland, proportion of county owned land and road density were compared using
maximum likelihood goodness-of-fit and classification accuracy. The classification
accuracy of the final logistic model was also tested against 15 pack territories and 15
unused areas outside of the H53 SA, which had not been used in developing the logistic

model.

24

Once the final model had been selected it was programmed into ARC/INFO and
tested by comparing areas predicted as being favorable wolf habitat with known wolf
pack territories in PWR in Wisconsin. This was accomplished by creating coverages of
both forested wetland and county owned lands (grouped with national forest land) in
ARC/INFO and calculating the proportion of these two variables across the landscape.
This same process was used to determine road density across the landscape.

The resulting values for each 30 m pixel across the landscape were entered into
the logistic regression model within the Grid module in ARC/INFO. This process
yielded an output GIS coverage with “habitat values” between 0 and 1 for each 30 m
pixel across the landscape (0 = the lowest quality wolf habitat and 1 = the highest quality
habitat). These values were converted to 0 — 100 to facilitate running the pixel
aggregation procedure described by Roloff and Haufler (1997). This process simulated
the placement of wolf pack territories across PWR in Wisconsin based on both the quality
of habitat as predicted by the logistic model and the spatial arrangement of that habitat.
This process randomly assigned seeds or starting pixels across PWR. It subsequently
aggregated pixels until the area being selected equaled the average size of wolf pack
territories analyzed in the H53SA (119 kmz). A minimum habitat value for inclusion in
an area being selected is set within this process. Pixels with habitat values below this
lower limit may be included in simulated territories, but these pixels will not contribute
toward the preset target size (average wolf pack territory size in this case). This
acknowledges the fact that while low quality habitat may be included in a territory, it will
contribute little to the value of the territory, and therefore territories comprised of low

quality habitat will generally be larger than territories comprised of high quality habitat.

25

The following guidelines were incorporated into this process. Seeds were not placed in
pixels with habitat values less than 10 to recognize that there is some lower bound of
habitat quality below which wolves would not attempt to establish a territory. The
process was set to generate 50 pack territories since this is the approximate number of
wolf pack territories estimated in PWR in the state. I set a lower habitat value limit of 30
for a pixel to contribute to the size of the area aggregated by the procedure. The level of
this lower limit was developed by calculating the average habitat value for pixels within
actual wolf pack territories as plotted by the WDNR (68.3) and subtracting one standard
deviation (38.7) from this value. I hypothesized that wolves would expand into lower
quality habitat as the population and pack density increased and the minimum habitat
value for this analysis should be set lower than means within currently occurring pack
territories.

The pixel aggregation process was iterated 10 times and the number of times each
pixel across PWR was included in a Simulated pack territory was calculated. This
created a frequency distribution with values ranging from 0-10. This coverage was used
to compare the probability of areas occupied by wolves being selected by the pixel
aggregation procedure to the probability of areas not occupied by wolves being selected.
The mean and frequency distributions of the pixel values (from the 0-10 distribution
output by the pixel aggregation procedure) of areas occupied by wolves and areas not
occupied by wolves were compared using a Wilcoxon rank-sum test and a Kolmogorov-
Smirnov test (01 = 0.10). These tests were conducted to determine if the combination of
the logistic regression model and the pixel aggregation technique were actually

differentiating between favorable and unfavorable wolf habitat in a non-random manner.

26

I assumed that if the pixel aggregation procedure was working correctly, the likelihood of
a pixel being selected (mean pixel value from the 0-10 frequency distribution) within
actual wolf pack territories should be higher than in areas that wolves are not using and
that the frequency distribution of individual pixel values should also be different.

The areas selected by the pixel aggregation technique were also converted into a
GIS coverage and used to clip the original coverage of habitat values derived from the
logistic regression model. The pixel values from this coverage were compared to the
pixel values from a coverage created by clipping the same logistic-generated habitat
value coverage with a coverage of actual wolf pack territories using a 2 sample t-test (V =
0.10). This was done to determine if the combination of the logistic regression model
and the pixel aggregation technique was selecting areas that are actually being used by
wolves. I assumed that if this combination was working as planned, the habitat values in
pixels within areas selected by the pixel aggregation procedure should not be different
than those in areas actually occupied by wolves.

The final step in this analysis was to run the pixel aggregation technique across
PWR with an initial number of seeds set to 500 and set to aggregate pixels to the average
size of wolf pack territories observed in the H53SA. This process saturated the landscape
with simulated territories and provided an estimate of the maximum number of wolf pack
territories possible within PWR based on the logistic model developed from my data in
the H53 SA. This estimated maximum number of packs was combined with the average
pack Size of wolves in Wisconsin (4.3 individuals, Wydeven and Cervantes 1997) to
estimate a carrying capacity for PWR. Past estimates of habitat-based carrying capacity

(Fuller et al. 1992, Mladenoff and Sickley 1998) have been developed by estimating the

27

total area of suitable or favorable habitat and dividing that by the average size of wolf
pack territories and interstices. This approach fails to take into account the spatial
arrangement of favorable habitat and may over estimate carrying capacity. For example,
if there is enough favorable habitat for 10 territories in an area but it is broken up into 20
patches, there will almost certainly be less than 10 territories established in the area.
Including the spatial pattern of favorable habitat when estimating carrying capacity yields
a more accurate estimate.
RESULTS

During the course of the study, 8,303 telemetry locations from 40 wolves in 15
packs were collected and analyzed. Three packs were present in the H53SA when the
study began in 1992. Additional packs were detected in 1993 (2), 1995 (4), 1997 (3), and
1999 (3). This led to 5 distinct years of pack initiation, which were used in analysis of
changes in pack territory characteristics with increasing territory density across time.
There was a general trend of packs established earlier in the study being located further
west in the study area (nearer the core Minnesota wolf population). Packs established
later in the study tended to be located further east (Figure 2.2). Average pack territory
size based on the 80% use isopleth was 1 19.1 km2 (range 9.5- 347.1 kmz).
Evaluation of Changes in Territory Characteristics With Increasing Population and
Territory Density

Pack territories established later in the study had higher road density than those
established early in the study (Table 2.1, Figure 2.3). Mean road densities calculated
using roads and trails ranged from 0.67 krn/km2 in packs established in 1992 to 1.37

km/km2 in packs established in 1999. Mean road densities calculated excluding trails

28

m 8:00.53
:89 30239: 88065 E038 28 8:35: he he» 88065 830 38-82 538mg»

m 1* 3 suite 5 82 beam m aafifi 2: e53 Eda eerie Seizes e 30> S a;
Z

5.58 «RE .I I

m egg 3

32 I

82 1.0.,
FILIE

32 I
82 l
89 I

8:33
seem me as,

 

‘——_-

 

29

Table 2.1 Mean road density within wolf pack territories established in different

years calculated from coverages using roads and trails and roads only in the Highway
53 Study Area in northwest WI 1992-2001.

 

 

roads and trails roads only

density (krn/kmz) density (km/kmz)
Year of Establishment (n) x (SE) x (SE)
1992(3) 0.67 (0.12) 0.35 (0.17)
1993(2) 0.99 (0.15) 0.54 (0.21)
1995(4) 0.90 (0.11) 0.67 (0.15)
1997(3) 1.15 (0.12) 1.01 (0.17)
1999(3) 1.37 (0.12) 1.09 (0.18)
All packs (13) 1.09 (0.11) 0.70 (0.13)

 

30

1992 1993 1995 1997 1999

 

1992

 

 

1993
1995

 

1997

 

1999

 

 

 

Fig. 2.3. Differences in road density within wolf pack territories established in
Different years in the Highway 53 Study Area in northwest Wisconsin 1992-
2001. Mean road density calculated using roads and trails was not different
within pack territories established in year groups connected by a line (p >
0.10 from Fisher LSD multiple comparison test). Groupings were the same
when road density was calculated using roads only (p > 0.10 from Fisher
LSD multiple comparison test).

31

from the analysis ranged from 0.35 km/km2 in packs established in 1992 to 1.09
km/km2 in packs established in 1999 (Table 2.1).

State owned land comprised a higher proportion (p < 0.001) of territories
established in 1992 (49%) than in territories established in all other years of the study
(a = 0.10 for Fisher LSD multiple comparison test). No other differences (p > 0.10)
in land ownership composition or land cover category based on year of territory
establishment were detected (Tables 2.2 and 2.3)

Differences in Measured Habitat Variables Between Pack Territories And
Random Unused Areas

Mean road density within pack territories was lower than in unused areas for roads
and trails (p < 0.001 , 1.06 and 1.57 km/kmz, respectively) and roads only (p < 0.001.
0.70 and 1.43 km/kmz, respectively) measures of road density (Table 2.4). County
owned land comprised a higher proportion (p< 0.001) of wolf territories (61%) than
unused areas (15%) (Table 2.4), while private lands comprised a lower proportion (p
< 0.001; 19%) of territories than of unused areas (56%) (Table 2.4). Lowland shrub
(p = 0.005), forested wetland (p = 0.005), and aspen (p = 0.073) cover types were all
more prevalent in wolf territories than in unused areas (Table 2.4). No other
measured variables were different between territories and unused areas (p > 0.10).
Landscape Level Habitat Suitability Model

A Pearson correlation analysis of all variables considered for inclusion in the logistic
regression analysis indicated that the roads only and roads and trails measures of road

density were correlated (r = 0.89, p < 0.001). Based on this result, the forward step

32

.Eofifiznflmo me fine» wcoEm mooceotE wEEEEB <>OZ< Eat BEES 33> Q m

 

 

mg ed :58 86 A58 86 ed o.o meow 3.582
:3 898 w; 898 86 93.8 v2 :98 go 398 So 22$
mg 92.8 3.0 92.8 as 5.8 $6 ans :3 3:: ”No @580

33v cos :3 :98 OS od od A38 23 3%
an 32 82 82 82 ~02

 

.ooom-moo_ 33085» “835.5: E 33 beam mm 3333 05 E Eon—ammfifimo
ESE“: .«o a?» 3 330cm mocoHEB find .203 bum we 2mg 32:30:: :oEmanoo bowoumo 3:80:30 655 Nd 2an

33

.EoEfigfimo mo 28» mcofiw mooceotfi wEEmeo <>OZ< 80¢ 36.5%. 33> Q <

 

 

$6 66 $6.8 26 $666 666 66 66 mean 35%?
56 86.8 $6 866V 86 3666 E6 26.8 666 86.8 :6 ant:
26 9:66 666 $266 26 566 $6 A866 56 9:66 66.6 980

:66v 2666 S6 6666 S6 66 66 $666 666 3%
(a 36. 82 32 ~62 ~62

 

666362 586023 amino: 5 82 roam mm kiwi 2: 5 €252:st

both—8 mo .89» 3 $3st motoxhou x03 :95 bum mo Hfimmv £89883 83.0.8an howofio 95555 USS Nd 033.

34

Table 2.4 Mean (SE) of habitat variables in wolf pack territories and corresponding
random unused areas in the Highway 53 Study Area in northwest Wisconsin 1992-
2001.

 

 

 

Territory Unused

Variable E 813A 3? 313A 1)”

Roads and trails (km/km“) 1.06 0.01 1.57 0.1 1 <0.001
Roads only (km/kmz) 0.70 0.11 1.43 0.12 <0.001
County ownershipC 0.61 0.01 0.15 0.01 <0.001
Private ownershipC 0.19 0.01 0.56 0.01 <0.001
Lowland shrubC 0.13 0.02 0.06 0.02 0.005
Forested wetlandC 0.14 0.01 0.07 0.01 0.005
AspenC 0.22 0.15 0.09 0.1 1 0.073
Shrublandc 0.04 0.01 0.01 0.01 0.321
Agriculturec 0.02 0.03 0.02 0.02 0.396
Mean Fractal Dimension 1.39 0.01 1.38 0.01 0.203
Edge DensityD 316.41 40.96 309.73 28.63 0.253
Mean Patch AreaE 2.44 0.32 2.46 0.48 0.533
Mean Patch Edge F 767.73 111.12 744.97 115.74 0.309
Shannon’s Diversity Index 2.38 0.17 2.42 0.21 0.719

 

A Values < 0.01 rounded to 0.01.
B Wilcoxon Rank-Sum test of medians.
C Values reported as the proportion of the total territory or random unused area classified in this
category.
Meters/Hectare.
E Hectares.
F Meters/Patch.

35

logistic regression analysis was performed twice, once including the roads only
density and once including the roads and trails density. In both cases, the stepwise
procedure converged on a model including the proportions of forested wetland,
proportions of county owned land, and the measure of road density.

Log likelihood 12 (-21n [likelihood ratio]) tests indicated that both models
improved significantly over a model including only the logistic generated constant
(12 = 28.12, 3df, p <0.001 roads and trails, [2 = 29.57, 3 df, p <0.001 no trails).
Both of these models correctly classified 93% of the sampled sites as wolf pack
territories (14 of 15) or unused areas (14 of 15) (observations with p < 0.50 classified
as unused by the model, those with p > 0.50 are classified as a territory for purposes
of testing classification accuracy). Based on the log likelihood tests, I used the
logistic model incorporating the roads only measure of road density. The logistic

equation for this model was:

2 —1.455 —3.643(R) + 4.691(C) + 36.623(F)
1+ 8 —1.455 —3.643(R) + 4.691(C) + 36.623(F)

 

with a logit of: p = 1.455 - 3.643(R) + 4.691(C) + 36.623(F) where R is the density of
roads in km/kmz, C is the proportion of county owned land, and F is the proportion of
forested wetland in the area of interest. This model correctly classified 14 of the 15
pack territories and 13 of the 15 unused areas outside of the H53SA. When this
model was incorporated into a GIS, it delineated 22,694 km2 of favorable wolf habitat
[p > 0.5 from the logistic model (Mladenoff et al. 1995)].

The comparison of the probability of areas known to be occupied by wolves to
those not occupied by wolves being selected by the pixel aggregation procedure

indicated that areas occupied by wolves had a higher probability of being selected as

36

suitable habitat than areas which are not occupied by wolves (p <0.001, mean
probability from 0 - 10 range, 1.6 for pack territories and 0.62 for unused areas).
This indicated that the pixel aggregation technique was selecting pixels in a non-
random manner and that it was more likely to select areas being occupied by wolves.

The comparison of the habitat values generated by the logistic model in actual
wolf pack territories to areas that the pixel aggregation technique selected as suitable
habitat indicated that the aggregation technique was selecting habitat similar to what
wolves are selecting. The mean habitat value (0-100 range) for pixels in actual wolf
pack territories was 74.8 (SE = 36.14). The mean pixel value in areas selected as
suitable habitat by the pixel aggregation procedure was 59.36 (SE = 29.18). This
difference was not statistically different, (p = 0.32) indicating that the pixel
aggregation worked as intended. The distribution of habitat values generated by the
logistic model compared to wolf pack territories in PWR delineated by the WDNR
from radio telemetry locations is presented in Fig 2.4.

When the pixel aggregation technique was run to saturate PWR with
simulated territories, it selected 66 areas as being suitable for wolf pack territories.
Combining this with the estimate of a mean pack size of 4.3 wolves, PWR in
Wisconsin can support approximately 284 wolves in resident packs.

DISCUSSION

The mean road density observed within wolf pack territories increased
throughout the course of this study and was higher overall than that reported by
Mladenoff et al. (1995, 1999) in Wisconsin. There are several possible explanations

for this relating to differing methodology. The first is a difference in the definitions

37

duets.
flab...» 83388 again—8 88 -82 $3833 363:8 3.5 82 3.8 mm 33.3: 3:: 628:8 38 as 88.65
.88. 88262 6362 e was 82:6 oaoo a 8 332 as; have _ 88 as? 3:3 5.3 :2. 63285 66 are

3:353
boats gown. _|ll_

       

 

93:9:
oo— fio‘so. 8 33>
«332.. 03397.53

38

of the term “road.” Mladenoff et al. (1995, 1999) used a more conservative definition
based on features represented by solid lines on USGS 1:100,000 scale quadrangles
and that did not include “unimproved forest roads.” I reported values for my measure
of roads only density using any road feature that was not classified as a trail in a GIS
coverage developed within the WISCLAND data. This coverage was also developed
from USGS 1:100,000 quadrangles, but likely included some features that would
have been classified as unimproved by Mladenoff et al. (1995, 1999). I chose this
measure for two reasons. First, these data are easily attainable by WDNR land
managers. Second, from personal observations in the H53SA, road features that
would likely be classified as “unimproved” are used to access otherwise roadless
areas on a regular basis, especially during the summer tourist and fall hunting
seasons. The benefits of using this measure of road density balanced the loss of
ability to make direct comparisons between the measures of road density presented by
Mladenoff et al. (1995, 1999) and my own.

There was also a difference in home range estimation technique used.
Mladenoff et al. (1995, 1999) used a Harmonic Mean estimator (Dixon and Chapman
1980). Concerns over the reliability of this technique have been raised (Worton 1987.
Worton 1989, White and Garrott 1990) and I chose to use a fixed kernel (Worton
1989) instead. This likely lead to different definitions of territory boundaries and
resultant differences in measures of road density and other measured variables.

The final difference in methodology was related to the resolution of land
cover data used. Mladenoff et al. (1995, 1999) used a land cover data set with a 16 ha

resolution. I chose to use land cover data with a 30 m resolution. This difference

39

could have contributed to observed differences in wolf land cover category usage
between studies. The impacts on calculated landscape pattern metrics due to different
resolution land cover data could also have led to the difference in the importance of
fractal dimension to the logistic models developed in both studies. These differences
in methodology should be kept in mind when viewing differences in the reported
results of this study and Mladenoff et al. (1995, 1999).
Changes in Territory Characteristics With Increasing Population and Territory
Density

The mean road density observed for this study within pack territories was
higher than any in the Great Lakes region other than that reported by Merrill (2000)
(1.42 km/kmz) from Camp Ripley military reservation in central Minnesota. Merrill
(2000) proposed that the persistent occupation of this area of higher than normal road
density was possibly due to lower than normal speed limits and generally favorable
human attitudes toward wolves. It seems likely that wolf occupation of Camp Ripley
and areas of similarly high road densities in the H53SA is due not only to favorable
human attitudes toward wolves, but also to increasing tolerance of roads by wolves.
Due to their territorial nature dispersing wolves are constantly exploring areas
unoccupied by resident packs. As the landscape fills up with resident packs, the
unoccupied areas available for colonization by newly forming packs become less
preferred. In spite of the generally higher road densities and resultant higher levels of
human activity in many of these recently colonized areas, with the current level of
human tolerance and legal protection wolves appear to be establishing territories and

persisting in them. While wolves seem to prefer areas with low road densities and

40

low levels of human contact to establish territories, they will establish territories in
areas of higher road density and human activity if these are the only areas available to
them. The currently high levels of human tolerance and legal protection seem to
allow wolves to survive in these areas, which in the past would have been viewed as
marginal or unsuitable as wolf habitat. The increasing road densities within pack
territories established later in the study as the wolf population increased and the high
overall mean road density within territories indicate that the use of areas previously
viewed as unsuitable or unfavorable is occurring in the H53 SA. Even several authors
who identified road density as the most important measure of wolf habitat suitability
(Mech 1995, Mladenoff et al. 1995, 1999) acknowledge that wolves are expanding
into more human dominated areas and that this may require the revision of our
perceptions of what is suitable wolf habitat.

Our understanding of wolf habitat suitability since the mid 19803 has largely
been based on two conflicting paradigms. The first is based on the widespread
distribution of wolves in North America before European settlement and their habitat
generalist and highly adaptable nature (Mech 1970, 1995). The other has been the
paradigm that road density is the best measure of habitat suitability for wolves, with
higher road density indicating lower habitat quality and probability of occupation by
wolves (Thiel 1985, Jensen et al. 1986, Mech 1995, Mladenoff et a1. 1995, 1999).
The factor that allows both of these contradictory paradigms to be valid is the effect
of human activity on wolf populations. While wolves are very adaptable in their
habitat requirements, they generally do not inhabit areas of high road density due to

potential conflicts with humans. Mladenoff et al. (1995) and Mladenoff and Sickley

41

(1998) proposed that “roads per se are not avoided by wolves but serve as indicators
of human contact and thus the likelihood of deliberate or accidental human caused
mortality.” Fuller et al. (1992) proposed that where humans tolerate wolves, prey
density is likely the main limiting factor in wolf populations. This again indicates
that human activity is the main limiting factor on wolf populations, but without this
limitation, wolves are very adaptable in their habitat use.

It seems likely that past views of wolves as a wildemess species which needs
large roadless areas to survive are due to data on wolf distributions being gathered in
periods when human tolerance of wolves was low or in the period when wolves were
recolonizing formerly occupied areas as human tolerance increased. In short, wolves
were found only in wilderness areas because they were extripated in areas more
accessible to humans before public attitudes toward wolves began to improve in the
19705 (Mech 1995).

The increase in mean road density within pack territories established later in
this study is likely due to wolves taking advantage of the present relatively high level
of human tolerance of wolves by moving into areas with high levels of human
activity. High levels of human tolerance allow wolves to express their highly
adaptable nature and colonize areas with a high probability of human contact, but
which provide otherwise high quality habitat (e.g., high prey density, good denning
habitat). The higher mean road density in all pack territories within this study
compared to those observed by Mladenoff et al. (1995, 1999) in Wisconsin is likely
due to this same trend of wolves continuing to move into more human dominated

areas as the wolf population increases.

42

Differences in Measured Habitat Variables Between Pack Territories and
Random Unused Areas

The higher proportions of forested wetland observed in pack territories
compared to unused areas supports past findings in the area by Mladenoff et al.
(1995). It also supports the results of a compositional analysis (Aebischer et al. 1993)
conducted in the study area (Chapter 3, this document) that found forested wetland
was highly selected by wolves at both the landscape and within-territory levels of
analysis. The importance of forested wetland to wolves is likely related to both the
avoidance of human disturbance and the value of this cover type to white-tailed deer
(Odocoileus virginianus), which is the preferred prey item of wolves in Wisconsin
throughout the year (Mandemack 1983).

Areas of forested wetland are generally not hospitable to human activity that
requires movement away from roads due to difficulty in walking, high populations of
biting insects in the summer, and the general lack of human development. Wolves
are likely selecting forested wetland both when establishing territories and when
pursuing life needs within territories at least in part due to relatively low levels of
human activity. As wolves colonize more human dominated areas, tracts of forested
wetland may serve as refuges of low human activity compared to the territory as a
whole. Areas classified as lowland shrub also comprised a greater proportion of
territories than unused areas. The apparent importance of these areas to wolves is
also likely due to low levels of human access and activity in them.

Lowland conifer stands (classified within the forested wetland category in this

study) are an important component of deer habitat, especially during severe winters,

43

providing thermal cover and lower snow depths than surrounding areas (Venue 1965,
Blouch 1984). The high use of these areas by white-tailed deer likely leads to
increased use by wolves and the resulting importance of forested wetlands areas to
wolves.

The higher proportion of aspen in pack territories than in unused areas is also
likely tied to white tail-deer habitat use. Aspen has been identified as the most
favorable forest type for deer production in the northern Great Lakes (Byelich et al.
1972). It is also a preferred habitat of Wisconsin deer (Habeck and Curtis 1959) and
was identified as dominating the diet of deer in northern Wisconsin during the early
and late summer by McCaffrey et al. (1974). This preference for aspen as a food
source likely leads to the high level of deer use observed by McCaffrey and Creed
(1969) during the summer when deer trails and tracks were 2-5 times more abundant
in aspen than in northern hardwood habitat types. The high use of aspen by deer,
therefore, leads to high levels of hunting by wolves in aspen areas and the resulting
importance of aspen in wolf territories.

County owned lands were more common in pack territories than in unused
areas, which supports the findings of Mladenoff et al. (1995) in northern Wisconsin.
They found that 39.3% of pack territories and 8.4% of unused areas were county
owned and that areas under county ownership contributed highly to favorable wolf
habitat rankings. The majority of county owned lands in the H53 SA are managed as
commercial forest. This leads to the maintenance of early succesional forests on large
portions of county owned lands. These early serial stage forests provide high quality

habitat for deer, which likely increases their attractiveness to wolves. County

44

ownership of an area generally precludes human development of the area, which also
likely increases the habitat quality of county owned land to wolves.

In contrast to county owned land, lands under miscellaneous private
ownership comprised lower proportions of wolf pack territories than unused areas.
This finding also supports Maldenoff et al. (1995) who reported 20.1% of pack
territories and 75.0% of unused areas were privately owned. Privately owned land
was highly correlated with road density and human population density by Mladenoff
et al. (1995). It, therefore, seems likely that the comparatively low use of privately
owned land is due to avoidance of human activity by wolves.

Landscape Level Habitat Suitability Model

The logistic based GIS model predicting suitable wolf habitat indicates that
there is still suitable wolf habitat that is unoccupied. This unoccupied habitat is
primarily in northeastern Wisconsin and was also identified by Mladenoff et al.
(1995). This indicates that there is the potential for the wolf population in the state to
continue to expand toward the target population of 350 individuals (WIDNR 1998).

The habitat suitability predictions created using this model and the pixel
aggregation procedure described by Roloff and Haufler (1997) provide managers with
a tool that allows the prediction of wolf occurrence based not only on habitat quality
but also on habitat juxtaposition. This predictive capability will prove useful not only
in managing habitat, but also in avoiding human/wolf conflict, and potentially
targeting areas for wolf harvest when a public harvest is conducted.

The importance of including the juxtaposition of habitat, as well as habitat

quality when estimating carrying capacity is illustrated by my data and predictions of

45

suitable wolf habitat in PWR. Average observed pack territory size in the H53SA
was 119 kmz. The logistic based model that I developed classified roughly 23,000
km2 of favorable habitat in PWR. This means that based solely on predicted habitat
quality, there is enough favorable habitat in PWR to support 193 packs. When the
juxtaposition of habitat (and the territorial nature of wolves) is included in this
calculation by incorporating the pixel aggregation procedure, only 66 areas where
classified as being suitable to support pack territories. This type of a difference can
have significant implications for population management, especially when estimating
carrying capacity for threatened or endangered species.
SUMMARY AND MANAGEMENT IMPLICATIONS

Evidence that at current levels of human tolerance/legal protection wolves are
successfully colonizing areas that are more human dominated and which have higher
road densities is growing (Fuller et al. 1992, Mech 1995, Merrill 2000, this study).
Wolf populations in these non-wilderness or semi-wild areas appear to be stable or
growing. This suggests that some areas that would have in the past been considered
as unsuitable wolf habitat (based primarily on road density) are in fact suitable. This
does not mean that road density is no longer a useful measure of wolf habitat
suitability. There is undoubtedly some level of road density (and the resultant level
of human activity/contact) above which wolves will generally not attempt to establish
territories or will not be able establish sustainable territories. With the wolf
population in the Great Lakes region continuing to increase, we likely have yet to
discover this upper limit of road density. Lacking a definitive upper road density

value for wolf colonization, I suggest that resource managers consider areas with less

46

1.5km\km2 as potential suitable wolf habitat given current levels of human
tolerance/legal protection. This figure is based on the highest mean road density
observed for a year of establishment group in this study (1.37 km/ km2 for packs
established in 1999). Packs established in 1999 have consistently produced pups and
appear viable 3 years after establishment. I saw no evidence that the packs
established in 1999 were encountering conditions which would likely lead to the
territories failing to persist or act as population sinks. While these packs were
monitored for less than two years after they were established, all 3 produced pups
both years and one of the three included the largest number of individuals of any pack
in the study area at the conclusion of the study. After only two years of monitoring,
the perception that these packs are experiencing no greater stress due to road density
than any other packs in the study area is preliminary. Nonetheless, I observed no
evidence that the road density within these packs was the maximum level tolerated by
wolves to establish sustainable territories.

Setting an upper road density limit for potential wolf habitat will aid managers
in two ways. It will help to avoid expending resources on wolf management in areas
that are likely not quality habitat. It will also aid in delineating areas of high potential
risk of human/wolf conflict. For example, if wolves move into an area with a road
density above the upper limit and a high likelihood of human contact and human/wolf
conflict, this area would be a good candidate for wolf population control if and when
it needs to be implemented. Since the WDNR wolf population target was set at 350
at least in part in an attempt to maintain the state’s wolf population at or below human

social carrying capacity, the need to control the population seems likely at some point

47

in the future. If this becomes necessary, being able to select areas to target for wolf
removal will be necessary. Areas, which wolves have colonized, and which have
higher than acceptable road densities seem like good regions to target for population
control measures, whether they are conducted by resource management agencies or
are some type of a public harvest.

Forested wetlands are an important component of wolf territories in the
H53SA. This cover type comprised larger portions of pack territories than unused
areas and was also the most selected cover type by wolves at the landscape and
within-territory level (Chapter 3, this document). This selection is probably due to
the relatively low levels of human activity in most forested wetland areas and the
importance of forested wetland to white tailed deer as winter yarding areas. Forested
wetlands are areas that managers should consider protecting if they are attempting to
maintain high quality wolf habitat.

Aspen was also highly selected by wolves at the landscape and within-
territory levels (Chapter 3, this document) as well as comprising higher proportions of
pack territories than unused areas. This indicates that aspen also has value to wolves
(likely due to its importance to white-tailed deer) and management for high quality
wolf habitat should include maintaining aspen as a component of the area. This is
especially important considering the maturation of the state’s aspen stands since the
mid-19803. The percentage of forested stands in the state classified as aspen-birch
dropped from 67% in 1983 to 57% in 1996 (Schmidt 1996). Thus, by the late 19905

over 40% of the state’s aspen was in stands that were dominated by other forest types

48

and were in the final stages of converting to these other forest types due to

succession.

49

LITERATURE CITED

Aebischer, N. J ., P. A. Robertson, and E. Kenward. 1993. Compositional analysis of
habitat use from animal radio-tracking data. Ecology 74:13] 3-1325.

Ballard, W. B., and J. R. Dau. 1983. Characteristics of Gray Wolf, (Canis lupus),
den and rendezvous sites in south-central Alaska. Canadian Field Naturalist
97:299-302.

Blouch, R. I. 1984. Northern Great Lakes and Ontario forests. Pages 391-410 in L.
K. Halls, ed. White-tailed deer ecology and management. Wildlife
Management Institute. Stack Pole Books. Harrisburg Pennsylvania, USA.

Byelich, J. D., J. L. Cook, and R. I. Blouch. 1972. Management for deer. Pages
120-125 In Aspen: symposium proceedings. General Technical Report, NC-l.

Dixon, K. R., and J. A. Chapman. 1980. Harmonic mean measure of animal activity
areas. Ecology 61 : 1040-1044.

Elkie, P. C., R. S. Rempel, and A. P. Carr. 1999. Patch analyst’s user manual, 3 tool
for quantifying landscape structure. NWST technical manual TM-002.

Fuller, T. K. 1989. Population dynamics of wolves in north-central Minnesota.
Wildlife Monographs 105.

_, and W. J. Snow. 1988. Estimating winter wolf densities using radiotelemetry
data. Wildlife Society Bulletin 16:367-370.

_, W. E. Berg, G. I. Radde, M. S. Lenarz, and G. B. Joselyn. 1992. A history
and current estimate of wolf numbers and distribution in Minnesota. Wildlife

Society Bulletin 20:42-55.

50

Haebeck, J. R. and, J. T. Curtis. 1959. Forest cover and deer population densities in
early northern Wisconsin. Transactions Wisconsin Academy Sciences, Arts,
and Letters. 48:49-56.

Hooge, P. N., and B. Eichenlaub. 1997. Animal movement extension to Arcview.
Version 1.1 Alaska Biological Science Center, US. Geological Survey,
Anchorage, Alaska.

Jensen, W. F., T. K. Fuller, and W. L. Roinson. 1986. Wolf (Cam's lupus)
distribution on the Ontario-Michigan border near Sault Ste. Marie.
Canadian Field Naturalist 100:363-366.

Kohn, B. E., J. Frair, D. Unger, T. Gehring, D. Shelley, E. Anderson, and J.
Ashbrenner. 1997. Impacts of highway development on northwestern
Wisconsin timber wolves, preliminary findings-May, 1992 through
November, 1996. Wisconsin Department of Natural Resources. Madison,
Wisconsin, USA.

Lenth, R.V. 1981. On finding the source of a signal. Technometrics 23: 149-1 54.

Mandelbroldt, B. B. 1997. The fractal geometry of nature. W. H. Freeman, New
York, New York, USA.

Mandemack, B. A. 1983. Food habits of Wisconsin timber wolves. Thesis,
University of Wisconsin-Eau Claire. Eau Claire, Wisconsin, USA.

Manly, B. F ., L. L. McDonald, and D. L. Thomas. 1993. Resource selection by
animals, statistical design and analysis for field studies. Chapman and Hall,

London, United Kingdom.

51

McCaffrey, K. R. and W. A. Creed. 1969. Significance of forest openings to deer in
northern Wisconsin. Wisconsin Department of Natural Resources Technical
Bulletin 44. 104pp.

McCaffrey, K. R., J. Tranetzki, and J. Piechura. 1974. Summer foods of deer in
northern Wisconsin. Journal of Wildlife Management 38:215-219.

Mech, L. D. 1970. The wolf: the ecology and behavior of an endangered species.
Natural History Press, Garden City, New York, USA.

_. 1995. The challenge and opportunity of recovering wolf populations.

Conservation Biology 9:270-278.

Merril, S. B. 2000. Road densities and gray wolf, Canis lupus, habitat suitability: an
exception. Canadian Field Naturalist 114:312-313.

Mladenoff, D. J., T. A. Sickley, R. G. Haight, and A. P. Wydeven. 1995. A regional
landscape analysis and prediction of favorable gray wolf habitat in the
northern Great Lakes region. Conservation Biology 92279-294.

_, and T. A. Sickley. 1998. Assesing potential gray wolf restoration in the

northeastern United States: a spatial prediction of favorable habitat and potential
population levels. Journal of Wildlife Management 62: 1-10.

_, T. A. Sickley, and A. P. Wydeven. 1999. Predicting gray wolf landscape

recolonization: logistic regression models vs. new field data. Ecological

Applications 9:37-34.

52

Reynolds, T. D., and J. W. Laundre. 1990. Time intervals for estimating
pronghom and coyote home ranges and daily movements. Journal of Wildlife
Management 54:316-322.

Roloff, G. J ., and J. B. Haufler. 1997. Establishing population viability planning
objectives based on habitat potentials. Wildlife Society Bulletin 25: 895-904.

Schmidt, T. L. 1996. Wisconsin Forest Statistics, 1996. North Central Forest
Experiment Station Resource Bulletin NC-183. Saint Paul Minnesota, USA.

Schultz, R. N., A. P. Wydeven, and R. A. Megown. 1996. Injury levels with
five types of leg-hold traps in Wisconsin. Pages 38-39 in 14‘h Midwest
F urbearer Workshop, Ironwood, Michigan, USA.

Seaman, D. E., J. J. Millspaugh, B. J. Kemohan, G. C. Brundige, K. J. Raedeke, and
R. A. Gitzen. 1999. Effects of sample size on kernel home range estimates.
Journal of Wildlife Management 63:739-747.

Shannon, C. E., and W. Weaver. 1949. The mathematical theory of communication.
University of Illinois Press, Urbana, Illinois, USA.

Shelly, D. P., and E. M. Anderson. 1995. Final report: impacts of US Highway
53 expansion on timber wolves—baseline data. University of Wisconsin
—Stevens Point, Stevens Point, Wisconsin, USA.

Thiel, R. P. 1985. Relationship between road densities and wolf habitat suitability in

Wisconsin. American Midland Naturalist 133:404-407.
Verrne, L J. 1965. Swamp conifer deeryards in northern Michigan: their ecology and

management. Journal of Forestry 63:523-529.

53

Vila, C., J. Castroviejo, and V. Urios. 1993. The Iberian wolf in Spain. Pages 104-
109 in C. Promberger and W. Schroeder, Editors. Wolves in Europe: status
and perspectives. Munich Wildlife Society, Ettal, Germany.

White, G. C. and R. A. Garrott. 1990. Analysis of wildlife radio-tracking data.
Academic Press, San Diego, California, USA.

Wisconsin Department of Natural Resources. 1999. Wisconsin wolf management
plan. Madison, Wisconsin, USA.

Worton, B. J. 1987. A review of models of home range for animal movement.
Ecological Modeling. 38:277—298.

_. 1989. Kernel methods for estimating the utilization distribution in home-range
studies. Ecology 70: 164-168.

Wydeven, A. P. and N. M. Cervantes. 1997. Recovery of the timber wolf in
Wisconsin, performance report, 1 July 1996- 30 June 1997. Wisconsin
Endangered Species Report #1 17. Madison, Wisconsin, USA.

Zar, J. H. 1999. Biostatistical analysis. Prentice Hall, Upper Saddle River, New

Jersey, USA.

54

CHAPTER 3
COVER TYPE AND LAND OWNERSHIP SELECTION OF
RECOLONIZING GRAY WOLVES IN NORTHWEST WISCONSIN
INTRODUCTION

Past research on wolves has focused on relating resource selection to features
including land cover, road density, prey base, and land ownership patterns (Mladenoff et
al. 1995) at the landscape level. Likely due the wide range of habitats originally
occupied by wolves and their overall generalist nature (Mech 1995), little research has
been conducted on cover type selection by wolves within territories. Although the habitat
generalist nature of wolves allows them to successfully utilize various cover types,
differences in cover type composition across a landscape do influence where wolves
establish territories (Mladenoff et al. 1995). Therefore, information about cover type
selection is imponant for this generalist species more as a tool in predicting where wolves
will occur in the future than as a tool in protecting or providing suitable habitat. This is
especially important in semi-wild landscapes due to the high potential for conflicts with
humans as wolves continue to expand their range. An increased ability to predict areas of
future occupation within these areas allows for a more accurate assessment of the
potential for human/wolf conflicts in semi-wild landscapes.

Another measurement of habitat quality that needs more investigation in semi-
wild landscapes is road density. Since the mid-19805 road density has been the most
commonly used measure of habitat quality for wolves, with higher road densities
associated with lower quality habitat (Mech 1995). It is generally accepted (Mladenoff et

al. 1995, Mladenoff and Sickley 1998) that while wolf mortalities do occur on and along

55

roads, increased human access into wolf occupied habitat and the resulting increased
direct killing of wolves by humans and potential for disease and parasite occurrence
(Mech and Goyal 1993), rather than vehicle caused mortality is the mechanism
responsible for the negative influence of increasing road density on wolf habitat
suitability. Wolf use of roads and trails as travel and hunting corridors has also been
documented (Thompson 1952, Mech 1970, Gehring 1995) indicating that roads may have
value to wolves. Based on past research there are still questions regarding whether roads
are actually being avoided or whether they have value to wolves and the avoidance
observed is actually of human activity promoted by roads.

This study examined selection of 14 land cover and 5 ownership categories by
gray wolves. It also examined the use of areas in proximity to roads by wolves during
daylight hours. The goal of these analyses was three-fold. My first objective was to
examine and quantify year-long selection of land cover and ownership categories by
recolonizing gray wolves in the H53 SA at the landscape (placement of territories within
the H53SA) and within-territory levels. The second objective was to identify land cover
categories that were being highly selected by wolves. These cover types are presumably
of importance to wolves and identifying them should allow resource managers to identify
key wolf habitat areas. Since wolves are using these areas heavily, identifying them will
also identify areas that have high probabilities of human/wolf conflict. A third objective
was to examine whether wolves use areas in proximity to roads in proportion to
availability, indicating selection or avoidance. This information can prove useful in
further understanding the actual mechanism that drives the documented negative effects

of road density on wolf habitat quality (Mladenoff et al. 1995, 1999).

56

METHODS
Capture, Handling, and Radiotelemetry

Wolves were trapped by WDNR personnel using modified #14 Newhouse leg-
hold traps (Schultz et al. 1996) during May-August in 1995-2000. Captured wolves were
immobilized using a mixture of ketamine/xylazine administered with a jabstick. Any
trap-related injuries were treated, blood samples were taken, an injection of antibiotics
was administered, and animals were weighed. Animals > 15kg were also fitted with
radio collars (Telonics Inc., Mesa, Arizona) and uniquely numbered ear tags. Yohimbine
was administered as a reversal agent for the xylazine and the animals were monitored
until they could move away from the trap site under their own power.

Radio-collared animals were located 1-5 times per week during 1995-2001using
telemetry from fixed-wing aircraft (Ballard and Dau 1983, Fuller 1989) or a vehicle-
mounted 5—element Yagi directional antenna until radio failure, death of the animal, or a
permanent movement out of the study area. The majority (>95%) of locations were
collected during daylight hours and a minimum of 3 bearings were collected for each
location. Triangulations for locations collected on the ground where solved using a Lenth
estimator (Lenth 1981) in the program Locate ll (Pacer, Turvo Nova Scotia, Canada).
Ground locations were collected during the course of data collection for this study and
two other graduate projects conducted within the H53SA during 1995-1997. The
accuracy of telemetry locations I collected was tested as described by White and Garrot
(1990). In addition to average error ellipse (3.8 ha), I also calculated the average error in
straight line distance from the actual location of the test transmitters to the estimated

locations (123 m). No estimates of telemetry accuracy were reported for the locations

57

collected in 1995-1997. Since Shelly and Anderson (1995) found no significant
difference between the accuracy of aerial and ground locations in the study area using the
same telemetry techniques and equipment, locations were pooled for territory estimation
and resource selection analyses. Locations obtained for an individual wolf were at least 4
hours apart to maintain independence (Reynolds and Laundre 1990). Date, time,
Universal Transverse Mercator (UTM) coordinates, ear tag number, and pack were
recorded for each telemetry location. Animals were classified as belonging to individual
packs based on trapping location and movement patterns.
Territory Boundary Estimation

Pack territory boundaries were estimated using a fixed kernel method (Worton
1989) based on an 80% isopleth using the animal movement analysis extension (Hooge
and Eichenlaub 1997) in ArcView (Environmental Systems Research Institute, Redlands,
CA). A minimum of 30 locations (Fuller and Snow 1988, Seaman et al. 1999) was used
in estimating territory boundaries. Only territories of packs which had been monitored
for at least 6 months and which had produced pups were included in the analysis.
Locations from all radio-marked animals within a pack during the course of the study
were used in territory boundary estimation.
Determination of Territory Land Cover and Land Ownership Composition

Territory boundaries were overlaid on a 30 m resolution LANDSAT land cover
data layer created in 1992 from Thematic Mapper imagery and a land ownership data
layer provided by D. Mladenoff and T. Sickley (University of Wisconsin Dept. of
Forestry) in ArcView. The land cover layer was provided by the WDNR and categories

had been classified within the WISCLAND land cover scheme

58

(http://www.dnr.state.wi.us/org/at/et/geo/data/wlc.htm). The original 28 land cover
categories present within the H53 SA in this layer were consolidated into 14 categories
(Appendix A) to lessen the occurrence of cover categories with small values for use and
availability. This consolidation was conducted based on the following 3 criteria. Any
cover type that represented less than 0.1% of both territories and the total study area and
which was present in less than 4 pack territories was excluded from analysis. Any cover
type category which, based on traditional paradigms of wolf ecology would not be used
by wolves (i.e., open water and urban) was excluded from the analysis. Similar cover
types that represented little variability in resources to wolves were combined (i.e., red oak
Quercus rubra and northern pin oak Quercus ellipsoidalis were combined).

The area of each territory and proportion of the total area of the territory within
each land cover and ownership category were calculated from the consolidated land cover
and ownership categories. The resulting land cover and ownership proportions were used
in a Compositional Analysis (CA) (Aebischer et al. 1993) to analyze year-long landscape
level and within-territory land cover and ownership selection. The calculations for CA
were conducted using ReSelect software (http://ces.iisc.emet.in/hpg/envis/resdocl 120.html).
CA was also used to analyze within-territory land cover selection during “summer” and
“winter” seasonal periods. CA involves a use vs. availability approach to delineate and
quantify selection, while avoiding the problem of non-independence of proportions
calculated for use and availability in many habitat selection techniques (Aebischer et al.
1993).

At the landscape level of analysis, the proportion of the study area within each

land cover or ownership category represented availability while the proportion of each

59

land cover or ownership category within territories represented use. At the within-
territory level of analysis, the proportion of a territory within each land cover or
ownership category represented availability while use was represented by the proportion
of the total number of wolf telemetry locations within each land cover or ownership
category. Use for the within-territory analyses was calculated for entire year, as well as
“summer” (15 April-14 September) and “winter” (15 September-14 April) seasonal
periods to facilitate analysis of potential differences in selection due to seasonal changes
in behavior. These periods were defined by Fuller (1989) corresponding to changes in
pack cohesiveness related to denning in northcentral Minnesota and corresponded well
with observed wolf behavior in the H53 SA.

My use of CA varied in one way from that described by Aebischer et al. (1993).
They suggested using individual radio-marked animals as the sample unit when
conducting CA. Due to the territorial nature of wolves and the resultant restrictions on
movement, independence of telemetry locations, and habitat use, I used a territory (all
marked animals in a pack) as the experimental unit for these analyses. Alpha levels for
both overall tests of selection and pair-wise comparisons of land cover categories within
compositional analysis were 0.10.

The examination of wolf use of areas in proximity to roads was based on
comparisons of used areas to their relative availability. To accomplish this, roads within
territories were buffered at arbitrarily set distances of 25, 50, 100, and 250 m. The
proportions of the area of territories within these road buffers were calculated. The mean
proportion of territories within these buffers was compared to the mean proportion of the

number of telemetry locations falling within buffers in each territory using a Mann-

60

Whitney U test (Zar 1999) (a = 0.10) to provide an indication of the selection or
avoidance of areas near roads. There is an extensive system of trails in the H53 SA used
for both forestry harvest and recreation. This trail system allows access to otherwise
roadless areas and is used heavily during the summer by all terrain vehicle riders and
during the winter by snowmobile users. Since increasing human access to wolf habitat is
generally regarded as a negative influence for wolf sustainability, I included this trail
system in my definition of roads for this analysis. With this addition, all features
identified as any class of road or trail was included in the analysis.
RESULTS
Selection of Land Cover Category

Five-thousand one-hundred and two telemetry locations, from 40 wolves in 14
packs (Figure 3.1), were analyzed. CA indicated that there was differential selection of
land cover types occurring at both the landscape (p < 0.001) and within-territory (p <
0.001) levels when telemetry locations (use) were considered as one year-long set of data
with no allowance for seasonal changes. The three cover types that comprised the largest
proportions of the study area were aspen, mixed/other broad-leaved deciduous, and
lowland shrub (21.5%, 14.1%, and 10.8% of the study area respectively) (Table 3.1). Not
only were these the three most common cover types in the study area, but they were also
three of the top four most selected cover types at the landscape level (Table 3.2). Jack
pine (Pinus banksiana), shrubland, and agricultural were the three least selected of the
three least common cover types within territories (1.8% and 1.3%) (Table 3.2). While
there were some differences between landscape and within-territory selection rankings,

forested wetland and aspen were two of the top three most selected cover types at both

61

 

Fig. 3.1 Wolf pack territories in the Highway 53 Study Area in northwest
Wisconsin and adjacent Minnesota 1995-2001

62

._o>o_ 38.6.25 05 8 mix—a5 fiaoEmon—Eoo E @2333: he 3:308 05 we? >3ng on: was none 555 «03 been 33 25 mo :5ch o
._o>o_ boqufiE—EB 05 «a mix—man 3.85838 5 3:36:63 me 8:32: 05 026 83 :

._o>u_ unaumv§_ 05 «a $9288 3838988 E 3: we 83.38 05 £3 25 850 was :23 5 6386.52 :6 £55 83 05 mo Boobs m
._o>o_ b95336;

2: E max—mam 382.6383 S 3.: mo 0.8808 05 83 095 850 98— noao E 332 82302 32:28 me 8956 05 .3 Beach <

 

63

 

 

ed N: mm mud n; _m~3_=otw<
5m cm; 3‘ 56 h; EESPEm
6..., m: 66 86 2 6&6 cos
3 666 3 36 a6 2&2
m; omd m4 Ned m4 many—£88 8:6onva
ad cud 0N No; 9N #60
ad mwd Wm mud mlm wad—$80
w.~ 36 NN amd a; 0:3 3M
<6 cm; Ce 3; cs msoeomioohsosfloov 3622
o.m on; Wm amd a; 3888 we? EmmEEm
12 :1” 0.2 NNM 03 macs—zone 328—6695 558352
w.o_ on; 3; Sam 6.: page. wan—Bod
m. _N wfim Yam SUN mAN ao%<
m.w $1 mtg mud oAN wad—83 Bump—om
mm m mm m howuaao 550 v53
0 22:89:00 m :oEmomEoo < mergers“ :53, 090
686 35% .x. beach: e\e 850 E 65382

30822 X.

 

.fiooméoa 5808:,» 603.30:
E seven—om 09$ :33 98— .«o mam—«:6 3536888 6 5 com: 893 e260 3 05 £53» 28382 @0822 mo owficooeeq 05

98 mew—858 x03 :95 £53, 98 602 33m mm .3235 05 E 536888 Eowofio 560 was “coupon :32 ._.m 2an

.0852 .«o 0m: 82888.: no :8 :E0>o .8 Good v moodm "an m $~ N m

.0952 ho a: 585.8: «o :2 :83 he :85 v a 90% ":0 EN ~ g

 

 

o v E _08::otw<

E I C cam—nanm

I S 2 088 0.03.

n m : x00

Wm n 3 0:8 By—

5 m o gobtcofimsoscmofiv @0032

m .o_ m fl w van—$80

m6 m_ N. 0302

S E o 30.82800 0050:0822

m6 w m 320008 00>» 80w808m

We a v 0.3032020 ©0200—6005 0050:0082

Wm m m c0am<

v o m 98% van—Bog

~ ~ _ cam—00>» BEBE
0w80>a 00—88 <88 _0>0_

wcflcmm DBEBéMSS/ 0908984 >5ng 80>oo @804

 

Aoowéoi 880835 60355: E «02 38m mm @335 05 E 8203 haw 8m 8% 0m:
mac—-30» so @003 $0380 mo m_0>0_ £8 8.“ 888 8500.8 wcmnaoqfltoo new 80:00.8 bouwcgéfiés Ea _0>0_ 0903882
mo mix—mam 38206388 05 8 00m: m0tow0§0 550 was. DZ<AUm§> @003 80308 $8805 0820:8800 .m.m 030%

64

levels (Table 3.2).

The analysis of seasonal land cover selection also indicated that cover type use
was not random (p < 0.001 for both “winter” and “summer” periods). Several differences
in selection rankings between “summer” and “winter” periods (i.e., oak, lowland shrub)
as well as between seasonal periods and the entire year (oak) were also detected (Table
3.3, Appendix D, Appendix E). Forested wetland and aspen, which where the two most
selected cover types in the year-long analysis, where also in the top 4 in both “summer”
and “winter” (Table 3.3). In contrast, oak was the third most selected cover type in the
year-long analysis and the most selected type in the “summer” analysis, but dropped to
twelfth out of 14 in the “winter” period. The selection of lowland shrub was also
different among “winter”, “summer” and the year-long period. It was the second most
selected cover type during “winter”, but the eighth most selected during “summer” and
the sixth most selected during the entire year. Grassland and maple where ranked among
the 4 least selected cover categories in all 3 seasonal periods.

Selection of Land Ownership Category

Compositional analysis indicated that the use of land ownership classifications was not
different from random at either the landscape (p =O.1 18) or within-territory (p =O.508)
levels (Table 3.4). The H53SA was primarily privately (61%) or county (21%) owned.
While the majority of the area within territories was also within these two categories.
county-owned lands (53%) were more prevalent than privately owned lands (28%). A

similar trend was evident with telemetry locations; 61% of locations occurred on county-

owned land and 21% occurred on privately owned-land (Table 3.4).

65

Table 3.3. Compositional analysis selection rankings of consolidated WISCLAND based
land cover types for year-long and seasonal periods at the within-territory level of

analysis for gray wolves in the Highway 53 Study Area in northwest Wisconsin 1995-
2001.

 

 

Year-long Winter Summer
Land cover category rank rankA rankB
Forested wetland l 1 2
Aspen 2 3 4
Oak 3 12 1
Agricultural 4 5 3
Mixed deciduous/coniferous 5 9 6
Lowland shrub 6 2 8
Red pine 7 14 10
Emergent wet meadow 8 7 7
Mixed/other broad-leaved deciduous 9 4 5
Jack pine 10 8 l l
Shrubland 1 1 6 14
Maple 12 13 9
Grassland 13 10 13
Mixed/other coniferous l4 1 1 12

 

A Winter = 15 September- 14 April.
B Summer = 15 April-l4 September.

66

Table 3.4 Land ownership category composition within the Highway 53 Study Area and
wolf pack territories and the percent of telemetry locations within each category in
northwest Wisconsin, 1995-2001 .

 

% of total telemetry

 

 

% study area % territory locations in land

composition composition cover category
Ownership Category 5? A7 SE X SE
County 21.6 53.6 8.10 61.2 8.61
National Park 0.5 0.3 0.15 0.9 0.46
Private 61.6 27.2 4.71 21.5 4.91
State 11.7 10.5 4.67 12.3 5.96
Industrial Forest 5.0 8.4 3.63 3.7 2.01

 

67

Use of Buffers Around Roads

Approximately 4 -33% of the area within wolf pack territories and 2—28% of
telemetry locations were within the buffers around roads and trails (Table 3.5). The
percent of telemetry locations within buffers of 25, 50, 100, and 250 m around roads and
trails was lower (p = 0.002, 0.067, 0.040, 0.032 respectively) than the percent of the total
area of territories within these buffers (Table 3.5).
DISCUSSION

It is generally accepted that wolves are habitat generalists (Mech 1970), originally
occupying every habitat in the northern hemisphere where large ungulates were present
(Mech 1995). It has also been documented that road density (as a measure of increased
human access and activity) is a useful measure of wolf habitat suitability (Mladenoff et
al. 1995, 1999) and that even in legally protected populations human activity is generally
the leading cause of death for adult wolves (Fuller 1989, Wydeven et al. 1995). In areas
where human tolerance for wolves is high and human caused mortality is low, the density
of large ungulate prey is the primary limiting factor on wolf populations (Fuller et al.
1992). Therefore, it seems reasonable that any analysis of habitat (cover type) use and
preference must consider both the potential impacts of cover type on the level of human
activity and on major prey species usage. Discussion of the selection or avoidance of
individual cover types by wolves will relate how a given cover type affects either or both
of these factors and wolf use of the cover type.
Selection of Land Cover Category

Wolf pack territories have been shown to be positively correlated with

68

Table 3.5. Percent of the Highway 53 Study Area and mean percent of territories within
buffers of varying distances around roads and trails compared to percent of telemetry
locations within these same buffers in northwest Wisconsin 1995-2001.

 

 

 

% of territories % of wolf telemetry
within road buffers locations within

Road Buffer % of H53SA within (Available) road buffers (Used)

Distance road buffer 217 SE )7 SE P"1
25m 4.95 3.69 0.30 2.23 0.52 0.002
50m 9.72 6.74 0.76 5.38 0.80 0.067

100m 18.55 13.08 1.45 10.02 1.53 0.040
250m 41.15 32.63 2.67 27.52 4.49 0.032

 

" Mean available values for all road buffer categories were greater than mean used values for all buffer
categories when analyzed in a Mann-Whitney U test V = 0.10, Ho: Available > Used.

69

forested wetland in the northern Great Lakes with territories containing higher
proportions of forested wetland than areas unoccupied by wolves (Mladenoff et al. 1995).
The results of the CA that I conducted confirm this finding. Forested wetland was the
most selected cover type at both the landscape and within-territory level when examined
with no allowance for seasonal differences. It was also the most selected cover type
during the winter seasonal period and the second most selected during the summer
seasonal period. This affinity for forested wetland areas is likely as much a product of
the lack of human access and activity in these areas as it is to the actual vegetation
composition of the area.

Road density has been recognized as an important measurement of wolf habitat
suitability since the mid 19805 with higher road density indicating lower wolf habitat
suitability (Thiel 1985, Jensen et al. 1986, Mech et al. 1988, Mladenoff et al. 1995). It
has been stated “roads per se are not avoided by wolves but serve as indicators of human
contact and thus the likelihood of deliberate or accidental human caused mortality”
(Mladenoff et a1. 1995, Mladenoff and Sickley 1998). It has been shown that wolves may
use roads and trails as travel corridors under certain conditions (Pimlott et al. 1969,
Thurber et al. 1994, Gehring 1995). While this indicates that roads may have value to
wolves, Thurber et al. (1994) found that wolves avoided roads with relatively high human
use, opting to use closed roads as travel corridors. Thus, human activity and not road
density, per se, is likely what wolves are avoiding. While forested wetland areas within
the H53 SA had higher densities of roads and trails (2.48 km/kmz) than the entire study
area (0.99 km/kmz), actual human activity in these areas is possibly lower than in areas

dominated by other cover types due to the difficulties of movement away from roads.

70

Forested wetland was also consistently highly selected across seasonal periods.
This was probably due to a combination of low human activity in these areas and
secondary factors including the importance of this cover type to prey species especially
white-tailed deer (Odocoileus virginianus). In the H53SA, white-tailed deer are the
preferred prey item throughout the year (Mandemack 1983). Lowland conifer stands
(classified within the forested wetland category in this study) are an important component
of deer habitat during severe winters, providing thermal cover and lower snow depths
than surrounding areas (Verme 1965, Blouch 1984). It has also been suggested that
yarding behavior in deer may be an anti-predator strategy (Nelson and Mech 1981)
evolved to compensate for the increased vulnerability to wolves during the winter
months. In either case, the increased use of these swamp conifer areas by white-tailed
deer may lead to increased use by wolves and an increased selection of the forested
wetland cover type.

Aspen was highly selected by wolves at both the landscape and within-territory
levels. It was also highly selected across all seasonal periods. The high use of aspen
areas is also likely tied to the white-tailed deer population. Aspen has been identified as
the northern Great Lakes’ leading deer producing forest type (Byelich et al. 1972). It is
also a preferred habitat of Wisconsin deer (Habeck and Curtis 1959) and was identified as
dominating the diet of deer in northern Wisconsin during the early and late summer by
McCaffrey et al. (1974). This preference for aspen as a food source likely leads to the
high level of deer use observed by McCaffrey and Creed (1969) during the summer when

deer trails and tracks were 2-5 times more abundant in aspen than in northern hardwood

71

habitat types. The high use of aspen by deer likely leads to high levels of hunting by
wolves in aspen areas and the resulting high cover type selection ranking.

The one cover type that showed a large difference in selection ranking between
the landscape and within-territory levels of selection was agricultural. At the landscape
level agricultural lands were the least selected of all cover types. In contrast, agricultural
lands were the fourth most highly selected within-territories. The avoidance of
agricultural areas at the landscape level (territory placement) is consistent with the
findings of Mladenoff et al. (1995, 1999). They found that wolf pack territories
contained lower proportions of agricultural land than corresponding unoccupied areas.
This avoidance is possibly due to the relatively high levels of human activity in
agricultural areas. The higher selection of agricultural areas within territories is likely
related to prey activity patterns. Agriculture in the H53SA is generally practiced on
relatively small parcels of land (X’= 3.7 ha). It also tends to be interspersed with forested
lands that are frequently harvested. This mixture often leads to high populations of
preferred prey species, including white-tailed deer. Agricultural practices maintain edge
and open areas providing favorable wildlife habitat in close proximity to cover for forest
wildlife. While wolves tend to avoid areas with high proportions of agricultural land
when establishing territories, once a territory is established, the relatively small
proportions of agricultural lands within the territory are used presumably due to high use
of these areas by prey species.

There were two cover type categories that showed large differences in selection
rankings between “summer” and “winter” periods. The biggest difference in selection

rankings between seasonal periods was observed for oak. It was the highest ranked cover

72

type during “summer” and the twelfth most selected during “winter”. The low ranking of
oak during the “winter” is likely related to the relatively low use of these areas by deer
during the winter months. Areas of oak provide little thermal cover during the winter in
comparison to the conifer-dominated lowland areas where deer tend to yard. Non-conifer
areas, including areas of oak also tend to have deeper snow depths than areas used as
deeryards due to a much more open overstory during the winter (Verme 1965, Ozoga
1968, Moen 1976). These factors lead to low quality winter cover in areas dominated by
oak and deer avoid areas of poor cover during the winter even if these areas provide
plentiful browse (Verme 1965). The relatively low use of oak areas by deer during the
winter (i.e. lower prey densities for wolves) likely leads to corresponding lower use by
wolves. The movement of deer out of deeryards and back into more upland areas in
spring likely raises the prey density in oak areas during the summer. The use of oak areas
by deer may be related to mast production during the later portions of the “summer”
period (as defined in this study). This may influence wolves to use oak areas more
heavily during the summer than during the winter. Another factor that may be partly
responsible for the high selection ranking of oak during the “summer” period is the use of
these areas as denning sites. Oak in the H53 SA tends to be associated with upland areas
having light, more sandy soil types. Wolves in the H53SA dig dens in these lighter soils
(Unger 1998) and may be spending more time in areas of oak during the summer due to
denning behavior.
Selection of Land Ownership Category

The failure of the CA to identify any selection of land ownership category is not

surprising. While some trends relating land ownership to biotic characteristics do seem

73

apparent, (e. g., county owned lands tend to be more forested than privately owned lands)
the correlations between land ownership and other biotic factors including land cover
category are apparently not strong enough to effect selection of ownership category by
wolves. The lack of consistent fine scale links between land ownership category and
biotic factors affecting wolf use limits the usefulness of land ownership category as a
predictor of wolf use of an area.

While CA indicated that the selection of ownership category was not different
from random at either the landscape or within-territory levels of selection, the raw data
did support past findings in the area. Both Mladenoff et al. (1995) and I found that
county owned lands were the most common ownership category within wolf pack
territories and that the percentage of county owned lands was higher in pack territories
than in the entire study area. The reverse was true in both studies with private lands
comprising higher proportions of the study areas than pack territories. These results are
likely due to the low human development of and activity within county owned lands in
comparison to privately owned lands.

Use of Buffers Around Roads

My examination of wolf use of areas around roads was limited by the lack of
telemetry locations during periods of darkness, but it does seem to indicate that wolves
are avoiding areas within 250 m of roads during periods of daylight. Vehicular traffic
and other activities along roads and trails are generally heavier during daylight periods
and the avoidance of these areas during the day is not surprising. While most use of
roads and trails by wolves likely takes place between sunset and sunrise, the avoidance of

roads during the day could potentially limit wolf habitat use and movements if the

74

proportion of the landscape in proximity to roads and trails becomes too high. Over 40%
(418,764 of 1,017,528 ha) of the H53SA is within 250 m of a road or a trail; wolves are
avoiding nearly one half of the area during daylight periods.

While wolves are most active during the night in the summer, they travel and hunt
extensively during the day in the winter (Mech 1970). The avoidance of roads and trails
could particularly constrain wolf activities in areas of high road densities during the
winter. The avoidance of roads during daylight may be one of the underlying reasons
that past authors (Thiel 1985, Jensen et al. 1986, Mech et al. 1988, Mladenoff et al. 1995)
have found an upper limit to road density in areas occupied by wolves. If wolves are
avoiding areas around roads during daylight hours and much of normal wolf activity
during the winter takes place during daylight, there is likely a level of road density that
will not allow enough freedom of movement for wolves to satisfy their life needs during
winter. This may be a specific example of the type of mechanism that leads to the
avoidance of areas of high road density when wolves are establishing territories. The
avoidance of areas around roads and trails does not seem to be limiting the wolf
population in the H53 SA, but it is likely that some level of road density would cause this
avoidance to limit the population.

SUMMARY AND MANAGEMENT IMPLICATIONS

The status of any wolf population is dependant on two primary factors (Fuller
1989). These are: 1) the level of human tolerance of and or legal protection of wolves
and 2) given human tolerance, the density of large ungulate prey species, which in

Wisconsin primarily consist of white-tailed deer. Any discussion of management issues

75

for wolves must focus on these two factors at least as heavily as any more habitat specific
issues.

The gray wolf population in Wisconsin will be reclassified under the ESA within
the foreseeable future. This will lead to lower levels of legal protection and eventually to
the total removal of legal protection. Human tolerance of wolves will become even more
important in maintaining viable wolf populations after legal protection for them is
removed. While the need to maintain a minimum biologically viable population should
set a lower limit for a target population, social carrying capacity must also be considered
when setting population goals.

Current forestry harvest practices are maintaining quality white-tailed deer habitat
throughout much of wolf range within Wisconsin (Mladenoff et al. 1997). Deer prefer
early succesional forests with ample openings (Fuller et al. 1992, Mladenoff and Steams
1993, Fuller 1995). The abundance of this type of habitat maintained by current forest
harvest practices promotes a high density of white-tailed deer, which in turn supports
growing populations of wolves. Thus, where wolves are tolerated by humans and or
legally protected, maintaining current forestry harvesting practices constitutes productive
habitat management for wolves. While maintaining high populations of deer is almost
inarguably good for wolf populations it is not, as discussed by Mladenoff et al. (1997)
good for some other components of the ecosystem including regeneration of vegetation
species highly preferred by deer as browse (Anderson and Loucks 1979, Verme and
Johnson 1986, Alverson et. al. 1988, Tilghman 1989, Pregitzer, 1990).

The apparent connection between the avoidance of humans and the habitat use of

prey species and the cover type selection of wolves observed in this study, support the

76

paradigm that “wolves are habitat generalists that are not habitat specific to a vegetation
or ecosystem type in the sense of sensitivity to a particular habitat structure” (Mladenoff
et al. 1995). Since wolf management does not need to focus on habitat creation or
preservation in a conventional sense, management efforts should focus on maintaining
high levels of public tolerance for wolves and quality habitat for prey species, especially
white-tailed deer.

While recent literature indicates that wolves are successfully colonizing areas
with road densities higher than past perceptions would have regarded as suitable habitat
(Merrill 2000, Chapter 2 this document), my findings on the avoidance of areas within
250m of roads suggest that road density may still be a limiting factor in wolf habitat
suitability. If wolves are avoiding areas near roads, there is likely a level of road density
which will place too much of a potential territory within the area being avoided for the
territory to be viable. If a primary management goal is maintaining quality wolf habitat,
it may become necessary to set upper limits on road density to avoid this situation.
Average road density within pack territories in the H53SA during 1992-2001 was 1.09
km/km2 (Chapter 2, this document). Territories established in 1999 had an average road
density of 1.37 km/km2 (Chapter 2, this document). Packs in territories with both these
levels of road density appear to be viable and therefore, at least in northern Wisconsin,
there seems to be no need to limit road densities to levels below 1.09-1.37 km/kmz.
While I observed no signs that a road density of 1.37 km/km2 was approaching the upper
limit of road density for a viable pack territory, areas with road density greater than this

may prove to be unsuitable.

77

LITERATURE CITED

Aebischer, N. J ., P. A. Robertson, and E. Kenward. 1993. Compositional analysis of
habitat use from animal radio-tracking data. Ecology 74:1313-1325.

Alverson, W. S., D. M. Waller, and S. L. Solheim. 1988. Forests too deer: edge
effect in northern Wisconsin. Conservation Biology 2:348-358.

Anderson, R. C., and O. L. Loucks. 1979. White-tailed deer (Odocoileus
virginianus) influence on structure and composition of Tsuga Canadensis
forests. Journal of Applied Ecology 16:855-861.

Ballard, W. B., and J. R. Dau. 1983. Characteristics of Gray Wolf, (Canis lupus),
den and rendezvous sites in south-central Alaska. Canadian Field Naturalist
97:299-302.

Blouch, R. I. 1984. Northern Great Lakes and Ontario forests. Pages 391-410 in L.
K. Halls, ed. White-tailed deer ecology and management. Wildlife
Management Institute. Stack Pole Books. Harrisburg Pennsylvania. USA.

Byelich, J. D., J. L. Cook, and R. I. Blouch. 1972. Management for deer. Pages 120-
125 in Aspen: symposium proceedings. General Technical Report, NC-l.

Fuller, T. K. 1989. Population dynamics of wolves in north-central Minnesota.
Wildlife Monographs 105. 4 1 pp.

_ 1995. Guidelines for gray wolf management in the northern Great Lakes
region. International Wolf Center technical publication NR 271. Ely,
Minnesota. USA.

_, and W. J. Snow. 1988. Estimating winter wolf densities using radiotelemetry

data. Wildlife Society Bulletin 16:367-370.

78

_, W. E. Berg., G. I. Radde, M. S. Lenarz, and G. B. Joselyn. 1992. A history
and current estimate of wolf numbers and distribution in Minnesota. Wildlife
Society Bulletin 20:42-55.

Gehring, T. M. 1995. Winter wolf movements in northwestern Wisconsin and
east-central Minnesota: a quantitative approach. Thesis, University of
Wisconsin-Stevens Point, Stevens Point, Wisconsin, USA. 131 pp.

Haebeck, J. R. and, J. T. Curtis. 1959. Forest cover and deer population densities in
early northern Wisconsin. Transactions Wisconsin Academy Sciences, Arts,
and Letters. 48:49-56.

Hooge, P. N., and B. Eichenlaub. 1997. Animal movement extension to Arcview.
Version 1.1 Alaska Biological Science Center, US. Geological Survey,
Anchorage, Alaska.

Jensen, W. F., T. K. Fuller, and W. L. Roinson. 1986. Wolf (Cam's lupus)
distribution on the Ontario-Michigan border near Sault Ste. Marie. Canadian Field
Naturalist 100:363-366.

Lenth, R.V. 1981. On finding the source of a signal. Technometrics 23:149-154.

Mandemack, BA. 1983. Food habits of Wisconsin timber wolves. Thesis,
University of Wisconsin-Eau Claire. Eau Claire, Wisconsin, USA.

McCaffrey, K. R. and W. A. Creed. 1969. Significance of forest openings to deer in
northern Wisconsin. Wisconsin Department of Natural Resources Technical
Bulletin 44. 104pp.

McCaffrey, K. R., J. Tranetzki, and J. Piechura. 1974. Summer foods of deer in

northern Wisconsin. Journal of Wildlife Management 38:215-219.

79

Mech, L. D. 1970. The wolf: the ecology and behavior of an endangered species.
Natural History Press, Garden City, New York, USA.

_, and S. M. Goyal. 1993. Canine parvovirus effect on wolf population change
and pup survival. Journal of Wildlife Diseases 29:330-333.

_, S. H. Fritts, G. Radde, amd W. J. Paul. 1988. Wolf distribution in
Minnesota relative to road density. Wildlife Society Bulletin 16285-88.

Merril, S. B. 2000. Road densities and gray wolf, Canis lupus, habitat suitability: an
exception. Canadian Field Naturalist 114:312-313.

Mladenoff, D. J ., and F. Steams. 1993. Eastern hemlock regeneration and deer
browsing in the northern Great Lakes Region: a re-examination and model
simulation. Conservation Biology 7:889-900.

_, T. A. Sickley, R. G. Haight, and A. P. Wydeven. 1995. A regional
landscape analysis and prediction of favorable gray wolf habitat in the
northern Great Lakes region. Conservation Biology 9:279-294.

_, R. G. Haight, T. A. Sickley, and A. P. Wydeven. 1997. Causes and
implications of species restoration in an altered landscape: a spatial projection
of wolf population recovery. Bioscience 47:21-32.

_, and T. A. Sickley. 1998. Assesing potential gray wolf restoration in the
northeastern United States: a spatial prediction of favorable habitat and potential
population levels. Journal of Wildlife Management 62:1-10.

_, T. A. Sickley, and A. P. Wydeven. 1999. Predicting gray wolf landscape
recolonization: logistic regression models vs. new field data. Ecological

Applications 9:37-34.

80

u.
ls"

Moen, A. N. 1976. Energy conservation by white-tailed deer in the winter. Ecology
57:192-198.

Nelson, M. E., and L. D. Mech. 1981. Deer social organization and wolf predation in
Northeastern Minnesota. Wildlife Monographs 77. 53pp.

Ozoga, J. J. 1968. Variations in microclimate in a conifer swamp deeryard in
northern Michigan. Journal of Wildlife Management 32:574-585.

Pimlott, D. H., J. A. Shannon, and G. B. Kolenosky. 1969. The ecology of the
timber wolf in Algonquin Park. Ontario Department of Lands and Forests
Research Report 87. 92pp.

Pregitzer, K., 1990. The ecology of northern white-cedar. Pages 8-14 in Proceedings
of northern white-cedar in Michigan workshop. Sault Sainte Marie, Michigan.

Reynolds, T. D., and J. W. Laundre. 1990. Time intervals for estimating
pronghom and coyote home ranges and daily movements. Journal of Wildlife
Management 54:316-322.

Schultz, R. N., A. P. Wydeven, and R. A. Megown. 1996. Injury levels with
five types of leg-hold traps in Wisconsin. Pages 38-39 in 14lh Midwest
Furbearer Workshop, Ironwood, Michigan, USA.

Seaman, D. B., J. J. Millspaugh, B. J. Kemohan, G. C. Brundige, K. J. Raedeke, and
R. A. Gitzen. 1999. Effects of sample size on kernel home range estimates.
Journal of Wildlife Management 63:739-747.

Shelly, D. P., and E. M. Anderson. 1995. Final report: impacts of US Highway
53 expansion on timber wolves—baseline data. University of Wisconsin

—Stevens Point, Stevens Point, Wisconsin, USA.

81

Thiel, R. P. 1985. Relationship between road densities and wolf habitat suitability in
Wisconsin. American Midland Naturalist 133:404-407.

Thompson, D. Q. 1952. Travel, range, and food habits of timber wolves in
Wisconsin. Journal of Mammalogy 33:429-442.

Thurber, J. M., R. 0. Peterson, T. R. Drummer, and S. A. Thomasma. 1994. Gray
wolf response to refuge boundaries and roads in Alaska. Wildlife Society
Bulletin 22:61-81.

Tilghman, N. G. 1989. Impacts of white-tailed deer on forest regeneration in
northwest Pennsylvania. Journal of Wildlife Management 53:524-532.

Unger, D. E. 1998. A multi-scale analysis of timber wolf den and rendezvous sites in
northwestern Wisconsin and east-central Minnesota. Thesis. University of
Wisconsin-Stevens Point, Stevens Point, Wisconsin, USA.

Verme, L J. 1965. Swamp conifer deeryards in northern Michigan: their ecology
and management. Journal of Forestry 63:523-529.

_, and W. F. Johnson. 1986. Regeneration of northern white cedar deeryards in
Upper Michigan. Journal of Wildlife Management. 50:307-313.

White, G. C. and R. A. Garrot. 1990. Analysis of wildlife radio-tracking data.
Academic Press, Inc. San Diego, California, USA.

Wilson, M. A. 1999. Public attitudes towards wolves in Wisconsin. Pages 66-70 in
Wisconsin wolf management plan. Wisconsin Department of Natural
Resources Madison, Wisconsin, USA.

Worton, B. J. 1989. Kernel methods for estimating the utilization distribution in

home-range studies. Ecology 70: 164-168.

82

Wydeven, A. P., R. N. Schultz, and R. P. Thiel. 1995. Monitoring of a recovering
gray wolf population in Wisconsin. 1979-1991. Pages 147-156 in L. N. Carbyn, S.
H. Fritss, and D. R. Seip editors. Ecology and conservation of wolves in a
changing world. Canadian Circumpolar Institute. Occasional publication No 35.
Zar, J. H. 1999. Biostatistical analysis. Prentice Hall, Upper Saddle River, New

Jersey, USA.

83

1‘. {2.411%
f‘v‘
A -

rm
4' a

Appendix A. Original and consolidated land cover categories within the Highway 53
Study Area in northwest Wisconsin.

 

 

 

 

 

 

 

 

 

 

 

 

WISCLAND code Description Consolidated code
100 URBAN/DEVELOPED 100

High Intensity
104 Low Intensity
110 AGRICULTURE 110
1 13 Corn
1 18 Other row crops
124 Forage crops
150 GRASSLAND 150
162 JACK PINE (Pinus banksiana) 162
163 RED PINE (P. resinosa) 163
166 WHITE SPRUCE (Picea glauca) Excluded from analysis
173 M1XED/OTHER CONlFEROUS 173
176 ASPEN (Populus spp.) 176
177 OAK (Quercus spp.) 177
179 Northern pin oak (Q. paluslris)
180 Red oak (Q. rubra)
183 MAPLE (Acer spp.) 183
185 Sugar Maple (Acer saccharum)
187 MIXED/OTHER BROAD 187

LEAVED DECIDUOUS
190 MIXED 190
DECIDUOUS/CONIFEROUS

200 OPEN WATER Excluded from analysis
21 l EMERGENT/WET MEADOW 21 1
212 Floating aquatic vegetation
217 LOWLAND SHRUB 217
218 Broad-leaved deciduous
219 Broad-leaved evergreen
220 Needle-leaved
222 FORESTED WETLAND 222
223 Broad-leaved deciduous
229 Coniferous
234 Mixed deciduous/coniferous
240 BARREN Excluded from analysis
250 SHRUBLAND 250

 

84

 

Appendix B. Compositional analysis based pair-wise comparisons of landscape level
land cover category selection by gray wolves in the Highway 53 Study Area of northwest
Wisconsin 1995-2001”.

110 150 I62 163 I73 176 177 183 187 190 211 217 222

150 +*

162 +

163 +*

173 +*

176 +*

177 +*

183 +*

187 +*

190 +"‘

211 +* + - +

217 +* + + + + +*

222 +* +* + +* + +* +*
_* an: at:

 

250 + - -*

+ indicates category in left column was preferred over category in top row.
- indicates category in top row was preferred over category in the left column.
* indicates preference was significant based on a 1 distribution at a = 0.05.

3 Definitions of WISCLAND land cover category codes:

1 10 Agricultural

150 Grassland

162 Jack Pine

163 Red pine

173 Mixed/other coniferous

176 Aspen

177 Oak

183 Maple

187 Mixed/other broad-leaved deciduous
190 Mixed/deciduous coniferous
21 l Emergent/wet meadow

217 Lowland shrub

222 Forested wetland

250 Shrubland

85

Appendix C. Compositional analysis based pair-wise comparisons of within-territory
land cover selection by gray wolves in the Highway 53 Study Area of northwest
Wisconsin 1 995-200 1 A.

110 150 162 163 173 176 177 183 187 190 211 217 222
150 "
162 '
163 '

173 -
176 +*

177 +*

183
187
190

21 1 ' '
217 - + + + - -
222 +* +* +* +* *+ +* + +*
250 - + - - + - - + -
+ indicates category in left column was preferred over category in top row.
- indicates category in top row was preferred over category in the left column.

" indicates preference was a significant based on a (distribution at a = 0.05.
B Definitions of WISCLAND land cover category codes:

1 10 Agricultural

150 Grassland

162 Jack Pine

163 Red pine

173 Mixed/other coniferous

176 Aspen

177 Oak

183 Maple

187 Mixed/other broad-leaved deciduous
190 Mixed/deciduous coniferous
211 Emergent/wet meadow

217 Lowland shrub

222 Forested wetland

250 Shrubland

   

 

86

Appendix D. Compositional analysis based pair-wise comparisons of within-territory
land cover selection by gray wolves in the Highway 53 Study Area of northwest
Wisconsin during “summer” seasonal periods (15 April-14 September) 1995-2001”.

110 150 162 163 173 176 177 183 187 190 211 217 222
150
162
163
173
176
180
183
187
190
211
217

222 '+

250 -* - -

+ indicates category in left column was preferred over category in top row.
- indicates category in top row was preferred over category in the left column.
"' indicates preference was a significant based on a t distribution at a = 0.05.
C Definitions of WISCLAND land cover category codes:

1 10 Agricultural

150 Grassland

162 Jack Pine

163 Red pine

173 Mixed/other coniferous

176 Aspen

177 Oak

183 Maple

187 Mixed/other broad-leaved deciduous
190 Mixed/deciduous coniferous
21 l Emergent/wet meadow

217 Lowland shrub

222 Forested wetland

250 Shrubland

+
*

 

+
+
+
+
+
+
+
+

+— +— +— +— 4- 4— +-

4— +-

 

87

Appendix E. Compositional analysis based pair-wise comparisons of within—territory
land cover selection by gray wolves in the Highway 53 Study Area of northwest
Wisconsin during “winter” seasonal periods (15 September - 14 April) 1995-2001”.

110 150 162 163 173 176 177 183 187 190 211 217 222
150 '
162
163
I73
176
177
183
187

190 "

21 1 - - - + - +

217 + +* + + + + + +* + +

222 +* +* +* +31: +* +* +* +* +=lt +*

250 - + - - + - - + - +
+ indicates category in left column was preferred over category in top row.

- indicates category in top row was preferred over category in the left column.

* indicates preference was based on a 1 distribution ata = 0.05.
8 Definitions of WISCLAND land cover category codes:

1 10 Agricultural

150 Grassland

162 Jack Pine

163 Red pine

173 Mixed/other coniferous

176 Aspen

177 Oak

183 Maple

187 Mixed/other broad-leaved deciduous
190 Mixed/deciduous coniferous
211 Emergent/wet meadow

217 Lowland shrub

222 Forested wetland

250 Shrubland

+
+
+
+
+
+
+

 

88

CHAPTER 4
GRAY WOLF DEN SITE SELECTION IN NORTHWEST
WISCONSIN: ADDITIONAL DATA
INTRODUCTION
Much effort has been focused on investigating biotic and abiotic attributes of gray wolf
(Canis lupus) dens and den site selection (Joslin 1966, Joslin 1967, Mech 1970, Peterson
1977, Ballard and Dau 1983, Unger 1998). This is likely due to the importance of den

sites to overall wolf ecology and population dynamics. The location of den sites may

1
l
I

determine the reproductive success of each wolf pack (Harrington and Mech 1982) and
have a direct influence on overall population dynamics of the wolf population in an area.
Investigations have examined both pack-level and micro-scale selection of den
sites. Ballard and Dau (1983), Gehring (1995), and Unger (1998) reported that den sites
tended to be located in the approximate center of territories. Stephenson and Johnson
(1973) and Ballard and Dau (1983) examined the distance between the dens of adjacent
packs and reported that the average distance between dens was approximately 40 - 45
km. Mech (1970) and Fuller (1983) noted that den sites appeared to be selected based on
slope, sandy soil, and aspect. Lawhead (1983) also examined micro-scale selection and

reported that depth to permafrost, active soil layer thickness, and drainage were important

 

attributes influencing den site location.
The importance of den success and den site placement led the WDNR to
investigate den site characteristics and placement (Unger 1998) as a component of the

overall H53 Expansion Project (Kohn et a1. 2000). The study by Unger (1998) used a

geographic information system (GIS) to analyze den site locations within annual

89

territories in relation to 1) an arbitrarily defined inner core of the territory, 2) roads, and
3) land cover categories. He also conducted a logistic regression analysis on the
variables collected at dens and corresponding random sites to determine which variables
most influenced den site selection. Unger (1998) found that wolves were selecting a
central core of a territory for dens. This inner core was centered in the territory and was
the same shape as the territory boundary, but encompassed only 25% of the territory’s
area. Unger (1998) hypothesized that the prevention of interpack strife and optimal
foraging behavior might explain this behavior. He also reported that at the microscale

level, wolves selected for areas of sandy soil and steeper slope, possibly for ease of

 

digging and drainage purposes.

While providing useful insight into wolf den site placement and characteristics,
Unger’s (1998) investigation was conducted on a relatively small number of dens (n =
13) due to the relatively small number of den sites across any landscape and the difficulty
in locating them. When research examining other aspects of wolf ecology (Chapters 2
and 3, this document) in the H53SA began, the WDNR requested that effort be directed
at locating and delineating additional den sites as part of the continuing wolf population
monitoring program. With this request in mind, I attempted to locate wolf den sites

during the summers of 1998-2000 and to investigate den site location and characteristics

 

in a manner similar to that used by Unger (1998).
The objectives of this study were to: 1) locate and delineate gray wolf den sites
within the H53 SA, 2) determine which landscape level factors (placement within territory

central core, presence/absence of and density of roads near dens) most influence wolf den

90

site selection, and 3) determine if wolves are selecting land cover categories in a non-
random manner when establishing dens.
METHODS
Capture, Handling, and Radiotelemetry

All methods follow those used by Unger (1998) unless noted. Wolves were
trapped by WDNR personnel using modified #14 Newhouse leg-hold traps (Schultz et al.
1996) during May-August in 1997-2000. Captured wolves were immobilized using a
mixture of ketamine hydrochloride/xylazine hydrochloride administered with a jabstick.
Any trap-related injuries were treated, blood samples were taken, an injection of
antibiotics was administered, and animals were weighed. Animals > 15kg were also
fitted with radio collars (Telonics Inc., Mesa, Arizona) and uniquely numbered ear tags.
Yohimbine was administered as reversal agent for the xylazine and the animals were
monitored until they could move away from the trap site under their own power.

Radio-collared animals were located 1-5 times per week during 1997-2000 using
telemetry from fixed-wing aircraft (Ballard and Dau 1983, Fuller 1989) or a vehicle-
mounted S-element Yagi directional antenna until radio failure, death of the animal, or a
permanent movement out of the study area. The majority (>90%) of locations were
collected during daylight hours and a minimum of 3 bearings were collected for each
location. Triangulations for locations collected on the ground where solved using a Lenth
estimator (Lenth 1981) in the program Locate II (Pacer, Turo Nova Scotia, Canada).
Date, time, Universal Transverse Mercator (UTM) coordinates, ear tag number, and pack
were recorded for each telemetry location. Animals were classified as belonging to

individual packs based on trapping location and movement patterns.

91

 

 

Den Site Location and Delineation

Areas that showed a tight cluster of telemetry locations (generally >3 locations in
a 1 km2 area) between February and April were searched for dens between July and
September of the same year. The locations of dens were recorded using a hand-held
global positioning system (GPS) (Garmin International, Lenexa, Kansas) and mapped
using ArcView (Environmental Systems Research Incorporated, Redlands, CA).

Annual territory boundaries were estimated using the 95% minimum convex
polygon method (MCP) (Mohr 1947) in the animal movement analysis extension in
ArcView (Hooge and Eichenlaub 1997). The MCP estimation technique was used
instead of the fixed kernel used in chapters 2 and 3 to avoid the potential bias caused by
examining telemetry locations centered on the den. Due to the algorithms used in a
kernel technique, the clusters of locations around dens would have made it more likely
that dens were located in the center of estimated territories. A 95% MCP was used to
maximize the distribution of wolf locations while allowing outliers to be eliminated.
Unger (1998) used a 99% MCP, but with territory estimation generally based on less than
100 telemetry locations, I believed that a 99% MCP allowed for the removal of an
unacceptably small number of potential outlier points.

Territories were required to have at least 30 locations (Fuller and Snow 1988)
obtained in at least 6 different months of the year to be included in analysis. Territories
were estimated based on yearly telemetry locations. A year was defined as 1 June to 30
May. This led to the calculation of territory boundaries based on telemetry locations
from approximately the time of den site abandonment the previous year through the

denning period the year of analysis.

92

 

Random sites corresponding to each den site were generated within territories
based on 1 random site per 4,000 ha of territory (Unger 1998). All variables examined at
den sites were also analyzed at random sites.

Analysis of Variables Influencing Den Site Selection

To examine whether wolves were placing dens near the center of their territories
as observed by Ballard and Dau (1983), Gehring (1995), and Unger (1998), I generated
an area the same shape as the estimated territory boundary, but occupying 25% of the
area of the entire territory and centered in the territory; hereafter referred to as core areas.
Core areas were created in ArcView by calculating the size of an area encompassing 25%
of the total area of territories. Territory boundaries were negatively buffered (buffers
were inside the boundary, not outside) at a distance that resulted in a polygon that
encompassed 25% of the area of the territory. Location of dens and random points were
recorded as occurring within this core area or outside of it. This technique was used by
Unger (1998) and the selection of a core encompassing 25% of the area of a territory was
an arbitrary choice.

When examining the affect of roads on den placement, Unger (1998) created a
1000 m radius buffer zone around the den or random point and overlaid these sites with a
roads coverage in ArcView. I chose instead to base the size of this buffer on the average
distance from the den an alpha female would travel (1212 111) during the denning period
(1 April — 1 July) based on radio telemetry locations. This distance was based on the one
confirmed alpha female monitored during the study. While this technique relied on data
from only one individual, I believed that it was more tightly tied to the biology of the

species and equally as valid as any arbitrarily set distance. The roads coverage used in

93

l _
l

‘T‘

 

this analysis was based on U. S. Geologic Survey 1:100,000 scale Digital Line Graphs
and included all road features that were not classified as trails. Presence or absence of
roads within buffers around dens or random points was recorded. Road density (km/kmz)
was also calculated by determining the straight-line distance of all road segments within a
territory in ArcView and dividing by the area of the territory.

A Fisher’s Exact test (a = 0.10) was used to test for differences in the frequency
of den or random site location within the 25% inner core of territories. This test (or =
0.10) was also used to examine differences in the occurrence of roads within the buffer
around dens and random sites. A two sample T-test (or = 0.10) was used to test for
differences in road densities within buffers around dens and random sites.

When analyzing the selection of land cover type category around den and random
sites, Unger (1998) created an arbitrarily defined 400 ha plot centered on the den or
random site. He judged this area to represent the area that wolves would most likely use
during the denning period. I chose instead to base the size of this buffer on the average
distance from the den an alpha female would travel (1212 m) during the denning period
(1 April — 1 July) based on radio telemetry locations. Buffers of this size were created
around each den and random point in ArcView.

The proportions of buffers around each den and random point classified as
grassland, pine, mixed/other conifer, aspen, oak, maple, mixed/other broad leaved
deciduous, mixed deciduous/coniferous, emergent wet meadow, lowland shrub, forested
wetland, and shrubland within the WDNR’s GIS WISCLAND land cover classification

system (http://wwwdnr.statewi.us/org/at/et/geo/data/wlc.htm) was calculated. These 12

 

categories were consolidated from the original 28 categories present within the H53SA.

94

This consolidation was conducted based on the following 3 criteria. Any cover type that
represented less than 0.10% of all territories being examined and which was present in
the den and random site buffers in less than 4 pack territories was excluded from analysis.
Any cover type category, which based on traditional paradigms of wolf ecology, would
not be used by wolves for denning (i.e., open water and urban) was excluded from the
analysis. Similar cover types that represented little variability in resources to wolves
were combined (i.e., red oak Quercus rubra and northern pin oak Quercus ellipsoidalis).
The proportion of each cover type category was examined for differences between

buffers around dens and random unused points with a T-test (or = 0.10).

The calculated proportions of each land cover category, presence or absence of
and density of roads within 1212 m of den or random site, and placement within or
outside the 25% inner core of the territory were all entered into a forward step logistic
regression analysis to firrther determine which measured variables influence den site
selection. All non-categorical and non-binomial variables were analyzed with 3 Pearson
rank correlation analysis to identify variables that were correlated before entering them
into the logistic analysis.

A compositional analysis (CA) (Aebischer et al. 1993) was conducted (on = 0.10)

to examine selection of the 12 consolidated cover types in areas around dens and random

 

sites. CA involves a use vs. availability approach to delineate and quantify selection,
while avoiding the problem of non-independence of proportions calculated for use and
availability in many habitat selection techniques. The proportions of each cover type in
the buffers around den sites represented “use” in this analysis. The proportion of the

entire territory classified as each cover type category of interest represented

 

95

 

“availability.” These proportions were calculated by determining the area of each pack
territory in a land cover category within ArcView and dividing by the total area of the
territory. The calculations for CA were conducted using ReSelect software

(http://ces.iisc.ernct.in/hpg/envis/resdocl 120.html).

 

RESULTS
Den Site Location and Delineation

Nine dens were located during 1998-2000. The locations of 5 additional dens
from 1993-1996 that could not be analyzed during the Unger (1998) study were provided
by D. Unger (Alderson Broaddus College, Phillipi, VA.), resulting in 14 den sites for
analysis. Forty-three (1-7 per territory based on 1/4000 ha) random unused points
corresponding to den sites were generated. Using a 1212m radius to buffer dens and
random sites resulted in buffers of 461 ha for analysis of land cover category.
Variables Influencing Den Site Selection

Den sites (11 of 14, 78%) were more likely (p = 0.002) to fall within the 25%
inner core of a territory than random points (15 of 44, 35%). Dens (7 of 14, 50%) were
also less likely (p = 0.029) to occur within 1212 m or less of a road than random points
(33 of 43, 76%). Average road density within buffers around dens was less (p = 0.095)
than around random points (0.37km/km2 vs. 0.6lkm/km2, respectively).

Analysis of the proportion of land covers within buffers around dens and random
points indicated that pine, oak, and shrub categories comprised higher proportions of
buffers around dens sites (Table 4.1). This analysis also indicated that lowland shrub

comprised lower proportions of buffers around den sites than random sites (Table 4.1).

96

Table 4.1 Land cover category composition in 1212m radius buffers around gray wolf
dens (n =14) and corresponding random sites (11 =43) in the H53SA in northwest
Wisconsin 1993-2000.

 

x % composition

 

x % composition (SE) of random point A
Land cover category (SE) of den buffers buffers P
Oak 3.3 (1.1) 3.0 (1.0) 0.014
Pine 8.1 (2.9) 5.6 (1.6) 0.071
Shrub 6.3 (2.5) 3.6 (1.5) 0.093
Lowland shrub 10.1 (2.3) 16.0 (1.8) 0.068
Aspen 22.1 (2.2) 26.9 (2.2) 0.194
Forested wetland 12.1 (1.8) 12.1 (1.0) 0.579
Mixed/other coniferous 1.4 (0.4) 1.4 (0.3) 0.286
Grassland 6.7 (0.9) 6.9 (1.2) 0.331
Maple 1.5 (0.4) 1.5 (0.4) 0.661
Emergent wet meadow 2.5 (0.9) 3.1 (0.7) 0.743
Mixed deciduous/coniferous 3.9 (1.1) 5.8 (1.0) 0.736
Mixed/other deciduous 15.1 (3.9) 13.9 (1 .8) 0.949

 

A Significant at or = 0.10 t-test

97

 

terminal-q
5.: I

 

 

No other differences in land cover category composition between den and random sites
were observed (Table 4.1).

The Pearson rank correlation matrix of variables being considered for inclusion in
the logistic regression indicated that no variables were highly correlated (r in all cases <
0.5). When all landscape and land cover type variables were entered into the forward
step logistic procedure, the final model output consisted of the logistic generated
intercept and binomial response for location within or outside of the 25% inner core of

the territory. A backward step regression procedure resulted in the same combination of

—2.268 +1.959(C)
1- 2.268 +1.959(C)

 

variables. The equation for this model was: p =

With a logit of: p = -2.268 +1.959 (C) where C is the binomial response of within or
outside the 25% inner core of the territory. Tests of classification accuracy for logistic
models involve converting the output of the logistic equation for each observation to a
binomial (0/1, success/failure, den/random) response through the use of a cutpoint (c)
(Hosmer and Lemeshow 2000) between 0.0 and 1.0. If the output of the logistic
regression equation for an individual observation falls below c, the observation is
predicted to be in group 0 (failure, in this case a random unused point). If the logistic
output is above this c, the observation is predicted to be part of group 1 (success, in this
case a den site). The actual groups that each individual observation belongs to are
compared to the predicted groups to determine the classification accuracy of the logistic
model. The ability to vary the 0 between the 0.0-1.0 probability interval allowed me to
determine the optimal value of c for successfully classifying used and unused sites

(Pereiara and Itami 1991). The maximum classification accuracy for the logistic equation

98

1"

let-'-

 

 

based on inner 25% core location was reached at a c of 0.40. Using this c, the model
correctly classified 11 of 14 den sites (79%) and 29 of 43 random sites (67%) for an
overall classification accuracy of 70%.

Results from the CA indicated that mixed deciduous, aspen, and lowland shrub
categories were the three most selected cover types and that maple, grassland, and oak
were the three least selected cover types by wolves (Table 4.2 and Appendix B).
DISCUSSION

My findings that wolves tend to place dens near the center of their territory
support the findings of Ballard and Dau (1983), Gehring (1995), and Unger (1998), but
refute those of Ciucci and Mech (1992) who found that within an area encompassing the
inner 60% of monitored territories, dens were located randomly. The use of the 60%
inner core by Ciucci and Mech (1992) compared to the inner 25% inner core in this study
and by Unger (1998) makes direct comparisons of results problematic.

Ciucci and Mech (1992) suggested that the tendency to place dens in the center
of territories was related to the comparatively larger territory sizes of the Alaskan wolves
observed by Ballard and Dau (1983). Wolves must repeatedly bring food sources to the
den to support the pups (Mech 1970, Groebner 1991) and have been shown to travel long
straight line distances when bringing prey to the den (Mech 1970, Young and Goldman
1944). Therefore, the benefits of a centrally located den to travel and foraging efficiency
when supporting the alpha female and pups at the den seem evident, but may not be
entirely dependent on territory size. Both Unger (1998) and I documented the tendency
to locate dens in the center of territories even though the observed territory size in both

studies (202km/km2 and 142 km/kmz, respectively) were comparable to those observed in

99

 

Table 4.2. Compositional analysis rankings (1 most preferred, 12 least preferred) of
land cover category selection at gray wolf den sites within the Highway 53 Study
Area in northwest Wisconsin 1993-2000.

 

Cover Type Category CA Ranking
Mixed/other broad-leaved deciduous 1
Aspen 2
Lowland shrub 3
Emergent wet meadow 4
Mixed deciduous/coniferous 5
Shrubland 6
Mixed/other coniferous 7
Forested wetland 8

Pine 9 "
Oak 10
Grassland 11
Maple 12

 

 

 

100

 

Minnesota (Van Ballenberghe et al. 1975, Fritts and Mech 1981, Berge and Kuehn 1982,
Fuller 1989) and smaller than those observed in Alaska (Peterson et al. 1984).

The other hypothesis explaining the tendency of wolves to place dens in the
center of territories is that this tendency is a mechanism to avoid contact and conflict with
wolves from other packs (Ciucci and Mech 1992, Unger 1998). Past authors have shown
that buffer zones between packs are often the site of interpack aggression and occasional
mortality (Peters and Mech 1975, F ritts and Mech 1981). Ciucci and Mech (1992) also
found that neighboring packs had a major influence on territory use and the distribution

of kills and hypothesized that adjacent packs could also influence den location. My data

 

seem to support the hypothesis that wolves tend to place dens in the center of territories
because this lowers the probability of wolves not in the natal pack encountering the den.
Of the 3 dens that were not located in the 25% inner core of the territory, 2 were in
territories, which to my knowledge had no other territories directly abutting them. The
lack of the threat of aggression from wolves directly neighboring these packs may have
allowed placement of the den in other than the center of the territory.

The lower road densities in buffers around pack territories than in buffers around

 

random sites is consistent with the documented avoidance of roads related to overall wolf

ecology (Mladenoff et a1. 1995, 1999, Chapters 2 and 3, this document) as well as past

 

research in the HSBSA (Unger 1998). The avoidance of roads during such a sensitive
segment of the wolf life cycle as denning, is not surprising since wolves avoid roads and
the associated human activity when establishing territories (Mladenoff et al. 1995, 1999,

Chapters 2 and 3, this document).

 

101

The results from CA and the comparison of land cover category composition
between den and random sites indicated that wolves are selecting for cover types
indicative of well-drained upland soils. This is consistent with Unger’s (1998) findings
in the same area and is likely related to the nearly exclusive use of excavated burrows as
dens in the H53SA. Unger (1998) observed that 11 of 12 dens he located were burrows
and all of the dens analyzed in this study were burrows. Excavating a den is presumably
easier and more efficient in well-drained upland soils than more lowland areas.
SUMMARY AND MANAGEMENT IMPLICATIONS

Disturbance of den sites may have an effect on the denning behavior of wolves
(Stephenson 1974, Joslin 1966, 1967). This coupled with the importance of the den site
and denning behavior of wolves on the local wolf population (Harrington and Mech
1982) emphasizes the importance of being able to predict potential den sites. If potential
denning areas can be identified, limitations on highly disruptive human activities during
the denning season can be imposed in these areas. Areas with a high likelihood of den
site usage could also potentially be protected by limiting access on a longer-terrn basis or
by limiting road density as management priorities dictate.

Based on the results of this study and that of Unger (1998), the most useful
indicator of wolf den site placement is location within the inner 25% core of a pack
territory. Delineating this core area requires the estimation of the boundaries of the
territory of interest. This is easiest when radio telemetry locations on the pack are
available, but as the wolf population in the upper Great Lakes region continues to
increase and is reclassified under the Endangered Species Act, progressively smaller

percentages of the population will be monitored via radio telemetry. Although this

102

increases the effort required to determine the inner 25% core, the WDNR already
estimates the boundaries of approximately 35% of the pack territories in the state using
techniques other than radio telemetry (i.e., track surveys, howling surveys) as part of their
population monitoring program (A. Wydeven, J. Wydenhoff, WDNR, pers. commun).
This indicates that territory boundaries and the inner 25% core of a territory can be
determined during normal management activities.

Once the inner 25% core of a territory has been located, other factors affecting
den site selection (land cover type category, road density) can be used to further refine
predictions of likely den sites. This insight into potential denning areas will prove useful
whether the objective of determining the den site is to restrict access to the area. locate
the den for research purposes, or any other management activity.

The majority of cattle depredations by wolves in the H53SA during the course of
my study occurred during the denning period. This was apparently due to the
combination of wolves foraging to bring constant food sources back to the den and the
availability of young and relatively vulnerable cattle during this period. Unger (1998)
suggested that the tendency of wolves to place dens in the 25% inner core of a territory
could prove useful to managers when dealing with depredation problems. He suggested
“Given that wolves may be selecting the inner core of their territory for a den site to
minimize interpack conflict, recorded wolf howling and/or distribution of wolf scat may
deter wolves from placing the den in an undesirable location,” the undesirable location
referring in this case to den placement in proximity to cattle farms. Additionally, farms

located within the 25% inner core of territories could be targeted for educational

103

programs and funding for measures targeted at preventing depredation such as improved

cattle enclosures, guarding dogs, and wolf exclusion techniques.

104

LITERATURE CITED

Aebischer, N. J ., P. A. Robertson, and E. Kenward. 1993. Compositional analysis of
habitat use from animal radio-tracking data. Ecology 74:1313-1325.

Ballard, W. B., and J. R. Dau. 1983. Characteristics of Gray Wolf, (Canis lupus),
den and rendezvous sites in south-central Alaska. Canadian Field Naturalist
97:299-302.

Berge, W. E., and D. W. Kuehn. 1982. Ecology of wolves in north-central Minnesota.
Pages 4-11 in F. H. Harrington and P. C. Paguet, editors. Wolves of the world;
perspectives on behavior, ecology, and conservation. Noyes, Park Ridge, New
Jersey, USA.

Ciucci, P., and L. D. Mech. 1992. Selection of wolf dens in relation to winter territories
of northeastern Minnesota. Journal of Mammalogy 73:899-905.

Fritts, S. H., and L. D. Mech. 1981. Dynamics, movements, and feeding ecology of a
newly protected wolf population in northwestern Minnesota. Wildlife
Monographs 80.

Fuller, T. K. 1983. Denning behaviors of wolves in north-central Minnesota.
American Midland Naturalist 121:188-193.

_, and W. J. Snow. 1988. Estimating winter wolf densities using radiotelemetry
data. Wildlife Society Bulletin 16:367-370.

1989. Population dynamics of wolves in north-central Minnesota.

Wildlife Monographs 105.

105

Gehring, T. M. 1995. Winter wolf movements in northwestern Wisconsin and
east-central Minnesota: a quantitative approach. Thesis. University of
Wisconsin-Stevens Pt. Stevens Point, Wisconsin, USA.

Groebner, D. J. 1991. Summer movement rates of a pair of northeastern Minnesota
timber wolves. Thesis. Northern Michigan University. Marquette, Michigan,
USA.

Harrington, F. H., and L. D. Mech. 1982. Patterns of homesite attendance in two
Minnesota wolf packs. Pages 81-105 in Harrington, F. H. and P. C. Paquits
editors. Wolves of the World. Noyes Publishing. Park Ridge, New Jersey, USA.

Hooge, P. N., and B. Eichenlaub. 1997. Animal movement extension to Arcview. Version

1.1 Alaska Biological Science Center, US. Geological Survey, Anchorage,
Alaska.

Hosmer, D. W., and S. Lemeshow. 2000. Second Edition. Applied logistic regression.
John Wiley and Sons. New York, New York, USA.

Joslin, P. W. B. 1966. Summer activities of two timber wolf packs in Algonquin Park.
Thesis. University of Toronto, Ontario, Canada.

__ 1967. Movements and home sites of timber wolves in Algonquin Park.

American Zoologist 7:279-288.

Kohn, B. E., J. L. Frair, D. E. Unger, T. M. Gehring, D. P. Shelley, E. M. Anderson,
and P. Keenlance. 2000. Impacts of the US Highway 53 expansion project on
wolves in northwestern Wisconsin. Wisconsin Department of Natural Resources,

Rhinelander, Wisconsin, USA.

106

‘71!""3‘

 

Lawhead, B. E. 1983. Wolf den site characteristics in the Nelchina Basin, Alaska.
Thesis. University of Alaska-Fairbanks, Fairbanks, Alaska, USA.

Lenth, R.V. 1981. On finding the source of a signal. Technometrics 23:149.

Mech, L. D. 1970. The wolf: the ecology and behavior of an endangered species.
Natural History Press, Garden City, New York, USA.

Mladenoff, D. J ., T. A. Sickley, R.G. Haight, and A. P. Wydeven. 1995. A regional
landscape analysis and prediction of favorable gray wolf habitat in the northern
Great Lakes region. Conservation Biology 9:279-294.

_, T. A. Sickley, and A. P. Wydeven. 1999. Predicting gray wolf landscape
recolonization: logistic regression models vs. new field data. Ecological
Applications 9:37-34.

Mohr, C. O. 1947. Table of equivalent populations of North American mammals.
American Midland Naturalist 37:223-249.

Peters, R. P. and L. D. Mech. 1975. Scent-marking in wolves. American Scientist
63:628-63 7.

Peterson, R. O. 1977. Wolf ecology and prey relationships on Isle Royale. National Park
Service. Scientific Monograph Series 11.

_, R. 0., J. D. Woolington, and T. N. Bailey. 1984. Wolves of the Kenai Peninsula,
Alaska. Wildlife Monographs 88.

Pereiara, J. M., and R. M. Itami. 1991. G18 based habitat modeling using logistic
multiple regression: A study of the Mount Graham red squirrel. Photogrammic

Engineering and Remote Sensing 57:1475-1486.

107

Schultz, R. N., A. P. Wydeven, and R. A. Megown. 1996. Injury levels with five types
of leg-hold traps in Wisconsin. Pages 38-39 in 14th Midwest Furbearer
Workshop, Ironwood, Michigan, USA.

Stephenson, R. O. 1974. Characteristics of wolf den sites. Alaska Department of
Fish and Game. Pittman-Robertson Final Report, W-l7-2 through W-17-6.

Stephenson, R. O. and L. Johnson. 1974. Wolf report. Pittman Robertson Project final
report, W-17-4. Job No. 14.3R through 14.7R. Alaska Department of Fish and ,
Game.

Unger, D. E. 1998. A multi-scale analysis of timber wolf den and rendezvous sites in

 

northwestern Wisconsin and east-central Minnesota. Thesis. University of
Wisconsin-Stevens Point, Stevens Point, Wisconsin, USA.

Van Ballenberghe, V. A., Erickson, and D. Byman. 1975. Ecology of the timber wolf in
northeastern Minnesota. Wildlife Monographs 43.

Young, S. P., and E. A. Goldman. 1944. The wolves of North America. Wildlife

Institute, Washington, DC. USA.

 

 

108

 

Appendix A. Original and consolidated land cover categories used for compositional
analysis in 1212m buffers around den and random points within the Highway 53
Study Area in northwest Wisconsin.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

WISCLAND code Description Consolidated code
100 URBAN/DEVELOPED Excluded from analysis
High Intensity
104 Low Intensity
110 AGRICULTURE Excluded from analysis
1 13 Com
1 18 Other row crops
124 Forage crops
150 GRASSLAND 150
160 PlNE
162 Jack Pine (Pinus banksiana) 162
163 Red Pine (P. resinosa)
166 WHITE SPRUCE (Picea glauca) Excluded from analysis
173 MIXED/OTHER CONIFEROUS 173
176 ASPEN (Populus spp.) 176
177 OAK (Quercus spp.) 177
179 Northern pin oak (Q. palustris)
180 Red oak (Q. rubra)
183 MAPLE (Acer spp.) 183
185 Sugar Maple (Acer saccharum)
187 MIXED/OTHER BROAD 187
LEAVED DECIDUOUS
190 MIXED 190
DECIDUOUS/CONIFEROUS
200 OPEN WATER Excluded from analysis
211 EMERGENT/WET MEADOW 211
212 Floating aquatic vegetation
217 LOWLAND SHRUB 217
218 Broad-leaved deciduous
219 Broad-leaved evergreen
220 Needle-leaved
222 FORESTED WETLAND 222
223 Broad-leaved deciduous
229 Coniferous
234 Mixed deciduous/coniferous
240 BARREN Excluded from analysis
250 SHRUBLAND 250

 

109

 

Appendix B. Compositional analysis based pair-wise comparisons of land cover category

by denning gray wolves in the Highway 53 Study Area of northwest Wisconsin 1993-

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2000A.
LandcoverB
category 150 162 173 176 180 183 187 190 211 217 222
162 +
173 + +
176 +* + +
180 + - - -*
183 - - -* -* '-
187 +* + + + +* +*
190 + + + - + +* -
2“ +* + + - + +* - +
217 + + + - + +* - + +
222 + + - - + +* - - - -
250 + + + - + +* - - - - +

 

 

 

 

 

 

 

 

 

 

 

A + indicates category in left column was preferred over category in top row.
- indicates category in top row was preferred over category in the left column.
* indicates preference was a significant based on a t distribution at a = 0.05.
8 Definitions of WISCLAND land cover category codes:

150 Grassland

162 Pine

173 Mixed/other coniferous

176 Aspen

180 Oak

183 Maple

187 Mixed/other broad-leaved deciduous
190 Mixed/deciduous coniferous
211 Emergent/wet meadow

217 Lowland shrub

222 Forested wetland

250 Shrubland

110

 

CHAPTER 5
SYNTHESIS AND FUTURE RESEARCH DIRECTION
SYNTHESIS

This final chapter is intended to provide a place to recap and consolidate the

 

information presented in the previous chapters of this dissertation. It is also intended to
provide me the opportunity to present some thoughts and or hypotheses on the

implications of these data and their potential impacts on management actions. The final

'3,

segment of this chapter will present suggestions for future research directions or topics
that have been suggested by the results of the research conducted during this project.
Human tolerance of or legal protection of wolves and the density of large
ungulate prey species are the two major limiting factors on most wolf population (Fuller
1989). Current forestry practices in Wisconsin are maintaining white-tailed deer
populations at densities that can support wolf populations larger than the target
population set by WDNR (Mladenoff et al. 1997, Wisconsin Department of Natural

Resources 1999). Therefore, no change in forest management practices is currently

 

required to support ungulate densities that can maintain healthy wolf populations in the

state. However, maintaining current white-tailed deer populations within the state will 5"
have negative affects on other aspects of forest ecology, especially regeneration of
browse species preferred by deer browse (Anderson and Loucks 1979, Verme and 34‘
Johnson 1986, Alverson et. al. 1988, Tilghman 1989, Pregitzer, 1990). These problems

should be considered when decisions regarding deer population management and the

resultant effect on wolf populations are made.

111

Managing wolf populations at a level that promotes high levels of human
tolerance is a more difficult task than managing for adequate prey populations. The wolf
population in Wisconsin has consistently increased during the past 10 years (Wisconsin
Department of Natural Resources 1999) and likely will exceed the target population set
by WDNR in the near future. This target population was set at least in part in an attempt
to maintain the wolf population below social carrying capacity and therefore to maintain
high levels of public tolerance of wolves. If the wolf population is to be maintained at or
below this level after ESA reclassification, it is likely that individuals will have to be
removed from the population. This can be accomplished by paying professional hunters
or trappers to cull animals or by implementing some form of public harvest. While the
logistics of a public harvest would likely be problematic, the potential monetary savings
and more importantly the potential gain in support for healthy wolf populations among
the hunting and trapping public make a public harvest a viable management option. Due
to the potential difficulties in establishing a public harvest of wolves (resistance by some
segments of the public, and logistic concerns including harvest levels, harvest methods,
and harvest zones) I recommend that WDNR begin exploring the possibility of a public
harvest. Beginning to explore the possibility of and examine solutions to the difficulties
of establishing a public harvest now will avoid lag time after ESA reclassification when a
public harvest may be a needed population management option.

As predicted by their habitat generalist nature and normal avoidance of humans,
land cover selection by wolves appears to be driven by the acquisition of prey and the
avoidance of human activity. During this study, wolves selected forested wetland and

lowland shrub when placing territories on the landscape. Areas of forested wetland are

112

 

generally not hospitable to human activity that requires movement away from roads due
to difficulty in walking, high biting insect populations in the summer, and the general
lack of human development. This suggests that the selection of forested wetland areas is
as much a result of low human activity and the corresponding lower probability of human
disturbance or contact as a result of the cover type itself. This is also the likely causal
factor behind lowland shrub, and emergent wet meadow being in the top five most
selected cover types at the landscape level of selection. The selection of forested
wetland areas by wolves may also have been influenced by prey densities in these areas
due to the typical use of lowland conifer areas as yarding areas by white-tailed deer in
severe winters. Wolves also highly selected aspen at the landscape level and within
territories, presumably at least in part because of the high use of aspen areas by white-
tailed deer and the resultant access to prey in these areas. Therefore, observed
preferences in land cover category by wolves may actually reflect the importance of these
land cover types to white-tailed deer, which is the main prey species of wolves in
Wisconsin (Mandemack, 1983).

Under current levels of human tolerance and legal protection, wolves appear to be
expressing their adaptable nature and colonizing areas that in the past would have been
viewed as unsuitable. While this will likely lead to larger wolf populations that are less
likely to experience catastrophic population declines that might force reevaluation of
current efforts to reclassify them under the ESA, there are also potential negative effects
of this expansion of wolf populations and range. It is generally accepted that human
tolerance and the resulting level of human induced mortality is a major limiting factor on

wolf population levels. As the wolf population in the Great Lakes region expands its

113

geographic range into more human dominated areas, the potential for human/wolf
conflict will increase. This in turn may lower levels of human tolerance and have
negative impacts on the wolf population in the region, which may require resource
agencies to adjust current management practices.
Future Research Directions

As with most research projects, the results I observed brought to mind
additional questions. Investigating the answers to these questions could provide useful
insight into wolf resource use and selection in Wisconsin. The following is a list of those
questions and brief suggestions on how to address them.

The GIS model predicting favorable wolf habitat in PWR within Wisconsin
generally matches existing wolf pack territories in the state (Chapter 2). However, there
are 5 territories that occur in areas that are classified by the model as being in the lowest
habitat suitability category. Assuming these territories are persistent, it seems that there
must be some characteristic(s) of these areas that set them apart from other areas that the
model classifies as being low suitability wolf habitat. Since wolves appear to be
successfully using these areas, research into what sets them apart from other low
suitability areas could provide insight into where wolves may occur in the future. Since a
major component of the model is road density (as a surrogate for human activity), it is
likely that wolves establishing territories in areas classified as containing low suitability
habitat will have a higher likelihood of contact and conflict with humans. Increasing our
understanding of what leads wolves to establish territories in these “high risk” areas will

allow us to minimize the potential for human/wolf conflict.

114

In an attempt to avoid excessive “fishing” within my data and produce the most
parsimonious model in Chapter 2, I only included 5 land cover categories in the analysis.
The selection of land cover categories to be included in analysis was based on my
understanding of their importance in wolf ecology. However, these 5 categories only
encompassed approximately 55% of monitored pack territories and 25% of examined
unused areas. It is possible that land cover categories occupying the rest of the landscape
may have an impact on wolf territory placement. If this is true, a reanalysis of the data
and reformulation of the logistic model underlying my GIS-based habitat model could
provide a model with more predictive power.

In the logistic analysis of factors influencing den site placement (Chapter 4),
location within the inner 25% core of a territory was identified as the main variable
influencing den site placement. As noted in that chapter, the use of a 25% inner core
versus any other measure of core area was based on past research in the H53SA (Unger
1998) and was arbitrary. Examining the affect of core areas of varying sizes could

provide further insight into wolf den site placement.

115

LITERATURE CITED

Alverson, W. S., D. M. Waller, and S. L. Solheim. 1988. Forests too deer: edge effect in
northern Wisconsin. Conservation Biology 2:348-358. 7

Anderson, R. C., and O. L. Loucks. 1979. White-tailed deer (Odocoileus virginianus)
influence on structure and composition of Tsuga Canadensis forests. Journal of
Applied Ecology 16:855-861.

Mandemack, BA. 1983. Food habits of Wisconsin timber wolves. Thesis,

University of Wisconsin-Eau Claire. Eau Claire, Wisconsin, USA.

Pregitzer, K., 1990. The ecology of northern white-cedar. Pages 8-14 in Proceedings of
northern white-cedar in Michigan workshop. Sault Sainte Marie, Michigan.

Tilghman, N. G. 1989. Impacts of white-tailed deer on forest regeneration in northwest
Pennsylvania. Journal of Wildlife Management 53:524-532.

Unger, D. E. 1998. A multi-scale analysis of timber wolf den and rendezvous sites in
northwestem Wisconsin and east-central Minnesota. Thesis. University of
Wisconsin-Stevens Point, Stevens Point, Wisconsin, USA.

Verme, L J ., and W. F. Johnson. 1986. Regeneration of northern white cedar deeryards
in Upper Michigan. Journal of Wildlife Management. 50:307-313.

Wisconsin Department of Natural Resources. 1999. Wisconsin wolf management plan.

Madison, Wisconsin, USA.

116

' 1+ljljljljllyjjgnjjlgi1